36 THE INTERFEROMETRY OF 



telescope. But one-half of this displacement is available, as the fringes 

 increase in size (usually with rotation) from fine vertical hair-lines to a nearly 

 horizontal maximum, and then abruptly vanish. This is one-half of the 

 complete cycle. 



If we regard the component beams, a b c and a'b'c', as being of the width 

 of the pencil diffracted by the slit of the collimator, it is clear that the maxi- 

 mum size of fringes will occur when c and c' are as near together as possible ; 

 furthermore, that as M moves toward P r , c continually approaches c r , until 

 b drops off (as it were) from the right-angled edge of the prism P'. To get 

 the best conditions i.e., the largest fringes c must therefore also be moved 

 up to the edge of P and very sharp-angled prisms be used at both P and P' . 

 The largest fringes (lines about 10 times the DiDz distance) obtained with 

 the right-angled prism were often not very strong, though otherwise satis- 

 factory. Much of the light of both spectra does not therefore interfere, being 

 different in origin. 



Results very similar to the present were described long ago* and found 

 with two identical half -gratings, coplanar and parallel as to rulings, etc., 

 when one grating was displaced normally to its plane relative to the other. 

 The edges of the two gratings must be close together, but even then the 

 fringes remain small and the available paths also. Strong, large fringes, but 

 with small paths, were obtained by the later method t of two identical trans- 

 mitting gratings, superposed. 



If the prism P' is right-angled (a special case of fig. 21), it may be rotated 

 as in figure 23, so that the rays c and c' pass off towards the observer. They 

 are then regarded through a prism-grating G and a telescope at T. This 

 method admits of much easier adjustment. With the component beams 

 a b, a'b', coplanar, horizontal, and of about equal length in the absence of 

 the prism P', the latter is now inserted with its edge vertical (rotation) and 

 the white slit images in T (without G) superposed, horizontally and vertically. 

 G is then added and the micrometer at M or N manipulated till the fringes 

 appear. As above, they are largest when c and c' are as nearly as possible 

 coincident and vanish as horizontal fringes at the maximum; for the effective 

 parts of c and c' are component halves of the same diffracted beam from the 

 slit. It is interesting to observe, seeing that interference also occurs when 

 one of the superposed spectra is inverted on a line parallel to its length, that 

 such diffraction is demonstrable in case of homogeneous light, even when a 

 slit is absent. Both beams must be nearly at the edge of P' in order that 

 strong, large fringes may be seen. 



The case of figure 22 was subsequently again tried on the large interferom- 

 eter, the distance P to MN being about 2 meters. G, in these experiments, 

 was a concave grating and T a strong lens near the principal focus of G. The 

 adjustment for long distances is not easy. The equilateral triangle of rays, 

 a, a', b f , b, should be first carefully leveled, the edges of P and P' being on 



* Phil. Mag., xxn, pp. 118-129, 1911; Carnegie Inst. Wash. Pub. 149, chap. vi. 

 t Physical Review, VH, p. 587, 1916; Science, XLII, p. 841, 1915. 



