n external connection, which serves
to investigate the effect of the tube under various conditions. It is
referred to chiefly to suggest a line of experiment followed.

Since the bombardment against the stem containing the leading-in wire
is due to the inductive action of the latter upon the rarefied gas, it
is of advantage to reduce this action as far as practicable by
employing a very thin wire, surrounded by a very thick insulation of
glass or other material, and by making the wire passing through the
rarefied gas as short as practicable. To combine these features I
employ a large tube T (Fig. 21), which protrudes into the bulb to some
distance, and carries on the top a very short glass stem s, into which
is sealed the leading-in wire w, and I protect the top of the glass
stem against the heat by a small, aluminium tube a and a layer of mica
underneath the same, as usual. The wire w, passing through the large
tube to the outside of the bulb, should be well insulated--with a
glass tube, for instance--and the space between ought to be filled out
with some excellent insulator. Among many insulating powders I have
tried, I have found that mica powder is the best to employ. If this
precaution is not taken, the tube T, protruding into the bulb, will
surely be cracked in consequence of the heating by the brushes which
are apt to form in the upper part of the tube, near the exhausted
globe, especially if the vacuum be excellent, and therefore the
potential necessary to operate the lamp very high.

Fig. 22 illustrates a similar arrangement, with a large tube T
protruding in to the part of the bulb containing the refractors button
m. In this case the wire leading from the outside into the bulb is
omitted, the energy required being supplied through condenser coatings
CC. The insulating packing P should in this construction be tightly
fitting to the glass, and rather wide, or otherwise the discharge
might avoid passing through the wire w, which connects the inside
condenser coating to the incandescent button m. The molecular
bombardment against the glass stem in the bulb is a source of great
trouble. As illustration I will cite a phenomenon only too frequently
and unwillingly observed. A bulb, preferably a large one, may be
taken, and a good conducting body, such as a piece of carbon, may be
mounted in it upon a platinum wire sealed in the glass stem. The bulb
may be exhausted to a fairly high degree, nearly to the point when