ON THE MEASUREMENTS OF WAVE-LENGTHS. 
89 
The curve is a very regular one, and might be drawn from a very few accurate 
points. 
The value of the error of -gFoobh °f an inch in measurement for different parts of 
the curve for prism spectra, in terms of wave-length, is given in the following table :— 
Portion of curve between the 
wave-lengtbs given. 
4780 to 4440 
4440 to 3990 
3990 to 3G00 
3600 to 3200 
3200 to 2800 
2800 to 2400 
2400 to 2020 
Value of error in tenth-metres 
of 2 -Aj o-th of an inch. 
IT to PO 
1*0 to 07 
07 to 0-5 
0-5 to 0-3 
0-3 to 0-25 
0-25 to 076 
076 to 0-08 
Well defined lines can with certainty be measured to oo ofh inch ; for weak and 
indistinct lines the error of measurement amounts to Tc/ooth, but even with this error 
most satisfactory determinations can be obtained from the curve. 
Great care is necessary in measuring these prismatic spectra. It is difficult to 
accurately adjust the cross hairs to the lines, owing to the latter being somewhat, 
though very slightly, curved; consequently the readings may not be quite true with 
reference to the tin and cadmium lines, but either a little too small or a little too 
large. 
The danger of this error is reduced to a minimum if the photographs are taken 
with the electrodes rather close together, or where only strong lines are to be observed 
by making the spark cross the slit. 
In the list of lines given in the tables for each metal, the greatest care has been 
taken to eliminate those due to foreign metals. This has been done as completely as 
possible by taking a large number of photographs of the spectra of several elements 
for the purpose of comparison. In cases of doubt, spectra of very strong solutions of 
the purest salts were taken. It has been found that lines due to impurities generally 
present a decidedly different character to any of the lines due to the metal under 
examination. This will be referred to again further on. The wave-length of every 
line identified with a given metal has been determined either from the grating or 
prismatic spectrum. In the present state of our knowledge this has been thought 
important. The short lines of cadmium and zinc, for instance, have not been 
measured by previous observers; they are, however, a very characteristic feature in 
each spectrum. 
Certain weak lines do not appear in the diffraction spectra, but are plainly visible 
in those obtained by means of a prism. Such lines have been measured with the 
interpolation curve, and are distinguished in the following tables by their wave¬ 
lengths being printed in italic figures. 
MDCCCLXX XIV. N 
