( 352 ) 



but since for rarefied gases n differs little from unity, even for the 

 anomalously dispersed rays which we consider, Rt\ may be neglected 

 with regard to 1 and we may write 



o^-liK ^"> 



R — 



ds 



For every kind of light q is consequently inversely proportional 

 to the density gradient of the vapour in the direction perpendicular 

 to that of propagation. 



An estimate of the magnitude of the density gradient existing, in 

 our experiments, between A and B, may be obtained in two ways. 

 It may namely be inferred from the produced difference of tempe- 

 rature, or from formula (2). 



The temperature difference between A and B would have been 



pretty easy to determine thermo-electrically ; up to the present, 



however, I had no opportunity to make the necessary arrangement. 



Besides, the relation between the density distribution in the space, 



passed by the rays, and the temperatures of A and B cannot be so 



very simple, since we have to deal not with two parallel planes but 



with tubes, from which moreover hang many drops of liquid sodium. 



dL ^ 

 The second method at once gives an average value of — for 



ds 



n — i 

 the space passed by the rays. It requires a knowledge of R = 



for a kind of ray for which in our experiments also q has been 

 determined. 



Now Wood (Phil. Mag. [6], 8, p. 319) gives a table for the values 

 of n for rays from the immediate vicinity of the Z)-lines. These data, 

 however, refer to saturated sodium vapour of 644°; but we may 

 deduce from them the values of n for vapour of 390^ by means of 

 the table which he gives in his paper on page 317. 



For, when we heat from 389° to 508\ the refractive power of the 



vapour (measured by the number of passing interference fringes of 



98 

 helium light X = 5875) becomes — = 11 times greater, and at fur- 



50 

 ther heating from 508° to 644" again — ^ 12,5 times greater (now 



found by interference measurement with light fi-om the mercury line 

 A = 5461) ; hence from 390^ to 644° the refractive power increases 

 in ratio of 1 to 11 X 12,5 = 137. 



Since now for rays, situated at 0,4 ANGSTRÖM-unit from the D- 



