( 350 ) 
shifted from the position they should occupy with an accurate quarter- 
wave apparatus, by an amount, which in the red is nearly 11% of 
the distance of two succeeding bands. In the green and violet this 
amount becomes 9 resp. 10.5%. 
5. Much smaller deviations gives a recently obtained rhomb, as 
is readily seen by comparison of the photographs 2 and 3. 
A somewhat higher accuracy was still obtained by interchanging 
the prism of the spectroscope used by a more dispersive one. It now, 
however, became impossible to photograph the whole spectrum at 
one operation. Fig. 4 and fig. 5 have been obtained with the old 
FRESNEL-rhomb and the higher dispersion. With the last dispersion 
and the new rhomb one can without measurements scarcely decide 
upon the sense of the error in the relative position of the two band 
systems (see fig. 6 and fig. 7). The deviation of a band in one 
field, from the centre of two bands in the other, never exceeds 
3.2% of the distance of the bands. (The deviations are 1.7, 2.6 and 
3.2% for the red, green and violet). 
Let the deviation, defined in this manner, be p percent, then the 
corresponding error of the phase becomes X 360° = 1.8 .p degrees. 
For the green p is about 3°/ 0 , hence 5°.4. An error of this amount 
in the phase difference of 90°, which exists between the two linear 
components into which circularly polarized light can be resolved, 
may easily be shown to have no influence in the case of the 
measurements of intensity described in $ 7. 
Let completely circularly polarized light be incident upon the 
rhomb, then the light after reflexion by the rhomb, giving a retar¬ 
dation of 90° diminished or increased with a small angle d, may 
be represented by: 
y — acos(nt + d) J' 
Hence 
sc 2 — 2ar y cos 6 y* — a? sin * d. 
The principal axes of this ellipse, become: | a |/2 sin d and a |/2, 
hence their ratio 4 sin d. 
When <f=6°, then ±sind= 0.0522. 
The intensity of the light leaving a Nicol with its plane of 
vibration perpendicular to the major axis of the ellipse, becomes 
(0.0522)* = 0.0027. 
As the minimum intensity, which under the circumstances of our 
observations may be recognized, will appear to be of the order 0.01, 
