( 353 ) 



lines ^) we have n — 1 = ± 0.36, (as the average of three values 

 taken from Wood's table on page 319), we ought to have with 

 sodium vapour at 390^ for the same kind of rays 



0.36 



n—\ = = 0.0026. 



137 



The density A at 390° is, according Jewett, 0.0000016, hence 



n~\ 0.0026 



R = — — = = 1600. 



A 0.0000016 



Then from formula (2) follows 



dA 1 1 



— = — = = 0,0000002. 



ds Rq 1600 X 3000 



Dispersion bands in the spectra of terrestrial sources. 



It is very probable that, when metals evaporate in the electric 

 arc, values of the density gradient are found in the neighbourhood 

 of the carbons that are more than a thousand times greater than the 

 feeble density gradient in our tube with rarefied sodium vapour *). 



The radius of cur\ature will, therefore, in these cases be over a 

 thousand times smaller than 30 meters and so may be no more 

 than a few centimetres or even less. A short path through the vapour 

 mass is then already sufticient to alter the direction of certain rays 

 very perceptibly. 



If now an image of the carbon points is produced on the slit of 

 a spectroscope, then this is a inire image only as far as it is formed 

 by rays that have been little refracted in the arc, but the rays which 

 undergo anomalous dispersion do not contribute to it. Light of this 

 latter kind, coming from the crater, may be lacking in the image 

 of the crater and on the other hand penetrate the slit between the 

 images of the carbon points. Thus in ordinary spectroscopic obser- 

 vations, not only broadening of absorption lines, but also of emission 

 lines, must often to a considerable extent be attributed to anomalous 

 dispersion. 



1) The spectrum e in our plate shows that the extremities of the peaks corre- 

 spond pretty well to light of this wave-length; for they approach the D-lines to 

 a distance which certainly is no more than V15 of the distance of the D-lines 

 which amounts to G Angstr. units. For these rays the opening of the diaphragm 

 was 1 cm. distant from the optical axis. 



') If we e. g. put the vapour density of the metal in the crater, where it boils, 

 at 0.001, the density of the vapour outside the arc at a distance of 1 cm. from 

 the crater, at Ü.0001, then we have already an average gradient 5000 times as 

 large as that used in our experiments. 



