92 BRIDGMAN. 



conditions, and obviate irregularity due to poor thermal conduction 

 of any transmitting medium which might get into the cracks between 

 specimen and capsule, the capsule was cemented into the specimen with 

 deKhotinski cement, which has a considerably higher thermal conduc- 

 tivity than the liquid. 



It will be noticed that the method of connection described above 

 amounts essentially to grounding one end of the heating element on 

 the sample itself, there being effectively a junction between the metal 

 of the specimen and copper. When a heating current is passed there 

 will be a generation of Peltier heat at the junction in addition to the 

 Joulean heat in the element itself. This effect reverses in sign when 

 the direction of the heating current reverses and so may be eliminated. 

 However in the experiment as actually performed the independence of 

 the reading of the thermo-couple of the direction of the heating cur- 

 rent was used as a criterion for absence of electrical leakage between 

 the heating and the thermal current circuit, so that elimination of the 

 Peltier heat by reversal of the heating current was not desirable. For 

 most of the metals the effect is very small and may be neglected, but 

 for antimony and bismuth the effect becomes inconveniently large, 

 rising to a change in the apparent magnitude of the thermal conduc- 

 tivity of 25% with reversal of heating current. For these two metals 

 the ground to the capsule was discarded, and two independent leads 

 to the heating element were used. 



The thermo-couple was constructed in the same way as those for 

 the cylindrical specimen ; a length of copper, then a piece of constantan 

 1 inch long, and then copper again. This was bent into the form of a 

 hairpin, whose two legs were thrust through two diametral holes 2 mm. 

 apart in the specimen. The junctions were located on the axis of the 

 specimen. After wrapping the wires around the specimen, as will be 

 described in a later paragraph, the two copper ends of the hairpin 

 were then soldered to the two copper thermo-couple terminals of the 

 three terminal plug. The thermo-e.m.f. of the couple obviously gives 

 the difference of temperature of the two junctions points. 



In order to avoid error from heat leak along the wires of the couple 

 it was necessary to make these of much smaller wire than in the case 

 of the cylindrical specimens. The wire used was 0.004 inch diameter, 

 and the hole in which the couple was placed was 0.007 inch. The 

 couples were brought to a diameter of 0.005 inch by the coating of 

 enamel. There is an outstanding possible play of the couples in the 

 holes of 0.002 inch, which might lead to maximum errors in the ap- 

 parent thermal conductivity of 5%. This is probably the chief source 



