1904.] at the Temperature of Boiling Hydrogen. 245 



value below which no further lowering of the temperature seems 

 to reduce it • and (2) that the parabolic connection between tem- 

 perature and resistance is no longer tenable at very low temperatures. 



Of the different thermometers constructed on the electric resistance 

 principle, fifteen were serviceable throughout the investigations ; the 

 others broke or failed from various causes. The metals employed were 

 platinum, gold, silver, copper, palladium, iron, nickel, and two alloys, 

 platinum-rhodium, and German silver. The metals were supplied by 

 Messrs. Johnson and Matthey, and the late Sir W. C. Roberts- Austen 

 of the Mint. Every endeavour was made to attain the highest purity 

 in the samples. In the Bakerian Lecture a table* was given containing 

 the constants of seven of these thermometers, and in the annexed Table I 

 similar results are given for the remaining eight. 



The observed resistances, after all corrections were made, were 

 reduced both by Callendar's and by Dickson's methods, and the close 

 accord of their results is apparent. When these temperatures are 

 compared with the results given by the hydrogen thermometer, they 

 confirm the general inferences above referred to. In the Callendar 

 method the a was determined from the resistances at 100° C. and 

 0° C, and the 8 was then obtained from the resistance and corre- 

 sponding observed temperature of liquid oxygen boiling under atmo- 

 spheric pressure. The same data were used in the Dickson formula, 

 except in a few cases where it was thought worth while to employ 

 the method of least squares, in which cases these data were included. 



In Table I the suffix number attached to the name of each metal 

 indicates a particular thermometer in my notes of the observations, and 

 may be taken as a rough index of the chronological order in which the 

 observations were made. 



Platinum, gold, silver, and copper show a remarkable agreement 

 between the two methods of reduction. The later specimens of the 

 metals of each of these groups are purer than the earlier ones. In 

 the platinum and gold groups we notice that the Centigrade tempera- 

 ture H, at which the resistance would vanish, rises with the purity. 

 This is still seen in copper, but something of the reverse appears 

 in the case of silver (perhaps this may be explained later). However, 

 the general rule is again apparent in palladium. With regard to 

 Callendar's coefficients, the temperature coefficient a increases with 

 the purity of the metal, while the difference coefficient 8 diminishes ; 

 nevertheless some uncertainty is apparent in the platinum thermo- 

 meter. Alloys appear to have much smaller values of a and much 

 larger values of 8. On the other hand the magnetic metals have 

 greater a's and very much increased S's. 



A marked peculiarity in the magnetic metals is that the difference 

 coefficient 8 is negative, that is, the temperatures in metal-degrees 



* Loc. tit. . - . 



