Inventing Temperature

by Hasok Chang

  credible step in that direction was taken by the Anglo-Indian antiquarian James Prinsep (1799-1840), who was also the assay master at the Calcutta mint.46 He began with a bold condemnation (1828, 79-81): "If all the experiments had been recorded, which at different times must undoubtedly have been made on the subject of Pyrometry … the catalogue would consist principally of abortive attempts, if not of decided failures." Dulong and Petit's work was valuable, but only for relatively low temperatures. The Wedgwood pyrometer was the only instrument applicable in the higher heats produced by furnaces, but "a slight practical acquaintance with metals and crucibles" was sufficient to teach one that Wedgwood's results were not reliable. Prinsep thought that Daniell's more recent work was much more promising, but still he saw some problems in the design of Daniell's instrument, which he thought were manifested in the lack of "a desirable accordance in the result of different trials."47

  Prinsep first tried to construct an air thermometer with a cast-iron reservoir. After experiencing various technical difficulties, Prinsep finally opted for a much more expensive solution (1828, 87-89): "a retort or bulb of pure gold, weighing about 6,500 grains troy [about 420 g], containing nearly ten cubic inches of air." This gold-based instrument was robust, but he recognized two problems of principle. First, the thermal expansion of gold was not well known, so it was difficult to correct for the errors arising from the expansion of the vessel. Second, he was not so convinced about the correctness of "the absolute law of gaseous expansion" in the pyrometric range, either. There were also practical difficulties, most of which were common to all air thermometers. Nonetheless, Prinsep (1828, 95) carried out some elaborate measurements and concluded that his results were unequivocal on certain important points, particularly on the melting point of silver: "[T]hese experiments … are sufficiently trustworthy to warrant a reduction in the tabular melting point of pure silver of at least 400 degrees [Fahrenheit] below the determination of Mr. Daniell, while they indisputably prove the superiority of that gentleman's thermometric table as contrasted with that of Mr. Wedgwood."

  When Prinsep discarded iron and went for gold, he was not only compromising on economy but the range as well, as iron could withstand a much higher degree of heat than gold. Given what people knew about metals at that time, there was only one hope: platinum. But, as noted earlier, the handling of platinum was still a very difficult art in the early nineteenth century. In fact J. G. Schmidt (1805) in Moldavia had already proposed making a pyrometer with air enclosed in a platinum container, but there is no indication that he ever executed this idea; Guyton (1811b, 103-104) could not imagine making such a contraption without soldering platinum plates,

  46. See Encyclopaedia Britannica, 11th ed., for brief biographical information about Prinsep.

  47. There is some irony in Prinsep's attack on Wedgwood. The prime example he gave of the unreliability of the Wedgwood pyrometer was the overly high melting point of silver, particularly the fact it was placed above the melting point of copper. As Prinsep noted, Wedgwood had put forward these erroneous silver and copper melting points "on the authority of Mr. Alchorne," who had performed the experiments for Wedgwood (see also Wedgwood 1782, 319). What kind of authority was Alchorne? He was the assay master at the Tower of London, a man with a bit more than "a slight practical acquaintance with metals and crucibles"!

  end p.138

  which would make the instrument only as robust as the soldering material. Similarly, according to Prinsep (1828, 81), Andrew Ure had recommended "an air thermometer made of platina," and even got such an instrument made for sale, but no reports of any experiments done with them could be found.

  By 1836, however, Pouillet managed to construct an air pyrometer with the reservoir made out of a single piece of platinum. I have already discussed Pouillet's work at the low temperature end in "Consolidating the Freezing Point of Mercury," and in fact one of the air thermometers he used for that work had originally been constructed for pyrometric purposes. The platinum-based air thermometer was capable of recording temperatures well over 1000°C (about 1830°F). As shown in table 3.2, the melting points of metals that Pouillet (1836, 789) obtained by this means were mostly quite consistent with values obtained by Daniell with his platinum pyrometer. Having the air thermometer readings available to such high temperatures also aided the development of calorimetry because it enabled specific heat measurements at high temperatures. Pouillet (1836, 785-786) reported that the specific heat of platinum increased steadily, going from 0.0335 around 100°C (212°F) to 0.0398 around 1600°C (2912°F). This knowledge allowed him to estimate the melting point of iron by water calorimetry, by putting a piece of platinum in the same heat that melted iron, and then performing calorimetry on the platinum piece. The resulting value was 1500-1600°C (roughly 2700-2900°F) for the melting point of iron.

