EEPORT OF THE SECRETARY. 69 



that there will be sufficiently close correspondence between such false lines in all the 

 cylindrics employed, so that they will appear in the final composite. 



Of late another method of combining the results to be derived from the separate 

 curves has come into use at the observatory, which is briefly as follows: 



1. Each of the several observers independently superposes the set of curves under 

 investigation, and from the general appearance and size of the deflections selects 

 those which he considers to probably correspond to genuine absorption bands, and 

 marks them relatively to their importance and to the evidence of their genuineness. 



2. He then measures upon the comparator the positions of his selected deflections 

 on each curve separately without sight of the others. 



3. The results thus obtained are averaged, throwing out such deflections as appear 

 from the divergence of the measurements to be accidental. 



4. The mean results of the separate observers are compared and reduced, indicat- 

 ing the extent to which the observers agree with regard to the importance and verity 

 of the different deflections. 



In view of the limitations of the composite process, it is proposed that it be kept 

 subordinate to that just described, and only used in the preparation of a spectrum 

 map in which the exact place of each line appearing shall be independently deter- 

 mined. 



The multiplicity and injurious character of the accidental deflections continued 

 painfully evident throughout the holographic harvest season of the year (the 

 months of September, October, and November). Fine days were also frequently lost 

 after cold weather came on, owing to the rapid "drift" of the galvanometer needle, 

 caused by the changing temperature conditions incident to the warming of the 

 observatory. 



In these circumstances the work of taking holographs was suspended for some 

 months, and the following important improvements were, after many experiments, 

 adopted, which have separately and in concert been signally successful in reducing 

 these sources of error : 



1. In extension of the principle introduced at Allegheny, a battery of sixty storage 

 cells connected in parallel replaced the four cells formerly employed, and effected 

 great reduction of small accidental deflections of the galvanometer needle. 



2. A specially constructed rheostat, entirely surrounded by a water jacket and pro- 

 vided with a slide wire adjustment oi>erated mechanically from without, replaced 

 the resistance box before used to balance the bolometer strips. In this instrument 

 only a single pair of coils of the same resistance as the particular bolometer employed 

 is used. The number of contacts between dissimilar metals is reduced to a mini- 

 mum, and all are kept from fluctuations of temperature by a continuous current of 

 water. 



3. The galvanometer, formerly supported upon a system of flat stones and rubber, 

 has been suspended by three wires from a tripod, after the manner of Julius. 1 This 

 method of support is found to decidedly reduce the vibrations of the galvanometer 

 needle due to ground tremors. 



4. The double-walled room which surrounds the spectrometer has been extended 

 to include the galvanometer also; and thus temperature fluctuations about the gal- 

 vanometer control magnets havo been reduced, to the great diminution of the 

 " drift." 



5. The whole observatory has been equipped with an automatic system of temper- 

 ature regulation, and the heat is supplied both night and day from steam boilers in 

 the basement of the Smithsonian main building. So successful has this system 

 proved that during the cold months the whole fluctuations of temperature within 

 the spectrometer room are reduced to little more than a tenth of a degree. 



6. A new galvanometer needle has been constructed with a considerably higher 

 degree of astaticism and somewhat greater sensitiveness than the one before in use. 



As the result of the improvements mentioned above, the "drift," heretofore so 



1 Wiedemann's Annalen, 56, 151. 



