652 PROCEEDINGS OP THE AMERICAN ACADEMY. 



account of the absorption beyond wave-length 4180, a point near the 

 limit of the violet. 



As this sharply bounded region of total absorption is approached, 

 the dispersion varies rapidly. According to the law suggested by Kerr, 

 the double refraction should increase as the inverse square-root of the 

 wave-length. This curve is an equilateral hyperbola whose axes are 

 the axes of co-ordinates. It is shown dotted in Figure 1. Each hyper- 

 bola was calculated by starting with the observed value of the double 

 refraction for the given temperature at wave-length 5890. But the 

 curve of double refraction observed in the present experiments is much 

 steeper than the hyperbola suggested by Kerr, as the region of total 

 absorption is approached. This raises the question whether the approach 

 to an absorption band may not be the controlling factor rather than the 

 change of wave-length. 



The possible effect of an absorption band is briefly considered mathe- 

 matically by N^culcea, 15 in his summary of the mathematical papers on 

 the general phenomenon of electric double refraction by Pockels, 10 and 

 Voigt.l7 



Method. 



A means of measuring double refraction, and at the same time of 

 taking into account the dispersion, is furnished by Babinet's com- 

 pensator. 



When this is viewed by plane-polarized light and an analyzing Nicol, 

 both properly oriented, a symmetrical system of straight dark fringes 

 appears, of which the central one is achromatic. If a doubly refracting 

 body, also properly oriented, be introduced between the compensator 

 and the source of light, the fringes are shifted through a distance x, 

 proportional to the double refraction. By analyzing the dark central 

 fringe with a spectroscope whose slit cuts the fringe at right angles, 

 a spectrum is formed traversed by a dark band, and any variation in x 

 with variation in color is made evident. If the dispersion of double re- 

 fraction in the material under observation be the same as that of the 

 compensator, namely of quartz, the straight dark band in the spectrum 

 will shift parallel to itself. If the dispersion be different from that of 

 quartz, the shifted band will be a curve plotting the difference. By the 



15 Le Phenomene de Kerr, p. 60. (Par E. Neculcea ; C. Naud, Editeur ; Paris, 

 1902.) 



16 Gi'.tt. Nachr. Math. Pliys. Klasse, 2, 102 (1896). 



" Wied. Ann., 69, 297 (1899); Ann. d. Pliys. (Jan., 1901). 



