Series System in the Spectrum of Gold. 23 



It may be noticed, tliat there are in the foregoing many 

 instances where displacement on the sequent produces a 

 nebulous line, whilst on the limit it is sharp. We cannot say, 

 however, whether this is a general rule. To indicate that the 

 given lines are at least near positions for normal lines, the 

 following values of the denominators calculated from them 

 are given: 1-911560, 2-926185, 3929447, 4'93l517, 5-930130, 

 6*936977, 7-929360.. 



P and S Series. 



It has generally been assumed that the strong pair in the 

 ultra violet, whose W.N. are (10) 411 7294, (10)37357'62 

 with separation 3815*32, form the first P doublet, in analogy 

 with the strong violet pairs in Cu and Ag. Although at 

 first sight this assumption would appear natural, it is easy to 

 show that it cannot be sustained. In the first place — as was 

 pointed out by Kayser and Hunge in their original measures, 

 — they are not in step with the Cu P and Ag P lines. The 

 P l lines would run Cu, 30782 ; Ag, 30471 : An, 41172 with 

 Au much larger instead of somewhat smaller than Ag. But 

 the matter is definitely settled against the supposition when 

 the question of the limit is considered. The rule that 

 S(oo)=jo(l) is always so closely observed that it may bo 

 considered as absolute. The rule that P(co ) =s(l), although 

 never exactly fulfilled when s(l) is extrapolated from the 

 s(in) formula, is also so very nearly followed that it should 

 be regarded as holding very approximately. In the case 

 of gold S(oo)=29469-85 = N/(1-929145) 2 ." If 41172 is 

 AuP^l), P(co) = 70642 79 = N/(1-246015) 2 . The undoubted 

 S(2) lines (see below) give *(2) = N/(2-545878) 2 . It is 

 clear, therefore, that 5(1) cannot be N/(1'246) 2 or near it, so 

 that 41172 cannot be Pi(l), or belong to a principal series 

 of the ordinary type. Nevertheless, if regarded as belonging 

 to a series in which the separations diminish with ascending- 

 order, it is possible to arrange a set for whose real existence 

 as a definite series some evidence can be adduced, apart f rom 

 the fact that the set are reproducible by a formula of the 

 ordinary type. I give the evidence, such as it is, premising 

 that considerations to follow may throw doubt on its validity. 

 The letter P' will be used provisionally to denote the series, 

 as its type is undecided. 



The value of P(oc ) is 70642*79 + £, and the denominator 

 of the first sequent = 2-940001— 115*9f. We should, there- 

 fore, expect to findP'x(2) with a denominator near 3*94 . . . , 

 or wave number near 58000. Also we might hope to find 

 evidence in support by the combination line p(l)— jp(2). 



