The Evolution of the Stars 
[ch. VI 
may be of the same type as, but a generalisation of, the radioactive processes 
as they occur on earth. The circumstance that the atomic numbers calculated 
in the last chapter came out somewhere in the neighbourhood of the atomic 
numbers of the radioactive elements gave support to this view. 
Nevertheless, something more than mere radioactivity is necessary to 
account for the radiation of the stars. So far as is at present known, every 
radioactive process increases the number of atoms in the universe, whereas 
stellar radiation can only result from a decrease; it calls for actual annihilation 
of atoms, a process of which terrestrial radioactivity knows nothing. 
It would obviously be rash to base any very definite conclusions on the 
very uncertain calculations of atomic weight given in the last chapter. It is 
entirely possible that the true atomic weights of stellar 'matter are all less 
than that of uranium, and that the atoms which undergo annihilation in the 
stars are merely isotopes of the heavier of the terrestrial elements. 
Russell’s two Theories of Stellar Evolution. 
154. The possibility that the observed spectral sequence corresponded to 
evolutionary development in the chemical composition of the stars, was in effect 
ruled out when Hertzsprung discovered in 1905 that red stars fell into the 
two distinct classes which he designated as giants and dwarfs. The giant red 
stars, of far higher luminosity than Sirius, shewed the lines of heavy elements 
in their spectra as well as conspicuous bands which were identified with the 
bands of titanium oxide. If the chemical evolution of the stars was from 
hydrogen to more complex elements, it was absurd to find titanium, of atomic 
number 22 , figuring prominently in the atmospheres of the youngest of the 
stars, not to mention strontium (37), yttrium (38) and barium (56). 
Russell emphasized this in 1913, when he published his first diagram of the 
distribution of spectral types by absolute magnitude, and based a theory of 
stellar evolution upon it*. 
In place of the reversed 7 formation which characterises the true Russell 
diagram (cf. fig. 6 , p. 62), Russell’s first diagram seemed to shew the formation 
exhibited in fig. 14. The most luminous stars were distributed over all spectral 
types along the range PQ, with the least luminous at R. 
This led Russell to propound a theory that the curve PQR represented an 
evolutionary sequence, the spectrum of the normal star first advancing from 
type M to type B or A and then receding again to type M. 
It was easy to find a theoretical justification for this hypothesis. We have 
seen how a star’s density can be calculated from its observed luminosity and 
surface temperatures. Giants of types M, K and G are found to have mean 
densities of the order of 0 ' 000002 , 0'0005 and 0'004 respectively. In general 
* American Association for the Advancement of Science, 1913 meeting, and Nature, xcm. (1914), 
pp. 227, 252.