DISPERSION IN ARTIFICIAL DOUBLE REFRACTION. 297 



There arc indications, however, in this glass, of a systematic dip at 5500, and a 

 subsequent rise. Various glasses also show signs of a peak in the red, between 

 6200 and 6500. None of these, however, are more than mere indications, and it is 

 only the curves for 3453, and in particular fig. 10, on which any safe deductions and 

 measurements can be based. 



If we refer to the table of 14 we see that 3453 and 3296 are very much alike in 

 chemical composition. Apart from this, no relation between this effect and chemical 

 composition can be predicated. It seems almost certain that boric acid has nothing 

 to do with it. The glasses richest in BjO 3 do not show the effect. K,O can hardly 

 be the explanation, or 3749 should show the effect more strongly. It seems not at 

 all unlikely that a small impurity, such as magnesium or zinc oxide, may be the 

 cause of the result. It is noteworthy that the only glasses which seem to show the 

 effect at all definitely, are precisely those which contain MgO and ZnO, and that the 

 one which really shows the effect in a measurable manner contains quite a respectable 

 percentage of MgO. 



21. Possible Explanation by Absorption Bands. 



The shape of the curve of residuals resembles the curve of index of refraction 

 plotted to either period or wave-length when we pass through an absorption band. 

 This suggests that the effect may be due to some faint or latent absorption band of 

 the glass in the visible spectrum, which band corresponds to a period active in 

 producing the artificial double-refraction. 



Following, as before, DRUDE ('Theory of Optics,' cap. V.), we have, ft being the 

 index of refraction, and K the co-efficient of absorption, 



p. 3 (1 ix) 3 = terms not depending upon the absorption band 



1- , + Q/{1-M/X-(W} ...... ..- (23), 



where X,, is the wave-length of the absorption band, a is a coefficient which increases 

 with the absorption, and Q is a coefficient depending on the arrangement and number 

 of the electrons. 

 This leads to 



{i- 



Now suppose the stress T to leave X, and a unaltered, and to alter Q. 



r = 2 



= (a 



whence, eliminating rf/f/c/T, 



)}'+A r l - (24). 



VOL. OCVII. A. 2 Q 



