November 14, 1895] 



NATURE 



41 



had on three occasions been removed and replaced by a 

 one. 

 It the close of these operations (December 1,9, 1894) the 

 le of FI had risen to 101*003, ^" increase of 0*36 per cent. 

 It should be remembered that in each determination the 

 stances were raised to 50° or 100° above their freezing-points 

 rfore the observations were taken ; for example, the freezing- 

 int of potassium sulphate is given as 1066° C, but it is certain 

 It when determining this point the pyrometer was previously 

 ed to a temperature considerably exceeding 1100°. A study 

 of the original table will show that the rate of increase in FI 

 diminishes with use. 



As this question of constancy is of vital importance, Messrs. 

 Heycock and Neville have given me permission to state that 

 they have used only one pyrometer during a continuous series of 

 high-temperature determinations extending over two months of 

 the past summer. When the account of their work is published, 

 it will be found that although the number of the observations on 

 the freezing-points of alloys exceeds some hundreds, the 

 pyrometer is as efficient now as at the commencement of their 

 work. Its FI on July 28 was 100-148, on August 20, 100-357, 

 and on nearly all the intervening days it had been immersed in 

 molten metal at temperatures between 900° and 1000° for five or 

 six hours at a time. A determination of the freezing-point of 

 copper, made at the close of the above series of experiments, 

 gave a value practically identical with that previously published. 

 Apart from the slight change in FI, above illustrated, there 

 is abundant evidence that when completely protected from the 

 action of furnace gases the platinum wire 

 undergoes no change. Space does not per- 

 mit the accumulation of further evidence, 

 but full information will be found in the 

 papers already referred to. 



II. Description of the Apparatus. 



The facts dwelt upon in Section I. show 

 that if the methods of platinum thermo- 

 metry are adopted, the measurement of 

 temperature becomes a question of the 

 measurement of electrical resistance, and 

 there are few physical quantities which 

 can, if due precautions are taken, be mea- 

 sured with greater accuracy than the 

 resistance of a conductor. The Kew 

 apparatus, therefore, may be regarded as 

 designed for the accurate measurement of 

 the resistance of a platinum wire, and some 

 of the contrivances introduced with the c 

 object of securing greater accuracy are, I 

 believe, peculiar to this apparatus. 



The designs were drawn up by Prof. Callendar and myself, 

 after consultation with Mr. Horace Darwin, and the apparatus 

 was constructed by the Cambridge Scientific Instrument Com- 

 pany, Ltd., under the personal direction of Mr. Pye. 



Fig. I is a diagrammatic view of the connections. 



The coils S, and S, are of equal resistance (about 5 ohms), Q 

 is a set of resistance coils, A B a bridge-wire, and K a thermo- 

 electric key. When the resistances between Cj and C2 and Pi 

 and Pj are equal, the bridge is balanced if the resistance at Q 

 is zero, and the contact-maker H is at the mark O near the 

 centre of the bridge- wire. The scale of this wire is so graduated 

 that if the reading to right or left of O be added or subtracted 

 from r (the resistance at Q), the result gives the value of P - C 

 where P is the resistance between P^ and Pj, and C the resist- 

 anccvbetween Cj and Cj. 



Now P = / -f C/, where / is the platinum coil resistance, and 

 C/ the resistance of the leads to that coil, including the thick 

 platinum wires which run down the thermometer stem. An 

 equal pair of leads run from C, Cj to similar thick platinum 

 wires in the thermometer stem, which are connected to- 

 gether at the lower extremities, but have no contact with the 

 coil. 



Thus r ± OH = / -h Cp - C ; and therefore if C/ = C, we 

 get ;> = r ± OH. 



The leads Cp and C are everywhere bound together except in 

 the thermometer stem, where they are parallel and adjacent, 

 being held in position by their mica discs, hence changes in C/ 

 and C caused by changes in temperature do not aflfect the 

 resulting value of /, and thus the readings are independent of 



the thermometer stem-temperature — a matter of great import- 

 ance at high temperatures.^ 



A certain amount of stem immersion is, however, necessary, 

 for the lower extremities of the leads must be heated to the bulb 

 temperature, otherwise they would, by conduction, cool the 

 extremities of the coil ; this is an additional reason for forming 

 the leads of platinum, which has a low thermal conductivity. 



A preliminary series of experiments led to the conclusion that 

 a certain quality of white marble had superior insulating proper- 

 ties to ebonite — the material generally used for the tops of 

 resistance-boxes. This superiority was partially due to its 

 non -hygroscopic properties ; for example, I placed slabs of the 

 best ebonite, black marble, and this white marble in an ice-safe 

 for some time. I then removed them one by one to the warm 

 laboratory, and tested them under similar conditions with a 

 " pressure " of 100 volts. The insulating powers of the ebonite 

 and black marble fell off alarmingly, while the white marble 

 was but little affected. 



Some difficulty -.vas experienced by the makers in devising a 

 satisfactory method of attachment between the marble and the 

 many brass connections, <S:c., but this difficulty was at length 

 overcome. The coil and bridge-wire were constructed from one 

 sample of platinum silver. The coil of a platinum thermometer 

 was replaced by a specimen of the wire (diameter -008 in. ) from 

 which the coils were formed, and which had been subjected to 

 the same process of annealing. Its temperature coefficient was 

 then determined with great care over the range 15° to 25° C. 

 (nitrogen scale), and was found to be -000260 in terms of the 



resistance at 20° C. I proposed to keep the box when in use 

 at a temperature near 20° C, for it is wise, when feasible, to 

 maintain all measuring instruments at a temperature exceeding 

 that of an ordinary room, for two reasons: (i) it is gener- 

 ally easy to raise the temperature of the apparatus above that of 

 the air, whereas it is extremely difficult to keep it at a lower 

 temperature ; (2) when the temperature of the apparatus ex- 

 ceeds that of the room, all its surfaces are kept dry, and also 

 more dust-free than would otherwise be the case. 



The greatest difficulty encountered in resistance measure- 

 ments is (according to my experience) uncertainty as to the 

 actual temperature of the coils. If a resistance-box is placed 

 in a tank, it is true that five sides of it can be maintained at a 

 constant temperature, but the top is necessarily exposed ; and 

 since all the coils are ultimately connected with the top, their 

 temperature at times differs considerably from that of the tank, 



1 The absolute equality of C and C/ is not essential ; both are small as 



compared with / ( - is always less than— ), thus C/ - C is a very small 



^p SO' 



fraction of /, and it is only the temperature change in C/ - C that affects 

 the measurements. The total resistance of the thick copper leads from the 

 box to the therriiometer is so small, and they are subject to such comp.ara- 

 tively slight ch.-inges of temperature, that the temperature change of their 

 difference may almost certainly be neglected. The greater part of C/ and 

 C is the resistance of the platinum stem leads, which aic certainly exposed 

 to considerable temperature changes. If, however, they are made of the 

 same platinum as the coil, then any irregularity has nearly the same effect 

 xs an alteration in the original length of the coil, and does not appreciably 



affect the ratio „' or tbe'values of//. In all carefully constructed thw- 



mometers the value of p/ — C may be regarded a> zero. 



NO. 1359, VOL. 53] 



