1880.] 



On the Atomic Weight of Aluminium . 



575 



In examining beforehand the purity of the specially prepared metal 

 nsed, warDed by Dumas' previous results as to gaseous occlusion by 

 silver, of which due account was taken in the paper, I did not neglect 

 to test in like manner the aluminium, but obtained an entirely negative 

 result. As, however, I heated the metal in a Sprengel vacuum only 

 to the highest temperature which a hard Bohemian glass tube would 

 bear, while M. Dumas says that gas is only given off suddenly " vers 

 le rouge hlanc" I have since seeing his paper repeated the experiment 

 in a porcelain tube, and at this higher temperature, obtained by means 

 of an excellent gas furnace, with a small remnant of the same material 

 used in the atomic weight determinations. 



4*783 grms. of the metal thus treated gave a small bubble of gas, 

 which measured but 75 cub. centim. at 24° C. and 752 millims. 

 pressure. Transferred to a miniature eudiometer it was exploded 

 with oxygen, and seemed to be pure hydrogen. This is but little more 

 than one-third the quantity of gas obtained by Dumas, but his experi- 

 ments and my own with silver prove that the amount occluded varies 

 with the conditions under which the metal has been fused, and two 

 circumstances in the preparation of my aluminium probably tended to 

 reduce the proportion of hydrogen taken up — namely, that, fearing 

 possible contamination of the metal by its alloying itself with sodium 

 at the moment of reduction from the bromide, I used a considerable 

 excess of the latter, and that, not only were the ingots of sodium wiped 

 free of naphtha with a cloth, but the outside surface was pared off 

 with a knife, and only quite clear and solid pieces of the alkaline metal 

 were used for the reduction. 



The above volume of hydrogen only represents a weight of 0-000061 

 grm. ; the corresponding weight from the largest amount of aluminium 

 used in the atomic weight experiments of Series 3, A (in which 

 the gaseous hydrogen evolved was measured), would be but 0*000010 

 grm., and the corresponding weight from the largest amount of alumi- 

 nium used in Series 3, B (in which the hydrogen evolved was burned 

 and weighed as water), would be "00006 7 grm. — both quantities too 

 small to be determined by the balance used. 



Even if it were possible to correct for this occluded hydrogen the 

 weight of metallic aluminium taken, the greatest error in the volume of 

 hydrogen obtained in Series 3, A, would only represent 012 cub. 

 centim., which would be barely measurable in the flask used, and 

 would affect the atomic weight deduced for aluminium to an extent 

 less than one-tenth of the probable error of the mean result. And, in 

 Series 3, B, the greatest error in the weight of the water produced by 

 combustion of the hydrogen would be 0*0006 grm., measurable by the 

 balance, but affecting the result by less than three-fourths of the 

 probable error of the mean. 



Hence it appears that, although there is undoubtedly a constant 



