68 PROCEEDINGS OF THE AMERICAN ACADEMY. 



These values have no meaning except for the particular process by 

 which they were obtained, but they show plainly how great a chance for 

 error existed when the tube was exhausted completely as soon as ignition 

 ceased. 



In the next analysis the procedure was exactly similar to that in 

 Analysis 4, except that during cooling the values indicated in the table 

 were maintained. No essential loss of weight took place after the third 

 heating, although several additional ignitions were carried out for the 

 sake of certainty. 10.58678 grams (in vacuum) of cobaltous oxide yielded 

 8.32611 grams of cobalt, corresponding to an atomic weight 58.929. 



In the face of this unsatisfactory result, there seemed nothing further 

 to do except to attempt the detection of the presence of impurities. 

 Three possibilities were open : the oxide might contain moisture, enough 

 nitrate might have been carried down by the cobaltous hydrate to cause 

 a small amount of nitrogen and oxygen to be occluded by the oxide, or 

 else some higher oxide of cobalt might be present. 



It was easily shown that cobaltous oxide is as a matter of fact slightly 

 hygroscopic. Several grams exposed for twenty-four hours to an atmos- 

 phere saturated with moisture gained five milligrams in weight, most of 

 which was lost in a short time iti a desiccator. It was satisfactorily 

 proved, however, by suitable experiments which need not be recounted, 

 that not enough moisture could have crept in during any of the processes 

 used to cause any important error. 



In order to prove the absence of included gases, a specimen of oxide 

 which had been treated in the same manner as the material used in 

 Analysis 5 was dissolved in hydrochloric acid in an apparatus for measur- 

 ing the gases evolved during the solution.* No gas was given off; but 

 upon treating the solution with potassic iodide and starch, an unmistak- 

 able liberation of iodine was detected. This showed conclusively that 

 some higher oxide of cobalt had remained undecomposed, for pure cobalt- 



* These Proceedings, XXXIII. 403, and Jour. Chem. Soc. Trans., LXXI. 559. 



