324. Mr. J. C. M‘Connel’s Notes on 
If the arrangements permit of taking readings when the 
plane of polarization is inclined at 90° to its two first positions, 
we can, by taking the mean of all four readings, get rid of 
the error depending on the squares of small quantities, as is 
evident from (8). This is what Lord Rayleigh has done in 
his recent measurements of the electromagnetic rotation in 
bisulphide of carbon. Fig. 3. 
The formula (6) applies to almost 
every mode of cutting a Nicol, ex- 
cept the important case when the 
ends are cut off square to the length, 
the case of the so-called “ flat-ended 
Nicol.” This case requires the in- 
vestigation to be modified. The 
peculiarity consists in N lying very 
close to P’. 
v= 90°—NP’P+NQ’Q—XQ’Q, 
as before ; 
and by (4) and (5), 
tan (90°—XQ’Q) 
_ 4 i aS 
igen wo) aveine 
If we take N as the point from 
which to measure 8, P, and Q, will 
coincide with N. 
We have now to find NP’P—NQ’Q. The process is pre- 
cisely similar to one already performed in finding equation (8) ; 
and we obtain 
>. 
PpPp2— QQ” 
NP’P—NQ’/Q= 5 tan NP’/P 
2 
—] 
F “Fa a ; 
a ap w— 
=la—meB, say. 
Taking XN =63°, w=1'52, we obtain 
[=°33,; -m='83. 
So the numbers are almost exactly the same as in the case of 
the ordinary Nicol. 
As a test of the accuracy of the above calculations, I made 
a few observations on a flat-ended Nicol. 1 found that, turn- 
ing the Nicol through 30’, about an axis in the principal plane 
ee 
