220 
2 f= 1, in consequence of which the following equation is left: 
eve | (<l).cos 4am dm dl ET (cl) sin ov dm dda.) (BB) 
0 
a 
0 
In this Z(cl) and //(c/) are registered functions, which are zero 
for great values of / according to equations (21) and (22). When 
we wish to examine the finer structure of practically monochromatic 
light, tbe Z and Z' curves for white light must first be determined. 
The distance of the zero points N and N' to the initial points of 
the curves follows namely sharply from this. The prism is now 
again put in the same place as before at the beginning of the 
measurement *), and in another photographic plate the Z and Z' 
curves for less compound light are now registered; now the zero 
points N and N' lie at the same distances from the initial points 
as before; without the previous measurements they would, however, 
hardly have been recognizable; for absolutely monochromatic light 
there would even have been no difference to be detected at all 
between the maxima or between the zero points. 
The white light, however, has as it were, gauged the photographic 
plate. A procedure as deseribed above is rendered possible through 
the fact that the zero points are known, even for the most strictly 
monochromatic light, and it enables us here too at least to closely 
approximate the required function x(m). In practice this approximation 
cannot be carried to an extreme, because at too great differences 
of path the beams of light become incoherent on account of the 
irregular vibration of the source of light, a drawback that will 
adhere to every method that rests on interference. 
This limit for the dissolving power of an interferential spectros- 
cope, can of course not be exceeded here either. But all the data, 
with which interference phenomena can furnish us about the com- 
position of apparently monochromatic light are used mathematically 
in the simplest way according to the method discussed, and lead 
to the drawing up of a function, which resembles the required 
distribution of energy g(x) most closely *). 
MicuELson first works up his data to a curve of visibility V, and 
concludes from this to the value of C* +S’, but has now to assume 
S= 0, ie. a symmetrical distribution of energy in order to be able 
1) In this use can eg. be made of the distinct movement of the photographic plate. 
2) To make the approximation still closer, we should have to make the source 
of light vibrate more regularly, or possess an entirely different instrument, which 
does not rest on interference. 
