DESCRIPTION OF APPARATUS AND METHODS. 29 



where one must shift back and forth in repeating any observations. 

 Since the CO2 band at 4.40 /m has been determined so accurately by mini- 

 mum deviation methods by Paschen^ and by Stewart/ the operation of 

 setting for minimum deviation with the small spectrometer was repeated 

 until this band came exactly at 4.40/*. With the large spectrometer this 

 band was found exactly at 4.40 /x for the minimum and non-minimum 

 deviation methods. The maximum is so sharp that it could be accu- 

 rately located to two or three seconds of arc, with the large spectro- 

 meter, and, since no difference could be detected in the position and in 

 sharpness of the large absorption bands, it appears that it makes no 

 difference which method is used.^ 



Ransohoff,^ assuming a lack of minimum deviation of 3' at y.y fx, com- 

 puted a change in deviation of 11", or 0.0012 /x, which agrees with the 

 theoretical work of Wadsworth.^ 



As an illustration of the difficulty in establishing the absolute value 

 of any absorption band, my experiences with the large spectrometer 

 may be cited. Thinking that a marked band of carbon tetrachloride, 

 found by Paschen at 6.45 fi, would be an excellent check on the present 

 work, it was examined and found to consist of two bands, of which his 

 is the mean value. 



THE CHEMICALS USED. 



In the beginning of this investigation it was never suspected that it 

 would grow to the present magnitude ; neither was the source of the 

 chemicals considered of the greatest importance. But with the increase 

 in my knowledge of the processes involved in manufacturing the chem- 

 icals, it became evident that the label " C. P." was of little significance, 

 and it is considered thus by chemists. Obviously it would have been 

 impracticable to purify all the chemicals used. The alternative was 

 to duplicate the work by examining the same compound manufactured 

 by different firms. In the " transmission curves," for many of the 

 compounds thus examined, the duplicate curves are given, and it will 

 be noticed that, disregarding their general appearance, due to the dif- 

 ference in the thickness of the films, which is of no consequence, rarely 

 IS there a difference in the spectra of the two samples. This means 

 that the two samples of a compound were equally pure, or what amounts 

 to the same thing, that they contained the same impurity if any was 



'Paschen, loc. cit. ^Stewart, loc. cit. 



^Helmholtz : Physiol. Optok., p, 260, shows that for purity of color bands and 

 sharpness of Frauenhofer lines it is not necessary that the rays are homocentric. 

 *Ransoho£f, loc. cit. 

 ^Wadsworth : Astrophysical Jour., 2, p. 264, 1894. 



