ATOMIC WEIGHTS 371 



G.39735 1.21644 19.015 



2.17450 .41332 19.008 



1.57903 .29966 18.977 



Mean, 18.9996, ± .0046 



Hence W= 183.60. 



Combined with Taylor's mean, 19. 0125, ±.0034, the general mean is 

 19. 0073, ±.0027. This ratio, however, is affected by constant errors, as 

 Smith and Exner have shown. There is not only a possibility of action 

 of the sodium carbonate njoon the glass bnlb, but also a loss due to slight 

 decomposition of the carbonate itself at the temperature employed in 

 the experiments. Smith and Exner therefore discard the method as too 

 inaccurate. 



The work done by Smith and Desi ' probably ought to be considered 

 in connection with that of Pennington and Smith on the trioxide. 

 Smith and Desi started with tungsten trioxide, freed from molybdenum 

 by means of gaseous hydrochloric acid. This material was reduced in 

 a stream of carefully purified hydrogen, and the water formed was col- 

 lected in a calcium chloride tube and weighed. To the results found I 

 add the percentage of water obtained from 100 parts of WO3. Vacuum 

 weights are given: 



.0008 



Hence W = 184.70. This method is also criticized by Smith and Ex- 

 ner, and rejected. 



Still another method for determining the atomic weight of tungsten 

 was tested by Thomas,' also in Smith's laboratory. Sodium tungstate, 

 Na2W04.2HoO, was dehydrated between 180° and 200°, and the per- 

 centage of water so determined. In this series of experiments the tung- 

 state contained traces of carbonate and silicate. With purer material other 

 determinations were made between 268° and 295°, and these were divided 



1 Read before Amer. Phil. Soc, Nov. 2, 1S94. 



- Journ. Amer. Chem. Soc. 21, 373. 1899. Thomas cites some work on tungsten trioxide, but 

 his figures appear in Hardin's series. 



