( 216 ) 



nothing is visible of (lie density gradients existing round the cooled 

 tube B, because light of this wave-length travels along straight lines 

 through the vapour. 



Proceeding to / 5870, we are still at such a distance from D % , 



\—n 



that the value of = R {n representing the index of retraction, 



A the density of the vapour) is moderate. Steep gradients of the 



density are required to make the rays deviate sufficiently for missing 



the slit in Q, and such gradients are only to be found very near 



the surface of the tube B. We therefore obtain the image ,?, in 



which B appears surrounded by a narrow dark region. 



The third photograph, y, has been made with ). 5877. For these 



1-n 

 waves the expression is greater than for ). 5870, so that smaller 



values of the gradient suffice to give to the rays a perceptible in- 

 curvation. The result is a broader dark region all round B l ). 



The photographs d and £ have been secured with the camera slit 

 on A 5881 and a 5885 respectively. This time the tube A has been 

 cooled instead of B. We see the dark "aureole" grow as the wave- 

 length we are using approaches %d., = 5890. Getting nearer still, the 

 whole field would finally become dark. 



Similar results are obtained if we approach D x from the side of 



n— 1 , 



the greater wave-lengths, thus using waves for which — — - has 



increasing values. 



By a slight change in the arrangement of our experiment we 

 may obtain the opposite effect, to wit, that merely rays, suffering 

 anomalous refraction, do enter the spectroheliograph, whereas the 

 normally refracted light is prevented from reaching the slit. We 

 have only to open the slit in P very wide, and to put a vertical 

 bar (a match for instance) in the middle of it, the image of which 

 now falls exactly on the slit in Q. Under these circumstances 

 light, issuing from the divided opening in P, can only be transmitted 

 by Q if it has been deflected in the vapour. 



In this way the photographs g, i\ and 6 were obtained, the second 

 slit being set on ;. 5884, ;. 5886, ;. 5888 respectively. If there had 

 been no density gradients, the whole field would have shown dark; 

 the bright regions, however, now prove the existence of the gradients. 

 When taking § and % the tube B, and when taking 6, the tube A 

 was cooled. 



!) In this image the lower right corner was cut off by a rubber tube accidentally 

 crossing the path of the beam. 



