REVERSED AND NON-REVERSED SPECTRA. 



37 



the median line. With G placed at the proper distance, the two spectra seen 

 at T will usually be quite distinct in the field. They should show the shadow 

 of the black line across the slit, at the same level in the spectra. The longi- 

 tudinal axis of the spectra may then be made collinear by slightly tilting the 

 edge of P' to the vertical, on a horizontal axis, with the adjusting screws. 

 M and N are then rotated on a vertical axis till the D lines coincide. Small 

 changes may be completed at M and N. The fringes when found are usually 

 strong, but very fine, less than the D\Dz distance in width. I have been able 

 to increase them to a width of 2DiD 2 , but they are then faint. The two 

 illuminated strips on the grating may even be an inch apart, but the fringes 

 are as usual large when this distance is the smallest attainable (virtual 



24 



23 



a 



/ 



a 



coincidence). The grating may be moved fore and aft without effect. As 

 N is moved on its micrometer, the interferences are first seen as vertical 

 hair-like striations, which gradually enlarge, rotate, and vanish just before 

 reaching the horizontal and at maximum size. The range of displacement did 

 not exceed 0.15 cm. for this rotation of 90, so that the total displacement 

 for 1 80 of rotation would be about 3 mm. Since N and T are close together, 

 the manipulation is convenient here, but with another lens at T' the phenom- 

 enon could be traced further on the M side. 



To secure a smaller angle of incidence and reflection, 5/2, at M, figure 24, 

 the combination of a silvered 20 prism P and a 30 prism P' was tested. M 

 and N are the opaque mirrors, G the concave grating with its focus at T for 

 inspection by a strong lens. L is the incident beam of white sunlight from the 

 collimator, which is split into the component pencils abed and a'b'c'd' and 

 interfere at T. The results, however, were about the same as above, the 

 range of displacement at M for 90 of rotation of fringes being about 0.15 cm. 

 As a and b make angles <p and <p f with the line of symmetry LL', 



was about 10. 



At a subsequent opportunity I made further trials with the paired prisms 

 of 20 and 30, but failed to increase the fringes above about DiD^/2 width. 



