REVERSED AND NON-REVERSED SPECTRA. 83 



It is convenient to describe the homogeneous fringes first. White light 

 must be absent, the wide field full yellow, and the longitudinal and side edges 

 of the two slit images sharply superposed. When the fringes appear they will 

 usually be oblique; but they may be made vertical by rotating the grating 

 on an axis normal to its face. If the grating is in the symmetrical position of 

 figure 59, the size of fringes is an intermediate minimum. To enlarge them, 

 curiously enough, the grating must be slightly rotated, either way, on a ver- 

 tical axis. The fringes then pass through a maximum of size at a definite 

 angle on either side of the minimum. In such a case they also appear rapidly 

 to become irregular and their perturbation is naturally enhanced. They con- 

 tain a double periodicity, which will presently be carefully examined. 



Fore-and-aft motion of the grating has no effect. In displacing the mirror 

 at M on the micrometer, the fringes remain visible for an excursion of at least 

 0.7 cm. In fact, in case of a strong telescope and wide slit they were not lost 

 for a micrometer displacement of over i cm., i.e., much over 30,000 wave- 

 lengths of path-difference. As a rule, the fringes are strong only in part of 

 the yellow field, and in such a case the center of intensity moves with the 

 displacement of M across the slit image, to disappear at the edges, as in the 

 usual cases of displacement interferometry. Slight non-coincidence of the 

 horizontal edges of the slit images slightly rotates the fringes, but they soon 

 vanish completely. Slight rotation of the grating around the vertical axis 

 distributes the fringes more evenly over the field, the proper setting being 

 determined by trial. Displacement by aid of a compensator of glass gave the 

 usual results. 



Later I returned to the experiments with sodium light and with the grating 

 rotated around a vertical axis to the right or left and out of the symmetrical 

 position of figure 59. In each case the fringes passed through maximum size 

 at an angle of asymmetry of about 5 or 10 from a normal position. Beyond 

 or below this they diminish in size. Naturally, to bring the fringes to the 

 center of the field, the micrometer screw at M or N had to be adjusted for 

 path-difference, as in displacement interferometry generally. 



The details of the interference patterns obtained were in astonishing variety. 

 Suppose that by rotating the grating around an axis normal to its face the 

 fringes are made nearly but not quite vertical at the beginning. Then on rota- 

 ting the grating around a vertical axis into the position for maximum size 

 just specified, the standard type of large fringes seen are of the appearance 

 shown in figure 6oa. In other words, they look and behave like independent, 

 thick, twisted cords, hung side by side. The evolution of these independent 

 parallel striations of fringes may be detected on rotating the mirror M or N 

 around a vertical axis, thus moving one slit image in definite amounts, micro- 

 metrically, over the other, horizontal edges remaining superposed. As the 

 one slit image passes in this way across the other, the original type, figure 606, 

 apparently continuous, breaks up and enlarges into the type c by the rotation 

 of its parts. Thus the successive lengths of the continuous fringe b behave 

 like a series of magnetic needles, each rotating on its own pivot. These may 



