Polarization of Sky Light. 



103 



used twenty, so that, a priori, the resulting error might be 

 expected to be much smaller in my case. Still I thought it 

 desirable to apply some test to see how great the error was. 

 So I prepared a diaphragm with an aperture about half the 

 apparent breadth of the Nicol, and took observations of the 

 sky with and without the diaphragm, with the five plates 

 inclined the same way as the three. The observations with 

 such a small aperture were very difficult ; but I satisfied my- 

 self that the change produced by the diaphragm, at Gr = 50° 

 did not exceed '2^ , and was probably less. With the actual 

 aperture of the Nicol the error is certainly less than 1°, 

 and when the two plates are inclined the other way the error 

 should be negligible. Thus the mean error should be less than 

 \°, and for G = 50° I have made no correction. For G=60° 

 it was impossible to determine the error in this way. But 

 some correction is probably necessary. So I have added '01 

 to the value of r, being partly guided by the value of r deduced 

 from FresneFs laws. 



In the following Table the first line gives the mean observed 

 value of ^, the second the co'-responding value of tan^^, i. e. r, 

 the third and fourth the greatest and least observed values of 

 tan^ <^, the fifth the value of r calculated from Fresnel's laws 

 on the assumption that /i = l'52^ and the sixth the value of r 

 finally adopted. The angle of incidence on the three plates 

 was found by a careful measure to be 47^°, and this angle was 

 used in the calculation. 



G 



0° 



20° 



30° 



40° 



50° 



60° 





(A 



38°-5 

 •630 





34°-9 

 •485 



30°-4 

 •345 



26°^3 

 •244 



22° 

 •163 



Mean observed r ... 



Greatest observed r . 



■669 





•489 



•368 



•260 



•175 



Least observed r ... 



•598 





•481 



•330 



•228 



•147 



Calculated r 



•603 

 •630 



•541 

 •571 



•457 



•485 



•370 

 •345 



•267 

 •244 



•188 

 •173 



Adopted r 





The intermediate values were obtained by interpolation, 

 using second differences where necessary. By calculating on 

 FresnePs tlieory the values of the ratios for three plates under 

 the incidences 47^° and 49|^° respectively, I found that to 

 make this Table apply to the readings at St. Moritz five per 

 cent, had to be subtracted from the values of r. 



The discrepancy between the observed values and those 

 calculated from Fresnel's laws seems to exceed probable errors 



