KLAUS HENTSCHEL * The discovery of the redshift of solar Fraunhofer lines by Rowland and Jewell in Baltimore around 1890 ONE OF THE most important optical discoveries in the 19th century was the detection of sharp dark lines in the sun's spectrum by William Hyde Wollaston in 1802. In 1814, Joseph Fraunhofer increased the resolution of spectral observations enormously by observing through a telescope; thus he was the first to give a detailed and scaled map show- ing about 350 dark lines, soon to be called Fraunhofer lines, distri- buted over the entire visible spectrum of the sun. 1 These lines were of particular interest not only to Fraunhofer, but also to David Brewster and John Herschel, since they always appeared at the same position in the spectrum. Fraunhofer was interested in the lines as markers to determine the dispersive powers of prisms made out of different types *Depanment of History of Science, Gottingen University, Humboldtallee II, 3400 Gottingen, Germany. For comments on earlier versions of this paper, I would like to thank: Gerd Grassholf, Jost Lemmerich, the DPG history group, Dresden 1991, the edi- tors, and my wife Ann. For archival help and permission to quote from the Rowland papers as well as the Johns Hopkins University Collection 137, both kept at the Eisenhower Library, Johns Hopkins University, my thanks go to Joan Grattan, Cynthia H. Requardt and Marion Suesil. The following abbreviations are used: AA, Astronomy and astrophysicS; AJP, American journal of physiCS; AJS, American Journal of science; AP J, Astrophysical Journal; AP, An- nales de physique; AS, Annals of science; BMNAS, National Academy of Science, BIO- graphical memoirs; CRAS, Academie des Sciences, Paris, Comptes rendus; DAB, DictIOn- ary of American biography; DSB, DictIOnary of sClenllfic bIOgraphy; JHUC, Johns Hop- kins University, Circulars; JHUC, Johns Hopkins University Collection; JP, Journal de physique theorique et appliquee; JRE, Jahrbuch der RadlOakllritdt und Elektromk; PA, American Academy of Ans and Sciences, Proceedmgs; PM, Philosophical magazme; PRSL, Royal Society, London, Proceedings; PTRS, Royal Society, London, Philosophical transactions; PZ, Physikalische Zellschrifi; RP: Rowland papers, MS6, SpecIal Collec- tions, The Milton S. Eisenhower Library, Johns Hopkins University, Baltimore, MA; TC, Technology and culture; ViA, VIStas in astronom}", Zp, ZellschTifi far PhYSIk; ZwPh, ZeilSchrifi far wissenschafillche Photographle. I. J. Fraunhofer, "Bestimmung des Brechungs- und Farbenzerstreuungs-Vermogens verschiedener Glasanen, in Bezug auf die Vervollkommung achromahscher Fernroh- reI!)," Akademie der Wissenschaften, Munich, DenkschTifien, 5 (1814/15). 193-226, esp. plate II. HSPS. 23:2 (1993) 220 HENTSCHEL of glass in order to improve his achromatic telescopes; a precise deter· mination of the angle of deflection for one of these dark lines would measure their dispersion. The merely useful phenomenon became a field of research of its own after Gustav Kirchhoff and Robert Bunsen showed that each element has a characteristic spectrum in the Bunsen flame. A further acceleration of instrument development followed tpe birth of the disciplines of spectral analysis and spectroscopy, tpe former looking specifically for any interdependency of the chemical nature of light emitting or absorbing bodies and their spectra,z tbe latter measuring and further analyzing these spectra, especially with respect to series characteristics.3 Progress in spectroscopy in the second half of the 19th century rested on the improvement of diffraction gratings. 4 David Rittenhouse's primitive grating of 1785 and Fraunhofer's first grating of 1821 were made of wires stretched parallel to each other in the notches of a long screw.s Later, Fraunhofer introduced the technique of scratching paraliel lines with a diamond point onto a glass surface coated with a very thin layer of gold.6 This technique was consider- ably improved and extended to the scratching of metal surfaces, which gave rise to diffraction spectra by reflection. The much higher regular- ity of ruled metal over ruled glass surfaces gave metal the advantage despite the low intensity of light reflected from them. Every introduc- tion of technological improvements into the art of ruling gratings caused a jump in the spectroscopic resolution )"jlJ),,: converted into today's units, measurements made before 1850 had 0 )" = I A, Nobert's carefully ruled glass gratings led to an tenfold increase in resolution around 1860 (lJ )" = 0.1 A ), Rutherfurd's gratings improved it once again by a factor of 2 to 5 in the late 1860s and 18705, and Rowland's concave gratings allowed for 0 A < 0.01 A. 2. Frank A.J.L James, "The establishment ofspectro-<:hemica/ analysis as a practical method of qualitative analysis, 1854-1861," Ambix, 30(1983), 3D-53. 3. Cf. Heinrich Konen, entries "Spektralanalyse" and "Spektroskopie," in Handw6rterbuch der Naturwissenschafien, 1 st ed. (lena, 1913) 9, 205-214, 222-251, esp. 223, for a general characterization of the aims of both disciplines. 4. For an overview of the development of technology for the ruling of gratings as well as scratching into surfaces for other purposes before Rowland, see Deborah Jean Warner, "Rowland's gratings: Contemporary technology," ViA, 29 (1986), 125-130, on 125f. and references therein. S. Cf. Th.D. Cope, "The Rittenhouse diffraction grating," Franklin Institute, Journal, 214 (1932), 99-104, resp. Fraunhofer (ref. I). Fraunhofer's grating consisted of 260 parallel wires. 6. Cf. David Brewster as quoted in Warner (ref. 4), on 126 for his acknowledgment of Fraunhofer's "superior powers and means of investigation." REDSHIFT 221 Table I Development of spectroscopes in the 19th centurl Name Year Instrument Size (em) No. of Lines L.ne separallon (an) Fraunbofer 1814 Prism and telescope Fraunbofer 1821 First d.ffracllon grarins; 3,600 0.0048 parallel wires; later scratebes .n a sold-coated sJass plate Noben 1851 D.amond scratcbes in glass ca 2.5 ca 1.000 0.0025 SlelDbe.1 1860 F1lDt glass pnsms, 4-6 in sequence (On,che.ner) 1864 G rallng by N oben 1.38 3.000 0.00046 (Angstrom) 1864 G ratIng by N obert ca 2 ca 4,500 0.00046 RUlberfurd 1868 Grating 10 glass and metal ca 5 ca 20,000 000025 RUlberfurd 1881 Grallng.n metal for MendenbaU 4.44.4 ca 30,000 0.00015 Rowland 1882 Concave grating ca 7.2 ca 45,000 0.00017 Rowland 1896 Concave grallng ca 14.5 ca 110,000 0.00015 Mlcbelson 1907 Concave grallng 22·11 ca 110,000 0.00010 a. Names in parentbeses stand for observers wbo d.d nOt manufacture tbelr own spectro- scopes but obtained tbem from otber instrument makers. The table shows that with Rowland the total number of lines in the surface of the grating increased while the distance of neighboring lines was nearly constant, though its regularity improved. Because the highest resolution of a grating Ala A is equal to the product of total number of lines and the order of interference m, and since Rowland's gratings allowed observation in the third and even fourth order of the spectrum, his gratings obtained resolutions of up to 400,000, which was then considered as the practical limit of resolution for precision spectroscopy.7 Prism spectrographs and the earlier gratings could resolve at best two spectral lines distant by 1/40th the separation between two sodium D-lines; Rowland's gratings had a resolution 2.5 times as great. The instrument revolution in spectroscopy achieved by Rowland was immense, "a new departure in spectrum-analysis," "one of the greatest inventions ever made in spectroscopy."8 In the 20th century, the continued use of refined gratings, for example, by Zeeman at Leiden and Kayser at Bonn, was supplemented increasingly by interferometric methods, as produced among others by Michelson, Fabry, Perot, and Benoit.9 Thus, grosso modo, we have three phases of measurement in spectroscopy: 7. Konen (ref. 3), 227. 8. Quotes from, resp., "President's address" (p. 48 I) in H.A. Rowland, "Remarks on the award of the Rumford Medals," PA, 19 (1883/84), 482-483, and Edward Charles Cyrill Baly, Spectroscopy, 3rd ed. (London, 1929). 1,28. 9. See A.A. Michelson, "On the application of mterference methods to speCtroSCOPIC measurements," PM, 30 (189 I), 338-346, and 34 (1892), 280-299; cf. Michelson, Benoit, Fabry and Perot (ref. 155), Weber (ref. 169). 222 HENTSCHEL l. Prisms (-1823, later only for special applications such as stellar spec- troscopy) 2. Gratings (1823-1906, and later as a supplement to other methods) 3. Interferometers (from 1895 on, more frequently from 1901 on, espe- cially for the definition of a primary and a few secondary standard wave lengths) with resolutions of up to 800,000. The subject of this paper, the discovery of the redshift of Fraunhofer lines in the sun's spectrum, took place at the height of precision spec- troscopy dominated by Rowland's gratings. I. ROWLAND'S REVOLUTIONS Henry Augustus Rowland (1848-1901) graduated as a civil engineer from the Rensselaer Polytechnic Institute at Troy, New York, in 1870. 10 He became instructor in physics in 1872 and assistant pro- fessor of physics in 1874." In 1875, he became the first professor of physics at the newly founded Johns Hopkins University, a position which he held until his premature death in 190 I. For the planning of his laboratory, he visited James Clark Maxwell in England and Her- mann von Helmholtz in Berlin, working in the latter's laboratories for four months.12 With the instruments and equipment he bought for more than $6,000, his laboratory became the best-equipped in the United States, and attracted many students. Furthermore, a well- equipped workshop in which new apparatus could be produced was closely connected with his laboratory. Despite reports of some deficiencies as a pedagogue, between 1879 and 190 I Rowland had 165 graduate students and 45 Ph.D. students, thirty of whom received stars in American men of science. 13 Rowland's widest influence probably came through his gratings, which were used by all the important spectroscopists of the late 19th 10. About Rowland cf. obituaries and biographical sketches by R.T.G. 10 Nature. 64 (1901), 16-17; T.e. Mendenhall, BMNAS, 5 (1905), 115-140; 1.5. Ames, SCIence, N.S., 13 (1901), 681-684, and DAB. 16 (1935), 198-199; H.F. Reid, APJ. 28 (1941), 117-119; H. Crew, AJP, 17 (1949),576-577, D.l. Kevles, DSB, 11112 (1981), 577-579; A.D. Moore, Scientific American, 246 (1982), 118-126; S. Rezneck, "The educaUon of an American physicist-Henry August Rowland," AJP. 28 (1960), 155-162. II. According to Biographical Record of the Officers and Graduates of the Rensselaer Polytechfliclnstitute 1824-1886 (Troy, 1887), 115, 117, 164. 12. S. Rezneck, "An American physicist's year in Europe, Henry Rowland," AJP, 30 (1962), 877-886; John David Miller, "Rowland and the nature of electric currents." Isis, 63 (1972), 5-27. 13. RC)bert Hugh Kargon, "Henry Rowland and the physics discipline 10 America," ViA, 29 (986), 131-136; cf. "List of scientific apparatus," Harvard University, Llbrar), Bulletin. 