82 BRIDGMAN. 



For the ratio of the resistance of Hquid to that of sohd at the melting 

 point Bernini found 1.337, and Northrup 1.44. The value which I 

 found above was 1.451, in good agreement with that of Northrup. 

 The agreement of my high pressure values with that found at atmo- 

 spheric pressure makes it probable that in my work the formation of 

 cavities did not cause any essential error at atmospheric pressure. 



It will be noticed that throughout, my values are considerably 

 closer to those of Northrup than Bernini. It is probable that in spite 

 of his efforts, Bernini did not succeed in eliminating the effect of cavi- 

 ties, or that there was some other source of consistent error. 



Potassium. Material from the same lot was used as that used 

 previously for the measurements of the coordinates of the melting 

 curve.* The purity was high, as shown by the sharpness of freezing. 

 Resistance measurements under pressure were made on both the liquid 

 and the solid. The mechanical softness of potassium, and its much 

 greater chemical activity made it infeasible to use the solid in the form 

 of bare wire, as had been possible with sodium and lithium. The 

 metal had to be contained, therefore, in a glass capillary. This is 

 to be regretted, but its much greater softness makes any error intro- 

 duced by the restraining action of the glass much less than in the 

 case of the other metals. Special examination was made of the mag- 

 nitude of the error introduced in this way by measuring the differ- 

 ence betw'een the resistance under increasing and decreasing pressure 

 at 28°. Very noticeable differences between the ascending and 

 descending curves were found, corresponding to a maximum difference 

 of mean pressure in the metal and the surrounding liquid of 100 kg. 

 The maximum discrepancy occurred at the highest pressure, where 

 it would be expected that the viscosity of the metal would be the 

 greatest. The pressure difference of 100 kg. was estimated from the 

 differences of resistance, assuming that the stresses in the metal had 

 the same effect on resistance as a hydrostatic pressure. Of course 

 there actually were stresses in the metal of a shearing nature, and it 

 is exceedingly unlikely that the effect of such stresses is equivalent to 

 a h^'drostatic pressure, so that it is probable that the glass capillary 

 was called on to support a stress difference of considerably more than 

 100 kg. The mean of readings with ascending and descending pres- 

 sure were taken as the correct value. There were, however, consider- 

 able irregularities, doubtless due to the irregTilar response of the glass 

 capillary to the heterogeneous strains in it. At higher temperatures 

 the differences between the readings with ascending and descending 

 pressure became much less than at 28°, as one would expect, but it was 



