OF ARTS AND SCIENCES. 247 



of temperature of the prisms between successive exposures. It is too 

 great to be attributed to motion of the star in the line of sight. 



The measurements have all been reduced to a common zero point 

 by subtracting the reading of the H line from each, and adding 5.000 

 to avoid negative readings. The H line was selected, since it is com- 

 mon to nearly all the spectra, and is so well marked that an accurate 

 setting can be made on it. This more tlian compensates for its 

 breadth, which renders it an inconvenient starting-point in the solar 

 spectrum. Any correction for error in setting on this line can better 

 be applied to the final results than to the original measurements. 



Since Jupiter and the Moon shine by reflected light, it may be 

 assumed that their spectra are identical with that of the Sun, and that 

 the wave-lengths of the lines in their spectra may therefore be taken 

 from a map of the solar spectrum. On account of errors in the rela- 

 tive prominence of different lines, errors in identification may occur 

 in comparisons with maps made by hand. The photograph of the 

 diffraction spectrum by Dr. Henry Draper (Amer. Journ. Sci., CVI. 

 401) has therefore been used as the standard to which these measure- 

 ments are referred. For wave-lengths too great to be contained in 

 this map, Angstrom's Map (^Recherches stir le Spectre Solaire, Berlin, 

 1869) has been employed. Comparison has also been made with the 

 map of Cornu (Ann. de Vlllcole Normale Superieure, Series II., Vol. III.), 

 and those of Vogel {Piihlicationen des astrophysiknlischcn Observa- 

 toriums zu Potsdam, I. 13,1, II. 83). The unit employed throughout 

 is the ten-millionth of a millimeter, and the results are only carried to 

 single units. The best maps of the solar spectrum differ by as much 

 as one or two of these units in the ultra violet portion, and a greater 

 precision than this in stellar spectra is obviously at present unattain- 

 able. 



To pass from screw readings to wave-lengths points were constructed 

 with the readings of the spectra of the Moon on Plate 1 9 as abscis- 

 sas and the corresponding wave-lengths as ordinates. This plate was 

 selected since the lines appear to be more numerous and better defined 

 than in the other photographs. A smooth curve was then passed through 

 these points. The Avave-lengths corresponding to each half-turn of the 

 screw according to this curve are given in Table II., columns 1 and 2. 

 The relation between these quantities may be very closely represented 

 by the formula, A. ^ 3672 -f- 39 n -|- 4«^ in which A denotes the 

 wave-leniith. and n the number of turns of the screw. The values 

 computed by this formula are given in the third column, and the ob- 

 served minus the computed values in the fourth column. 



