of the Vibrations of Polarized Light by Diffraction. 327 
image than the lower, then if the grating entirely cover the 
object-glass, as was always the case in the above experiments, 
the upper diffracted image will be too bright. And if this image 
be polarized horizontally, 6 will be found too great ; if vertically, 
too small. To render the experiment perfect, it is therefere neces- 
sary to turn the rock-crystal through an angle of 180°, and take 
the mean of the two values of 6 so obtaimed. If this precaution ™ 
be neglected, very considerable errors may be introduced, espe- 
cially when smoke-gratings are employed, and I imagine it is 
this that has misled Holtzmann. He observed, in the case of a 
smoke-grating, that for a diffraction of 20° there was a very con- 
siderable difference in the brightness of the horizontal and verti- 
cally-polarized images. This is nearly always so with gratings 
of this description: the upper or the lower image appears the 
brightest, without reference to the position of the plane of polari-. 
zation. With a perfectly accurate grating, M. Holtzmann would 
not have been able to distinguish the slight difference that really 
does exist. 
_ My first experiments were made with a gold grating (1000 
bars to the Paris inch). Light polarized at an angle of 45° 
with the vertical, when diffracted with this grating, gave two 
images, of which neither could be made entirely to disappear for 
any position of the rock-crystal ; and this was still more evident 
when the grating was placed obliquely. The diffracted light must. 
therefore either have been elliptically polarized,or have been partly 
converted into ordinary light. That the former was the case, 
I inferred from the fact that elliptically-polarized light could be 
converted by diffraction into circularly and plane-polarized. If, 
for example, I passed light polarized at the angle « through a 
Fresnel’s parallelopiped, whose reflecting surfaces made the angle 
45° with the vertical, the angle a could be so chosen that one 
image in the telescope could be made entirely to disappear, or, 
on the other hand, so that the two images, on turning the rock- 
crystal, always retained the same intensity. : fire 
By- measurements made in this manner, I convinced myself. 
that the phenomenon is essentially the same as that which 
accompanies the ordinary reflexion of light from polished metal. 
surfaces, and that the effect of the diffracted light is imper- 
ceptible in comparison with that reflected from the edges. 
I now provided myself with various smoke-gratings. Polished 
glass surfaces were smoked with burning camphor, and then 
treated with a few drops of oil of turpentine to fix the smoke to 
the glass. These were then divided by means of a machine into 
bars only an inch long (2, 5, 10, and 16 to the millim.). 
With these gratings I no longer observed any elliptic polari- 
zation. I found no observable differences in the results for the 