11-12, 302-304, 350-353, esp. 351-353, for the Baltimore Physical laborato- ry. REDSHIFT 223 and early 20th centuries: 14 Henri Deslandres and Alfred Cornu (who later worked on band spectra using Rowland's gratings), Pieter Zee- man (who employed Rowland's gratings in the experiments that led to the discovery of the influence of magnetic fields on spectra in 1896),15 Carl Runge, August Kundt, the Vogel brothers, Heinrich Kayser, and Friedrich Paschen,'6 Janne Rydberg (Lund),17 Arthur Schuster (Man- chester), George Higgs (Liverpool), several investigators at Cam- bridge,18 and one at the Royal University of Ireland. 19 George E. Hale used a Rowland grating from his early days at the Kenwood Observa- tory.20 Charles Edward St. John, Arthur Scott King and their col- leagues used one later at the Mount Wilson Observatory, Frank Wads- worth and others had one specially ruled for the Yerkes and Allegheny Observatory,21 W.F. Meggers, K. Bums and others had theirs at the National Bureau of Standards, Washington, and, of course, Albert A. Michelson got one for Chicago. By 1895, Rowland had sold more than 100 of his gratings to spec- troscopists all over the world at a price determined by production costS.22 By January 1901, sales totaled more than $13,000, which represents between 250 and 300 gratings sold at cost to physical and chemical laboratories as well as to astronomical observatories all over the world, not counting those gratings given away for free. 14. William McGucken, Nmeteenth-century spectroscopy. Del'elopment of the under· standing of spectra (Baltimore, 1969), esp. 135. IS. Pieter Zeeman, Researches in magneto-opllcs (London, 1913), on 9ff. 16. Roben Bezler, "Zur Geschichte des grossen Rowland-Gltters am Physlkahschen Institut der Universitat Tiibingen," Bausteme zur Tabinger Vn/l'ersllatsgeschlchte, 3 (1987),141-178. 17. See J.R. Rydberg, "On a cenain asymmetry in Prof. Rowland's concave grat- ings," PM, (5) 35 (1893),190-199, also in AA, 12,439-448. 18. Catalogue 5 of the Whipple Museum of the History of SCience lists Rowland grat- ings in the former possession of the Cavendish laboratory, the Institute of Astronomy, the Depanment of Physical Chemistry and of R.S. Whipple. 19. W.E. Adeney and J. Carson, "On the mounting of the large Rowland spectrome- ter in the Royal University ofireland," PM (5),46 (1898),223-227. 20. Horace W. Babcock, "Diffraction gratings at the Mount Wilson Observatory," ViA, 29 (1986), 153-174, and Physics today, 39 (1986), 34-42, on 34, reponing that Hale later transferred his plane Rowland grating to the Mount Wilson Observatory. 21. Cf. Frank L.O. Wadswonh, "On the aberration of the concave grating, when viewed as an objective spectroscope," PM, 6 (1903), 119-156, on 121. 22. See Miller (ref. 12), n. 2, Kevles (ref. 10), 579, Warner (ref. 4), 129; the manufac- ture of the gratings and quality testing was mostly carried out by Lewis E. Jewell, while distribution was left in the hands of Brashear. Cf. John Alfred Brashear: The autoblOgra· phy of a man who loved the stars, ed. W. Lucien Scaife (Boston, 1925), 76, for funher customers of Rowland gratings. 224 HENTSCHEL On receiving the gold and silver Rumford medals from the Ameri- can Academy of Arts and Sciences in 1883/4, he disclosed how his interest in ruling gratings arose:23 My attention was first called to the construction of dividing-engines by an inspection of a dividing-engine constructed by Professor W.A. Rogers, at Waltham, in this State [Massachusetts]. On returning to Bal- timore, I devoted much time to the general problem of such machines; and, through the liberality of the trustees of the Johns Hopkins Univer- sity, I was enabled to construct an engine. Ruling technology Like his predecessors, Rowland employed a "ruling engine" (figure I), which guided a sharp, carefully chosen and mounted diamond point over a coated glass plate or metal surface.24 After one grating line was ruled, a mechanism raised the diamond point and shifted the grating surface a short distance, whereupon the diamond ruled the next line. The straightness of the lines was easily achieved by guiding the diamond along two parallel metallic rails; the positioning of the diamond point after each ruling by exactly the same distance, no more than 0.00015 em in Rowland's machine, constituted a greater chal- lenge. From studies carried out by William August Rogers (1832-1898) at the Harvard College Observatory, Rowland knew about the main sources of error in the gratings of Friedrich Adolph Nobert (1806- 1907) and Lewis Morris Rutherfurd (1816-1892).25 According to Rogers, even the best diffraction gratings of his time were subject to three classes of errors:26 23. Rowland (ref. 8), 482. 24. Hugo Schroeder, "Ueber die Verwendung des Diamanten in der PriizislOns- Mechanik," Zellschrift fur Instrumentenkunde, 7 (1887),261-269,339-347; J.S. Ames, "Henry August Rowland," Johns Hopkms University, Alumm magazine, Jan 1916,92- 99, on 96. Rowland's assistant Jewell became his expert for the choosing and purchase of diamonds, usually bought at Tiffany's, New York. 25. Cf. Edward W. Morley, "Memoir of William August Rogers," BMNAS, 4 (1899), 187-199; W. Rollmann, "Friedrich Adolph Nobert," Naturwissenschafthcher Vereine van Nue-Vorpommern und Rugen im Greifswald, Mltthellungen, 15 (1884), 38-58; G.LE. Turner, "The contributions to science of F.A. Nobert," Institute of Physics, Bul· letin. 18 (1967), 338-348; B.A. Gould, "Memoir of Lewis Morris Rutherfurd," BMNAS. 3 (1895),417-441. 26. William A. Rogers, "On the first results from a new diffraction ruling engine," AJS 19 (1880), 54-59, on 54; Morley (ref. 25); D.J. Warner, "Lewis M. Rutherfurd: Pioneer astronomical photographer and spectroscopist." Te, 12 (1971), 190-216. on 214f. REDSHIFT 225 1. accidental errors of single subdivisions, mainly due to the irregular motion of the ruling diamond upon a non-homogenous metal 2. systematic, more precisely periodic errors, being a function of one revolution of the ruling screw 3. errors dependent upon the position of the nut upon the screw, equivalent to a varying pitch The most surprising error was the periodic error in the separation of the lines; Rogers proved its existence in Nobert's, Rutherfurd's, and his own gratings and showed it to be of the order of magnitude of 1120,000 inch.27 He attributed the unwanted periodicity not to the screw itself but to its mounting, but wherever it came from, periodic errors in diffraction gratings caused trouble. They caused the appear- ance of additional faint lines in the spectrum, the so-called "ghosts," which could easily be (and often were!) mistaken for true lines. The versatile mathematician, scientist, and philosopher Charles Sanders Peirce (1839-1914) confirmed the suspicion that the periodic errors in the line separation f caused ghosts through a rigorous mathematical analysis and subsequent experimental checks using vari- ous Rutherfurd gratings.28 Rowland's first challenge was to eliminate this type of error. Another troublesome factor was irregular variation of f, which, according to Rowland, resulted in a general loss in sharp- ness of the spectral image. But Rowland was quite sure that his grat- ings were too precise for ghosts haunting other experimenters: "The ghosts are very weak in most of my gratings. "29 In any case, he thought he could distinguish between artifacts of a badly manufac- tured grating and real lines. We learn more in an article about screws that Rowland wrote for the 9th edition of the Encyloptfdia Bri/an- nica.30 According to this text, "ghost lines" tended to change their positions relative to other lines, while true lines only changed in their scale, but not in their relative positions, under a certain adjustment of his spectrometer. Apparently, this way of discriminating the "good" 27. W.A. Rogers, "On a possible explanation of the method employed by Nobert In ruling his test plates," PA, 11 (1875), 237-255, on 243 for a description of the method employed to measure directly the magnitude of the periodIC error. Rogers preferred the Nobert grating over Rutherfurd's. 28. C.S. Peirce, "Note on the progress of experiments for comparing a wave length with a metre," AJS, 18 (1879),51; H.A. Rowland, "On concave gratings for optical pur- poses," PM, 16 (1883), 197-210, on 198, citIng Peirce's more detaIled paper In the American journal of mathematics, 1879; A.A. Michelson, "On the spectra of Imperfect gratings," APJ, 18(1903), 278-286. 29. H.A. Rowland, "A few notes on the use of gratings," JHUC. 8 (1889),73-74. 30. H.A. Rowland, "Screws," Encyciopadla BTIlanmca (9th edn.), 21 (1884), 552- 553, reprinted in The phySical papers of Henry Augustus Rowland (Baltimore, 1902), 506-511. 226 HENTSCHEL from the "bad" lines worked better than the old method of distin- guishing between them, namely changing the order of the lines observed. Ghosts revealed themselves by appearing in all orders of the spectrum:3l They [the ghosts] never cause any trouble, as they are easily recognized and never appear in the solar spectrum. In some cases the higher orders of ghosts are quite as apparent as those of the first order .... Hence, to avoid them, obtain magnification by increasing the focal distances instead of going to the higher orders. These are rare hints. Observation routines were seldom mentioned in textbooks or scientific articles but only acquired through practical work in the laboratory under the supervision of a skillful teacher. 32 In all ruling machines made after Nobert, the transport of the grat- ing under the ruling diamond point was activated by the turning of a screw by an exact number of degrees.33 Rowland realized that special care had to be invested in the design and production of the screw and in its proper installation (with the aid of many adjustment screws) to get the necessary degree of uniformity of the line separations. 34 The screws, up to 25 cm (10 inches) long, were made out of special flawless steel in a painstaking process that could take up to 14 days. The ra .... screw, held in a nut, was then continuously tightened in a process that could take up to another 14 days, again under conditions of constant temperature with continuous removal of friction heat through liquid grinding materials (such as emery powder and oil or optical rouge) and with a regular switch of the direction of the screw in the nut eve!") ten minutes throughout the whole procedure to avoid any asymmetries in the screw driving. Rowland claimed that for a screw produced in this manner, "there was not an error of half a wave-length, although the screw was nine inches long," indeed a remarkable precision for a mechanically produced object. 35 31. Rowland, Papers (ref. 30), on 519. 32. Cf. Ames (ref. 24), on 97: "I have seen Rowland stand by the machine wllh a screw driver in his hand looking at the specimen of ruling and then say 'I think I'll tI> this.' Then he would poke his screw driver in, doing something which would be Impos· sible for anyone else to understand clearly; and the chances were that after one or two such attacks on the machine it would work all right. When I would ask him what he had done, and why he had done it, he was never able to explain fully. The truth was that hIS knowledge of machines of all kinds was in part a process of instinct." 