340 Dr. J. Kerr's Experiments on the 



rature a little above 100° C, is laid along the remote edge of 

 the plate for a measured time, 3 to 6 seconds. The effect in 

 the refractor is as distinct and regular as possible, and indi- 

 cates a relative retardation of the pencil which is close to the 

 edge. Though the temperature of the source is higher than 

 in the former experiment, the effects are much less intense, 

 the range of displacement of the fringes being much smaller 

 (one third to two thirds of a fringe- width), and their changes 

 of inclination much less marked. To explain this decrease of 

 effect, we have to remember that the compression along the 

 length of the heated plate is much more intense at the heated 

 edge than at the opposite edge, and also much more concen- 

 trated, much more rapid in its rate of decrease from the lateral 

 surface inwards : we have to remember also that, in the old 

 experiment, there is a considerable inequality of temperatures, 

 (with some consequent inequality of indices), for the two 

 pencils. Upon the whole, this twefth proposition appears to 

 me to be proved experimentally beyond objection. 



Proposition XIII. 



To specify the wave-surface in directionally strained glass. 



18. As the line of strain is an axis of physical symmetry, 

 we may assume that the directions of the three principal elas- 

 ticities are respectively parallel and perpendicular to that line, 

 and that the elasticities in directions perpendicular to it are 

 equal. Hence we infer, by the principles of Fresnel's theory, 

 that the wave- surface in directionally strained glass is similar 

 to that in uniaxal crystals. From Brewster's property also 

 (art. 6) it follows that the wave-surface is of the negative 

 class (the spheroid oblate) in the case of compression, and of 

 the positive class (the spheroid prolate) in the case of tension. 

 The results of the preceding experiments carry us a step 

 further; they enable us to connect definitely the wave-surfaces 

 in the medium strained and unstrained. Our final hypothesis 

 may be stated in simple terms, by reference to the adjacent 

 figure, which contains five curves of the second degree, upon 

 rectangular axes A B and C D through their common centre. 

 The required sheets of wave-surface are represented by these 

 curves in principal section, that is, in plane central section 

 along the line of strain A B ; so that the sheets are generated 

 by revolution of the curves round A B. 



(1) Strain zero. The wave-surface for this case is not in 

 question, but is required as a term of comparison : it is repre- 

 sented, for a given time of propagation, by the mean curve in 

 the diagram, the primitive circle. 



