44 H. Rubens and B. W. Snow on the Refraction of 



A study of this curve shows that the dispersion in sylvite, 

 which in the visible spectrum is only slightly inferior to that 

 in rock-salt, decreases in a similar manner but far more rapidly 

 than in the latter mineral, so that at wave-length \=8/«6 the 

 dispersion is only about one third of the corresponding 

 dispersion in rock-salt. 



Notwithstanding the great durability of this material, and 

 its permanence in moist air as well as its almost perfect trans- 

 parency to thermal radiations, the exceedingly rapid decrease 

 in the dispersive power of sylvite renders this substance not 

 so well adapted for experiments involving the use of prisms 

 as rock-salt, whose surfaces are only with difficulty kept 

 perfect. In the construction of condensing lenses this 

 difficulty does not occur. 



Fluor it e. 



The prism here examined is the same one which was used 

 in the former investigation. The value of the refracting 

 angle was redetermined and was found to agree very closely 

 with the observations previously made. 



For a long time we tried in vain to measure beyond wave- 

 length \ = 3 # 5/u,, the energy-spectrum produced by the fluorite 

 prism. The results of the previous observations show the 

 cause of our failure to be due to the fact that, after a region 

 of comparatively feeble dispersion, the dispersive power of 

 fluorite increases and the energy in this portion of the spec- 

 trum becomes proportionally weaker. In order to make 

 further advances, we were finally compelled to open wider 

 the slit of the spectrometer at those places where the radiant 

 energy sank below a measurable quantity. The repetition of 

 this device enabled us to reach a wave-length in the infra-red 

 greater than X=8/a. In the curve shown in fig. 3, which 

 represents the observed distribution of energy produced by 

 the fluorite prism, the slit was twice opened ; the first time 

 from 0*1 miliim. to 0*4 millim. when the arm of the bolometer 

 was at a deviation a = 30° 10', and a second time from 0*4 

 millim. to l'O millim. at an angle of deviation a = 28° 50''. 

 By this means the deflexions of the galvanometer were 

 increased fourfold and tenfold respectively. Owing to the 

 greatly increased dispersion and the corresponding increase 

 in the breadth of the interference-bands, this change in the 

 width of the slit did not materially interfere with the sharp- 

 ness of the bands in this region of the spectrum. Inasmuch 

 as only one side of the slit was movable, a correction had to be 

 applied to the reading of the arm of the bolometer when the 

 slit was opened. 



The distribution of energy, as shown in fig. 3, gives a curve 

 whose character is wholly different from the representation of 



