RICHARDS AND PARKKR. — ATOMIC WlOKiHT OF M.VGNESIUM. 71 



from a water blast was i)assed through the apparatus and then through a 

 weighed phosphorus pentoxide bulb fur ncaily two hours, without the 

 slighte.xt appi-eciable increase of weight of the pentoxide bulb. The same 

 test was applied to the ajjpai-atus for drying the hydrochloric acid gas, 

 with tlie same result. 



Willi the help of this important addition to the apparatus, another 

 series of deiermiuatious was now made. The somewhat lower result of 

 this series is undoubtedly due to the more perfect desiccation of the 

 eases : the agreement of the individual results is still not quite perfect, 

 but the series is undoubtedly far more reliable than the second. 



SERIKS 111. 



Fourth axd Final Series of Determinations. 



The apparatus was now put in the best possible order, and the phos- 

 phorous pentoxide tubes were recharged, in order to make ready for a 

 series of determinations in which the very highest exactness was to be 

 aimed at. The purest samples of material were used, and all other pre- 

 cautions, learned from previous work, were taken to insure accuracy. 

 The following determinations were consecutive, with the exception of one 

 between Nos. 15 and 16, which was spoiled by a slight accident. 



These results agree with one another as well as could possibly be 

 expected, for the difference between the extremes in the last series 

 corresponds to a difference of only one tenth of a milligram in the \vei<^ht 

 of the magnesic chloride. Since two wholly distinct samples of this salt 



