3o6 



NATURE 



[July 28, 1898 



Age or prolonged annealing may introduce an appreciable change 

 in physical properties. 



In addition to the work already referred to, a comparison has 

 been made at Sevres of the hydrogen scale with the natural 

 scales of low range glass-alcohol and glass-toluene thermometers. 

 At - 70° C. on the hydrogen scale the toluene temperature is 

 - 56° -63 C, as against about - 63° C. on the alcohol scale. 

 Notwithstanding its greater contraction of scale at low temper- 

 atures, toluene is preferred by Dr. Chappuis to alcohol, on the 

 ground that the latter is much more difficult to get of uniform 

 purity. At - 70° C. differences of as much as 1° were observed 

 in thermometers filled with alcohol supplied as pure by different 

 first-rate chemists. Even with toluene thermometers, o°'oi C. 

 appears the limit of accuracy to be hoped for. English alcohol 

 thermometers, I should explain, are not, as a rule, constructed to 

 give temperatures on the glass-alcohol scale. The degree 

 divisions are shortened as we go down the scale, in such a way 

 as to make the thermometer, when exposed to freezing mercury, 

 read - 37°*9 F. ; this being the air thermometer temperature for 

 freezing mercury according to Balfour Stewart's determination. 



The thermometric work at the German Reichsanstalt^ has 

 included the comparison of the verre dtir scale with that of 

 several German glasses, notably the Jena glasses 16111 and 59'". 

 The former glass is fairly similar in character to verre a'ur ; 

 the latter is a boro-silicate glass capable of resisting very high 

 temperatures, and showing exceptionally small depression of 

 zero. Thermometers made of it, with compressed gas above 

 the mercury to prevent boiling, supply a convenient means of 

 measuring temperatures up to 500° C. or even 550° C. In such 

 high temperature measurements it is often difficult to avoid 

 having a long mercury column emergent above the bath or other 

 source of heat whose temperature is in question. The con- 

 sequent error can be found apparently with great accuracy by 

 means of a special form of long bulb thermometer (" Faden- 

 thermometer"). Dr. Guillaume, who apparently anticipated 

 the Reichsanstalt observers in the idea, has curiously enough 

 found it foreshadowed in the Phil. Trans, for 1777. Using the 

 " Faden-thermometer," the Reichsanstalt observers apparently 

 claim an accuracy of o°' I C. in comparisons made in a well-stirred 

 bath at 500° C. They claim, however, an accuracy of o°'02 C. 

 in comparisons of Jena glass thermometers with the air thermo- 

 meter between 100° and 300° C. Until these results are con- 

 firmed or similar accuracy is claimed by the Sevres observers, 

 a chronicler may perhaps be pardoned an attitude of reserve. 



Even with the aid of compressed gas, the range that can be 

 covered by a mercury thermometer is somewhat limited, in view 

 of modern requirements ; and within that range there are many 

 cases in which other means of measuring temperature are prefer- 

 able. Nearly every property ot every natural substance is 

 modified by heat, so that the possible ways of measuring 

 temperature are practically innumerable. Several of the ways 

 that have been proposed for measuring high temperatures are 

 very ingenious and may have a great future before them ; but 

 the methods that have actually been utilised to an appreciable 

 extent are but few. Of these the two that have been most to 

 the front of late years have depended on the measurement of 

 electric resistance and electromotive force respectively. The 

 former method we may regard as embodied in the platinum- 

 resistance thermometer. Its introduction and the improvements 

 it has undergone are due mainly to Prof. Callendar and Mr. 

 E. H, Griffiths, while its application to the determination of 

 melting points of metals and alloys is largely due to Mr. C. T. 

 Heycock and Mr. Y. H. Neville. A clear description of the 

 necessary apparatus and the mode of graduating platinum 

 thermometers was given by Mr. Griffiths in Nature, November 

 J^95> P- 39- The essential fact is that a piece of platinum 

 wire, suitably protected, is exposed to the temperature it is 

 desired to measure, and its electrical resistance is found by a 

 Wheatstone's bridge method. If Rj be the resistance in steam, 

 Rq in ice, R at any other temperature, then 



pt= 100 (R - Ro)-^(Rl - Ro) 



is termed the "platinum temperature." In common use //is 

 employed only to deduce a quantity t, connected with it through 

 the relation 



t -pt = ^(t|loof - {tjioo)], 



where 5 is a constant, so chosen that t equals 444*53 when the 



1 Described in various memoirs in the Reichsanstalt's IVissens. Abhandl. 

 and in the Zeitschriftfiir Instrumentenkunde. 



platinum wire is at the temperature of the vapour of sulphur 

 boiling under standard pressure. 



