254 Prof. J. J. Thomson. [June 20, 



the discharge passes through them. Thus when the discharge first 

 passes through the vapour of ethyl alcohol, C 2 H 6 0, the spectrum on 

 the positive side of the plate is the candle spectrum, that on the 

 negative side the carbonic oxide spectrum. For some little time 

 after the discharge commenced I could not detect any hydrogen lines 

 on either side of the plate; after a timq, however, they appeared on 

 the negative side but not on the positive. If the discharge was kept 

 running for some time without letting a fresh supply of alcohol into 

 the tube, the " candle spectrum " on the positive side of the plate was 

 replaced by the CO spectrum, which now occurred on both sides of 

 the plate, accompanied on the negative side by the hydrogen spectrum. 

 This is the appearance presented by all the compounds of carbon, 

 oxygen, and hydrogen which I examined, when the spark had been 

 passing through them for a considerable time, and it is what would 

 occur if the vapour were decomposed by the spark into carbonic acid, 

 water, and hydrogen. 



The appearance of the candle spectrum on the positive side of the 

 plate with the CO on the negative was observed in many other cases. 

 Thus on sparking through a tube filled with CO I could not detect 

 any difference between the spectra on the two sides of the plate, but 

 when a little hydrogen was let into the tube the "candle spectrum" 

 appeared on the positive side of the plate, the carbonic oxide spectrum 

 on the negative. The same effect was observed in a tube filled with 

 cyanogen mixed with a little hydrogen. When the tube was filled 

 with the vapour of methyl alcohol, CH 3 OH, the candle spectrum 

 was on the positive side of the plate, the carbonic oxide and hydrogen 

 spectra on the negative ; with this vapour, unlike that of ethyl 

 alcohol, I could not detect any stage when the hydrogen spectrum 

 was absent. 



The first explanation which occurs to one of this phenomenon is 

 that it is owing to the potential gradient at the negative side of the 

 plate being steeper than that on the positive, so that we may imagine 

 we have a fierce spark on the negative side, a mild one on the posi- 

 tive, and that the fierce spark gives the CO spectrum, the mild one 

 the candle spectrum. There are, however, some phenomena which 

 seem inconsistent with this explanation : in the first place if the 

 current is reversed after flowing in one direction, traces of the former 

 spectra linger for some time at the sides of the plates, and, secondly, 

 if the difference is due to the greater decomposition at the negative 

 side of the plate, how is it that in the case of the vapour of ethyl 

 alcohol the hydrogen spectrum is not seen, at the commencement of 

 the discharge, on the negative side of the plate ? It only appears after 

 the discharge has passed through for some time, when hydrogen has 

 probably been set free by the decomposition of the vapour by the dis- 

 charge. If the absence of the candle spectrum from the negative side 



