EFFECT OF PRESSURE ON CONDUCTIVITY OF METALS. 103 



been flaws in the casting, as was indeed found to be the case one or 

 two other times when the longitudinal specimen was cast. For lead, 

 the correction to be applied to readings on the longitudinal specimens 

 for the lateral flow through the transmitting medium was unusually 

 large, reducing the observed efl^ect from 30% to 20%. 



The readings on the two best samples were sufficiently good to show 

 a departure from linearity with pressure, the slope becoming less at 

 the higher pressures. This is as would be expected. Part of the 

 departure from linearity is without doubt to be ascribed to the effect 

 of the transmitting medium, and since for most of the other metals 

 the accuracy was not great enough to show this departure from 

 linearity, it did not seem worth while to attempt to establish it more 

 exactly for lead. 



It will be seen that the two best runs agree pretty closely with 

 each other and with the mean of all the measurements, so that con- 

 siderable confidence may be given to the average value, 20.7%, as 

 the most probable effect of 12000 kg. It is to be remarked that this 

 change of thermal conductivity is markedly greater than that of 

 electrical conductivity under the same pressure, which is only 14.6%. 



The average pressure coefficient of thermal conductivity per kg. 

 change of pressure is 0.0000173. Lussana ^ found as the average 

 coefficient between atmospheric pressure and 2600 kg., 0.0000134. 

 His initial coeflficient was 0.0000164, and the final O.O4II6, showing 

 departure from linearity in the same direction as I found, but much 

 greater in amount. His results do not show, however, any departure 

 from linearity in this range in the relation between pressure and 

 electrical conductivity. 



Tin. A few preliminary measurements were made on tin with 

 "chemically pure" metal from Eimer and Amend, but all the results 

 of any value were obtained with Bureau of Standards tin, prepared 

 for the calibration of thermo-couples by means of the melting point. 

 The analysis of the tin was as follows: Pb .007%, Cu .003%, Fe .002%; 

 As, S trace; Sb, none; Tin (by difference) 99.988%, and the melting 

 point given by the Bureau's certificate was 231°. 9. 



Twelve different samples of tin were used, ten by the radial and two 

 by the longitudinal flow method. None of those specimens in which 

 a copper sheath was used gave results of any value whatever. The 

 reason for this has already been discussed; the effect is due to lack 

 of good thermal contact between the tin and the sheath. Neither was 

 it found possible to get any results of value with those specimens in 

 which the heating element was enclosed in a central silver tube. For 



