52 THE INTERFEROMETRY OF 



visibility in case of the grating is not warranted. In the second order of the 

 ruled grating or with a grating of higher dispersion (D = iy5Xio- 6 cm.) the 

 field was too dark for experiments of this kind. In this work, however, I 

 obtained the linear phenomenon for the first time, from the double diffraction 

 of a film grating. 



In conclusion, it is interesting to refer to a relation of the reversed and the 

 inverted spectra and their interferences. If in case of reversed spectra one 

 of the superposed pair is rotated 180 in its own plane, around an axis normal 

 to that plane and through the line of symmetry, the new pair of superposed 

 spectra is an inverted system. At the same time the interferences which are 

 ellipses in both experiments probably rotate their major axes 90. In the 

 case of reversed spectra this major axis is transverse, coinciding with the line 

 of symmetry in a given wave-length, and the ellipses are extremely eccentric, 

 whereas in the case of inverted spectra the major axis is probably longitudinal. 

 It is not unusual to obtain a single line running all the way from red to violet ; 

 but arrow-shaped forms never occur, so that the ellipses are rounded forms 

 and belong to distant centers. An adequate reason for the highly eccentric, 

 closely packed elliptic fringes of reversed spectra on their evolution from the 

 round ellipses of inverted spectra by rotation is yet to be given. 



23. Apparent lengths of uniform wave-trains. In 16 certain results were 

 given which made it seem plausible that the path-differences within which 

 interferences are producible (i.e., the apparent lengths of uniform wave- 

 trains) increase, as the dispersion to which the incident collimated white 

 light is subjected is made continually greater. Work with this quest in view 

 is reported in table 10, the plan being to produce the interferences by one and 

 the same method, but with a successive variation of the dispersion of spectra. 

 The method, figure 14, was first selected for this purpose, inasmuch as the use 

 of prisms and gratings of different dispersive power at P meets the require- 

 ments, while spectra of the first and second order are equally available. 



It is obvious that in work of this kind the spectra must be bright, otherwise 

 the fine lines will escape detection. Deficient values will thus be attained 

 if the spectra are too dark. Moreover, the results can not furnish data of 

 precision, since the exact instant at which fringes, continually decreasing in 

 size, have actually vanished, can not be fixed; and it is the fine fringes which 

 furnish a considerable amount of the displacement. The differences, however, 

 are so large that not only orders of values are clearly apparent, but the 

 ranges more than sufficiently so to substantiate the argument. 



It is possible that the method, figure 14, gives the half -ranges only, since 

 the efficient patches of light, figure 8, can not cross each other. The methods 

 applied will nevertheless be trustworthy, since they are identical, the same 

 telescope and other appurtenances being used throughout. Later the grating 

 method (fig. 3), suitably modified, will be used. Path-lengths of a meter or 

 more were usually admissible. 



In table 10 the first series of measurements is obtained with a 60 prism, 



