50 Recent Advances in Stellar Astronomy
Class G, (to which the Sun belongs) they are predomi-
nant, while the spark lines become weaker as do also
the lines of hydrogen. This change continues into Class
K, and the flame lines of the metals, (produced in a
flame, or an electric furnace, at relatively low tempera-
ture) become conspicuous. Finally, in Class M, heavy
absorption bands appear, which have been identified as
due to the presence in the star’s atmosphere of the vapor
of titanium oxide. Another set of bands, produced by
compounds of carbon, is prominent in Classes R and N.
What physical or chemical conditions are there behind
this succession of types? Are the atmospheres of the
various stars really as different in chemical composition
as their spectra appear to indicate? This is a tempting
hypothesis at first sight: but there are weighty objec-
tions to it.
First of all, the sequence of spectra is linear. In
passing from a given type to those which closely resemble
it, we have practically only two lines to follow—up or
down the series. It seems an inevitable deduction from
this that the spectrum of a star might be described in
mathematical language, as a function of a single variable.
Now the chemical constitution of an atmosphere is a
function of many variables. For all that we know, the
amounts of helium, oxygen, iron, hydrogen, and other
elements in it may vary independently. If variations
of this sort lay behind the spectral differences, there is
no apparent reason why the lines of different elements
should always show so definite a relation between their
behavior—or, indeed, any conspicuous relation at all.
Moreover, we know of pairs of stars, so close that
no telescope can separate them, whose duplicity is known
only because one component of the system eclipses the