210 Prof. H. L. Callendar on Platinum Thermometry. 



however, that they are too high, and that the results given 

 "by the pure wire are the more probable. The latter are 

 approximately a mean between the values of Violle 954°, and 

 Holborn and Wien 971°, and may be taken, in the pre- 

 sent state of the science of high-temperature measurement, 

 to be at least as probable as any other values, in spite of 

 the extrapolation from 445°, by which they are obtained. 



The extrapolation is not really so unreasonable as many 

 observers seem to think. The parabolic formula for resistance 

 variation has been verified for a great variety of cases, through 

 a very wide range, and with much greater accuracy than in 

 the case of many so-called laws of nature. For instance, a 

 similar formula, proposed by Tait and Avernarius, is often 

 regarded as the law of the thermocouple, but the deviations 

 of thermocouples from this law are far wider than those of the 

 most impure platinum thermometer. If we take a Pt-PtRh 

 thermocouple, and apply the S.B.P. method of reduction in 

 the same manner as in the case of a platinum thermometer, 

 taking the data, t= 100°, e=650 microvolts; £ = 445°, e = 

 2630 mv. ; we should find d= -7*4. At *=1000°C., e=9550 

 mv., the temperature on the scale of the thermocouple is 

 ^ = 1470°. The temperature calculated by the parabolic 

 formula is £ = 804°. Whence it will be seen that the devia- 

 tion from the formula is about ten times as great as in the 

 case of a very impure platinum wire. A cubic formula was 

 employed by Holborn and Wien to represent their observations 

 at hio-h temperatures with this thermocouple, but even this 

 formula differs by more than 20° from their observations at 

 150° C. It is, moreover, so unsatisfactory for extrapolation 

 that they preferred to adopt a rectilinear formula for deducing 

 temperatures above 1200° C. 



There are, however, more serious objections to the adoption 

 of the thermocouple, except to a limited extent, as a secondary 

 standard: — '(1) The scale of the thermocouple is seriously 

 affected, as shown by the observations of Holborn and Wien 

 and Barus at high temperatures, and of Fleming at low 

 temperatures, by variations in the quality of the platinum 

 wire and in the composition of the alloy. (2) The sen- 

 sitiveness of the Pt — PtEh thermocouple at moderate 

 temperatures is too small to permit of the attainment of 

 the order of accuracy generally required in standard work. 

 (3) No satisfactory method has yet been devised in the case 

 of the thermocouple for eliminating residual thermal effects 

 in other parts of the circuit, which materially limit* the 



* My present assistants, Prof. A. W. Porter, B.Sc, and Mr. N. Eumor- 

 fopoulos, B.Sc, whose work on Emissivity and Thermal Conductivity has 

 already in part been published in this Journal, employed this thermo- 



