zotero-db/storage/NANGQGKS/.zotero-ft-cache

Title Pages

Mapping the Spectrum: Techniques of Visual Representation in Research and Teaching
Prof. Dr. Klaus Hentschel
Print publication date: 2002 Print ISBN-13: 9780198509530 Published to Oxford Scholarship Online: January 2010 DOI: 10.1093/acprof:oso/9780198509530.001.0001

Title Pages
(p.i) Mapping The Spectrum (p.ii) (p.iii) Mapping the Spectrum

(p.iv) This book has been printed digitally and produced in a standard specification in order to ensure its continuing availability

Page 1 of 2

Great Clarendon Street, Oxford 0X2 6DP
Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide in
Oxford New York
Auckland Cape Town Dar es Salaam Hong Kong Karachi Kuala Lumpur Madrid Melbourne Mexico City Nairobi New Delhi Shanghai Taipei Toronto With offices in Argentina Austria Brazil Chile Czech Republic France Greece Guatemala Hungary Italy Japan South Korea Poland Portugal Singapore Switzerland Thailand Turkey Ukraine Vietnam
Oxford is a registered trade mark of Oxford University Press in the UK and in certain other countries

Title Pages
Published in the United States by Oxford University Press Inc., New York © Oxford University Press 2002 Not to be reprinted without permission The moral rights of the author have been asserted Database right Oxford University Press (maker) Reprinted 2009 Colour plates printed with the support from the Georg-Agricola-Gesellschaft zur Förderung der Geschichte der Naturwissenschaften und der Technik e. V. All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, or under terms agreed with the appropriate reprographics rights organization. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above You must not circulate this book in any other binding or cover And you must impose this same condition on any acquirer ISBN 978-0-19-850953-0 Cover illustration: The luminous, thermal, and chemical spectrum, as depicted in Robert Hunt’s hand-coloured plate from 1844. Printed and bound in Great Britain by CPI Antony Rowe, Chippenham and Eastbourne
Page 2 of 2

List of Illustrations
Mapping the Spectrum: Techniques of Visual Representation in Research and Teaching
Prof. Dr. Klaus Hentschel
Print publication date: 2002 Print ISBN-13: 9780198509530 Published to Oxford Scholarship Online: January 2010 DOI: 10.1093/acprof:oso/9780198509530.001.0001
(p.ix) List of Illustrations
Figures 2.1 Two drawings of spectrum observations by Leonardo 22 2.2 Delia Porta’s conceptualization of prismatic color as condensation, 1593 23 2.3 Descartes’s apparatus to measure the index of refraction, 1637 24 2.4 Descartes’s prismatic experiment with solar light 25 2.5 Newton’s experimentum crucis with the solar spectrum, 1672 27 2.6 A later variant from Newton’s correspondence, 1721 28 2.7 Newton’s optical analogue to the diatonic scale, 1704 29 2.8 Newton’s parallelogram representation of the solar spectrum, 1704 30 2.9 Erxleben’s seven homogenized primary colors, 1772 31 2.10 Hassenfratz’s absorption spectra as superimposed primary colors, 1808 32 2.11 Fraunhofer’s pencil drawing for his solar spectrum, c. 1814 35 2.12 Absorption spectra schematically drawn by Brewster, 1822 37 2.13 J. Herschel’s continuous curves of absorption in various media, 1830 38 2.14 David Alter’s tabular line count of spark spectra by color, 1854 40 2.15 Miller’s lithograph of molecular flame spectra, 1845 43 2.16 J.W. Draper’s diagram of continuous flame spectra, 1848 44 2.17 Swan’s comparison of the solar spectrum with hydrocarbon spectra, 1857 46 2.18 Bunsen’s spectra of alkali metals and alkaline earths, 1860 47 2.19 Diacon’s copper chloride spectrum, 1865 48 2.20 Bunsen and Kirchhoff’s second spectroscope, 1861 49 2.21 Comparison of spectra generated by flint and crown-glass prisms 50 2.22 Bunsen’s symbolic plot of emission spectra, 1863 52
Page 1 of 6

List of Illustrations
2.23 The spectrum of barium oxide compared with other compounds as plotted by Mitscherlich, 1864 53 2.24 Kirchhoff’s four-prism Steinheil spectroscope 54 2.25 Comparison of the scales of a prismatic and a diffraction spectrum 56 2.26 W. Herschel’s experimental setup, 1800 62 2.27 W. Herschel’s comparison of visual and thermometric intensity curves 63 2.28 J. Herschel’s thermograph and photograph of the solar spectrum 65 2.29 J. Herschel’s interpretation of the thermograph, 1840 66 2.30 Coexistence of luminous, thermal, and chemical spectra 68 2.31 Stokes’s mapping of the extreme violet and the “invisible region beyond” the optical solar spectrum, 1852 70 2.32 Lamansky’s plot of the energy distribution in the near infrared of the solar prismatic spectrum, 1872 73 2.33 Wiring diagrams for three types of bolometers 74 2.34 Langley’s experimental setup with both a prism and a grating, 1883 76 2.35 Langley’s graphic procedure for converting a prismatic spectrum into a wavelength plot, 1883 78 2.36 Langley’s graph of the solar spectrum into the infrared, up to 28 000 Å 79 3.1 Comparison drawings of the A group, 1861–1886 90 (p.x) 3.2 Browning’s automatic spectroscope with a six-prism chain 93 3.3 A typical lantern projection apparatus 96 3.4 Drawings of the A group between 1860 and 1881, magnified and redrawn to the same scale by Piazzi Smyth, 1882 99 3.5 Comparison of the solar spectrum at different times of day, 1871 102 3.6 Section of Thollon’s comparative atlas of the solar spectrum, 1890 103 3.7 Direct-vision rainband spectroscope 105 3.8 The rainband as seen in various weather conditions 107 4.1 Instruments and hand positioning for copper engraving 112 4.2 Comparison of line profiles in engraving and etching 112 4.3 Shading details of the Sr and Mn spectra in a steel engraving, 1864 113 4.4 Cross-hatching to render shades of gray in spectrum bands, 1871 114 4.5 Detail of Fraunhofer’s spectrum map with superimposed shading, 1814 116 4.6 Engraver’s roulette and sample use of it in Lecoq de Boisbaudran’s atlas of band spectra, 1874 118 4.7 Lecoq de Boisbaudran’s pencil sketch of the same motif, 1872 118 4.8 Lithographic line patterns drawn with various implements 122
Page 2 of 6

List of Illustrations
4.9 Two different types of drawing pens 123 4.10 Sample section of Kirchhoff’s map of the solar spectrum, 1861 125 4.11 A symbolic notation for the spectral sensitivity of collodion emulsions 131 4.12 Piazzi Smyth’s conventions for the representation of spectral line intensity 131 4.13 Section from Cornu’s map of the near-ultraviolet solar spectrum, 1874 134 4.14 Thollon’s second prism setup, 1879 136 4.15 Thollon’s final prism-chain design, 1879 138 4.16 Thollon’s final experimental setup, 1879 138 5.1 Reprint of Kirchhoff’s plates by the ‘precede Dulos’, 1864 146 5.2 Joseph Albert’s Munich Printing House, c. 1870 159 5.3 Photogalvanographic print of a sunspot photograph, 1862 161 5.4 Cartographic camera in military use, about 1885 165 5.5 Ruling machine for cartography 168 6.1 Microphotograph of a silver-bromide emulsion 186 6.2 E. Becquerel’s dark lines beyond the violet of the visible spectrum, 1842 195 6.3 J.W. Draper’s daguerreotype of the solar spectrum, 1842 197 6.4 Draper’s representation of the optical and tithonic solar spectrum, 1843 199 6.5 Hunt’s survey of contemporary photochemical experiments, 1844 200 6.6 The action of the solar spectrum on vegetable dyes, Somerville, 1846 202 6.7 Hunt’s color photograph of the solar spectrum, 1840 205 6.8 Mezzotint engraving of ultraviolet spark spectra of zinc and magnesium 209 6.9 Segment of Rutherfurd’s photograph of the solar spectrum, c. 1864 210 6.10 The violet and ultraviolet solar spectrum from Fraunhofer line G to the line complex R, 1862 212 6.11 Henry Draper’s wet collodion photograph and magnified pencil drawing of the solar diffraction spectrum, 1872 215 6.12 Albertype of H. Draper’s photograph of the solar spectrum, 1873 219 6.13 Photographic retouching procedures 222 6.14 Sample plate from Capron’s atlas of arc and spark spectra, 1877 224 6.15 Comparison of Lockyer’s ink drawing and photograph of the solar spectrum near H, 1881 225 6.16 A photomechanical print compared with its lithographic redrawing, 1890 228 (p.xi) 6.17 Detail from the second series of Rowland’s Photographic Map of the Normal Solar Spectrum, 1888 231
Page 3 of 6

List of Illustrations
6.18 Higgs’s mounting for a concave grating, 1894 241 6.19 Higgs’s photograph of the b group, 1894 242 6.20 Detail from Henry Crew’s photographic atlas of the zinc spectrum, 1895 245 7.1 H.W. Vogel’s small and large spectrographs 249 7.2 Comparative plot of the sensitivity of a dyeless gelatino-bromide plate, and enhanced sensitivities with various dyes 253 7.3 Abney’s comparison of Herschel’s thermograph with his own map of the infrared solar spectrum, 1880 257 7.4 The stepwise progress of photography of the infrared spectrum after 1900 261 7.5 Wedge spectrograph and wedge spectrograms of various types of photo graphic plates 263 7.6 Cyanine dyes with methine chains of increasing length 264 7.7 Fluctuations of the galvanometer needle connected to a holograph 268 7.8 Langley’s final bolometric curve of the infrared spectrum, 1900 269 7.9 Langley’s conversion of bolometric curves to normal spectra 271 7.10 Bolometric curve, 1929 273 7.11 Scheme of a photocell and electric amplification of the photocurrent 278 7.12 Diagram of Koch’s registering microphotometer, 1912 279 7.13 Photometric curve of a Zeeman-split spectrum line 281 7.14 Spectrophotometric wedge photographs 282 7.15 Utrecht apparatus for direct intensity recordings, 1938 285 7.16 Sample from the Utrecht Photometric Atlas of the Solar Spectrum, 1940 285 7.17 Transformation of a line profile into a microphotometer output function 287 7.18 Correction for the change in intensity of the continuous background 288 8.1 Balmer’s Pythagorean approach to series lines 297 8.2 Huggins’s representation of the hydrogen series in a Lyrae, 1879 299 8.3 Balmer’s geometric approach to series lines 300 8.4 Two sketches from the Balmer papers, c. 1884 301 8.5 Rydberg’s plot of the wave numbers of spectrum lines for the alkaline metal series, 1889 302 8.6 Lecoq de Boisbaudran’s pencil sketches of the N2 and BaCl spectra, 1872 309 8.7 Homologies in the spectra of Mg, Zn, and Cd, Hartley, 1883 314 8.8 Telescope mounting, micrometer screw, recording barrel, and prism setup for Piazzi Smyth’s measurements, 1883 315 8.9 Higgs’s wavelength plot of the oxygen band at 6900 A 322 8.10 Lowe’s atlas of the ultimate lines, 1928 331
Page 4 of 6

List of Illustrations
8.11 Diagram illustrating Gerlach’s relative determination of the concentration of an impurity in a sample, 1925 332 8.12 Practical example of identifying pairs of equally intense lines at various concentrations of Pb in Sn 335 8.13 Huggins’s star spectroscopes 345 8.14 Redshift of the hydrogen line H^ in the spectrum of Sirius, 1868 348 8.15 Stellar field photographed by the objective-prism method 352 8.16 A.J. Cannon inspecting a stellar spectrum photograph, c. 1940 354 9.1 General physics laboratory, MIT Boston campus, 1890s 372 9.2 Main physics lecture hall at MIT with posters of the spectrum, 1890s 373 9.3 Photographic dark room in the MIT laboratory of general physics 375 (p.xii) 9.4 Student exercise: wavelength determination in the sodium spectrum 380 9.5 Charting of absorption spectra in laboratory exercises in chemistry, c. 1890 384 9.6 Students at Wellesley College with physical laboratory instruments 387 9.7 Physics labs and Rowland concave grating with heliostat at Wellesley 387 9.8 Student drawings of spectra from Whiting’s lab session, c. 1890 389 9.9 Homemade fluid prism and box spectroscope for student exercises 407 9.10 Student observing the reversion of the sodium line in the flame 408 9.11 Homemade spectroscopic slit for observing Fraunhofer lines 408 10.1 Level diagram and frequency plot of the Balmer series, 1937 438 10.2 Graph with relative weights of symbolic and iconic signs in visual representations of emission spectra 1855–1955 440 10.3 Balmer’s drawing of a perspectivally shortened stairway, 1887 443 10.4 Balmer’s tangent method, 1884 444 10.5 Kayser’s composite photograph of the cyanogen band spectrum, 1889 456 10.6 Piazzi Smyth’s drawing of the Fraunhofer A and a bands, 1877 457 10.7 Instruction sheet on various techniques of drawing topographic maps, 1854 461
Color Plates Color plates appear between pages 34 and 35
I Fraunhofer’s map of the solar spectrum. Rare handcolored version and published black and white print, drawn and etched by Fraunhofer, 1814 II Top: Iris print of W.A. Miller’s lithographic plate of flame spectra of various metallic compounds, 1845
Page 5 of 6

List of Illustrations II Bottom: Bunsen’s chart of characteristic lines in emission spectra of alkali metals and alkaline earths, 1860 III Top: Brewster’s handcolored drawings of absorption spectra, 1823 III Bottom: Kirchhoff’s chromolithograph of the solar spectrum, printed on six stones, 1861 IV Top: Three of Secchi’s four classes of stellar spectra, engraved and colored by Dulos, 1870 IV Bottom: Color photographs of arc spectra obtained by Hermann Krone with the Lippmann process, 1892
Page 6 of 6

Tables
Mapping the Spectrum: Techniques of Visual Representation in Research and Teaching
Prof. Dr. Klaus Hentschel
Print publication date: 2002 Print ISBN-13: 9780198509530 Published to Oxford Scholarship Online: January 2010 DOI: 10.1093/acprof:oso/9780198509530.001.0001
(p.xiii) Tables
Tab. 3.1 Interplay between research objective and representational type 88 Tab. 5.1 Statistics on Berlin lithographers, copper engravers, and photographers according to the Be diner Adressbuch, 1836–1905 149 Tab. 5.2 Statistics of printers and engravers in London 1852–1900 150 Tab. 5.3 Photographic agencies and artisanal workshops in Berlin, 1875 152 Tab. 5.4 Survey of photomechanical printing techniques 156 Tab. 5.5 The various expenses connected with the production of a geographic map in 1793 170 Tab. 5.6 Cost comparison for copper and steel plates, 1861 172 Tab. 6.1 Main photographic processes, 1839–1879 177 Tab. 6.2 List of absorbents used by Rowland to filter parts of the spectrum for photography 233 Tab. 7.1 Survey of photometric atlases 1929–1981 286 Tab. 8.1 Pollok’s nomenclature for the sensitive spectrum lines of metals in solution, 1907 326 Tab. 8.2 Statistics on publications in spectrochemical analysis 1920–55 338 Tab. 8.3 Translation between the stellar spectrum classifications of Secchi, Vogel, and the Harvard System 353 Tab. 8.4 Survey of atlases of stellar spectra 1890–1978 361
Appendix 1 Survey of maps of the solar spectrum 1802–1900 465 Appendix 2 Survey of maps of terrestrial spectra 1835–1949 467
(p.xvi)
Page 1 of 2

Introduction
Mapping the Spectrum: Techniques of Visual Representation in Research and Teaching
Prof. Dr. Klaus Hentschel Print publication date: 2002 Print ISBN-13: 9780198509530 Published to Oxford Scholarship Online: January 2010 DOI: 10.1093/acprof:oso/9780198509530.001.0001
Introduction
Klaus Hentschel DOI:10.1093/acprof:oso/9780198509530.003.0001
Abstract and Keywords This introductory chapter begins with a survey of the rapidly expanding study of visual representations in science, followed by a discussion of spectroscopy as a prime example of a visual science culture. It describes ten historiographic levels of analysis, which are then documented in the remaining chapters. The mapping metaphor is analysed, and the rhetorics of spectra are studied. The chapter concludes with acknowledgments and a list of abbreviations for the twenty-five archives consulted.
Keywords: historiography, visual science cultures, mapping metaphor, rhetorics of spectra
1.1 The study of visual representations in science
Page 1 of 32

Introduction
In the last decade much attention has been devoted to nonverbal communication and visual representation in science.1 Pressed for an explanation of this upsurge in interest, both in sociologically oriented science studies and in the history of science, I would point to a confluence of several separate currents of research which call for more serious analysis of this dimension of scientific practice: (i) Several art historians have paved the way by bridging the gap between the fine arts and the sciences and by extending their studies into the realm of representations in the history of science.2 (ii) There is a growing number of studies by historians and sociologists of science and technology on the interplay between nonverbal communication and cognition—on the way one thinks, so to speak, with the hands and the eyes.3 (iii) There is a heightened awareness in each of these branches of the persuasive power of visual representations, which is backed up by a close examination of scientific controversies of the past and present.4 (iv) Even in histories of photography there is a broadening of scope. Traditionally focused on such standard sujets as portraiture, landscapes, closeup studies of the early pioneers or technical innovations, they now sometimes incorporate early scientific applications as a major, and indeed fascinating, historical factor.5 (v) Modern historians of cartography have started to reform their traditionally somewhat antiquarian approach. They now consider mapping as a process, and conceive of the resulting maps as systems communicating cartographic information.6 (vi) Finally, there are several articulate and welldocumented studies on the emergence and evolution of visual representations, especially in the geo- and biosciences, and also in astronomy.7
Page 2 of 32

Introduction
(p.2) The historical development of the graphical representation of data and objects had already been traced in some detail in a few isolated fields, such as statistics and botany, in the 1950s.8 But Martin Rudwick’s famous pioneering paper of 1976 on “The emergence of a visual language for geological science 1760–1840” pointed out particularly clearly that more was involved than development of progressively better and more suitable means of visualization. First of all, the use of visual representations in general is a contingent historical phenomenon, strikingly absent from some scientific traditions (Lagrange’s analytic mechanics being an often-cited example), yet absolutely indispensable in others like the spectroscopic terrain surveyed here. Second, new modes of visual representations do not emerge as isolated innovations but within the context of larger cultural packages in which they are securely wrapped. These packages include components of theory as well as of practice, as becomes readily clear once we look at—or better still, try to use—older systems of representation, such as an astrolab or a navicula. They are absolutely ‘opaque’ to a novice who has never been introduced to their underlying ideas. Nor does such a theoretical briefing make them immediately ‘transparent’.9 A hands-on initiation into the artifice and the practical skills necessary to handle them is still needed in order to see the phenomena through them. Experimental procedures and scientific instruments must recede into the background in order for the residual phenomena to emerge as objects independent of human intervention. The same holds for their representation. The reader must become versed in the appropriate use of representational devices such as histograms or bar graphs, pie charts, curves, plot functions, stereograms, cartograms, nomograms, etc.10 He or she must learn how to understand and read them. Initial resistance, misunderstandings, and controversies frequently occur in the history of science with the introduction of a new visualization technique. For instance, William Playfair (1759–1823), who first applied many now familiar graphical techniques to statistics, justified his use of a bar graph as follows:
This method has struck several persons as being fallacious because geometrical measurement has not any relation to money or to time, yet here it is made to represent both. The most familiar and simple answer to this objection is that if the money received by a single man in trade were all guineas and every evening he made a single pile of all the guineas received during the day, its height would be proportioned to the receipts of that day, so that by this plain operation time, proportion, and amount would be physically combined.11
He was right not to assume that his new technique would be immediately perceived as transparent. It was slow to take hold. Only the next generation, in the second half of the (p.3) nineteenth century, became aware of Playfair’s influence on the Continent.12 This reluctance to adopt new visual techniques is by no means limited to statistics. Historians have noted it in eighteenth-century scientific journals as well. Experimental graphs, illustrating the functional
Page 3 of 32

Introduction
dependency of one observable upon another, are found in a number of isolated cases but “never became commonplace in that age”.13 On the other hand, after such a technique has taken off, it remains alive and well for a long while, as the omnipresence of the pie chart in modern-day economics readily confirms. Once a technique has become established and its users fully accustomed to it, it often contributes substantially to the way problems are effectively handled and thought about. Exploded views of the inner workings of a complicated machine, first found in notebooks by Leonardo around 1500, are a good example; so are flowcharts used in engineering and programming. In fact, “graphs are especially suggestive when they re-evoke, stimulate, or revise the researcher’s visualphysical view of the process studied”. Rather than being perceived as a convention-ridden image, these nonverbal representations serve as “interactive sites” that the researcher is able to “see through” to the physical processes and materials selected or measured.14 Eugene S. Ferguson and Walter G. Vincenti have shown how nonverbal thinking allows craftsmen, designers and inventors to develop mental images of their machine’s workings, how they can reason their way through its successive stages of operation and how such intuition enables them to spot the critical phase or come up with an improvement.15 But thinking with (and in) pictures, e.g., drawing a diagram to organize thoughts rather than arguing in syllogistic form, is characteristic not only of the art of technology, but also of practitioners of other branches of science, such as mathematics, astronomy, geology, or botany. In Chapter 8, I give various examples of this ‘thinking’ with spectrum maps and spectrography—a thinking which often involved the search for patterns of lines belonging to a common series or band, or for homologies between different spectra. In reply to the likely objection that this pattern search was typical only of the nineteenth century, the following anecdote might be appended: When interviewed by Thomas S. Kuhn in 1967, the quantum physicist and chemist Friedrich Hund (1896–1997) explained how he had worked himself into the field of molecular band spectra, that is, before a systematic inventory in the style of Paschen-Gotze had been compiled for multiline spectra: “Well, I remember that I looked through much numerical data. I also remember that once I drew multiline spectra on sheets of millimeter paper about so long and stared fixedly at them.”16 Needless to say, with the advent of Bohr’s atomic model in 1913, and more so, with the rise of quantum mechanics in 1925, such a study method rather became the exception. But before this break in theoretical conceptions (which made it possible to explain what could formerly only be described), the wild-goose chase for patterns was a quite common activity. As evidence of this I will later discuss Johann Jacob Balmer’s and Henri Deslandres’s searches for coherent descriptions (p.4) of series or band spectra, Lecoq de Boisbaudran’s and G.L. Ciamician’s search for homolo-gies between spectra of different elements, and GJ. Stoney’s and Arthur Schuster’s work on the optical analogues of harmonics or overtone series (see § 8.2–8.4).
Page 4 of 32

