316 



Tike Value of the Silver Oun. 



28996-43 ±'3 is N/(1 + 34A)' 2 . The denominator is found to 

 be 1-944830+10, 



34A = 944830 ±10, 



A = 27789-12±-29. 



The agreement of the two independently arrived at values 

 is very striking. We shall be safe in regarding the most 

 probable value of A as 27789'10 with a possible error +*3. 



AsA = 666, 



8 =421-0470 ±-0045. 



Taking then the atomic weight of Ag to be 107*88 exact,, 

 the ratio g=S/(W/100) 2 is 



? = 361-7837±-0038. 



Thus q is given within 1 in 100,000. The old value ob- 

 tained in [III.] was 361'80±-1. 



The value of S, or of <?, is the same whatever scale of 

 wave-length measurement is used. It does, however, depend 

 on the correct value of N. The value 109675 in R.A. is 

 determined from the spectrum of H. The value determined 

 from He indicates a number from 25 to 45 larger. Bohr's 

 value, based on theoretical considerations only, gives a value 

 practically constant for all elements heavier than He, and 

 about 50 larger. Suppose then N is N + dN = N(l + f) 2 

 where £ 2 is negligible. In any particular case the value of 

 A has been calculated by throwing the limits of the doublet 

 into the form N/d 2 , N/i' 2 when d — d' = A. Now N/d 2 ... are 

 independent of N. Hence if N becomes N(l + f) 2 , d, d' 

 become d(l + 0, d'(l + ?) and A + dA = {d-d') (1 + J) = 



A(l + ». 



Hence 



dq 



d8 



8 



dA 



A 



Z=\ 



N 



•45xlO- 5 8N. 



Consequently 



^/ = -3618 x -45 X 10- 2 SN = -1628 x 10- 2 6N". 



The fact that q as determined directly is 361-8 + *1 within 

 error limits for all elements in which it can be determined 

 would tend to show that any variation in N must be < 100, 

 although the error limits are not sufficiently narrow in all 

 cases absolutely to prove this statement. For N= 109720, 

 dq = -07d2, and for Ag, 8 = 421\L322±-0045. 