33. Nobert's ruling engine was still based on a large circle divider. 34. For a survey of the contemporary technology of screw production, see Charles and John Jacob Holtzapfel, Turning and mechanical manipulatzon (5 vols., London 1881), 4, reprinted as J.J. Holtzapfel, Hand or simple turning: Principles and practice (New York, 1976). esp. chapt. 10. 35. Rowland (ref. 8). on 482. REDSHIFT 227 Rowland's screws typically had 20 threads to the inch, and were turned at a constant angle (1/720 of a full circle) by a toothed wheel; the grating thus moved about 1120· 11720 = 1114400 inch after each line was ruled. J6 A motor run by water power (because of its greater regularity and higher reliability than electric power), drove the pro- cess, which could take up to 14 days. Every effort was made to keep the room temperature constant, because even minute fluctuations might spoil the gratingY To Lord Rayleigh (as to many others), he wrote in March 1882 after realizing that his ruling engine really worked well and reliably:J8 I have just completed in our workshop a machine for ruling gratings and it is a great success, among gratings fully equal if not superior to Rutherfurd's and of larger size .... Rutherfurd could only make one good grating out of many, but my machine makes them as good as his best every time. Although he exaggerated-Jewell later reported the large number of wasted rulings eliminated after the quality test he conducted- Rowland rightly claimed the superiority of his gratings over those of Rutherfurd and other contemporaries. Their excellence was soon ac- knowledged by the scientific community.J9 Rowland's gratings thus excelled because of this unprecedented care in the manufacture of the screw and in its careful mounting, which Rowland found "even more difficult to make without error than the screw itself."4O His ruling engines were able to rule gratings of up to 110,000 lines into a metal surface with an accuracy of one mil- lionth of a millimeterY Although others soon tried to improve on Rowland,42 the only person to better his performance was Rowland himself. He built three ruling engines altogether, the first one, com- pleted the autumn of 1881, could rule 14,438 lines per inch; the 36. This line distance was considered by him to be the practical limit "with the ordi- nary conditions of ruling;" see ibid. 37. T.e. Mendenhall, "On the determination of the coefficient of expansIOn of a diffraction grating by means of the spectrum," AJS, 21 (1881),230-232. 38. Rowland to Lord Rayleigh, 6 Mar 1882, as quoted \0 Strutt I 36f. and in Warner (ref. 4), 128 (emphasis orig.); cf. Warner (ref. 26), 215. 39. Lord Rayleigh, "President's address," British Association for the Advancement of Science, Report 1884, 17: "the magnificent gratings of Rowland are a new power 10 the hands of the spectroscopists, and as triumphs of mechanical art seem to be little short of perfection." 40. H.A. Rowland, "Preliminary notice of the results accomphshed in the manufac- ture and theory of gratings for optical purposes," PM. 13 (1882), 469-474, on 471. 41. Cf. H. Kayser, Erinnerungen aus memem Leben, unpubl. typescnpt (1936). 187. 42. E.g., Ertel, Fraunhofer's successor in Munich, Thomas Grubb 10 Dubhn. and Adam Hilger in London; see Warner (ref. 4) for references. 228 HENTSCHEL second and third, of 1889 and 1894, ruled 20,000 and 15,020 lines per inch, respectively, over as much as 25 square inches, and each incor- porated improvements in the adjustment mechanism of the main screw and the carriage of the ruling diamond_ 43 Rowland used to say that "No mechanism operates perfectly-its design must make up for imperfections."44 An indication of the prac- tical application of this dictum appears from a description and evalua- tion of the carriage system given by one of Rowland's direct descen- dants in the grating art, John Strong:45 The grating grooves are ruled on the grating blank by repeated, straight- line strokes of a diamond point-a point guided by a carriage that spanned the blank. It is carried on divided cross-ways; guided by one sliding shoe on the right side of a rectangular-bar at one end of the car- riage, together with a second shoe on the other end of the carriage, bear- ing on the left side of another rectangular-bar, aligned and parallel. After Rowland's death, Professor Joseph Ames suggested to John Anderson that the shoes might better slide on the same side, right or left; Anderson never followed up the suggestion, because he realized the intricate compensation of imperfections granted by the symmetric arrangement. Strong again:46 In Rowland's arrangement, using opposite sides, the motion of the dia- mond midway between the two shoes becomes immune to lateral shifts due to the lubricating oil thickness, as long as the variations of the OIl film during the ruling stroke are equal. And the arrangement also makes the motion immune to wear. This clever built-in stability under unavoidable variations of the oil film and wear enabled Rowland to claim that the diamond point repeated a straight line stroke to within the incredible precision of half a wavelength of visible light. A similar remark applies to Rowland's overcoming of the mechanical difficulties in ruling on concave surfaces by allowing for a judicious tipping of the spherical blank on its car- riage resulting in a nearly invariant angle between the diamond and the ruled surface.47 43. See Rowland, Papers (ref. 30), Appendix; Cf. the detailed repon by J.S. Ames. "The present condition of Rowland's ruling machines," JHUC, 4 (1906), 62-65. Kevles (ref. 10), 578 even repons on a machine capable of ruling up to 43,000 lines per Inch but I could not find independent confirmation for this value. 44. Rowland, as quoted in John Strong, "Rowland's diffraction-gratIng an," VIA. 29 (1986), 137-142, on 137. 45. Ibid. 46. Ibid. Cf., however, Joh. Adolf Repsold, Zur Geschichte der Astronomlschen Messwerkzeuge (2 vols., Leipzig, 1908-14), 2, 140f, for critical remarks about the prismatic form of the guiding shoes. 47. Cf. Strong (ref. 44), 141; Ames, "A description of the dividing engines deSIgned by Professor Rowland," in Rowland, Papers (ref. 30),691-697, and plates 1-5. FI G. I Tr an sv er se s ec tio na l el ev at io n v ie w o f R ow la nd 's ru lin g en gi ne , sh ow in g fe ed -s cr ew (1 2), n u t (15 ), a dju - st ab le d ia m on d ho ld er (2 ) w ith d ia m on d po in t ( 1) , p la te to b e ru le d (6) a n d th e in tri ca te c o rr e c to r fr am e m ec ha ni cs . A m es (r ef. 4 7), p la te 5 . '" m 51 :r: ~ N N \0 230 HENTSCHEL FIG. 2 Rowland in front of his first (and smallest) ruling engine. Rowland (ref. 30), 43, Appendix. Rowland did not make public his procedures for testing his gratings and ruling engines. In this silence he copied what he believed was Nobert 's policy of keeping testing a "trade secret."48 In fact , Norbert had lifted the veil slightly in an obscure Prussian journal in 1845. but not enough to allow his contemporaries to reproduce his rulings easily.49 The same applies to Rowland's direct predecessor in the manufacturing of diffraction gratings, Lewis Morris Rutherfurd :50 48. Rogers (ref. 26), 238 : "Nobert has well kept the secret of his process;" John May- all , "Nobert 's ruling machine," Royal Society of Arts, Journal, 33 (1885). 707- 71 5: Mendenhall (ref. 10), 124. For Rutherfurd's techniques, see L.M. Rutherfurd. "On the construction of the spectroscope," AJS, 39 (1865), 129-132, and A.M. Mayer, " Spec- trum," Appleton's Cyclopedia (2nd edn.) , 15 (1878), 238-254, 243f; for Rowland 's, H.A. Rowland, " Preliminary notice on the results accomplished in the manufacture and theory of gratings for optical purposes," JHUC, 17 (1882), 248-249. also in PM. 13. (1882), 469-474, and in Nature, 26, (1882), 211-213, and J.S. Ames, "The concave grating in theory and practice," JHUC, 73 (1889), also in PM, 27, (1889), 369-384. 49. F.A. Nobert, "Ueber Kreistheilung im Allgemeinen und iiber einige. bei ei ner Kreistheilmaschine angewendete Verfahren zur Erziehung einer grossen Vollkom- menheit der Theilung derselben," Verein zur Bef6rderung des Gewerbefleisses in Preussen, Verhandlungen , (1845), 202-212; W. Rollmann (ref. 25), on 53; Rogers (ref. 26), 237 : " You properly ask me if I can reproduce these rulings. I frankly answer that I cannot." 50. Cf. B.A. Gould (ref. 25); Warner (ref. 26). - Iti ~ / . FI G . 3 V ie w o f R ow la nd ' s ru lin g e n gi ne , sh ow in g a ga in t he d ia m on d po in t ho ld er (2 ), its c a rr ia ge (4 ), th e di vi de d c ro ss w a y (5) , a n d th e gr at in g bl an k (6) . A m es (r ef. 47 ), pl at e II. '" m o V l :: t ~ IV ...., 232 HENTSCHEL Rowland himself remarked that "many mechanics in [America] and in France and Germany have sought to equal Mr Rutherfurd's gratings. but without success."51 Even if Rowland had wanted to transmit his methods, he could have done so only to someone with whom he was in daily practical collaboration. The only one who qualified was his mechanician Theodore C. Schneider, who died in the same year that Rowland did. The Baltimore Sun of April 17, 190 I, commented:52 A question which is being asked since the announcement of the death of Professor Rowland is, Will his art of making microscopically fine grat- ings on a concave surface for spectroscopes be lost with him, or was hiS work left in such conditions that other scientists will be able to take it up where he left off and continue to furnish the gratings which have practically revolutionized the art of spectroscopic analysis? Dr. Remsen said last night that he is sure the art died with Professor Rowland. About a month ago Mr. Theodore C. Schneider, the mechanic who was trained by Professor Rowland and who was the only man besides him who could construct the machine, died. Since Mr. Schneider's death Mr. Charles Childs, who had long been associated with Professor Rowland, had been learning the art. Mr. Childs has made rapid progress, but it is thought by those who are in a position to know that he has not yet gotten to the point where he can count upon success- fully carrying on the work without having the master mind to direct him. Mr. Schneider, skilled man that he was and working in harmony with the ideas of the inventor, would occasionally strike obstacles which were entirely beyond the range of mechanical skill and required the closest application of pure theoretical reasoning before the way could be discovered for the resumption of operations. It took about ten years, before Rowland's work on the ruling of grat- ings was taken up by John Anderson.53 Schneider did manage to transmit some information about Rowland's techniques to Heinrich Kayser, who tried to visit Rowland in Baltimore during his first trip to America. Rowland was out of 51. Rowland (ref. 40); cf. Henry Draper, "On diffraction spectrum photography." AJS, 6 (1873), 401-409, also in PM, 46, 417-425, and C.A. Young, "Note on the use of a diffraction grating as a substitute for the train of prism in a solar spectroscope," AlS. 5 (1873), 472-473, on 472: "the spectra furnished by these plates far exceed ID brilli- ance and definition anything of the kind ever before obtained." 