e not confined to the metals themselves; the _salts_ of these
metals yield the bands of the metals. Chemical union is ruptured by a
sufficiently high heat; the vapour of the metal is set free, and it
yields its characteristic bands. The chlorides of the metals are
particularly suitable for experiments of this character. Common salt,
for example, is a compound of chlorine and sodium; in the electric
lamp it yields the spectrum of the metal sodium. The chlorides of
copper, lithium, and strontium yield, in like manner, the bands of
these metals.

When, therefore, Bunsen and Kirchhoff, the illustrious founders of
_spectrum analysis_, after having established by an exhaustive
examination the spectra of all known substances, discovered a spectrum
containing bands different from any known bands, they immediately
inferred the existence of a new metal. They were operating at the time
upon a residue, obtained by evaporating one of the mineral waters of
Germany. In that water they knew the unknown metal was concealed, but
vast quantities of it had to be evaporated before a residue could be
obtained sufficiently large to enable ordinary chemistry to grapple
with the metal. They, however, hunted it down, and it now stands
among chemical substances as the metal _Rubidium_. They subsequently
discovered a second metal, which they called _Caesium_. Thus, having
first placed spectrum analysis on a sure foundation, they demonstrated
its capacity as an agent of discovery. Soon afterwards Mr. Crookes,
pursuing the same method, discovered the bright green band of
_Thallium_, and obtained the salts of the metal which yielded it. The
metal itself was first isolated in ingots by M. Lamy, a French
chemist.

All this relates to chemical discovery upon earth, where the materials
are in our own hands. But it was soon shown how spectrum analysis
might be applied to the investigation of the sun and stars; and this
result was reached through the solution of a problem which had been
long an enigma to natural philosophers. The scope and conquest of this
problem we must now endeavour to comprehend. A spectrum is _pure_ in
which the colours do not overlap each other. We purify the spectrum by
making our beam narrow, and by augmenting the number of our prisms.
When a pure spectrum of the sun has been obtained in this way, it is
found to be furrowed by innumerable dark lines. Four of them were
first seen by Dr. Wollaston, but they were afterwards multipli