e 18, but without making a perforation in the bottom of
the large tube. Heating and expanding by air pressure, first through the
large tube, then through the smaller tube and then again through the
large tube, will give a satisfactory finish to this part of the work.

[Illustration: Fig. 11]

The syphon tube is now joined on to the large tube as shown by _a_, Fig.
11, care being taken to seal the other end of the syphon tube before
joining. The details of the final and re-entrant joint of the syphon
tube are shown at the lower part of _a_. This join is made by expanding
the sealed end of the syphon tube into a small, thick-walled bulb, and
the bottom of this bulb is burst out by local heating and blowing; the
fragments of glass are removed and the edges made smooth by melting. A
similar operation is carried out on the side of the tube to which the
syphon tube is to be joined. This stage is shown by _a_. Now heat the
syphon tube at the upper bend until it is flexible, and press the bulb
at its end into the opening on the side of the other tube. Hold the
glass thus until the syphon is no longer flexible. The final join is
made by heating the two contacting surfaces, if necessary pressing the
edges in contact with the end of a turn-pin, fusing together and
expanding. The finished apparatus is shown by _c_.

_Electrodes._--A thin platinum wire may be sealed into a capillary tube
without any special precautions being necessary. The capillary tube may
be drawn out from the side of a larger tube by heating a spot on the
glass, touching with a glass rod and drawing the rod away; or the
exhaustion branch described on page 18 may be used for the introduction
of an electrode. It is convenient sometimes to carry out the exhaustion
through the same tube that will afterwards serve for the electrode. The
electrode wire is laid inside the branch before connecting to the
exhaustion pump. When exhaustion is completed the tube is heated until
the soft glass flows round the platinum and makes the seal air-tight.
The branch is now cut off close to the seal on the pump side, a loop is
made in the projecting end of the platinum wire, and the seal is
finished by melting the cut-off end.

Platinum is usually employed for such work, but if care is taken to
avoid oxidation it is not impossible to make fairly satisfactory seals
with clean iron or nickel wire. Hard rods of fine graphite, such as are
used in some pencils, may also be sealed into