REVERSED AND NON-REVERSED SPECTRA. 



35 



under conditions where the paths of the component rays may have any 

 length whatever. It is thus an essential extension to the method (fig. 21) 

 given in the preceding report (PP f , prisms; M, N, mirrors; Gp, Ives prism 

 grating; T, telescope), where the path-differences were essentially small and 

 the spectra reversed. 



In figure 22, P is the first prism cleaving the white beam L, diffracted by 

 the slit of the collimator. M and N are the opaque mirrors, the former on a 

 micrometer. For greater ease in adjustment, the second prism P' is here 

 right-angled, though this is otherwise inconvenient, since the angle 8 = 90 <p 

 is too large. The rays reflected from P' impinge normally on the reflecting 

 grating G (D = 200X10^) and are observed by a telescope at T. P, P', M, 

 and N are all provided with the usual three adjustment screws. P' must be 

 capable of being raised and lowered and moved fore and aft. The field is 



21 



22 



brilliantly illuminated. When the path-difference is sufficiently small,' the 

 fringes appear and cover the whole length of superposed spectra strongly. 

 They are displaced with rotation if M is moved normally to itself. 



As first obtained, the fringes were too closely packed for accurate measure- 

 ment. But the following example of the displacement e of the mirror M, 

 for successions of 40 fringes replacing each other at the sodium lines, shows 

 the order of value of results : 10^=1.55, 1.40, 1.60, 1.55 cm., so that per fringe 



o~ 6 cm. 



cm. 



The computed value would be (<p, the prism angle) 



X 58-93 __, 



oe = 



2 COS 5/2 2X.8l 



assuming 6 = 90 <p. The difference is due both to the small fringes, which 

 are difficult to count, and to the rough value of <5. The range of measurement 

 is small (if M only moves), not exceeding 1.6 mm. for a moderately strong 



