26 



NATURE 



[March 8, 19 17 



log-ues of Lockyer are discussed in relation to the 

 radial velocities, so far as they have been deter- 

 mined. For the ascending branch there are 57 

 such stars in the two catalogues, and for the 

 descending branch 47 stars. 



It should perhaps be recalled that the helium 

 stars constitute type B of the Harvard classifica- 

 tion, and are subdivided into classes Bo to Bg, 

 in accordance with variations in detail. In 

 Lockyer's system, seven classes of helium stars 

 are recognised, three on the ascending branch, 

 one at the apex, and three more on the descending 

 branch of the temperature curve, thus :• — 



Alnitamian 

 Crucian Acheinian ^ 



Taurian Algolian 



Kigelian Markabian y 



It was found by Campbell that after eliminating 

 the apparent velocity due to the sun's motion, 

 the helium stars showed a systematic positive 

 (receding) velocity of +4*07 km., the apex of the 

 sun's way being taken as 0=270°, 5= +30°. 

 This systematic error, which was designated 

 "K" by Campbell, has not yet been satisfactorily 

 accounted for, but it is such as would arise if, in 

 the helium stars, the lines were subject to a pres- 

 sure effect which caused them to be displaced 

 slightly to the red side of their normal positions; 

 the effect of such a displacement would clearly be 

 to superpose on the real radial motions a receding 

 velocity of about 4 km. for all the stars, irrespec- 

 tive of their positions on the celestial sphere. The 

 magnitude of K, and the sun's velocity, are deter- 

 mined on the supposition that, in the mean, the 

 stars are at rest with respect to the stellar system. 



Forming the helium stars on the ascending 

 branch of Lockyer's series into one group, and 

 those of the descending branch into another, Dr. 

 Herassimovitch proceeds in the usual way to deter- 

 mine the sun's velocity and the K term for each 

 group, using equations of the form 



VoCOS<^ + K = V, 

 where Vq is the velocity of the sun, ^ the angular 

 distance of the star from the apex of the sun's way, 

 V the observed radial velocity reduced to the sun 

 by correction for the earth's orbital motion, and 

 K the residual velocity. In each group, K is thus 

 the mean algebraic residual after eliminating the 

 solar motion. The results are as follows : — 



From 57 stars on the/K = + 6-32 km. ±1-50 km. 

 a.scending branch \ ¥„= -20 84 km. ±2-40 km. 

 rom 47 stars on the/K =+ 1-17 km. + ii36 km. 

 descending branch \ V,= -20-03 km. ±2-29 km. 



Thus, while the resulting velocity of the sun is 

 almost the same for the two groups, the values of 

 K are strikingly different. For the descending 

 branch, in fact, K almost disappears, while for the 

 ascending branch its value is considerably in excess 

 of that found by Campbell from all the helium 

 stars taken together. Lockyer's diflferentiation of 

 the ascending and descending branches thus re- 

 ceives substantial corroboration. 



It was already known from the work of Camp- 

 bell that the groups of stars giving the largest 

 NO. 2471, VOL. 99] 



values of K (i.e. the Harvard classes B, K, and 

 M ') are among the most distant, and Dr. 

 Herassimovitch has therefore further discussed, 

 the ascending and descending groups in relation 

 to the mean parallaxes. Applying Kapteyn's 

 formulae, it results that for the helium stars on 

 the ascending branch the probable mean parallax 

 is o'oo5'' +o"ooo9", and for the descending branch 

 o"oi2" +o'oo3o". By the same process, Camp- 

 bell has found for the Harvard classes B0-B5 a 

 mean parallax of 0006", and for the classes 

 B8— B9 a mean parallax of o'oiag", which are 

 nearly identical with the values now found for 

 the ascending and descending stars. A pre- 

 dominance of B0-B5 stars on the ascending 

 branch, and of B8— Bg stars on the descending 

 branch, would thus account for the observed differ- 

 ence in the mean parallaxes. But this cannot 

 account for the whole difference, for although; 

 there are actually a greater number of B8— B9 

 stars on the descending than on the ascending 

 branch of Lockyer's curve, the excess among the- 

 stars here considered is only four. It would seem,, 

 then, as in the general case, that the more distant 

 group, of helium stars gives the larger value of 

 the K term, and it is interesting to find that 

 Lockyer's criteria for spectroscopic classification 

 have so successfully withstood this further test.. 



Dr. Herassimovitch has also investigated the 

 ascending and descending groups in relation to- 

 the magnitudes of the stars involved. Omitting 

 those at the summit of the curve, there are 155 

 helium stars available for this part of the inquiry, 

 and the figures show that the stars on the de- 

 scending branch are in general fainter than those 

 on the ascending branch. This difference cannot 

 be explained entirely by the excess of classes- 

 B8— Bg on the descending branch, because within? 

 the limits of a given Harvard sub-group, say 

 B3— B5, there is the same increase in the number 

 of faint stars on the descending branch as when- 

 all the B stars are taken together. When cor- 

 rection is made to absolute magnitudes, by apply- 

 ing the mean parallaxes previously deduced, it 

 also appears that the stars of the descending 

 branch are in general fainter than those of the 

 ascending branch which fall in the same Harvard 

 sub-class, and are therefore presumably at the 

 same stage of temperature. 



Stars at the same heat level on opposite sides- 

 of the temperature curve probably have the same 

 intrinsic brightness, and if it be assumed that 

 the average masses are equal, it would follow 

 that the stars on the ascending branch must in 

 general be of greater volume and lower density 

 than those on the descending branch. This is 

 precisely the physical difference which is de- 

 manded by Lockyer's hypothesis, and also by 

 that of Russell, and it may reasonably be sup- 

 posed capable of explaining the soectroscopFc 

 differences which have enabled Lockver to sort 

 out the two classes. 



The nature of the K term remains obscure. If 

 the greater brightness of the stars of the ascend- 



1 K is practically zero for stars of classes A, F, and G. 