Introduction
1.2 Spectroscopy as a visual culture I argue that spectroscopists during the second half of the nineteenth century were extraordinarily visually oriented, and this was by no means restricted to their main representational device, the spectrum map. The instrument maker and optician Josef Fraunhofer, the astronomer Charles Piazzi Smyth, the physicist Alexander Herschel, and his father, the pioneer photographer John Herschel, or the astrophysicist Samuel Pierpont Langley—all were interested in many different visual fields at once, and most of these scientists were accomplished draftsmen from the outset. Langley, for instance, became famous not only for his holographs of the infrared spectrum (to which we return on p. 79), but also for his highly detailed drawings of solar-spot observations by eye, during moments of exceptionally good seeing. In the case of Piazzi Smyth, his activities are as diverse as the tri-angulation of South African districts, landscape painting, day-to-day or tourist sketching, the lithography or engraving of prominent architectural sites, documentary photography of the Egyptian pyramids or the Tenerife Dragon tree, or ‘instant photographs’ of the clouds above his retirement home in Clova, Ripon.17 His colorful records of solar and terrestrial spectra profited from his trained eye and his subtle mastery of the pen and the brush. Piazzi Smyth was conversant in most of the technical printing repertoire that the nineteenth century had to offer and was able to select the most appropriate one for each subject.18 What led me to study his case more closely was his invention of symbolic techniques for representing spectra, just one more aspect of the fascinatingly broad scope of this truly artistic astronomer’s œuvre. He is a particularly vivid proponent of what I call the ‘visual culture of spectroscopy’ towards the close of the nineteenth century, a culture that spans across several disciplines, among them chemistry and astrophysics, physics and photography, and even medicine and engineering (for more on disciplinary issues see here § 10.1, pp. 420f.).
The longevity of such visual cultures as subsets of science with their integrated or associated predispositions for—or against—certain techniques of representation is documented in Peter Galison’s recent studies, in which he contrasts an image tradition and a logic tradition within twentieth-century physics. The former depends heavily on the use and construal of visual images (of mimetic reconstructions of natural events such as cloud formation, particle tracks, scattering events, etc.), while the latter relies on the quantitative, statistical analysis of large numbers of events as registered by electronic circuitry and detectors with apictorial, numerical output. Both of these research styles co-evolved during most of this century and survived several rather abrupt breaks with theoretical conceptions. It is only recently, with digitized image analysis, as recorded by charge-coupled devices (p.5) (CCDs), that these two competing styles began to merge: the electric output generated by these photocells can be manipulated in computer logic but ultimately still produces images similar to conventional photographs.19
Page 5 of 32

Introduction
Seen from this aspect, the study of spectra in general and spectrum analysis in particular has always leaned heavily towards the image side of the image-logic dichotomy. Successful research with spectra required refined patternrecognition skills in order, for instance, to distinguish the spectra of different chemical elements or to detect the spectrum lines of trace elements in another substance. Spectroscopy thus provides us with excellent material for a case study of the mechanisms and problems encountered throughout the many stages of the learning process: How to distinguish between different basic patterns? How to make sense of the many subtle variations and superimposed effects (such as Doppler shifts, line broadenings, Zeeman splittings, etc.)? How to discriminate between artefacts of the representation at hand and ‘real’ effects at the very border of detectability? The data in spectroscopy were, first and foremost, spectral plates and maps plotting the distribution of lines and their relative intensities. Wavelength tables and lists, and classifications of line strengths and types were also used, of course, but only in conjunction with the maps, and rather for special purposes, such as metrology. What counted primarily was recognition of the Gestalt of a certain line grouping as it was depicted in Bunsen’s chart of characteristic lines or in Ångström’s map of the solar spectrum. The absolute placement of these lines in the Ångström wavelength scale, later recorded in tables to a precision of many decimal places, did not matter to the overwhelming majority of its practitioners. In fact, it is not easy to think of a contemporary scientific discipline with a heavier load of images for memorization. Thus it is no surprise that many traces remain of the procedure by which newcomers had to acquire these skills. One nice example is the huge wall-hanging posters that most of us have seen at school on the walls of our chemistry or physics lab. Because of its great importance during the late nineteenth and early twentieth century, spectroscopy achieved fairly broad dissemination down to the level of high-school education and popular science literature. One of the aims of this monograph is to study these modes of diffusion of knowledge and, in particular, the use of visual representations of spectra in this process. As we shall also see (on pp. 387ff.), along with the ubiquity of spectrum representations in the scientific workplace came a pedagogic move away from passive memorization to active drawing or recopying of characteristic spectra by the students, a point which I could document on the basis of numerous student notebooks preserved at one East-Coast college (Welles-ley) which—unlike most university archives—fortunately thought these student notebooks warranted their storage space.
Page 6 of 32

Introduction
The history of spectroscopy in general has already been treated by more than one of its own protagonists, but the agenda of the earliest studies was too much determined by polemical priority disputes, while the later ones were too often blatantly celebratory.20 In 1900, the doyen of German spectroscopy, Heinrich Kayser (1853–1940), started publishing a multivolume series of handbooks designed to cover everything ever published in (p.6) the field.21 This Handbuch der Spektroskopie certainly does fulfill its purpose as a compendious inventory, but it falls into the historiographic trap of streamlined Whig history, cutting short what has been declared as obsolete. Spectroscopy has also been repeatedly studied since the late 1960s by a few professional historians of science mostly interested in the interplay between spectroscopic results and the emergence of atomistic models of matter or spectrochemical applications.22 Following the lead of Frank A.J.L. James, the most recent studies complement these internalistic perspectives with an approach directed at the cultural history of the field and connections to the institutionalization of new subdisciplines like astrophysics.23 But to this day, whole research areas of considerable impact in today’s science—such as quantitative spectroscopy, for instance—have still to find their historian. Altogether, this somewhat cursory attention does not adequately reflect the immense importance of spectroscopy in the natural sciences of the nineteenth and twentieth centuries. The spectroscope has served not only chemists, but also physicists, astronomers, and other scientists. It was utilized as a telescope, a microscope, a speedometer, a thermometer, a tape measure, a clock, and a chemical detector for such minute quantities as a teaspoonful of salt in a swimming-pool full of water. By the turn of the century, spectroscopy had become the dominant specialty, numerically speaking, in many institutes of physics, particularly in the United States where scientists had been concentrating on “manipulating light” since the nineteenth century.24
Page 7 of 32

Introduction
The amount of nonverbal, visual elements of the primary literature and documentation incorporated in the secondary literature is particularly disappointing: For instance, McGucken’s standard history of nineteenth-century spectroscopy has only nine illustrations, seven of which depict spectra; but these are analyzed only in connection with the contemporary atomic and molecular models of matter. This negligence reflects two lamentable historiographic tendencies: (i) an overemphasis of theory, which leaves room for discussion of experiments only insofar as they somehow relate to the theoretical argument (as is the case with McGucken’s book which concentrates on the search for series relations), and (ii) a myopia toward the many facets of nonverbal primary sources. As has already been pointed out by Martin Rudwick, this second point may have something to do with the text-orientedness of the historian of science, who is trained to work, analyze, interpret and deconstruct texts, and who in turn ends up writing a text about these texts. Many lack the training and practice for deciphering and analyzing nontextual components also embedded in these source materials. This contrasts sharply with the great attention that has been devoted within the last decade to nonverbal communication and visual representation in science, both within sociologically oriented science studies and within the history of science.
However, one can argue that in the history of astronomy, visual representations of the (p.7) constellations and celestial phenomena have always enjoyed considerable attention by historians of science.25 For this branch of science, scientific photography seems to be particularly well documented, reflecting its enormous importance in the waning half of the nineteenth century.26 Photochemistry, on the other hand, despite its defining role in the progress of scientific photography, has not received its due from professional historians of science. The only surveys we have in this area are written by some of the actors themselves, such as Robert Hunt and Josef Maria Eder,27 but an in-depth historical analysis of the development of this field still remains to be written. In Chapter 6, 1 discuss the following main stages of this arduous research and development, which throughout that century resembled more an intricate art than a systematic science:
• explorations of the so-called ‘actinic’ spectrum, which could only be recorded by silver salts or other photosensitive substances (see here § 6.1–6.2); • Hermann Wilhelm Vogel’s discovery in 1873 that dye additives sensitized his silverbromide emulsions to red light over 5000 Å, which led to the development of ‘orthochromatic’ plates (cf. here p. 248); • the availability of ‘panchromatic’ plates in 1904, which were sensitized further up to 7000 Å with dyes like dicyanine;
Page 8 of 32

Introduction
• further advances into the infrared with other dyes like cryptocyanine in 1919 (up to 8200 Å), neocyanine in 1925 (up to 9100 Å), and later derivatives, extending the range to 13 500 Å by around 1934 (cf. here p. 261).
The papers of George Harrison, William Meggers, and Kenneth Mees provide documentation for close cooperation between industrial research on photographic emulsions, at companies such as Kodak or Ilford, and researchers at the National Bureau of Standards, the MIT Spectroscopy Lab, or the Mt. Wilson Solar Observatory.
Unlike other studies on visual representations in the history of astronomy, which have a strong penchant for objects of special controversy (such as, for instance, Percival Lovell’s infamous Martian canals, or Nasmyth’s discovery of solar granulation),28 this book will focus on the uncontroversial, accepted routine of the spectroscopist and the everyday practice of men and women who taught and published on the subject. Thus it proposes to contribute to the growing body of studies on scientific practice. It retraces interesting if not always spectacular cases that, taken together, should give a clear and representative image of the actual procedures followed in the observation, recording, and mapping of spectra. For this was by no means an enclave of a handful of specialists, but one of the busiest branches in the sciences, at least in the period from about 1860 to 1900. It attracted chemists and physicists, and reached far afield into such areas as blood analysis and steel production, to name just two early and important applications (discussed in § 9.8).
(p.8) The establishment of scientific photography notwithstanding, astronomers, chemists, physicists, and spectroscopists alike depended on others to get their drawings or photographs into print. And these too often completely obscure artisans are indeed historio-graphic aliens. We sorely need supplementary approaches from social history and labor history to shed light on the status, working conditions, and specific skills of engravers, lithographers, photographers, and other specialists involved in the transition of research data onto the published plate. ‘Who were the guys?’, a question posed a quarter of a century ago by the social historian of science Lewis Pyenson with regard to lesser known scientists, and which since has also been raised with regard to scientific instrument makers, is now a burning issue with respect to members of this printing culture, too. In Chapter 5, I survey the existing work on the material culture of the printing trade and then present a few specific cases (see here pp. 143ff.).
Page 9 of 32

Introduction
One of the finest recent studies perceptive of this aspect is Alex Soojung-Kim Pang’s article on Victorian representations of the solar corona, which goes well beyond the historiography in other papers by systematically examining the interplay between observing practices, drawing techniques, and the subsequent technologies used in printing.29 The present monograph continues this line of research in another area of representations, namely, the spectrum. Interestingly enough, spectra share several qualities with the solar corona, for instance, diffuseness; that is, it is not easy to observe and hence not easy to draw. In fact (as we shall see in Chapter 2), it took a long time for certain conventions to become established on how to draw or map a spectrum, and these conventions changed dramatically with the introduction of new printing technologies and with the interpolation of photographic techniques into both the recording and printing processes (see Chapters 4 and 6). Yet, the research contexts in which the corona and the spectrum were scrutinized could hardly be more different. Until the invention of Lyot’s coronagraph around 1930, the corona was observable only during the rare times of solar eclipse at remote locations to which instruments and observers had to be transported on often strenuous scientific expeditions. Spectra, however, are usually observed in the laboratory, under controlled conditions, with fairly simple apparatus, and without the confining time constraints imposed on eclipse observers.30 For the role that photography had gained by 1900, Pang’s study of astrophotography provides an even better comparison case, because both stellar and spectrum photography require considerable care and experience to transgress the many technological limits of the contemporary photoengraving. As Pang documents for the case of astrophotography at the newly founded Lick Observatory, and as is shown here for spectrum photography, it was but a fine line between “improvement or correction of plates and doctoring”, and considerable craft went into the correct choices of printing technique, contrast and tone, ink and paper.31 But even with the most qualified of illustrators, things could go terribly wrong if their efforts were not carefully matched with the intentions of the spectroscopist who had originally supplied the sketches and drawings on which the lithographs and engravings were to be based. Even then, the lithographer and engraver were inevitably better placed to add the final touches to a plate on its way to press, as one spectroscopist ruefully notes in 1882: (p.9)
as to the Royal Society’s engraver who doubted my Oxygen lines: I have often had occasion to complain of the plates in the Transactions. A notorious case for instance is the plate illustrating Dr. Huggins’ paper to the Society; the drawing is abominable and a disgrace to the Society, while the original drawing (which I have seen) is excellent + very clear indeed. But these things we cannot cure + must therefore endure.32
1.3 The mapping metaphor
Page 10 of 32

Introduction
As natural and indispensable as it may look to us nowadays, the use of visual representation is very much a historical phenomenon, strikingly absent from some scientific traditions as much as it is strikingly prominent in others. The introduction of a device as simple, in our view, as a scaled two-dimensional celestial map, for instance, does not occur before the fifteenth century. It is roughly at this time that the familiar geographic maps appeared: unlike the earlier symbolic mappœ mundi, representational devices now “mapped’ specific geographical positions according to a specific scale or projection. As Peter Whitfield, author of a recent catalogue with dozens of first-class reproductions of such maps, argues: “celestial maps as we know them […] were a product of the Renaissance sense of ordered space, the sense which also saw the development of perspective, terrestrial mapping and scientific diagrams.”33 Whitfield points out that Ptolemy’s Almagest was not passed on to posterity with any illustrations (except geometrical figures). Even someone like Tycho Brahe did not feel the need to produce a star map, because to him “this was merely a demonstration aid which the non-specialist without instrument might use to identify what he saw in the sky”. Tycho preferred to invest his energy in compiling a reliable and precise tabular star catalogue. The introduction of representational techniques often derives from a transfer from one field, in which it has already become a familiar fixture, into another where it has yet to prove its utility. The rediscovery of Ptolemy’s projection method in the early modern period seems to be one such link between the history of central perspective, cartography, and stellar maps.34
The structure of the star chart, the projection of a measured sphere, was dependent on the new language of cartography which appeared at the end of the fifteenth century. The key feature of that language, without which modern scientific mapping could not emerge, was the co-ordinate structure, the ordered space imposed by the grid of latitude and longitude, that was learned by Renaissance geographers from the revived works of Ptolemy. The new art of ‘cosmography,’ with its diagrams of the earth and the heavens, became a characteristic Renaissance pursuit. With the growth of map printing, all atlases from the later sixteenth century onwards included star charts of the northern and southern heavens, often with diagrams of cosmic structures, the geometry of eclipses, lunar phases and so on. The star chart, a scientific document just as the world map is, became a publishing genre, subject to the intellectual and commercial demands of the day.
Page 11 of 32

Introduction
(p.10) In § 5.3 I document the presence of similar transfers of printing techniques from cartographic maps to spectrum maps, many of which were actually printed at agencies specializing in the production of topographic maps, such as the Ordnance Survey and Johnston’s printing house in Edinburgh, or the Swedish Military Lithographic Printing Agency. In the cases of several of the most outstanding engravers and lithographers involved in the production of nineteenth-century spectrum maps, I could likewise document active involvement in cartographic and architectonic branches of the printing business (see pp. 168ff.). Furthermore, both cartographers and spectrographers crucially depended on high-precision instrumentation, often delivered by the same superb instrument-makers.35
This affinity does not stop here, though. As will be discussed further in one of the concluding sections (10.10), an unidentified spectrum also shares much with terra incognita: Both have to be mapped for orientational purposes, as well as in order to truly ‘know’ them. In both cases it is not clear from the outset how to overcome their inherent amorphousness, which allows several nonequivalent ways of ‘mapping’ them. It may sound strange to use the expression ‘mapping’, familiar in the context of geography or genetics, but as we shall see, it crops up in many quotes about spectra. So it is quite definitely an actor’s category, not some concoction of us latter-day historians.36 Svetlana Alpers already pointed out the necessity of distinguishing between ‘mapping’ in a narrower and broader sense of the term:
Used narrowly, mapping refers to a combination of pictorial format and descriptive interest that reveals a link between some landscapes and city views and those forms of geography that describe the worked in maps and topographical views. Used broadly, mapping characterizes an impulse to record or describe the land in pictures that was shared at the time by surveyors, artists, and printers, and the general public in the Netherlands.37
The identification of a deeply rooted “mapping impulse in Dutch Art” helped Alpers to integrate Dutch painting into the broader cultural context and mentality of this sea-faring nation in the seventeenth century. For us, too, it will serve an integrative function with respect to common attitudes, interests, and skill transfers between spectroscopy and cartography, between various scientific subdisciplines and certain branches of the graphic arts. Our case likewise ranks the ‘testimony of the eye’ above traditional authority, with a primarily descriptive mood reigning.
Page 12 of 32

Introduction
But there is no such thing as a unique visual description of any object. While the different cartographic conventions and projection techniques are well known, along with the debates over which to choose from among them,38 it is less known that there is likewise a rich repertoire of modes of spectrum representations (cf. here pp. 2Iff.). Once a given mode is chosen, the spectrum map may be enlarged, or parts of it zoomed into closer view, (p.11) as soon as the interior has been explored and charted more thoroughly. They can be extended once new knowledge of bordering regions has been won, and then condensed again to regain a better overview and to avoid the danger of getting lost in the maze of details in a highly magnified chart. In fact, this parallel with cartography goes even further with respect to printing techniques (as we shall see in § 5.3).
Incidentally, cartographers tend to distinguish quite clearly between maps and atlases. An atlas is defined as a “very specific intermingling of written cartographic texts whose whole is more than just the sum of its maps: the atlas is a symbol of both its maker’s professional status and the social worth of its owner (because of the greater financial capital involved in atlas production) and is also a metaphor for the encyclopedic sum of geographic knowledge”.39 Spectroscopic use of the two terms is much looser, however, with many selfdescribed spectrum maps no less costly to produce nor less anxiously awaited by practitioners than so-called spectrum atlases. Ångström, Eder and Valenta, or Hagenbach and Konen presumably called some of their publications atlases in order to indicate a particularly broad spectral range covered or larger numbers of lines tabulated in the accompanying text. But other atlases (such as Higgs [1894]) are devoid of such accompanying tables. Size may also have been a criterion, with most atlases being in folio format or even grand-folio; but again there are exceptions, such as Uhler [1907], Mees [1909], or Löwe [1928], all of which are in small octavo. Besides these parallels to the cartographer’s encyclopedic epitomes, there may also have been an—unconciously sought— elevation of status, some groping for recognition. But of much more importance than the name attached to the product of one’s labors was whether it was accepted by the community as a standard reference work. Ångström’s atlas achieved this just as much as did Rowland’s map.
Page 13 of 32

Introduction
‘Mapping’ in and of itself is a widely used visual metaphor, signaling that a broad region of scientific objects is subsumed within the ‘empire of knowledge.’ Anthropologists use the concept of mapping to denote the production of a “mnemonic aid, especially visual, which all cultures utilize in some form or another.” The word derives from the Latin mappa which originally meant ‘tablecloth’ or ‘napkin’ and later also became the root of mappœ mundi, (i.e., medieval maps of the world). As Alice Jenkins writes in her analysis of spatial imagery in nineteenth-century representations of science, “the process of mapping is part of the harmonization of knowledge: ‘cartography’ creates order.”40 Whether geographic terrain, brain tumors, cosmic radio sources, or the human genome, entering this keyword ‘mapping’ into any university library database will yield output from all these fields and more, spectroscopy being just a footnote. But the analogies, I claim, are much deeper than simply the way that scientists speak about their activities (which is Jenkins’s focus). They are also to be found in the way that these representations are employed in the various stages of scientific research practice (which is my focus). Initially, just as in its counterpart the imperial chart, a new spectral region is mapped exploratively, for reconnaissance purposes, so to speak. Surprisingly often, this is done without any real understanding of what it is that is being mapped: Langley aptly described (p.12) his procedure as a “long groping in the dark”.41 In fact, as has been pointed out with respect to early maps of radio sources, “the value of such a map was also independent of any particular conception of the sources. It would, however, facilitate further optical identification and thereby improve the likelihood of establishing the nature of the radio source”, or more generally, of the mapped object.42 Exchange the word ‘radio’ with the word ‘light’, and the above quote is also valid for spectroscopic maps throughout the nineteenth century. Thus a substantial part of this book deals with this early, explorative phase of map-making, before there was any deeper theoretical understanding of what causes these spectra (the step-by-step deciphering of the spectral code is discussed here in Chapter 8).
Page 14 of 32

Introduction
In the next stage of map-making, the map is supposed to inform other experts working on adjacent or similar territory about the new conquests. Finally, the map assumes the function of unambiguously dispersing ‘secure’, certified knowledge in order to further the depletion of the new terrain’s resources or potential applications (similar to the initial logistical planning stages for the exploitation of a fully circumscribed and subjugated colony). Seen from this angle, Langley’s plot of the infrared spectrum, and his clarion call upon each new advance is the staking of a territorial claim at the farthest frontier: the infrared wavelength region beyond 10 000 Å: that is, beyond what photographers before him had reached is henceforth Langley’s land. Likewise, the region below 1850 A in the near ultraviolet, where Victor Schumann first succeeded in obtaining photographic spectra by evacuating his spectroscope and preparing gelatine-free emulsions (see here p. 71) is henceforth Schumann’s region, which name it still bears today. The Lyman region lies further beyond, and so on with many other examples throughout the full electromagnetic spectrum. This parallel between cartographic and spectroscopic mapping goes deeper, though. As we shall see (on pp. 168ff.), there were multifarious links between both activities. Indeed, if we replace ‘earth’ with ‘spectra’ in the following quote about geological maps, the first president of the Royal Geological Society might just as well have been referring to spectrum maps:
Words following words in long succession, however ably selected those words may be, can never convey so distinct an idea of the visible forms of the earth as the first glance of a good Map. […] In the extent and variety of its resources, in rapidity of utterance, in the copiousness and completeness of the information it communicates, in precision, conciseness, perspicuity, in the hold it has upon the memory, in vividness of imagery and power of expression, in convenience of reference, in portability, in the combination of so many and such useful qualities, a Map has no rival.43
Page 15 of 32

Introduction
But the concept of mapping goes beyond the connotations of depicting, exploring, and offering orientation in unfamiliar terrain. Maps try to achieve more than a simple ‘representation’ of something that previously either had been observed directly or recorded indirectly. Maps are more than mere icons— in Peirce’s terminology those signs that bear a similarity to the depicted elements. They are also laden with symbols of various kinds (p.13) whose meaning is defined by convention. Think of a typical road map with its ‘key’ to explain symbols for various types of roads, railway tracks, rivers, churches, etc., and a ‘scale’ to indicate how to translate distances on the map into distances of the objects depicted. Sometimes a regular grid is superimposed onto the map so that each detail may be correlated with its respective longitude and latitude. Less often, indexical signs such as arrows point to features of special interest. All these features allow us to ‘read’ a map, not unlike how we read a book, as something “easily legible in a succession of fixations”. Thus in every map symbolic conventions are mixed with iconic elements. The systematic reason for this is that maps aim at something different from mere depiction or mirroring. As the art historian Ernst Gombrich put it: “Maps are normally designed to impart information about the invariant features of an area, in other words they leave ‘appearance’ on one side. […] In maps we want identicals to show as identical regardless of the angle from which we happen to look at them.”44 Throughout this book, spectrum representations reiterate this inbuilt tension between icon and symbol, between capturing the full specificity of the observed feature and rendering what is typical about it, between representing and abstracting. For each of the many features of spectrum lines, such as line intensity, width, sharpness, wing shape, satellites, background intensity, and foremost color, an adequate rendering had to be found. And as we shall see, quite often no single best solution existed but rather several, often coexisting possibilities, which set apart the various practitioners of spectroscopy. It is a central aim of this book to illuminate the full breadth of this phenomenology of spectrum maps and their associated practices, but also to understand the reasons behind changes in these representational practices and their interplay with technological, experimental, and theoretical developments. On p. 45, for instance, when we look into the early history of representations of terrestrial spectra (as opposed to the solar spectrum), we shall see a gradual transition from a merely narrational description of flame or line colors (such as Talbot’s) via a symbolic, tabular vertical arrangement (such as Alter’s or Wheatstone’s) to an iconic horizontal chart (such as Masson’s). We shall also see, however, that this development is not a one-way track; on the contrary, a diachronic analysis of visual representations of emission spectra will reveal a meandering between preferences for iconic and symbolic types of representations (see Fig. 10.2, p. 440 for a synthetic summary).
1.4 The rhetorics of spectra
Page 16 of 32

