MEMOIR II.] SPECTRUM ANALYSIS OF FLAMES. (53 



principles just pointed out in this method of horizontal 

 analysis, being of much greater width than the rest of 

 the spectrum, and recalling to the imagination the ap- 

 pearance of Saturn's ring when nearly closed and seen 

 through a telescope of moderate power. This ray, from 

 its superior breadth, must necessarily come from that 

 pale, tawny light which invests the bright part of the 

 flame. This, which is readily seen when the flame is 

 large, envelops the middle and upper parts, but cannot 

 so easily be detected low down. It is to be attributed 

 to the carbonic acid and steam that have risen at a high 

 temperature in the burning shell, and are escaping at a 

 degree above that of incandescence into the air, and are 

 mingled with oxygen diffusing from the air into them. 

 A similar tawny cloak surrounds the upper part of the 

 flame of a candle ; it answers to the oxidizing flame of 

 the blow-pipe, and yields Brewster's monochromatic yel- 

 low light. 



(A few years subsequently this yellow ray was dis- 

 covered by Swan to be due to sodium. It is now known 

 as the lines D. At the date of this memoir it was not 

 suspected that sodium is so universally diffused.) 



IV. Explanation of tlie nature of colored flames, show- 

 ing, for example, why carbonic oxide burns blue, and cy- 

 anogen red. 



To return now to carbonic oxide and cyanogen. Fig. 

 8, No. 1, represents the solar spectrum with its fixed 

 lines ; No. 3 represents the spectrum of carbonic oxide 

 burning in the air. It begins in the red region short of 

 the fixed line C, and terminates between the lines G and 

 H. It yields, therefore, rays of every color, and this in 

 accordance with the principles we have laid down ; but 

 when the relative quantity and force of the rays are es- 

 timated in comparison with the sunlight spectrum, the 



