396 
ASTRONOMY: H. N. RUSSELL 
widening due to physical conditions in its atmosphere (which are likely 
to affect some lines more than others). 
(d) Another promising field is found among the reddest stars. Curtiss 
makes the very interesting suggestion that the division of the spectral 
series into the branches G-K-M and G-R-N (or perhaps K-R-N) 
may be due to differences of chemical composition^ ^ — since it is known 
that the surface temperatures of these stars are low enough to permit 
the formation of chemical compounds. If this is true, the strength 
of the characteristic bands of titanium oxide or of carbon should depend 
upon the relative proportions of these elements, and show little correla- 
tion with the color index, or the extension of the spectrum in the violet, 
which depend primarily on the temperatures. There is already consid- 
erable evidence that this is actually the case, and it may be remarked 
that the star Epsilon Geminorum, which is of spectral class G5 has a 
color index ( + 1.52) almost equal to that of Classes M or R.^^ 'pj^jg 
star may be in the situation anticipated by Curtiss, in which an exact 
chemical equilibrium between carbon and titanium oxide suppresses 
the bands of both. 
Photography of the spectra of bright stars in the red, and even the 
near infra-red, is now practicable, and Merrill^^ has already obtained 
results of great interest and promise. Investigation of the spectra of 
the brightest stars with high dispersion is also profitable, as is shown 
by the work of Adams^'' upon the pressures which probably prevail in 
the atmospheres of Sirius, Procyon, and Arcturus. Fortunately, the 
stars brighter than the second magnitude afford examples both of giant 
and dwarf stars of almost every spectral class. 
2. (a) Almost equal in importance to the line absorption in stellar 
spectra is the distribution of intensity in the continuous background. The 
most complete and satisfactory method of studying this would be the 
direct measurement of the energy carried by different wave-lengths, but 
this has not yet been proved practicable. A first step has however been 
taken by Coblentz,^^ who has not only measured the total energy radia- 
tion of more than a hundred stars, but in some cases the percentage 
transmitted by a water cell, thus providing our first knowledge of 
stellar radiation in the infra-red. With the great reflectors just com- 
pleted, the determination of spectral energy curves for the brightest 
stars may be possible. 
The distribution of energy in the luminous region of the spectrum is 
however readily determinable. For the brighter stars, spectro-photo- 
metric methods can be employed, as in the visual work of Wilsing 
