70 REPORT OF THE SECRETARY. 



vexatious, has been reduced to such a degree that its whole amount for a week's time 

 often does not exceed 2 centimeters, or the amount which it had in the early experi- 

 ments in a single minute. This result, it must be remembered, is reached with the 

 galvanometer adjusted to such a sensitiveness that a millimeter deflection on the 

 scale corresponds to six ten-billionths of an ampere in the galvanometer circuit. Not 

 only is the drift thus practically eliminated, but the small deflections from accidental 

 causes which heretofore so obscured the deflections corresponding to real absorption 

 lines, while still very numerous, are now reduced to an amplitude rarely exceeding 

 two-tenths of a millimeter. Occasional deflections of a little greater magnitude occur, 

 but not often, and these can be readily eliminated in a comparison of curves. The 

 deflections now remaining are almost wholly due to ground tremors not entirely 

 eliminated by the system of suspension of the galvanometer. 



This large decrease of drift and incidental deflections brought out the defects in 

 the construction of most bolometers hitherto in use, and on analysis it was found 

 that the enhanced "drift" when these were employed was due to the hard rubber 

 insulation, and that small accidental deflections were caused by imperfect fastening 

 of the strips at the ends. A bolometer frame of metal with mica insulation was filled 

 at the observatory and has proved very satisfactory. 



Another source of drift not before noticed was found to be the falling of the spec- 

 trum upon the bolometer case during a run from A to £1, and it was entirely removed 

 by a suitable diaphragm. 



These observations on the drift and false deflections in the record may be suitably 

 closed by the remark that under your instructions of March, 1896, it has been the 

 invariable practice to make a short record curve, with- no heat falling on the bolom- 

 eter, both at the beginning and end of each holograph, to show the instrumental 

 conditions prevailing. Such records were occasionally taken in former years, and 

 were included in the report of the number of holographs taken. So many have this 

 year been taken, however, that it seemed misleading to include them in the count, 

 and they are accordingly entirely neglected in the table of photographs given below, 

 and only those curves which represent some part of the spectrum are there counted 

 as holographs. 



After the approach of hot weather rendered it impossible to continue uniform 

 temperature conditions, the work of taking holographs was again discontinued, and 

 extensive quantitative experiments were prosecuted to determine the magnitude of 

 the probable error arising from the many sources attendant upon the complicated 

 bolometric process. The results of these experiments revealed in several instances 

 greater error than was suspected, and led to important changes in the apparatus 

 which are not at the close of the period covered by this report entirely completed. 



A detailed accounb of the methods employed in determining the various sources of 

 error must be postponed till the appearance of the forthcoming publication; but it 

 will be proper in this place to give a brief summary of the investigations. 



Laying down as in the report for last year ending June 30, 1895, an accuracy of 

 one-tenth millimeter in position of a deflection upon the holographic curve (corre- 

 sponding under the convention of speeds adopted to six-tenths second of arc in 

 the spectrum, or to three-tenths second of arc on the circle) as the limit of accuracy 

 desired in the final results, this amount of error will be treated as in a sense the unit 

 of error in this discussion : 



I. Sources of error in the optical apparatus. 



1. Apparent displacements of the slit — 



a. Momentary, from ground tremors and from air currents of different density 

 Displacements now negligible, as a result of recent improvements. 



b. Periodic, from temperature and other changes in the supports of optical 

 iipparatus. Important errors from this cause were within the last month 

 covered by this report eliminated. 



