APPENDIX VI. 133 



which burns vaporized petroleum oil with a strong air draft. This is 

 no doubt the CO2 emission band, as he himself thinks. Unfortunately, 

 he does not give this band, using the lamp zvithout the mantle, so we 

 do not know whether the shift from 4.4 to 4.6 /x is due entirely to rise 

 of temperature, as found by Paschen (loc. cit.), or whether it is due, in 

 part, to the difference in the dispersion curves, as found by him and 

 by Rubens. 



Langley (loc. cit., p. 219) has also determined the dispersion of 

 fluorite. " The method employed consisted simply in taking a number 

 of holographs with a fluorite prism, comparing these with holographs 

 taken with rock salt, picking out common absorption lines, measuring 

 the position of these lines in the fluorite holographs, and, after reduc- 

 tion of these measures, comparing their results with those obtained for 

 the salt." The indices are all smaller than those recently found by 

 Paschen,^ who compared the fluorite prism directly with a grating. 



Now, this is just what one would expect, viz, that Langley's indices 

 of fluorite are too small, if the wave-lengths of his absorption lines are 

 too large; and the wave-lengths of his absorption lines are too large 

 beyond 2 p., if his indices of refraction of rock salt are too large, as will 

 be noticed in fig. 132. For, reasoning from the fact that the index of 

 refraction decreases as the wave-length increases, if we find the disper- 

 sion of fluorite by simply "picking out common absorption bands" whose 

 true wave-lengths are smaller than those found by Langley by means of 

 the rock-salt prism, using his indices of refraction of rock salt, then it 

 necessarily follows that, if the wave-lengths of the absorption lines are 

 too large, the fluorite indices will be too small. • The whole depends 

 upon the question of the indices of refraction of rock salt. 



From the general appearance of the dispersion curve of rock salt, 

 using the indices found by Rubens, it appears that the value, n= 1.5255 

 for X = 2.35 ij. may possibly be too low. The dispersion curve may be 

 a straighter line at 2.35 /x. than is indicated in fig. 132. If it be a straight 

 line, then the index is n == 1.5256 for A = 2.35 /x, or the wave-length is 

 A = 2.39 /x for 7i = 1.5255. 



It is not for me to say which of these two investigators' indices are 

 the more nearly correct. It will be sufficient to add that it is quite 

 evident that the Langley values do not harmonize matters, while by 

 straightening the Rubens curve at 2.33 /x (using m= 1.5256) the dis- 

 crepancy at 2.95 fx is entirely obliterated without aflfecting the wave- 



^Paschen : Ann. der Physik, 4, p. 302, 1901. 



