Metallie Vapours in an Exhausted Space. ok iar 
outside of the alkali metal which fills the bottom of the tube, 
since otherwise the action of the hot molten metal on the 
platinum wire and the glass causes the seal to crack. A is 
an iron or graphite anode. The length of the tube AB is 
from 5 to 10 inches. By shaking, the swinging anode EH 
and the platinum wire CD come into contact, and on separa- 
tion a small spark is produced, which ionizes the vapour in 
the tube, and in consequence of the proximity of the alkali 
metal cathode the ionization process is started at the latter. 
An arc is thus established between E and the alkali metal ; 
the small are EB acts then as a side-branch, and causes the 
main tube BA to start up. This may take place instant- 
aneously, or require a few seconds, according to the care 
with which the tube was exhausted. A direct contact of the 
auxiliary iron anode and the alkali metal would of course 
produce the same effect. Since, however, the alkali metal is 
introduced in solid form, and afterwards in melting contracts 
in volume, and since the arc causes evaporation of the alkali 
metal, the distance between E and the surface of the alkali 
metal changes, and arrangements would have to be made 
which would allow the iron to be brought in contact with 
the cathode at all these different distances. I found all such 
arrangements more complicated and less reliable in their 
action than the one shown above. A small tube T (not 
shown in the fig.) is provided at the lower part of the tube, 
through which the pieces of the alkali metal can be intro- 
duced, while, to prevent oxidation, hydrogen gas is allowed 
to pass through the tube. A stopcock on the pump-con- 
nexion is for this purpose to be connected with a hydrogen 
generator. After the alkali metal is introduced the tube T 
is closed on a sharp blast flame, and the exhaustion begun, 
When the vacuum is good the arc EB is started, and when 
the large amounts of gas evolved from the alkali metal have 
been carried away by the pump, the tube behaves exactly 
like a mercury tube and starts up almost as easily. ‘he 
exhaustion being made by means of a mercury pump, the 
discharge goes at the first moment through the mercury 
vapour that is present in the tube, but is soon replaced by a 
pure are of sodium or potassium vapours. I succeeded in 
having ares through potassium vapours 10 in. long and 
lasting for a few hours. The attack on the glass, in con- 
sequence of chemical action combined with a relatively high 
temperature, sooner or later puts an end to the experiment. 
The characteristics of the arc, except the spectrum, are 
essentially the same as those of the mercury arc. ' The elec- 
trodes possess the same properties, The potential drop 
12 