  After that whirlwind tour of early pyrometry, we can now come back to the question that we set out to answer: on what grounds did people decide that Wedgwood's temperature values were incorrect? For quite a while after the Wedgwood pyrometer was generally rejected, none of the available alternative pyrometric methods were clearly superior to Wedgwood's, either in principle or in practice. Le Chatelier's harsh retrospective judgment on nineteenth-century pyrometry is not entirely an exaggeration: Since Wedgwood, many have undertaken the measurement of high temperatures, but with varying success. Too indifferent to practical requirements, they have above all regarded the problem as a pretext for learned dissertations. The novelty and the originality of methods attracted them more than the precision of the results or the facility of the measurements. Also, up to the past few years, the confusion has been on the increase. The temperature of a steel kiln varied according to the different observers from 1,500° to 2,000°; that of the sun from 1,500° to 1,000,000°. First of all, let us point out the chief difficulty of the problem. Temperature is not a measurable quantity in the strict sense of the term. … It is evident that the number of thermometric scales may be indefinitely great; too often experimenters have considered it a matter of pride for each to have his own. (Le Chatelier and Boudouard 1901, 2-3)

  An inspection of the data collected in table 3.2 shows that by the middle of the nineteenth century the comparability of each of these methods had not been established: either the data were too scant for comparisons to take place or the measurements of the same phenomena obtained by the same method differed considerably from each other. In fact, in terms of comparability, it could easily be

  end p.139

  argued that the Wedgwood clay pyrometer was superior to alternative methods because the results obtained by this method by Wedgwood, Guyton, Clément and Desormes, and Pouillet were in close agreement with each other for most phenomena (see the first column of data in table 3.2). Guyton (1811a, 83-84) very ably defended the Wedgwood pyrometer and its comparability. Fourmi had just published an argument that the contraction of the Wedgwood pyrometric pieces was a function of the exposure time, as well as the temperature to which it was exposed. Guyton took Fourmi's own data and argued quite convincingly that they actually showed a remarkable degree of comparability between different trials with very different amounts of exposure time.48

  The non-Wedgwood methods did not agree all that well between themselves, either. However, as we can see at a glance in table 3.2, it was still very clear that the numbers produced by them tended to agree much more with each other than with Wedgwood's. As Guyton put the matter already in 1811: I believe that we can conclude that the values assigned by Wedgwood to the degrees of his pyrometric scale ought to be reduced considerably, and that all the known means of measuring heat contribute equally toward the establishment of that result, from the zero of the thermometer to the temperature of incandescent iron. (Guyton 1811b, 112)

  Beyond the melting point of gold, the clay pyrometer readings, even as recalibrated by Guyton, were distinctly far away from the range where the numbers produced by other methods tended to cluster.

  That is where matter
s stood for quite some time. The transition into the kinds of pyrometry that would be recognizable at all to modern physicists and engineers did not occur until the last decades of the nineteenth century.49 The most important basis of modern pyrometry is a quantitative knowledge of the radiation of heat and light from hot bodies, and of the variation of the electrical properties of matter with temperature. Such knowledge could not be gained without basing itself on previous knowledge gained by the types of pyrometry discussed in this chapter. Modern pyrometry was the next stage in the saga of the extension of the temperature

  48. Fourmi exposed various Wedgwood clay pieces to very high degrees of heat, around or beyond the melting point of cast iron, for repeated periods of 30 to 40 hours each. For instance, one piece shrank to the size corresponding to 146°W, after one period of exposure to a heat estimated by Fourmi at 145°W; two more exposures each at 145°W brought the piece only down to the size of 148°W; another exposure estimated at 150-151°W brought it to 151°W. Another piece (no. 20), on the other hand, contracted to 151°W after just one exposure at 150-151°W. Later commentators, however, have sided with Fourmi's verdict. Daniell (1821, 310) voiced the same opinion and that is also in line with the modern view, as indicated by Chaldecott (1975, 5).

