Fkbbuary 7, 1919] 



SCIENCE 



131 



period and phase of the orbital motion, and the 

 relative brightness of the secondary, are in 

 agreement with the photometric data, and (3) 

 by the success of the reverse process of photo- 

 metric investigation of promising spectro- 

 scopic binaries, which have in many cases re- 

 vealed variation of the eclipsing type, though 

 of small amplitude, but with the theoretical 

 period and phase. The fact that some spectro- 

 scopic binaries have proved not to be va- 

 riable is a further confirmation of the theory 

 (the orbital inclination being such that 

 eclipses fail to happen). 



The development by the speaker of simple 

 methods for computing the orbital elements of 

 these systems has extended our list so that, at 

 present, owing mainly to Shapley's industry, 

 orbits are available for about 100 eclipsing 

 pairs — a sufficient nimiber to allow of drawing 

 conclusions by statistical methods. These 

 stars show a strong concentration toward the 

 galactic equator, and the majority of them 

 have spectra of classes A and B, though there 

 are a number of class F and a few of classes 

 G and K. Their periods are usually less than 

 ten days, though two are known with periods 

 of about six and nine months, and several 

 others with periods about a month. 



'WTien a more careful study is made of their 

 variations it is found that in every case in 

 which decisive observations have been made, 

 the eclipsing companion is not a dark body, 

 but a self-luminous star. The maximum 

 difference in brightness between the com- 

 ponents of a pair is about four magnitudes, 

 which would not be considered very great in 

 the case of a visual binary. The secondary 

 minimum, due to the eclipse of the com- 

 panion by the principal star, is almost always 

 of observable depth, and has been found when- 

 ever properly looked for. 



In most of the cases which have so far been 

 studied, the faint companion is of greater 

 diameter than the brighter primary; but it 

 is very doubtful whether this represents the 

 general rule among close binary systems, for 

 pairs in which the fainter star is the smaller 

 can at best show but a small range of va- 

 riation, and few of them are likely to be dis- 



covered, especially as observers give the pref- 

 erence to the stars of large range of variation. 



It appears however to be an invariable rule 

 that the faint companion is always much 

 redder than the primary — which is clear evi- 

 dence that its faintness is due to lower tem- 

 peratiu-e. This is confirmed by Miss Cannon's 

 direct observation, that in U Cephei, a typical 

 system of this class, the spectrum of the 

 brighter star is of class A, while that of the 

 much fainter one which totally eclipses it at 

 minimum is of class K. Two stars separated 

 by a space equal only to their diameters, and 

 doubtless of common origin and equal age, 

 may therefore differ as widely in spectral type 

 as do Sirius and Arcturus. 



These systems offer the only direct method 

 at present available for finding the relation 

 between the color index and surface bright- 

 ness of a star. The investigation has so far 

 been complicated by uncertainties regarding 

 the color equation of both the visual and 

 photographic observations of such faint ob- 

 jects. When this difficulty has been sur- 

 mounted, as it soon should be, there is a good 

 prospect of being able to determine the sur- 

 face brightness of a star from its color index — 

 which would clear the way for determining 

 the linear diameters of all stars of known 

 parallax, and the angular diameters of all the 

 stars in the heavens. 



Of no less imfwrtance is the information 

 which eclipsing binaries alone can give about 

 the densities of the stars. The numerous 

 stars of this sort, and of spectra A and B, 

 are remarkably similar in density, averaging 

 about one sixth the density of the sim. The 

 fewer pairs of spectra F, G and K show a 

 much greater range of density — some being 

 denser than any of the stars of the first type, 

 while others are of very low density, in a few 

 cases less than 1/100,000 of the sun's. There 

 is also strong evidence tliat the large faint red 

 components of the systems mentioned above 

 are less massive, and hence much less dense, 

 than the small, bright, primaries of "earlier" 

 spectral type. The bearing of these facts on 

 stellar evolution is obvious. 



By combining this knowledge of the densi- 



