EFFECT OF PRESSURE ON CONDUCTIVITY OF METALS. 95 



was taken as the integral of the losses of short elepientary cylinders. 

 The loss from the upper end was calculated from the formula for the 

 flow between two concentric hemispheres, one having the radius of the 

 specimen, and the other the internal radius of the pressure cylinder. 

 The value obtained in this way is obviously too large, as most of the 

 pressure cylinder is situated further from the end of the specimen than 

 its internal radius. The lateral flow and the flow from the ends as 

 computed in this way involve the difference of temperature between 

 the source end of the specimen and the outer cylinder. This tempera- 

 ture difference may be expressed approximately in terms of the heat 

 input into the rod and the thermal conductivity of the rod. The 

 temperature difference may ndw be eliminated, giving a result in 

 terms of heat input and the conductivities of the liquid, /l2, and the 

 metal of the bar, ki. For the lateral loss of heat I obtained the formula 



11.7 ~Q, and for the loss from the end 12. — Q (this last is too high). 

 ki ki 



As an approximate value for the total loss through the transmitting 

 medium the value 20 — Q was used in the computations. 



The value of /1-2, the thermal conductivity of petroleum ether, was 

 determined by direct experiment, as will be described later. At 

 atmospheric pressure the value found was 0.0004. At a pressure of 

 12000 kg/cm^ the conductivity has increased 2.2 fold. The change in 

 the value of the lateral flow correction with pressure determines the 

 correction that must be applied to the apparent change of conductivity 

 of the metal under pressure in order to obtain the true change. The 

 magnitude of this correction evidently changes greatly with the metal 

 used. The details with regard to this correction will be discussed 

 under the different metals separately. 



In the longitudinal flow method there is no correction to be applied 

 for the change in dimensions of the heating element, as there was in 

 the radial flow method. The same arrangement of a shunt circuit 

 which eliminated the necessity for a correction for the changing re- 

 sistance of the heating element with the radial flow method was used 

 here also. There is a correction to be applied here, not necessary in 

 the radial flow method, for the change of dimensions of the specimen 

 under pressure. Under pressure the cross section becomes less and 

 the distance between the thermal junctions also becomes less, the first 

 b\' twice the linear compressibility and the second by the linear com- 

 pressibility. The sum of the two effects is a correction equal to the 



