2 SMITHSONmN MISCELLANEOUS COLLECTIONS VOL. 52 



was fed by a two-mirror coelostat with silver-on-glass mirrors. The 

 solar rays were not concentrated on the slit. Hence they represented 

 the integrated rays of the entire solar disk. 



The infra-red solar energy spectrum was recorded on moving photo- 

 graphic plates 8 X 24 in. in surface. The clockwork was arranged 

 so that 4 cm. of plate corresponds to 5' of spectrum, and the plate 

 passed the recording light-spot of the galvanometer at the rate of 

 2 cm. in I minute. 



The three prisms were set according to computations so that the 

 beam of rays of wave-length 1.05/x would pass through each one 

 of the prisms approximately in minimum deviation. This same setting 

 was continued unchanged in all the observations. The total deviation 

 of the rays of this wave-length was roughly 180° and the dispersion 

 from A to n was about 5° 25'. Hence we were obliged to use five 

 61 -cm. plates to cover the entire region with overlap sufficient for 

 identification. 



Generally three curves of each of the five regions were impressed 

 on a single photographic plate. Care was taken to arrange them 

 vertically in close superposition, so as to facilitate comparison. Plates 

 I to 5 give reproductions of some of the most satisfactory observations. 



Linear scales are drawn on plates i to 5 parallel to the direction 

 of motion of the recording photographic plate. They have numbers 

 closely agreeing with those of the extensive table 3 of linear measures 

 and wave-lengths, given below. In each group of three curves the 

 air-mass of observation decreases as between the several curves from 

 the bottom upward and in each curve (except in pi. 5) from left 

 towards right. In most plates there is a very considerable increase 

 of air-mass between the upper and lower curves. This will facilitate 

 the discrimination, by those interested, of solar and telluric lines. De- 

 tails of times of observation and air-mass and notes on the conditions 

 are given in table 2. 



A very considerable increase of detail appeared in these energy 

 curves when compared with those taken 30 years ago with a single 

 glass prism. In the A line, for instance, not only could the doubles 

 be recognized, but in many of them the individual components were 

 resolved separately in the energy curve. Some of the bands near wave- 

 length 0.82/A showed as many as five veridical lines in the new curves 

 where only one broad band could be distinguished in the older work. 



The identification of lines was done entirely by Mr. Freeman, and 

 in the following manner. A series of several holographs was super- 

 posed, either on millimeter cross-section paper or on a comparator 