52. "Death of Prof. Rowland," The Sun [Baltimore), 17 Apr 1901, p. 4, col. 2. "None to fill place. Science suffers irreparable loss in death of Prof. Rowland. Ho" great physicist died," ibid., p. 12, col. 1-3; and "Great men mourn," ibid., 4-5. 53. See H.D. and H.W. Babcock, "The ruling of diffraction gratings at the Mount Wilson Observatory," Optical Society of America, Journal, 41 (1951), 776-786; Strong (ref. 44), Babcock (ref. 20); R.F. Jarrell, "Gratings, production of," Encyclopedza of spec· troscopy (New York, 1960), 174. REDSHIFT 233 town. Schneider, however, was at the laboratory. Inspired by their common language (the mechanician was of German descent) and an immediate sympathy, Schneider told Kayser some of the tricks of the trade, at which he had been working since 1876 (Rowland knew Schneider from his student years at Rensselaer). Schneider had built the ruling engine and oversaw the manufacture of the gratings. 54 John Brashear produced and polished the spherically curved grating sur- faces onto which Schneider ruled the lines. 55 Among the details that Kayser learned from Schneider was the method of adjusting the ruling screw: 56 Schneider showed me all the installations necessary to manufacture the screw, the critical part, and he explained with admiration, how Rowland adjusted the setting of the screw. Its end must be exactly positioned to a millionth of a millimeter, which is achieved by a number of fine adjust- ment screws. The procedure is as follows: on a test plate, a certain number of lines is ruled with the ruling engine, say one thousand, then the grating surface is turned a bit, and now again about one thousand lines are ruled across the first set of lines. If the main screw is correctly adjusted, and thus the distances between the lines are everywhere the same insofar as they depend upon the position of the screw, then the points of intersection of both systems of parallels are on straight hnes too. But if the distances slightly vary, then the points of intersection fol- Iowa curve, one sees some form of moire. Anyone can see this and then knows that the adjustment of the ruling screw is not yet perfect. Row- land, however, looks at this moire for a few minutes, then he says: "Tum adjustment screw A by about one twentieth of a full tum to the right, screw B by about one tenth to the left. You might also tum screw D by about one fiftieth to the right." And then often the adjustment is complete. How he is able to decipher this from the moire is hard to grasp. Rowland and his collaborators had already reached such a level of refinement, that spectroscopists who tried to reproduce his results and did not have the privilege of having been one of his pupils, occasion- ally had serious trouble. Even the "master of light," Albert Abraham Michelson, who was attracted to the technological challenge of the rul- ing engine around 1904, later regretted "ever having got this bear by the tail."51 Here is what Johannes Hartmann of the Potsdam Observa- tory wrote in 1903 about Rowland: 58 54. Kayser (ref. 4 I), 190; Ames (ref. 43), 62. 55. Mendenhall (ref. 10), 125; Warner (ref. 4), 129; J.A. Brashear (ref. 22). 56. Kayser (ref. 4 I), 190f. 51. See Babcock (ref. 29), 154; cf. A.A. Michelson, "The ruling and performance of a ten inch diffraction grating," American Philosophical Society, Proceedmgs. 54 (19 I 5). 137-142. 58. J. Hartmann, "A reVlSlon of Rowland's system of wave-lengths," APJ. /8 (\903), 234 HENTSCHEL How Rowland obtained the screw value [in his readings of wavelengths] with sufficient accuracy for such long distances is not to be readily ascertained from his publications, which, indeed, contain so few data as to the measurements themselves that a test of them is impossible. Fig. 3. F,g. 5. AG. 4 "Moires" produced by the superposition of two rulings with a slight systematic error of one-sided increase of intervals amounting to 11240 mm. A. Cornu, "Sur les diverses methodes relatives a I'observation des proprietes appelees anomalies focales des reseaux diffringentes," eRAS, 116 (1893). 1421-1428, on 1426. Concave gratings During the winter of 1882/3, Rowland spent much time trying to account for the influence of irregular variation of the line distance in his gratings. Surprisingly, many gratings proved much more reliable than the theoretical estimates given by Peirce and himself had led him to expect, and the quality of the gratings often differed for different sections of the spectrum. Thomas Young, in a comment Rowland knew of, had observed that often the quality and sharpness of the image of a grating can be improved considerably by slightly bending its surface. 59 Perhaps the unavoidable distortions in the images of plane gratings might have been compensated by a slight curvature? If 167-190, on 169; cf. Konen (ref. 76), 792. 59. Rowland (ref. 29), 199. REDSHIFT 235 so, the flaw could be made a virtue by ruling gratings on curved sur- faces. 6o A further consideration may have strengthened Rowland's interest in following this direction. Whether spectroscopists used prisms or plane gratings, they had to employ at least two lenses to get sharp images: the so-called collimator, essentially to guide the light directly onto the prism or grating surface, and a second lens to project the diffracted light for observation or photography. Both lenses had to be corrected; the collimator for achromatism, the camera lens with its large focal plane for spherical aberration.61 Even in the setup devised by Littrow,62 which employed the same lens as collimator and projec- tor, there were serious disadvantages. The lenses absorbed parts of the spectrum, in particular in the infrared and ultraviolet regions: the dispersion of the glass varied too much to produce a well-scaled image: and other imperfections of the lenses gave rise to further distortions. These blemishes menaced high precision spectrometry. Accordingly, around 1882 Rowland broke with the practice of several generations of spectroscopists and decided to eliminate all lenses. A spectroscopic grating ruled on a concave surface works like a burning mirror, thus eliminating the problem of projection of the image of the spectrum. The collimator too became superfluous for a concave grating, since a simple slit, wide enough to allow the light from the source to illuminate the whole surface of the grating. was sufficient. Slit, concave grating, photographic plate, and a stable mounting were all Rowland needed for his new spectroscope. Rowland soon realized the optimal geometrical configuration of the three basic components of his apparatus: the slit (functioning as the source of light in the geometric analysis of the apparatus). the grating. and the photographic plate should all be installed on a large circle of diameter equal to the radius of curvature of the spherical grating.63 This insight was based on a detailed mathematical analysis of the optical charac- teristics of his new instrument. reported to the London Physical Society in 1882, which proved that automatic focusing occurred 60. Anon (ref. 52); Ames (ref. 24),95; Strong (ref. 44), I 38ff. 61. In spectrographs, the difference in the focal length for the different parts of the spectrum was compensated for by slIghtly bending the photo plates. hence the use of such thin glass plates (Jost Lemmench, personal commUnication). 62. Cf., C.F. Brackett, "Note on the Littrow form of spectroscope," AJS, 24 (1882). 60-61; Konen (ref. 3),223,226. 63. Rowland (ref. 28); W. Baily, "On the spectra formed by curved diffraction grat- ings," PM 15 (1883), 183-187; R.T. Glazebrook, "On curved diffraction-gratings," P.\f 15 (1883), 414-423; E. Mascart, "Sur les Tt!saux metalliques de M. H.-A. Rowland," Journal de phYSique, 2 (1883), 5-11; and Wadsworth, 'The modem spectroscope," APJ. 3 (1896),47-62, esp. 54-60. 236 HENTSCHEL FIG. 5 Photograph of a six-inch concave grating, especially manufactured for Zeeman by Rowland (central part of the upper circle) including adjustment screws and a turning lathe support. Zeeman (ref. 15), 10. whenever the slit, the grating with spherical curvature radius R, and the photographic plate resided on a circle of radius R12, and that the best optical image would be obtained if grating and photographic plate were nearly opposite one another. Rowland chose the following proportions for his spectroscopes: • a spherical radius of curvature of 21.6 feet (10.8 feet for the early gratings), which determined the size of the whole apparatus • a diameter of the spherically concave grating surface of up to 6 inches • a line density on the gratings of 7,200, later (around 1887) of 14,400 lines per inch and finally (1896) of 20,000 lines per inch • a micrometer run of 5 inches with a precision of 1/20,000 inch. With these dimensions Rowland could arrive at dispersions of one second of arc (=0.0012 inch) so that the sodium doublet had a width of about 4 mm (figure 6). Figure 6 shows a Rowland grating set up for business; notice the solid supports for all pieces of instrumentation. REDSHIFf 237 FIG. 6 A spectrograph in Zeeman's Amsterdam laboratory, with a radius of curvature of 3m and a Rowland concave grating of 4 inches and 52,000 lines. The grating is in the lower right comer of the picture; the slit at the left edge; and, in the center, the segment of the circle with the angle indicator and mounted camera (in front of the second dark surface from the left). The whole apparatus rests upon two steel supporting beams anchored to the walls of the laboratory building for maximum stability. Zeeman (ref. 15), 12. Rowland's geometrical analysis of 1883 had shown that he could photograph the whole spectrum by moving camera M and the grating GA along the circle with the diameter of the radius of curvature AM of the concave grating (see figure 7). The direct reflection of the light coming from the slit S would then show up at the point 0 symmetri- cally mirrored to S across the symmetry axis AM of his apparatus. Diffraction patterns of higher orders would appear in sharp focus in both directions from this point 0 on the circle SGAM . Opposite to the grating GA, near the point M of the circle, the dis- tances between the spectral lines are directly proportional to the wavelengths. Because of the very large diameter of the circle (21 .6 feet!) , this region of linearity, best suited for spectrometric high preci- sion measurements, was about 10 cm long. For a region of about 6 inches, Rowland estimated the obtainable accuracy to be as good as III ,000,000, and for about 18 inches still as high as 1/350,000. Further away from M, proportionality of distances of spectral lines to 238 HENTSCHEL , , FIG. 7 Observations of different orders of diffraction using a Rowland grating GA with radius of spherical curvature AM, light coming from the slit Sand registered in camera M moving along the circle SGAM. The numbers refer to the order of diffraction in the spectrum. Rowland (ref. 28). their wavelength no longer exists, because of the effects of the curva- ture of the circle SGAM, which distorts the images of the spectrum. 