6o SMITHSONIAN MISCELLANEOUS COLLECTIONS VOL. 68 



This amount can also be computed by Fresnel's and by Bouguer's 

 formulae for reflection and absorption, respectively, from the known 

 coefiicients of refraction '(Rubens)^ and transmission (Merritt).* 

 Several computations made before the idea of inserting the quartz 

 in this branch had occurred, gave this loss as about 19 per cent. Con- 

 sidering the uncertainties due to the great impurity of the spectrum 

 used for computing this loss, this is in sufficient agreement with the 

 mean value of 15 per cent found from subsequent curves like a and a'. 



Region b'c'd'e'. — For the energy of prismatic deviations belonging 

 to the region b'c'd'e' quartz is opaque. The energy recorded must 

 therefore belong to wave-lengths for which quartz is transparent. 

 That is to say, the observed energy is scattered from and belongs to 

 region a. There must be reflected from the quartz surfaces some 

 of the energy scattered into the second region (namely, an amouni 

 corresponding to the difference in energy between the areas a and a'). 

 Therefore the curve b'c'd'e' does not represent all the energy scat- 

 tered here from a' but requires to be increased by the mean observed 

 ratio a/ a' (or 1.18) in order to represent the total field light from 

 region a'.^ But even this is not sufficient to give the total field light 

 in the long- wave parts of the spectrum for it gives only that coming 

 from region a and it will be directly seen that the scattered energy 

 of even longer wave-lengths than 4 /^ is appreciable. The somewhat 

 complex determination of the correction for stray light, not trans- 

 missible by quartz, will now be considered. 



The two regions a and b'c'd'e' corrected as just described and 

 reduced to a more convenient scale are reproduced in figure 18 as a 

 and bi, c^, the latter magnified respectively a thousand fold and 

 three thousand fold relative to region a. With close approximation 

 the first curve may be considered as the energy curve of the region 

 producing the scattering shown in the second. The energy curve 

 of the region a, from which the energy is scattered, seems so sym- 

 metrical that it appears probable that the distribution of scattered 

 energy from it, b^c^ , would not be materially altered if all its energy 

 should be concentrated at the center of the region. Let it be assumed 

 that the whole energy of region a is concentrated in a central strip 

 4' of spectrum (i cm. of plate) wide as indicated by the dotted lines 

 in figure 18. Let this central strip be joined to the curve b^c^ by 



^ Rubens, Annalen der Physik und Qiemie, 54, p. 476, 1895. 

 ^ Merritt, Annalen der Physik und Qiemie, 55, p. 459, 1895. 

 ' As stated above the mean loss observed is 15 per cent. The reciprocal of 

 the 85 per cent remaining is 1.18. 



