440 PROCEEDINGS OP THE AMERICAN ACADEMY. 



the steadily rising transition temperature, did not lose its impurity fast 

 enough to make any considerable final yield a possibility. The com- 

 mercial salt sometimes gives results several degrees in error, and even 

 after eight recrystallizations one sample melted over half a degree below 

 the true point. Such material is of course wholly out of the question. 

 Perhaps some other isomorphous impurity beside sodic chloride was 

 present. Whatever may have been the cause of the trouble, this crude 

 material was entirely rejected in subsequent work, and all other samples 

 were made from pure hydrobromic acid or bromine and pure alkali. 



The next preparation of sodic bromide was made from purified sodic 

 carbonate and hydrobromic acid. The sodic carbonate was five times 

 recrystallized, and the bromine, from which the hydrobromic acid was 

 made, was freed from chlorine by solution in calcic bromide and repre- 

 cipitation by water. The bromine was converted into hydrobromic 

 acid with the help of carefully washed red phosphorus and water. 

 Hydrobromic acid, prepared in this way, contains phosphorus and some- 

 times traces of arsenic ; but three successive fractional distillations 

 easily removed these impurities. The purity of the product and the 

 residue was determined in each case by ammonium molybdate. The 

 first residue, after the distillation of about eight kilograms of fifty per 

 cent hydrobromic acid in the retort, yielded about thirty grams of 

 ignited phosphomolydenum oxide. The residue from the second dis- 

 tillation yielded only a quarter of a gram, and the residue from the 

 third distillation gave no precipitate whatever with the amnionic molyb- 

 date. The distillate was of course even more pure than this, because 

 the impurities concentrate in the residue. Iodine was expelled in the 

 first aqueous fractions by adding bromine water before each distillation. 

 No attempt was made to eliminate the iodine by adding zinc oxide 

 according to Stas, because this method seemed to us of doubtful efficac} r . 

 As will be shown later, the sodic bromide (sample II) prepared thus 

 was very pure, but it had not yet attained the greatest purity possible, 

 because the single treatment with calcic bromide is usually insufficient 

 to remove every trace of chlorine from the commercial bromine. A 

 number of other samples were made in this way, and according to the 

 amount of chlorine in the original crude bromine, the substance obtained 

 varied from a state of almost perfect purity to a state of contamination 

 with 0.05 per cent of its weight of sodic chloride. Protracted and 

 tedious quantitative precipitations of silver bromide, conducted with all 

 the accuracy of atomic weight determinations, were necessary to j>rove 

 this fact and explain slight irregularities in the transition temperatures 



