MEMOIR II.] SPECTRUM ANALYSIS OF FLAMES. Ql 



It might be supposed that in the familiar instance of 

 an oil lamp, if we put any check on the supply of the 

 air, and thereby check the intensity of the combustion, 

 we ought to produce a flame emitting rays of light the 

 refrangibility of which becomes less and less, and which, 

 from their being originally white, should pass through 

 various shades of orange, and end in a dull red. But 

 the compound nature of the burning vapor interferes 

 with that result ; for when a certain point is gained the 

 hydrogen, for the most part, alone burns, the carbon be- 

 ing set free as smoke, and such a flame cannot support 

 itself in strict accordance with the principle given. 



We must, then, search for other conditions under which 

 carbon is found which are free from this difficulty. Two 

 at once present themselves : they are carbonic oxide 

 and cyanogen gas. In the former the carbon is already 

 united with half the oxygen required for maximum ox- 

 idation : its complete combustion can therefore be carried 

 on with a limited supply of atmospheric air ; in the lat- 

 ter the carbon is united with nitrogen, which during the 

 combustion is set free, and interferes with the process by 

 cutting off the more complete access of the atmosphere. 



In place of the burning coal of the former experiments 

 I substituted a jet-pipe through which the various gases 

 might be made to pass, and the rays emitted by their 

 flames enter the telescope after passing through the slit 

 and prism. In this arrangement the slit should be hori- 

 zontal and not vertical. So far from it being immaterial 

 which of the two positions is selected, very great advan- 

 tages arise from the former. If the slit be vertical, the 

 prism, it is true, will separate the constituent colors from 

 one another, but it fails to show their relative positions. 

 If it be horizontal, the relative positions of the different 

 colors can be demonstrated, and it can be proved that a 

 horizontal section of a flame is in reality, as has been al- 



