Polarization of Sky Light. 101 



surface of the quartz, and a somewhat smaller proportion is lost 

 of that polarized in the principal plane than of that polarized 

 in the perpendicular plane ; but the difference is very small. 

 The former proportion is about 0*0460 and the latter 0"0472. 

 These are for a single reflexion. Therefore, owing to the 

 double reflexion, the tan^ </> should be multiplied or divided by 

 the factor 1'0025. This correction is quite negligible in such 

 ■work as the present. 



My standardizer consisted of a wooden framework to hold 

 the quartz plate, with a round hole over the quartz just large 

 enough to let the brass tube of the Nicol turn freely. To 

 the Nicol were attached two pointers for taking the readings 

 on opposite sides of a carefully graduated circle. The frame- 

 work w^as screwed onto the polarimeter, care being taken to 

 set the principal plane of the quartz as nearly as possible 

 parallel to the plane of rotation of the glass plates. The outer 

 end of the Nicol was closed with a glass cap to keep out dust; 

 but between the Nicol and the quartz or the quartz and the 

 pile of plates no glass intervened. The quartz plate was 

 nearly an inch thick. It was one of the pieces (plate iii.) I 

 used in a research on the wave-surface of quartz (Phil. Trans, 

 part i., 1886). 



The readings were taken by setting the pile at a known 

 angle of incidence and then adjusting the standardizer till the 

 polarization was neutralized. This could be done in four 

 difierent positions of the Nicol, corresponding to the four 

 quadrants, since tan^ ^ attains any given positive value 

 four times while ^ is changing from 0° to 360°. I gradu- 

 ated my circle suitably, two opposite points being called 0°, 

 the two rectangular points 90", and four intermediate points 

 45°, Averaging the readings in adjacent quadrants eliminated 

 any error in the position of the zero-point, while averaging 

 the readings in opposite quadrants eliminated, at least approxi- 

 mately, the error arising from the axis of rotation of the Nicol 

 not being truly parallel to the incident light*. I did not take 

 any large number of readings, as it seemed not worth while 

 to aim at any high degree of accuracy in standardizing obser- 

 vations of such a variable character as those I have given. 

 Moreover, in optical observations there is always some risk of 

 the errors being systematic. In the following Tables are given 

 the means of the readings in the four quadrants for different 

 values of the angle of incidence Gr on the five plates. The 

 upper line gives the readings when the plates were tilted to 

 the right, i. e, the same way as the fixed plates, the lower 



* See a paper on the use of Nicol's prism, Phil. Mag. May 1885. 