64 Because measurements in the second or third order are most reliable (they are still bright enough for clear observation of the spectral lines with a telescope, best suited for fine grained photography, and free of reflections and ghosts that vex in the first order), the slit typically stood somewhere between the grating and the camera (in figure 7, the camera M is in the second order of the spectrum). To cover the whole range of the spectrum, Rowland preferred to move the grating GA together with the camera M along the circle SGAM, keeping the slit S fixed, so that the camera and the grating always opposed each other and thus remained in the region of best approximation to linearity. Rowland's spectrograph and its simple geometry made the routines much easier for him than for his forerunners, who had to repeat the focussing of their images for each exposure anew. No wonder it received his enthusiastic praise: "nothing can exceed the beauty and simplicity of the concave grating when mounted on a movable bar .... Thus the work of days with any other apparatus becomes the work of hours with this."6s Later, some experimenters slightly varied Rowland's original procedure by keeping both the slit S and the grat- ing fixed and moving the camera along the circle. To do so, they had 64. Rowland (ref. 28), 203-204. 65. Ibid., 205. REDSHIFT 239 to use photographic plates bent with a radius of curvature of AM/2. This idea seems to go back to Sir William de Wiveleslie Abney (1844-1920); the first to practice it was Carl Runge (1856-1927); and later Kayser and his pupils employed it in their measurements of wavelength normals in the spectrum of iron.66 Much craftsmanship was required in the actual installation as well as the manufacture of Rowland gratings. The whole apparatus had to be anchored firmly. Kayser reported that the practical utility of the precious Rowland grating in his possession was largely diminished throughout his time in Berlin because of nearby traffic on the street just outside his laboratory:67 So I worked many a night; since it soon became apparent that during the daytime the vibrations in the building were much too strong; but even at night, not even one fifth of the exposures [taken with the Row- land grating) were usable. Because of this, a lot of material and time was wasted, and only when I made a new installation [of the Rowland grating) in my own institute at Hannover a couple of years later, could I earn the fruits of my labor .... [For the installation) a [steel) rail curved in the form of a semicircle with a diameter of 6 m was needed, which had to be manufactured with very high precision, that is, on a turning lathe. In Germany, a turning lathe of this size only existed at Krupp in Essen, where it was needed for armored turrets of tanks. Prof. Fuchs ... had contacts at Krupp, the library of which he supervised, and through him came the request to Krupp to supply such a rail. Krupp not only did this, he also funded the whole installation, which includes a massive cement foundation and other metal pieces, and sent us workers from his factory who installed the whole thing .... Ever since, it was possible to work With the large grating without wasting time, and to obtain accurate measurements. This remarkable statement again demonstrates the close interdepen- dence of measurement technique within science and industrial produc- tion processes. The accurate manufacture of the steel support was cru- cial for making effective use of the precision of Rowland's concave grating, because the slit, camera, and grating had to be on a very accu- rate circle, to guarantee the automatic focussing. Apart from vibra- tions, temperature variation also threatened high precision spectros- copy. Zeeman's new laboratory (built in 1921-23). had a room with a Rowland grating and a special control that held temperature constant to 0.01 degrees Celsius.68 66. Zeeman (ref. 15), 12; Kayser (ref. 41), 235f; H. Konen, "Uber dIe Kruppsche GIt- teraufstellung 1m physlkahschen Instltut der Universltiit Bonn." Z .... Ph. J (1903). 325- 342. 67. Kayser (ref. 41). 120. 235f; see also H. Konen (ref. 66). 68. G.c. Gerrits. "Zeeman." in Grote Nederlanders (Leiden. 1948). "Zeeman." 473- 501. on 480. 240 HENTSCHEL Kayser's unpublished autobiography makes clear the impact of Rowland's new concave gratings:69 I myself had begun to work spectroscopically, namely I wanted to deter- mine and measure spark spectra photographically. Only, I did not make much progress; there were no suitable instruments available. I only had some lenses and prisms at my disposal, and that it is not quite so easy to build a spectroscope with these everyone today knows. Then, in the year 1883 a paper appeared, by Rowland in Baltimore, in which he reported on the production and performance of his concave gratings, which had brought about the rapid developments within spectroscopic measurement. I showed this paper to Helmholtz and asked him whether he could not obtain such a grating through Rowland for the Physical Institute. Helmholtz said that I myself should write to Rowland in his name. Rowland, who had formerly worked under Helmholtz, responded in the friendliest manner to the request, and sent us a grating as a gift. We see here that Rowland was not only central as the inventor of the new type of grating, but also as its actual producer, through Brashear, his gratings were sent to institutes for chemistry, physics, astronomy. and astrophysics all over the world. 70 Owing to the small output of Rowland's shop of about 10 gratings per year, the possession of an authentic Rowland grating became one of the hallmarks of the quality of an institute. 71 Even among these gratings, a hierarchy of quality existed. based on comparisons made by Rowland himself. Kayser again:71 I received a letter from Rowland, in which he told me that he was send- mg one of his gratings, which was for sale, to an exhibition of instru- ments at Berlin. He informed me of this, because he would be very happy If this grating would end up in my hands, since I had given sufficient proof that I know how to use it. This was [according to Row- land) the second best grating ever made in his workshop; he was keep- ing the very best, of course, for his own use.-Since I was not in the position to buy the grating myself, I sent the letter to Helmholtz with the request that he might ask the [Prussian) Academy [of Science at Ber- lin] to buy it and then to lend it to me. And so it happened, and the 69. Kayser (ref. 41), on 119-120. Cf. Kayser's letters to Rowland, 31 Jul and 19 Nov 1882, II May and 4 Aug 1883, preserved in RP. 70. The use of at least one, often several Rowland gratings is acknowledged, e.g., by Kundt. H.W. Vogel, H.C. Vogel, GIeseler, Kayser, Konen, Runge, Paschen, Back, Higgs. Lockyer, Meggers, Hale, and many others. 71. Adeney and Carson (ref. 19),223. quote from a letter by Brashear: "You are very fortunate in getltng thIS grating. for no one knows when we will get another." Cf. note 96. 72. Kayser (ref. 41). 146. REDSHIFT 241 marvellous grating was put to extensive use, first by me in Hannover. then, once I got the permission to take it with me upon my move to Bonn, by me and my students [there]. Over a third of the approximately 150 spectroscopic studies carried out under my supervision at Bonn were made using this grating. The impact of Rowland's gratings was amplified by improvement in photography, especially dry plate processing, to which Rowland contributed an emulsion that enabled him "to photograph from the violet down to D line."73 The enlargement permitted by photography also helped by revealing many lines, hitherto supposed to be single. as double lines; Rowland published his results in two Photographic maps of the solar spectrum.74 As an obituarist summed Up:75 A new weapon was placed in the hands of spectroscopists; it became possible to photograph spectra directly without the use of prisms or lenses, and with a greatly increased dispersion and resolving power; the beautiful maps issued at a later date by Rowland himself and by Higgs of Liverpool are striking evidences of the value of the grating; the addi- tions to our knowledge arising from this one discovery are already enor- mous; much has been achieved which, without it. would have been impossible. This is also true of the minute shifts of spectral lines in the sun's spec- trum discovered around 1890 in Rowland's laboratory at Baltimore. New standards Rowland and his collaborators put together extensive tables of spectral lines, some 20,000 in all running from 2100 to 7100 A. No one then could put forth a plausible explanation of their origin and magnitude. Rowland's research program was a huge Baconian exercise in data taking. He referred the wavelengths of all the 20.000 lines in the solar spectrum he recorded to the absolute value of the primary reference line, the D\-component of the sodium D-doublet. The wavelength of this line had been measured independently by many different observers; notably, Angstrom and Thalen in Uppsala. Muller and Kempf in Potsdam, Kurlbaum in Berlin, Peirce at the U.S. Coast Survey, and, in 1887/8, Louis Bell in Rowland's laboratory. in a very careful determination. In 1887, Rowland took all these published 73. Rowland (ref. 8), 483; Jon Danus. Beyond l'lSlOn (Oxford. 1984). for the history of SCientific photography. 74. H.A. Rowland, Ph%graph,c map of the normal solar spec/rum (Baltimore, 1888), and Ph%graph,c map of the Band D lines and carbon bands of the solar spec/rum (Bal- timore, 1889). 75. R.T.G. (ref. 10), 16-17. 242 HENTSCHEL FIG.8 Sketch of the "Rowland Room" at the Physical Institute, Universit) of Tubingen, with Rowland's concave grating G installed at the upper right and the camera movable on the semi-circle AS covering several orders. Here the spectrum of helium is being tested under the influence of a large magnet. Bezler (ref. 16), 144. values for the O( component of the sodium doublet, attributed rela· tive weights to them, and put forward the weighted average as the definitive absolute value to serve as the conventional unit for all further precision measurements he would undertake. Rowland employed the Table 2a Rowland's averaging procedure to define an absolute value for '\[Na-D[: Observer Year '\[0 1] weight Angstrom and Thalen 1868/84 5895.81 I Muller and Kempf 1886 5896.25 2 Kurlbaum 1888 5895.90 2 Peirce 1879 5896.20 5 Bell 1887/88 5896.20 10 Rowland's average 1887 5896.156 a. From Rowland, "A new table of standard wave-lengths," PM. 36 (1893),49-50. REDSHIFf 243 "method of coincidences" using selected lines at regular intervals as a secondary standard; between these lines, he determined the wavelengths of all the others by interpolation, relying on the superb quality of his custom-made screws to secure accuracy of his microme- ters. The coincidence method rests on the supposition that a coin- cidence of two spectral lines ~'1 and X2 belonging to orders 111. and 1112 occurs if 111.X.