D O C U M E N T 2 4 0 M A R C H 1 9 2 6 2 4 9 In this case, too, a specific direction of incidence on the interference apparatus cor- responds to a specific direction of emission, only that the angle is reduced by the factor . Here, too, in theory, beams flared to any given width can be brought to interference. A diaphragm can be positioned in the plane of image B of K formed by l, which allows part of K to be picked out. This setup also enables one to perform the experiment first proposed by me,[10] by putting the 1/10-mm grating at B. Then one would have to try to glue lens L into the canal ray vessel but one could also glue in a glass pane and position the lens externally. This setup should also allow verification of the second-order Doppler effect re- quired by relativity theory, in that it is shown that the fringe shift does not depend exactly linearly on the canal ray velocity but that there is also a quadratic term in play.[11] Your determination of the canal ray velocity by comparison with the inten- sity at rest[12] should also work better this way. Better still for the relativity theory experiment would be to use a plane mirror S and to put it exactly perpendicular to the optical axis (autocollimation). If the tube were then adjusted so that the inter- ference patterns generated directly and through the mirror coincided exactly, then just the relativity effect would be left. It would amount to about 1/10 of a fringe width for your most rapid H rays. But now comes the most interesting thing: namely, a consequence demanded by the wave theory, of which it doubtful, however, if it is fulfilled. I ask for the inter- ferences provided by your setup, which differs from the one just discussed only in the positions of the lenses l and L. [13] canal ray [14] [15] telescope set to Here, two rays, σ1 and σ2, coming from one point of the canal ray beam, have dif- ferent colors, hence as a whole they cannot yield any clear interference. Also, the images of the wave planes emitted by a canal ray particle are not formed at the same location on the telescope’s image plane b. Nevertheless, according to the wave theory, clear interferences can form if, through a slight tilt of the mirrors s1 α′ α ----- f F --- = [13] [14] [15] 1 a -- - 1 b -+-- 1 f -=-- ∞