Metallic Vapours in an Exhausted Space. 101 
that in order that an are should be started, this primary 
ionization must be excited on the cathode. 
We are now able to complete the description of the expe- 
riments 1 and 2 by giving the most convenient way, before 
alluded to, for starting the arc BD in those experiments. It 
consists in making the tube AB a side-branch, making the 
connexions as shown in fig. 2, and then disconnecting A and 
applying the second source g’. 
The different methods by means of which a mercury arc 
can be started in an exhausted tube may be now summarized. 
1. Bringing the cathode into contact with the anode and 
separating the two electrodes. This is the ordinary method 
used in all arcs. . 
2. Application of high voltage to the two electrodes. 
3. The new method described in experiments 1 and 2, that is, 
bringing a cathode in contact with an already active electrode. 
4. Mere mechanical agitation of the mercury. 
Experiment 6.—If, in fig. 2, the tube BD is short, the tube 
well exhausted, the connexion to the electrode A can be taken 
off, and the are BD started by violent shaking of the tube. 
This is accounted for by the fact that the shaking of the 
mercury in an exhausted tube is accompanied by production 
of light, that is, of luminous ionized vapour, starting from 
the surface of mercury. 
5. Contact of the cathode with ionized vapour. Returning 
to experiment 1, we found a small current flowing between 
the electrodes A and C. This was carried by the ionized 
vapour supplied by the are BD. This current differs in mag- 
nitude according to the amount of ionized mercury vapour in 
the space between the electrodes. The eventual starting of 
the are AC can be explained by assuming that the presence 
of ionized vapour facilitates, under certain conditions, the 
starting of the ionization process on the cathode. Shaking of 
the mercury surface or the impact of the condensing mercury 
running back seems to be the more effective the larger the 
amount of ionized vapour present above the surface. In this 
connexion the following experiment is interesting. 
Experiment 7.—The glass tube ABCD (fig. 4) has three 
mercury cups, A, B, CU, and one graphite electrode, D. 
B and C are connected to a direct-current source g; A and 
D to another direct-current source yg’. The circuit AD 
being open, the are BC is started; in this arrangement 
the are surrounds the electrodes A and D of the other 
current. Closing the switch s in the circuit of AD usually 
establishes an are between Aand D. It is, however, possible, 
although not very easy, to realize such conditions that on 

