4oG Prof. C. Timiria/etf. [Apr. 30 r 



well-known method used by physicists for recomposing the prismatic 

 spectrum. We may thus combine the intensity of the prismatic spectrum 

 with the advantages of the diffraction spectrum, the gasometric results 

 being directly comparable. I used the very ingenious apparatus devised 

 by the late Jules Dubosq for the production of complementary colours. 

 It consists of a cylindrical lens and of a wedge-prism fixed on a plate 

 of glass. Instead of a spectrum we obtain two stripes of light of equal 

 area and of complementary colours. 



More recently I proposed a modification of this apparatus (skilfully 

 executed by M. Pellin) consisting of two glass plates with their respec- 

 tive prisms, which may be made to slide past one another and divide 

 the spectrum into three parts (see Plate 20, A, b). 



The glass tubes containing the leaves being immersed in mercury 

 (Plate 20, A, a), the spectrum must of course be divided in the vertical 

 sense, but in other cases it may be more convenient, as will be presently 

 seen, to adopt Kundt's method of dividing it in the horizontal sense 

 so that the two coloured surfaces may be put into direct contact 

 (Plate 20, B, c). As I have just said, this method combines the advan- 

 tages of the grating and of the prism, the difference of dispersion 

 being eliminated as in the diffraction spectrum, the intensity remaining 

 the same as in the prismatic. 



Moreover, in this last respect the method may present even an 

 advantage over the simple prismatic spectrum. If desirable the 

 spectrum used may be in a considerable degree impure (that is, 

 obtained by means of a broad slit), provided the stripes of light to 

 which the leaves are exposed are of a homogeneous colour and present 

 equal areas. We have but to place behind them a small Sorby- 

 Browning spectroscope (Plate 20, B, a) with a scale giving the wave- 

 lengths, and notwithstanding a certain degree of overlapping of the 

 portions of the spectrum corresponding to the coloured stripes, the 

 analytical results may be directly plotted on a normal spectrum scale. 

 Thus, thanks to this method, not only the error arising from the 

 difference of dispersion, but even if desirable a certain degree of 

 impurity of the spectrum, may be easily done away with and the 

 maximum intensity of light obtained. I may add that I still prefer 

 using a pure spectrum. 



The experiment undertaken to find the relative effect of the two 

 halves of the spectrum was conducted in the following manner. A 

 beam of light reflected by a large Foucault foeliostat (furnished by 

 M. Pellin) (fig. 9), and condensed on the slit by means of a lens of 

 25 cm. diameter, was decomposed by a direct-vision prism and recom- 

 posed into these two complementary bands of yellow and blue light. If 

 we admit that the limits of the effective rays in the visible spectrum 

 correspond to the wave-lengths of 700 and 400 millionths of a 

 millimetre, a line passing somewhere about the wave-length 550 



