778 Mr. S. G. Lusby on the 



rejected on account of their extreme volatility. In no case 

 were the variations in the critical voltage deduced from the 

 current curves outside the limits of experimental error. A 

 typical curve is shown in fig. 1, in which ordinat.es represent 

 current and abscissae the potential-difference between the 

 electrodes. 



Fig. 1. 

 Current in Pure and Salted Flame. 



5 ■ ' 



80 120 160 £00 



■PoTCNTt/u. DirrEftENCl //v l/oirs. 



A complete investigation of all the suitable salts of the 

 conducting metals was then made, experiments being made 

 at intervals of about 200°. The salts available for use at 

 each temperature of course varied according to the tem- 

 perature. As the temperature is reduced, the number of salts 

 suitable for exact work diminishes, owing to the fact that the 

 melting-points of many salts are above the flame temperature. 

 On the other hand, many salts which at high temperatures 

 are too volatile to employ prove very useful at low tem- 

 peratures. At 1150° it was rather difficult to work with 

 divalent metals (barium, strontium, and calcium) as their 

 conductivity is so low; the chlorides of barium and strontium 

 were the most reliable, whilst no calcium salt at all could be 

 used. The only calcium salt that could be emplo} r ed even 

 at the highest temperature was the chloride ; it was found 

 necessary to dry it immediately before use. Thus for any 

 generalization made about divalent metals there is not yevj 

 strong direct evidence in the case of calcium. No trivalent 

 salts could be found with conductivity high enough or vola- 

 tility low enough for measurements of mobility. If one could 

 be discovered, it would be possible to investigate the relation 

 between valency and mobility at low temperatures. 



