1892.] Spectra of the Flames of Metallic Compounds. 117 



lative care, ensured exactitude for purposes of comparison between 

 the behaviour of the magnetised and unmagnetised steel bars in the 

 corrosive fluid. 



An average of the twenty-nine experiments in Table I indicates an 

 increase of corrosion in the steel due to magnetic influence of about 

 3 per cent, under the conditions of experimentation. 



The steel bars were not highly magnetised, and I purposely ex- 

 posed them to the action of the solution for somewhat long periods 

 in the present experiments, as thereby, perhaps, affording a better 

 indication of the influence of magnetisation on general corrosion, the 

 almost immediate effect of magnetisation on the corrosion of steel 

 having been demonstrated in the electrical observations of Part II. 



It is probable that the deviation in the individual results of 

 Table I, in the extent of the loss by corrosion, may be attributed to 

 variation in the extent to which the several bars were magnetised. 

 The results recorded in the present paper indicate that magnetisation 

 exerts an effect, though small, on the extent of the corrosive action of 

 copper salts on iron and steel. This is probably owing to the local 

 currents, set up by magnetisation between the polar and central 

 portions of the bars, inducing somewhat greater chemical action. 



In some of the experiments with the copper solution which con- 

 tained the more highly magnetised bars, the copper solution was of a 

 perceptibly lighter colour towards the end of an experiment when 

 compared with the colour of the copper solution containing the 

 unmagnetised steel bar. 



VIII. <c Note on the Spectra of the Flames of some Metallic 

 Compounds." By G. D. LlVEiNG, M.A., F.R.S., Professor of 

 Chemistry, and J. DEWAR, LL.D., F.R.S., Jacksonian Pro- 

 fessor, University of Cambridge. Received June 3, 1892. 



[PLATE 5.] 



A study of the spectra of flames offers many points of interest. It is 

 long since A. Mitscherlich (Poggendorff's ' Annalen,' vol. 116, p. 499 ; 

 vol. 121, p. 459) showed that the spectra of flames are, for the 

 most part, those of compounds of the elements present, and contain 

 comparatively few rays proceeding directly from the elements them- 

 selves. But there are many questions still undecided. For example, 

 it is not known whether the vibrations which give the spectra of com- 

 pounds in flames are those which the molecules of the compounds in 

 question would assume under the action of a high temperature alone, 

 or whether they are not vibrations of a different order, arising during 



