360 POLARISATION. 



of the previous one, is given by combination 13. The margin 

 and centre in the horizontal cylinder here appear neutral ; the 

 former because an optic axis is perpendicular, the latter because 

 the two superposed ellipses intersect at right angles. With the dis- 

 tance from the middle line, however, these ellipses take up a more 

 upright position ; they now intersect acutely, and therefore give 

 rise to an effect which corresponds to the position I upon the 

 middle line. Towards the margin, however, this action becomes 

 gradually eliminated again, because the difference between the 

 major and the minor axis of the ellipses becomes ever smaller, 

 and at the margin is nil. We consequently get two stripes 

 illuminated by interference colours on either side of the middle 

 line, while the latter and the two margins act neutrally. 



If we now endeavour to reduce, where possible, the alternatives 

 which figure in the above table to a smaller number, the first 

 question is whether the direction of the optic axes may perhaps be 

 experimentally determined. Considered theoretically, the matter 

 is unusually simple, as is readily seen. It is only necessary to 

 find out, by inclination of the section to different sides, the two 

 positions in which the object acts as a single-refracting medium 

 that is, in which the optic axes are exactly perpendicular. Simi- 

 larly, on inclination of the horizontal cylinder, we shall discover 

 whether the optic axes lie in a tangential plane or not ; for, where 

 the former is the case, the margins, with a definite inclination 

 which corresponds to the perpendicular position of an optic axis, 

 must necessarily act neutrally. 



Practically, however, these rules have only a veiy subordinate 

 value, for the simple reason that observation is not possible with 

 the accuracy that we might d priori expect. Transverse sections 

 cannot in most cases be employed, and even marginal views fre- 

 quently enough give a very doubtful effect. Beyond this, the latter 

 afford only under the most favourable relations the data which are 

 requisite for the determination of the position of the axes. For 

 since the inclination of the ellipsoid of elasticity that is to say, 

 of its longitudinal axis L to the axis of the cylinder is in general 

 unknown, the positive or negative character of the elements of 

 the cylinder remains doubtful, inasmuch as the angle which 

 the optic axes together form cannot be approximately measured, 

 nor the presence of a single optic axis testified. The few 

 rules which may be of practical value, with regard to the 



