498 
element is a mean; that when we say that 
the atomic weight of oxygen is 16 we 
merely state that the average atomic weight 
is 16; and it is not inconceivable that a cer- 
tain number of molecules have a weight 
somewhat higher than 32, while a certain 
number have a lower weight. 
We therefore thought it necessary to test 
this question by direct experiment with 
some known gas ; and we chose nitrogen, as 
a good material with which to test the 
point. A much larger and more conven- 
ient apparatus for diffusing gases was built 
by Mr. Travers and myself, and a set of 
systematic diffusions of nitrogen was car- 
ried out. After thirty rounds, correspond- 
ing to 180 diffusions, the density of the 
nitrogen was unaltered, and that of the 
portion which should have diffused most 
slowly, had there been any difference in 
rate, was identical with that of the most 
quickly diffusing portion, 7. e., with that of 
the portion which passed first through the 
porous plug. This attempt, therefore, was 
unsuccessful; but it was worth carrying 
out, for it is now certain that it is not pos- 
sible to separate a gas of undoubted chem- 
ical unity into portions of different density 
by diffusion. And these experiments ren- 
dered it exceedingly improbable that the 
difference in density of the two frac- 
tions of helium was due to separation of 
light molecules of helium from heavy 
molecules. 
The apparatus used for diffusion had a 
capacity of about two litres. It was filled 
with helium, and the operation of diffusion 
was carried through thirty times. There 
were six reservoirs, each full of gas, and 
each was separated into two by diffusion. 
To the heavier portion of one lot the 
lighter portion of the next was added, and 
in this manner all six reservoirs were suc- 
cessively passed through the diffusion ap- 
paratus. This process was carried out 
thirty times, each of the six reservoirs hay- 
SCIENCE. 
[N.S. Von. VI. No. 144. 
ing had its gas diffused each time thus in- 
volving 180 diffusions. 
After this process the density of the more 
quickly diffusing gas was reduced to 2.02, 
while that of the less quickly diffusing had 
increased to 2.27. The light portion on re- 
diffusion hardly altered in density, while 
the heavier portion, when divided into 
three portions by diffusion, showed a con- 
siderable difference in density between the 
first third and the last third. A similar 
set of operations was carried out with a. 
fresh quantity of helium, in order to accu- 
mulate enough gas to obtain a sufficient. 
quantity for a second series of diffusions. 
The more quickly diffusing portions of both 
gases were mixed and re-diffused. The den- 
sity of the lightest portion of these gases. 
was 1.98, and after other 15 diffusions the 
density of the lightest portion had not de- 
creased. The end had been reached; it. 
was not possible to obtain a lighter portion 
by diffusion. The density of the main body 
of this gas is therefore 1.98 ; and its refrac- 
tivity, air being taken as unity, is 0.1245. 
The spectrum of this portion does not differ 
in any respect from the usual spectrum or 
helium. 
As re-diffusion does not alter the density 
or the refractivity of this gas, it is right to 
suppose that either one definite element has 
now been isolated or that if there are 
more elements than one present they pos- 
sess the same, or very nearly the same, den- 
sity and refractivity. There may be a 
group of elements, say three, like iron, co- 
balt and nickel ; but there is no proof that 
this idea is correct and the simplicity of the 
spectrum would be an argument against 
such a supposition. This substance, form- 
ing by far the larger part of the whole 
of the gas, must, in the present state of our 
knowledge, be regarded as pure helium. 
On the other hand, the heavier residue is 
easily altered in density by re-diffusion, 
and this would imply that it consists of a 
