REVERSED AND NON-REVERSED SPECTRA. 81 



In other words, in case of the rays nM, the violet is incident at a larger 

 angle at G' than the red, and but one color (yellow) can be diffracted along 

 G'T, whereas in case of the rays mN violet is incident at G' at a smaller angle 

 than red, and G' may thus be so placed that all rays are diffracted along G'T, 

 supposing the two gratings to be nearly identical as to dispersion. Figure 

 58, presently to be described, suggests the inclination of the successive verti- 

 cal planes in figure 57. 



One curious result deserves special mention. Each separate spectrum 

 (a or b, fig. 57, without superposition) shows very definite coarse stationary 

 interferences; i.e., the usual appearance of channeled spectra. The cause of 

 this long remained obscure to me, but will be explained in Chapter VI. The 

 gratings being of the reflecting type and the mirrors silvered on the front face, 

 there is no discernible cause for interferences. No film or set of parallel plates 

 enters into the experiments. If in figure 5 7 the grating G' is reflected at M 

 into G'i, and this image reflected in m into G'z, the phenomenon may be treated 

 as if the gratings were transmitting in a manner shown in figure 58. Here the 

 direction of the traces of the grating G and G', the mirrors m and M only 

 are given, together with the direction of the reflected images of G' in M 

 (G'i), and in m ('2). Then the violet (v) and red (r) rays from G impinge 

 on G'z virtually with a greater angle for v and a smaller one for r, as already 

 suggested. An enhanced spectrum must be produced beyond G'z- This 

 second spectrum is channeled. 



36. Experiments. Transmitting grating. Parallel rays. The chief diffi- 

 culty in the preceding experiments was the absence of sufficiently intense 

 homogeneous light. This may be obviated by using the transmitting grating. 

 But as two samples were not available (as in fig. 56), the simplified method of 

 figure 59 was tested, where but a single grating G is used. Here the light L 

 from collimator and slit impinges on the grating G and is diffracted to the 

 opaque mirrors M and AT. From here it is reflected to the corresponding 

 opaque mirrors m and n, to be again reflected to the grating G, and finally 

 diffracted along the line GT. The interferences are observed by the telescope 

 at T. In order that the undeviated white beam may not enter the telescope 

 annoyingly, the diffraction LG takes place in the lower half of the grating and 

 the mirrors are slightly inclined upward, so that the second diffraction GT 

 may occur in the upper half of the grating. To obviate glare in the field, the 

 beam LG is carried to the grating in an opaque tube and all undeviated light 

 is suitably screened off. The distances mn to G and G to MN were about a 

 meter each. 



The interferences were easily found. They are usually at an angle to the 

 vertical, but may be erected by rotating the grating on an axis normal to its 

 face. They were linear and exactly like the cases of Chapter I, probably in 

 consequence of the low dispersion of the grating used. Considerable mag- 

 nification at the telescope is thus admissible. 



The horizontal fringes traveling up or down are available for interferometry, 



