REVERSED AND NON-REVERSED SPECTRA. 



17 



TABLE 4. 



The second series here is practically the mean of the 

 two, though the reason for these large discrepancies is 

 not clear to me, even in consideration of the wedge- 

 shaped plates. The mean of the results may, however, 

 be used for computation. 



The path-increment introduced by the glass of thick- 

 ness 2 = 0.489 cm. and index of refraction n = 1.526, at 

 an angle of incidence i and refraction r for n fringes, 

 beginning at i = o, may be written (see fig. 7, where I is 

 the incident ray) 



/cos (i r) 

 e\ 

 N 



/i \ 



= en\ - i) 



\ cos r / 



cosr 



-0 



This is a cumbersome equation. If the angles i are small, the cosines may 

 be expanded and then approximately 



which, since i = nr nearly, may be further simplified to 



Thus for the second set (mean) 



4.8 

 6.4 



5-9 c 

 6.9 



6.8 

 6.5 cm. 



The wave-length thus comes out very much too large, but in consideration 

 of the inadequacy of the fiducial position, i = o, this is not unexpected. 

 Thus the probable values of i in the tables (computed from X correct) agree 

 with the third series. In addition to this the effect of slightly wedge-shaped 

 plates, etc., can not be ignored. For the first set (mean values) the results 

 are similar, being 



io 5 X = 6.8 7.0 6.9 



7.0 



7.0 cm. 



if computed by the approximate equation. This is again too large, but the 

 probable value of i computed from the correct X, as before, agrees nearly 

 with the first series of this set. 



