﻿952 Prof. C. Barus on the 



slit and the whole divergence in the collimator is due to the 

 diffraction of the slit. With optical plate none of these 

 difficulties would appear. 



Apart from drawbacks of this nature, which require patience 

 for their removal, the measurements (shown by little circles 

 in the curves) are quite satisfactory as a whole, as may be 

 seen in fig. 7. 



Different values of (N' c )e in the two series of fig. 6 are 

 due to the backlash in the micrometer*, as the mirror travelled 

 in opposed directions in these cases. If the observed mean 

 data are compared with the computed data for AN C , the 

 results are practically coincident except at the G line, where 

 the ellipses were too dark to be seen distinctly. Hence 

 equation (3) for N c may be accepted as correct. If we 

 write it 



N - = v-^-^i( ( ^- sin2I) - M t> • • (4) 



and put m 2 = p, 2 — sin 2 1, t r = log^, the equation becomes 



M-£)- (5) 



On integration and reduction this may be written 



£/xcosR = \(C—jXd\/\ 2 ),. . . . (6) 



where is a constant of integration. Unfortunately equation 

 (6) is too complicated for the computation of /i, when K c is 

 observed in terms of X throughout the spectrum ; for the 

 reference to the standard E line gives (sin I = constant) 



x/^-si^I/X-^l- sin*I/X E =-- f AN c ^. 

 If, however, I is very small, or R = 0, 



(?) 



Let/i=B(l/A, 2 — 1/Xe) as above. Then either equation (3) 

 or equation (fc) for 1 = 0, leads to 



(AN C ) =(N C -N E ) = 3^B(1/X 2 -1/X|). . . (9) 



This equation is to be tested in the following paragraph. 



10. Continued. Case of 1 = 0. — The limiting values of I in 

 equation (7) are interesting. They correspond to the form 

 of interferometer approached in fig. 8, where gg is the face of 



* N' refers to the arbitrary zero on the drum. 



