Vapour in the Visible and Ultra-violet Regions. 299 
covered with layers of cotton and asbestos board, and con- 
sistent results were at once obtained. There was still a source of 
trouble which for a long time remained undiscovered. On 
heating the tube it was found that the relative shifts changed 
as the temperature rose, becoming constant after a shift of 
about a dozen helium fringes. Repeated heating of the 
sodium to a high temperature, combined with pumping, 
lessened the discrepancies, but the trouble was never com- 
pletely eliminated. It is now believed that a small amount 
of hydrogen is held captive by the sodium vapour, condensing 
with it as the tube cools. When the tube is again heated, 
this occluded hydrogen is evolved with the first traces of 
sodium vapour, and modifies its refractivity. If we are work- 
ing with light on the same side of the D lines as the helium 
line, for which the refractive index of the vapour is less than 
unity, the evolution of this gas will oppose the displacement 
of the fringes, holding both sets back by practically the 
same amount. Suppose we are working with a wave-length 
for which the relative shift with respect to helium light is 
1:10. If we form a sufficient quantity of sodium vapour 
to shift the helium fringes through 30 fringe-widths, and the 
other set 3 widths, and an amount of hydrogen sufficient to 
give a shift of one fringe-width has been set free, it is 
obvious that our observed ratio will be 2:29, or in round 
numbers 1:15. The errors observed were always at the 
beginning of the series when the tube was heated, and at 
the end when it was cooled, and were always of such a nature 
as to be explained perfectly by the hypothesis which I have 
advanced. On this account we always disregarded the rela- 
tive values found when the sodium vapour was only present 
in very small quantities. 
Many weeks of work were necessary before all sources of 
error were eliminated and the best methods of working dis- 
covered. Asa sample of one of the troubles met and over- 
come, the following will serve. When working with light 
from the monochromatic illuminator of a wave-length very 
near that of the D lines, the fringes disappeared before shifts 
of more than one or two fringe-widths were observed. It 
had been our habit to set the mirrors of the instrument in 
such positions that the centre of the fringe system was seen 
in the telescopes. The disappearance was finally found to 
be due to the dispersion of the vapour, and by turning the 
serew of the instrument in such a way as to increase the 
length of the air path, they appeared again*. The most 
* This action of the vapour on the fringes will be discussed in the 
paper following the present one (p. 324). 
