314 Affinity and Fleat. 
Suppose that a body with little stability—such as hydro- 
chlorate of alumina—a body which is reduced to its ele- 
ments by the least heat, be introduced into the dialyzer, 
above the membrane or parchment-paper which serves as 
a filter with very close meshes. Below this filter is water, 
which I assume to be constantly pure. 
The hydrochlorate of alumina absorbs the more heat 
the greater the quantity of water in which it is dissolved, 
so that at a given moment it may be supposed to contain 
so much heat that if this heat became sensible and were 
applied to its elements, they would at once separate. At 
this moment hydrochloric acid becomes free, and hydrated 
alumina separates in particles extremely divided, which 
occupy all the liquid in which the separation is effected. 
Hence it ought to have all the properties of a dissolved 
substance. It will soon be seen that these properties are 
only apparent. | 
That is what takes place in the dialyzer ; moreover the 
hydrochloric acid passes through the filter, and alumina 
remains on the surface in the state of apparent solution, or 
in the colloidal state, to use the expression which Mr. Gra- 
ham has adopted. Yet .in practice this reaction does not 
take place exactly in this manner. | 
We may suppose the permeable membrane divided into 
two horizontal layers,—one the higher layer, where the 
hydrochlorate of alumina penetrates (this has little thick- 
ness), the other the lower layer, where water alone pene- 
trates. It is clear that in an apparatus of this kind the 
solution of hydrochlorate of alumina will very soon find a 
layer near the surface where the water, which is renewed — 
with rapidity, will be in considerable proportion as com- 
pared with hydrochlorate. In this layer the decomposition 
of the salt will then take place by indefinite diffusion (con- 
sequently by the effect of heat), the hydrated alumina will 
remain in the state of colloidal particles on the upper sur- 
face, and the hydrochloric acid will be carried away by the 
water. The colloidal layer thus produced will itself become 
a true filter, and the phenomena of decomposition may go 
on in its interior. 
Yet decomposition by diffusion cannot be complete; for 
it varies with the ratio of the quantity of water put in con- 
tact with hydrochlorate of alumina, and with the quantity 
of hydrochloric acid which this water contains. In other 
words, the quantity of water separated by diffusion will be 
the smaller the more this water is charged with hydro- 
