﻿fowle] 



ABSORPTION OF WATER VAPOR 



sold) terrestrial, the two lines thus determined should intersect on 

 the line of zero atmospheres. 1 



This condition is probably as well fulfilled as can be expected, con- 

 sidering the impurity of the spectrum. Probably the upper line of 

 each pair suffers little in its accuracy from this impurity; but as to 

 the lower line— the small dispersion, the slit width, the bolometer 

 width, the time of swing of the galvanometer needle, the speed with 

 which the spectrum is passed over the bolometer,- and the composite 

 character of the lines all tend to make the deflections too great. 

 Figure 2, plate 1, shows the composite character of 'l\ as shown in 

 detail holograms of 1898; and this may be compared with the same 

 band in figure 1 under the present conditions. Moreover the greater 

 the absorption in the band, the more it is influenced by these condi- 

 tions, and all tend to make the holographic ordinate at the bottom too 

 great, and to cause the lower line of each pair in the plots to reach 

 the line of zero atmospheres below the upper line. For days and 

 lines of very great absorption, the deflection becomes so small in 

 some of the greater bands that the error from the uncertainty of the 

 zero is an important additional factor. 



However, despite these limitations, as shown in plates 11 and in, 

 extrapolations to without the earth's atmosphere have filled up the 

 hands as follows : 



Table III 



1 The absorption, however, may not be solely terrestrial on account of the 

 possible superposition of some solar line as previously indicated. The depth 

 of some solar lines is quite great. The ratio of the energy at the bottom of 

 the C line to that indicated by a smooth curve across its top. is certainly less 

 than 0.70. For the H and K lines it is certainly less than 0.15. 



2 The difference of deviation from A tow 2 was 130'; the slit width sub- 

 tended from 10" to 15", the bolometer width 16"; the spectrum passed over 

 the bolometer at a speed of 20' in one minute of time; and the time of a 

 single swing of the galvanometer needle was about \Y 2 seconds. No correc- 

 tion has been made for these. 



3 See also similar data in columns five and six of Table IV. 