  49. Matousek (1990, 112-114) notes that electrical-resistance pyrometry was only proposed in 1871, by William Siemens; radiation pyrometry started with the Stefan-Boltzmann law, discovered around 1880; optical pyrometry was pioneered by Le Chatelier in 1892; thermoelectric pyrometry did not become reliable until the 1880, although its basic idea can be traced back to Seebeck's work in the 1820s. (We have seen that Pouillet began to gain confidence in the thermoelectric method in the low-temperature range in the 1830s; Melloni used it in the same period with great effect in his study of radiant heat, but not as a pyrometer.)

  end p.140

  scale, and I expect that its development involved the same kind of epistemic challenges as we are examining here.

  The Wedgwood pyrometer was discredited long before the establishment of the methods we now trust. All in all, it seems that the Wedgwood pyrometer met its demise through a gradual convergence of a host of other methods all lined up against it. But does such epistemic ganging up prove anything? The Wedgwood pyrometer continued to be used with practical benefit well into the nineteenth century. Even if we disregard Le Chatelier's retrospective hyperbole that the Wedgwood pyrometer was "the only guide in researches at high temperatures" for "nearly a century," we cannot dismiss the estimate in E. Péclet's 1843 textbook on the practical applications of heat that the instrument of "Vedgwood" was the most generally employed pyrometer, even sixty years after its invention.50 There are various other reports showing the uses of the Wedgwood pyrometer later in the century, too.51 What exactly was gained by declaring it to be incorrect, on the basis of a convergence of various other methods that were each insecure in themselves? These questions will be addressed more systematically in the analysis part.

  Analysis: The Extension of Concepts beyond Their Birth Domains

  [Physics] has come to see that thinking is merely a form of human activity … with no assurance whatever that an intellectual process has validity outside the range in which its validity has already been checked by experience.

  P. W. Bridgman, "The Struggle for Intellectual Integrity," 1955

  To make and describe scientific observations and measurements, we must make use of certain concepts and material instruments. These concepts and instruments embody certain regularities. In the first two chapters we have seen how the great difficulties involved in establishing such regularities can be overcome, at least to some extent. However, there are new challenges in extending these regularities to new domains of phenomena, so that the concepts can function usefully and meaningfully there. The double narrative given in the first part of this chapter gives

  50. See Le Chatelier and Boudouard 1901, 1; Péclet 1843, 1:4.

  51. According to Rostoker and Rostoker 1989, the Ordnance Manual for the Use of Officers in the United States (1841) gave the melting point of steel as 160°W; H. C. Osborn, also in America, used the Wedgwood pyrometer to help the manufacture of "blister" steel, as he reported in 1869. In France, Alphonse Salvétat (1857, 2:260), the chief of chemistry at the porcelain works at Sèvres, criticized the Wedgwood thermometer but still reported that the Wedgwood degrees for the melting points of gold, silver, and cast iron were sufficiently exact.

  end p.141

  ample illustrations of those challenges. Now I will give a more thorough and general analysis of this problem of the extension of concepts, in their measurement and in their meaning. I will start in "Travel Advisory from Percy Bridgman" with a more careful characterization of the challenge of extension, with the help of Percy Bridgman's ideas on operational analysis. In "Beyond Bridgman" I will argue that Bridgman's ideas need to be modified in order to avoid the reduction of meaning to measurement, which makes it impossible to question the validity of proposed conceptual extensions. After these preliminary steps, "Strategies for Metrological Extension" and "Mutual Grounding as a Growth Strategy" will present "mutual grounding" as a strategy of extension that can help knowledge grow in the absence of previously established standards.

