376 Messrs. du Bois and Rubens on Refraction 



deviation, as remarked above. In the plane of the spectro- 

 meter's slit a small spectrum from an arc lamp was projected. 

 By slightly turning the collimator the slit could be adjusted 

 to given positions in the spectrum. For this purpose the 

 divided circle had been calibrated by observations with the 

 spectral lines of Li, Na, and H, and the readings correspond- 

 ing to the wave-lengths of Li a, D, F, and Gr noted once for 

 all. The refractive index was calculated by equation (5) (§10). 

 We begin by giving two full sets of observations in Table 

 III., which explains itself. We worked with three prisms of 

 each metal, in order to get our results freed as much as 

 possible from the individuality of the prisms and from other 

 sources of error. 



§ 18. The results are given in Table IV. ; the numbers in 

 prominent type are the arithmetic means of the indices ob- 

 tained for the three prisms. Beneath these we moreover give 

 figures headed "means B ;" these were calculated by giving 

 different weight to the indices derived from different prisms 

 proportional to the refracting angle of the latter. As the num- 

 bers B show no systematic differences against A, the results are 

 really independent of the refracting angle (compare A, p. 264). 

 The means B are of no further use, as the observations with 

 the steeper prisms are for various reasons not more trustworthy 

 than those with the flatter ones. The indices for " red " 

 light are those obtained above (Table II.) ; they correspond 

 to a wave-length of about 64 , 4 x 10 -6 centim. * 



§ 19. Discussion. — A comparison of some of our numbers 

 with those found by Prof. Kundt for Ni and Fe with light 

 only approximately homogeneous (A, p. 266) leads to the 

 following remarks : — 



1. For Ni the coincidence is sufficiently near, considering 

 the preliminary character of Prof. Kundt's measurements and 

 the uncertainty about the wave-length which has to be 

 attributed to his " blue " light ; 



2. For Fe deviations appear, rather insignificant at the blue 

 end but increasing towards the red. 



It is of course at present impossible to decide whether these 

 deviations may be explained by Kundt's prisms not having 

 been free from oxide (lesser index, normal dispersion A, p. 267) 

 or by his electrolytic iron baths not having been chemically 

 pure. Considering the improvements since introduced into 

 the method, mainly by Prof. Kundt himself, we believe our 



* du Bois, Wied. Ann. xxxi. p. 956 (1887). 



