of the double refraction in strained glass. 57 



ficiently homogeneous. A rather delicate preliminary adjustment 

 was therefore necessitated, the glass being shifted slightly several 

 times until, on examining in the ordinary way between crossed 

 Nicols, a fairly large region in the centre of the block showed 

 a uniform tint when the block was loaded, even when the tint of 

 passage was reached. The aperture of the pencil which was to be 

 examined was then narrowed by means of a diaphragm so that the 

 rays were restricted to pass through the region of the block where 

 the tint had been found to be uniform. The straining apparatus 

 was then left untouched, save for the putting on of the weights, 

 and these were applied with great care, so as to avoid any jar. 



Finally by placing the crossed Nicols horizontal and vertical 

 and then loading the block, it was found that, over the region of 

 uniform tint, the field remained very sensibly dark. 



It was assumed therefore that throughout this region, the 

 stress should be taken vertical and, at all events in planes perpen- 

 dicular to the paths of the rays, uniform. It is easy to see that 

 variations in intensity (but not in direction) of the stress along 

 the paths of the rays will not affect the results, the total effect 

 being the sum of the effects of elementary laminae perpendicular 

 to the path of the light, and depending only upon the mean stress. 



3. When the block B is unloaded, no light is visible in the 

 telescope T if the Nicols A and N be crossed. But when the 

 block B is under vertical pressure, the light breaks up into two 

 pencils, one horizontally and the other vertically polarized, which 

 suffer a difference of retardation in passing through the block B. 

 The light is now no longer quenched and a spectrum appears in 

 the field of T. This spectrum consisted of two continuous bands 

 and a bright line spectrum in between, being in fact the spectra of 

 the two carbons and of the arc. 



The difference of retardation of the two rays passing through 

 the block B is (see first paper loc. cit. p. 481) 



r = GtT, 



C being what I have called the stress-optical coefficient, t the 

 thickness of the glass, T the mean stress. 



If a, b, c be the breadth, thickness and height of the block B, 



CP 



.'. r = — . 

 a 



Now when this difference of retardation is equal to an integral 

 multiple of the wave-length, light of that particular wave-length 

 is again quenched. 



The spectrum should therefore be crossed by a dark band 



