ATOMIC WEIGHTS 373 



This combination is evidently of very little significance. It includes 

 data which are confessedly defective, and which do not tend to com- 

 pensation of errors. The abnormally high value derived from ratio 1, 

 which dominates the combination, is due to the excessive weight given 

 to the determinations by Pennington and Smith, which Smith himself 

 has discarded. If, in place of ratio 1 we take the determinations of 

 Smith and Exner alone, namely, WO3: W : : 100: 79.3169,±.0007, we 

 have the more trustworthy value, W = 184.075, ±.0064. This, combined 

 with the value from ratio 7, also due to Smith and Exner, gives a general 

 mean of W= 184.092, ±.0046. This seems to be the most probable value 

 now available, and it is checked by the fact, already pointed out, that 

 the two ratios of Smith and Exner, combined, give a good value for the 

 atomic weight of chlorine. 



UEANIUM. 



The earlier attempts to determine the atomic weight of uranium were 

 all vitiated by the erroneous supposition that uranous oxide was really 

 the metal. The supposition, of course, does not affect the weighings 

 and analytical data which were obtained, although these, from their 

 discordance with each other and with later and better results, have now 

 only a historical value. 



For present purposes the determinations made by Berzelius,^ by Arf- 

 vedson," and by Marchand'' may be left quite out of account. Berzelius 

 employed various methods, while the others relied upon estimating the 

 percentage of oxygen lost upon the reduction of UgOg to UO,. Eammels- 

 berg's* results also, although very suggestive, need no full discussion. 

 He analyzed the green chloride, UCI4; effected the synthesis of uranyl 

 sulphate from uranous oxide; determined the amount of residue left 

 upon the ignition of the sodio and bario-uranic acetates; estimated the 

 quantity of magnesium uranate formed from a known weight of UOo, 

 and attempted also to fix the ratio between the green and the black 

 oxides. His figures vary so widely that they could count for little in 

 the establishing of any general mean; and, moreover, they lead to esti- 

 mates of the atomic weight which are mostly below the true value. For 

 instance, twelve lots of UaOg from several different sources were reduced 

 to UO, by heating in hydrogen. The percentages of loss varied from 3.83 

 to 4.67, the mean being 4.121. These figures give values for the atomic 



^ Schweigg. Journ., 22, 336. 1818. Poggend. Annalen, 1, 359. 1S23. 

 ^Poggend. Annalen, 1, 245. Berz. Jahr., 3, 120. 1822. 

 3 Jonrn. prakt. Chem., 23, 497. 1841. 



*Poggend. Annalen, 55, 318, 1842; 56, 125, 1842; 59, 9, 1843; 66, 91, 1845. Journ. prakt. Chem., 

 29, 324. 



