Prof. Challis on a Theory of the Dispersion of Liglit. 275 



the double line D apply to the mean position between the com- 

 ponents. The refractive index corresponding to the measure 

 1989*5 has been altered conjecturally, the given value (1*63] 33) 

 havingbeen the same as that corresponding to the measure 2005*0. 



Respecting the numbers in the last four columns, it is to be 

 stated that they express actual lengths in millimetres multiplied 

 by 10 8 . It will hence be seen that the differences between the 

 calculated and observed values of A, are generally very small. The 

 larger differences occur so exceptionally that they must plainly be 

 referred to errors of the data from observation. This is especially 

 the case with respect to the rays whose measures by Kirchhoff 

 are 1135-0, 21 198, 24160, and 24365, and the ray designated 

 by the letter c Leaving out of account the discordant results 

 for the ray 1135*0, there seems to be a systematic variation be- 

 tween the calculated and observed wave-lengths in the part of 

 the spectrum from B to E, but not nearly in the same degree in 

 any other part. Also it is to be noticed that there is a close 

 agreement between the results from the two calculations, the 

 difference in no case exceeding 26, excepting in the first three 

 comparisons, for which the differences are respectively 117, 80, 

 and 40. This circumstance might be supposed to indicate a 

 discrepancy in the data for the rays B and C. 



In order to clear up this point, I went through for the seven 

 principal rays the same calculations as those by which the Table 

 above was constructed, only using, instead of Ditscheiner's values 



o 



of X, those given by Angstrom in his Recherches sur le Spectre 

 solaire, pp. 31 & 32. The results in the two preliminary calcu- 

 lations of the constants A', B', C were 



log A' = 1*0870469, logB' = 0*3399332, C' = 7*343192; 

 log A'= 1*0576341, logB' = 0*2657302, C' = 7*028368. 



The excesses of the calculated values of X resulted as follows : — 



Ray. 



i Angstrom's 



Excess of calculated wave-length. 



Former 



length. 



By first ( By second 



Mean. 



mean. 







calculation, calculation. 





B 



6S671 



+191 



+95 



+58 



C 



. 65621 



-131 



-65 



-40 



D 



58921 



-108 



- 67 



-87 



-89 



E 



. 52691 







- 4 



- 2 



- 5 



F 



. 48607 



+ 13 







+ 6 



+ 11 



G 



43073 







+ 3 



+ 2 



-12 



H 



39681 



- 32 



-16 



+ 8 



Hence it appears that the excesses for the rays B, C, D follow 