Introduction
Each of these different modes of representing spectra carries with it a very specific rhetoric, something by no means limited to the verbal component of historical sources.45 Mapping, in particular, involves selection and omission, simplification, classification, the creation of hierarchies and symbols. All of these procedures have or may have a rhetorical effect. Even the plainest spectroscopic map—apparently devoid of extraneous ornamentation—carries with it at least one agenda: to convince the users of its correctness and objectivity, its selfevident factuality. Henry Draper’s first photograph of a solar spectrum is a striking example. He presented this product of a diffraction grating in photomechanical reproduction as “the (p.14) work of the sun itself, absolutely untouched”. We shall see in § 6.7 whether this actually agreed with the manipulations of this image by himself and his printer. Draper’s written comments on his spectrum plate were ‘rhetoric’ in the derogatory sense, even approaching intentional deception or trickery, but we should not limit our definition of rhetorics to such extreme cases. Even without this textual commentary, Draper’s oversize plate in itself also carried rhetorics of its own (on the following, cf. Fig. 6.12 on p. 219). It displays the full spectrum strip in moderate magnification together with an enlarged section, presumably to show more clearly the high density and fine gradations of the recorded lines. A reproduction of the corresponding part from Ångström’s lithographic atlas was supposed to underscore further that photography records far more lines than visual observation.
The rhetoric in many of the examples of visual representations presented in this book constitutes much more than simple accuracy and austerity of design. The choice of scale, frequency (or wavelength) range, mode of representation and printing technique, along with many other characteristics contributes to the overall appearance of the map and its influence or persuasive power. Samuel P. Langley’s interesting dual representation of the infrared spectrum (see Fig. 2.36 on p. 79) shows that spectrum maps and atlases played a crucial role not only in scientific documentation but also in public presentation. His plot of the infrared part of the solar spectrum, which is nine times as long as the visible part, must have been impressive for students like Whiting’s, who were shown this plot (cf. here the quote from one of her students’ notebooks on p. 390): Here was a vast new territory to explore. We might compare this with early maps of North America when the Wild West was still largely unmapped terrain, or with subliminal but highly effective Eurocentrism at work in the mercator projection of world maps in a typical college atlas.
Page 17 of 32

Introduction
The persuasiveness of such a visual representation also has much to do with issues of accepted visual conventions and aesthetics, again largely unexplored areas as far as scientific images are concerned. In § 10.9, I reflect upon aesthetic assessments voiced by the historical actors with respect to graphs, maps, and other visual representations of spectra. I argue that from 1860 on, spectroscopy played a significant part in establishing a more visually oriented science culture, quite in step with other fields, which were drawing away from a rigid textual orientation at roughly the same time.46 The nonverbal and nonnumeri-cal types of spectral representation used by teachers of spectroscopy played a crucial role in conveying the skill of pattern recognition. The Bunsen chart of the characteristic spectra of the alkaline metals (which figures so prominently in the research contexts of § 2.4, pp. 47ff., and §8.1) thus also serves as an icon for a didactic style: The pupils were made thoroughly familiar with these characteristic patterns in laboratory exercises designed for identification of the elements in the Bunsen-burner flame (cf. § 9.2). Such spectra were displayed on printed posters which, after 1860, suddenly began to appear in their thousands in chemical, physical, and technological laboratories and classrooms throughout the world. Emission line spectra of the chemical elements thus served as a spectrochemical alphabet, with the (p.15) spectra of unknown compounds analogous to words. These complex spectra, interpreted as superimposed images, then had to be analytically decomposed into their constituent elementary images in a visual process by which the lines of identified elements are subtracted from the complex compound spectrum. It was basically a mechanical alphabet—this analogy has come up repeatedly since the eighteenth century47 but was never to become as productive in spectrum analysis as in the physico-chemical sciences. The same ‘letters’ later also resurfaced in the spectra of celestial bodies such as stars or nebulae, allowing one to ‘spell out’ their chemical constituents despite the immense distances involved.
Page 18 of 32

Introduction
Concerning the periodization of visual representations, Lorraine Daston and Peter Gal-ison have suggested a specific tripartite division of such sources in a deliberately cross-disciplinary fashion.48 By studying a much narrower field in depth, taking into account not only the published images but, where possible, also the unpublished drafts or the larger pools of photographs from which the final ones were pulled, I attempt a systematic survey and analysis of these nonverbal sources. Starting with a brief survey of the earliest records of spectrum representations, a more detailed discussion fixes the beginnings at shortly after 1800, with both Wollaston and Fraunhofer finding indications of nontrivial substructure within the solar spectrum. This rich tradition of mapping the solar spectrum, both lithographically and photographically has its height towards the end of the nineteenth century. I trace the roots of spectrum analysis and qualitative spectroscopy to the older research strands of physical optics and chemical flame-color analysis and discuss the emergence of these fields as veritable research areas, stimulated by the publications of Bunsen and Kirchhoff. I do not continue this particular historical line after the transition to Bohr’s quantum theory in 1913, because the remainder of this story has often been told. Stellar spectroscopy, which only took off seriously in the 1860s (see here § 8.8, pp. 343ff.), and quantitative spectroscopy, which had an even later breakthrough in the late 1920s and 1930s, are consequently followed up somewhat further into the late 1940s, with some tabular summaries concerning their later rapid expansion. The overall dynamics of these areas of research, which is so strikingly different from the temporal development of visual representations, necessitate these differing time spans.
1.5 The structure of this book During my earlier researches on the interplay of instrumentation, experiment, and theory in solar and terrestrial spectroscopy,49 I was struck by the impressive variety of techniques used since 1800 to depict this phenomenon in all its subtlety and complexity. This diversity inspired me to go through atlases, maps, plates, and other illustrations in spectroscopic and astrophysical publications to study systematically the “emergence of a visual language” for spectra.50 Following a generally diachronic vein, my aim is to combine a historical investigation of the changing modes of spectral representations with systematic considerations (p.16) about the interplay between research and printing, on the one hand, and research and teaching, on the other. Along the way, we encounter other branches that prove indispensable for dealing in a satisfactory manner with ‘line matters’—to borrow the lithographer’s term. For not only physicists, astronomers, and chemists were occupied with spectral representations, but also engravers, lithographers, photographers, printers, and other such artisans, and for each of these groups, questions of social identity have to be settled. For each we must know how they acquired their skills, what are their norms, their reputations, and how did they interact among themselves. As their social status changed over time, so did their everyday routine in the
Page 19 of 32

Introduction
wake of research and printing innovations that developed rapidly over the course of the nineteenth century. The goal of this study is to write the history of representations of spectra in as broad and balanced a way possible, focusing as much on the internal development of the scientific disciplines involved, as on the underlying material cultures. As much attention will be given to changes in the cognitive framework in which spectra were discussed, as to the visual skills needed to find the patterns considered relevant throughout the decades, and the manual skills needed to fix them as spectral characteristics identifiable and recognizable to any newcomer in the field. My sources are figures, plates, maps, and photographs from the published literature of several disciplines, especially physics, astronomy, and chemistry, but also from engravers and printers manuals, photography handbooks, correspondence between members of the various groups involved, and—where possible—sketches and drawings from laboratory notebooks. Among the fruits of my research in two dozen archives were quite a few documents allowing a step-by-step reconstruction of the different stages of the illustration process, leading from the initial observation through recording to final publication. For instance, in Fig. 6.11, the reader can examine Henry Draper’s collodion photograph right next to his pencil drawing of the same region of the spectrum, and in the subsequent Fig. 6.12, compare a proof of Bierstadt’s Albertype reproduction of it. All this material I found glued into one of Draper’s laboratory notebooks, together with notes on the back-andforth between researcher and printer (see here p. 213). But this full documentation of so close an interaction is admittedly highly unusual. In many cases a considerable amount of footwork was needed to come up with anything beyond the printer’s or the engraver’s last name—quite definitely, the social history of the printing trade, in particular, those branches connected to scientific illustrations, still deserves much more attention.51
Although in the early years of the nineteenth century depictions of the spectrum were quite rare, from about 1860 spectroscopic diagrams became quite frequent. An initial survey of which kinds of representation were dominant when is given in Chapter 2. It also outlines the gradual emergence of certain conventions in the mapping of spectra. In Chapter 3, 1 concentrate on the stepwise enlargement and narrowing in on interesting spectral segments, which was directly linked with progressive improvements in the scientific instrumentation used for observation and recording. I go into considerable detail about instruments of particular relevance to the production of spectrum maps (such as Fraunhofer’s, (p.17) Henry Draper’s, or Rowland’s experimental setups).52 The interplay between the chosen modes of representation and the available printing technologies is thematized in Chapter 4 with respect to engraving and lithography and in Chapter 6 with respect to photography.
Page 20 of 32

Introduction In the 1870s and 1880s the growing importance of photography with the development of sensitive dry-plate processes and manageable techniques of photomechanical reproduction raises several other interesting points that are explored in Chapter 6. In contrast to the popular conception of photography as the most naturalistic form of visual representation, supported by the passionate rhetoric of its early ardent advocates (cf. § 6.2), 1 show in § 5.2 that at least until the mid-1870s, but to a lesser degree throughout the nineteenth century, the contemporaneous photographic and photomechanical techniques had crippling drawbacks which delayed full-fledged displacement of the traditional techniques of recording and reproduction for almost two decades, when the first photographic maps of the solar spectrum appeared (§ 6.9). During the transition period very often both traditional and modern illustrative techniques were used side by side, compensating each other’s weaknesses in a complementary fashion (cf. again § 6.4–6.8).
Page 21 of 32

Introduction
Having thus dealt with the different modes of representation in the first part, the book then turns to a discussion of the use of such spectrum plates and maps in research (Chapter 8) and teaching (Chapter 9). The discussion of the research applications centers around areas in which the search for patterns played a major role, namely spectrum analysis (§8.1), series, harmonics, and homology identification (§ 8.2–8.4), band spectra (§ 8.5–8.6), quantitative and stellar spectroscopy (§ 8.7–8.8). Chapter 9 presents the way spectroscopy was taught at high schools, colleges, and universities and its dissemination in popular treatises to other audiences like apothecaries, chemists, physicians, etc. Special emphasis is given to documenting the actual use of visual representations of spectra in lectures and laboratory exercises (§ 9.4–9.6). By means of a systematic comparison of the pertinent sections on spectroscopy and related issues in textbooks, laboratory manuals, and other material (§ 9.1), I am able to reconstruct the many levels of this education and pinpoint the crucial importance of nonverbal communication in acquiring pattern recognition skills, which were constitutive for spectroscopy during the nineteenth century. I also compare various local and national differences in the curriculum, picking out examples such as E.C. Pickering’s courses at MIT (in § 9.2), those offered at the Harvard Student Astronomical Laboratory (§ 9.7), or Sarah Whiting’s courses at Wellesley College (§ 9.5). The last is a particularly interesting case, because unlike most other institutions of higher learning this women’s college has preserved the lecture and laboratory notes of many of their students, some of which were later employed at the Harvard College Observatory as computers and aids in the classification of stellar spectra. While Whiting’s emphasis lay on transmitting the skill of spectroscopic pattern recognition, Lockyer’s teaching at South Kensington proposed to convey the skills needed for spectral mapping (§ 9.4). The gradual evolution in the selected topics and teaching methods is studied diachronically for the case of physics education at the Ecole Polytechnique in France roughly between 1860 and 1920, on the basis of preserved hectographed course manuals drafted by four generations of spectroscopists (§ 9.9).
(p.18) Both for the research and the teaching contexts, the aesthetic appeal of spectra on their viewers can scarcely be exaggerated. In § 10.9 I discuss this phenomenon with special regard to the language that spectroscopists used to describe fluted band spectra. Afterwards, I show how aesthetic motives also formed an important undercurrent in Louis Thollon’s endeavors at what I term ‘spectroscopic portraiture’, i.e., a rendering of the overall appearance, the Gestalt of a spectrum range; in § 3.5 we see how broader strata of the population perceived spectroscopy. Farmers, meteorologists, and military men, for instance, all were interested in weather forecasting applications of the so-called rainband in the sky spectrum, and we look at some reflections upon their experiences with the handy pocket spectroscope.
Page 22 of 32

Introduction In a sense the end of the conventional depiction of spectra is marked by the rise of photometric methods of registration which had the great advantage of yielding direct information not only about the precise placements of the spectrum lines, but also about their profiles, which formerly could only be inferred from the earlier modes or representation. Because of the great importance of this new technique for twentieth-century research, the main body of my historical discussion ends with a brief chapter on photometric techniques (§ 7.4). Bohr’s atomic model of 1913, and quantum mechanics of 1925 totally transformed the approach toward spectra and spectrum analysis, by introducing theoretical explanations for what hitherto had only been described phenomenologically. Classical spectrum maps gave way to schematic term diagrams and photometric curves. Such radical breaks in the modes of representation also invite reflections about the most appropriate periodization (see the discussion of this point in § 10.8). How compulsory were the preferences for certain modes of representation? And how was the transition from one of these modes to the next made? In the final chapter, I summarize my findings concerning what I refer to as a ‘visual culture’ of spectroscopy from the late nineteenth century. Dominated by the search for patterns, visual representations were handled very much in the spirit of a morphologically oriented natural history. I then contrast the older and later modes of dealing with and representing the spectrum. 1.6 Acknowledgments
Page 23 of 32

Introduction
First and foremost, I have to thank the Dibner Institute for the History of Science and Technology in Cambridge, Massachusetts, most notably its directors, Jed Z. Buchwald and Evelyn Simha, for providing me with a Dibner resident fellowship in the academic year 1996/97. Their warm hospitality, the intellectual stimulus, and the helpful support from the technical staff at the institute facilitated concentrated work during the early phase of this project. Secondly, the American Institute of Physics funded a research trip to their archive in the Niels Bohr Library, which proved a rich source of both written and visual material. I am very grateful to the staff and its director Spencer Weart for helping me to make full use of their holdings. A substantial amount of time was spent in the archives at Harvard University and MIT, and I am indebted to the staff members for their enduring help with my many inquiries. After my return to Göttingen in 1997,I continued working on this book while teaching as assistant professor at the Institute for History of Science of the Georg-AugustUniversity. In particular, I would like to thank Elizabeth Eck, Ralf Haubrich, and Gerhard Rammer for their support. It would lead too far afield to list all the other institutions that I have visited or contacted—the list of abbreviations of the archives, from whose holdings I quote in the following, will have to suffice. Not included in the mentioned list (p.19) are the libraries from which I obtained the texts listed in the bibliography. The majority of the references were traced in the following libraries: Staats- und Universitätsbibliöthek Gottingen; Technical University, Berlin; Widener Library, Harvard; MIT Hayden Memorial Library and Burndy Library; Library of the Harvard/Smithsonian Center for Astrophysics (the latter four in Cambridge, Mass.); Library of Congress, and Library of the Naval Observatory, both Washington, DC, and finally the Niels Bohr Library of the American Institute of Physics at College Park, Maryland. The staff at Gottingen and at the MIT Humanities Library also had to bear with my numerous interlibrary loan orders for the remaining texts.
Page 24 of 32

Introduction
A substantial part of my study hinges upon access to relics of the past of yet another kind: historical instruments and their various material recordings of spectra, i.e., photographs, drawings, prints, and other forms of unpublished visual documentation. The most exciting set, relating to the work of Henry Draper, was made available to me by Deborah Warner at the National Museum of American History. I thank her as well as her colleague Peggy Kidwell very much for having shown me so many of these treasures in their instrument and photograph collections. The keeper of historical collections at the MIT Museum, Michael Yeates, was also very helpful in showing me the remaining visual documentation of MIT’s physics and chemistry laboratories in the late nineteenth century. At Wellesley College Archives, Mrs Wilma R. Slaight and her colleagues were very helpful in providing me with materials pertaining to Sarah Frances Whiting’s teaching of spectroscopy, a superb collection of student material. At the Royal Observatory in Edinburgh, Deputy Librarian Shona McEachern, her assistant Dawn Anderson, and the librarian A.R. MacDonald very kindly gave me access to the archival holdings relating to the work of Charles Piazzi Smyth, which includes his correspondence, notebooks, spectrum drawings, and photographic plates. Staff members at the archives of the École Polytechnique in Palaiseau near Paris kindly showed me various unpublished course manuals and practice session exercise lists, and librarians of the Observatoire de la Côte d’Azur and of the Lick Observatory Archives helpfully sent requested copies of materials and answered other questions.
Concerning the printing techniques used in the production of Fraunhofer’s solar spectrum map, Mrs Marjorie Cohn, Curator of Prints and her colleagues at the Fogg Art Museum, Harvard University, were so kind as to examine this plate microscopically and to discuss the manufacture of the print with me in May 1997. The archivists at the Liverpool Record Office tracked down two obituary notices on George Higgs in local newspapers, and Professors Alan Bowden, David Edwards, and Martin Suggett in Liverpool kindly pointed me to further information about the man. The staff of the Massachusetts Institute of Technology Special Archives, in particular Ms Fran O’Donell, were very helpful in locating a set of Crew’s spectrum photographs among their holdings and providing me with a reproduction; likewise, staff members of the Wolbach Library at the Harvard-Smithsonian Center for Astrophysics provided me with photographs of plates from Higgs’s spectrum atlas.
Page 25 of 32

Introduction

Furthermore, many historians of science and technology have helped and supported me while I was writing this text, either by constructive criticism of selected parts of earlier versions, or by contributing good ideas and helpful hints as to what might be interesting to incorporate. While acknowledgment is given in the annotation wherever it was possible to nail down such help to specific points, let me just list those to whom I feel especially indebted: Bruno Belhoste, Hermann A. and Mary T. Brück, Olivier Darrigol, (p.20) Michael Hoskin, Kevin L. Johnson, Andreas Kleinert, Jost Lemmerich, Andrea Loettgers, Falk Müller, Kathryn M. Olesko, Donald Osterbrock, Alan Shapiro, and Roger Stuewer.

What would a book on spectrum representations be without color pictures? I am indebted to the Georg-Agricola-Gesellschaft for a generous grant which enabled me to include four color plates. I would also like to thank Sonke Adlung, Anja Tschortner, and Richard Lawrence for their help in preparing the manuscript for the press.

Parts of this book have already appeared in the form of journal articles. For the permission to reproduce these texts in extended and updated form in this book, I am indebted to Science History Publications (for parts of Chapter 6) Taylor and Francis (for parts of § 4.3 and 4.5) and Birkhauser Verlag (parts of Chapter 9).

Last but not least, I want to thank my wife, Ann, for all her support, ranging from assistance in archives to correction proofreading, translation of the French and German quotes, and revision of my English for this monograph.

1.7 Archival abbreviations The following abbreviations are used for more frequently mentioned institutions. I am grateful to the archives and libraries whose unpublished sources I have cited for granting permission to reproduce the relevant passages and illustrations from among their holdings.

Page 26 of 32

AASP: Archives de l’Acadèmie des Sciences, Paris (Mme Christiane Demeulenaere-Douyére) AEP: Archives of the École Polytechnique, Palaiseau near Paris AIP: Niels Bohr Library, American Institute of Physics, College Park, Maryland ANP: Archives Nationales, Paris ArP: Archives de Paris, Direction des services (Mme Marie-Andrée Corcuff) BNC: Bibliothèque Nationale, Département de Cartes et Plans, Paris

Introduction

Page 27 of 32

BNE: Bibliothèque Nationale, Département des Estampes et de Photographie, Paris BNT: Bibliothèque Nationale, Tolbiac, Paris BöB: Basel, Offentliche Bibliothek CUL: Cambridge University Library, Cambridge, England DMM: Deutsches Museum, München HUA: Harvard University Archive, Cambridge, Massachusetts HUBL: Handschriftenabteilung der Universitätsbibliothek Leipzig JHUA: The Johns Hopkins University Archive, Baltimore, Maryland LAB: Landesarchiv Berlin LOC: Library of Congress, Manuscripts Division, Washington, DC MITA: Massachusetts Institute of Technology Archive, Cambridge, Massachusetts MITM: Massachusetts Institute of Technology Museum, Cambridge, Massachusetts NMAH: National Museum of American History, Smithsonian Institution, Washington DC RAS: Royal Astronomical Society, London RCW: Royal Collection, Windsor Castle ROE: Royal Observatory Edinburgh RS: Royal Society, London RSE: Royal Society, Edinburgh SAdK: Stiftung Archiv der Akademie der Kiinste, Berlin SPK: Staatsbibliothek Preussischer Kulturbesitz, Berlin

Introduction
SUBG: Handschriftenabteilung der Staats- und Universitatsbibliothek Gottingen WCA: Wellesley College Archives, Wellesley, Massachusetts WCSP: Wellesley College Library, Special Collections, Wellesley, Massachusetts.
Notes: (1) For overviews of more recent work on representation in science, see, e.g. Lynch and Woolgar (ed.) [1988], Mazzolini (ed.) [1993], Rheinberger [1994], and Pang [1997b].
(2) See, e.g., Edgerton [1984], [1991] chap. 7 on Galileo’s representations of the Moon’s surface, or Kemp [1990] on the history of linear perspective.
(3) See, e.g., Ferguson [1977], [1992], Vincenti [1990] as well as Latour [1986]. Cf. also Arnheim’s [1969] earlier claims about intuitive thinking (anschauliches Denken), and Kaufmann [1980] for a concise survey of the various psychological theories of human cognition.
(4) See, e.g., Knorr-Cetina in Lynch and Woolgar (ed.) [1988], Dennis [1989], and Harwood [1989] on Hooke’s Micrographia, Schaffer in Gooding etal. (ed.) [1989], Hetherington [1988] on various astronomical issues, Secord [1986] on the Cambrian-Silurian dispute, and Rudwick’s studies of other geological controversies.
(5) See, e.g., Darius [1984] for an anthology of nice samples, Thomas (ed.) [1997] for some well-illustrated analytical essays, Schaaf [1979]-[1992] for pathbreaking studies on scientific photographers like John Herschel, Fox Talbot, and Piazzi Smyth, and Schaaf (ed.) [1994], [1996] for editions of pertinent primary materials.
(6) Compare, e.g., Crone [1953] with Woodward (ed.) [1975], [1987] for the change in introductory texts on the history of cartography; cf. Woodward [1974], Blakemore and Harley [1980] for historiographic surveys.
(7) I am thinking of Rudwick [1976] on the geosciences, Blum [1993] on American zoology, Pang [1994/95], [1996] on solar eclipses, and Pang [1997a] as well as his contribution in Lenoir (ed.) [1998] on stellar photography.
(8) On statistics see Funkhouser [1938], Royston [1956]; on botany see Blunt and Steam [1951], Nissen [1951], and recently Nickelsen [2000]; for more general surveys cf. also Nissen [1950], Ford [1992], Mazzolini (ed.) [1993].
Page 28 of 32

Introduction
(9) On the definition of ‘transparency’ in this epistemological sense see the contributions by David Gooding and Thomas Nickles, in Gooding et al. (ed.) [1989] pp. 14f., 216f., 302f. 31 Of.
(10) For a glossary of these and several other terms see Funkhouser [1938] pp. 364–8. For a critical survey from the point of view of a modern user of these representational devices, see Tufte [1983]. Cf. also Hankins [1999] pp. 52f. on the term ‘graph,’ and on nomograms in particular.
(11) Playfair (1801), quoted from Royston [1956] p. 242. Royston (as well as Shields [1938]) suggested that Play-fair transferred this idea of plotting graphs from the context of steam engine production. He had been employed as a draftsman by Boulton & Watt since 1780, which at that time was already using indicator diagrams to test the efficiency of their machines. For a contextual analysis cf. Brain [1996] chap. 1.
(12) On Playfair’s reception see, for instance, Funkhouser [1938] pp. 292ff.
(13) Quote from Tilling [1975] P- 194; cf. also Shields [1937]. The earliest examples found relate to thermometer and barometer readings recorded by Rømer and Lambert.
(14) See Krohn [1991] p. 197, and footnote 9 above on epistemological ‘transparency’. Cf. also Lynch’s or KnorrCetina’s shop-talk analysis, confirming this point that researchers ‘see’ their objects, rather than images thereof.
(15) See here footnote 3.
(16) F. Hund in an interview for the Archive for History of Quantum Physics (AIP), session of 26 June 1963, transcript p. 13. Unless otherwise indicated, all English translations in this book are by Ann M. Hentschel.
(17) On Piazzi Smyth’s wide-ranging talents and his emphasis on visual aspects, see Warner [1983], where he is portrayed as an ‘astronomer-artist’, as well as Schaaf [1979b], [1980/81], Thomas (ed.) [1997] pp. 88–91.
(18) See Smyth [1843/46] for detailed criticism of various published illustrations of nebulae by William Herschel (1811, 1814) and John Herschel (1834); he also comments there on his mezzotint techniques for rendering Maclear’s observations of Halley’s comet in 1835. Smyth’s paper is reprinted, with annotation and illustrations added, in Hentschel and Wittmann (eds.) [2000]. According to Warner [1983] p. 113, Piazzi had “apparent authority on the processes of engraving, aquatinting and mezzotinting”, and from 1845 on also in lithographing.
Page 29 of 32

