ATMOSPHERIC ATTEXUATIOX OF ULTRA-VIOLET LIGETT. 367 



and as the effect of scattering may well be important, these relative 

 intensities were plotted with the computed curves for molecular 

 scattering. The details of the procedure follow. 



Procedure. — Over a hundred photographs of the ultra-violet spec- 

 trum of a Cd spark were taken between January 191G and November 

 1917, at varying distances from the source. Nearly all the work was 

 done on the clearest nights so as to avoid the long exposures caused by 

 haze. The spark between Cd terminals in parallel with a suitable 

 capacity was produced by a half kw. transformer on a 60 cycle, 110 

 volts circuit. The apparatus for photographing the spectra con- 

 sisted of an objective prism arrangement, — a box containing a large 

 60° quartz prism, a quartz lens about S cm. in diameter, of approxi- 

 mately one meter focal length, mounted directly back of the prism, 

 and a holder for the photographic plate. An open tube about 60 cm. 

 long, 10 cm. in diameter was mounted in front of the prism and 

 several diaphragms were inserted along the path of the light to prevent 

 stray radiation from reaching the photographic plate. A small 

 telescope was attached permanently to the box to aid in getting the 

 apparatus in line with the distant source. The adjustment, once 

 made, was not changed until a number of photographs had been taken. 



The images are recorded as small spots as shown in the accompany- 

 ing photographs. (1) shows the spectrum of the Cd spark X3612- 

 X2573 through 160 meters of air obtained by a 10 seconds exposure. 

 (2), (3), and (4) were all taken on the same plate on an exceptionally 

 clear night through 2300 meters of air, the exposure being 15, 30 and 

 60 minutes respectively. A number of exposures were then made 

 through an air path of SOOO meters. On several plates of 4 hours 

 exposure X2573 was recorded through this long path of air but the 

 images were very faint. Strutt has already published photographs 

 of mercury vapor spectrum through about 6500 meters which show the 

 same result. 



Although many photographs werQ. taken, only a few nights in a 

 year are really satisfactory for this work. A short rain followed bv a 

 dry Northwest wind seemed to leave the air in the best condition. 



A large number of exposures were made in the laboratory also with 

 an air path of two or three meters. The same plate, Seed 30, emulsion 

 15511, and the same development were used throughout the work. 

 By using a diaphragm in the collimator the exposure times as well as 

 the opacities given in (2), (3) and (4) were approximately reproduced. 

 In the case of plates for the five mile air path, it was found that the less 

 refrangible end of the spectrum was over-exposed so badly in order to 



