80 " BRIDGMAN. 



pressure measurements, and found only later that it had been previ- 

 ously proposed by Niven ^ in 1905, particularly for measuring the 

 conductivity of comparatively poor conductors. 



The method was used with many modifications, as will be described 

 later. Its essential weakness is that it demands that the specimen be 

 perfectly homogeneous ; this is difficult to obtain in sufficient perfection 

 in a casting, or even in commercial drawn metal. The method worked 

 best for lead, which is fairly easy to cast, is a comparatively poor 

 conductor, and is not crystalline, so that inhomogeneities due to 

 crystallization during cooling were not important. The method 

 failed for copper and nickel. 



The second method was a longitudinal flow^ method. Heat is put 

 into the end of a rod, the other end of which is connected with a mas- 

 sive copper block in intimate contact with the pressure cylinder, so 

 that we have essentially a rod with a source at one end and a sink at 

 the other. The difference of temperature between two points on the 

 axis of the rod is measured with a thermo-couple. In contradistinc- 

 tion to the other method there is here an important correction due to 

 the lateral loss of heat to the fluid by which pressure is transmitted. 

 An elementary discussion shows that in order to reduce this correction 

 to the smallest value it is necessary that the dimensions of the specimen 

 be very small. The specimens actually used were only \ inch in 

 diameter, a little over 1 cm. long, and the two thermal junctions were 

 2 mm. apart. The correction for lateral loss is evidently smallest for 

 those metals with the greatest conductivity, so that this method 

 supplements the other, working best for those metals for which the 

 first method gave the poorest results. The magnitude of the correc- 

 tion for lateral loss may vary from 20% of the total effect for bismuth, 

 to about 1% for copper and silver. The most disagreeable feature of 

 this correction is that it changes with pressure because of the change 

 of conductivit;\' of the transmitting medium under pressure. This 

 effect is large and has to be independently determined. The thermal 

 conductivity of petroleum ether, which was the medium used in all 

 the later part of this work, increases 2.2 fold under 12000 kg/cm^. 

 The determination of the pressure effect on petroleum ether will be 

 described in detail later. 



Discussion of further details of these two metliods is reserved until 

 later. 



The electrical measurements involved in either method were es- 

 sentially the same in character, and could be made with the same 

 apparatus. The measurements necessary were of heat input, which 



