1899.] on Measuring Extreme Temperatures. 107 



gas-thermometer at high temperatures in respect of all the various 

 sources of error above mentioned. (1) It was observed that the 

 volume of the bulb underwent continuous changes, chiefly in the 

 direction of contraction, and that the shrinkage was not symmetrical, 

 being apparently greater in the circumference than in the length of 

 the cylinder. (2) To prevent leakage, and to close the pores of the 

 material, it is necessary to have the porcelain bulb glazed both inside 

 and out. The glaze becomes sticky, and begins to run at a tempera- 

 ture below 1200° C, and the bulb begins to yield slightly and con- 

 tinuously to pressure above this point. (3) With some gases there 

 appear to be slight traces of chemical action or occlusion of the gas 

 by the walls of the bulb at high temperatures. It is for this reason 

 preferable to use the inert gases nitrogen or argon as the thermometric 

 material. In any case, the limit of high temperature measurement 

 would be reached when either the gas, or the material of the bulb, 

 began to dissociate or decompose. Deville and Troost, employing 

 C0 2 for filling the porcelain bulb, found the temperature of the B.P. 

 of zinc nearly 150° higher than with air or hydrogen. This they 

 attributed to a partial dissociation of the C0 2 at the temperature as 

 low as 930° C. Some experiments made by the writer appeared, how- 

 ever, to indicate that the effect was due to chemical action between 

 the gas and the porcelain. 



For these and other reasons it appears very doubtful whether any 

 improvement or extension of range can be expected from the use of 

 glazed porcelain. If an attempt is made to employ any of the more 

 refractory kinds of fire-clay, there is the difficulty of finding a suit- 

 able glaze, and of eliminating leakage and porosity. The writer 

 suggested the use of bulbs of fused silica some years ago (Proc. Iron 

 and Steel Institute, 1892), and endeavoured to get such bulbs con- 

 structed, but without success. This material possesses many of the 

 requisite qualities, but is for this very reason extremely difficult to 

 work. Metallic bulbs of platinum or platinum-iridium are by far 

 the most perfect in respect of constancy of volume, regularity of 

 expansion, and facility of accurate construction ; but unfortunately, 

 as Deville and Troost showed, they have such an inveterate tendency 

 for occluding or dissolving gases at high temperatures, that the use 

 of metallic bulbs has been practically discontinued, in spite of their 

 obvious advantages in other respects. 



After making many vain experiments, the writer was forced to 

 the conclusion that the ordinary bulb-methods did not promise any 

 satisfactory solution of the problem of extending the range of the 

 gas-thermometer, and that it was necessary to attempt a radically new 

 departure. The optical method, depending on the measurement of 

 the refractivity of a gas at high temperatures, and the acoustical 

 method, depending on the observation of the wave-length of sound, 

 although of great theoretical interest, did not appear to promise 

 sufficient delicacy of measurement or facility of practical application. 

 Experiments were therefore made on the methods of effusion and 



