er, and to determine the consequent displacement of the
fringes. E. Mach and J. Arbes have used a method depending on total
reflection (Drude's _Theory of Optics_, p. 394).

[Illustration: FIG. 4.--Anomalous Dispersion of Sodium Vapour.]

A very remarkable example of anomalous dispersion, which was first
observed by A. Kundt, is that exhibited by the vapour of sodium. It
has not been found practicable to make a prism of this vapour in the
ordinary way by enclosing it in a glass vessel of the required shape,
as sodium vapour attacks glass, quickly rendering it opaque. A. E.
Becquerel, however, investigated the character of the dispersion by
using prism-shaped flames strongly coloured with sodium. But the best
way of exhibiting the effect is by making use of a remarkable property
of sodium vapour discovered by R. W. Wood and employed for this
purpose in a very ingenious manner. He found that when sodium is
heated in a hard glass tube, the vapour which is formed is
extraordinarily cohesive, only slowly spreading out in a cloud with
well-defined borders, which can be rendered visible by placing the
tube in front of a sodium flame, against which the cloud appears
black. If a long glass tube with plane ends, and containing some
pellets of sodium is heated in the middle by a row of burners, the
cool ends remain practically vacuous and do not become obscured. The
sodium vapour in the middle is very dense on the heated side, the
density diminishing rapidly towards the upper part of the tube, so
that, although not prismatic in form, it refracts like a prism owing
to the variation in density. Thus if a horizontal slit is illuminated
by an arc lamp, and the light-rendered parallel by a collimating
lens--is transmitted through the sodium tube and focused on the
vertical slit of a spectroscope, the effect of the sodium vapour is to
produce its refraction spectrum vertically on the slit. The image of
this seen through the glass prism of the spectroscope will appear as
in fig. 4. The whole of the light, with the exception of a small part
in the neighbourhood of the D lines, is practically undeviated, so
that it illuminates only a very short piece of the slit and is spread
out into the ordinary spectrum. But the light of slightly greater
wave-length than the D lines, being refracted strongly downward by the
sodium vapour, illuminates the bottom of the slit; while that