phosphorescence begins to appear.

[Illustration: FIG. 21.--IMPROVED BULB WITH NON-CONDUCTING BUTTON.]

[Illustration: FIG. 22.--TYPE OF BULB WITHOUT LEADING-IN WIRE.]

When the bulb is connected with the coil, the piece of carbon, if
small, may become highly incandescent at first, but its brightness
immediately diminishes, and then the discharge may break through the
glass somewhere in the middle of the stem, in the form of bright
sparks, in spite of the fact that the platinum wire is in good
electrical connection with the rarefied gas through the piece of
carbon or metal at the top. The first sparks are singularly bright,
recalling those drawn from a clear surface of mercury. But, as they
heat the glass rapidly, they, of course, lose their brightness, and
cease when the glass at the ruptured place becomes incandescent, or
generally sufficiently hot to conduct. When observed for the first
time the phenomenon must appear very curious, and shows in a striking
manner how radically different alternate currents, or impulses, of
high frequency behave, as compared with steady currents, or currents
of low frequency. With such currents--namely, the latter--the
phenomenon would of course not occur. When frequencies such as are
obtained by mechanical means are used, I think that the rupture of the
glass is more or less the consequence of the bombardment, which warms
it up and impairs its insulating power; but with frequencies
obtainable with condensers I have no doubt that the glass may give way
without previous heating. Although this appears most singular at
first, it is in reality what we might expect to occur. The energy
supplied to the wire leading into the bulb is given off partly by
direct action through the carbon button, and partly by inductive
action through the glass surrounding the wire. The case is thus
analogous to that in which a condenser shunted by a conductor of low
resistance is connected to a source of alternating currents. As long
as the frequencies are low, the conductor gets the most, and the
condenser is perfectly safe: but when the frequency becomes excessive,
the _role_ of the conductor may become quite insignificant. In the
latter case the difference of potential at the terminals of the
condenser may become so great as to rupture the dielectric,
notwithstanding the fact that the terminals are joined by a conductor
of low resistance.

[Illustration: FIG. 23.--EFFECT PRODUCED BY A RUBY DROP.]

It is,