of the Solar Spectrum. 243 



This prism was an equilateral one, each side being 36 millims. 

 long. The prism from end to end was 44 millims. As it was 

 impossible to stick tinfoil upon this prism without injuring the 

 incident surface, a sheet of brass was employed instead, in which 

 a slit 3 millims. wide was cut, and which was placed close before 

 the incident surface of the rock-salt prism. 



In the following Table the results of two series of experiments 

 are given, each series having been performed in exactly the same 

 manner as that described in speaking of the glass prism. 



Taking the mean of the two numbers in each case, and mul- 

 tiplying it by a factor, which gives the product 10 for the red, we 

 have, — 



In the blue 3-7 



In the yellow 7*9 



In the red lO'O 



1" in the invisible .... 13*2 

 3'" „ „ .... 15-9 



4'" „ „ .... 13-2 



6'" „ „ .... 1-7 



The fact that the deflections for the visible portions of the 

 spectrum in the case of the rock-salt prism were smaller than 

 when the glass prism was employed, is clearly entirely owing to 

 the fact that the surfaces of the rock-salt could not be so per- 

 fectly ground and polished as those of glass, and, further, to the 

 existence of a slight turbidity in one part of the rock-salt prism. 



It is true that these numbers do not quite correspond with 

 those for glass, even in the visible portion of the spectrum ; 

 nevertheless the discrepancies, which certainly result wholly 

 from errors of observation, are not of a kind to justify us in 

 calling in question the identity of the curves of thermal intensity 

 in the visible portions of the spectra derived from the glass and 

 rock-salt prisms. (It is of course impossible here to take into 

 account the differences of partial dispersion for glass and rock- 

 salt.) Beyond the red boundary of the visible spectrum, how- 

 ever, the two curves of intensity separate widely from one 



