OF ARTS AND SCIENCES. 25 



in each series, the letter a is attached to all the columns measured with 

 the polarimeter, b to those measured on the optical circle with 

 Babinet's wedges, and c to those in which the point of disappearance 

 of Savart's bands was found. 



It will be noticed that no observations are given of the polarization 

 of a beam transmitted by one surface of glass. There seemed to be no 

 easy method of measuring this quantity. It might be done by making 

 a series of prisms of such angles that when the light was incident on 

 one face at 10°, 20°, 30°, &c., the refracted ray would strike normally 

 on the second face. The effect of the latter would then be nothing, so 

 that the polarization would in this case be entirely due to the first 

 surface. 



We proceed now to discuss the results given in Table XIII. Exam- 

 ining the observations on a single surface of glass, we see that their 

 concordance is much greater than in the observations of sky polariza- 

 tion, although two quite distinct methods were employed. The column 

 headed " Theor." gives the theoretical polarization as given in Table VI. 

 The observed polarization is somewhat too great for angles less than 

 57°, and too small for greater angles. The difference may be explained 

 by the fact that the index of refraction was somewhat less than 1.55, 

 hence the angle of total polarization less than 57°. The same results 

 are shown in Fig. 6, in which abscissas represent angles of incidence, 

 and ordinates polarization. In the case of the light reflected by two 

 surf.ices. Fig. 7, the agreement is also very close. The same may be 

 said for the refracted beam for angles below 80°. It is difficult to 

 observe the polarization at greater angles, as the light then passes so 

 obliquely through the glass. There seems, however, to be a decided 

 excess of the observed over the theoretical polarization. Two series 

 only, of observations on the light reflected by four plates of glass, 

 Fig. 8, were taken, as they seem to agree sufficiently well with each 

 other, and with theory. The case of the light transmitted by four plates 

 of glass has special importance from its application to the polarimeter 

 of Arago. For although, of course, any other number of plates might 

 be used, yet this number, since the eclipse of 1871, seems to have been 

 more frequently employed. Six concordant series are given, obtained 

 by three distinct methods, and all agree in showing a marked divergence 

 from theory for angles greater than 60°, the observed being greater 

 than the computed polarization. 



The observations on twenty surfaces, as might be expected, present 

 still greater discrepancies. As regards the reflected ray, Fig. 9, series 

 37 b agrees pretty well with theory, but gives a much less result than 



VOL. 1. 4 



