384 



red, which increases on ap|)ioiicliing the sun's limb, may be con- 

 sidered in connection with these inferences). 



Let us now imagine our small spectral region to contain tioo 

 neighbouring sharp absorption lines, then we have: 

 ,.-1 = K-1) + («,-!) + (n,-l), 



where n, — 1 is again assumed to be nearly constant, and the other 

 two terms are strongly variable with A. In the region between the 

 lines, (?z, — 1) and (n, — J) have o|)posite signs (cf. Fig. \,b, where 

 the three terms are represented separately). The resultant n — 1 =f{i) 



Fig la 



Fig 2a 



Fig Sa 



Fiy 4„ 



F,a lb 



S(»,-') 



|Z('.-')I 



r..SM' 



n-l n-l 



Fig. ia^ib. 



shows a point of inflexion there (Fig. 2,b) ; and, owing to the 

 opposite signs of (n, — l)and(«, — 1), the modulus |(w, — 1)+(?J, — 1)-|- 

 -|-(?i, — 1)1 is smaller tiian |(n, — 1) -|- (n, — 1)| in the left section, 

 and smaller than \{?i„ — 1) -|- (n, — 1)| in the right section of the 

 interval (cf. Fig. 3,6), so that on the two sides of the refraction 

 lines that face each other the weakening of the light is less than 

 it would be if the lines stood wide apart. The "centres of gravity" 

 of two neighbouring refraction lines are, therefore, a little more 