=1112X2· He started with the value of prominent lines in the spectrum relative to his absolute standard NaD.; he then obtained some fifteen reliable reference lines throughout the spectrum and interpolated between them within one order of the spectrum only.76 Rowland claimed an improvement of a factor of ten over the pre- vious efforts of precision spectroscopy. especially the tables of Angstrom,77 and he estimated the accuracy of his own values one in a million for the visible part of the spectrum. 78 Sometimes, dropping the cloak of Anglosaxon understatement, he said that no greater per- fection was possible: 79 Thus I have constructed a table of about one thousand lines, more or less, which are intertwined with each other in an immense number of ways. They have been tested in every way I can think of during eight or nine years, and have stood all the tests; and I think I can present the results to the world with confidence that the results of the relative meas- ures will never be altered very much. I believe that no systematic error in the relative wavelengths of more than about ±O.OI [A] exists any- where except in the red end as we approach [the Fraunhofer line] A. Rowland always stressed the intricate interweaving of all his measured wavelengths through many built-in checks, most notably the method of coincidence, which made use of the fact that all orders of the spec- trum were focussed simultaneously in his apparatus. 80 He did not merely compile interpolated wavelengths, but established a "system of 76. H. Kochen, "Der rote Tell des Elsenbogenspektrums." ZwPh, 5 (1907). 285-299. on 290, and Konen. "Wellenlangenmessang." Handbuch der PhYSik. 19 (1928). 777-801. 792, for the coincidence method. 77. H.A. Rowland, "On the relative wave-lengths of the hnes of the solar spectrum." AJS, 33(1887),182-190. on 183, also in PM. 22. 257-265. 78. Rowland (ref. 28), 204f. 79. H.A. Rowland, "A new table of standard wave-lengths," Amencan Academy of Arts and Sciences, MemOIrs, 11 (1896), 101-186. on 105; see also Rowland (ref. 28). 209. 80. Rowland (ref. 79), 102f. 244 HENTSCHEL wave-length standards." This system remained essentially und isputed in the 19th century:81 Through the use of his concave gratings, Rowland succeeded in estab- lishing a system of wavelength normals [for the arc and the sun's spec· trum], the relative accuracy of which was estimated to be some thousandths of a ten millionth of a millimeter, and which in fact fo rmed the basis for all wavelength measurements until the year 1906 within the fields of physics and astrophysics. Even many decades after Rowland's publications, updates, or as they were called, revisions, of Rowland's tables were published in 1928 and then again in 1966.82 It was within this then recently establi shed stan- dard system of solar and laboratory wavelengths that, in 1890, the red shifts of the Fraunhofer lines in the sun 's spectrum were discovered. FIG. 9 Short section of a photograph of the solar Fraunhofer spectrum made by (and signed) " H.A. Rowland, Phot[ographer];" the two darkest lines are the sodium D lines. Reproduced from Moore (ref. 10), 124. 81 . Konen (ref. 67), 780. 82. Charles Edward St. John et ai. , Revision of Rowland's tables of solar spectrum wave lengths with an extension to the present limit of the infra-red (Washington, 1928): Charlotte Emma Moore, M.G.J. Minnaen, and J. Houtgast, The solar spectrum 2935 Ii to 8770 Ii; second revision of Rowland's preliminary table of solar spectrum ware· lengths (Washington, (966). REDSHIFT 245 2. REDSHIFTS A key assumption of Rowland and the spectroscopists of his day was the invariabilily of the position of the lines in lhe spectrum, or, the precise coincidence of the emission lines in their laboratory spectra with the corresponding absorption lines in the Fraunhofer spectra from sunlight (figure 10). FIG. 10 Coincidence of some of the bright lines of iron with some Fraunhofer lines. Lockyer (ref. 121), 268. How far was the assumption of absolute coincidence justified? How much did it predetermine the outcome of measurements? 0 standard independent of Rowland's measurements existed; in a sense. he did not measure but rather defined the wavelengths. The foregoing diagram shows that, with his concave gratings, Rowland had just reached the point where the red and violet shifts in the sun's spectrum became detectable; before him observers could not have seen the effect. 0.1 .. Ditsch(o,ncr .. van der Wllhgen Angstrom .. Kurlbaum . .. . Z6l1ner 0.025 • Mu ller u Kempf . . ... ' . 0.01 FIG. II Accuracy obtained in spectrometry, 1864-1896; the order of magi- tude for red and violet shifts in the sun's Fraunhofer spectrum is indicated by the shaded area. Author's drawing. 246 HENTSCHEL 3. THE DISCOVERY OF REDSHIFTS IN THE SOLAR SPECTRUM While working on the Preliminary table of solar spectrum ware· lengths, Rowland encountered a problem whenever he obtained photo- graphic plates showing both the Fraunhofer lines in the sun's spectrum and a comparison spectrum of emission lines of glowing gases from the source in his laboratory. Instead of finding the expected precise coincidences of the sort as illustrated in figure 10, but on a much more detailed scale, he nearly always found minute shifts:83 In every plate having a solar and metallic spectrum upon it, there IS often-indeed always-a slight displacement. This is due either to some slight displacement of the apparatus in changing from one spectrum to the other, or to the fact that the solar and the electric light pass through the slit and fall on the grating differently. In all cases an attempt was made to eliminate it by exposing on the solar spectrum, both before and after the are, but there still remained a displacement of 11100 to 11200 division of Angstrom, which was determined and corrected for b) measuring the difference between the metallic and coinciding solar Jines. selecting a great number of them, if possible. The effect proved to be too persistent to be simply ignored. There· fore he tried to eliminate it. He exposed the lower third of his plate to the sun's spectrum, then the middle third to light from a laborator) arc, then the upper third to sunlight. Since this procedure was designed to exclude accidental shifts owing to minute changes in the relative position of slit, grating, and camera between the exposures of solar and arc spectra, he expected perfect coincidences. The effect nonetheless persisted. Now he had no choice but to mention it in hiS major publication and to invent a notation for his tables to indicate those lines where these shifts had occurred most conspicuously:84 However, it is not always possible to correctly assign the exact pOSition. and consequently there are probably many errors in the pOSitIOns assigned. Where the solar line is too strong to be due entirely to the ele· ment with which it is identified, it is represented thus: -Fe, and mdl' cates that the iron line is coincident with the red side of the solar Ime. the origin of the rest of the line being unknown. In keeping with this convention, he denoted a slight shift of the Fraunhofer line in the solar spectrum relative to his laboratory spectra toward the red by a minus sign to the left of the element symbol. In the case of a shift to the violet, the minus sign was placed at the right 83. Rowland (ref. 79), 116; cf. Rowland (ref. 77), 186. 84. Rowland, Preliminary table of solar spectrum wave·lengths (Chicago. 18961 signed Dec 1894, on 6. REDSHIFT 247 side of the element symbol of each line in his tables of solar spectrum wavelengths. Both cases occurred, as figure 12 illustrates. IlIIc,,"uy II",-II""Y \V:avc- 1cnl:lIl Sul,.,.I:lIICC alill \\':I\C ICIlI.:III SIlII .. I~II,"C alld ('I':U.U •. lcr ( 1 •. II.llh' ------- 311 6:·';SIl C 00 .1 IiI'7·CJCJI, ( .? (lIlO 3:l62.54 1 C 000 ,SI,II.060 (" - " .! ]1l62.62 7 C? .: .!I)(,}'. '7/ I; uo 3862.7 27 lSIIS . .:61 ( 0 .,lI62.S27 C 000 ,lihli . .!7': (" " 3!)62·~c)7 C 00 :;SI'~·';51 000 3!)62.C)02 0 .\III·!{.S:\') C 3l163·0';1 C 00 llihS.6:! 5 C <10 3S6 1-11.\ C 00 1~h11.700 C n 3S63.20 1 , IK6::1·7l1S C uO 3S63·3.;1 000 N ,ShS.!) 7 J C , 3!!61·5.U C I N 3tihti ·,)4 1 C 0 3S6 1.0SS C 00 jSllI).17c) C 0 3S0 3.73.; C 06 - 3~1)()·30S C 1 3S63.S1~ c· , .. .1 11o')·444 C oN d? .lSIl3.S!H! Fe .lSc,C)·SJ 1 ( . . ISlq.006 a I .ltiO') C,')2 I L ( . .ISC,.;./11 C a 1/ I!!Oc)·74 5 C .1SG.;.2.;1l ~I .. ·C 1 .!So,).SoS C 3Mo .;.';JK , C l!!ol).C)oo C 00 3Mo.;.0:6 C .1::l70 .oS3 C·Co 1:--1 .l!!6-1.720 C 00 ,S70.20 4 C 0 386,;.S02 C 00 .l1l70 .2tlc) C· IN ]S6S·00S V , N .->.3-dependence was soon tested in experiments by Duffield and Rossi, both of whom claimed to confirm it, although their measurements scarcely favored the dependence on >.3 over one on >.2. Once discovered, the pressure dependence of spectral lines became the object of a research industry. The more hands at work, the more complicated the results. Franz Exner, Eduard Haschek, and Heinrich Mache claimed that the wavelengths of spark spectra depended strongly upon the density of metal vapors and also upon the unknown 144. J. Wi Ising, "Theoretical considerations respecting the dependencies of wave length on pressure which Messrs Humphreys and Mohler have observed in the arc· spectra of cenain elements," APJ, 7 (1898),317-329; C. Godfrey, "Note on Professor Wilsing's anicle on the effect of pressure on wave-length," AP J, 8 (1898), 114. 145. W.J. Humphreys, "An attempt to find the cause of the width and of the pressure shift of spectrum lines," APJ, 23 (1906), 233-247; J. Larmor, "Note on displacement of spectral lines," AP J, 26 (1907), 120-122. 146. O.W. Richardson, "A theory of the displacement of spectral lines produced b) pressure," PM, /4 (1907), 557-578, on 558f.; cf. W.J. Humphreys, "Bericht tiber die Verschiebung von Spktrakkinien durch Druck," JRE. 5 (1908), 324-374. REDSHIFT 269 conditions of discharge in the production of the spark,147 while others insisted on the independence of these parameters, "as long as meas- urements are carried out correctly."148 More and more factors appeared to influence the spectrum lines: density and temperature of the gases, electric and magnetic fields at the emitter, capacity of the electric devices used in producing the sparks, and so on. Earlier measurements that had not taken into account these param- eters or had not fixed their values were suddenly rendered unreliable or useless. The investigators, usually immunized against questions about their certified data, had to admit the limits of their measure- ments, took to criticizing one another's accuracy and to hunting down the errors in instrument design and experimental procedures of their rivals. 