426 DR. W. M. HICKS: A CRITICAL STUDY OF SPECTRAL SERIES. 



belong to the modified i/ 1} v a which D satellites have already shown in Kr and X, whilst 

 5371, 2641 or thereabouts belong to the normal separations depending on displacements 

 in the S(o) alone. Moreover, another very frequent separation is 5631, connected 

 with other sets as triplets with a v. 2 in the neighbourhood of 2800. This at once 

 suggests the analogue of the 1864 F series of X. 



A first quite definite starting point, from the material at disposal, is found by a 

 search for lines of the F and F type, or the twin AB sets. The limit A belongs to 

 a D sequence, which from analogy with Kr and X should be expected to be of the 

 order n = 30000. Now in the observed spectrum there is a long gap between 27671 and 

 32031, within which such limit must be. That no lines should be found near this 

 limit is to be expected. If, however, such double sets exist we should expect to find 

 sets of lines with exactly the same separations on either side of this gap. 

 Unfortunately there are only four lines on the violet side, but one such set is found. 



They are 



(l)26669'23 (0)33259'05 



897-13 896-91 



(2)27566'36 (0) 3236214 



The corresponding limit should be the mean of either of the two corresponding lines, 

 viz., 2996414 or ...4'25, say 29964'20"05. The possible observed errors in these 

 lines are not large and the practically exact equality of the two separations is strong 

 evidence of the reality of the suspected connection. But any doubt on this point 

 must be removed when it is noted that by RYDBERG'S tables, the separation 897 is 

 that due to two denominators 578, 6 '7 8, whilst if the denominators are calculated 

 using 29964 as limit the same values are found. The two results are quite independent. 

 By a further use of RYDBERG'S tables it is possible to find approximate positions for 

 other lines of the F system, the F being quite beyond the observed region in the ultra- 

 violet. Such lines are found at (5) 15846, (4)22310'4, (8)25171 for m = 2, 3, 4. 

 Again, connected with 15846, are (3) 21488, (6) 24296'9 giving separations 5642, 2809. 

 The lines 22310, 21487 are due to C. R.'s copper electrodes and are subject to considerable 

 possible errors d\ = lA, dn = 5, or even more. These strikingly correspond to the 

 1864 sets in XF. 



The limit 29964'20 must be very accurate and subject only to any systematic errors 

 in R. and R.'s measurements. This is shown by the exactness of the observed 

 separations in sets so far removed from one another as 26669 and 32362. Using this 

 limit with the lines 15846, 26669'23 for m = 2 and 5 the calculated formula is 



'20-N /{ 

 / I 



= 29964'20-N m+ 757457 + 



m 



The two lines used may be regarded as having possible errors dn = 2 and '7 

 respectively and any consequent errors in the constants will scarcely affect the 



