OF THE ATOMIC WEIGHT OF ALUMINUM. 
1013 
1 . That each process used should be as simple as possible, and should involve as 
little as possible of known liability to error. 
2 . That different and independent processes should be resorted to as the means of 
checking each other’s results, even though it may fairly be assumed that one is more 
advantageous than another. 
3. That each process should be carried out with quantities of material differing 
considerably from each other in successive experiments. 
4. That only such other atomic weights should be involved as may be counted 
among those already known with the nearest approach to accuracy. 
The most scrupulous care was taken in the purification and examination of all the 
reagents used, and as far as possible vessels of platinum or of hard porcelain were 
substituted for those of glass. 
Means and method of weighing employed .—For the weighings an excellent balance, 
of Becker’s construction, was employed. It was in perfect order, carefully adjusted 
(especially as regards centre of gravity of beam with average load to be carried), and 
would bear safely 200 grms. in each pan, giving when thus loaded a deflection of the 
index to the extent of 1-j; division of the scale over which it moves for a difference 
of weight of ‘0001 grm. All weighings were made by the well-known method of 
observing the vibrations of the index on either side the position of rest. In one series 
of experiments absolute weights were required, i.e., real equality of weight between 
the quantities of matter dealt with and the standards of weight with which they were 
compared; in these cases the method of “double weighing” was made use of, so as to 
eliminate any error arising from inequality in length of the arms of the balance. In 
view of this need, in connexion with a part of the research, for absolute weights, 
directly comparable with those used by Begnault in his determination of the density 
of hydrogen, I applied to my friend J. E. Hilgard, Esq., in charge of the office of the 
United States’ Coast Survey at Washington, for a comparison of a kilogramme with a 
weight of the same denomination belonging to the Coast Survey, the value of which 
latter weight is accurately known in terms of the original “ kilogramme of the 
Archives” at Paris. He kindly had this comparison made, and sent me the results in 
detail, showing that my weight was 8T milligrammes heavier than the “ star kilo¬ 
gramme” which is the standard of reference at Washington (both in vacuo), with an 
uncertainty of comparison not exceeding '1 milligramme, while the “star kilogramme” 
is certified to as agreeing with the normal “ kilogramme of the Archives” within 
IT milligramme. I had already a 10-gramme weight, professedly normal, but, as it 
turned out, too light by a very minute fraction of its value, and with these two, 
checked against each other, a full series of comparisons was made of all the other weights 
to be employed, the specific gravity of each piece being determined before its final 
comparison as to weight, so that the real values might all be referred to a vacuum by 
calculation of the buoyancy in air. Determinations of the specific gravity of all 
materials and vessels which had to be weighed were also made, and, the barometer 
