88 BRIDGMAN. 



rounding is due to impurity ; in the early days of this kind of measure- 

 ment there was room for honest question whether the absolutely pure 

 substance would show premature rounding or not. 



While it cannot be claimed that the measurements above meet all 

 these requirements, it is evident that any error from cavities must be 

 negligible when freezing takes place under thousands of atmospheres, 

 and as far as purity goes, the potassium used above never showed any 

 preliminary rounding, thus bearing out the observations on the same 

 material on the melting curve. The melting of Bernini's sample was 

 not abrupt, and took place at 62.04°, nearly 0.5° below that of mine. 

 Northrup records the value 63.5° for the melting point of his specimen; 

 this is so high that it seems that it must be due to errors in tempera- 

 ture measurement. 



The value which I found for the temperature coefficient of resistance 

 at atmospheric pressure is considerably lower than that of either 

 Bernini or Northrup. The first found 0.00601 between 0° and 50°, 

 and the latter found 0.0058, in terms of the resistance at 0°C. between 

 20° and 50°. I found the relation to be linear, and the value of the 

 coefficient, corrected for the expansion of the glass, 0.00512 between 0° 

 and 50°. My value for the liquid is, on the other hand, larger than 

 that of Bernini. I find for the mean coefficient between 95° and 130° 

 0.00403 of the resistance at 95°, and Bernini gives between 90° and 

 100° 0.00358 of the value at 95°. Northrup's value, reduced to 

 fractional parts of the resistance at 95°, is 0.00342. 



In using the Table, caution should be employed not to force it 

 beyond its accuracy. In particular, too much importance should 

 not be attached to the variation of the differences with temperature. 

 If at any time in theoretical work it should be important to know 

 exactly the variations of pressure and temperature coefficients with 

 pressure and temperature for small ranges of pressure, this work 

 should be repeated with apparatus capable of measuring pressures 

 with greater sensitiveness, and the pressures should not be pushed so 

 high as to introduce irregularities in the glass. In this work the 

 smallest pressure steps were 1000 kg., and no readings were made 

 between and 1000 kg. Where the changes are so large in an interval 

 of 1000 kg. it is entirely possible that some essential detail of behavior 

 may have been overlooked, or smoothed out in constructing the Table. 



Magnesium. In the previous paper on the resistance of metals 

 under pressure it was possible to give only very rough values for the 

 pressure coefficient of magnesium. Difficulty was previously found 

 in making good connections because of the impossibility of soldering 



