460 Prof. W. M. Hicks on 



values for m — 2, 3, the formulae for Si and D n are 



S 31523-48-N/{». + -431291- '^^}\ 



/( -091 7731 2 



J m + -989023--^-^ , 



or if the order in S be taken as m— ^, 



S = 31523-48 -N U m-*5 + *924719 — ~~^ \ ■ 



The calculated Si (5) =27783*86 is sustained by two linked 

 lines, viz. 



u. (6 n)3512'19 = 28464-24- 680*68 = 27783-56, 

 (2r0 3688-60. u = 27102-99 + 680-68 = 27783-67. 



The latter is the line allocated by K. K. to D n (4) which, 

 we have seen cannot be correct. The calculated value 

 for D n (4) is « = 27107-52. An arc line by C. T. at 

 n = 27104-60 would be (2 6\) D u , also by C. T. at n = 27117*17 

 is 9'63 ahead and as 2 5 shifts 9-88 it would be (-2S)D n . 

 These are therefore possibilities. Hasbach gives X = 

 3687-5 I. A. or ?< = 27ll0'17-7-3 clX R.A. His excitation 

 may have produced the normal set. 



lite P series. — There is a clear set for P(l), strong, with 

 the normal separation, showing reversals, and the Zeeman 

 patterns belonging to P l5 P 2 . With the value of S x (go ) just 

 obtained the limit P (go ) is known with the same exactness. 

 But although the observed region should include several, 

 succeeding orders it is difficult to allocate them with certainty. 

 There is a large number of weak lines where P (2) should 

 occur. Like the S and D series the P series appear as if 

 frayed out and any summation lines lie far down in the 

 ultra-violet beyond any observed region. There is a line at 

 49363'70±4*9 (X=2025) showing reversal and in about the 

 proper position for l 3 ! (2). P 2 (2) should be about 60 above 

 it, but there is no trace of it. It gives for the combination 

 p 1 (1) — p 1 (2), n = 18580-93d=4'9 which may be the observed 

 spark line (E. Y.) (3) 18579*61. Another spark line at 

 n — (3)18772*87 might correspond to p 2 (1) — p 2 (2) giving 

 the separation of the P (2) lines as 55' 10, in order but rather 

 small. The formula, calculated from this line as P 5 (2) with 

 P(l) and the known limit gives lines for succeeding orders 

 which have not been observed, nor does the calculated 

 sequence for P (3) give any p(l)—p{?>) combinations. If, 

 however, the weak line r t =(l)49069-64±2-4 be taken as 

 F 1 (*2j with P 2 (2) not observed, the resulting formula gives 



