(O RESEARCHES UPON ATOMIC WEIGHTS. 



through dry bromine in a small bulb. This apparatus was constructed entirely 

 of glass with ground joints. The tube which conducted the gases into the cru- 

 cible passed through a Rose crucible-cover of glazed porcelain in all experiments 

 except analyses 28 to 31, in which a quartz cover was employed. The quartz 

 crucibles were always contained in large porcelain crucibles while being heated. 

 They remained almost absolutely constant in weight during the experiments. 

 The bromine was in each case a portion of the sample from which the silver 

 bromide had been made. 



Next the bromide was heated barely to fusion in a slow current of chlorine, 

 generated by the action of hydrochloric acid upon manganese dioxide, and 

 dried by means of concentrated sulphuric acid. The apparatus for this pur- 

 pose also was constructed wholly of glass. When the bromine was apparently 

 completely displaced, the silver chloride was heated in the air for a few min- 

 utes to expel dissolved chlorine, and then was cooled and weighed. A repetition 

 of the heating in chlorine seldom affected the weight of the salt more than a 

 few hundredths of a milligram, although occasionally a third heating was 

 necessary to effect this result. 



That no loss of silver chloride by volatilization took place is certain for two 

 reasons. In the first place the cover of the crucible and the delivery-tube for 

 the bromine, when rinsed with ammonia and the solution treated with a slight 

 excess of hydrochloric acid, gave no visible opalescence in the nephelometer. 

 In the second place the weight of the chloride became constant without diffi- 

 culty. It has already been shown that silver chloride which has been fused 

 in chlorine, if subsequently heated in air, retains no excess of chlorine.^ 



The following vacuum corrections were applied: silver bromide, -}- 0.00C041 ; 

 silver chloride -J- 0.00007 1.^ (See table on page 61.) 



RESULTS AND DISCUSSION. 



Aside from the close agreement of all the results of Series I, the fact is to be 

 emphasized that of the last seven analyses, which were consecutive, only two 

 differ from the average of the series, 79.916, by as much as o.ooi unit. Fur- 

 thermore, there is no evidence of any dissimilarity in the different preparations 

 of bromine. Material which has received only two distillations from a bromide 

 gives values no lower than bromine which has been thus treated four times. 

 The various specimens of silver also show no difference in purity. 



In the case of Series II, the extreme variation of the results is only 0.004 unit, 

 and only one of the 13 experiments yielded a value which differs from the aver- 

 age by more than o.ooi unit. 



Finally, the difference between the averages of Series I and II is only 0.0007 

 unit. It is extremely unlikely that constant errors could have affected both 



^ Baxter: Proc.Amer. Acad., 40, 4^2 {igo^); Joitr. Amer.Chem.Soc.,26,i5gi; Zeit.anorg. 

 Chem., 43, 29. (See page 103.) 

 ' See page 41. 



