( 349 ) 



case, liowever, there certainly exist currents in the mass of vapour 

 which cause the distribution of density to be less regular. Also when 

 A and B are at equal temperatures, we sometimes see the sodium 

 lines slightly broadened ; it stands to reason to attribute this also to 

 refraction in such accidental irregularities. 



That spectral lines possess some breadth is commonly ascribed 

 either to motion of the light-emitting molecules in the line of sight 

 or to changes in the vibrational period of the electrons by the col- 

 lisions of the molecules. We now have a third cause : anomalous 

 dispersion in the absorbing medium. The whole series of phenomena, 

 observed in our sodium tube, corroborates the opinion that this latter 

 cause must in general be regarded as by far the most important. It 

 will appear that this conclusion holds not only for dark but also for 

 bright spectral lines. 



If the slit in the diaphragm P is made much broader towards p', 

 this has no influence on the spectrum as long as A and B are at 

 the surrounding temperature. The Z)-lines appear as in «, PI. I. If 

 now A is cooled below this temperature, B raised above it, the 

 dark D-lines only broaden in the direction of the shorter wave- 

 lengths, while at the side of the longer wave-lengths the intensity of 

 the light is even increased, since now also anomalously bent rays 

 from the radiation field j)' can reach the point S through the slit Q. 

 (see j5, PI. I). The spectrum /9 passes into y when the temperature 

 difference between .1 and B is made to change its sign or also when the 

 original temperature difference is maintained and the slit in P is made 

 much broader towards j) instead of towards //. A small shifting of 

 the whole diaphragm P (starting from the condition in which it was 

 when taking /?) so that S falls exactly in the shadow, causes the 

 spectrum ö to appear, which makes the impression of an emission 

 spectrum of sodium with slightly shifted lines, although it is evidently 

 only due to rays from the field p' whicii have undergone anomalous 

 dispersion in the vapour. 



Let us now return to the diaphragm P with a narrow slit placed 

 on the optical axis. (A piece of glass coated with tinfoil in which 

 a slit was cut out, was generally used). The spectrum then shows 

 broad bands when there is a density gradient between A and B. 

 If beside the slit an opening is cut in the tinfoil, a group of rays 

 of definite refractivity (and consequently also of definite wave-lengths) 

 is given an opportunity to reach S through Q, and a bright spot is 

 formed in the dark band, the shape of which depends on the shape 

 of the opening in the tinfoil, but is by no means identical with it. 



23* 



