May 14, 1914] 
NATURE 281 
RELATIONS. BETWEEN THE SPECTRA AND 
OTHER CHARACTERISTICS OF THE 
STARS. * 
Lvs 
“FS proceed farther we must have recourse to the 
study of eclipsing variable stars. Methods for 
computing their orbits have been developed at Prince- 
ton during the last few 
years,*° the main motive for 
the investigation being the 
astrophysical importance of 
the results. Dr. Shapley, using 
the methods devised by the 
speaker, has obtained 
elements for eighty-seven such 
systems," for each one of 
which the density of the com- 
ponents may be calculated. 
The values here employed are 
those which result from the 
assumption that the stars pre- 
sent discs darkened toward 
the edge, like the sun, but to 
a still higher degree, and the 
principal uncertainty of the 
results (which in any case 
cannot be very serious) arises 
from our present ignorance of 
the actual degree of this 
darkening. For our present 
purpose, they may be best 
utilised by computing the 
absolute magnitudes which 
the brighter component of 
each system would have if its 
mass and surface brightness 
were equal to those of the 
sun, leaving outstanding the 
differences due to density 
alone. 
The results for the eighty 
eclipsing variable the elements 
and spectra of which are 
known are plotted in Fig. 4, 
on the same system as in the 
preceding figures. The black 
dots represent those stars for 
which the photometric data 
are most trustworthy, the 
open circles those of less pre- 
cision. We are once more 
greatly indebted to Prot. 
Pickering and Miss Cannon 
for information regarding the 
spectra of these stars. To the 
absolute magnitude 4-0 on 
this diagram corresponds a 
density 1/3 that of the sun; 
torsaeunis: to 2-0, 1/45 5"t0 
1-0, 1/180 of the sun’s density, 
and so on. 
This diagram bears at first 
sight but small resemblance to the previous 
ones, but a _ litile study brings out severai 
important things. . First, though the majority 
~ An address delivered before a joint meeting of the Astronomical and 
Astrophysical Society of America and Section A of the American Association 
for the Advancement of Science, at Atlanta, Georgia, December 30, IQ13, 
with a few additions, by Prof. H. N. Russell. Continued from p. 258. 7 
*0 Astrophys Jour., vol. xxxv., p. 315, and vol. xxxvi., pp. 54, 239, 395; 
1912. 
*l Astrophys. Jour., vol. xxxviii., pp. 159-73, 19013. 
NO. 2324, VOL. 93] 
of these eclipsing variables are of Class BA 
every class from B to K is represented, and there 
are eight stars of Class G or redder. Secondly, all 
but one of these eighty stars are less dense than the 
sun, though but few of them are of less than 1/100 
the sun’s density. Thirdly, the stars of Classes A 
and B are fairly similar to one another in density, 
the great majority having densities between 1/3 and 
1/45 of the sun’s; those of Class F show a compact 
oe 
ne 
a 
Fic, 4. 
group of high density, and an isolated star of low 
density ; but in Classes G and K the range of density 
is enormously great—from nearly twice that of the 
sun (W Ursz Majoris) to one-millionth of the sun’s 
| density (W Crucis, at the top of the diagram). 
Fourthly, among the stars of density less than 1/200 
that of the sun (corresponding to about +1-om. on 
the diagram), only one of the seventy-one stars of 
Classes B to F appears, while four of the eight stars 
