REVERSED AND NON-REVERSED SPECTRA. 



49 



at c", c'", to be joined in the telescope at T. The internal reflection being 

 total, with the rays entering and leaving at right angles to the faces and re- 

 quiring no silvering of the latter, I made use of it for the following experi- 

 ments : M and N are on micrometers with the screw in the direction normal 

 to their faces. P, M, N, P' must all be adjustable. After preliminary meas- 

 urement for equal distances, the fringes were found without trouble. They 

 were strong but fine, beginning with vertical hair-lines and gradually rotating 

 as they grew coarser, till they rather abruptly vanished. The displacement 

 of the M mirror did not exceed 0.6 mm., nor the rotation 30. The spectra 

 being non-reversed, the fringes covered the whole field. Nevertheless these 

 lines must probably be regarded as arcs of circles or ellipses with enor- 

 mously distant centers. In fact, the appearance of the whole of the elliptic 

 symmetry, in the preceding experiments ( 21) with gratings, is also to 

 be associated with a slight difference of length of two overlapping spectra. 

 This is necessarily the case, since the two gratings G and G f , figure 25, have 

 never quite the same constant. The third grating must therefore produce 

 two spectra, the one slightly incremented and the other decremented in length, 

 respectively, as compared with the case for white light. 



One would naturally suppose that the abrupt evanescence of fringes was 

 due to the escape of the b beam at the edge of the prism P'; but this is not 

 possible, as the mirror M was traveling toward the rear. Furthermore, the 

 fore-and-aft motion of the prism P' over several millimeters had scarcely any 

 effect on the fringes. This is unexpected; for the rays, c, c', are compelled to 

 approach or recede from each other by this motion. Finally, the sodium 

 doublets may be moved at some distance (many times their breadth) apart 

 without destroying the fringes. They are often most distinct when the D 

 lines are not superposed. The same is also true for the longitudinal axes, 

 though to a less degree. 



These features are therefore peculiar. The rays c c' were about 5 mm. 

 apart. Unfortunately the faces of the prisms were optically inadequate, so 

 that the sodium lines were not sharp. For this reason no results were obtained 

 with homogeneous light and a wide slit. 



