OF ARTS AND SCIENCES. 289 



paratively easy ; for the solar spectrum can be used to identify the 

 lines of the metals, and to ascertain their wave-lengths. It is only in 

 the extreme infra red region and in the ultra violet that such observa- 

 tions become difficult. In these regions we must trust to photography 

 to reproduce by long exposures of the sensitive plate the feeble lines 

 of metals which may manifest themselves there. In the infra red 

 region, as far as wave-length 10,000, it is possible to photograph the 

 solar lines, and we can compare the spectra of such metallic lines 

 as may exist between the A line and the limit 10,000 with the solar 

 spectrum. Beyond this limit, and beyond wave-length 2800 in the 

 violet, the solar spectrum disappears, and the problem of measuring 

 the wave-length of metallic lines which extend beyond these limits 

 becomes a difficult one. 



Besides the resolution of the difficulty of measuring the wave- 

 lengths of the invisible rays of light with proper accuracy, the meas- 

 urement of such wave-lengths is destined to prove a crucial test for 

 various theories which must arise in the progress of physical science. 

 The lines of the metals are exceedingly numerous in the ultra violet 

 region, far more so than in the infra red region. If there are any har- 

 monic relations between the wave-lengths of the spectra of metals, 

 it is here that one might expect to observe such relations. Indeed, 

 Professor Griiuwald of Prague has lately enunciated a remarkable 

 hypothesis upon the relations between the wave-lengths of so-called 

 elements, and finds in the observations of various observers in the 

 ultra violet a strong confirmation of his hypothesis. In any theo- 

 retical work upon the grouping of spectral lines, it is of fundamental 

 importance that the wave-lengths of the lines should be determined 

 with as great accuracy as possible. The coincidence of metallic lines 

 with solar lines is at the best a doubtful piece of evidence. This 

 evidence is of moment only when the number of coincidences becomes 

 great, and is accompanied by characteristic grouping. A mistake of a 

 wave-length in the question of position is sufficient to destroy the 

 support which the author of any hypothesis might claim for it. 



Conditions for Accuracy of Measurement. 



All measurements of wave-lengths hitherto published have been 

 made by the old method of angular measurements with a spectrometer. 

 We say old, for the use of Rowland's concave grating with its peculiar 

 mounting must be characterized as a new method and a new departure 

 in measurements of wave-lengths. The observation of wave-lengths 

 of metallic spectra by the eye is most laborious, and the photo- 

 VOL. XXIII. (n. s. xv.) 19 



