207 
incident beam sla is divided into a reflected and a transmitted one. 
The reflected one follows the path abcdea /f, the transmitted one 
‚the path aedebaf. In some experiments the flame of a MrKèr- 
burner, in which a platinum spoon with common salt was placed, 
was put between e and d. In another case a greater gradient of 
density of tbe sodium vapour was obtained by introducing sodium 
into an iron tube connected with an air-pump and closed on both 
sides with glass plates. By heating this tube at the bottom and 
cooling it at the top, the desired density distribution could be obtained. 
With the flame between e and d the phenomenon was observed 
reproduced on the Plate in Fig. 1. 
Fig. 2 represents the interference phenomenon with a sodium 
prism of greater density. Both photos were taken with a spectro- 
scope with one glass prism; the first is enlarged 7 times, the second 
4 times. In Fig. 1 the two absorption D-lines are visible. In fig. 2 
the whole region round the D-lines has disappeared. 
It will not be attempted here to give a full explanation ; it would 
claim more space than we can afford to give it. Even without the 
sodium vapour the explanation of the dependence of the interference 
fringes on the mutual position of the five reflecting planes and on 
the thickness of the slightly silvered mirror is a rather complicated 
problem, which has not been treated in detail as yet for so far as 
we know. . 
We shall confine ourselves to a few remarks on circumstances 
that play a role in the appearance of the phenomenon. 
A point to which we draw attention is the particularity that the 
rays which travel in opposite directions, are deflected by a different 
amount after their passage through the sodium prism. For if the 
vapour prism is placed between e and d, ray aed will have to 
travel over the long path dcba/, before it reaches the object glass 
of the telescope, abcd on the other hand only the short path eaf. 
In. this connection we may still mention that nothing has ever 
appeared of a dependence of the velocity of propagation of light 
on the intensity. We assume that this does not exist, else this 
dependence would already give rise to phase differences in our 
experiment, because one ray is weakened at the beginning of its 
way, another at the end. 
The following auxiliary experiment is also of importance. We 
introduce a screen with a narrow aperture between / and a in 
order to insulate as, it were, a ray of light. In the principal experiment 
an image of the interference fringes was thrown on the slit of the 
spectroscope by the aid of the lens /; now, however, we increase 
14* 
