480 A. P. Matkews 



cules in a gram mol, gave the result (w 2 &) 2/3 = 3.20 X io~ 37 

 which is very close to the figures already obtained from 

 methane and hydrogen, and somewhat less, as it ought to be, 

 than the mean value of 3.45 X io~ 37 which was obtained when 

 residual valence was disregarded. In view of these facts I 

 believe we may assume that 3.201 X io~ 37 is the real value 

 of C, although the theoretical basis of this relationship is still 

 lacking, and proceed with the calculation of the total valence 

 of a molecule on that assumption. 



The value of M 2 K is less satisfactory. It is here that the 

 main uncertainty of the calculation lies. As I have already 

 discussed the methods of calculating this value in my paper 

 on the internal pressures of liquids I will not go into the 

 question at this time further than to point out two or three 

 considerations bearing on the probable accuracy of those 

 figures. In all the formulas for "a" which have so far been 

 proposed certain assumptions have been made. The one 

 ordinarily made in van der Waals' method of computing "a" 

 is that b c = V c /3- In the various methods I have proposed 

 for the computation quite different assumptions have been 

 made in the different formulas, but nevertheless these formulas 

 have all given results which are not widely different if the 

 uncertainty of some of the experimental data are considered. 

 Nevertheless, the formulas do not always give exactly the same 

 values as they should if all the assumptions and data were 

 rigorously correct. The computation of the cohesion from the 

 latent heat of vaporization should give a correct result since 

 the assumptions made here are less radical than in any of the 

 other methods. Now this method generally gives a value for 

 "a" lower, in some cases 5 percent lower, than that com- 

 puted from the critical data. But I have not been able to 

 attach more importance to this deviation for the reason that 

 the computation must be made close to the critical tempera- 

 ture, within a fraction of a degree of it, and a very slight error 

 in the difference of the vapor and liquid densities would make 

 a very large error in "a." That x the formulas proposed for 

 "a" are possibly not entirely accurate may be shown also 