The investigations of Callendar, Griffiths, Heycock and 

 Neville show that the values obtained in this way for /, over a 

 range of at least 1000° C., are very close in different samples of 

 platinum wire, so that / represents temperature on what is at 

 least very approximately a definite fixed scale. Further, Prof. 

 Callendar found that the scale so arrived at approximates very 

 closely to that of the air thermometer (at constant pressure) over 

 at least the range 0° to 600° C. ; whilst the values of / obtained 

 by Messrs. Heycock and Neville for the melting points of silver, 

 gold and copper, lie pretty close to the corresponding air 

 temperature results obtained by Holborn and Wien at the 

 Reichsanstalt. 



If the wire of all platinum thermometers possessed the same 

 value of 8, then every platinum thermometer would give the 

 same pt when exposed to the same temperature t. We should 

 then have a definite independent platinum scale, precisely as we 

 now have a definite verre dur mercury scale between 0° and 

 100° C. 



In reality, however, 5 varies considerably — over at least 25 

 per cent. — in existing platinum thermometers, so that the present 

 use of the term " platinum temperature" is open to criticism. 



The question as to what is the best formula for use in 

 platinum thermometry has been discussed by Mr. Hamilton 

 Dickson recently in the Phil. Mag. (December 1897, p. 445, 

 and June 1898). After considering Prof. Callendar's principal 

 formulae, and others suggested at one time or another by him 

 and Mr. Griffiths, Mr. Dickson decides in favour of the species 



CR + af = p{t+ b), 



where a, b, p are constants, and t is the temperature answering 

 to a resistance R in the platinum wire. Mr. Dickson applies 

 this formula to Prof. Callendar's original comparison with the 

 air thermometer, to certain melting point determinations by 

 Prof. Callendar and Mr. Griffiths, and to low temperature com- 

 parisons by Profs. Dewar and Fleming and by Messrs Holborn 

 and Wien. 



Determining the constants in the several cases by the method 

 of least squares, he finds the probable divergence of observed 

 and calculated values to be of the order o°"25 C. 



The formula approved by Mr. Dickson is really of the type 



t = a + bY^ + fR2, 



employed previously by Holborn and Wien in discussing observ- 

 ations made by them at the German Reichsanstalt. These 

 gentlemen, perhaps owing to their less exact method of deter- 

 mining the constants, claim for their formula accuracy only of 

 the order 1° C. Their comparison with the air thermometer 

 extended down to -190° C, so that it seems in any case a 

 valuable tribute to the suitability of platinum thermometers for 

 the measurement of low temperatures. 



At high temperatures Holborn and Wien's experience of the 

 platinum thermometer was not very favourable, the wire show- 

 ing appreciable permanent changes. As Mr. Griffiths, how- 

 ever, points out, these changes occurred at temperatures to 

 which platinum thermometers of the type he approves have fre- 

 quently been exposed without any apparent ill effect. The 

 preference expressed by Holborn and Wien for thermo-electric 

 methods thus perhaps carries less weight than it might seem to 

 deserve at first sight. It would certainly appear, as pointed out 

 by Mr. Griffiths in Nature, vol. liii. p. 390, that the deter- 

 minations of the melting point of copper, about 1080° C, by 

 Heycock and Neville, with a variety of different platinum ther- 

 mometers, agree considerably better amongst themselves than 

 the corresponding results obtained by Holborn and Wien with 

 thermo-couples. 



Be this as it may, there can be no doubt that thermo-couples 

 are very convenient instruments for high temperature measure- 

 ments, and they have had hitherto a considerably wider use than 

 platinum thermometers. 



The physical quantity whose variations in the thermocouple 

 give temperature variations, is the total electromotive force in a 

 circuit. The mainly active part of the circuit consists of two 

 metals, one of whose common junctions is usually kept at a 

 known fixed temperature, the other being exposed to the tem- 

 perature it is desired to measure. The most widely used couple 

 of late years has been Le Chatelier's, in which one metal is 

 platinum, the other an alloy of platinum with rhodium (10 per 

 cent, rhodium). The substitution of iridium for rhodium is not 



NO. 1500, VOL. 58] 



