80 THE INTERFEROMETRY OF 



erect them by rotating G' on an axis normal to its face. A difficulty, however, 

 existed in retaining the fringes with a flickering arc. It will be seen that in 

 this case the line LG moves over a small angle in all directions with the bright 

 spot on the positive carbon, so that the angle of incidence is varied, and with 

 it the angle of diffraction 6 at G. All this is magnified by reflection from the 

 miiTors. Moreover, unless the collimator lens is very near G, the illuminated 

 part or bright line on G is displaced right and left. Path-difference between 

 GnNG' and GmMG' is thus modified. If the faces of the mirrors are not all 

 quite in a vertical plane or parallel to the same plane, the up-and-down play 

 of the arc will mar the longitudinal coincidence of the two superposed spectra, 

 and hence the interferences will vanish. Thus they appear and disappear 

 periodically, depending on the accidental position of the bright spot of the 

 arc; and if this annoyance is to be avoided, sunlight or a steady light must 

 be used. The phenomenon and the spectra were not nearly so bright as 

 when observed with the transmitting grating, a result probably due both 

 to the additional reflections (particularly those at the grating) and to the 

 high dispersion. 



In other respects the behavior was the same as that described in Chapters 

 I and II, though the strip of fringes for reversed spectra seemed to be some- 

 what broader, probably owing to the increased dispersion and hence the greater 

 breadth of adequately homogeneous spectrum light. The linear phenome- 

 non, moreover, consisted of two or more black lines alternating with bright, 

 whereas a single black line was the characteristic feature above. When dif- 

 ferent strips of the grating G are used (the illumination should not be more 

 then 0.5 cm. wide), considerable fore-and-aft displacement at the mirror M 

 is necessary. 



The adjustment for crossed rays Mn and Nm, figure 57, is subject to new 

 conditions. In case of white light and a narrow slit, the dispersion produced 

 by G is at least partially annulled by G' instead of being incremented ; for the 

 change of the angle of incidence here compensates the changes of the angle 

 of diffraction. Thus if sini'v sin dv = \v/D for violet and sin^Y sin 6r = 

 \ T /D for red, and if sin iv=\v/D and sin ir = \ r /D, then sin =sin 9r = o. 



A sharp, white image of the slit may thus be seen for the reflection from each 

 mirror M and N, or the images may be colored if but a part of the spectrum 

 is reflected from M and N. The system of two gratings, G and G', tends to 

 become achromatic. It would seem to follow, therefore, that in general 

 homogeneous light and a wide slit would have to be used, but this introduces 

 additional annoyances, inasmuch as the transverse axes of the spectra (sodium 

 lines), which are to coincide, are not visible, but must be replaced inade- 

 quately by the edges of the slit. The experiment is thus (particularly in view 

 of the faint illumination seen in the telescope) difficult, and in a laboratory 

 not free from agitation, or in the absence of a good mercury lamp of intense 

 homogeneous light, it did not seem worth while to spend much time on it. 

 Moreover, a similar investigation will presently be made with a transmitting 

 grating. 



