( 344 ) 



• 



light alternately travelled along different paths through that flame. 

 With these relative displacements of beam and flame the rays of the 

 anomalously dispersed light were much more bent, on account of the 

 uneven distribution of the sodium vapour, than tiie other rays of the 

 spectrum ; absorption and emission changed relatively little. The 

 result was, that the distribution of the light in the neighbourhood of 

 Dj and D^ could be made very strongly asymmetrical, which could 

 easily be explained in all details as the result of curvature of the 

 rays. The existence of "dispersion bands" was thus proved beyond 

 doubt. 



But the pure efl"ect of emission and absorption was not absolutely 

 constant in these experiments and concerning the density of the sodium 

 vapour in the different parts of the flame only conjectures could be 

 made. Moreover, the whirling ascent of the hot gases caused all rays, 

 also those which suffered no anomalous dispersion, sensibly to deviate 

 from the straight line, so that the phenomena were too complicate 

 and variable to show the effect of dispersion strictly separated from 

 that of emission and absorption. 



So our object was to obtain a mass of vapour as homogeneous as 

 possible and, besides, an arrangement that would allow us to bring 

 about arbitrarily, in this vapour, local differences of density in such 

 a manner, that the average density was not materially altered. The 

 absorbing power might then be regarded as constant. At the same 

 time it would be desirable to investigate the vapour at a relatively 

 low temperature, so that its emission spectrum had not to be 

 reckoned with. 



In a series of fine investigations on the refractive power and the 

 fluorescence of sodium vapour R. W. Wood^) caused the vapour to 

 be developed in an electrically heated vacuum tube. It appeared 

 possible, by adjusting the current, to keep the density of the vapour 

 very constant. Availing myself of this experience I made the following 

 arrangement for the investigation of dispersion bands. 



Apparatus. 



NN' (see fig. 1) is a nickel tube of 60 centimetres length, 5.5 cms. 

 diameter and 0,07 cm. thickness. Its middle part, having a length 

 of 30 cms., is placed inside an electrical furnace of Heraeus (pattern 

 E 3). Over its extremities covers are placed, the edges of which fit 

 into circular rims, soldered to the tube, and which consequently 



1) R W. Wood, Phil. Mag. [6], 3, p. 128; G, p. 362. 



