234 PROFESSORS G. D. LIVEING AND J. DEWAR 



The wave-lengths of this last group of lines are all too large by about 0'35 as com- 

 pared with BELL'S, and they seem to be affected by some common error, but whence it 

 arises we do not know. Our measures of lilies in the visible spectrum, where, of 

 course, there are fewer sources of error, agree very closely with BELL'S. Thus our 

 measures of the blue lines gave the figures for X 5085 '26, 4799*40, and 4677*59, and 

 these, when multiplied by the factor 1*00016, come very close indeed to BELL'S 

 numbers, which are 5086*09, 4800*15, and 4678*39. On the whole, we are inclined to 

 think that a really good goniometer with a plane grating and a telescope of moderate 

 dimensions, focal length 8 or 9 decimeters, will give extremely accurate wave-lengths, 

 while the greater angular aperture of such a telescope gives it a considerable advantage 

 in point of light over the concave grating used by BELL. The method of measuring 

 wave-lengths by the coincidences of lines in spectra of different orders, for which a 

 concave reflecting grating is admirably adapted, could not be easily applied to the 

 ultra-violet spectra of cobalt and nickel, because the lines are so crowded that the 

 overlapping of two or three spectra, all in focus together, would produce a complication 

 which it would be nearly impossible to unravel, except by dispersing the spectra in a 

 direction at right angles to the dispersion produced by the grating. The chromatic 

 aberration of our quartz lenses is a positive advantage in dissipating the light of the 

 spectra of those orders which are not under examimation. 



For the determination of the cobalt and nickel lines specimens of those metals were 

 prepared so far spectroscopically pure that the spark between fragments of the cobalt 

 showed none of the characteristic strong lines of nickel, and the spark between pieces 

 of the nickel showed none of the characteristic strong lines of cobalt. As the metals 

 after reduction were fused with an oxyhydrogen blowpipe in lime crucibles, they were 

 not free from all traces of other metals. For the arc lines much labour in the purifi- 

 cation of the metals would have been wholly thrown away, because a variety of metals 

 are present in the carbon electrodes as well as in the limestone used for crucibles. In 

 the arc, therefore, we used samples sold as " pure," and identified the lines either by 

 their coincidence with spark lines photographed at the same time through a part of 

 the slit, or by their making their appearance, or being notably strengthened, on the 

 introduction of the metal into the arc. The list of arc lines is much less complete 

 than the list of spark lines, because weak lines in the arc are more easily overlooked 

 in a photograph crowded with lines, and when noticed their origin is with difficulty 

 identified. The wave-lengths of the spark and arc lines which were not measured 

 directly by means of a grating were determined by interpolation from photographs 

 of refraction spectra. In the highest region the copper lines were used as lines of 

 reference in this interpolation. For the direct determination of the wave-lengths of 

 the nickel lines about 170 photographs were taken, measured, and the results reduced ; 

 for the cobalt lines about 200. In many cases several lines could be measured on the 

 same plate, but we have rarely been satisfied without getting two or more independent 

 measures of the deviation for each line, and in many cases the measures have been 



