532 



SCIENCE 



[N. S. Vol. XXXIV. No. 877 



be constant and was made the basis of the experi- 

 ment. The apparatus devised consists of a carbon 

 rod, mounted in a spindle and made to rotate 

 between the poles of a right-angled are fed by a 

 20-ampere direct current at 40 volts. The end of 

 the rod is bored to receive charges of the metal 

 up to .5 gram. The inside of the bore is projected 

 by a lens upon the slit of a concave grating 

 spectrograph. 



The calcium was introduced into series of tubes 

 in two different ways: (1) The tubes were filled 

 with a supersaturated solution of calciimi chloride 

 diluted in multiples of two; (2) uniform slivers 

 of metallic calcium each of weight about 2 mgr. 

 were weighed out into charges of from 5 to 

 100 mgr. and placed in the heating tubes. Each 

 tube was charged only once, and two to five ex- 

 posures of 20^ each were made with it. About 800 

 photographs were taken with 225 tubes. The tem- 

 perature of the bore was determined from a curve 

 constructed with currents and melting points of 

 refractory metals as arguments. The approximate 

 effect of the calcium impurities in the carbon tube 

 and the air was obtained and allowed for. 



Conclusion: The lines g, X 4,227, and H, X 3,968, 

 require for reversal vapor-densities that are about 

 1/7 and 3/2 as great, respectively, as that neces- 

 sary for a reversal of K, X 3,933, when the vapor 

 is observed at a temperature of 2500° C. and at 

 atmospheric pressure. 



(To be published in the Astrophysical Journal.) 



Sadial Velocity of Salley's Comet as Derived 



from a Spectrogram: Edwin B. Erost. (Read 



by Dr. Slocum.) 



The radial velocity of Halley's comet on May 

 24, 1910, as determined by the writer from the 

 displacements of the Eraunhofer lines, was 

 + 55 km. per second. The velocity calculated 

 from the ephemeris of the comet agreed with this 

 value within a kilometer. 

 Note on the Magnitudes of the Stars in the 



Cluster, Messier 3: S. I. Bailey. 



Certain globular clusters appear to be composed 

 of two groups of relatively bright and faint stars, 

 with few stars of intermediate magnitudes. This 

 appearance has been noted by Pickering, Palmer, 

 Eath and the writer. 



A recent plate taken by Eitchey on Mount Wil- 

 son with the 60-inch reflector, having an exposure 

 of four hours, shows extremely faint stars. An 

 enlargement of this plate, loaned for this purpose 

 by the acting director of the Solar Observatory, 

 was used for the present discussion. The faintest 



stars shown appear to be of magnitude 21.5. The 

 number of stars is probably not less than 30,000. 

 An enlargement was made with a reseau dividing 

 the photograph into 1,280 squares. The magni- 

 tudes of all the stars in 160 of the squares were 

 determined by comparison with the sequence se- 

 lected for the study of the variables in the cluster. 

 The number of stars thus measured was 2,542. 

 The scale of magnitudes as extended to the fainter 

 stars may be somewhat in error. 



Grouped by magnitudes, the number of stars of 

 each magnitude from 13.5 to 21.5 is as follows: 

 2, 6, 16, 31, 43, 221, 576, 1,071 and 573; the 

 total light derived from the groups in percentages 

 are: 11, 20, 15, 11, 6, 12, 13, 10 and 2. The 

 light of stars fainter than magnitude 21.5, if 

 such exist, is probably negligible. The stars of 

 the four groups, 13.5 to 16.5, give 57 per cent, of 

 the light of all the stars measured, and those of 

 the four groups, 18.5 to 21.5, 37 per cent. The 

 group at 17.5 may be regarded as intermediate, 

 and here the light is a minimum. The total light 

 of the 2,542 stars whose magnitude was deter- 

 mined is approximately equal to that of a star of 

 magnitude 10.4. 



From the relation between the known magnitude 

 of the whole cluster regarded as a single object, 

 and that portion studied in the present paper, 

 assuming that the distribution of the stars by 

 magnitudes is the same throughout, the whole 

 number of stars in the cluster may be derived. 

 This number will be in error, if the scale of mag- 

 nitudes employed is incorrect. By an independent 

 determination of the number of stars by count, 

 the error in the scale of magnitudes, if any, may 

 be determined. 



Tables of Effective Wave-lengths of Lines in 

 Stellar Spectra: S. Albrecht. (Eead by Mr. 

 W. E. Harper.) 



This paper is a report of progress in the work, 

 begim in 1906, of determining the effective wave- 

 lengths of spectrum lines in stellar spectra. The 

 writer has determined tables of wave-lengths for 

 each of the main divisions of the Draper classifica- 

 tion (as modified by Miss Cannon) from types B 

 to M, both inclusive. The part of the spectrum 

 covered is from X 4,236 to X 4,655. The work is 

 based on measures made, while at the Lick Ob- 

 servatory, on spectrograms taken with (a) the 

 Northern Mills three-prism spectrograph having 

 X 4,500 central with Ti-spark comparisons; (6) 

 the Southern Mills three-prism spectrograph, 

 X 4,450 Central and Ti-spark comparisons; (c) the 



