470 Prof. W. M. Hicks on 



by K.R. The Zeeman effects have been observed for Cu 

 D 12 , D 22 by Uunge and Paschen and for AuD n , D 22 by 

 Hartmann, and agree with those for the given allocations. 



The satellite separations mnst depend on displacements in 

 the sequent of a multiple of ouns. The observed separations 

 cannot be in error by as much as dn = 'l. The denominators 

 of the sequents, calculated from the limits D(x>) as deter- 

 mined above and in previous papers in this journal (30642*60 

 for Ag, 29469-85 for An) are : 



Cu. 



d, 1-496751 



1500 



Ag. 



An. 



1-490214 



1-502485 



3053 



9578 



1-487161 



1-492907 



These differences can only be subject to uncertainties 

 not greater than 2 or 3 units, due to error observations, 

 calculation errors of *5 with possibility of 1 on a difference, 

 and a small error in the limits. Now with no uncertainties 

 we find 



Cu. Ag. Au. 



10J8 = 1497-9 1\ 8 = 3052-4 6| 8 = 9195-9 



There is thus perfect agreement within these extremely 

 narrow variation limits for copper and silver, but for gold 

 there is definitely no multiple relation. At the same time 

 the relation is found to hold so universally that there must 

 be some reason for the deviation here. The most natural 

 place to seek it is a change in the other sequent, viz. D(oo ), 

 and we find that a displacement of —28 in this, completely 

 explains the deviation — numerically at least. The oun dis- 

 placement on D(oo) alters the limit by 2*69. Hence — 2S X 

 increases it by 5*38. The d^ sequent calculated from this 

 must therefore be increased by the same amount, and the 

 denominator is then found to be 1*502402, thus altering 

 the mantissa difference to 9495 or the exact 6f 8 = 9495*9. 

 It may be noted that the D u line only appears to take 

 its normal relative intensity in gold. It suggests the 

 possibility that similar displacements are the normal rule. 

 It would explain certain other anomalies between F sepa- 

 rations and d satellites, as for instance the anomalous 

 /satellites referred to above (?n=6, 7 on p. 466). Here the 

 observed separation of 625 is not the true sequent separation 

 which is 620*03, and it is the latter which should be expected 

 as the corresponding F doublet separation. 



