90 MESSRS. A. SMITHELLS, H. M. DAWSON, AND II. A. WILSON : ELECTRICAL 



From the analogy stated to exist between dilute solutions of solids and matter in the 

 gaseous state, and from his own theory that in dilute solutions electrolytes are in 

 greater or less degree dissociated into their ions, ARRHENIUS was led to suppose tli.it 

 the vapour of electrolytes distributed in small concentration throughout a gaS would 

 likewise be electrolytically dissociated. He considers his e.xponmi>nt;il results to 

 harmonise with this view. He supposes that when an alkali salt is introduced into 

 a coal-gas flame the large excess of water vapour present converts the salt into a 

 hydroxide, in accordance with the following equation : 



M'X' + H,0 = M'OH + HX'. 



The hydroxide is then supposed to dissociate to a certain extent into its ions. 



Now, according to the electrolytic theory of solution, a free ion may, in virtue of 

 its electric charge, 1)6 characterised by properties totally different from those pertaining 

 to the ordinary chemical atom. Thus, in a dilute solution of sodium chloride, sodium 

 ions may persist in presence of water. If we accept the results and views of 

 AURHENIUS, we might suppose that in a flame coloured by an alkali salt, the metal is 

 liberated as an ion, and as such may persist in a strongly oxidising medium of flame 

 gases. Such an explanation would avoid the difficulties attending the more usually 

 adopted views. 



Another consideration appeared to favour the hypothesis suggested to us by the 

 results of ARRHENIUS. According to him the conductivity of a salt vapour is propor- 

 tional to the square root of its concentration in the flame. Now GOUY has shown 

 (' Ann. Chim. Phys.,' 18, 5, 1879), that the luminosity of a flame coloured by an 

 alkali salt is also within certain limits nearly proportional to the square root of its 

 concentration in the flame. 



The parallelism of these numerical relationships would obviously find a simple 

 explanation in the event of the luminosity and electrical conductivity being both 

 dependent on the presence of free ions. 



The importance of these deductions as affecting spectrum analysis decided us to 

 undertake an experimental investigation of the subject, and we were the more 

 inclined to this from the belief that we had at hand an apparatus capable of giving 

 accurate results. Besides this, a close examination of the results of ARKHENIUS 

 revealed some apparent discrepancies that detract from the weight of his conclusions.* 



The apparatus that we designed to employ was that used in other investigations 

 of flame (SMITHELLS, ' Phil. Mag.' (V.), 39, 123 (1895)). This apparatus permits of 

 the wide separation of the two cones that constitute the non-luminous flame of a 

 Bunsen burner, and it was thought that the interconal space which, in the apparatus, 

 is quite shielded from draughts, would afford a particularly favourable means of 



*r 

 * In his paper (loe. rit.) on p. 33, ARRHENIUS, dealing with the relation of conductivity to electromotive 



force, gives a set of galvanometer readings for an electromotive force of -2 Daniell. Later in the paper, 

 on p. 36, when dealing with the relation of electromotive force to concentration, another set of readings, 



