1904.] SMITH AND EXNER — ATOMIC WEIGHT OF TUNGSTEN. 189 



The trioxide used in experiments i to 4, inclusive, was obtained 

 by gently igniting the ammonium salt in a porcelain crucible. 

 That used in experiments 5 and 6 was strongly heated in a porce- 

 lain crucible. In 7 and 8 the ammonium tungstate was heated for 

 one hour in a double platinum crucible. The oxide in experiment 

 7 had been heated two hours in the same kind of crucible, while in 

 experiment 10 the ignition continued for two hours in the double 

 platinum crucible. The gradual rise in the value to 183.85 by 

 protracted heating could surely not be due to the expulsion of vola- 

 tile matter, for there was no change in weight after the first hour of 

 ignition in the double crucible. Evidently the oxide absorbed 

 impurities which led to the rise in the atomic weight. Accordingly, 

 samples of ammonium paratungstate were ignited under conditions 

 as nearly similar as possible in crucibles of platinum and of porce- 

 lain. The values from the oxide in the crucibles of porcelain were 

 higher than those from the oxide made in platinum crucibles, show- 

 ing in all probability, that the oxide took up more foreign material 

 from the porcelain than from the platinum. Therefore, the mere 

 ignition of the ammonium salt in vessels such as have been 

 described drew in sources of error. These would, of course, have 

 to be eliminated if the method was to be tested upon its own 

 merits. It was sought to accomplish this by igniting thoroughly 

 dry ammonium paratungstate and ascertain the loss (water and 

 ammonia) sustained by different amounts, which resulted in dis- 

 covering that the percentage of volatile matter could be obtained 

 to within less than 0.01%, which would answer for the purpose of 

 atomic weight determination. And therefore, in the actual experi- 

 ments, the ammonium paratungstate was weighed out directly into 

 the flask, it being only necessary to make the proper calculations to 

 arrive at the amount of trioxide which was thus used. Six deter- 

 minations were made; the results in the atomic value varied from 

 183.4 to 183.81. The early explanation for the lack of concor- 

 dance, if the method was not faulty, would be to suppose that the 

 action of the soda upon the glass would withdraw varying amounts 

 of silica, and there would follow, of course, the liberation of corres- 

 ponding amounts of carbon dioxide. If this really occasioned the 

 error, it was hoped that the substitution of a platinum bulb, similar 

 to the glass vessel, would lead to success. This was done. The 

 experiments were performed as before, with slight modifications 

 where it was considered advisable and advantageous. The atomic 



