ON THE ULTRA-VIOLET SPECTRA OF THE ELEMENTS. 235 



made in more than one order of the spectra. Owing to difference in the strength of 

 the spectra on the two sides of the normal to the grating, it sometimes happened that 

 faint lines appeared in one of a pair of photograplis, that is on one side of the normal, 

 but not on the other. In such cases, when the reading of the circle for the normal 

 position of the grating (which is liable to vary with every adjustment of focus) could 

 be accurately determined by the help of stronger lines, which could be measured in 

 both photographs, the wave-lengths of the faint lines have been calculated from the 

 deviation as measured on one side of the normal. Such a measurement, though it 

 gives a valuable result, is plainly not quite independent. 



From the close chemical relationship between cobalt and nickel we should have 

 expected that their spectra would closely resemble one another. In regard to the 

 large number of lines which they exhibit, they certainly resemble each other, and 

 resemble iron ; and the resemblance goes a little further, inasmuch as the lines of all 

 three spectra are much more crowded in certain regions than in others, and the 

 crowded regions are approximately the same for all three. But, beyond such a general 

 resemblance, we have been unable to trace any definite correspondence in the spectra. 

 The number of lines of cobalt which according to our measurements have wave-lengths 

 identical with those of nickel lines is small, but in a record of 580 lines of cobalt and 

 400 of nickel it would be surprising if there were not many close coincidences, and, in 

 fact, we note forty-six cases where lines of cobalt do not appear to differ in wave-length 

 from lines of nickel by more than a tenth of a tenth-metre. Now, if the cobalt lines 

 were uniformly distributed over the whole region mapped and the nickel lines distri- 

 buted at random amongst them, it would be an even chance that twenty-six nickel 

 lines would not be more than a tenth of a tenth-metre distant from the nearest cobalt 

 lines. A glance at the map will, however, show that the cobalt lines are by no means 

 evenly distributed, and in the regions where they are most closely packed the nickel 

 lines are also for the most part closely packed. Hence, the chance of merely accidental 

 coincidences is very much greater than that above mentioned ; and we find that in 

 the region between the wave-lengths 2250 and 2550 the number of lines of the two 

 metals, which, as measured, are not more than a tenth of a tenth-metre distant from 

 one another is twenty-five, more than half the whole number of coincidences to that 

 degree of approximation. On the whole we are unable to conclude that the coin- 

 cidences are more than fortuitous. It should be observed that we have not yet had 

 the opportunity of comparing the spectra of the two metals photographed on the 

 same plate with high dispersion. On many of our plates the iron spark has been 

 photographed simultaneously with the spectrum of another metal, and we have noted 

 on the table of wave-lengths the cases in which these photographs show an unresolved 

 coincidence between an iron line and a cobalt or nickel line. These photographs, 

 however, have all been taken with a prismatic spectroscope, and it is probable that a 

 higher dispersion might resolve some of these coincidences. 



2 H 2 



