RICHARDS AND BAXTER. ATOMIC WEIGHT OF COBALT. 71 



Tlie four investigations which had no parallel in the case of nickel 

 were those of Sommaruga, Weselsky, Reoimler, and Hempel and Thiele. 

 The first of these need not claim serious attention, for there can be no 

 doubt that the result (60.) is far too high. Weselsky's contribution was 

 very similar to Lee's, which has been treated in Richard and Cushman's 

 criticism. Since the work was obviously crude and incomplete, we must 

 ascribe its accurate average result (58.96) to a compensation of errors. 



Two more investigations remain to be considered. In 1891, when the 

 " Gnomium" dispute was at its height, Remmler entered the fray. He 

 fractionated cobaltous hydrate by treating a large quantity of this precipi- 

 tate with strong ammonia. Each fraction was subjected to a protracted 

 purification, which, although efTective in removing heavy metals, intro- 

 duced more impurity than it removed. His final solution of cobaltous 

 nitrate, from which he formed the oxide by ignition, contained the alkali 

 and silica taken up by at least six successive evaporations to dryness in 

 glass or porcelain, in three of which the solution was ammoniacal. With 

 such hopelessly contaminated material, it is no wonder that the results 

 from twenty-four fractions varied between limits which were nearly one 

 per cent apart. As his method of igniting the monoxide to constant 

 weight in an inert gas was subject to the same errors which have been 

 discussed in the case of Russell and Zimmermann, no further comment is 

 necessary. The average of twenty-four determinations was 58.80. 



Hempel and Thiele, in 1895, at first attempted to find the ratio 

 CoO : Co from analyses of the monoxide. This part of their research 

 has been already discussed. 



The remainder of their work consisted in the quantitative preparation 

 and analysis of cobaltous chloride. They were unable to prepare it in a 

 wholly anhydrous condition, and realized this fact; hence the part of 

 their calculation depending upon the weight of the chloride was unsatis- 

 factory. The final step of their investigation was the weighing of the 

 argentic chloride obtainable from the chloride combined with a known 

 weight of cobalt. Although not wholly satisfactory, this part of their 

 research far excelled the earlier work of a similar kind (Dumas) and 

 yielded one of the most trustworthy values for the atomic weight of 

 cobalt (59.91 to 59.94) which had been obtained at that time. 



It is clear, from a careful consideration of the details just referred to, 

 that all the wide variations in the older results are explicable. Hence 

 no serious argument exists against the atomic weight of cobalt (58.995, 

 if = 16) determined by our analyses of cobaltous bromide. Since 

 moreover the still more recent results with the chloride (less than 59.05) 



