DISPERSION IN ARTIFICIAL DOUBLE REFRACTION. 303 



If we refer to the diagram published in a paper previously referred to (' Camb. 

 Phil. Soc. Proc.,' vol. XII., p. 335) showing dependence of C upon percentage of PbO, 

 we see that until this percentage reaches about 40, PbO has little influence on the 

 stress-optical effect. 



We are therefore to look at the two remaining principal constituents, Bj0 3 and 

 K 8 O, for the cause of variations in GO- 



Looking at Table VI., in which the glasses are arranged according to descending 

 order of magnitude of C , we see that the four glasses with the high C u all contain a 

 percentage of B,0 3 which is considerable. On the other hand, the three glasses with 

 a low C all contain a comparatively high percentage of K a O. We conclude that 

 either BjO 3 raises C , or K a O lowers it, or both. Also, looking at 2783, which is the 

 lowest of the seven, we notice that it contains the least percentage of BjO 3 , but not 

 the highest percentage of K,O. This suggests that BjO 3 is more efficient in raising 

 GO than K,O in lowering it. That the effect of B,O 3 must be predominant is other- 

 wise evident from the fact that the order of the percentages of B,0 3 is the same as 

 the order of magnitude of C , with one exception (1809). 



The last column of Table VI. shows the values of (percentage of BjO s ) (percentage 

 of K a O). This places 1809 in the sequence, but throws out 3749. Also the marked 

 difference between the four first and the three last is shown very clearly. 



The glasses do not form a sufficiently regular series to enable us to go further and 

 to determine exactly the law of dependence of G u upon the percentages of BjO 3 and 

 of K S O in the glass. But the increase of B.,O 3 certainly increases C, and the 

 increase of K 2 O probably decreases it; and Bj0 3 seems to be at least twice as 

 efficient as K a O. 



With regard to the mean values of AO, they all appear to be of the same order of 

 magnitude. At all events no definite dependence of AO upon the composition can be 

 traced. For percentages of B,O 3 not exceeding 35 and of K a O not exceeding 25 it 

 would seem that Ay is independent of the composition of the glass or that the 

 dispersion is, in every case, proportional to the double refraction. 



24. Failure of HOOKE'S Law for Glass 2783. 



Before concluding, the peculiar phenomena shown by glass 2783 require explanation. 



This glass showed a progressive increase with increase of load in rAA/AW r for 

 both tension and pressure. 



Also when the load was increased the band, which was straight and vertical for 

 moderate loads, became curved, being convex towards the red as shown in fig. 12. 

 The load was eventually increased to about 59 kilogrammes when one of the glasses 

 broke. The baud was observed when the glass was on the point of rupture and it 

 exhibited a decided V shape as drawn. 



