Natufe of Dielectric Capacity. 11 



that any minor error in 43' 6 becomes relatively large when 

 all accumulated on the residual 5*6. For this reason then 

 the values of B and of B 1 / 3 in Table III. are to be taken as 

 only rough approximations. In view of the proportionality 

 between es and B 1/3 in the Li family, both giving the series 

 2, 3, 4, 5, 6, and the sharpness with which F r /2 or es in the 

 Be family gives the series 4, 5, 6, 7, 8, I think it is the best 

 course to take B in this family as given by the formula 

 B 1 /3 = 0-32 (4, 5, 6, 7, 8). The values of B thus obtained are 

 given in Table IV. as B calc, and are used for the study of 

 (6) with = 2-31, D = 0-456 in the Be family of metals in 

 their compounds as set forth in Table IV. 







Table IV. 









Be. 



Mg. 



Ca. 



Sr. 



Ba. 



B calc 



.... 2-097 



4-096 



7-077 



11-24 



1677 



(*-l)B 



.... 6-6 



69 



101 



133 



16-1 



p 



... 4-29 



5-86 

 2-685 



565 



2-427 



7'80 

 2-184 



8-19 



N 



... 4-147 



1-960 



N calc 



... 415 



2-77 



2-40 



2-11 



1-97 



K 



... 17-2 



7-21 



5-89 



4-77 



3-84 



From this Table and II. it appears that (6) gives a satisfac- 

 tory account of N, and therefore of K, for these two families 

 of metals in their compounds. In the other families of metals 

 the data are imperfect, but when obtained they will show 

 how the transition to the case of non-metals occurs. A com- 

 parison of C in these two families of metals is important, 

 and also a comparison of D. The ratio of the two values of 

 C is 2-31/0-835 = 2-77. Now on the simplest theoretical 

 grounds we should expect this ratio to be 2, for, if we turn 

 back to the reasoning by which (5) was established, we see 

 that two pairs of special electrons in the atom will cause G 

 to have double the value due to one. Moreover, in the Be 

 family is to C in the Li family as 0'53/0*32 is to 1-20/0-65, 

 that is as 1 to 1*114. On these two accounts then C in the 

 Be family ought to be 2 x 1*114= 2*23 times C in the Li 

 family. This theoretical 2*23 falls short of the actual 2*77 

 just found from the data. Perhaps the two special electron 

 pairs in the Be family exercise a mutual influence which 

 should be included in the theory. As regards D we see that 

 D/C = i/2a\ and in the Be family has the value 0-456/2*31, 

 in the Li family 0*735/0*835, which bear to one another the 

 ratio 1 to 4*46. But if i the inertia of an electron is the 

 same in both families, this result implies that cr 3 in the Be 

 family is 4-46 times cr 3 in the Li family. It is only acci- 