  Travel Advisory from Percy Bridgman

  The scientists discussed in the narrative part of this chapter were explorers into unknown territories—some of them literally, and all of them metaphorically. There were clear dangers and mirages awaiting them in the new lands. Their journeys would have been much easier with a travel advisory from a knowledgeable authority, but there were no such authorities then. However, there is no reason why we should not retrace, analyze, and reconsider their steps, thinking about how they could have avoided certain pitfalls, where else they might have gone, or how they might have reached the same destinations by more advisable routes. There will be fresh understanding and new discoveries reached by such considerations.

  In our own journey we should seek the help of any guides available, and I cannot imagine a better one than Percy Williams Bridgman (1882-1961) (fig. 3.6), the reluctant creator of "operationalism,"52 whose pioneering work in the physics of high pressures was rewarded with a Nobel Prize in 1946. His chief scientific contribution was made possible by technical prowess: in his Harvard laboratory Bridgman created pressures that were nearly 100 times higher than anyone else had reached before him and investigated the novel behavior of various materials under such high pressures. But as Gerald Holton has noted, Bridgman was placed in a predicament by his own achievements: at such extreme pressures, all previously known pressure gauges broke down; how was he even to know what level of pressures he had in fact reached?53 That was just the same sort of pitfall as Braun fell into by his success with freezing mixtures, which made the mercury thermometer

  52. Bridgman denied that he ever intended to create a rigid and systematic philosophy. In a conference session devoted to the discussion of his ideas in 1953, he complained: "As I listened to the papers I felt that I have only a historical connection with this thing called 'operationalism.' In short, I feel that I have created a Frankenstein, which has certainly got away from me. I abhor the word operationalism or operationism, which seems to imply a dogma, or at least a thesis of some kind. The thing I have envisaged is too simple to be dignified by so pretentious a name; rather, it is an attitude or point of view generated by continued practice of operational analysis" (Bridgman in Frank 1954, 74-75).

  53. See the entry on Bridgman in The Dictionary of Scientific Biography, 2:457-461, by Edwin C. Kemble, Francis Birch, and Gerald Holton. For a lengthier treatment of Bridgman's life and work viewed within broader contexts, see Walter 1990.

  end p.142

  Figure 3.6. Percy Williams Bridgman. Courtesy of the Harvard University Archives.

  inoperable. The situation was even starker for Bridgman because he kept breaking his own pressure records, creating the need to establish pressure measures fit for a succession of higher and higher p
ressures. Therefore it is no surprise that Bridgman thought seriously about the groundlessness of concepts where no methods were available for their measurement.

  end p.143

  With the additional stimulus of pondering about the lessons from Albert Einstein's definition of simultaneity in his special theory of relativity, Bridgman formulated the philosophical technique of operational analysis, which always sought to ground meaning in measurement. The operational point of view was first set out systematically in his 1927 Logic of Modern Physics and became very influential among practicing physicists and various thinkers inspired by the tradition of American pragmatism or the new philosophy of logical positivism. In the rest of this chapter we shall see that Bridgman had much careful warning to offer us about extending concepts beyond the domains in which they were born. Bridgman's concern about the definition and meaning of scientific concepts was also forged in the general climate of shock suffered by the early twentieth-century physicists from a barrage of phenomena and conceptions that were entirely alien to everyday expectations, culminating with quantum mechanics and its "Copenhagen" interpretation. In a popular article, Bridgman (1929, 444) wrote: "[I]f we sufficiently extend our range we shall find that nature is intrinsically and in its elements neither understandable nor subject to law."

  But as we have seen in the narrative, the challenges of the unknown were amply present even in much more prosaic circumstances. Bridgman was aware of the ubiquity of the problem and chose to open his discussion of operational analysis with the example of the most mundane of all scientific concepts: length. He was both fascinated and concerned by the fact that "essential physical limitations" forced us to use different operations in measuring the same concept in different realms of phenomena. Length is measured with a ruler only when we are dealing with dimensions that are comparable to our human bodies, and when the objects of measurement are moving slowly relative to the measurer. Astronomical lengths or distances are measured in terms of the amount of time that light takes to travel, and that is also the procedure taken up in Einstein's theorizing in special relativity; "the space of astronomy is not a physical space of meter sticks, but is a space of light waves" (Bridgman 1927, 67).