﻿582 Dr. Marshall Watts and Mr. Wilkinson on 



entered upon a study of the subject, and the question was 

 supposed to have been get at rest. 



But in 1900, the late Sir G. Stokes suggested that the 

 Swan spectrum belonged to carbonic oxide, and this view was 

 adopted by Prof. Smithells, who in a paper in the Phil. Mag. 

 for April 1901 has sought to establish it by a critical study 

 of the spectra of carbon compounds generally and of flame 

 spectra in particular. 



He sets aside the evidence of many observers that the 

 Swan spectrum is obtainable by the electric discharge in 

 hydrocarbons, carbon chloride, and carbon sulphide, by the 

 supposition that all these different materials contained oxygen 

 as an impurity. He points to the difficulty of removing films 

 of air or moisture from glass, the occlusion of gases by 

 electrodes, the oxidizing character of glass itself ; and 

 there is indeed no doubt that the evidence obtained from 

 exhausted tubes must be received with much caution ; but 

 the case would seem to be different when the spark is taken 

 in gases at ordinary pressure, and the results will be still less 

 open to objection on account of possible contamination by 

 impurities when the spark is taken in the liquid carbon 

 compound itself. We beg to offer the results of some experi- 

 ments we have made on the spectra of the spark in various 

 liquids containing carbon and not oxygen, as far as it lay in 

 our power to secure the absence of this element, as a contri- 

 bution towards the determination of the chemical origin of 

 the " Swan" spectrum. 



By the " Swan " spectrum we mean the band-spectrum as 

 obtained at the base of a Bunsen flame, and more brightly 

 by the combustion of hydrocarbons such as olefiant gas with 

 oxygen in a " mixed " blowpipe jet. In the visible portion 

 of the spectrum we have five groups of bands each made up 

 of several " edges " sharply defined on the red side and fading 

 away on the more refrangible side. The wave-lengths of 

 these edges are as follows : — 



f 6188-1 

 I 6120-0 

 Group a, <j 6052-0 



5999-0 



5955-0 



f 5635-4 



I 5585-5 



Group y <{ 5540-9 



I 5500-0 

 I 5470-0 



