air is slightly rarefied and conducting, then true conduction losses
occur also. In such case, of course, considerable energy may be
dissipated into space even with a steady potential, or with impulses
of low frequency, if the density is very great.

When the gas is at very low pressure, an electrode is heated more
because higher speeds can be reached. If the gas around the electrode
is strongly compressed, the displacements, and consequently the
speeds, are very small, and the heating is insignificant. But if in
such case the frequency could be sufficiently increased, the electrode
would be brought to a high temperature as well as if the gas were at
very low pressure; in fact, exhausting the bulb is only necessary
because we cannot produce (and possibly not convey) currents of the
required frequency.

Returning to the subject of electrode lamps, it is obviously of
advantage in such a lamp to confine as much as possible the heat to
the electrode by preventing the circulation of the gas in the bulb. If
a very small bulb be taken, it would confine the heat better than a
large one, but it might not be of sufficient capacity to be operated
from the coil, or, if so, the glass might get too hot. A simple way to
improve in this direction is to employ a globe of the required size,
but to place a small bulb, the diameter of which is properly
estimated, over the refractory button contained in the globe. This
arrangement is illustrated in Fig. 28.

[Illustration: FIG. 28.--LAMP WITH AUXILIARY BULB FOR CONFINING THE
ACTION TO THE CENTRE.]

The globe L has in this case a large neck n, allowing the small bulb b