Introduction
(19) See Galison [1997]; cf. DeVorkin [1985] and Smith and Tatarcwicz [1985] on the technological history of these devices, and Edgerton and Lynch [1988] on the aesthetics of digital image processing.
(20) See. e.g., Kirchhoff [1863] vs. Stewart [1863], Diacon [1867], Stokes [1876]. Kayser [1900], [1909], [1911], Junkes [1962], Dingle [1963], Brand [1995]. Cf. also James [1985a] for a critique of this Whig historiography.
(21) See Kayser [1900], [1902], [1905], [1908], [1910], [1912], [1924], [1930], [1932], [1934]. For the background of Kayser’s handbook, see also his autobiography [1936]. About his life and work see Crew [1941], Freiesleben [1973], and here pp. 244 and 365.
(22) See, e.g., McGucken [1969], Sutton [1972], [1976], [1986], Maier [1964/81]; James [1983], [19856], [1986], [1988].
(23) See James [1981], [1995], Lankford [1981], [1997], DeVorkin and Kenat [1983], and James [1985a], Hentschel [2000] for historiographic surveys.
(24) See Forman et al. [1975], Weart in Reingold (ed.) [1979] pp. 300f., and Lankford [1997]. On Rowland’s “school of light” specifically, see also Kargon [1986] and Sweetnam [2000] chaps. 1, 3, and 8.
(25) Pang [1997b] p. 155 makes this point in his useful literature survey on visual representation, citing Edgerton [1984], [1991], Winkler and van Helden [1992], Lankford [1981], Rothermel [1993] and others.
(26) See, e.g., Eder [1945], Lankford [1984], Schaaf [1990], [1992], Thomas (ed.) [1997] as well as other references here in footnote 163, p. 213.
(27) See Hunt [1844a], [1852], Abney [1874], [1878c], Eder [1884], Mees [1961]. The short remarks on the development of photochemistry found, e.g., in H. and A. Gernsheim [1955] chap. 23, for instance, are mostly based on these pioneering texts.
(28) See, e.g., Hetherington [1988] chap. 5 and further sources cited there on pp. 135f. Bartholomew [1976].
(29) See Pang [1994/95]; cf. also Becker [2000].
(30) On solar-eclipse expeditions, see Pang [1996]. On the coronagraph, see Hufbauer [1994].
(31) See Pang [1997] and here Chapter 4, § 6.7–6.8.
(32) Arthur Schuster to Charles P. Smyth, 14 July 1882 (ROE, 15.67, folder S); the plate in question is most likely pi. xxxiii illustrating Huggins [1868a].
Page 30 of 32

Introduction
(33) Quote from Whitfield [1995] p. ix; on the following see idem, p. 67.
(34) On this point see, e.g., Edgerton [1975] and Whitfield [1995] pp. 1, 61–3, quoted in the following main text. Strangely enough, Ptolemy himself designed and probably made a celestial globe, but never took the step of using his projection method to create a two-dimensional map. On another such transfer, see here footnote 11. p. 2.
(35) On the use of Ramsden’s graduated circle and engineer’s chain in the 1784 survey from London to Dover see Brown [1949] pp. 257–9; on Fraunhofer and Utzschneider’s theodolites see Jackson [2000] and here p. 36.
(36) Just to give one example here: “The process of mapping infra-red spectra […] is, at best, a slow and tedious one. […] The work involves two distinct kinds of activity, viz. mapping the spectra, by means of a series of curves, and studying them.” Coblentz [1905] pp. 4, 15.
(37) Alpers [1983] p. 147; cf. also here § 10.10 on ‘Taking the mapping metaphor seriously’.
(38) See, e.g., Snyder [1993], Monmonier [1991] chap. 2, [1995] chap. 1 on the controversy over the Peters and Mercator projection.
(39) Edney [1993] p. 57.
(40) See Jenkins [1998] p. 190. Cf. also the Oxford English Dictionary, 2nd edn, vol. 9, pp. 348–51 on its etymology, with the first documentable use of ‘mappe’ in English dated to 1527.
(41) Langley [1882b ]p. 587.
(42) Quote from Mulkay and Edge in Lemaine (ed.) [1976] p. 163. Quite in line with the analogy, Langley’s infrared map does not distinguish between solar and terrestrial spectrum lines, and as we shall see on p. 268, Langley even had a hard time distinguishing between ‘real’ lines and the artefacts of his new instrument.
(43) George Bellas Greenough (1841), quoted by Secord [1986] p. 29.
(44) Gombrich [1975] pp. 127, 146.
(45) On the rhetoric of visual representations in science see. e.g. Lynch and Woolgar (ed.) [1988], or Latour [1986]; for cartographic examples cf. Harley [1989], Monmonier (1991] on “how to lie with maps”. or Monmonier [1995] chap. 5 on the use of maps in the continental drift controversy.
Page 31 of 32

Introduction
(46) See, e.g., Felix Klein in Klein and Riecke (ed.) [1904] or Herbert Mehrtens, Moderne Sprache Mathematik, Frankfurt: Suhrkamp, 1990, pp. 60–84 on Klein’s preference for mathematical models as visualization aids; or Suzanne L. Marchand, Down from Olympus. Archeology and Philhellenism in Germany, 1750–1970, Princeton University Press, 1996, pp. 142–51, on the increasing importance of Anschauung and visual aids such as maps, posters, and slides, in the teaching of the arts, classical history, and archeology (my thanks to Kathryn Olesko for pointing out this latter cultural parallel). (47) See, e.g., Ferguson [1977] p. 835 or [1992] chap. 5 for examples of such sets of machine ‘elements’ used in the contexts of teaching and museums. (48) See Daston and Galison [1992], Galison [1998], and here the discussion in § 10.8 on p. 450. (49) See Hentschel [1993], [1996], [1997c], [1998] for the results of these studies dating back to 1990. (50) This is, of course, an allusion to Rudwick [1976]. (51) The studies by Courboin [1914] and Dyson [1984] provide a useful orientation for engraving in the contexts of the French and British fine arts in the eighteenth and nineteenth centuries. See also Wakeman [1973] about Victorian book illustration, and K. and A. Hentschel [2001] for an in-depth case study on one Parisian engraver. (52) The history of the various types of spectroscopes as scientific instruments is a quite neglected area of historiography—see Bennett [1984], Hearnshaw [1986] chap. 1, and Wolfschmidt [1998] for brief surveys, as well as Austin [1993], Warner [1993], for more detailed studies on particular types of spectroscopes.
Page 32 of 32

The Spectrum in Historical Context
Mapping the Spectrum: Techniques of Visual Representation in Research and Teaching
Prof. Dr. Klaus Hentschel Print publication date: 2002 Print ISBN-13: 9780198509530 Published to Oxford Scholarship Online: January 2010 DOI: 10.1093/acprof:oso/9780198509530.001.0001
The Spectrum in Historical Context
Klaus Hentschel DOI:10.1093/acprof:oso/9780198509530.003.0002
Abstract and Keywords This chapter presents a historical survey of the analysis of spectra, starting with Leonardo da Vinci's observations of spectra from a glass of water. Newton's ‘new theory of light’ is treated along with his introduction of the prism as a research instrument. The long dominance of the Newtonian color scheme is discussed. Wollaston's and Fraunhofer's discoveries of dark lines in the solar spectrum are documented with hitherto unpublished material from the Fraunhofer papers in Munich and Berlin. William Herschel's thermography and John Herschel's maps of photographically obtained traces of spectra are covered as well as new instruments for exploring heat radiation, such as Nobili's and Melloni's thermopiles based on the thermoelectric effect.
Keywords: Leonardo da Vinci, Isaac Newton, Joseph Fraunhofer, William H. Wollaston, William Herschel, John Herschel, Leopoldo Nobili, Macedonio Melloni, thermography, thermopile
2.1 “The Phenomena of Colours”
Page 1 of 81

The Spectrum in Historical Context
The rainbow, or the natural decomposition of sunlight into its spectral components, has been a source of fascination for all human cultures.1 But concerted experimental investigation of this phenomenon is of a much younger date, partly because reasonably clear glass or other translucent dispersive media are needed to generate a spectrum.2 Glass of such quality was first manufactured in Roman times. In the first century AD the natural philosopher Lucius Annaeus Seneca first compared the colors of the rainbow with those created by con-ically shaped glass rods. During the twelfth and thirteenth centuries, interest in optics was revived again at Islamic centers of learning, soon to be followed by Christian ones. The Polish monk Witelo used translucent hexagonal crystals of quartz to examine color phenomena, covering three of the six surfaces in opaque wax. More commonly, water-filled globes were used. The German Dominican Dietrich von Freiberg, the English Franciscan Roger Bacon, the Egyptian astronomer and mathematician Ibn al Haita̠m, and the Persian Kamāl al-Dīn al Fārisī all used such artificial raindrops to study the distribution of colors.3 In the ninth century Venice became famous for its highly translucent, untainted cristallo glass—it is probably no accident that spectacles, or eyeglasses, were invented in Italy around 1300.4 The technology of eyeglass production (initially convex lenses, 200 years later also concave ones to correct myopia) spread to other regions, notably the Netherlands where the telescope and microscope were invented around 1600. One of the first systematic experiments to include not only brief verbal descriptions but also a drawing of the spectrum was conducted by Leonardo da Vinci (1452–1519).5 In one of his notebooks he described “rainbow colors” formed by the edges of air bubbles in a transparent glass of water. By observing these colors in direct transmission, Leonardo was able to exclude any influence of the eye in the projection of these colors onto the floor (cf. Fig. 2.1).
Prismatically shaped pieces of glass had long been known to produce colored light and were used for this purpose, for example, in Venetian lusters. According to the Oxford English Dictionary, the term ‘prism’ derives from the Greek πρἱσαα, meaning: “a thing (p.22) sawn”. In accordance with Euclid’s geometry it is defined as “a solid figure of which the two ends are similar, equal, and parallel rectilineal figures, and the sides parallelograms.” The first documented use of the term in optics is found in Henry Peacham’s Gentlemans Exercise of 1612, where he mentions a “most pleasant and delightfull experiment […] in a three square cristal prisme, wherin you shal perceiue the blew to be outmost next to that the red.”6 In a German edition of Delia Porta’s (c. 1535–1615) Magia Naturalis which appeared in 1715, the term ‘Prisma’ was still treated as an unusual terminus technicus typeset in italics, carrying the definition “dreyecktes GlaB”.7
Page 2 of 81

The Spectrum in Historical Context

Thus it was only during the

second third of the seventeenth

century that these curious

triangular pieces of glass

—“fool’s paradise” as they were

then also referred to because

they “transform the colours of

things into a thousand shapes”8

—advanced from a curiosity, a

mere toy, to a standard

instrument of research. The Bohemian professor of medicine Jan (p.23) Marcus Marci (1595–1667),9 for instance, used prisms to simulate the rainbow. Thirty-six of the theorems in his book Thaumantias from 1648 were devoted to the “iris trigonia”, the ‘prismatic rainbow’. He noticed the one-to-

Fig. 2.1. Two drawings of spectrum observations by Leonardo. Left: spectrum observed in transmission through a water glass with air bubbles. Right: spectrum observed in projection upon the floor. From a Leonardo manuscript (RCW, no. 19150 r), reproduced by permission: The Royal Collection © 2001, Her Majesty Queen Elizabeth II.

one correspondence between

spectral color and angle of refraction.10 But these acute observations were

interpreted according to quite traditional premises rooted in the Aristotelian

conception of color as constituting a specific mixture of light and darkness.

White light was a simple substance, whereas the spectral colors were

aberrations, or imperfections, possibly related to a change in the light’s density

(condensatio). Consequently, Marci’s explanation for the appearance of the

various colors in a beam that had passed through a prism was a difference in the

degree of condensation, caused by the tapering thickness of the piece of cut

glass.11 Why color fringes appeared preferentially along the edges of lenses or

prisms was settled by the supposition that there the ‘colored rays’ (radü

colorigeni) were less often superimposed by the colorless and more intense

image-generating rays (specie objecti)12 A similar interpretation of color as a

condensation phenomenon is also found in Giambattista della Porta’s treatise De

Iride et Colore of 1593. Figure 2.2 shows rays of light, emitted from the Sun (G),

traversing different thicknesses of glass, with violet emerging at R, blue at S,

green at T, and red at V. (The parallelism of the refracted rays reveals Della

Porta’s ignorance of dispersion!)

Page 3 of 81

The Spectrum in Historical Context

(p.24) In his Dioptrique of

1637 René Descartes (1596–

1650) used prism-shaped glass

for his precise determination of

the index of refraction of glass

(cf. Fig. 2.3). A fine pencil of

light was generated by means of

holes A and L through two

wooden props mounted at E and

F. The glass sample was cut into

the shape of a triangular prism

and placed against the second

prop. Because its face QR was

Fig. 2.2. Delia Porta’s conceptualization

flush against the prop, no

of color generated by refraction through

refraction occurred where the

a prism ABCE. The density of the light is

light entered the prism, and the altered variously as it traverses different

refracted ray along the line BI

thicknesses of glass. From Della Porta

clearly indicated the refractive

[1593] p. 223.

power of the glass: the higher

its refractive index, the shorter

the distance PI. The point of refraction R was quite close to the tip of the prism

because that way the light ray was less likely to encounter inhomogeneities in

the glass and would be less strongly absorbed. No scale is given in this

illustration but the distance FL cannot have been much larger than 5 cm

because of the inherent limitations in the contemporary production of

homogeneous glass.13

Fig. 2.3. Descartes’s apparatus to measure the index of refraction. From Descartes [1637] p. 137 (reproduced in Descartes [1982] vol. 6, p. 212).

Page 4 of 81

The Spectrum in Historical Context
This experiment was designed to obtain just one value for the refractive index. Quite in line with this, Descartes also used a single index of refraction for water, namely, 187 : 120, for his famous calculations of the rainbow in another essay illustrating the fruitfulness of his treatise on method, Les Météores, which appeared in the same year. Nevertheless, he must have noticed that various colors were refracted slightly differently. This is confirmed by another illustration in Les Météores where he described a prismatic experiment (cf. Fig. 2.4) in which solar light is directed through a prism NMP and a spectrum HF is projected onto a surface perpendicular to the slit DE parallel to the second face of the prism. In this illustration, the divergence of the rays coming from both limbs of the Sun is greatly exaggerated: practically all of them hit face NM of the prism orthogonally, and the light refraction only occurs upon leaving the prism. Descartes noted that no clear spectrum appeared if the opening (I'ouverture), or slit, as we would say, between D and E was too large. He inferred that, unlike normal light refraction, the spectral colors could only be formed along a dark border or edge. According to his mechanistic model of light, the spectrum colors were generated by a kind of spin imparted to the light particles hitting the border zone by E or D of the slit: those hitting the left end D rotated in one direction and thus exhibited one color (red) at F, while the others were turned (p.25) the other way round by the surface at E (and hence were transformed into blue-violet at H). The dashed middle ray hits neither border of the slit, but is nevertheless refracted to the point G. Thus the yellowish area near G was the effect of normal refraction of the Sun’s rays. Descartes’s mechanistic model of light was needed to explain why “darkness or a boundary to the light is necessary”14 for creating the other colors of the spectrum such as red and blueviolet at the areas F and H, respectively. Color conceived as a superimposed rotation (vertigo) is also found in contemporary texts of other natural philosophers such as Thomas Hobbes (1588–1679), who may possibly have been inspired by Descartes’s model of tournoiement.15
Page 5 of 81

The Spectrum in Historical Context

While occupied with the

“grinding of Optick glasses

other than spherical” in the

mid-1660s, Isaac Newton

(1642–1727) became intrigued

by the phenomonon which we

now call ‘chromatic aberration’:

even the most perfectly shaped

lens does not focus all colors of

a well-defined beam of light at

the same spot, and this greatly

reduces the definition of the

image created by refracting

telescopes. A first look at a

bicolored thread through a

cheap prism reportedly bought

at a fair, already revealed to

him that the blue rays were refracted more than the red ones. Newton then initiated a series of more systematic experiments, using a triangular glass prism inserted in a small hole in the window shutter of a darkened room, to “try

Fig. 2.4. Descartes s prismatic experiment with solar light. Coming from ABC, the ray is projected vertically onto face NM of the prism, refracted from the opposite face, and falls onto the screen PF. From Descartes [1637] p. 224 or [1982] vol. 6, p. 330.

therewith the celebrated

Phenomena of Colours”.16 Having studied the (p.26) optical theories of

Descartes and Hooke, he had doubts about the accepted modification theories of

color, and his experiments amplified this skepticism. Quite in line with his

general inclination towards an atomistic conception of matter, Newton

conceptualized white light as a heterogeneous mixture of light globuli.

According to his mechanistic model, which is documented in various

manuscripts but not publicized until later, these globuli, distinguished by their

different refrangibilities, stimulate the human eye to see specific colors. In his

‘New theory of light and colors’, which he first published in 1672, Newton was

intent on restricting his presentation to propositions gathered from phenomena

by induction. He thought he had shown conclusively that a white pencil of light

contains rays of various colors that are separable by refraction in prisms,

raindrops, or more generally, at the interface between two media of different

refractive index. According to his mechanistic model, prisms thus act like

separators, and colored glass or other colored matter act like filters, admitting

only certain light globuli through while blocking others out. To please the induc-

tivist Baconian spirit of the newly founded Royal Society, and also to immunize

his findings against potential criticism, Newton omitted most of these

Page 6 of 81

The Spectrum in Historical Context
mechanistic assumptions, however, and decided to present the experiments in a way that made his theory emerge as an ‘inevitable’ outcome.17
The term ‘spectre’ conjures up notions of phantoms and unreal objects of thought. It was also used to refer to the apparitions exhibited in the then popular camera obscura—not unlike the darkened chamber Newton used for his investigations. Another root of the term ‘spectrum’ introduced by Newton to describe the colorful oblong image of the Sun produced by his prism (colorum prismatis) in his article on the ‘New Theory about Light and Colours’ of 1672 is the Latin verb specto for seeing or watching.18 By varying the distance between a moveable screen and the prism, he verified that the rays continued to propagate along straight lines and not in some curved manner, as Descartes’s model of color deriving from a spinning motion of light globuli would have suggested. Placing an inverted prism right next to the first prism restored a circular image of the entrance hole on the screen. This indicated that the spectrum was not due to imperfections in the prism, in which case it would have elongated further.19 When he generated a spectrum at least five times as long as it was wide with a good prism positioned up to twenty-two feet away from the screen, he was able to demonstrate that no further ‘division’ of the colors was possible after the ‘analysis’ of light in the first prism: Similar to a chemical reduction which, once completed, cannot yield anything more elementary, when light of a selected (p.27) color traverses two or more prisms in sequence there is no further alteration in the color—at least in principle, i.e., if the experimental procedure is followed rigidly and skillfully.20 Not anticipating the many problems his contemporaries would have in replicating precisely this experiment, around 1672 Newton saw it as a particularly striking demonstration of the elementary nature of colored light and the composite nature of white light.
Light is a confused aggregate of Rays endued with all sorts of Colors, as they are promiscuously darted from the various parts of luminous bodies. And of such a confused aggregate, as I said, is generated Whiteness, if there be a due proportion of the Ingredients; but if any one predominate, the Light must incline to that colour […].
For, of the Rays, constituting the incident light, since those which differ in Colour proportionally differ in Refrangibility, they by their unequal refractions must be severed and dispersed into an oblong form in an orderly succession from the least refracted Scarlet to the most refracted Violet.21
This experiment flatly contradicted—and in Newton’s opinion refuted—the common conception of normal white light as elementary and colors as its derivatives. So Newton called this the ‘crucial’ experiment in his earlier optical papers, which claim he later retracted, however, because of its controversial nature.22
Page 7 of 81

The Spectrum in Historical Context

(p.28)
Not all of these colored rays could satisfy Snel’s law of refraction with a given medium and one index of refraction (or as Newton put it, with one degree of refrangibility). Newton could resolve this apparent paradox by showing experimentally that each color on its own fulfilled Snel’s law. The price he had to pay was varying the index of refraction with the color of each ray. For instance, in his calculations of the rainbow, it ranged between an approximated 108 : 81 for red rays and 109 : 81 for violet ones, as a consequence of refraction in water droplets.23 As described most completely in Newton’s second paper on light and colors from 1675, prompted by Hooke’s persistent criticism, but also in prop. II, prob. I of book one of his later Opticks, he saw an analogy between the perception of light by the eye and the perception of sound by the ear. In his earlier lectiones opticae of 1670–72, he had divided the spectrum into five principal colors. Now he added orange and indigo, and argued that this doctrine of seven primary and simple colors, namely red, orange, yellow, green, blue, indigo, and violet, was also based on this harmonic analogy.24

Fig. 2.5. Newton’s experimentum crucis with two prisms in sequence. The light from a small circular opening F in the window-shutter is led through a condenser lens onto a fairly large prism ABC. The spectrum formed is projected onto a vertical screen DE and a small part of it is thrown onto a second screen at de, in which a small circular hole g admits rays from only one prismatic color onto another prism abc. The dispersed light is then projected onto the back wall MN and does not show any further dispersion, regardless of the color admitted through the hole in the projecting screen. Diagram from Newton [1704b] p. 47.
Fig. 2.6. A later variant of Newton’s experimentum crucis from his correspondence (1721): The sketch differs from the schematic textbook presentation in that it depicts a condenser lens in front of the first prism instead of a second screen (DE with hole G) right after it.

Page 8 of 81

The Spectrum in Historical Context

(p.29) As Fig. 2.7 illustrates, Newton tried to interpret this color sequence as analogous to the diatonic scale in music: d–e (whole step), e–f (half step), f–g (whole step), g–a (whole step), a–b (whole step), b–c (half step) c–d (whole step). Accordingly, orange and indigo were semitones of smaller extension than the other primary colors in the spectrum.

Newton’s representation of the

spectrum, which prevailed for

the next 100 years, was a

“parallelogram of light, with

circular ends, in which the

seven colours gradually shaded

into each other without any

interruption”.25 Figure 2.8

shows how Newton’s parallelogram representation harmonized the continuous color change with a presumed finite set of primary colors. Superposed circles indicate the respective positions of the primary colors in the continuous spectrum. The

Fig. 2.7. Newton’s optical analogue to the diatonic scale. The relative elongation of the various color ranges according to Newton’s harmonic analogy (from Newton [1704b] p. 127), juxtaposed with a schematic comparison keyboard: the two halftones EF and BC correspond to the smaller intervals μκ and δβ.

smaller the circular hole is

through which the solar light is

screened, the less two adjacent circles overlap and the better the

“heterogeneous Rays of compound Light” are separated.

