l bulbs which are alike,
only one is exhausted to a low and the other to a very high degree.
When connected to the coil, the filament in the former glows uniformly
throughout all its length; whereas in the latter, that portion of the
filament which is in the centre of the bulb glows far more intensely
than the rest. A curious point is that the phenomenon occurs even if
two filaments are mounted in a bulb, each being connected to one
terminal of the coil, and, what is still more curious, if they be very
near together, provided the vacuum be very high. I noted in
experiments with such bulbs that the filaments would give way usually
at a certain point, and in the first trials I attributed it to a
defect in the carbon. But when the phenomenon occurred many times in
succession I recognized its real cause.

In order to bring a refractory body inclosed in a bulb to
incandescence, it is desirable, on account of economy, that all the
energy supplied to the bulb from the source should reach without loss
the body to be heated; from there, and from nowhere else, it should be
radiated. It is, of course, out of the question to reach this
theoretical result, but it is possible by a proper construction of the
illuminating device to approximate it more or less.

For many reasons, the refractory body is placed in the centre of the
bulb, and it is usually supported on a glass stem containing the
leading-in wire. As the potential of this wire is alternated, the
rarefied gas surrounding the stem is acted upon inductively, and the
glass stem is violently bombarded and heated. In this manner by far
the greater portion of the energy supplied to the bulb--especially
when exceedingly high frequencies are used--may be lost for the
purpose contemplated. To obviate this loss, or at least to reduce it
to a minimum, I usually screen the rarefied gas surrounding the stem
from the inductive action of the leading-in wire by providing the stem
with a tube or coating of conducting material. It seems beyond doubt
that the best among metals to employ for this purpose is aluminium, on
account of its many remarkable properties. Its only fault is that it
is easily fusible, and, therefore, its distance from the incandescing
body should be properly estimated. Usually, a thin tube, of a diameter
somewhat smaller than that of the glass stem, is made of the finest
aluminium sheet, and slipped on the stem. The tube is conveniently
prepared by wrapping around a rod