172 



SCIENCE. 



[N. S. Vol. XII. No. 292. 



work of a 12-incli telescope only 5- to 7-fold. 

 A 36-inch reflector, suitably mounted, and in 

 a dry climate, would probably be more effi- 

 cient thana 36-inch refractor in this problem. 

 Success in determining stellar velocities 

 depends upon a great many elements, of 

 which the following were all that time per- 

 mitted a mention. 



I. A suitable compromise between the 

 various optical and mechanical features — 

 many of which conflict — to secure a power- 

 ful and efiQcient spectrograph. 



II. Suitable prisms and lenses. The re- 

 quirements demanded by the line of sight 

 tests are so severe that satisfactory lenses 

 and prisms are difficult to obtain. Since a 

 change to a new prism-train usually re- 

 quires many changes in the mounting, the 

 prisms should be constructed and thor- 

 oughly tested before the rest of the instru- 

 ment is designed. 



The day for using the ordinary double 

 camera lens is past. Even the triple lenses 

 do not fulfill the difficult spectrograph's re- 

 quirements. 



III. Absence of differential flexure dur- 

 ing the exposure time : the optical angles 

 must remain constant. This is very largely 

 a problem in mechanical design. The 

 elimination of appreciable flexure effects 

 during long exposures is difficult, but not 

 impossible. Existing spectrographs, in- 

 cluding those just now completed, are 

 wrongly designed, in that they are sup- 

 ported only at one end. The heavy prism- 

 end projects out into space, unsupported. 

 The spectrograph is held out ' at arm's 

 length,' so to speak. Now the spectro- 

 graph is an instrument complete in itself, 

 and need not be rigidly connected with the 

 telescope tube. A framework connected with 

 the telescope could be arranged to support 

 the spectrograph, in position, both near its 

 upper and its lower ends ; care being taken 

 that strains in the supporting framework 

 should not be communicated to the spectro- 



graph. This form of support would permit 

 a more economical distribution of material 

 in the spectrograph, and should decrease 

 the flexure effects several fold. 



IV. The Mill's spectrograph has been en- 

 closed in a wooden box case, lined with 

 thick hair felt, carrying on its interior sur- 

 face about 30 feet of German silver wire- 

 As the temperature within the case falls, a 

 storage battery is connected with the re- 

 sistance wire. The heat generated tends to 

 keep the instrument, and especially the 

 prisms, at a constant temperature — a most 

 important desideratum. 



V. An accurate method of guiding dur- 

 ing the exposure. 



VI. The use of the finer dark lines 

 throughout the spectrum, rather than of 

 the heavier lines which have heretofore 

 been the ones measured. 



VII. Long experience in making the 

 measures. 



An exposure of one hour is required for 

 a star of the 5th photographic magnitude, 

 as given in the Draper Catalogue. 



At the Lick Observatory two or more 

 plates have been secured for each of about 

 300 stars. For the best stars, such as Po- 

 laris, the probable error of a single deter- 

 mination of velocity is about one-third of a 

 kilometer per second. For such stars the 

 plates seem to be essentially perfect, and to 

 secure greater accuracy it appears that a 

 more powerful spectrograph is required. 



Even with perfect plates the observer 

 must be on his guard against systematic 

 errors arising from changes in his personal 

 habit of measurement. 



Of the 300 stars observed, 22 have been 

 found at the Lick Observatory to be binary, 

 in addition to three previously detected 

 from the same list by Belopolsky. This 

 leads us to the important conclusion that 

 one star in twelve is a spectroscopic binary. 

 And it is probable that others of the stars 

 observed are binaries as yet undetected. 



