in the Atmospheres of certain Fixed Stars. 



190 



pressure against the minimum number of necessary rigid points of 

 support. 



The object-glasses of the spectroscope were made by Brashier, and 

 are all excellent. The three dense prisms were also made by Brashier ; 

 their definition is very fine, but the glass is rather yellow in colour, and 

 produces great absorption of rays more refrangible than H y . The 

 single prism, by Steinheil, gives excellent definition, and the glass is 

 much whiter than in Brashier's prisms. The optical constants of the 

 prisms have not yet been determined. 



The mounting was constructed by the Cambridge Scientific Instru- 

 ment Company to my designs, in the most careful and satisfactory 

 way. I am greatly indebted to Mr. Horace Darwin for much care in 

 supervision of the work and for some very ingenious and important 

 improvements in detail which he carried out. 



Above all I am indebted to Mr. H. F. Newall, who has taken infi- 

 nite trouble in making and testing the permanent adjustments of the 

 instrument and in supervising the arrangement of its final details. To 

 him I owe the fact that the instrument arrived at the Cape practically 

 in perfect adjustment and ready for work. After a series of prelimi- 

 nary focussing trials by Ne wall's method,* a number of photographs of 

 star spectra were made with the three-prism box and long telescope. 

 The present paper deals chiefly with the results of measures of a 

 photograph of the spectrum of ft Crucis and of a comparison iron spec- 

 trum obtained on March 15. The plate was exposed to the comparison 

 spectrum of iron immediately before and immediately after the expo- 

 sure for the star spectrum. (See Plate 8.) 



Lines of the iron spectrum cover the whole exposed length of the 

 plate from Fe X 4187*99 to 4563*99, the linear interval on the plate 

 between these lines being 70*367 mm. 



As a preliminary step, the intervals between successive pairs of iron 

 lines were measured with the micrometer of the old Eepsold astro- 

 photographic measuring apparatus.! If 



As is the interval between two adjoining lines in terms of revolu- 

 tions of the micrometer screw, 

 X 1 and X 2 their respective wave-lengths, 

 Kn = i(X\ + X 2 ) and AA. = X 1 - X 2) 



I found, to my surprise, on computing AA/As for many different values 

 of X mi and plotting these on millimetre paper with AA/As as ordinate 

 and X m as abscissa, that within the limits of error of plotting and 

 observation the resulting curve was practically a straight line : in other 

 words, the screw values can be represented by 



* ' Monthly Notices, E.A.S.,' vol. 57, p. 572. 



f Described by Bakhuyzen, 4 Bulletin du Congres Astrograpliique,' toI. 1, p. 169. 



