DISTRIBUTION OF INTENSITY IN BROADENED SPECTRUM LINES. 
481 
The form of the curve for H a , and in fact of all curves obtained, is not consistent 
with the supposition that g is the same for all components. There are two 
alternatives to consider in a very simple proof. If the separations are very minute, 
all points on the contour not between two components—or in other words all points 
except very close to the vertex, are outside all the component axes, and if q were 
constant throughout, all the boundary beyond a small distance from the vertex 
would take the form 
I a e pytan “ = (l 0 + 2 2 I r cosh go- r )e- 9 ", 
and be entirely straight. This is not the case. In the second place, if the 
separations were comparable with Stark’s, the initial part of the curve would be 
j^pytana _ + 2 2 I r cosh xq . e -9 ' 7 ', 
r = 1 
and dyfdx = — q/p tan a. at x = 0 , whereas near y = 0 in the final part of the curve 
outside all the component axes, the slope is again — g/p tan a, The initial and final 
slopes should therefore be equal whatever the nature of the kinks. This does not 
occur, and we must therefore conclude that the rates of attenuation of components 
are different. The same conclusion can be deduced in other ways from the curves. 
The rate of attenuation of the central component is given by the initial slope in all 
cases, if the separations are not so extremely small as to make this an indefinite 
quantity. From the photographs of H a in the present experiments we find as the 
mean of several measurements, 
p tan a lq = 0'057, 
whence q — 16‘2 for the central component, since p tan a = 0'922. 
It has been taken for granted that the separations in H a are not minute, in 
accordance with later work. The detailed description of H a will follow, our present 
object being the derivation of results applicable in general to the whole series of 
curves, with H a as a convenient illustration. 
The values of q decrease as the separation of the components increases. This can 
be deduced from the fact that the final slope of all the curves to the axis of x (the 
base of the photograph) is smaller than the initial slope. For the final slope is 
derived from the final branch, 
and 
dy _ 
I c e p1jUna = I 0 e~ X9o + 2 2 2 l„ cosh q n * n . e~ iq \ 
r = 1 
a 
-p tan = -j I o g 0 e _x9 ° + 2 2 I„g» cosh q n <r n . e~ yq ' \\\ I u e x<?0 + 2 2 I„ cosh q n a n e xq ‘ 
