134 PROCEEDINGS OF THE AMERICAN ACADEMY. 



lu the table on page 135, each horizontal line reports a complete 

 experiment, which usually required a day for preparation and another 

 for accomplishment. At first the copper cells were multiplied ; but later 

 when the silver seemed to be quite as uncertain as the copper, the nobler 

 metal received its full share of representation. Some of the earlier 

 experiments failed for one reason or another, and these failures are not 

 included in the table. The last six determinations were consecutive ; 

 and in these the conditions of experiment were much more satisfactory 

 than in the earlier ones ; for much practice was naturally acquired in the 

 course of the work. In all the experiments excepting Nos. 1 and 2 the 

 cupric electrolyte was boiled and cooled to zero in an inert atmosphere 

 before the electrolysis; and in all excepting Nos. 1, 2, 3, and 6 the cop- 

 per electrolysis was conducted in an atmosphere of hydrogen. 



Attention is again called to the fact that each corrected value for the 

 copper deposit is computed from two deposits, one on a large and one on 

 a small electrode. The exposed area of the large electrode was always 

 about 50 square centimeters ; and in all cases, except those indicated 

 by the signs ft below, the area of the smaller plate was half that of 

 the larger. The current densities varied slightly, but were always 

 less than 0.01 ampere per square centimeter on the smaller plate, 

 the total current strength usually being about 0.12 ampere for the 

 system. When the total current strength was kept below this limit, 

 no systematic variation due to its changes coidd be detected in the re- 

 sults; hence the individual figures are omitted in this published account. 

 In order to prevent saturation with cuprous salt, the volume of the 

 electrolyte used in each cell was fairly large, being usually about 200 

 cubic centimeters. 



The corrected average value from the table for the atomic weight of 

 copper, G3.5G3, has a "probable error " of 0.004 as far as the accidental 

 inaccuracies of manipulation are concerned. It has already been shown, 

 however, that a value obtained in this way must correspond to a de- 

 posit of copper slightly too great ; for the mode of correction does not 

 take account of the growing, although slight, presence of cuprous salts. 

 On the other hand, the values computed from the directly observed 

 quantities of copper must be too small. The average observed values, 

 calculated from the sums at the bottoms of the respective columns, are 

 Cu = 63.525 from the larger plates, and Cu = G3.547 from the smaller 

 plates. It is evident then, that the true electrochemical atomic weight 

 of copper, when compared with the silver deposited in the ordinary 

 silver voltameter, must lie between the limits 63.563 and 63.547. Jt 