Page 9 of 81

The Spectrum in Historical Context
Yet there is a strange inconsistency in this representation. A sufficiently small hole will cause the overlapping area between any two adjacent circles to vanish. But the spectrum exhibits intermediary colors between the primary colors, not white gaps. Although all these intermediary colors are interpreted as a superpositioning of two (or more) primary colors, the lower parallelogram representation fails to reflect this. In his Opticks26 Newton did his best to smooth over this point and, astonishingly, none of his many followers seems (p. 30) to have noticed this incongruency between the conceptual and representational assumptions. In principle—if we ignore diffraction effects for a moment—this homogenization of the spectrum from a reduction in the size of the aperture can be carried on indefinitely. It would then actually imply an indefinitely large number of different rays rather than a finite set of primary colors. Although this point was made by one of France’s foremost Newtonian physicists, Jean Baptiste Biot (1774–1862), in his treatise on experimental and mathematical physics,27 many of Newton’s epigones were quite evidently misled by the rhetorics of the parallelogram representation. To give just one example: the Göttingen naturalist Johann Christian Polykarp Erxleben (1744–1777) took Newton’s visual scheme at face value:
The colorful image is composed of as many circles as there are colors in it, one of which is red, another orange, etc., the last violet, which fuse into each other in the colored stripe. Each of these circles is an image of the Sun in light of a different refrangibility, so they cannot all fall on a single spot. But just because these circles or images of the Sun are so large, they fuse into each other and so one can make them smaller by holding a raised glass between the prism and the hole in the window shutter; then each simple [ray of] light separately assumes the form of small round disks rowed on top of each other.28
Erxleben even went about illustrating this idealized separation of the seven primary colors in a figure (see below) for his influential textbook, and it withstood Georg Christoph Lichtenberg’s many revisions. Erxleben’s unfortunate diagram bore the brunt of Johann (p.31) Wolfgang von Goethe’s (1749–1832) derision, whose own alternative conception of light and color in his Farbenlehre is beyond the scope of this book.29
In [Erxleben’s] figure, now, the seven rays are placed nice and neatly on top of each other as seven circlets, just as though some one had once seen them like that; the connecting dashes that Newton had wisely always added are left out.
Page 10 of 81

The Spectrum in Historical Context

Newton’s visual scheme continued to dominate visual representations of the spectrum throughout the eighteenth century. In 1808 the polytechnicien Jean Henri Hassenfratz (1755–1827) still depicted the absorption spectra of various colored glass plates and liquids interposed in a narrow beam of sunlight as overlapping circles of primary colors (cf. Fig. 2.10). Even though he declared in the introduction to his long paper that white light actually consists of an infinite number of colors, not just seven or three, his visual representation of his findings remained confined within the strictures of this Newtonian divisioning into primary colors.30

Fig. 2.8. Newton’s parallelogram representation of the solar spectrum, a superpositioning of circles indicating the refraction of primary colors. The lower diagram illustrates improved color separation with smaller hole diameter. From Newton [1704b] p. 65.
Fig. 2.9. To the right of Goethe’s [1810b] § 246 comment, (Leopoldina edn, ser. I, vol. 5, p. 92): Erxleben’s misleading representation of seven homogenized primary colors, from Erxleben [1772a–c] pl. VI, fig. 75.

Despite the persistence of Newton’s visual scheme, around the turn of the eighteenth and nineteenth centuries his hypothesis about the seven primary colors came under renewed attack by advocates of a three-color theory.31 It was even harshly censured for “the mystic number seven, the child of judicial astrology”, being “devoid of foundation, and too plainly betray[ing] a tincture of the mysticism of the age”.32 Others, most notably Hassenfratz, were eager to point out that “la théorie de l'immortel physicien anglais” only intended to divide the spectrum into seven parts for practical reasons, and that Newton actually considered the spectrum as composed of an infinitude of different colors that fuse into each other in infinitely fine nuances.33 All the same, the visual representations of the spectrum remained comparatively uninteresting, for no further structure was distinguished apart from the color sequence from red to violet. This changed only in the early nineteenth century, but as we shall soon see, the transition was gradual and the Newtonian heritage of analyzing the spectrum primarily in terms of color theory prevailed for quite some time. Nice evidence (p.32) for this thesis is provided by Johann Benedict Listing (1808–1882), a Göttingen specialist in physiological optics, who as late as 1866 still defended a standardized division of seven color regions in the visible part of the spectrum.34

Page 11 of 81

The Spectrum in Historical Context

The longevity of the Newtonian

color scheme is a good example

of a phenomenon which we

shall encounter repeatedly

throughout this study, namely

the tenacity of representational

forms despite drastic

transformations in the

theoretical and conceptual

contexts from which they had emerged.35

Fig. 2.10. Hassenfratz’s absorption spectra of colored glass represented as

2.2 The dark lines In 1802 a gentleman scientist observed dark lines in the solar spectrum. The retired physician

superimposed circles or ellipses of the seven Newtonian primary colors of light (color code is along the top). From Hassenfratz [1808].

William Hyde Wollaston (1766–

1828), who financed his broad-

ranging experimental researches from his discovery of a practical method for

working platinum, made this observation while conducting an “examination of

refractive and dispersive powers” of a flint-glass prism. He interpreted these

lines as natural boundaries between four color zones:

I cannot conclude these observations on dispersion, without remarking that the colours into which a beam of white light is separable by refraction, appear to me to be neither 7, as they usually are seen in the rainbow, nor reducible by any means (that I can find) to 3, as some persons have conceived; but that, by employing a very narrow pencil (p.33) of light, 4 primary divisions of the prismatic spectrum may be seen, with a degree of distinctness that, I believe, has not been described nor observed before.36

The lines which Wollaston designated A and B defined the boundaries of red, while D and E demarcated the violet from the blue (between C and D). The least clearly defined area, yellowish green, extended from B to C. Two more lines near C could not be ascribed any clear borderline position between the color zones, so Wollaston assigned them the lowercase labels f and g. In the position of minimum refraction in which, as Wollaston put it, “the colours are most clearly divided”, he found that the intervals AB, BC, CD, and DE related to each other in the proportions 16, 23, 36, and 25, but these relations closely depended on the angle of incidence and on the type of prism used.37

Page 12 of 81

The Spectrum in Historical Context
For the first time, someone had noticed an internal structure in the spectrum other than the sequence of rainbow colors. This discovery had been triggered by the combined use of improved glass prisms and a very narrow slit.38 How nontrivial this finding was is perhaps best illustrated by an anecdote reported many years later, after this observation had been repeated independently by Fraunhofer in Munich,39 followed by hundreds of others. By 1823 several lengthy and detailed descriptions of how to generate and observe the famous phenomenon were available, yet even experienced optical experimenters seemed to have trouble, at first, in replicating Wollaston’s and Fraunhofer’s findings. Charles Babbage reported about a visit he paid to John Herschel in his mansion in Slough:
Conversing with Mr Herschel on the dark lines seen in the solar spectrum by Fraunhofer, he inquired whether I had seen them; and on my replying in the negative, and expressing a desire to see them, he mentioned the extreme difficulty he had had, even with Fraunhofer’s description in his hand and the long time which it had cost him in detecting them. My friend then added, “I will prepare the apparatus, and put you in such a position that they shall be visible, and yet you shall look for them and not find them: after which, while you remain in the same position, I will instruct you how to see them, and you shall see them, and not merely wonder you did not see them before, but you shall find it impossible to look at the spectrum without seeing them.”
On looking as I was directed, notwithstanding the previous warning, I did not see them, and after some time I inquired how they might be seen, when the prediction of Mr Herschel was completely fulfilled.40
(p.34) This episode is the more striking since Babbage did not even have to set up the instrumentation himself but was presented with the fully prepared phenomenon. What he lacked was thus not so much instrumental skill in producing the phenomenon as the knowledge about what exactly to look for. Knowing “how to see” the extremely fine irregularly spaced lines at right angles to the spectrum’s color gradient was essential, given the low resolution of Herschel’s home-made spectroscope. Lacking the experimental knack of a Herschel, it was a nearly insurmountable task to reproduce the phenomenon without knowing how to handle prisms, how they are mounted so as to ensure a symmetric passage of light through them, or how to construct a regular and sufficiently narrow slit.41
2.3 Early modes of representing the spectrum
Page 13 of 81

The Spectrum in Historical Context
The earliest representation of the solar spectrum deserving the term ‘map’ was drawn and engraved onto a copper plate by the self-educated optician and instrument maker Joseph Fraunhofer (1787–1826) in 1814/15.42 (See Fig. 2.11 and color Plate I.) Fraunhofer had learnt this printing technique in his youth as a court glazier’s apprentice.43 Like Wollaston, Fraunhofer also attached alphabetical labels to some of the most prominent dark lines in the solar spectrum (from A to H and a to b: see Fraunhofer’s early pencil sketch, Fig. 2.11), but he explicitly rejected the Englishman’s opinion that they were color boundaries and thus moved away from color theory.44 Altogether, Fraunhofer counted a total of 574 lines between B and H alone, but in his published map he depicted only about 350 of them, taking considerable pains to represent accurately not only their relative positions and internal structure (the yellow D line is clearly depicted as a double line), but also to indicate their comparative intensities: compare the bold D or F lines with the very subtle lines right next to them in Plate I (cf. the analysis of the printing techniques for his map on p. 116). On top he added a curve indicating the relative intensities of the various spectral colors.45
PLATE I (next page). Left: Rare handcolored version of Fraunhofer’s 1814 map of the solar spectrum. From the archive of the Deutsches Museum, Munich, map collection, STO 1107, cabinet 39, shelf 03. Right: Published version of Fraunhofer’s map of the solar spectrum, carrying the inscription “ge-zeichnet u. geatzt von J. Fraunhofer” (drawn and etched by J.F.). Original length of both: 36.5 cm. (For detailed commentary on the printing, labeling and the additional intensity curve from Fraunhofer [1815a] pl. II, see here pp. 34ff. and 116.)
Captions of color plates II—IV (preceding pages)
PLATE II. Above: Lithograph (with superimposed colors by iris printing) of the flame spectra of various metallic compounds (cf. here pp. 42f., and p. 124 on the askew color borders typical of iris prints). From Miller [1845]. Below: Characteristic lines in the emission spectra of alkali metals and alkaline earths, with the solar spectrum at the top as reference (cf. here pp. 47f.). Chromolithograph by W. Creuzbauer’s printing house in Karlsruhe, original width 16.6 cm (cf. also fig. 2.18 on p. 47 for the black and white version and further commentary). From Bunsen [1860] plate V.
PLATE III. Above: Samples of eight handcolored drawings of various absorption spectra (cf. here p. 37), approximately original size. From Brewster [1822/23a] plate XXVII. Below: Chromolithograph printed on six stones with various greens to render different line intensities (cf. here pp. 125). From Kirchhoff [1861/62a] plate II.
Page 14 of 81

The Spectrum in Historical Context PLATE IV. Above: The first three of Secchi’s four classes of stellar spectra, engraved by P. Dulos, printed by Sarazin imp. (Paris). From Secchi [1870a] pi. II. Below: Samples of arc spectrum photographs obtained with Lippmann’s process by Hermann Krone in Leipzig. 1892 (cf. here p. 208). Photograph from Deutsches Museum, Munich, Bildstelle. (p.35) Already in an early pencil sketch reproduced below, but also in the black-and-white printed plate, Fraunhofer tried to convey the decrease in overall light intensity by means of shading near both ends of the spectrum (cf. Fig. 4.5, p. 116 for a close-up of the printed version).
Fig. 2.11. Undated handdrawn pencil sketch of the solar spectrum, with all major lines labeled and darkening at both ends of the spectrum indicated by pencil hachure. originally 27 cm long. From DMM. Fraunhofer papers, folder NL 14-52.
Page 15 of 81

The Spectrum in Historical Context
In the main body of the text he simply cited refraction angles for some of the major lines. This dual documentation of qualitative and quantitative aspects was sufficient for the purpose Fraunhofer had in mind when he drew his map. He just needed markers for specific colors in the visible spectrum (using the Newtonian color terms) for his refractive index measurements of glass samples. Fraunhofer’s angular measurements were taken with a precision theodolite, ruled on silver in 10 arc-second intervals, and an achromatic telescope. The two verniers helped provide a precision of one arc second so he could (p.36) determine the index of refraction to six decimal places.46 In contrast to this, previous measurements of angular diffraction had been limited to an accuracy of less than 10 to 15 arc minutes because of the continuous change of the colors in the spectrum.47 Fraunhofer’s spectrum map must thus be seen in the context of the applied research done at the optical and mechanical manufactory of Utzschneider and Reichenbach at the former monastery in Benediktbeuern.48 As Myles Jackson has pointed out, Bavarian makers of scientific instruments had profited from the Napoleonic occupation and the order by Maximilian IV to conduct an up-to-date geodetic survey of Bavaria employing the techniques of the Bureau topographique. Meanwhile their British competitors, who had dominated the field in the eighteenth century, were being hampered by high taxes on glass.49 Fraunhofer’s ‘Optical Institute’ supplied large amounts of bubble-free and striae-free achromatic glass for optical precision instruments such as telescopes and also responded to the high demand for such geodetic instruments as theodolites. It was ideally located to take advantage of an already experienced labor force. For glass making was not a foreign practice at the Benedictine monastery even before its secularization in 1803. The 48 people involved in the production of optical glass in Benediktbeuern included twenty polishers, two tube drawers, one heater, five turners, one glass pourer, one assistant optician, and numerous other skilled artisans.50 Within this context of optical glass production, spectrum lines remained important for determining refractive indices and dispersion formulas throughout the nineteenth century.51
Page 16 of 81

The Spectrum in Historical Context
Once Fraunhofer’s discovery became more widely known through reprintings of his results and translations into English and French in 1823, the dark lines in the solar spectrum and the corresponding bright lines in flame spectra began to attract the attention of researchers, sporadically already in the 1830s, and extensively since the late 1850s. For such studies, Fraunhofer’s shorthand symbols were clearly not adequate. Nor were the vague color terms used in the early 1820s, such as, for instance, those in a paper on a monochromatic source of light by the Scottish physicist Sir David Brewster (1781–1868). Its spectrum is described as “a fine homogeneous yellow, which, when analyzed by the prism, exhibited faint traces of green and blue, but not a single ray of red or orange light”.52 John F.W. Herschel’s report on “the colours of the prismatic spectrum exhibited by certain (p.37) flames” such as muriates of strontia (i.e., chlorides of strontium oxide) and of lime, copper nitrate, and boric acid, published in the same year 1822, is not any clearer, nor are Fox Talbot’s descriptions of “experiments on coloured flames” written four years later, in which he confined himself to a simple line count in the different color regions of sodium, potassium, and strontium salts held in the flame.53 Both Brewster and J. Herschel had embarked on a study of the selective absorption of various media not so much out of interest in the spectrum per se. They were rather looking for perfectly monochromatic sources of light as a precondition for highly defined refractive index measurements. We have to remember that this research program was started shortly before Fraunhofer’s contributions became well known in England, and that this strategy of isolating light of a distinct color seemea very promising.54 But Brewster must have felt the inappropriateness of verbal descriptions for colorful absorption phenomena. In a more detailed report from 1822 for the Transactions of the Royal Society of Edinburgh he appended a hand-colored plate of spectrum strips (red, yellow, green, blue, and lavender), with the respective absorption areas blackened out (see Fig. 2.12 and color Plate III).
Page 17 of 81

The Spectrum in Historical Context

Even a black-and-white reproduction of this plate indicates by the uniformity of the shades of gray that no effort was made to represent the gradual color transitions in the continuous spectrum, nor to exhibit anything even approaching a dark ‘line’. Even the dark stripes are very wide and undifferentiated. But for

Fig. 2.12. The first eight of 12 absorption spectra as drawn by Brewster. Nos. 2 and 3 represent those of different kinds of blue glass, nos. 4 and 5 a combination of the former of greater thicknesses; nos. 6 and 7 represent the spectrum of light sent through a sky-blue paste and a very thick piece of green glass, respectively. From Brewster [1822/23a] pl. XXVII, reproduced in color on Plate III.

Brewster this was no fault,

because the only thing that mattered then was the extension of each of the

various colored areas: the broader the blackened parts were, and the fewer the

luminous areas, the better suited it was for his purpose of finding a

monochromatic source of light. In the same year in which (p.38) his paper was

published (1823), Brewster hit upon Fraunhofer’s publication and immediately

translated and republished it in the Edinburgh Philosophical Journal.

Interestingly, he deviated from Fraunhofer’s original title by adding “with an

account of the lines or streaks which cross the spectrum”,55 obviously still not

quite decided about what to call the dark parts in Fraunhofer’s spectrum, as his

earlier drawing suggests. Other commentators of Fraunhofer’s discovery

compared his dark lines with the black stripes in interference patterns, thus

hypothetically linking Fraunhofer’s discovery with Thomas Young’s and Augustin

Fresnel’s discoveries in physical optics and their revival of a wave theory of

light.56

The 1820s saw the introduction of another representational form of the observed spectrum as well. In his survey article on light, written for the Encyclopædia Metropolitana in 1827, John F. W. Herschel publicized his idea of describing the variation of light absorption with color in noncrystalline media, such as liquids and gases, by means of a curve. He plotted the amount of transmitted light as a function of refrangibility for various thicknesses of the absorbing material.57

Page 18 of 81

The Spectrum in Historical Context

The schematic representations obtained (cf. Fig. 2.13) were not based on measurements; they were rather a “geometrical picture of the action of the medium on the spectrum.”58 Nonetheless, they nicely illustrate the color intensification of absorbent media as their thickness is increased (left case with green glass), as well as the change in

Fig. 2.13. Herschel’s continuous curves of absorption in various media as a function of thickness. Left: medium with maximum transmission in the green such as green glass; center: yellow glass; right: medium with two different maxima of transmission in the extreme red and green. In all cases the red end of the spectrum is to the left, the violet to the right. From J. Herschel [1828/30] pl. 7, figs. 113–15.

perceived color (dichroism) if

the medium has more than one transmission maximum (right case). The center

of attention for both Herschel and Brewster was the hotly debated question of

which theory of light best accommodated the newly discovered effects of

polarization, interference, and the observed dependency of refrangibility on

color (dispersion).59

(p.39) Some observers continued to quote angular refractions or equivalently refractive indices until the late 1860s,60 but these values (e.g., the angle 51°11′55” and the index n D = 1.33397) were not only awkward to use but obviously dependent on the angle of refraction and material composition of the specific prism used in the experiment (see Fig. 2.21). Another recording method simply entailed counting the lines in the intervals delimited by the main Fraunhofer lines. This was done especially when the quality of a representation comparable to Fraunhofer’s or later Kirchhoff’s spectral map was at issue.61 Given the thousands of spectrum lines revealed by a spectroscope even of only moderate resolution, which were available by the mid-nineteenth century, this task was far less easy than might be supposed. Especially when the observer wanted not only to count each line, but to locate it roughly as well, visual observation of the spectrum had physiological constraints:

Page 19 of 81

The Spectrum in Historical Context
With the first attempts already, it became evident that it is extraordinarily difficult to indicate the number of lines in one part of the spectrum at the same time as their location. For at first glance, one sees only the stronger and very black lines, whereas when one looks at the colored stripes for a longer time until the eye has adjusted to a specific type of light, almost in every single segment of the spectrum on which the eye focuses, a large number of lines of differing light intensities becomes apparent, none of which or only individual ones of which, of a different quality, had been perceptible before. So if one combines observation simultaneously with measurement of the refractive angle, then many lines that the observer had been able to see quite clearly upon concentrated examination of the spectrum vanish because the eye must adjust itself in rapid succession to different types of light.62
Consequently, line counting and the measurement of relative distances between lines soon evolved into separate branches of spectroscopic research. The observer to whom we owe the above description of the eye’s ability to adjust so quickly from one color region to another decided to concentrate first on the precise determination of relative distances between the main Fraunhofer lines and other characteristic lines. He then changed over to a detailed examination of the fainter lines in the much smaller intervals between the strong markers. Sections of the spectrum with very many regularly spaced lines lying close together essentially required little more than a line count. The data thus obtained were then transformed into sketches of a great number of small segments of the spectrum made to the same scale, to form a composite map of the spectrum.63
In 1854 a passionate amateur experimenter with home-made instruments, and inventor of several electrical gadgets including a proto-telegraph, also tried to distinguish the “physical properties of light produced by the combustion of different metals in the electric spark.” This physician, David Alter (1807–1881)64 of Freeport, Pennsylvania, drew up a (p.40) table of the seven Newtonian primary colors (from left to right) and listed twelve chemical elements (vertically, see Fig. 2.14). In the resulting 84 spaces he entered up to four vertical lines for the number of prominent spectral lines or bands (effectively somewhat like Roman numerals). So in a sense he simply counted how many such lines there were per color region for each spectrum.
Page 20 of 81

The Spectrum in Historical Context

His motivation for devising this

recording approach arose from

a desire to emend the limited

verbal description of flame

colors then in common use.

Even the most refined

vocabulary fell short when

trying to discriminate between

the color of flames or arcs of

elements like silver and

thallium: the arc spectra of

these two metals both not only

were green but happened to

exhibit “the same shade of

green.” However, Tyndall identified two green bands in silver and only one in thallium, situated “midway between the

Fig. 2.14. David Alter’s representation of spark spectra as a tabular line count in seven colors. From Alter [1854] p. 55.

two greens of the silver.”65 Thus

a line or band count did improve upon the qualitative description. In the

accompanying text Alter tried to specify the lines entered in his table further by

adding descriptions like “faint yellow line”, or “in the orange a very bright band,

one of yellow and one of green—two faint bands in the blue which are not always

seen”. But these verbal descriptions remained diffuse and ultimately inadequate.

While this sector-count representation, as I would like to call it, gave at least

some information about the overall distribution of bold lines in the spectrum of

the dozen elements examined, it was far from satisfactory in conveying their

features. No effort was made to register the relative intervals between the lines,

not to speak of their intensities. Authors positively struggled to express the

subtle intensity distinctions: the “faintest suspicion of a line” as opposed to:

“more decided line, fine line, decided line, more decided line, still more” or

“strongest line” against one “almost as strong”.66 After 1860 Alter, with the

support of a few others, tried to portray himself as a (p.41) “veteran

spectrochemical analyst” or even as “the discoverer of spectrum analysis”.67 But

a closer look at what he actually says in his publications about “coloured bands”

reveals that he was still far away from finding a suitably unambiguous

description for characteristic spectra of the chemical elements. Clearly, a better

means of referring to a specific line within the spectral ranges of several

hundred units of Å (in later nomenclature) was needed.

