1905.] On the Platinum ana Normal Scales of Tcm^rat..^. 535 



Taking the value for 8, viz., 1-90, obtained by comparing the thermo- 

 meters in liquid air, we obtain for A the value 2'64, introducing 

 error of less than 0-1. 



an 



to Maintain Constant Temperatures below the Freezing Point of 

 /-. AVe have not succeeded in making comparisons of our 

 thermometers at temperatures other than - 78 and the temperature of 

 liquid air. The use of liquefied gases such as nitrous oxide and 

 ethylene in quantities up to 100 c.c. did not give sufficiently constant 

 temperatures, and larger quantities could not be procured except at a 

 prohibitive cost. Attempts were also made to compare the thermo- 

 meters in baths of melting solids, such as ethyl acetate, but without 



success. 



Comparison of the Thermometers at the Transition Point of Sodium 

 Sulphate. Richards and Wells* have shown that the transition point of 

 .sodium sulphate decahydrate to the anhydrous salt may be considered 

 as a standard temperature in thermometry. They measured this 

 temperature on three Baudin thermometers and on one Tonnelot 

 -thermometer, and found, as the mean of results which differed among 

 themselves by O'Ol , the value 32*383 on the normal scale. 



For the purpose of measuring the transition temperature on the 

 platinum scale, we prepared some pure sodium sulphate, dried the 

 crystals, and ground them to powder. The powder was heated in a 

 beaker in a water-bath at about 40 C. till the mass became pasty, 

 and was then transferred to the vessel surrounding the thermometer, 

 which was enclosed in a tube containing paraffin oil (p. 532). The 

 vessel containing the sulphate was a glass cylinder, 8 cm. wide and 

 20 cm. deep, with a round bottom. The cylinder was supported in a large 

 thermostat, and it was found that when the temperature of the latter 

 was kept at 32 '6 C. the temperature of the sodium sulphate remained 

 quite steady, and was not affected by stirring. 



The value of 3 corresponding to the mean value of A is 1'98, and it 

 will be observed that this is identical with that obtained at - 190, and 

 very close to the value at - 78 C., viz., 1'84. Indeed, it appears that 

 .a parabolic formula may be applied to the calculation of A between 30' 

 .and - 190 C. with a maximum error, probably not greater than 

 0-2 C. at about - 100 C. This result is mainly dependent on the 



* ' Zeit. Phys. Chemie,' 1898, vol. 26, p. 690 ; 1903, voh 43, p. 465. 



2 R 2 



