38 
Proceedings of the Koyal Society of Edinburgh. [Sess. 
V 
In the case of C 2 H 5 Br and H 2 0 it is the molecular weight that is 
given by w. The two entries marked CuK and CuL are independent 
determinations for the same band — in the one case based on observations 
made with the K radiations, and in the other case on observations made 
with the L radiations. The second last column gives the number of 
electrons per atom, or, in the case of a compound, per molecule, calculated 
by formula (8), and the last column the same quantity calculated by 
formula (6). The third, fourth, and fifth columns give the data used 
for the calculations by formula (8). In using (8), \ was determined on 
both sides of the band, and half the difference taken as \ — A m ; \ and 
\ m are, of course, measured in 10“ 8 cms. Strictly speaking, formula (8) 
is not applicable, as the bands have not the simple form required by 
theory; but in practice it appears to give better results than the other 
formula. The difference between the last two columns is due to there 
not being sufficient points on the curves to determine their shapes with 
accuracy; it was necessary to extrapolate, and I apparently did this 
in a different manner each time. 
Almost ten years ago * I applied a formula similar to (8) to the absorp- 
tion bands of aniline colouring matters and inorganic salts. I found then 
that in the aniline colouring matters there was one electron per atom 
concerned in the production of each absorption band, and the values 
obtained for inorganic solutions suggested ions. But in many cases 
the results suggested neither atoms nor ions ; the absorption band corre- 
sponded apparently not to the vibrations of a single electron in the atom, 
but to a degree of freedom of a group of electrons or ions. 
The first seven results, which are all for K bands, do not differ much 
from I. Their mean value according to (8) is *972, and according to (6) 
is T01. Hence in this case the absorption band is due to the vibration 
of one electron per atom or per molecule. The results in the case of Pt 
and Au are for L bands. There are not sufficient observations to speak 
with certainty, but these bands appear to be double. The estimate by (8) 
was made for one component on this assumption ; the estimate by (6) was 
made for the whole band. Hence the discrepancy in this case is greater 
than it looks; there may be, however, one electron to each component. 
The results in the case of A1 and H 2 0 are for J bands, and suggest on 
first sight that there is only one active electron in each 481 atoms and 
each 1124 molecules respectively. There is, however, a much more 
plausible interpretation than this. If we assume that the band is due 
to a charged hydrogen atom or an a particle, i.e. a helium atom with two 
* Proc. Roy. Soc A, 82, 1909, p. 606. 