Page 21 of 81

The Spectrum in Historical Context
Sir David Brewster suggested an alphanumerical reference system to refer to the prominent lines between the few major lines that Fraunhofer had already labeled with upper or lowercase letters of the alphabet. About 200 stronger intermediate lines between these Fraunhofer lines were marked with natural numbers, beginning with 1 at the reddest end of each interval between two Fraunhofer lines: hence Dl, for instance, denoted the first dark line next to the yellow D line, while C26 denoted the closest dark line at the other side (towards the red end of the spectrum).68 The many other fainter lines fared no better, though, in this notation system, which could be hardly more specific than: ‘inbetween line C26 and Dl’. The structural problem with this kind of labeling system was that the number of known lines was continually rising. So no matter how many lines were assigned ‘names’, there would always be lines ‘in-between’ that were not directly identifiable. To overcome this problem, Stokes proposed an alternate nomenclature to indicate the position of a chosen spectrum line with respect to any given pair of Fraunhofer lines. He divided the distance between any two Fraunhofer lines, say between lines D and E, into 100 equal parts and then indicated a position between D and E, 27/100 units away from D, as ‘D 27 E’, and a position 50 units from the interval G–H beyond the last visible line H as G H ½. The size of these units thus clearly varied with each interval between the Fraunhofer lines.69 As this alphanumeric classification system was being suggested, however, the transition to a more strictly pictorial mapping of emission and absorption spectra was already underway.
Charles Wheatstone’s (1802–1875) “table of bright lines” in the spark spectra generated from six different types of metallic electrodes is sometimes referred to as the first ‘naturalistic’ drawing of emission spectra. But unlike its reprint of 1861, Wheatstone’s original contribution for the Dublin meeting of the British Association for the Advancement of Science in 1835 did not include any published drawing of the spectrum, just narrative descriptions of “definite rays, separated by dark intervals from each other”.70 The ‘reprint’ includes a figure structurally halfway in-between David Alter’s table (shown above in Fig. 2.14) and spectrum representations common after 1860. This illustration, in which each spectrum is still represented vertically, not horizontally, must be seen in the context of the priority dispute over the discovery of spectrum analysis—a point unfortunately missed in most standard accounts of the history of spectroscopy.71
Page 22 of 81

The Spectrum in Historical Context (p.42) In 1845 the professor of chemistry at King’s College in London, William Allen Miller72 (1817–1870), published the results of his experiments and observations of the flame spectra of metallic compounds like copper chloride dissolved in an alcohol solution before being brought into the flame. To better illustrate the “intricate spectra” thus obtained, Miller chose to supplement his paper with two chromolithographed maps, one of which is reproduced here as Fig. 2.15 and in color on plate III. Both consisted of five color zones (red, orange, yellow, green, and blue) whose borders also nearly coincided with the major Fraunhofer lines from B to F. The distribution of spectrum lines and bands was indicated by more or less strongly shaded regions. The presence of the Fraunhofer lines as orientational aids offered—at least in principle—a fairly accurate way to determine the relative distances and positions of the respective lines. Nevertheless, Miller cautioned against overestimating the precision of his prints, which had a total width of 15 cm and thus an average dispersion of c. 200 Å per cm: “No pretensions to accuracy are made in these sketches; they are simply intended to convey an idea of the general position and grouping of the lines. […] the comparison, though not rigidly accurate, being still very nearly so, and perfectly sufficient for my purpose”.73
Page 23 of 81

The Spectrum in Historical Context
As Miller divulged to his readers at the outset, the purpose of his research had originally been to discover certain recurrent patterns in the “general arrangement of lines” in spectra of different but chemically related materials. Even though this aim, which is actually quite in line with the much later search for ‘homologous spectra’ (cf. here pp. 308ff.), was not met, Miller decided to publish his results anyway, if only to inform other researchers about which classes of substances did not work for such a research program.74 Interpretation of his findings was thus relegated to the indefinite future—all the more reason at least to provide a sufficiently accurate phenomenological description of them. One indication that Miller did succeed in this much is the positive assessment of commentators even in our century. Despite the acknowledged inaccuracies, Miller’s plates were still “strikingly realistic pen and ink drawings”.75 The London-based chemist was critized by his contemporaries, though, in particular for the low-temperature spirit flame he used, which was unable to produce the full spectrum of many of his samples, if at all. They were also disturbed by the fact that the alcohol solution superimposed its own spectrum on the examined spectra.76 Thus, (p.43) when John William Draper (1811–1882),77 professor of chemistry at the University of New York, published his report on the production of light by chemical action in 1848, that is, three years after Miller’s paper, his figure of the spectra of various flame spectra (with few exceptions) just makes use of rectangular stripes to indicate the different extensions of these continuous spectra mapped relative to the Fraunhofer lines in the solar spectrum. However, we do still see a rather symbolic depiction of what we today would identify as the two broad cyanogen band spectra (below G and between F and G), as well as interruptions of the continuous spectrum at regularly increasing intervals (in the spectral region B to E). In a blow-pipe spectrum, generated when additional oxygen is blown into the flame, thereby substantially increasing its temperature, Draper even noticed something that we would now identify as emission lines: “the spectrum was divided into five well-marked regions, separated from one another by inactive spaces; in short, I saw five distinct images of the blue cone, one yellow, two greens, one blue, and one violet.”78
Page 24 of 81

The Spectrum in Historical Context

Altogether, a comparison of

Draper’s diagram (Fig. 2.16)

with Miller’s lithograph from

three years earlier (Fig. 2.15)

shows how far ahead of its time

Miller’s illustration was in the

(p.44) mid-1840s. This more

naturalistic visual

representation of emission and

absorption spectra, which

includes the specific positions

of dark and bright lines as well as a faithful depiction of their relative distances and intensities, was taken up by several other researchers in the succeeding years. In 1851, Antoine-Philibert Masson (1806–1860), who taught physics at the Lycée Louis-leGrand and at the École Centrale

Fig. 2.15. W.A. Miller’s representation of molecular flame spectra. No. 7: solar comparison spectrum with major Fraunhofer lines; no. 8: CuCl2; no. 9: H3BO3; no. 10: SrNO3; no. 11: CaCI2: no. 12: BaCl2. Lithograph by J. Basire. From W.A. Miller [1845a] pl. II (cf. also here Plate II (top) for the coloring of this plate by iris printing.

des Arts et Manufactures,

decided to embellish his description of metallic spark spectra with little black-

and-white sketches of about 7 cm length. They essentially consisted of

rectangular white boxes separated by stripes of different widths indicating the

positions, color range, and to some extent the strengths of the main lines.

Masson’s ‘tableau’ actually is a very detailed projection of the various spectra

recorded with the aid of a camera lucida.79 Compared with Alter’s table

Masson’s representation endeavors to provide more detailed information on the

positioning of various lines within each rectangular box. It is much more iconic

and less symbolic than its predecessor map. Had Masson followed up his

iconography, he might have ended up with equivalents of Bunsen’s plates

displaying the characteristic spectrum lines of the (p.45) various chemical

elements. But the fact is: he didn't. Like several of his contemporary physicist

colleagues, he too was only interested in the nature of the electric spark, and not

in spectra per se, which were nothing more than a noteworthy side-effect.

Page 25 of 81

The Spectrum in Historical Context

What we thus see in the early

history of representations of

terrestrial emission spectra (as

opposed to the solar spectrum)

is a gradual transition from a

merely narrational description

of flame or line colors (such as

Wheatstone’s) to a symbolic,

tabular vertical arrangement

(such as Alter’s) to an iconic

horizontal chart (such as

Masson’s), which then leads to

Fig. 2.16. J.W. Draper’s diagram of

the more extensive maps of

continuous flame spectra (nos. 2–9) and

larger spectral regions. Since

of one discontinuous line spectrum of an

many individual researchers are oxygen enriched blow-pipe cone (no. 10),

involved, working within their

all drawn relative to the solar spectrum

various ‘scientific communities’

with the main Fraunhofer lines (no. 1).

or in isolation in private

Woodcut. From J.W. Draper [1848] p. 107.

laboratories, and in many

different national and local

contexts, it cannot be expected that this transition be governed by any strict

rules of the type ‘no iconic mapping before 1854’ or ‘no tabular arrangement

after 1851’. We are rather dealing with a gradual shift of preferences for one

representational style over another, with considerable overlapping and a certain

inertia especially with respect to the terminology used to denote them.80 While

in solar spectroscopy this transition to iconic depictions of the spectrum

occurred already in the early nineteenth century (to a large extent from the

immense impact of Fraunhofer’s map on his contemporaries), in terrestrial

spectroscopy a proper mapping of the relative distances between spectrum lines

only became common in the mid-1850s.

2.4 The tardy emergence of spectrum analysis There is a reason for the relatively late emergence of a ‘mapping impulse’ (to borrow a term coined by the art historian Svetlana Alpers) in the analysis of flame, arc, and spark spectra. It is true that such pioneers as Fox Talbot and Masson came quite close to discovering spectrochemical analysis. Talbot, for instance, had already pointed out that the two red tints of lithia and strontia flames, hardly distinguishable with the unassisted eye, could be readily kept apart if one looked at their spectrum.

Page 26 of 81

The Spectrum in Historical Context
The strontia flame exhibits a great number of red rays well separated from each other by dark intervals, not to mention an orange and a very definite bright blue ray. The lithia exhibits one single red ray. Hence I hesitate not to say that optical analysis can distinguish the minutest portions of these two substances from each other with as much certainty, if not more, than any other known method.81 But despite such farsighted prophesies, prior to the late 1850s it was far from clear that each element had its own characteristic spectrum, and a lot of evidence actually seemed to speak against this assumption at the time. For instance, virtually all spectra exhibited a strong yellow line. Swan’s findings seemed to suggest that different hydrocarbon compounds could not be distinguished by their respective spectra. Likewise, Masson’s publications, which superficially look so much like an anticipation of spectrum analysis,82 merely established (p.46) a set of four strong lines (which he called α, β, γ, and δ) as common to all eight spectra sketched in his ‘tableau’.83 To confuse things even further, some elements displayed completely different spectra in the flame than in the electric arc or spark. By no means all chemists had prisms at their disposal. Such were a more frequent item in physical cabinets, but physicists generally were not yet interested in problems of chemical analysis. Finally, even if they had been, it was far from clear at the time what the fundamental unit would be with which to correlate the various spectra. Chemistry was still grappling with the question of atomic structure, and many alternative models of the atom were intensely discussed throughout the nineteenth century. Supporters of Prout’s hypothesis, according to which all atoms were composed of packets of light hydrogen atoms, were pitted against people like Lockyer who were looking for even smaller constituents of the chemical elements.84
Page 27 of 81

The Spectrum in Historical Context

The problem of too low flame temperatures was overcome in the middle of the 1850s with the invention of the Bunsen burner. The colorless flame produced by this gas burner yielded temperatures of up to 1800°C.85 One of the first scientists to use this new device for spectroscopic analysis of various hydrocarbon compounds was the Scottish physico-chemist William Swan (1818– 1894). His result was that for all spectra related to substances of the chemical type CrHs (e.g., wax, oil, tallow, turpentine) or of the form CrHsOt (e.g., ether, alcohol, and glycerine), the same five colored bands and half a dozen characteristic bright lines appeared, once the slit was adjusted to the blue base of their flames. Indeed, this spectrum was identical with the one from the blowpipe, “though that little instrument gives it more neatly, clearly, and without the swamping effect on its best characteristics of the dense continuous spectrum derived from the more or less yellow light in the upper parts of most kinds of simple lamp flame.”86 This surprising similarity in the spectra of dissimilar chemical compounds made Swan curious about how their common lines compare with those of sunlight. But his comparative plot (Fig. 2.17) showed disappointingly few coincidences: only in the case of the line α could he verify a precise coincidence.

(p.47) As disappointing as

Swan’s result of 1857 might

have been, methodologically it

was decisive. A systematic

application of a hypothetical

Fig. 2.17. Swan’s comparison of the solar

one-to-one correspondence

spectrum (top) with the common lines of

between each chemical element various hydrocarbon spectra (bottom).

and a characteristic spectrum

Copper engraving, 1857. From Swan

generated in the high-

[1853/57b] pl. I.

temperature flame of a Bunsen

burner triggered the sudden

breakthrough in 1859. It was born of a fruitful collaboration between an

analytical chemist, Robert Wilhelm Bunsen (1811–1899), who could prepare

unusually pure samples, and a physicist well-versed in optics, Gustav Robert

Kirchhoff (1824–1887), both of them at the university of Heidelberg.87

Page 28 of 81

The Spectrum in Historical Context

Bunsen’s celebrated color plate

of alkali metal and alkaline

earth spectra, which was

promptly supplemented in 1860

and 1861 by the newly

discovered elements of

rubidium and caesium (see

color plate II and Fig. 2.18 for a

later black-and-white version),

became a model of its kind: It

was reprinted repeatedly in

virtually every introductory text

on spectrum (p.48) analysis,

from the small-scale versions in

sales catalogues of instrument

makers to the wall-hanging poster size. I would venture to say, it was the most frequently reprinted scientific illustration in the second half of the nineteenth century. Thenceforward, the rapidly growing community of spectroscopists proliferated representations of other spectra not depicted in Bunsen’s celebrated 1860 plate of emission spectra. For instance, Bunsen’s map inspired his pupil Rudolph Theodor Sim[m]ler

Fig. 2.18. The spectra of alkali metals and alkaline earths, including the newly discovered elements rubidium and caesium, as depicted by Bunsen and Kirchhoff, here reproduced in a later black-and-white version wood-engraved by Dulos and published in Secchi [1870] p. 230 (for the original color plate from Bunsen [1860] pl. V, limited to Ka, Na, Li, Sr, Ca, and Ba, see here Plate II). Only the characteristic lines of each spectrum are depicted, no scale is given, and the solar spectrum is added at the top for orientation.

(1833–1874) to search for

further applications of his

teacher’s method, and to plot the spectra of copper salts, manganese, and boric

acid in a similar fashion.88 But, along with countless others who may have had

the same idea, he soon had to acknowledge that many metals (such as Mg, Al,

Fe, Mn, Co, Ni, Cr, U, and Zn) and their compounds did not exhibit distinctive

spectra when held in the Bunsen-burner flame or were dominated by much

stronger, already familiar lines.89 Bunsen’s map likewise clearly served as a

model for H.F. Brasack in Halle in his choice of the scale and mode of

representation for his map of 14 metals published in 1866.90

Page 29 of 81

The Spectrum in Historical Context

Chemists like Alexander Mitscherlich in Berlin or Émile Jules Diacon (1827– 1893) in Montpellier, were skeptical of the universal validity of Kirchhoff’s and Bunsen’s thesis that each element had its own unique spectrum.91 But even they brought forward evidence to the contrary in a form that clearly followed the standards set by Bunsen’s charts (see, e.g., Figs. 2.19 or 2.23). Spectra of chemical compounds like metallic oxides, chlorides, bromides, or iodides, were soon called band spectra to distinguish their channeled appearance from those of the Bunsen and Kirchhoff type, called line spectra, with isolated lines lacking obvious regularity in arrangement. Thereafter chemists continued to plot a considerable number of absorption spectra with broader bands distinguishing different absorbers.

2.5 Numerical scales in spectrum maps

It is important to point out,

though, that neither Bunsen’s chart nor other lithographed maps of the early 1860s provided any numerical scale: they simply offered the solar spectrum (p.49) as a gauge, or other arbitrary lines as orientational aids.92 In the

Fig. 2.19. The copper chloride spectrum depicted by Diacon. This copper engraving by Dulos (cf. here p. 143) shows the several broad bands with different ruling distances to indicate the various intensities. Its total length is 16.4 cm. From Diacon [1865] pl. I, fig. 1.

second publication of 1861 by

these two Heidelberg professors we learn more about how they had made their

original drawings: Bunsen had certainly used a numerical scale as a guide. Their

second spectroscope (cf. Fig. 2.20), which was a much improved design by the

Munich optician Carl August Steinheil93 (1801–1870), projected a custom-made transparent scale94 that was inserted in the slit of a third collimating tube onto

one face of the same prism that decomposed the light coming from the flame. By

reflection off the prism surface, the image of this scale could be seen together

with the flame spectrum. With this device two different spectra could easily be

compared in addition, by means of a reflecting prism placed in front of the

micrometric slit.95 Thus instead of having to take angular measurements with

reference to a graduated circle (such as in Fraunhofer’s theodolite

arrangement), angles were simply read off this scale.

Page 30 of 81

The Spectrum in Historical Context

But while such a scale was used

as an internal reference aid, it

was not considered

unreasonable to exclude it from

the published plate.96 First of

all, the choice of the origin for

this scale was purely

conventional. Bunsen happened

to choose the value 50 for

Fraunhofer’s (p.50) D line,

since this line was most easily

identifiable in almost every

spectrum.97 The scale units, as seen through the eye-piece of the spectroscope, were even more arbitrary, because they were dependent not only on the unit chosen for the transparent scale γ, but also on the spectroscope’s focal distance. Even if the scale γ and all the optical parameters could have

Fig. 2.20. Bunsen and Kirchhoff’s second spectroscope, built by C.A. Steinheil in Munich. From Bunsen and Kirchhoff [1860/61b] plate. The scale at δ is projected onto the prism at P, whence it is reflected into the observing telescope B. The observer thus sees the scale superimposed onthe flame spectrum passing through tube from slit ε.

been reproduced perfectly by

other spectroscopes, two observers with different instruments would

nevertheless still have disagreed on the numerical values of certain lines,

because prismatic dispersion, and with it the relative distances between

spectrum lines, depends crucially on the specific type of prism glass used in the

spectroscopes.

Page 31 of 81

Fig. 2.21. Comparison of spectra generated by flint and crown-glass prisms. From Schellen [1870/72b] p. 229.

The Spectrum in Historical Context
As Fig. 2.21 demonstrates,98 a prism of flint glass produces not only a broader spectrum, but also relatively less red and more blue-violet than a prism made of crown glass. Thus, essentially: “Every prism gives a different map of the spectrum, nor when we find a band or line by the prism have we any means of fixing the absolute place, except by a reference to the normal or wavelength scale, or to one derived from it”.99 As Simon Schaffer has argued, this “irrationality of dispersion” was quite instrumental in some of the debates on Newton’s crucial experiments in the late seventeenth and early eighteenth centuries. The continental scientists used different kinds of glass than the British, which precluded unequivocal comparisons of the refraction spectra obtained.100 There is another aspect to Bunsen’s chromolithographic representation of emission spectra beyond the omission of a scale. A controversy between Bunsen and two American chemists reveals particularly clearly how conscious Bunsen was of deviating— (p.51)
not without good reason[—]from the much more accurate means of measurement used by physicists in the determination of the refractive indices of transparent bodies, inasmuch as the chemist, for whom our apparatus is specially designed, does not require so much an exact knowledge of the absolute position of the single lines in the spectrum as he needs to be able to observe quickly and easily, especially when lines have to be recognized which only flash for a moment.101 Late in 1862, the chemists S.W. Johnson and O.D. Allen from New Haven had criticized Bunsen for incompletely and incorrectly mapping the spectrum of caesium. The two Yale chemists reported having found “without difficulty seven more lines” than the eleven depicted in the map published by Bunsen and Kirchhoff in 1861 and they additionally disagreed about the precise placement of some of the lines figured by their German colleagues.102
Page 32 of 81

The Spectrum in Historical Context
In reply to these charges Bunsen argued as follows: While a physicist’s ultimate aim is to work towards numerically accurate absolute measures and complete catalogues of wavelengths, a chemist is interested in quite different information, namely the characteristic Gestalt of the spectrum lines of a given element. Therefore, in order to reduce the complexity of real spectra and to turn the given illustrations into a more easily recognizable pattern distinct from those of other elements, Bunsen had purposely omitted many of the weaker and less remarkable features: “we did not endeavour to catalogue the lines completely, but to represent as truly as possible the characteristic appearances of the spectra of the several substances. Owing to the difficulty of accurately distributing many shades of the same colour in chromolithographic printing, the weaker lines have necessarily been omitted in our drawings.”103 This selfimposed restriction to strong lines should not be understood as a defect, but as a virtue of Bunsen’s mapping, because his goal was not comprehensive inventorization but a fixing of each element’s most characteristic lines to allow its unequivocal identification. Bunsen’s chart thus only depicted the basic fingerprint, so to speak, of each chemical element.104
To the extent that the limitations of chromolithography had hampered Bunsen’s visual representation, this was a call for an alternate way of depicting spectra. Bunsen conceded this point by devising a quite different method of plotting the relative distances and intensities of spectrum lines and continuous spectra in the plate accompanying his rebuttal (cf. Fig. 2.22). The most prominent feature of the whole graph was the scale, printed as on a centimeter ruler, and gauged, as before, at NaD = 50, with the prism placed at the angle of minimum deviation for all readings. Where Bunsen observed a spectrum line, band or, occasionally, a continuous spectrum with his spectroscope, he drew a black blob of proportionate breadth and length to symbolize the relative intensity of the spectral feature.
(p.52)

Page 33 of 81

Fig. 2.22. Bunsen’s symbolic plot of emission spectra and continuous spectra. The scale indicates readings from Bunsen’s spectroscope, gauged at NaD = 50 at the abscissae of the black spots; their ordinates represent the intensity of the features. Woodcut. From Bunsen [1863d] pl. V, figs. 4–5.

The Spectrum in Historical Context
This partly iconic spectral coding was very simple to draw and, unlike chromolithographic plates, was easy to print as well. A simple woodcut would do the job (cf. here § 4.4), while otherwise several lithographic stages were necessary. This symbolic technique was used widely since,105 particularly by chemists to render absorption spectra, which often exhibit broad absorption bands and other extended features that are thus easily captured. As late as 1881, William Marshall Watts (1844–1919), an experienced spectroscopist specialized in mapping the carbon spectrum and its various compounds, wrote: “There is no better plan of noting the peculiarities of a spectrum than that employed by Bunsen, in which each bright line is represented by a black mark on the paper whose height represents the intensity of the line.”106
Even though Kirchhoff and Bunsen recognized that, for prism spectroscopes, numerical scales were “arbitrary”, many others, especially chemists, were much less aware of this. In a whole series of experimental investigations, Alexander Mitscherlich (1836–1918), who had just submitted his dissertation to the University of Berlin, where his father Eilhart Mitscherlich lectured on chemistry,107 corroborated his claim that the spectra of metallic compounds differ from those of the pure metals. His lithographed maps display the spectra directly above each other to drive home his point. His mode of charting the spectra combined characteristics from both representational techniques introduced by Bunsen: an iconic depiction of individual lines and their intervals, and a condensed comparative chart focusing on the distinguishing features in the spectra (cf. Fig. 2.23):
Mitscherlich’s spectrum charts are also a good illustration of the interplay between techniques of representation and original research: these comparative plots revealed recurring characteristic lines in the various compound spectra of a single metal. For instance, Mitscherlich claimed that the distances between the characteristic lines for barium varied with the atomic weights of the compounds (cf. the shift to the right of the two prominent lines in the spectra of BaF2, BaCl2, BaBr2, and BaJ2, with an increase in distance from 3 scale units to 3.9, 5.2, and 7.3 in the last four spectra of Fig. 2.23. According to their atomic weights, the relation between BaCl2, BaBr2, and BaJ2 would have implied (p.53) 104/148.5 = 5.5 and 104/195.5 = 7.3, respectively). Mitscherlich also searched for similar relations by comparing the spectra of various metallic oxides. He found several local similarities but he eventually had to admit that his drawings provided only weak support.108
Page 34 of 81

The Spectrum in Historical Context

Despite the failure of

Mitscherlich’s efforts to reveal

essential clues about the

structure and relationship of

various elements, his systematic

comparison of compound

spectra followed the trend away

from a merely qualitative depiction of the Gestalt of spectra towards a more precise, quantitative rendering of their line distributions. Bunsen’s comparative and scaled blackand-white woodcut chart of metallic spectra and Kirchhoff’s famous map of the solar spectrum, to which we now turn, sparked the transition

Fig. 2.23. The spectrum of barium oxide compared with those of strontium and calcium oxide, and with barium fluoride, chloride, bromide, and iodide, according to Mitscherlich. The numerical scale underneath each spectrum is arbitrary, whereas the letters D, E, F. and b indicate the positions of strong Fraunhofer lines. Stone engraving by A. Schütze (Berlin). From Mitscherlich [1864a] pl. VI.

toward representations of

chemical spectra with an

associated numerical scale, irrespective of whether or not it was deemed

arbitrary. In 1861, right after the birth of spectrum analysis, Kirchhoff undertook

to draw a map of the solar spectrum. The coincidence of bright emission lines

and dark absorption lines now permitted a reading of the dark Fraunhofer lines

as indicators of the presence of chemical elements in the solar atmosphere,

suddenly giving this endeavor new relevance. To obtain maximum resolution,

Kirchhoff used another Steinheil spectroscope with a four-prism chain designed

according to his own specifications (Fig. 2.24).

How to convey the excitement he must have felt upon first seeing the solar spectrum in such unprecedented detail and intensity? One Munich instrument manufacturer exclaimed in 1862: “The splendor of Kirchhoff’s solar spectrum is enthralling!”109 A small sample segment from Kirchhoff’s published map is reproduced here as Fig. 4.10 (p. 125). It included (p.54) chemical identification of some of the Fraunhofer lines that he could correlate with 463 emission lines of laboratory-generated spark spectra, with a juxtaposed numerical scale that ran from about 1000 (D) to 2250 (past F) in the map of 1861. The continuation of the map in 1862 extended this scale down to 381 (near A) towards the red end, and up to 2875 (near G)—note that Kirchhoff’s scale runs contrary to the later wavelength scales. Why now include an arbitrary scale?

Page 35 of 81

The Spectrum in Historical Context

Above the drawing of the

spectrum I have placed a

scale divided into

millimetres, and having an

arbitrary commencement.

