J8 MKSSKS. A. S.MITIIKI.LS. 11. M. 1'AWSON, AND 11. A. WILSON: F.I.KCTKICAL 



would appear, to solution of the glass. The quantity dissolved is, of course, small, 

 but still quite sufficient to make itself felt. The conductivity of the flame gas. s 

 alone, is, it must be remembered, extremely small in comparison with that of a flame 

 containing the spray of -rdoV normal solution of a potassium salt.* 



The following table gives the conductivity of the flame gases with distilled water 

 spray for several electromotive forces : 



E.M.F. in volts . . 5'66 2'52 1-09 '803 '521 '230. 

 Current .... '547 '313 '226 '177 '118 "056. 



These results are plotted in Curve V, p. 110, the ordinates being multiplied by 10. 



The above values were checked from time to time during our work. The agree- 

 ment was always well within the limits of accuracy required, and consequently we 

 have used these values in all cases to represent the conductivity of the free flame, 

 subtracting them from the gross readings given by salt solutions to obtain the 

 conductivity of the salt vapour itself. 



AKRHENIUS, in his experiments on this subject, noticed an increased conductivity 

 of the free flame immediately after a salt solution had been sprayed. He attributed 

 it to a deposition of salt on the electrodes, the deposit remaining on the electrodes for 

 some time after the spray of salt solution had been stopped. The effect in question 

 was not important in the case of salts of the alkali metals on account of their ready 

 volatility. 



In our experiments we noticed this effect, but we believe that in our apparatus with 

 the salts used by us it was fully accounted for by the fact that after the spray of salt 

 had been replaced by one of water, salt spray lingered for some time in the apparatus 

 between the sprayer and the flame tube. We noticed that the flame remained coloured 

 below the electrodes, and that as the colour faded the galvanometer deflection fell 

 steadily to the normal value. We have therefore used the normal value as the true 

 correction. Even if our explanation of the higher value, found immediately after 

 stopping the salt spray, is incomplete, the arbitrariness in our case of taking any other 

 value than the normal one as the correction, would forbid us attempting further 

 refinement, where, as a matter of fact, the difference involved could not seriously 

 affect our final results. 



Unipolar Conduction. 



It has long been known that unipolar conduction is shown to a marked extent in 

 the case of flame gases, that is to say the current passes from one electrode to the 

 other more easily in one direction than in the reverse direction. The following table 



*SCHAU,ER has shown ('Zeit. Phys. Cliem.,' 25, 497, 1898) that pure water acts so rapidly on glass 

 vessels as to forbid their use in conductivity experiments, but that with salt solutions, even of very small 

 concentration, the solvent action is not sufficient to introduce sensible error. 



