347 
§ 24. For brevity, let us speak of "dispersion electrons”; by 
this expression of course those electrons are meant which, in a gi¬ 
ven region of the spectrum, are chiefly effective in modifying the 
refractivity of the medium. Suppose for a moment that the number 
of dispersion electrons in a molecule of a given gaseous body can 
be estimated. If a is that number and if (as in §§ 1. and 10.) M 
is the number of molecules contained in unit volume of the gas 
at 0° C. and 760 millims., then by formula (1), § 15., 
a A = 
3 tic 2 
(i) 
The right-hand member does not depend on the particular nature 
of the gas; so that, according to the simple form of theory from 
which (1) of § 15. is derived, the product a A ought to be an uni¬ 
versal constant. 
Thus, if the values A 1: zl 2 of the constant for a pair of gases 
are known, we can calculate from 
a i/ a 2 ^ 2/^1 ( 2 ) 
the relative numbers of the dispersion electrons contained in 
their molecules. We should expect therefore two such values A ly 
A 2 to be always in the ratio of two integral numbers. This con¬ 
clusion is corroborated in a simple manner by the results we have 
obtained. Thus for hydrogen the probable value of A is very nearly 
8 , for oxygen it is about 3,8 and for carbon dioxide about 2,1;, 
in round numbers: — 
a co 2 : a o 2 ‘ a H 2 ~ 8 *. 4 : 2. 
This example immediately suggests the view discussed at great 
length by Drude x ), chiefly in connection with the Theory of Dis¬ 
persion in solid bodies, and only partially adopted by him, namely: 
that the number of dispersion electrons in a molecule is given by 
the aggregate number (say v ) of units of valency, or "bonds”, by 
which the atoms contained in the molecule are assumed to be lin¬ 
ked. Without laying undue stress on this hypothesis (which has 
been shown by Drude to be untenable in many cases) we shall 
!) Annalen d. Physik, Bd. 14., p. 677. 1904. 