This is, in the first place,

useful for the purpose of

obtaining an easy means of

nomenclature for the lines.

[…] By means of this scale

we are likewise enabled to

specify with a greater degree

of accuracy the positions in

the spectrum where no dark

Fig. 2.24. Kirchhoff’s four-prism

lines occur. A relation

Steinheil spectroscope. From Kirchhoff

between the numbers on the

[1861/62a], pl. III.

scale corresponding to the

individual lines and the

refractive indices of my prisms for these lines does not exist, because the

prisms were sometimes placed more exactly than at other times at the

angle of minimum deviation for the particular rays.110

Thus Kirchhoff’s scale was nothing more than a convenient labeling system that had to be used in conjunction with his spectrum map. Orientation within the spectrum had been (p.55) made one step easier, but each observer still had to assess visually the similarity between the mapped groups of lines and what he saw in the ocular of his own spectroscope. A ‘subjective’ criterion still decided on the identification of a given line instead of some ‘objective’ one, such as a numerical coincidence of tabulated values with scale values read off the instruments. Despite this drawback Kirchhoff’s scale was readily accepted by the scientific community. As an anonymous referee put it in his review of the English translation of Kirchhoff’s memoir by Henry Enfield Roscoe (1833–1915) in 1862: “provisionally at least, Kirchhoff’s scale is certain to be adopted. The whole science would be thrown into great confusion were different scales, each with its own zero, adopted by different observers, like the several systems of longitude; we earnestly deprecate such a proceeding.”111

2.6 Prisms versus diffraction gratings Aside from the differences between various types of glass prisms and prism spectroscopes, there was another “inherent defect in the prismatic spectrum”:

Page 36 of 81

The Spectrum in Historical Context in the same spectrum, from the very circumstance of their greater refrangibility, those [rays] in the violet will be relatively more separated from each other than those in the red. […] The result of this increased separation in the more refrangible regions is to give an apparent dilution to them, while the lesser refrangible regions are concentrated. […] When, therefore, we obtain a prismatic impression on any sensitive surface, it is very far from representing the true character of the phenomena. The action which ought to be concentrated in a lesser space at the more refrangible region is spread over a greater, and with that augmentation an apparent diminution of the amount of action is perceived. This, of course, should make the maximum point vary, spread out unduly the violet end, and dilute the true effect. The different regions of the prismatic spectrum cannot be fairly compared with one another.112
This graduated ‘dilution’ of the prismatic spectrum is illustrated in Fig. 2.25. However, as Draper indicates here, there was an alternative way to assign a numerical value to a specific spectrum line which promised to yield a materialindependent, ‘normal’ wavelength scale.
Page 37 of 81

The Spectrum in Historical Context
In a paper published in 1823, Joseph Fraunhofer had shown mathematically that a perfectly regular grid of lines would yield a spectrum by constructive interference of wave-fronts differing by integral multiples of one wavelength. He had even made some initial attempts to build such a diffraction grating, as they came to be called. They were made of wire wound along finely tooled screw threads to form a fine, regular grid. In a later design, closely parallel lines were scratched with a diamond onto a gold-plated glass surface. But these primitive gratings still could not compete against prisms.113 Furthermore, the poor intensity of the resulting spectrum precluded their application in stellar spectroscopy; the (p.56) superposition of different orders of interference, and the appearance of pseudolines (so-called ghosts) beside prominent ‘real’ lines were further disadvantages of diffraction gratings. Nevertheless, improvements in ruling technology in the 1860s and 1870s, achieved by the instrument makers Friedrich Adolph Nobert (1806–81) in Pomerania, William A. Rogers (1832–98) at Harvard, and Lewis M. Rutherfurd (1816–92) in New York, changed all that. Diffraction gratings began to replace prism spectroscopes in metrological tasks requiring instrument-independent information about the ‘normal’ position of spectrum lines or other spectroscopic features. I cannot dwell here on the history of the production of these diffraction gratings,114 but significant progress was made in the evenness of line ruling, the equidistancing of adjacent lines and the total width of a ruled surface S, which determines the resolution λ/Δλ ∼ n · S/ ε (n is the order in which one observes, typically the second or third). The small Nobert gratings in the late 1860s (used by Ångström, van der Willigen, Mascart, and Cornu) were replaced by Rutherfurd’s in the mid and late 1870s, followed by Henry Augustus Rowland’s concave gratings invented in 1882. The latter relieved the experimenter of collimator lenses or other optical components and thus simplified spectroscopic measurement considerably. They reached a theoretical resolution λ/Δλ of up to 400 000, more than three times the resolution of the best prism gratings available, marking the high point of nineteenth-century diffraction grating technology.115 Given certain preconditions (such as correct mounting and adjustment and, for a reflection grating, proper placement of the photographic plate opposite the grating, etc.), Rowland’s concave gratings produced a normal spectrum at high dispersion and essentially free of ghost lines. They even provided the additional feature of considerable overlap among adjacent orders of the spectrum, which allowed internal verification of wavelength measurements by the coincidence method: a wavelength λ1 in the order n 1 and a wavelength λ2 in the order n 2 were coincident in the photograph if n 1 · λ 1 = n 2λ2. Because of all of these features, these concave gratings were widely hailed as a labor-saving device, and access to one of these (p.57) precision instruments became an essential prerequisite for spectrometric work at the research front. When Rowland reported on the possibilities opened up by his new technology at the Physical Society in London, James Dewar (1842–1923) commented:
Page 38 of 81

The Spectrum in Historical Context

We have heard from

Professor Rowland that he

can do as much in an hour as

has been done hitherto in

three years. I struggle with a

very mixed feeling of elation

and depression. Elation for

the wonderful gain to science

and depression for myself,

for I have been at work for

three years in mapping the

ultra violet.116

Fig. 2.25. Comparison of the scales of a

In cases where high-resolution diffraction gratings were not

prismatic and a diffraction spectrum. From Löwe [1925a] p. 7.

obtainable or not fully

applicable to the specific

spectral range under study, this normal spectrum had to be reconstructed from

the data using a graphical method of conversion.117 Thus we see that the visual

mode of representation of spectra, in turn, generated a need for graphic modes

of data analysis. Once ‘normal’ spectrum maps such as Ångström’s in 1868

became firmly established—“classically accurate and chemically expounded” as

one contemporary put it118—older maps and spectral tables that had been

plotted to other scales were converted into the new wavelength system. In 1868,

for instance, both the Astronomer Royal George Bidell Airy in Greenwich and

Josiah Willard Gibbs at Yale published formulas and tables for converting the

‘old’ Kirchhoff scale into the ‘new’ Ångström scale.119 Likewise, in 1864 Leaner

Ditscheiner in Vienna transposed the ‘old’ Fraunhofer angular refraction values

into the new wavelength system and in 1869 Gibbs converted Huggins’s

arbitrary scale, strongly criticizing Airy’s conversion of Kirchhoff’s scale as

utterly useless for all scientific purposes.120 The same procedure was later

reiterated when Rowland’s system of wavelength normals replaced the older

‘normal maps’ in the 1880s.

Page 39 of 81

The Spectrum in Historical Context
It would be deceptive to think that with the introduction of the ‘normal’ wavelength system everything was stabilized: the half-life of each system of wavelengths, often associated with its own form of visual representation, i.e., the standard spectral map, remained incredibly short. To stay with the example of solar spectrum maps: Ångström’s map and wavelength values were superseded by Rowland’s “preliminary” wavelength standards, obtained with the best of his concave gratings. Rowland’s standards, in turn, were officially revised twice, the first revision appearing in 1928, and the second as late as 1966, not to mention the many minor corrections and alterations to his system under the ægis of the International Union for Co-Operation in Solar Research from 1904 on, and then, after World War I, under the guidance of the International Astronomical Union (IAU).121 Closer examination (p.58) of Rowland’s laboratory notebooks in Baltimore reveals the considerable part played by known wavelength values in new measurements—if only as orientational markers and as an easy means of checking the basic reliability of the measurements.122
The emerging preference for maps generated by diffraction gratings displaying the normal spectrum, and line positions measured in terms of wavelengths, went hand in hand with a dispute over which unit to use for these wavelengths. French authors pled for the metric system and quoted their data in fractions of a millimeter, or equivalently, of one Å (defined as 10−10m), in commemoration of the pioneering spectrum map published in 1868 by Anders Jonas Ångström (1814–1874).123 Ångström’s access to a high-quality grating made by the Pomeranian instrument maker Nobert (see above p. 56) meant that he did not have to resort to prisms and could plot his spectrum directly from observation. Consequently, his map was the first of the ‘normal’ spectrum with the interstices between spectrum lines directly proportional to wavelength. He quoted wavelengths in units of 10−7mm.124 Ångström’s publication consisted of two volumes: the text volume included extensive wavelength tables, already attaining an accuracy of two digits after the decimal point, combined with detailed calculations concerning his wavelength standard as compared with a meter standard.125 In a separate plate volume, six plates displayed the solar spectrum from B to H2 together with a wavelength scale in units of a tenmillionth of a millimeter and chemical identification for an estimated 5 % of the total 792 lines plotted on his map. At the time of publication, the accuracy of his line-position measurements was such that he could often do without most post decimal-point digits for rough identification. For instance, a wavelength of 5567 Å still clearly designated an unfamiliar new line that Ångström had observed in the spectrum of the aurora borealis.126 However, rapid spectroscopic advances based on improved diffraction gratings soon required more decimal places to describe a line adequately, which made this labeling convention somewhat cumbersome. Simple inertia guaranteed the continued use of this unit by spectroscopists, who all used Ångström’s map and tables, as well as later ones based on his standards, but occasionally alternatives were suggested.127 The
Page 40 of 81

The Spectrum in Historical Context
British inch, in particular, had some staunch advocates, who simply refused (p. 59) to quote “wavelengths in modern French terms, adopted by Ångström in the latter years of his life, vice the ‘inches’ of his renowned and heroic Scandinavian forefathers”.128 But the British inch, measuring 2.54 cm, is considerably larger than the millimeter and wavelengths quoted in inches would only lead to smaller numbers with longer strings of post-decimal digits. On the other hand, it was advantageous if one decided to plot spectra not according to wavelength λ, but according to frequency v = c/λ, or equivalently, according to wave number n — 1/ λ, essentially indicating the number of waves per unit length. A prominent spectrum line such as 5888.98 Å for the sodium DI line in Ångström’s system thus translated into 43 085 in Piazzi Smyth’s wave-number scale.129 Note that with the change from wavelengths to their inverse, the color sequence of the spectrum maps was also inverted: Ångström and his followers counted from small to large wavelengths, i.e., from the violet (left) to the red (right), while Charles Piazzi Smyth (1819–1900), Astronomer Royal for Scotland, plotted from low to high wave numbers, i.e., from the red (left) to the violet (right), “exactly suitable to Fraunhofer’s now expugnable order of lettering the chief lines of the solar spectrum from Red A as the beginning to Violet H as the end.” While prismatic spectra compressed the red end and diffraction spectra the opposite end of the visible spectrum (cf. here Fig. 2.25), a frequency or wave-number plot promised to achieve a “most desirable mean between the oppositely exaggerated views of Prisms on the one side, and gratings on the other”.130 The wave-number (i.e., frequency) plot was particularly useful in the line-rich blue and violet regions because there it amplified distances between spectrum lines and thus decompressed line groups and bands “miserably cramped by the untoward qualities” of the wavelength scale adopted by Ångström.131 Thus, all spectrum maps made by this Astronomer Royal for Scotland were labeled in his idiosyncratic scale of number of waves per British inch, despite repeated cautions by friends that this scale would adversely affect the popularity of his maps outside Great Britain.132 Because of the prevalence of Ångström -based spectrum maps and the lack of fans of the British inch on the continent, (p.60) Piazzi Smyth’s convention remained somewhat idiosyncratic but it had another argument in its favor. The abscissa on his maps indicated wave numbers rather than wavelengths, and since wave numbers were directly proportional to frequency, his maps proved particularly useful to those more theoretically inclined minds who were looking for overtone series or other regular patterns in line and band spectra. (For more about this endeavor, which became quite popular in the last quarter of the nineteenth century, see here § 8.5f.)
Page 41 of 81

The Spectrum in Historical Context
Leaving aside these concerns about precision and theoretical adequacy of a spectrum scale, practical considerations also played a role. At least until the introduction of blazed gratings in the 1930s, whose grooves are cut in such a way as to concentrate the diffracted light in one preferred direction, diffraction gratings simply could not be used for many practical applications of spectroscopy, either because the intensity of the resulting diffraction spectrum was too low, or because the light source itself was too dim to allow examination with the less handy grating spectrographs. As late as 1921, both the director and the head of the research department of the Parisian municipal laboratory of chemistry complained about the available ‘normal’ spectrum atlases:
Unfortunately, this spectrum comes from a grating and its appearance is very different from the spectrum provided by a standard spectrograph; the dissimilarity is so great that the operator is left in an inextricable dilemma, finding it impossible to attribute wavelengths to the principal lines of its spectrum.133 Thus, despite the undisputed state of perfection of ‘normal’ spectrum representations, a definite need for their prismatic counterparts remained, which other spectroscopists soon hastened to satisfy.134 Seen from this angle, the story of spectrum representations is not a simple gradual replacement of one outmoded form of representation (i.e., prismatic spectra) with a newer one (i.e., grating-generated ‘normal’ spectra). The coexistence of different spectroscopic practices led to a parallelism in contemporary atlas types and plates. As the foregoing quote also shows, translating between the different images was by no means easy: considering the striking difference in appearance of a single line group in the two types of representations, I think it is fair to speak here of different ‘spectro-scopic domains’,135 or different visual subcultures within a broadly framed visual culture of spectroscopy. 2.7 Extension of the spectral range and its interpretation
Page 42 of 81

The Spectrum in Historical Context
Thus far I have been discussing improvements in scientific instrumentation that led to an increase in accuracy of spectroscopic measurements. Let us now turn to another important research strand that is intertwined with the former: exploration of the spectrum at and beyond the red and the violet. This agenda dates back to the discovery around 1800 of new types of radiation in these invisible spectral regions.136 It must be pointed out that it was by no means clear from the outset that, essentially, only one spectrum was involved, and a convenient framing concept such as electromagnetic radiation was, of course, only to emerge (p.61) with the work of Maxwell and Hertz in the second half of the nineteenth century. Prior to that, researchers often believed they had found a completely new type of radiation with its own spectrum, apparently overlapping neighboring ones that were thought to be produced by different entities. In the following I will concentrate on the exploration of what we now call the infrared spectrum. The research on so-called chemical rays (nowadays referred to as the ultraviolet spectrum) will be discussed in § 2.8 on the phosphorogenic spectrum and in § 6.4 in the context of the development of scientific photography.
2.7.1 William Herschel’s heat intensity curve William Herschel137 (1738–1822) is often hailed as the discoverer of infrared radiation in 1800. He himself was divided about how to interpret his results. At one point they seemed to suggest that the spectrum simply continued beyond the red end of the visible spectrum, and at other times they rather suggested a fundamentally new type of heat ray with qualities distinct from those of visible light. After empirically confirming the validity of the laws of reflection and refraction for these rays, Herschel initially opted in favor of a close similarity between heat rays and light. But their totally different properties of transmission and absorption in various media spoke against it: “we have a direct and simple proof, in the case of the red glass, that the rays of light are transmitted, while those of heat are stopped, and that thus they have nothing in common but a certain equal degree of refrangibility”.138 The red filter thus acted as an auxiliary to separate the two types of rays present in natural radiation from the Sun. In addition, the diffuse scattering of light from unpolished surfaces was totally different from that of radiant heat: while thick black velvet scattered heat rays very effectively, it absorbed light nearly completely; on the other hand, gold-leaf paper scattered light rays very well, but ‘invisible light’ much less.139
Page 43 of 81

The Spectrum in Historical Context In another well-known experiment, Herschel took three blackened thermometers and exposed one for five minutes each in different color zones of the spectrum, the other two thermometers remaining just outside the illuminated area to test for systematic errors. After recording the temperature readings systematically as a function of position, he found that the maximum of the heat spectrum was located outside the visible spectrum, beyond the red end (see Fig. 2.26). This was the more surprising since it conflicted with earlier determinations of the heating effect of various spectral colors.140 Herschel’s finding also implied that as the heat intensity curve rose against the refrangibility when moving, say, from the yellow region to the red end of the spectrum, the corresponding curve of visual intensity decreased in the same interval (see Fig. 2.27).141 This curve marks the transition from an earlier mode of inquiry into heat rays, mostly by means of mirrors and somewhat analogous to geometrical optics, to a research program (p.62) that could appropriately be called the spectroscopy of radiant heat because of the emphasis on its prismatic decomposition. As John Heilbron notes, this figure was also one of the first comparative visual displays of the results of quantitative measurements ever published.142 Herschel saw the different placements of the maxima, but particularly the partially counterposing tendencies of the photometric and thermometric curves, as indicative of a fundamentally different entity, ‘invisible light,’ giving rise to these thermal effects. (p.63)
Page 44 of 81

The Spectrum in Historical Context

Now when these [two curves] are compared, it appears that those who would have the rays of heat do also the office of light, must be obliged to maintain the following arbitrary and revolting propositions; viz,

Fig. 2.26. Herschel’s experimental setup with three thermometers, one on a projected solar spectrum and two placed beyond the spectral strip for comparison measurements. Copper engraving by J. Basire. From W. Herschel [1800b], plate xi, facing p. 292.

that a set of rays conveying

heat, should all at once, in a

certain part of the spectrum,

begin to give a small degree

of light; that this newly

acquired power of

illumination should increase,

while the power of heating is

on the decline; that when the

illuminating principle is

come to a maximum, it

should in its turn also,

Fig. 2.27. Herschel’s comparative plot of

decline very rapidly, and

visual R and thermometric S intensity

vanish at the same time with

curves of the spectrum. From W. Herschel

the power of heating. How

[1800d] pi. xx, facing p. 538.

can effects that are so

opposite be ascribed to the

same cause? First of all, heat without light; next to this, decreasing heat

but increasing light, then again, decreasing heat and decreasing light.143

His contemporaries were also undecided: John Leslie (1766–1832) and Christian Ernst Wünsch (1744–1828) believed that it was only a question of inadequately filtered optical rays, while Henry Charles Englefield (1752–1822), Thomas Young, and later also Jacques Étienne Bérard (1789–1869) took Herschel’s view that there is thermometric action beyond the red end of the spectrum, although Bérard located the maximum within the visible part of the red.144 The precise location of the maximum of the red region became the (p.64) subject of heated controversy,145 which was only resolved in 1819 with Thomas Johann Seebeck’s (1770–1831) demonstration that the maximum was dependent on the prism material.146 Similarly, the heterogeneity of characteristic features and the polarity in their respective physical actions were the main arguments inducing researchers, John William Draper among them, to postulate yet another independent agency, namely chemical rays, at the other end of the visible spectrum.147

Page 45 of 81

The Spectrum in Historical Context
When Johann Wilhelm Ritter148 (1776–1810) discovered the chemical action of invisible light on silver salts in 1801, he interpreted his finding, true to the Romantic natural philosophy then in vogue, as evidence of polarity in nature. He thought he had found the counterpart to William Herschel’s heat rays at the red end of the visible spectrum: aside from the diametrical opposition with respect to its placement at the violet end, Ritter’s rays were also associated with cold (vs. heat effects at the other extreme), with chemical reduction (or desoxydation, as Ritter called their chemical action), and with a stiffening effect on solids (as opposed to the loosening induced by heat rays).149 Ritter’s interpretation of his findings rapidly became known also in the English-speaking world via translations and abstracts. Even though Herschel was not influenced by Romantic natural philosophy, this argument nevertheless strengthened his own conviction about the ontological distinctness of heat rays: Aside from the betterknown visible spectrum, we now had—or so it seemed to the majority of Herschel’s and Ritter’s contemporaries—two new natural kinds: heat rays and chemical rays.
2.7.2 John Herschel’s thermograph In early 1840, William Herschel’s son, John William Frederick, developed what is called the thermograph technique for rendering heat radiation visible.150 He thoroughly blackened one side of a very thin sheet of white tissue paper with the smoke of a candle, then exposed the white surface to the solar spectrum generated by a heliostat and two flint-glass prisms of 45° in the position of minimum deviation. Moistening the tissue paper with alcohol, he could observe that where thermic rays collected more intensely the paper dried more rapidly than elsewhere. Heat spots formed as a result, or thermic images of the Sun that “traced out their extent and the law of their distribution by a whiteness so induced on the general blackness which the whole surface acquires by the absorption of the liquid into (p.65) the pores of the paper.”151 These conspicuous and intense patches were only “transiently visible” with this technique, but there was enough time to jot down a good drawing of their features as they intensified and evolved over time, before the alcohol evaporated completely away. Later Herschel succeeded in fixing the image by adding a dye to the alcohol, residues of which remained on the paper in greater concentrations where the alcohol had evaporated away more quickly.152 A photographic impression of the exposed spectrum intensifies just marginally towards both ends. The ‘thermic spectrum’ in Fig. 2.28, by comparison, builds up on one side only, in the infrared, with the exposure time increasing linearly from nos. 1 to 5. The big spot Y widens towards β, followed by the spots γ, δ, and the barely visible spot ε at the far left. Herschel’s series of sketches was redrawn and published as a stipple engraving by James Basire for the Philosophical Transactions of the Royal Society.
Page 46 of 81

The Spectrum in Historical Context

At first, the contemporaries were unsure about how to interpret these results. From Herschel’s correspondence we learn that he privately favored regarding these “heat spots” as evidencing a partial absorption by the terrestrial atmosphere of an otherwise unbroken (p.66) thermic spectrum. He thus imagined a continuous unabsorbed spectral range of radiant heat intermittently registered as a series of isolated spots.

Fig. 2.28. J. Herschel’s thermograph and photograph of the solar spectrum at five different exposure times. Stipple engraving by James Basire. From J. Herschel [1843a] pl. I, fig. 9.

I have made some very curious thermographic experiments lately which lead me to think that not much more than half the Sun’s radiant heat reaches the Earth’s surface, being absorbed in the atmosphere[—]at least I interpret the insulation of two heat spots at a great distance beyond the spectrum—thus [then follows Fig. 2.29].153

Melloni suggested that Herschel had merely recorded the discontinuous thermal absorption of his flint-glass prism.154 In March 1842, the American chemist John William Draper confirmed this “want of continuity” in the least refrangible part of the spectrum using the daguerreotype recording technique. He found not five but three strikingly dark zones beyond the red end of the visible spectrum which he called α, β, and γ, respectively. This (p.67) he considered its extremity beyond which no radiation could exist.155 Later, these features were rediscovered and explored further with the aid of sensitive alcohol thermometers, thermomultipliers, phosphorescence techniques, and finally photography.156

Page 47 of 81

The Spectrum in Historical Context

All in all, interpretation of these findings was far less straightforward than we may be inclined to think. It is hardly conceivable for us that these newer lines might be anything other than natural extensions of the visible spectrum beyond the red and violet. Some pioneer researchers (e.g., W. Herschel and J.W. Draper) did entertain this theoretical model for a while, but from the available evidence of the physical action on thermal, optical, and

Fig. 2.29. J. Herschel’s interpretation of the thermograph, 1840: Y signifies the position of Fraunhofer’s line F in the yellow, A is labeled as the luminous spectrum, B the chemical spectrum, extending further beyond the violet, C “the complete or unabsorbed thermic spectrum continuing past the red end of the visible spectrum, and C' the thermic spectrum incomplete, as it exhibits itself to the eye in my experiments”. Letter to J.W. Lubbock, 9 April 1849 (RS, HS 22.43). By permission of the President and Council of the Royal Society.

chemical detectors in the

different spectral regions, most chose another interpretation: They postulated

other types of radiation. According to both J. Herschel and J.W. Draper, solar

radiation and the radiation emitted from luminous terrestrial sources was a

heterogeneous mixture of visible, actinic, and calorific rays. The spectrum

generated by a prism or a diffraction grating and photographed or explored by

other means was understood as the superposition of three components: an

optical spectrum more or less within the confines of Fraunhofer’s lines A and

H,157 an actinic or chemical spectrum with its maximum beyond the violet end of the visible spectrum,158 and a thermic or calorific spectrum with its maximum below Fraunhofer’s line A.159 The experimental problem facing these

pioneers was how to disentangle these different components of normal radiation.

