REVERSED AND NON-REVERSED SPECTRA. 



137 



must coincide horizontally and vertically. They do not in this case probably 

 coincide in the fore-and-aft direction, for the plates, etc., are not optically 

 flat. When the slit images are separated at the same horizontal level into 

 two fine parallel lines, the complementary fringes in fact become Fresnellian 

 fringes, finer as the slit images are more separated and as the ocular is more 

 rearward or forward. This is precisely what should occur. We may conclude, 

 therefore, that the complementary fringes are Fresnellian interferences of two 

 slit images and that the central hyperbolic forms are due to outstanding front 

 and rear positions of the two slit images, which seem to coincide in the field 

 of the telescope. Differentiated from these, the achromatic fringes are refer- 

 able to the colors of thin plates. I have, in fact, also succeeded in obtaining 

 the complementary fringes in the shape of broad, straight, gorgeously colored 

 vertical bars, without suggestion of hyperbolic contour. 



An attempt was made to get quantitative estimates of the passage of fringes 

 on rotating the paired mirrors over an angle a. To control the small angles 

 the device, figure 95, was improvised and did good service. Here e is the 

 tangent screw of a divided circle 6 inches in diameter. It is surrounded by 

 a snug annulus of cork and holds the brass ring /, on whose surface a coarse 

 screw-thread has been cut. Near this and with its axis in parallel is a quarter- 

 inch screw a and socket (not shown) controlled by the disk b. A strong linen 

 thread c is looped once around / in the grooves of the screw and once or twice 

 around a, the string being normal to the cylinders and kept taut by two 

 small weights, g about a half ounce and h about an ounce. The head b may 

 be turned either way and the angle read off in minutes on the head of the 

 tangent screw e. 



The theoretical value, apart from glass-paths and other corrections, should 

 be, per fringe vanishing, 



where R is the radius of rotation corresponding to the angle A a and X the 

 mean wave-length of light. In the given adjustment, 2R=io cm. was the 

 normal distance apart of the two interfering beams. Hence 



X 6oXio- 

 Aa = = - = 6Xicr 6 radians or 1.2 



2K 10 



per vanishing fringe. As the change of glass-path of one beam would have 

 to be deducted from 2R, a somewhat larger value would be anticipated. 

 Testing the complementary fringes (white light), the passage of about 25 



