688 



REPORT 1892. 



Table I. 



Mean atomic weight B = 10-921. 

 Probable error ± -010. 



Method II. 



Distillation with methyl alcohol and hydrochloric acid. 



A -weighed quantity of anhydrous borax -was dissolved in a small quantity of 

 water, alcohol added, and an excess of pure hydrochloric acid ; it was then dis- 

 tilled on a water-bath with addition of methyl alcohol, till boracic acid could no 

 longer be detected in the distillate by the flame test. The flask was then dried in 

 an air-bath at a temperature of about 100-110° C. in a slow current of air. The 

 temperature was then raised to about 3.50° for two or three hours to drive off the 

 last traces of water ; the flask was then left to cool, and weighed after it had stood 

 for some time in the balance case. 



The weight of salt obtained from the horax was thus determined : the sodium 

 present was calculated ; from this number the amount of borax corresponding to 

 two atomic proportions of sodium was found; this number less I^slJO, gave B,, 

 and hence the atomic weight of boron. 



In the first series of experiments a flask of soft glass was used. The results 

 obtained were not very concordant, for the glass was attacked, replacement of 

 oxygen by chlorine took place, resulting in the formation of silica and sodium 

 chloride from sodium silicate, for on dissolving the salt in water an insoluhle resi- 

 due of silica was left. 



The results are given in Table II. ; the mean atomic weight is 10-952, and is 

 not far from that obtained in the next series of distillations. 



Table II. 



Mean atomic weight B = 10-952. 

 Probable error ± -010. 



Second Series. 



To avoid the error introduced by the action of the hydrochloric acid and alcohol 

 upon the soft glass, a flask made of combustion tubing was used. Ii was found 

 that this glass was hardly acted upon, but when the salt was dissolved in ^yater a 

 very slight residue of silica was left. In order to check the results, the chlorine 

 in the sodium chloride was estimated with silver nitrate in the usual manner ; the 



