1030 
PROFESSOR J. W. MALLET OR A REVISION 
leads to the suspicion that such constant error as may have been involved tended in 
this direction, and if so that the resulting atomic weight would be made to appear 
somewhat too low. The extent of error from this cause, however, if it existed at 
all, must have been extremely small. 
D irect results of second set of experiments of third series. —The results of these last 
experiments were— 
B.—Hydrogen weighed as water. 
I.—2T704 grms. of A1 gave 2T661 grms. of H 3 0. 
II.—2-9355 „ „ 2-9292 „ 
III.—5-2632 „ „ 5-2562 „ 
Calcidation of results. 
In calculating the atomic weight of aluminum from the data furnished by the above 
described experiments, the atomic weights assumed for the other elements involved 
are those which result from the researches of Stas and the previous investigation by 
Dumas and Stas of the composition of water, namely— 
0=15-961 S = 31-996 N=14-010 
In regard to silver and bromine a difficulty arises from the fact that the relation 
between these elements was determined by Stas with metallic silver which, as 
Dumas'" has pointed out, contained in all probability occluded oxygen. It appears 
from Dumas’ experiments and from mine, that the quantity of oxygen which may 
be so retained varies with the conditions under which the metal is fused, and it is 
impossible now to ascertain precisely how much was present in that used by Stas, 
while the correction to be applied on this account, though small, is not inappreciable 
in its effect upon the atomic weight of the aluminum. Omitting to extract in the 
Sprengel vacuum the occluded oxygen from the silver used in my experiments would 
not have secured identical condition of the silver with that of the metal used by Stas, 
since the circumstances of fusion and cooling would probably not have been altogether 
the same, and it seemed best to use silver fully purified in this respect, so that my 
results might be directly comparable with any obtained in the future, since this source 
of error once pointed out ought not to be hereafter neglected. I have therefore used 
as the atomic weights of silver and bromine the numbers obtained by Stas (from 
his experiments on silver bromide and bromate), recalculated on the assumption 
that the metal employed by him would have yielded 57 c.c. (reduced to 0° C., and 
760 m.m.) or 82 mgrm. of oxygen per kilogramme, this being the quantity obtained 
by Dumas from silver treated as nearly as possible as was in all likelihood that which 
Stas employed. This has the advantage of reducing the remaining error to that only 
which depends on the difference between the real amount of oxygen which was 
* Loc. cit . 
