The Dispersion and Absorption of Fnchsin y 



rotated through about 20° to produce a shift of one band, 

 was therefore used, instead of the usual thick glass ones, and 

 witli this no such disturbing effect was observed, since an infini- 

 tesimal movement of this would not visibly affect the bands. To 

 get around the difficulty of mica having three different indices 

 of refraction, the piece of mica was cut so that one of its axes of 

 elasticity was the axis of rotation. By using plane polarized light, 

 and making the plane of polarization parallel to the axis of rota- 

 tion of the mica, the retardation introduced by the mica when at 

 different angles was always proportional to just the one index 

 of refraction. It was found incidentally thattlie introduction of 

 the Nicol prism, N , fig. 2, made the bands much more distinct. 

 On rotating the Nicol prism through a right angle, the bands 

 became more confused than with the Nicol out. 



The compensation C, fig. i, was made by cutting a thin piece 

 of mica in two and attaching one of the pieces to a device by 

 means of which it could be rotated through an angle and the 

 angle read off. These two pieces of mica were placed in the 

 same path as the fuchsin, the fixed piece and the fuchsin in the 

 lower half of the path, and the movable piece in the upper half, 

 the line of separation of the mica precisely on a level with the 

 edge of the fuchsin ; in both upper and lower halves of the other 

 half was a very thin film of fuchsin that served as an absorbing 

 screen. When both pieces of mica were at right angles to the 

 beam of light, their effect was small, because they affected the 

 air bands alike, being of equal thickness. It could be insured 

 that they were equally thick by removing the fuchsin and observ- 

 ing whether the upper and lower spectral bands coincided when 

 both pieces of mica were in the same plane. 



It is usual to calibrate a compensator by turning it through 

 different angles, and noting the angles corresponding to dif- 

 ferent numbers of bands displacement. However, for this work, 

 as it was always less than a single band displacement that had 

 to be measured, it did not help to know the angle corresponding 

 to one band displacement or two bands displacement, because 

 the greatest change of curvature of the calibration curve comes 

 betv/een zero and one. 



107 



