196 * PROCEEDINGS OF THE AMERICAN ACADEMY 



of perfect purity may be regarded as almost an impossibility. One 

 by one the usual cupric and cuprous compounds were rejected as un- 

 suitable for the foundation of a new research, until finally, after much 

 study, cupric bromide was selected on account of the simplicity and 

 exactness of its analysis, and moreover because by establishing beyond 

 doubt the relation of copper to bromine we connect its atomic weight 

 with precise values previously determined in this laboratory. 



The balance used throughout the research was the one used by 

 Professor Cooke in his admirable work upon the atomic weight of 

 antimony, and was kindly loaned by him for the present purpose. A 

 long-armed balance of the best type, manufactured by Becker of New 

 York, it was very perceptibly turned by one twentieth of a milligram 

 with a load of fifty grams in each pan, and showed unusual constancy 

 in its indications. It was kept in a large quiet room of nearly con- 

 stant temperature, and was protected from unequal heating and from* 

 air currents by suitable curtains. The weights were made by Sarto- 

 rius in Gottingen, the larger ones being of gold-plated brass, and the 

 fractions of a gram of platinum. Three days were spent in their com- 

 parison, according to a method somewhat similar to that described by 

 Crookes,* and the very small corrections thus found were tabulated 

 and applied to all results. 



Before weighing, every object was allowed to remain in a desiccator 

 near the balance, for at least seven hours. Every weight was deter- 

 mined at least thrice, each time according to a different method of 

 swinging, and all precautions for accuracy were taken which are usual 

 in such work. In the first experiments the weighings were made by 

 substitution, but the conditions were so uniformly constant and the 

 balance so manifestly reliable that this procedure seemed a useless 

 expenditure of time. In fact, the probable error of the ordinary 

 method of weighing under such circumstances is very much less than 

 other possible errors of an investigation like the present. It is certain 

 that the total inaccuracy of any weighing, due to all possible causes 

 combined, could not have exceeded seven or eight one-hundredths of 

 a milligram, and was without doubt usually very much less than that 

 quantity. All weigliiiigs were of course reduced to a vacuum standard, 

 allowing for the air displaced by the weights. Since these were stan- 

 dardized with reference to a brass ten-gram piece, it is evident that the 

 fractions of a gram must be calculated as brass, and not as platinum. 



* Fresenius's Zeit8chrift,vi. 431. 



