38 H. Rubens and B. W. Snow on the Refraction oj 



I. Rock-salt. 



We had at our disposal a prism of this mineral having 

 a triangular base 3 J centim. on each side and 4 J centim. in 

 height. Before being used the prism was freshly polished 

 and its refracting angle redetermined. The observations 

 with the bolometer gave the energy spectrum represented in 

 fig. 1, PI. II. The positions of the maxima and minima 

 were corrected, as in the paper cited above, with the aid of 

 the enveloping curve, whereby the points of contact of the 

 two curves were used without further modification as the 

 characteristic points in question. As the theory shows, this 

 method gives a closer approximation to the quantities required 

 than the method by construction given in the former paper. 

 But little weight, however, is to be attached to the superiority 

 of this modification, as both methods lead to results which are 

 identical to the fourth decimal place. 



Inasmuch as the present investigation was undertaken ex- 

 pressly for the purpose of extending measurements as far as 

 possible into the infra-red, we were compelled to use a com- 

 paratively thick layer of air for reflecting the interfering 

 beams of light, which brought the interference-bands in the 

 spectrum very near together. It was quickly found, how- 

 ever, that even the narrow width of the bolometer and the 

 impurity of the spectrum, caused by the aberration of the 

 lenses, placed a limit beyond which the further reduction of 

 the breadth of the interference-bands could not be carried. 

 With the feeble dispersion of the materials used, this limit 

 was practically reached when the visible spectrum was crossed 

 by seven or eight interference-bands, which gave a value to 

 the constant K = 2d cos i of about 8 # 5/a. According to this, 

 the minimum of the first order, which is the farthest possible 

 attainable point in the infra-red, has a wave-length \=^8'5/jl. 

 Then follow the maximum of the second order and the corre- 

 sponding minimum, which have wave-lengths X = 5 , 7fi and 

 \ = 4'3 / u respectively. Although the curvature of the curve 

 of dispersion in this region is slight, it seems to us, neverthe- 

 less, desirable to add, for greater accuracy in our measure- 

 ments, other possible points to the small number already 

 obtained. In order to attain this end, we found it advan- 

 tageous to use not only the corrected positions of the maxima 

 and minima and their corresponding wave-lengths for plotting 

 the curve of dispersion, but also the points of intersection of 

 the energy curve G=/(«) (see fig. 1) with the curve of 

 mean energy K=/(a)'^ since the wave-lengths corresponding 

 to the abscissae of these points are easily calculated. This 



