222. 
from two curves that can be registered, it making no difference 
whether the distribution of energy is a symmetrical or an asym- 
metrical function; in the former case the registered curves are 
mutually dependent, in the latter case they are independent. 
It is noteworthy that the metal reflections in the instrument of 
fig. 7 are not identical. The r-component has been reflected against 
the silver layer at an angle of 45° in air, and the d-component at 
a smaller angle in glass. Probably the phase-shiftings occurring here 
will be the same; otherwise MicnerLsoN could not have drawn up 
equation (4). If this supposition should be incorrect, fig. 8 represents 
an arrangement in which this evil has 
been remedied; the silver layers at Zand 
Z are now only struck at an angle of 
45°. Now the front planes of the prisms 
must be transparent both at the top and 
at the bottom, and the cylinder lenses 
must be above the plane of the drawing. 
The silvering in Z (fig. 8) and in C 
(fig. 7) must be such that the intensities 
of the light of the beams entering the 
telescope are equal, in spite of the fact that one has undergone 
reflections in the lenses. The intensities also can be kept equal by the 
aid of a glass plate G*). When the intensities of the beams in the 
modified interferometer differ from that in the original one, one of 
the terms in equation (25) must be provided with a factor of pro- 
portionality, which is only to be determined experimentally. 
We have scrupulously taken care that the interfering beams did 
not get a relative phase shifting with respect to each other in con- 
sequence of unequal metallic or total reflections, so that the phase 
shifting, which one beam obtains in advance of the other, has only 
Z 
Fig. 8. 
Jt 
been obtained in the focal line of the cylinder lenses E large), 
and does, therefore, not depend on the wavelength. This renders 
the instrument efficacious for all possible sources of light and for 
spectrum regions of any given extent, supposing the lenses to be 
made achromatic ®). The result of the method, expressed in equation 
(25), is therefore perfectly exact for all possible spectrum regions 
1) With such a plate in the arrangement of fig. 7 we must make p= gq. 
4) That the photographic plate QQ’ in fig. 7 may be adjusted once for all. A 
possible dispersion would not have influence on the constant phase shifting „either. 
