The dispersion employed in any normal map of the spectrum may be
expressed by its scale, that is, by the ratio of the wave length as
represented to the actual wave length. It will be more convenient to
divide these ratios by one million, to avoid the large numbers otherwise
involved. If one millionth of a millimeter is taken as the unit of wave
length, the length of this unit on the map in millimeters will give the
same measure of the dispersion as that just described. When the map is
not normal, the dispersion of course varies in different parts. It
increases rapidly toward the violet end when the spectrum is formed by a
prism. Accordingly, in this case the dispersion given will be that of
the point whose wave length is 400.

This point lies near the middle of the photographic spectrum when a
prism is used, and is not far from the H line. The dispersion may
accordingly be found with sufficient accuracy by measuring the interval
between the H and K lines, and dividing the result in millimeters by
3.4, since the difference in their wave lengths equals this quantity.
The following examples serve to illustrate the dispersion expressed in
this way: Angstrom, Cornu, 10; Draper, photographer of normal solar
spectrum, 3.1 and 5.2; Rowland, 23, 33, and 46; Draper, stellar spectra,
0.16; Huggins, 0.1.

The most rapid plates are needed in this work, other considerations
being generally of less importance. Accordingly, the Allen and Rowell
extra quick plates have been used until recently. It was found, however,
that they were surpassed by the Seed plates No. 21, which were
accordingly substituted for them early in December. Recognizing the
importance of supplying this demand for the most sensitive plates
possible, the Seed Company have recently succeeded in making still more
sensitive plates, which we are now using. The limit does not seem to be