104 Dr. E. Weintraub on the Are in 
pump, and in view of the influence of the non-conductive 
mercury vapour on the stability of the arc recorded later, I 
consider it highly probable that the presence of mercury 
vapour in itself is sufficient to prevent the starting of the 
lamp. The behaviour is, of course, quite different when the 
impressed voltage is high enough to split up the mercury 
molecules. In that case the heating of the tube and the 
mercury facilitates the starting. The same is, however, the 
case with the presence, in small quantities, of foreign gases; 
so that the analogy between the behaviour of inert mercury 
vapour and other non-conductive gases toward high, as well 
as low, voltages is complete. 
Starting from the idea that the presence of slight traces 
of foreign gases or mercury vapour affects only the rate of 
propagation of the ionization, I was looking for means to 
accelerate the latter, and so make the starting of the are sure 
at any time, with cold and warm tubes. This means has 
been found in a thin carbon filament of high resistance, 
suspended to the anode and reaching within a short distance 
(3-6 ins.) of the mercury cathode. Such a lamp, if well 
exhausted and if care has been taken to drive out by heating 
the gases of the filament and anode, starts up at any time 
instantaneously. 
Eaperiment 8.—The lamp is represented in fig. 5. After 
the tube has been well exhausted and the mercury introduced 
into the cups, the are is started in the side-branch, the 
tube being connected to the pump. The arc in the main 
tube starts up first between the mercury cathode and the 
lower end of the filament. The filament is in this way 
heated up by the current which flows through that are and 
the gases set free are carried away by the pump. When 
most of the gases occluded in the filament have been driven off. 
the arc fills out the whole tube and goes to the piece of 
graphite D, which serves as anode. By the heat generated 
at the anode the gases occluded in it are expelled. When 
most of the gases have been driven out and the tube shows a 
behaviour like that described above (instantaneous starting at 
any time with a small current in the side-branch), the lamp 
can be sealed off and the experiments repeated with it at any 
time. Care has to be taken, however, never to have an arc 
in the tube with higher current than that at which it was 
exhausted. 
The réle of the carbon filament is not perfectly clear. It 
certainly helps the starting by shortening the distance between 
the cathode and anode, since when the ionization reaches the 
end of the filament a small current is already established 
