Rotatory Dispersion of Neod-gmium. 273 
length 5790 moves distinctly towards the blue, while rotating 
the nicol in the opposite direction causes a motion towards 
the red. 
If we attempt to explain this behaviour of the band by 
assuming a certain distribution of plane-polarized light in it, 
we find that we have a discontinuity in the position of the 
plane of polarization at the centre of the band. It appears 
to me that the more probable state of things is as indicated 
in fig. 2, the polarization at the centre of the band being 
circular, passing to the state of plane polarization (oppositely 
inclined) on each side of the band through intermediate 
ellipses. This means of course that the two circular com- 
ponents of the plane vibration, which by their unequal 
velocity in the magnetized medium give rise to the rotation 
of the plane of polarization, are unequally absorbed, a phe- 
nomenon which has been observed by Cotton in the case of 
naturally active bodies. 
From a very careful study of the spectrum and the 
direction of the shift of the dark and light regions in it, I am 
inclined to regard the rotations as indicated in fig. 2, the 
incident vibrations being vertical. 
With a thicker film very strong anomalous rotation can 
be observed in the green. After the restoration of the light 
by the magnetic field, if the analysing nicol be turned in the 
direction in which the plane of polarization has been rotated, 
a broad dark band moves from the blue-green region towards 
the absorption-band in the yellow, becoming narrower as it 
moves along, until at the moment when it is at the edge of 
the absorption-band it is so narrow as to be barely visible. 
This indicates of course that the curve of the magnetic rota- 
tion becomes very steep as we approach the absorption-band 
from the blue side, precisely as in the case of sodium vapour 
(fig. 3). On the other side of the absorption-band I am 
unable to determine how the curve runs. The rotation is 
apparently about the same as in the blue-green region, and 
of the same sign, but no certain trace of any moving band 
could be discerned as the nicol was rotated. The curve is 
evidently not as steep on the red side, or the band of ex- 
tinction could certainly be detected in the neighbourhood of 
the absorption-band. So far as could be determined, rotation 
of the nicol extinguished the red and orange portions of the 
spectrum simultaneously, the faint suggestion of a dark band 
coming in from the red side being in all probability an 
illusion, for as is well known extinction appears to begin 
at the faintly visible end of a spectrum, even when it is 
extinguished uniformlv. 
