82 BRIDGMAN. 



A correction also had to be applied for the change in thermal e.m.f. 

 of the couple under pressure. The couple was of copper-constantan, 

 and the corrections due to pressure may be taken from data already 

 published.* The total correction at 12000 kg. due to this effect is 

 0.8%; the correction is to be subtracted from the apparent thermal 

 conductivity. 



It will be noticed that this method differs from those previously 

 used in determining the effect of pressure on electrical conductivity 

 or thermal e.m.f. in that it is not a differential method, but is direct. 

 A number proportional to the total thermal conductivity is determined 

 at different pressures, and from these data the effect of pressure is 

 computed. As originally planned I had intended to make the method 

 differential, measuring the difference of thermal conductivity between 

 one sample exposed to the pressure and a similar one not exposed. 

 This can be simply done by opposing the two thermo-couples of the 

 two specimens. It soon appeared, however, that the regularit}- of the 

 results was not sufficient to justify this refinement, and the simpler 

 direct method was used. The direct method demands somewhat 

 greater accuracy in the resistances and the comparison standard cell, 

 which, however, was easy to attain. 



Since the primary interest of these measurements was in the pro- 

 portional changes of thermal conductivity produced by pressure, a 

 highly accurate knowledge of some of the absolute data was not 

 essential. For instance, it is much easier to compare within 0.1% the 

 two potential taps for the heating unit and its shunt than to measure 

 the resistance of either accurately to 10~^ ohms, which would have 

 been demanded by 0.1% on the absolute resistance. 



We return now to a discussion of the details of the two different 

 methods. First the radial flow method will be described. 



The difficulties were chiefly mechanical, involved in getting the 

 specimen into proper shape, and ensuring the right boundary condi- 

 tions. The method demands not only that the heat input take place 

 accurately on the axis, but most of all that the exterior surface of the 

 specimen be maintained at constant temperature. The difficulty of 

 this last requirement can be seen when it is considered that the thermal 

 conductivity of copper, for example, is 3000 times greater than that 

 of the petroleum ether by means of which pressure is transmitted to 

 the specimen. There must of necessity be some crack between the 

 walls of the pressure cylinder and the specimen; even if this crack is 

 made vanishingly small at atmospheric pressure, it assumes quite 

 appreciable proportions under 12000 kg. owing to the elastic deforma- 



