573 



NA TURE 



[October 28, 1922 



The Galactic System. 1 



By Dr. Harlow Shapley. 



II. 



IX the first pari oi tins article the main character- 

 ol the globular and open clusters were dis- 

 cussed, and it was shown how the determination of their 

 distances led to the proposal of extremely great dimen- 

 sions lor tin- galactic system. A theory of the origin 

 and structure of the Galaxy also seems to be indicated 

 by the observations. 



Objections to Proposed Scale of the Galaxy. 



Although the new values of galactic dimensions 

 have been widely accepted by astronomers, at least 

 qualitatively, they have been openly challenged by 

 some. Without questioning my values (which may 

 indeed have been unknown to him), Prof. Charlier 

 published a few years ago provisional cluster distances 

 that are of a wholly different order of magnitude from 

 those I derive. 8 He had, in effect, affiliated the 

 globular clusters with the local system of B stars. I 

 believe he has now accepted the larger values of the 

 distances. 9 



An extensive critical examination of my methods 

 and results has been made by Prof. Curtis. His 

 discussion and my reply have been published together 

 in Bulletin No. n of the National Research Council. 



Dr. Schouten has attempted to derive the distances 

 of clusters by assuming that the frequency of absolute 

 magnitude (the luminosity-curve) is the same in glob- 

 ular clusters and in the neighbourhood of the sun. 10 

 The method is questionable for several reasons : (1) 

 All spectral types are lumped together by Schouten 

 regardless of our present knowledge of the peculiar 

 relation of type and luminosity in the globular clusters. 

 (2) The giant-dwarf phenomenon is essentially ignored 

 in the method. (3) The observed luminosity-curves 

 in -lobular clusters do not conform with the law 

 assumed. 11 (4) It is certainly improbable that the 

 stage of evolution in any given cluster is closely com- 

 parable with the average of the many stages represented 

 by the heterogeneous mixture around the sun. (5) As 

 applied, the luminosity-curve method involves dan- 

 gerous extrapolation, for we know the frequency of 

 magnitudes for only the very brightest stars in 

 clusters. 12 



I in j the necessarily fragmentary luminosity-curves 

 for a few clusters, Schouten finds distances averaging 

 about one-eighth the values I have computed. 



Recently, Kapteyn and van Rhijn published a 

 valuable paper on the proper motions of Cepheid 

 variable stars of the short-period sub-type. 13 It 

 is generally accepted that Cepheid variables of long 

 period are giant stars ; and from the simultaneous 

 occurrence of the long-period and short-period Cepheids 

 in globular clusters, 14 I have assumed of course 

 that the short-period Cepheids, which occur most 

 frequentlv in clusters and serve in one of the methods 

 of estimating the distances, are also giant stars. Kap- 

 teyn and van Rhijn, on the other hand, have computed, 

 from the large values of the proper motion, small 



1 Continued from p. 547. 

 NO. 2765, VOL. I IO] 



distances for the cluster-type variables, and therefore 

 low luminosities. They conclude that Cepheid vari- 

 ables of this sub-type may be dwarfs, both near the 

 sun and in clusters ; and, by assuming that the long- 

 period Cepheid variables in clusters are abnormalities 

 and the short-period Cepheids are normal, they assert 

 that the clusters may be at less than one-seventh the 

 distances I place them. But Kapteyn and van Rhijn 

 appear to have overlooked the decisive factor that the 

 known radial velocities of these short-period Cepheids 

 are remarkably high 15 (much too high, apparently, 

 for the application of the method they use ; 16 and 

 therefore that the large proper motions they deduce 

 and the wide distribution in galactic latitude are 

 almost certainly the result of exceptionally high 

 velocities in space, rather than an indication of near- 

 ness and low luminosity. Similarly, the long-period 

 variables of spectral type M are giants at maximum, 

 are widely distributed in galactic latitude, and have 

 high space velocities. 



As I )r. Crommelin has hinted recently in the Obser- 

 vatory, 11 a sufficient answer to those who would 

 reduce the distance of clusters to one-fifth or one-tenth 

 the values proposed, is that apparently they do not 

 consider fully the dire consequences of such reduction 

 on a vast body of other astronomical data that is now 

 generally accepted. If the distances I give are not 

 greatly reduced or increased these troubles do not 

 arise ; all that we know of the colours, spectra, mag- 

 nitudes, and motions in the clusters, and of the clusters, 

 then fits in well with our general stock of astronomical 

 fact and theory. 



Before we knew much about the character of stars 

 in clusters we were not restricted by observation or 

 theory from placing the clusters at whatever distances 

 we liked. But now, if we alter the present distances 

 by the amount Curtis, Schouten, and Kapteyn and 

 van Rhijn suggest, we immediately set up peculiarities 

 and discordances in great numbers. For example, 

 among other difficulties evoked by such changes, we 

 would seriously question the general applicability of 

 the spectroscopic method of determining luminosities 

 and distances; 18 we would introduce confusion into 

 Russell's and Eddington's theories which now so 

 happily conform with physical laws and observational 

 results, in clusters as well as outside ; we would over- 

 throw the period-luminosity law of Cepheid variation. 

 Sooner or later it may be necessary to divide or multiply 

 by 1 -5 (Dr. Crommelin suggests 2) the distances I have 

 computed for the clusters ; any larger factor will entail 

 alterations elsewhere that now appear improbably large. 



It seems to me that a better line of attack on the 

 proposed scale of the Galaxy would be to question the 

 apparent magnitudes rather than the absolute lumin- 

 osities. The latter, as we have seen, are supported 

 by too much evidence of a varied nature to yield easily. 

 Moreover the values of the absolute luminosities for 

 the stars in clusters come within the range of our usual 

 experience, whereas the apparent magnitudes (and 

 distances) of cluster stars are quite different from those 

 of stars we ordinarily treat. 