149 4. PROBLEMS OF ST ANDARDIZA TION It will only be noted here that due to individual errors and the unevenly distributed differences between Rowland and International A, a reduc- tion with an accuracy of better than 0.01 A cannot be obtained. ThiS deficiency is also strengthened by the fact that there are too few Row- land wavelength normals for iron, and most observers have therefore supplemented them by using the system of Fraunhofer lines. These lines in tum are loaded with other types of errors, such as pressure shifts, gravitational redshifts, etc., so that to most observers, the "Rowland system" does not at all constitute an identifiable system of measure- ments. ISO With only slight exaggeration it might be said that each research school had its own system. 147. E. Haschek and H. Mache, "On the pressure in the spark," APJ, 9 (1899), 347- 357, and 12 (1900),50-51; F. Exner and E. Haschek, "Uber die Verschiebung der Spek- trallinien," Akademie der Wissenschaften, Vienna. Math.- Phys. K1asse, Sltzungsber- iehte. lJ6:2a (1907), 323-341. 148. 1.M. Eder and E. Valenta, "Unveranderhchkelt der Wellen lange im Funken- und Bogenspektrum des Zmks," ibid., lJ2:2a (1903), 1291-1304, and G.W. Midd1ekauf, "The effect of capacity and self-induction upon wave-length in the spark spectrum," APJ, 21 (1905), 116-123; H. Kayser, "Ole Veranderhchkeit der Wellen- hingen im Funkenspektrum," ZwPh. 3 (1905), 308-310, and "Die Konstanz von Wellenlangen von Spektrallinien," ZwPh. 5 (1907), 304-308, bnefly defended the view that the shifts in wavelengths were fictitious. 149. 1.M. Eder and E. Valenta, "The mvariability of the the wave-lengths in the spark and arc spectrum of zinc," APJ, 19 (1903), 251-262; Haschek and Mache (ref. 147); 1.F. Mohler, "Pressure m the electric spark," APJ, 10 (1899),202-206. 150. Konen (ref. 76), 797. 270 HENTSCHEL The realization of the potential influences of many parameters hitherto regarded as irrelevant led to the general demand for new stan- dards. William Marshall Watts (at Manchester), Heinrich Kaiser (at Bonn), Eder and Valenta, and Exner and Haschek (both groups in Vienna) and Fabry and Perot (later together with Buisson, at Mar- seilles) drew up their own spectrum tables, all expensively published, and all assigning slightly different values to the wavelengths of spectral lines}SI A new type of research arose, consisting in the systematic comparison of the various standard tables and in the search for for- mulas for the conversion of their values. ls2 Fabry and Buisson called a line 4427.313 A, Rowland called it 4427.482 A, and Kayser called it 4427.314 A:1S3 The conviction which had steadily been gaining ground for a long time past, that Rowland's wavelength system, otherwise quite accurate, which has been in use for the last twenty years as the exclusive basis of all spectroscopic research, is with respect to their absolute values subject to quite considerable errors, has thus received full confirmation; it has thus become apparent that a thoroughgoing reassessment of these values is necessary, using either Michelson's or some other similar interference method. Michelson's application of interferometric methods to the definition of the unit of length led to a radically new way of defining the meter and thus to an absolute basis for precision spectroscopy. Michelson's measurements of 1895, and Benoit, Fabry, and Perot's of 151. W.M. Walts, Index o/spectra (London, 1872), as well as its Appendix M (Man- chester, 1902) and Appendix A.A. (Wesclift', 1931); H. Kayser, "Standard lines m the arc spectrum of iron," AP J, /3 (1900), 329ft', and Handbuch der Spektroskopie, 5 (1910). 446ff., 7:1 (1924), 405ff.; Ch. Fabry and A. Perot, "Measures of absolute wave-lengths In the solar spectrum and in the spectrum of iron," APJ, 15 (1902), 73-96, 261-273; F. Exner and E. Haschek, Wellen/angen-Tabellen fur Spektra/ana/ytische Untersuchungen (2 vols., Vienna, 1902-04), and Die Spektren der E/emente bei norma/em Druck (2 vols .. Vienna, 1911-12); I.M. Eder and E. Valenta (ref. 149), Atlas typischer Spektren (Vienna, 1911), and Wellen/angenmessungen des Lichtes (Braunschweig, 1926). 152. Jewell (ref. 97); Gustav Eberhard, "Systematic errors in the wave-lengths of the lines of Rowland's solar spectrum," APJ, 17 (1903), 141-144; H. Kayser (refs. 148 and 152); "Bericht tiber den gegenwartigen Stand der Wellenlangenmessungen," ZwPh, 12 (1913), 296-308; Kuno Behner, "~r das Bogenspektrum des Titans von A = 7496 bis A = 2273," ZwPh, 23 (1925), 323-342, Hartmann (ref. 104), "The correction of the stan- dards of wave-lengths," AP J, 20 (1904), 41-48; and "Tabellen fUr das Rowlandsche und das intemationale Wellenlangensystem," Gesellschaft der Wissenschaften, G6ltingen, Abhandlungen, 10:2, 1-78; Konen (ref. 76), 781. 153. Values from H. Kayser, "Standards of third order of wave-lengths on the mter- national system," APJ, 32 (1910), 217-225; quote from Bernhard. Hasselberg, speech at the presentation of the Nobel prize to A.A. Michelson in 1907, in Nobel Lectures 1901- 1925 (Amsterdam, 1967), 162. REDSHIFT 271 1907 and 1913, which gave selected wavelengths widely distributed in the solar spectrum to a precision of about 0.00 I A, forced a drastic revision of Rowland's earlier work. Rowland had believed that the absolute values he had given in 1889 were correct to one part in one hundred thousand, and that the relative errors, that is, the ratios of any two wavelengths, should not contain errors exceeding one part in a million. ls4 But in 1893, Michelson's interferometric determination of the standard meter, found Rowland in error by about one part in 30,000.1SS Although this result drastically reduced the accuracy of Rowland's determinations, astrophysicists easily adapted themselves to it by multiplying his numbers by a factor close to unity. In any case, they cared more about the relative values of spectral lines, which would not alter by a rescaling of the absolute values. The situation changed again, when Charles Fabry and Alfred Perot made interferential measurements upon approximately thirty solar lines between 4643 A and 6471 A in 1901. 156 They compared their wavelengths against the standard of the red cadmium line: relying on their high-precision absolute measurement of one line, they calculated the others relative to it. If Rowland's numbers were relatively correct, their ratios for all lines should be the same, or very nearly so, as those of Fabry and Perot. In fact, the ratio not only varied for each line, but also showed systematic tendencies (figures 16 and 17): The ratios have nearly the same value for lines close together, but they differ for lines from widely different parts of the sun's spectrum, reaching eight parts in a million, more than eight times the limit of error for relative 154. H.A. Rowland, "Table of standard wave-lengths," PM, 27 (1889), 479-484; cf. eh. Fabry and H. Buisson, "Wave-length measurements for the establishment of a sys- tem of spectroscopic standards," APJ. 28 (1908),169-196, on 170; William Fredenck Meggers, "Standard wave-lengths," Optical Society of America, Journal. 5 (1921). 308- 322, on 309. ISS. A.A. Michelson, "Comparaison du metre mternatlonal avec la longueur d'onde de la lumiere du cadmium." CRAS. 116 (1893), 790-794, "Les methodes mterferen- tielles en metrologie," JP, 3 (1894), 5-22, and "Determmation expenmentale de la valeur du metre en longueurs d'ondes lumineuses," Bureau International des POids et des Mesures, Traveaux et memO/res, 11 (1895), 1-85; J.-R. Benoit, "Application des phenomenes d'interferences a des determmatlons metrologiques," JP, 7 (1898), 57-68, and "De la preCISIOn dans la determination des longueurs en metrologie," Congres International de Physique, Rapports, J (1900), 30-77; J .R. Benoit, Ch. Fabry and A. Perot, "A redetermination of the lengths of the red cadmIUm line," APJ. 26 (1907), 378-380, "Nouvelle determination du rapport des longueurs d'onde fondamentales avec l'unite metrique," Bureau International des Poids et des Mesures, Traveaux et memoires, IS (1913), 1-134, and "Observations," Ibid., i-cxlvi. 156. Ch. Fabry and A. Perot, "Longueurs d'onde de quelques raies du fer," CRAS, 132 (1901), 1264-1266, and "Mesures de longueurs d'onde en valeur absolu," AP. 25 (1902),98-139. 272 HENTSCHEL 1.0000550 l0000;300 450 . ~ ~ . . " . , . . \ -.,-t-t---f-~ . .:-I 50n .J Fig. 5 . I I-f-+ 600 B~ FIG. 16 Variation of Rowland's relative wavelengths as measured by Fabry and Perot in 1902 (ref. 156), 136. }, (R.~ }, (M. and K.) 0.999988 87 86 S5 84 83 82 81 80 79 i8 ii j6 75 0·999974 1\ \ 4500 i\ \ \ " " 1.\ \ /\ \ II I I i \ i / I I , : I I 1 I , 1 i _\ I ! ! J ( \ , f I 'J .-\ I 1£\ { ' I '. ! ! i '-; ill ! \"" I I I ". i \: ,: \J 1 ! \i -'j . : j : -j-' J ,-2i i 1/ , V K VI I V . I I , I 1 5000 5500 6000 6500 WAVE LE:-IGTIIS. },IR.) " (R.) }, ()1. and K,) }, (F. and P.) },(R.) }, (F. and P.) 1 1.000039 38 37 36 35 34 33 32 31 30 29 .00aazS FIG. 17 Correction curve for Rowland's wavelengths, 1903. Eberhard (ref. 152), 143, REDSHIFT 273 values that Rowland assigned to his values. The relative error of Rowland's determinations appeared to be ten times larger than he had claimed. To save Rowland's results, spectroscopists devised an elaborate set of conversion tables.157 Around 1904, Kayser claimed to have found that the errors in Rowland's determinations arose from systematic errors of the ruling of the gratings. ,s8 Others preferred other causes, such as disturbances in the apparatus or the failure to correct for pos- sible Doppler shifts and temperature effects.,s9 Slowly, a consensus developed that the grating showed systematic deficiencies for the determination of wavelengths over large intervals. Consequently, no reduction of Rowland's values would be possible with an absolute accuracy of more than a few percent of one A.160 This limitation of Rowland's system was officially acknowledged by the commission for wavelength standards of the International Union for Cooperation in Solar Research at its first meeting in 1904. Spectroscopists needed a new reliable primary standard linking their measurements with metrology, and a better, different method of measuring major reference lines at regular distances not too far apart from each other, the so-called secondary standards. Further reference lines between these secondary standards would be required for routine measurements in the laboratories all over the world (tertiary slall- dards). Gratings could be used for the determination of lines between the reference lines, because the errors induced by gratings could be neglected over intervals less than 50 A. "The grating, which is an excellent dispersive piece, is well adapted for measurements made by interpolation within a narrow interval, but is unsuitable either for absolute measurements or for the comparison of widely separated lines."161 The community of astrophysicists demanded a new system of stan- dards of wavelengths, if possible derived from artificial sources to avoid problems of the sun's physics. '62 The problem became so urgent that it prompted formation of the International Union for Co- Operation in Solar Research. '63 But only at the second meeting of the 157. Konen (ref. 76), 797f. 158. Kayser (ref. 148), and (ref. 151t); cf. Kochen (ref. 76). 