346 Messrs. Trowbridge and Sabine on Wave-lengths 



wave-lengths can be determined by simple linear measure- 

 ment. Moreover, it is desirable, as we have said, that the 

 underlying spectra should be brought to the same focus as the 

 overlying ; for by this means we can compare the wave-lengths 

 of lines in the spectra of different orders, and halve our errors. 

 It is true that some confusion results from having the metallic 

 lines in the spectra of different orders photographed upon the 

 same plate ; but a little experience enables one to separate the 

 lines with comparative ease, and the gain in accuracy com- 

 pensates for the additional trouble. 



The apparatus which best answers the requisitions we have 

 pointed out is that of the concave grating of Rowland, with 

 its peculiar mounting, which has been fully described in the 

 Phil. Mag. [5] vol. xvi. (1883) p. 197. 



Objects of the Present Investigation. 

 The conclusion of the work of the Committee of the British 

 Association on the tabulation' of metallic spectra seemed to us 

 to require a survey of the work, which must be done in the 

 future in order to perfect and correct the work of the past. 

 We have therefore examined the tables given by the com- 

 mittee in order to see what lacunar could be supplied, and to 

 point out the directions for routine work which may afford 

 material for future generalizations. In the pursuance of this 

 work, we have been compelled to examine the accuracy of 

 measurements of wave-lengths hitherto made in the ultra- 

 violet. With the aid of the new Map of the Solar Spectrum 

 published by Professor Rowland it is very easy to determine 

 the wave-length of metallic lines in the visible spectra of 

 metals ; for it is merely necessary to photograph a portion of 

 the solar spectrum upon the same plate as that which receives 

 the spectra of the metals under consideration, and then to refer 

 to the published map. We have already remarked, that even 

 a superficial examination of hitherto published catalogues of 

 wave-lengths of metallic spectra will show that distinguished 

 observers differ in their determinations by one or two wave- 

 lengths. The task of remeasuring the wave-lengths of metallic 

 lines is a very great one, and approaches in character the 

 routine work now prosecuted in astronomical observatories in 

 the redetermination of star-places, the photometric intensities 

 of stars, and the classification of star-spectra. In our present 

 work we have confined our attention to ultra-violet spectra. 

 Since the solar spectrum disappears in the neighbourhood of 

 wave-length 2800, the task of identification of wave-lengths 

 becomes a very serious one. To replace the solar spectrum 

 we must refer the lines of metallic spectra to carefully 



