Eartlis Motion on the Rotation of a Refracted Ray, 593 



incidence of 70°, the azimuth o£ the plane of polarization of 

 a ray could be varied at will. These were then mounted, 

 one at a time, between the polarizer and the analyser. 

 Starting with an azimuth of 45°, the analyser was adjusted 

 for minimum intensity, being shifted to receive the light 

 axially, since, on inserting the wedge, the direction of the 

 ray was changed. A second plate and its mounting was then 

 inserted and adjusted so as to receive the ray from the first 

 one in an azimuth of 45°, the wedge-shaped plate being 

 turned in its own plane upon its ring-carrier until the 

 multiply reflected images were thrown out of the field of 

 view. This necessitated a somewhat spirally-shaped arrange- 

 ment of the system when the analyser and all the plates 

 were finally mounted. A thin strip mica compensator * was 

 placed after the polarizing half-shade system, to compensate 

 for any slight amount of elliptic polarization which might be 

 produced by the plates themselves. This was quite effectual 

 and, with three or four plates, moderately close settings 

 could be made with the plane-polarizing half -shade system. 

 However, with white light, the colour dispersion due to the 

 rotation prevented anything like the sensibility required by 

 the experiment. I attempted to compensate for this by 

 means of quartz, as did Fizeau, but still found that, with 

 white light, the sensibility was far too small, I then tried 

 sunlight, using one of Wulfing's spectral " sifters " to obtain 

 homogeneous light. But here my intensity was too low. 

 A powerful mercury lamp gave similar results. I then saw 

 that white light and complete compensation of the dispersion, 

 due to the rotation of the plates (this rotation was approxi- 

 mately through one quadrant), would be necessary to attain the 

 required sensibility. Only moderate pains had been taken in 

 grinding the plates, and I found that, with all ten plates in, 

 the astigmatism was so large that the definition of the " half- 

 shade ^ was greatly reduced. I therefore decided on a 

 different arrangement which, though apparently unlike that 

 of Fizeau, involved the principle of his experiment and, 

 hence, would be valid in a repetition of the test. This con- 

 sisted in combining pairs of plates into the " crossed " system 

 mentioned above. Thus (fig. 1) if in the equation 



tan a 



tan ft = -p s , 



cos (i — r) 



(ft — a) is the rotation experienced by a ray after refraction 

 at a single surface, 8(fi—a) would be the increment, say, due 



* Phys. Review, Feb. 1904, p. 70. 

 Phil Mag. S. 6. Vol. 10. No. 59. Nov. 1905. 2 S 