Herschel had demonstrated chemical action of the solar spectrum far beyond the

extreme red rays. Likewise Herschel and Melloni had both demonstrated the

generation of heat in the visible part of the spectrum.160 Evidently no clear-cut

separation between the different ontological domains was feasible, because the

various effects of the different types of radiation were not bounded by well-

defined spectral limits but variant only in magnitude. In the following Fig. 2.30,

the three strips illustrate the spectra of light, heat, and actinism positioned

relative to the Fraunhofer lines in the optical spectrum. The middle strip is

actually a redrawing of Herschel’s thermograph of 1840 (cf. here Fig. 2.28), with

its discontinuous features marked (p.68) α to ε, while the third, the actinic

spectrum, clearly exhibits the gap in the yellow-orange region so typical of

contemporaneous photographs. The dotted curved lines indicate the points of

maximum and minimum effect for the three spectral types: The curve with its

maximum at yellow refers to visible light, while the maximum at α refers to the

thermal spectrum. The actinic curve is split into two, with one broad maximum

in the violet region, and the other in the extreme red.

Page 48 of 81

The Spectrum in Historical Context

Traces of this spectral taxonomy so different from ours are found in the writings of well-informed and established scientists up to the 1870s.161

2.8 The phosphorogenic spectrum

Since the 1840s researchers

had been aware of another

Fig. 2.30. The coexistence of three

spectral type in addition to the

different types of spectra: luminous,

three distinguished in Fig. 2.30. thermal, and chemical. From Hunt

It is labeled at the top of the

[1844b], unnumbered handcolored plate,

figure as ‘fluorescent rays’. This reproduced here in color on the dust

“spectre phosphorogenique”

jacket.

was observed using substances

like Canton’s phosphorus or

Bolognian stone162 as detectors. Although, like the chemical spectrum, the

resulting image initially could not be fixed permanently, it could be made visible

temporarily by projection onto a fluorescent fluid (such as quinine) in a flat glass

container or onto a specially prepared screen coated with a powdered

phosphorescent substance that was glued to the paper (p.69) with gum

arabic.163 Without magnification it was difficult to distinguish well-defined lines

from the diffuse bands, and even more difficult to obtain accurate angular

deflections while having to work in the dark. Consequently, the first published

drawings of the phos-phorogenic spectra of calcium sulphide and barium

sulphide published by Antoine César Becquerel (1788–1878) in the early 1840s

show only a few luminous bands of considerable extension from the Fraunhofer

line F onwards into the violet and beyond.164 But eventually his son, Alexandre

Edmond Becquerel (1820–1891), who succeeded him as director of the Parisian

Museum d'histoire naturelle, made some progress.165 He and others in his

research tradition managed to “prove the existence of the same lines in the

[solar] spectrum formed by these [phosphorogenic] rays as in the luminous and

chemical spectra”.166 This coincidence was an important argument in favor of

the basic unity of the three—or four—different kinds of spectra, something

which had already been suspected by the elder Becquerel in 1823 but which

only became experimentally demonstrable around the mid-nineteenth

century.167

Page 49 of 81

The Spectrum in Historical Context

In 1852, Sir George Gabriel Stokes (1819–1903), Lucasian professor of mathematics at Cambridge University, decided to follow up an idea of John Herschel’s to substitute the fluorescent screen (prepared with quinine sulphate) with a phosphorescent one. Using a spectroscope, which incorporated a chain of three or four Fraunhofer prisms, he observed the action of the spectrum projected onto this screen and plotted “fixed lines of the solar spectrum in the extreme violet and in the invisible region beyond”. While waiting to deliver a lecture at the Royal Institution in 1853, he was experimenting with electric light discharged from their powerful Leyden jars and dispersed by high-quality quartz apparatus when he made the chance discovery that quartz optics absorbs far less light towards the violet end of the spectrum than conventional glass prisms and lenses. These two innovations allowed Stokes to extend the known spectrum considerably beyond the Fraunhofer line H into the region invisible to the unaided eye. He found out that the extent of this new region amounted to no less than six to eight times the length of the visible spectrum. Accordingly he coined the term “long spectrum”. Figure 2.31 illustrates his attempt to form natural groups of the many new lines observed. In continuing the labeling he was not at all (p.70) sure how the lines detected with his new visual method related to the lines and line-groups already spotted in earlier photographic work. So he decided not to use the capital letters employed by Draper in 1843, choosing instead lowercase letters, from k to p (in a later publication, up to s).168

This decision to resort to a new

nomenclature illustrates very

nicely the general problem of

gauging empirical findings

obtained with different types of instruments. Although both Draper and Stokes explored the solar spectrum in the same region of refrangibility, namely, beyond its violet end, Stokes simply did not recognize Draper’s spectrum in what he saw with his technique. And who could guarantee that the

Fig. 2.31. Stokes’s mapping of the extreme violet and the “invisible region beyond” the optical solar spectrum, ending at H. Steel engraving by J. Basire. From Stokes [1852a] pl. XXV. Cf. also Jamin [1858–66a] vol. 3, p. 483 for a woodcut illustrating an observer’s visual impression of these faintly luminescent spectra seen in complete darkness.

rays detectable with his visual

techniques really were the same as those that Draper had found with his

photochemical method? As long as interpretation of these different types of rays

remained unclear, there was no proof that the two types of spectra ought to be

identical. Thus first explorations into uncharted spectral regions often led to

incompatible, at best, untranslatable results, depending on the investigative

technique being employed.

Page 50 of 81

The Spectrum in Historical Context
By a similar technique, Wilhelm Eisenlohr (1799–1872) traced the solar spectrum further into the ultraviolet region, as it came to be called by the mid-1850s. For wavelength measurements he used a diffraction grating which had been ruled on smoked glass. And in 1862 Stokes mapped the ultraviolet emission and absorption spectra of several substances recorded on screens made of uranium glass or a coating of uranium phosphate powder, this time reaching as far as the increasing opacity of air for shorter wavelengths would allow.169 Three years later, E. Becquerel mapped the phosphorescence emission bands in 15 different solids relative to the main Fraunhofer lines of the visible spectrum with the aid of a standard spectroscope mounting.170 (p.71) Another important improvement was introduced in 1874 by the Genevan physicist Jacques-Louis Soret, who thought of placing a fluorescent substance like uranium glass in the focal plane of a standard spectroscope and observing the thus generated phosphoro-genic spectrum with another telescope slightly tilted with respect to the axis of the first.171 In 1883 Henri Becquerel found out how to modify this technique for the infrared region of the spectrum. This was not so easy since infrared rays have the strange property of annihilating rather than augmenting the phosphorescence of substances like Balmain’s pigment or zinc sulphide. What he did was to expose the phosphorescent plate to intense blue light for about 15–30 seconds, and immediately afterwards to the infrared part of the spectrum for one to two minutes. Then for a brief time, he could see traces of the infrared spectrum lines in the form of dark stripes on the green phosphorescing screen. In this way he was able to draw a map of the nearinfrared spectrum of half a dozen elements and six other substances.172 Even though he could reach wavelengths up to 13 000 Å, Becquerel’s ocular measurements were very inaccurate because the light intensity of the zincsulphide screens was very low. As a result, many of his readings could not be verified by later researchers.173
Page 51 of 81

The Spectrum in Historical Context
At roughly the same time, J.W. Draper finally succeeded in photographing phosphoro-genic spectra. The basic idea was to bring the photographic plate into direct contact with the phosphorescent screen after the latter had been exposed to the spectrum for a while, typically a few minutes. This plate would then record each phosphorescent area and allow more precise readings at a later time. But the rapid dimming after exposure prevented Draper from obtaining good results. Even the Fraunhofer lines of the solar spectrum did not show clearly because of the inherent blurriness of his ‘phosphorographs’. His explanation for this was a lateral spreading of the luminescence on the photographic plate into the dark areas of the Fraunhofer lines. In 1888 Eugen Lommel (1837–1899) at the University of Munich finally overcame this limitation by following Henri Becquerel’s example and confining himself to working in the regions where exposure to the spectrum counteracted the luminescence of the phosphorescent screen.174 However, as a cursory inspection of Fig. 6.16 (on p. 228) will confirm, this technique of phosphoro-photography could not compete in terms of line definition and sharpness with direct photography of other segments of the spectrum. Nor are any permanent traces left in this incursion into the ultraviolet region with Stokes’s fluorescent screens, which is but one step more direct. It was not until the 1890s that Victor Schumann (1841–1913) in Leipzig realized that the gelatine on the photographic plate was absorbing the ultraviolet rays. By devising special plates (later called Schumann plates) with virtually no gelatine content, he was able to remove one of the obstacles that had prevented photography from reaching below 1850 Å. Furthermore, he eliminated the absorption in air by constructing a vacuum spectrograph with lenses and prism made of white fluorspar. But it was a nerve-racking fight against air leaks in his spectrograph and other hurdles, and it was at the expense of his health that he obtained the first photographs of the hydrogen spectrum in the region around 1620 Å, which was later named (p.72) after him. He eventually even attained wavelengths down to c. 1000 Å.175 The only other way to avoid the absorption of the ultraviolet part of the solar or stellar spectra beyond the cut-off below 2900 Å caused by atmospheric ozone was to photograph it outside the terrestrial atmosphere. That became possible after World War II with V-rockets.176
2.9 New instruments for exploring the heat spectrum
Page 52 of 81

The Spectrum in Historical Context
With the development of better thermal detectors, good progress was made in exploring the heat spectrum well beyond the red end of the visible spectrum. The first such invention was the thermopile, designed by Leopoldo Nobili (1784– 1835) on the basis of the thermoelectric effect, which Thomas Johann Seebeck had discovered two years earlier. Accordingly, the change in resistance caused by a change in temperature of a pair of bars of antimony and bismuth effectively converted radiant heat into electric signals, which could be amplified and measured with a sensitive galvanometer.177 In the hands of Macedonio Melloni178 (1798–1854), who improved the sensitivity and reliability of this instrument considerably, the thermopile became a powerful tool. With it Melloni systematically measured the so-called ‘diathermancy’, or heat conductivity of various substances.179 Melloni was also the first explorer of the heat spectrum to be thoroughly convinced that he was roaming in the outer fringes of the visible spectrum rather than in completely alien ranges of another kind of ray, as most of his contemporaries (in particular the two Herschels and J.W. Draper) had come to assume.180
In late 1871 a Russian physicist and physiologist also used the thermopile in conjunction with various prisms made of flint glass, rock salt, or filled with fluid carbon bisulphide, to map the heat intensity in the solar prismatic spectrum. Sergei Iwanowitsch Lamansky (1841-?)181 was working in Heidelberg under Hermann von Helmholtz’s guidance at the time. His plot (Fig. 2.32) of the heat spectrum’s intensity as a function of frequency shows a (p.73) pretty smooth curve coming to a maximum beyond the red end of the visible spectrum (labeled ‘Ende’ in the figure), and then three successive dips in an overall rapidly decreasing intensity curve.182

Page 53 of 81

Fig. 2.32. Lamansky’s plot of the energy distribution in the solar prismatic spectrum as registered by a thermopile. From Lamansky [1872] pl. V. fig. 2.

The Spectrum in Historical Context

The second important innovation was the ‘actinic balance’ or bolometer as it was soon called, which also made use of the sensitivity of electrical resistance to thermal change. The bolometer consisted of two identical blackened strips of thin metal and two identical bridge coils installed in a Wheatstone bridge (cf. Fig. 2.33) with a connected galvanometer. For high temperature sensitivity (in the order of magnitude of 10−6°C per mm deflection of the galvanometer), one usually chose platinum, which has a high resistance-temperature coefficient, a small specific heat, and low heat conductivity. Besides having these suitable physical characteristics, this material could also be transformed into strips of the required thinness of 0.002 mm. This was done using so-called Wollaston wire of 0.1 mm diameter made of a silver sleeve around a platinum core of 0.0125 mm. After hammering the wire fiat, the silver sleeve was etched off to expose an extremely thin platinum strip.183

Briefly, measurement with such a bolometer proceeded as follows: By removing a screen, one of the two platinum strips was exposed to the thermal radiation of a predetermined part of the spectrum. Since the resistance of this arm changes, the deflection of the sensitive galvanometer needle directly indicated variations in the energy emitted by the relevant part of the spectrum. When a quartz prism was used to disperse the solar radiation—or (p.74) better still, a rock-salt or fluoride prism, which absorbs less of the incoming rays—and the spectrum was slowly guided across the platinum strip, the corresponding deflections of the galvanometer needle indicated the thermal energy as it was distributed throughout the spectrum. For the bolometer’s inventor, Samuel Pierpont Langley184 (1834–1906) it was thus possible to plot the spectral maxima and minima of the solar heat radiation.185

This self-made man and, since

1890, head of the newly

founded Astrophysical

Observatory of the Smithsonian

Institution in Washington, had at his disposal very large rocksalt prisms for the best possible dispersion. Most spectacular among them was his large 60°

Fig. 2.33. Wiring diagram of Langley’s bolometer (L) [1881a,b], Snow [1892] and Aschkinass [1896] (S & A), and Julius (J) [1892/93]. From Coblentz [1908] p. 444.

prism, which John Brashear in

Pittsburgh had cut from one of the biggest natural pieces of rock salt ever found.

Traditionally, to get from the directly observable prismatic angular deviation of a certain spectral line to its wavelength λ, one used so-called dispersion formulas which describe the change of refractive index n with wavelength λ.186 The simplest and thus most widely used formula had been suggested by Cauchy already in 1836:

Page 54 of 81

The Spectrum in Historical Context
with a, b, and c constants calculated by the least squares method. Alternatively, one could also use the so-called Redtenbacher formula:
the Briot formula:
(p.75) or the so-called Hartmann formula:
with an additional adjustable constant α depending on the prism material. It was usually defined at α≈ 1.2 for average crown and flint glass or even simply set equal to 1 for purposes of approximation. The Hartmann formula had the advantage of being easily resolvable to compute the wavelength λ:
with λ0 a constant for each spectroscope, which hence had to be measured only once. Moreover, eqn (2.1), with 1 /α set equal to one, allowed replacing the refractive index n by any directly observed quantity linearly dependent upon it, such as angular deviations, or micrometer readings.187 However, all these dispersion formulas were empirically derived from data taken from the visible part of the spectrum. As soon as researchers such as Langley or Schumann forged further into the infrared or ultraviolet, it was unclear to what extent one could rely on a naive extrapolation into unknown spectral regions, and which of these incompatible formulas beyond the visible spectrum one should choose. While taking measurements of the solar heat spectrum on top of Mount Whitney in September 1881, Langley “came upon a hitherto unknown cold band whose [angular] deviation indicated a (probably) very great wavelength.” Applying Cauchy’s formula to this band, subsequently designated Ω, Langley arrived at a troublesome result, “the formula declaring that no such index of refraction as I had measured was possible in the prism in question.”188 No one knew the dispersion curves for prisms of rock salt, quartz, or glass in the wavelength ranges that Langley eventually reached (up to 5 μ = 50 000 Å). Because simple extrapolation from the dispersion curves plotted according to other dispersion formulas would not do either,189 Langley was forced to use the only instrument that furnished a wavelength-scaled spectrum, a diffraction grating. But as mentioned earlier (cf. p. 56), concave gratings partially superimpose succeeding orders of interference: “overlapping spectra and feeble heat make the use of the grating too difficult”.190 That is why he devised the following setup combining both prism and grating (cf. Fig. 2.34). As we shall see, this gave him a definite wavelength value for crucial lines or bands, independently of these interpolation formulas.
Page 55 of 81

The Spectrum in Historical Context

The incoming radiation was focused by a mirror M to pass through a slit S1 and fall onto a small Rowland concave grating G. Depending on the adjustable position of a second slit S2, a selected portion of the grating’s spectrum was guided either by a flint-glass lens L1 or by suitably oriented mirrors onto a big rock-salt prism P before reaching the bolometer B.

(p.76)

At any given angle of deviation off the Rowland grating, this arrangement caused two or three orders ni, of the solar spectrum to be diffracted toward the prism with their wavelengths λi satisfying the relation

Fig. 2.34. Langley’s experimental setup with both a prism P and a grating G to disentangle different orders of interference in the heat spectrum. From Langley [1883d] fig. 1.

This ‘coincidence method’, which Rowland had just introduced in conjunction with the development of his concave gratings,191 thus linked a known wavelength, say, δ3 in the third-order visible portion of the spectrum with perhaps hitherto unknown wavelengths δ1 and δ2 in the infrared.192 However convenient for wavelength measurements, concave gratings also had their drawbacks. The bolometer could not simply be put near L1, because then it would register the integrated heat intensity of all orders of the spectrum diffracted in that particular direction. This is one of the main reasons why gratings found so little application in the early explorations of the infrared. In Langley’s clever arrangement, however, this composite of rays of various diffractive orders is dispersed again by the prism into beams of radiation in three distinctly separate spectral ranges (i.e., ultraviolet, visible, and infrared). Only one of these is focused by the second lens L2 and finally reaches the bolometer B. By slowly changing the position of the slit S2 and scrutinizing the visible portion of the spectrum formed by the rock-salt prism for certain prominent features, such as the sodium D line, one can unequivocally identify certain wavelengths δ2 in the order n 2, which (by means of eqn 2.2) then allows a precise gauging of unknown wavelengths δ1 of the order n 1 in the infrared part of the spectrum. When this procedure is followed for a sufficient number of well-defined rays, an empirical dispersion curve can be drawn for the (p.77) prism at hand, which allows determination of all other wavelengths by standard methods of interpolation.193
Page 56 of 81

The Spectrum in Historical Context
With this empirical dispersion curve, Langley was able to gauge the dispersive powers of his rock-salt and fluor-spar prisms and thus to work with them directly. This was essential for the fainter lines and bands because of the generally weak intensity of spectra produced by gratings. But Langley had another problem: what his bolometer recorded in conjunction with its sensitive galvanometer was the heat intensity in a certain small section of the spectrum decomposed by his rock-salt prism. However, the red region in this prismatic spectrum was too compressed compared with a normal spectrum (i.e., proportionate to wavelength, cf. here Figs. 2.25 and 2.36). To get the proper heat distribution, he had to convert this intensity curve into one plotted as a function of the wavelength λ. In order to do this correctly, he had to make sure that the total amount of energy was the same in both representations overall as well as for each spectral region. This constraint led him to a suitable graphical conversion method that dates back to J.H.J. Müller’s examination of the thermic action of solar heat rays in 1858 (on the following cf. Fig. 2.35).
In a two-dimensional coordinate system, the directly measured thermal energy distribution CD of the prism is plotted along the x axis. Straight above it is drawn the empirical dispersion curve EF (obtained from the prism-grating combination discussed above). Further to the left of the gauging curve, the x axis is divided into an arbitrary number of equal units (just four in the left part of Fig. 2.35) representing the normal wavelength scale. The corresponding points on the prismatic plot are found by drawing a horizontal line from each of the equidistant points on the y axis until it meets the gauging curve, and then drawing a vertical line down from each of these intersecting points until it reaches the x axis. Note that the nonlinearity of the gauging curve makes the points corresponding to the wavelengths of 0.4, 1.0, 1.6, 2.2, and 2.8 μ on the prismatic scale no longer equidistant: the large interval between 0.4 and 1.0 μ on the x axis has shrunk, whereas the compressed portion in the infrared part of the prismatic spectrum between 2.2 and 2.8 μ has expanded.
If one follows this procedure for a larger set of equidistant points than the four intervals depicted in the left part of Fig. 2.35, it is also possible to find the height of the thermal energy curve AB. The right part of the figure shows how the constraint of equal thermal energy for any two definite wavelengths is implemented in Müller’s graphic conversion method. Several other persons were also involved in Langley’s adaptation of this method, most notably Langley’s assistant James Edward Keeler (1857–1900).194
Page 57 of 81

The Spectrum in Historical Context

The thermal energy is the gray shaded area ab of the prismatic plot, bounded by the two verticals at wavelengths λ1 and λ2 This surface area has to be equal to the gray shaded area cd of its corresponding normal energy curve. In practice, both a and c have to be (p.78) sufficiently small in order to get satisfactorily smooth curves—in the infinitesimal limit, the ratio of d : b is equal to tan ϕ, with ϕ being the angle formed by the tangent to EF at their point of intersection. It is important to note that the maximum of the heat-intensity curve will also ‘shift’ (from 1.0 to 0.5 in the above example).

A comparison of Langley’s prismatic spectrum from 1883 (Fig. 2.36, top) with the normal spectrum constructed from it (bottom) shows how much the known optical spectrum was extended into the infrared region. The range that the visible part occupied shrinks from over 50 % of the prismatic spectrum to less than a fifth of the total range in the standard wavelength plot (then ending at 28 000 Å).

Fig. 2.35. Langley’s graphic procedure for converting a prismatic spectrum into a wavelength plot (for details cf. main text). From Langley [1883d] figs. 3–4.

By comparing several such infrared spectra taken from solar light at various times of the day and thus through differing atmospheric thicknesses, Langley could also demonstrate that the major troughs in the spectrum, already indicated in Herschel’s thermograph (see above p. 65), were due to absorption in the Earth’s atmosphere. Later research proved that Langley’s intricate way of determining the wavelengths of solar spectrum lines and bands was riddled with averaging uncertainties. Langley’s estimates in the region 8500 to 11 000 Å deviated by about 4 Å, and for a few lines much more.195 Nevertheless much territory had been gained since the first groping explorations of the dark thermal fringe of the spectrum.

(p.79)

Page 58 of 81

The Spectrum in Historical Context

Notes:
(1) The colorful story about the changing interpretations of the rainbow is told by Boyer [1959], with references to primary texts.

(2) The technology of glass-

making goes back to the third

millennium BC. But the early

Mesopotamian and Egyptian

cultures only had opaque types

of colored glass, which were often used as imitation gem stones.

Fig. 2.36. Langley’s graph of the solar spectrum into the infrared, up to 28 000 Å; above in the prismatic mode, and

(3) See Seneca’s Naturales

below in the reconstructed ‘normal’

Quaestiones, book 1, part 7, § 1; mode. From Langley [1883a] pl. III.

book X of Witelo’s Perspectiva

(c. 1270, copied many times in

manuscript and later included in F. Risner’s Opticae Thesaurus Vitellonis of

1572); Roger Bacon’s Opus Majus (c. 1266/67); and Wiedemann [1912] on Kamāl

al-Dīn.

(4) See, e.g., Rosen [1956].

(5) On the following see the manuscript (RCW, no. 19149 r); cf. also Richter [1883], no. 288 for a translation of Leonardo’s commentary.

(6) Quoted from the OED, 2nd edn (1989), vol. 12, p. 510, original spelling.

(7) See Delia Porta [1558d] p. 954. The respective passage is omitted in an English translation from 1658 of the same book (based on the second Latin edition of 1589).

(8) The quote is given in OED from a translation of Johann Amos Comenius, The Gate of the Latine Tongue Unlocked, London, 1656, §480, p. 139; Schaffer [1989] p. 73 also quotes Thomas White (1654) and Kenelm Digby (1669) about “Fools Paradises”.

(9) This Czech physician, mathematician, and natural philosopher on the medical faculty of the University of Prague, was offered a chair at the University of Oxford in 1662 and membership in the Royal Society soon after its official founding; see, e.g., Kuba [1974], Marek [1998]. On Marci’s influence on Barrow and Newton see also Rosenfeld [1932], Marek [1969] pp. 392ff., and Hall [1993] pp. 21f.

Page 59 of 81