xxi.] EXPERIMENTS. 293 



abundance of dense vapour which conies over at a higher 

 temperature with a spectrum of a certain kind. 



Now in our experiment we deal not with a compound body in 

 the ordinary sense, but with a so-called elementary body. 

 The question is, Will this make any difference ; or, rather, shall 

 we get similar differences ? 



FIG. 98. Hypothetical spectra obtained on distilling a successively increasing 

 temperatures a mixture of light and heavy hydrocarbons. 



The experimental work has followed two distinct lines. I 

 shall refer in detail to the results obtained along each. 



Experiments based on the First Consideration. 



The first method of investigation adopted consists in 

 volatilising those substances which give us flame spectra in a 

 Bunsen flame and passing a strong spark through the flame, first 

 during the process of volatilisation, and then after the tempera- 

 ture of the flame has produced all the simplification it is capable 

 of producing. 



The results have been very striking; the puzzles which a 

 comparison of flame spectra and the Fraunhofer lines has set us 

 find, I think, a solution ; while the genesis of spectra is made 

 much more clear. 1 



To take an instance, the flame spectrum of sodium gives us, 



1 I allude more especially to the production of triplets, their change into 

 quartets, and in all probability into flutings, and to the vanishing of flutings into 

 lines, by increasing the rate of dissociation. 



