DISPERSION IN ARTIFICIAL DOUBLE REFRACTION. 301 



showing in its unstressed condition strong absorption bands in the visible spectrum 

 would exhibit this effect in a most marked manner. A small slab of didymium glasp 

 was obtained and examined under direct pressure with the apparatus described in 

 the previous paper in ' Camb. Phil. Soc. Proc.,' already referred to several times. The 

 ol)servation8 were not very precise, and it is hoped to repeat them with the flexure 

 apparatus, when larger slal can be obtained.* Such as they were, however, they 

 gave a negative result. The l>and due to stress passed through the double absorption 

 band in the orange without showing any marked irregularity. 



It seems, therefore, that the didymium electrons which produce the alworption 

 band are not affected by the stress in the way described. It appears possible that 

 the didymium particles really float about, as it were, in the glass, in a free state, like 

 particles in suspension in a fluid, and that they cannot be influenced to any great 

 extent by stress applied to the glass. Further research in this direction is in 

 progress. 



22. Determination of Absolute Values of C. 



Although the experiments were primarily undertaken to show the dispersion 

 effects, it is desirable to know also the absolute values of the stress-optical coefficients. 

 These are not given by the experiments as described in 2, because the differences of 

 level i Zj and z^h cannot be measured with sufficient accuracy. 



To determine absolute values a second slit is used, of which the height is h + b.h. 

 The two slits were cut in the same diaphragm, so that A/4 is easily measured once 

 for all. 



Referring to formula (7), the relative retardation due to the second slit, 



where C', = stress-optical coefficient of beam F for the wave-length for which the 

 retardation is R'. 

 Also 



R = (3M,C/2V)[>i-z,+<r(zi-A)] ....... (30). 



Hence 



RVC^-R/C^ -3M/TA///2V ....... (31). 







If therefore we know the ratio CX, : C 3 we can find the alolute values of either. 

 If R', R correspond to kinds of n lb order, 



R = nX, R' = nX'. 

 Therefore 



nfX'/C'.-X/C,) = -3M,rAA/26, a . 



[*JVofe added April 3, 1907. Since writing the above the experiment has Iwen repeated under the more 

 accurate conditions, and the negative result has been confirmed. If the effect exists in the didymium 

 glass it is certainly small.] 



