THE MEASUREMENT OF TEMPERATURE. 77 



changes in the external pressure l and the Paris, and some 

 platinum standards. A summary of the results will be found 

 in paper (13), and they but more firmly establish the con- 

 clusions previously arrived at. 



There are indications that the formula (d) gives the rela- 

 tion between the air and platinum scales at very low tempera- 

 tures. In a communication to the Phil. Mag. (10) it was 

 pointed out that we can find at what platinum temperature the 

 resistance of platinum will vanish by placing o for R in the 



formula , D x 



( K - K ) 



'*- Rr-R7 = IO ° 



and then by using the value of S peculiar to the wire and 

 assuming the parabolic formula to hold over so large a 

 range, we can deduce the corresponding value of t. The 

 method was applied to all those thermometers whose con- 

 stants had previously been published and the mean result 

 gave - 2 y 2> - 9°C. The experiments of Professors Dewar and 

 Fleming (12) led them to the conclusion that the electrical 

 resistance of platinum wire would vanish at absolute zero, 

 and thus we have what seems to me very strong evidence 

 of the accuracy of the (a 7 ) formula at low temperature. 2 



As regards temperatures exceeding 7oo°C. there is no 

 reason to suppose that the departure of the t - pt curve 

 from the parabola becomes important. It is difficult to 

 carry the comparison with the air thermometer above the 

 temperature already obtained. What we require for high 

 temperatures, however, is not so much the measurement on 

 the air scale, but some practical standard to which they may 

 be referred, and even if the relation between / and pt at 



1 Callendar, Roy. Soc. Proc, January, 1891. 



2 In cases where a high order of accuracy is not a necessity, a yet 

 simpler mode of graduation is thus suggested. Having found the value of 

 R x and R if we assume R = o when t = - 2737°C. we can obtain an ap- 

 proximate value of 8 and thus dispense with the observations in sulphur 

 vapour. The results obtainable by this simple method are but approxi- 

 mate. In seven thermometers whose constants were given in paper (10) 

 the greatest error caused by the adoption of this method amounted at 150 

 to "44 C., but if the thermometers are intended for use between o° and 

 ioo°C., the probable error would not exceed *o5°C. 



