COMPARISON OF PLATINUM AND OAS THERMOMETERS. 103 



theory what difference should he found between the results in the two case**. 

 CALLENDAR and GRIFFITHS point out, however, that the few observations they made, 

 using their instrument as a constant-volume thermometer, gave a result for the 

 sulphur point about half a degree higher than that found on the constant-pressure 

 scale. If confirmed, this would account for more than half of the difference between 

 the two results. 



It may be of interest to calculate what differences there would l)e between tempe- 

 ratures expressed on CALLENDAR'S air scale and the same temperatures on the 

 nitrogen scale Iwused on the adoption of our value for the boiling-point of sulphur, and 

 on the validity of the 8 formula. The adoption of the new value for the sulphur 

 boiling- point 074 higher than that of CALLENDAR and GRIFFITHS would raise 

 a 8 of T500 to 1-5423. The differences between the temperatures deduced by 

 admitting the validity of the parabolic formula in each case are shown in the follow- 

 ing table : 



T,.,. ... -50 25* 50 75 100 200" 400' 600 1000 



T,*.,-!,., . +0'-03 0' -0'-008 -O'-Oll -0-008 + 0-09 +0-57 + 1-5 +5M. 



From the results of our comparisons we might calculate a formula giving the magni- 

 tude of a small corrective term to l>e applied to the temperatures as deduced by the 

 parabolic formula to reduce them to the scale of our nitrogen thermometer. This 

 correction, however, is not the same for the different platinum thermometers we used, 

 and an examination of the differences given in Column VIII. of the tables shows that 

 in some places the corrective terms for the two thermometers differ by a quantity of 

 about the same order as the corrections themselves. 



Since, as we have previously explained, our own nitrogen scale is somewhat arbitrary, 

 and its relation to the normal scale of the hydrogen thermometer is only known over 

 a small part of the range covered by the experiments, we would suggest that, for the 

 present, temperatures deduced by the platinum thermometer should l>e reduced by a 

 parabolic formula. The results thus obtained can always be recalculated and expressed 

 on any scale which may subsequently be adopted as the standard scale for high 

 temperatures. 



Although we found it impossible to use hydrogen at high temperatures in our gas 

 thermometer with glass reservoir owing to some chemical action taking place between 

 it and the glass, yet it is quite possible that a suitable material may be found for the 

 construction of a thermometer reservoir in which this gas may be employed at high 

 temperatures. 



Until further investigations have been made as to the relations of the various gas 

 scales at high temperatures, and as to the influence of the initial pressure and the 

 effect of impurities and traces of water vapour in the gases employed, and until 

 exact determinations have been made up to high temperatures of the coefficient of 

 expansion of the material used as thermometric reservoir, we think that for the 



