78 BRIDGMAN. 



on the thermal conductivity of eight metals and four alloys. Except 

 for two of the alloys, he finds that the thermal conductivity always 

 increases under pressure, and the change is of such a magnitude that 

 the Wiedemann-Franz ratio remains nearly constant. The classical 

 conception of the mechanism of conduction would prepare one to 

 expect this result. I have not been able to verify the results of Lus- 

 sana, but find that the conductivity of more than half my metals 

 decreases instead of increases under pressure. This is perhaps a result 

 not to be expected, and must have considerable significance for our 

 picture of the mechanism. With regard to the differences between 

 Lussana's results and my own, I think that Lussana's work is open to 

 serious criticism in several particulars, and I have gone into this in 

 detail later. 



Measurements of thermal conductivity are known to be among the 

 most difficult in physics (witness the disagreement among different 

 observers as to the sign of the temperature coefficient of thermal 

 conductivity), and I was accordingly prepared for considerable diffi- 

 culty in measuring changes, which might be expected to be of the order 

 of a few per cent, in a quantity which is itself so hard to measure. The 

 difficulties were of course enhanced by the technical necessity of mak- 

 ing the apparatus so small that it could be enclosed in a pressure 

 chamber. The anticipated difficulties were encountered; these 

 measurements are by far the most difficult of any that I have at- 

 tempted under pressure. The accuracy is accordingly much less than 

 that possible in such measurements as of electrical resistance under 

 pressure, for example. Certainly not more than two significant 

 figures can be claimed for the pressure coefficient of thermal conduc- 

 tivity. If the work were to be repeated now it would be possible to 

 improve considerably on the accuracy, or at least the presentability, 

 of much of it. But I believe that the results which follow are essen- 

 tially correct in their large features, and that the use which can at 

 present be made of such data does not justify the expenditure of more 

 time. 



Description of Apparatus and Method. 



The choice of methods was much restricted by the limitations of 

 space imposed b}' the necessity of getting the specimen and all the* 

 attachments into a pressure cylinder, and by the necessity of using 

 only four electrical leads to the heating element and thermo-couples. 

 Two methods were used ; these were both of the same general character 



