46 
DR. A. SCHUSTER ON THE SPECTRA OF METALLOIDS. 
No measurements are of any value unless they are given in wave-lengths. It has 
been suggested to use the air lines as reference lines, between which we might inter¬ 
polate any known line. It seems to me that a much better choice might be made. 
The air lines are not spread evenly through the spectrum ; they are by no means sharp 
under high dispersion, and their wave-lengths have only been obtained by interpolation 
I do not think that these possess the requisite degree of accuracy to be used again as 
reference lines. It should be our endeavour to make our measurements as homogeneous 
o 
as possible, and as long as Angstrom’s solar maps is the standard work for wave-length 
measurements we should refer all our measurements to that map. Thalen has 
adopted this plan in his excellent series of measurements, but the method which he 
used presents serious difficulties when the light to he examined is weak. 
I have used as reference lines the bright lines of such metals as are present in the 
sun, the wave-lengths of which can therefore be directly found on Angstrom’s map. 
The lines of iron are very useful in this respect. 
The interpolation formula which is usually employed is based on the fact that the 
difference in refractive indices of two lines varies nearly as the difference between the 
inverse squares of the wave-lengths. If reference lines can be found sufficiently close, 
the formula does pretty well ; but I find that for the dispersion which I have used, 
the formula is no longer applicable when the distance between the reference lines is 
about ten times the distance between the two sodium lines. As it is not always 
possible to find good reference fines within that distance, I have generally used three 
reference fines, and taken the inverse fourth powers of the wave-lengths into account. 
I am not quite certain that even then the errors of interpolation did not exceed, in some 
cases, slightly the errors of observation. The reason why the ordinary interpolation 
formula cannot be used, does not fie so much in the fact that the law of the inverse 
square does not hold accurately for refractive indices (although this, of course, is also 
true), but that, except in automatic spectroscopes, our measurements are not propor¬ 
tional to refractive indices. 
It is difficult to say to what degree of accuracy measurements are made, yet I trust 
that, except for the weakest fines, the error of my measurements will be seldom found 
to exceed 0‘5 x th metre. 
After these remarks I pass to the description of the various spectra of oxygen. I 
distinguish four such spectra. - At the lowest temperature at winch oxygen becomes 
luminous it gives a continuous spectrum. As the temperature is raised, the continuous 
spectrum changes successively into two fine spectra, which, as I have already men¬ 
tioned, I call the compound fine spectrum and the elementary fine spectrum. It is 
one of the principal objects of this paper to show that these two fine spectra, which 
have been much confounded with each other, have a separate existence ; and that the 
generation of one involves the destruction of the other. The fourth spectrum is 
that which is seen at the negative pole. 
