MR. CHARLES DK \\ ATTKVII.LE ON FLAME SPECTRA. 163 



the point at which the discharge current leaves the electrode and enters the air 

 which plays the most important part in the production of the spectrum, and not the 

 general temperature pf the spark itself. It is obvious that the greater the resistance 

 to the passage of the current at this point the greater will l)e the amount of heat 

 1 Unrated there. The experiments of Professor HARTLEY* and M. ScHENK.t who 

 olwerved that what are considered to be the high-temperature lines of the spark 

 spectrum disappeared when the electrodes were heated to incandescence, do not prove 

 that this was the real reason ; for, since the electrodes were heated artificially, the 

 air l>etween them would IK? rendered conducting. Hence there would be less 

 resistance to the passage of the discharge, and therefore a less amount of heat 

 produced there, so that the high-temperature lines would be expected to disappear 

 under the circumstances. 



We cannot suppose that the oscillations themselves play any part of electrical 

 nature in the production of the lines in the spectrum of the oscillatory spark, since 

 the only effect of the self-induction is to suppress existing lines and not to introduce 

 new lines in the spectrum. If certain lines are increased in brilliancy by the 

 introduction of self-induction, as in the case of the spectra of metals of the iron 

 group, we may conclude, with Dr. HKMSALECH, that here the effect of the oscillations 

 is to raise the temperature of the vapour already formed lietween the electrodes by 

 the initial discharge.! Since the amount of heat lilxjrated in this initial discharge is 

 less than it would IK- in the case of the non-oscillatory discharge, the vapour will be 

 in a state of less advanced dissociation ; but if the amount of heat Migrated in the 

 succeeding discharges is more than sufficient to compensate for the loss by radiation, 

 which must be considerable on account of the high temperature of the vapour, then 

 the temperature of the vapour will be increased and consequently also the intensity 

 of some of the lines. 



It seems to me that a satisfactory explanation of the phenomena I have olworved 

 may be obtained by considering them as the results simply of variations of tempe- 

 rature. Of course, it cannot l>e denied that spectra may l)e formed by other means 

 than a pure and simple elevation of temperature, but these methods ultimately 

 reduce themselves to chemical transformations and thermal changes. This tempe- 

 rature hypothesis, which accounts in the simplest way for the production of the 

 spectra, finds considerable support in the fact that the stare give spectra which are 

 similar in their constitution to those obtained in the laboratory by electrical methods. 

 It is difficult to imagine any other but a purely thermal cause for the luminosity of 

 stars. 



In conclusion, I wish to express my l)est thanks to Professor SCHUSTER for his 



* ' Phil. Trans.,' 1882. 



t ' Astrophys. Journal' 14, pp. 116-135, 1901. 



I Loc. tit., p. 11. 



Y 2 