159. Janet Tucker Howell, "The fundamental law of the grating," APJ, 39 (1914), 230-242; Jewell (ref. 97); F. Goos, "Standard wave-lengths In the arc spectrum of iron," APJ, 35 (1912), 221-232, and "A further contribution towards the establishment of a normal system of wave-lengths In the arc spectrum of Iron," APJ, 38 (1913), 141-157. 160. Konen (re;. 76), 779ff. 161. Fabry and Buisson (ref. 154), 171. 162. Cf. G.E. Hale, "Co-operatlon In solar research," APJ, 20 (1904), 306-312. 163. See the proceedIngs of the first meetIng reported in APJ. 20(1904), 301ff. 274 HENTSCHEL International Union in Oxford in 1905 did the body take decisions toward achieving its aim practically:'64 I. A line suitable for high-precision interferometry should be selected as a so-called primary standard of wavelength, defining once and for all the unit in which all other wavelengths are to be measured, differing as little as possible from 10-10 meters and to be called the Angstrom in honor of Anders Jonas Angstrom. 2. Additional lines should be selected from throughout the solar spec- trum as further reference lines, the so-called secondary standards, to be measured by interferometric methods relative to the primary stand- ard. '6S 3. For everyday use, further easily reproducible lines should be selected in intervals of about 50 A, the tertiary standards, by careful interpola- tion between the secondary standards. Because nearly 4000 A were to be covered in the visible and UV-range of the spectrum, about one hun- dred secondary standards, chosen mostly from the Fe-spectrum, were needed. 166 4. A good spectral source in the laboratory would be an electric arc operating at about 6 to 10 amperes. The community agreed on the red cadmium line as the primary standard, as the narrowest line then known, upon Michelson's recom- mendation based on his research on its fine structure by means of 164. Reported in the Transactions of the International Union, J (1906), 2301f., and Fabry and Buisson (ref. 154), 172. 165. Fabry and Buisson (ref. 154); Paul Eversheim, "Determination of wave-lengths of light for the establishment of a standard system," API, 26 (1907), 172-190, "Meas- urement of wave-lengths of standard iron lines," API. 31 (1910), 76-77, and "Wellenliingennormale im Eisenspektrum," AP, 36 (1911), 1071-1076, 45 (1914), 454- 456; A.H. Pfund, "A redetermination of the wave-lengths of standard iron lines," APJ. 27 (1908), 197-211. 166. Kayser, Handbuch. 7:1 (ref. 151); Kochen (ref. 76); E.J. Evans, "The arc spec- trum of iron A 6855 to A 7412," API. 29 (1909), 157-163; Franz Papenfus, "Ole Brauchbarkeit der Koinzidenzmethode zur Messung von Wellenliingen," ZwPh. 9 (1911), 332-346, 349-360; F. Goos (ref. 159); C.E. St. John and L.W. Ware, "Tertiary standards with the plane grating: The testing and selection of standards," APJ. 36 (1912), 14-53, 39 (1914), 5-28; Keivtn Bums, "The arc spectrum of iron," Lick Obser- vatory, Bulletm. 8 (1913), no. 247, 27-42; Ludwig Janicki, "Wellenliingennormalen dnUer Ordnung aus dem Bogenspektrum des Eisens," ZwPh. 13 (1914), 173-185; Hein- nch Viefhaus, "Ein Beitrag zur Bestimmung tertiiirer Normalen," ZwPh. 13 (1914). 209-234. 245-264; Sophie HoeJtzenbein, "Messungen im Bogenspektrum des Elsens zwecks Bestimmung tertiiirer Normalen," ZwPh, 16 (1916),225-253; H. Werner, "Mes- sung von Wellenliingennormalen 1m international en System filr den roten Spektral- bereIch," AP, 44 (1914),289-296; F. Goos, "Wellenliingen aus dem Bogenspektrum des Eisens im mternationa1en System," Astronomische Nachrichten. 199 (1914), 33-44; H. Plckhan. Untersuchungen des Systems der Elsennormalen (Ph.D. thesis; University of Milnster, 1918), Fnedrich Milller, "Beitrag zur Aufstellung des Systems internatlOnaler Wellenliingen," ZwPh. 22(1922), 1-20. REDSHIFT 275 interferometry. 167 At its third meeting, the International Union for Co-Operation in Solar Physics (lUCSP) attributed to this line the value 6438.4696 A in dry air under a normal pressure of 760 mm mercury at 15·C on the basis of two nearly concordant measurements made by Michelson in 1895 and by Benoit, Fabry and Perot in 1907. 168 The accuracy, about one part in ten million, bettered Rowland's determinations by a factor of 100. Later research into the fine structure of the red Cd-line showed that further progress in high precision spectroscopy would depend on the choice of yet another. sharper spectral line, such as A 5649 or A 5570 of the inert gas kryp- ton, which have a width of only 0.006 A and a limiting order of interference of about 600,000 at ordinary temperatures. 169 Several observers then started to measure a group of about 80 lines using interferometric methods and a pairwise comparison with the pri- mary standard. The independent measurements of A.H. Pfund. Paul Eversheim, and Fabry and Buisson were submitted to the fourth meet- ing of the International Union in 1910, which decided to average their very close values and to adopt the averages as the secondary stan- dardsYo Furthermore, in 1922 the International Astronomical Union adopted a supplementary system of 20 neon lines as secondary stan- dardsYI Several groups of spectroscopists measured the tertiary standards during World War I, especially in Bonn (by doctoral students of Kayser) and at the U.S. National Bureau of Standards (Meggers. Burns, Kiess). In 1922 the International Astronomical Union adopted a system of 302 iron arc lines carefully interpolated between eighty secondary standards previously adopted.172 During his work on tertiary standards in the iron arc spectra. Fritz Goos discovered the so-called pole effect in 1913: Depending on where the slit of a spectrometer is focussed between the two poles of the electric are, the laboratory emission wavelengths change their shape and shift by as much as 0.1 A '. Certain iron lines, particularly 167. International Union for Co-OperatIOn in Solar Research. Transacl/ons. 1 (Man- chester, 1906), 80ff., and 2 (1908), 109ff. 168. The two values differed by less than I part in 16 million (= 0.0003 A); cf. IUCSP, TransactIOns. 2 (1908). 109ff.; Konen (ref. 76), 790ff. 169. Anton Peter Weber, "Eme neue Methode h6chster Genauigkeit zur inter- ferometrischen Wellenhingenmessung und lhre erstmahge Anwendung zur Vorbestim- mung der filr den deutschen Anschluss des Meters an Llchtwellen vorgeschlagenen Kryptonlinien," PZ, 29 (1928), 233-239; Konen (ref. 76), 780. 170. APJ. 32 (1910), 215ff., and 33 (1911). 85ff.; H. Kayser. Handuch der Spektrosko- pic. 6 (Leipzig, 1912). 171. IUCSP, Transactions. 1, 35ff.; Meggers (ref. 154), 311 f. 172. IUCSP, TransactIOns. 1,35ff. 276 HENTSCHEL sensitive to pressure, had slightly different wavelengths in the center of the arc than near the negative pole. It took some time to establish the existence of this effect beyond doubt. t73 In the end, however, there was no choice but to amend the recommendations of the International Union with further details of the electric circuits involved, the precise distance between the two poles of the electric arc, and the point on which to focus the slit of the spectrometer; temperature and pressure varied too much over the length of the arc to allow spectroscopists to measure where they pieasedY4 All earlier measurements in which these parameters had not been specified clearly enough to recalibrate had become more or less worthless. Frustration prevailed. I75 In connection with the testing of Einstein's prediction of a gravita- tional redshift, the American astrophysicist Charles Edward St. John (1857-1935) realized the need for a revised table of wavelengths of solar spectral lines. Without such a revision, Einstein's prediction could not be tested, since the test required undisputed and sharp values for both solar and laboratory wavelengths free from any other effects. 116 In 1920, at the very beginning of the work that resulted in the Mount Wilson Tables of 1928, which covered the whole range from>. = 2975 A to 10200 A, St. John wrote about the endeavors of Rowland more than 20 years earlier: 177 It has long been recognized that the wave-lengths of Rowland's PrelimI- nary Table of Solar Spectrum Wave-Lengths, owing to an error in hiS primary standard, do not represent absolute values in the e.G.S. system and that the errors in the relative wave-lengths due to the method of coincidence used in passing from his primary standard are roughly periodic. It was the opinion of the solar physicists at the Brussels meet- ing of the International Astronomical Union in 1919 that the time had arrived when consideration should be given to the preparation of a table of solar wave-lengths based upon the international system. 173. Goos (ref. 159); Thomas Royds, "An investigation of the displacement of un- symmetrical lines under different conditions of the electric are," Kodaikanal Observato- ry, Bul/elm, 40 (1914),83-94; e.E. St. John and H.D. Babcock, "A study of the pole effect in the iron are," APJ. 42 (1915),251; "The elimination of the pole-effect from the source for secondary standards of wave-length," AP J, 46 (1917), 138- I 66. 174. See IUeSp, Transactions. 4 (1914), 58f.; Meggers (ref. 154),312: 6mm arc, 6 Amp for wavelengths greater than 4000 A, for others 4 Amp or less, a potential of 220 Volt, Iron rods of 7mm diameter and the choice of the axial pan in the center of the light source plus the restriction to iron lines of pressure dependency class a-d (Mt. WIl- son classification). 175. Kayser (ref. 41), 248f, 267f. 176. Hentschel (ref. 138); cf. John Earman and Oark Glymour, "The gravitational redshift as a test of general relativity," Studies in history and philosophy of SCience, J J (1980),251-278. 177. Carnegie InstitutIOn of Washington, Yearbook, /9 (1920), 228; cf. St. John. Moore, Ware, Adams, and Babcock (ref. 82). REDSHIFT 277 Despite the need to improve Rowland's tables of solar wavelengths, the astrophysical community did not lose its respect for his life's work. When the second revision of Rowland's tables was to be published, Marcel Gilles Jozef Minnaert, himself involved in a photometric Atlas of the sun's spectrum as another complement to Rowland's Tables, wrote: 178 What we have felt ever and ever again in the course of these years, that is the deepest admiration for Henry Rowland, who accomplished a simi- lar enterprise 70 years ago, with so much less technical means, and whose work is stiII now a marvel of perfection. That an unanticipated effect-solar redshift-was discovered during Rowland's decade-long efforts to establish solar wavelength measure- ments to eight digits, seemed at first to constitute a major challenge to precision spectroscopy; instead, this search for the next decimal turned into a virtue for high-precision physics. 178. M.GJ. Minnaert. "Forty years of solar spectroscopy," in C. de Jager, ed., Solar spectrum symposIUm (Dordrecht. 1963), 3-25. Minnaert and Houtgast (ref. 82) used a Michelson grating of 12 cm width.