12 Prof. W. M. Hicks on the 



with those for m = 1,2. All the set appear except the weak 

 line D 12 , and it is at the same time instructive to notice that 

 there appear no signs of displaced lines. For ??z = 4, on the 

 contrary, there are possible representatives only of D 22 and 

 D 22 , but there are clear evidences of extensive displacements. 

 The calculated value for D n is 25034*05. This has a u link 

 to (4)3637-4*87, for which the Zeeman pattern has been 

 observed by Purvis. It is represented by 0/6-J, the proper 

 pattern for a D n line. Displacement certainly alters the 

 Zeeman pattern — witness, as an instance, the different forms 

 for satellite lines. If this indication that linkage leaves the 

 Z-patterns unaffected is sustained by further evidence, it 

 would be of fundamental importance for spectral theory. 

 The normal line 20534 itself appears split up into a number 

 of others all showing displacements by an oun multiple in 

 the limit. We find in succession the lines in the following- 

 list, in which are given the suggested displacements, and the 

 O-C values on the supposition that the true D u (4) should 

 be 25034*24 :— 







O-C 



(1) 25012*07 



(2*,)Dii 



•11 



25034*05] 



[Dn] 



•03 



(1) 25045*39 



(-Si)Dn 



-*06 



(1) 25055*69 



(-280 D n 



•01 



(In) 25077*80 



(-4801)11 



-•08 



The two lines suggested in the Table for D 22 , D 22 are 

 probably not the normal, but may be displaced. They would 

 give a satellite separation of 43*40, much too large to be in 

 step with those for m=l,2, 3. Here 28 o\ would produce a 

 separation of 17*5. The general disintegration of the D(4) 

 set would seem to be indicated by the existence of fragments 

 of other sets in the neighbourhood. For instance, we find 



(3w) 25170-53 29472-88 (lr) 33775-23 



381370 381452 



(1) 28984-23 33286-99 (1) 37589-75 



Here 25170 has the normal limit. 



In m = 5, D 22 is represented by two displaced lines, viz., 



(5) 36374-88 as D 22 (-3S) gives ... 70*52, 

 (In) 36367*70 as D 22 (28) gives ... 70*60, 



but the first is D n (4) — u, and is too strong to represent D 22 . 

 It forms a merel}' numerical coincidence. 



