458 PROCEEDINGS OP THE AMERICAN ACADEMY. 



sible for the escape of the last traces of water and hydrochloric acid. 

 Moreover since the mother liquor contained almost as much chlorine as 

 the potassic chloride, weight for weight, the inclusion of a trace would 

 not greatly affect the result. 



The results for the potassic chloride are given in the foregoing table. 

 The mean value 39.139 agrees very closely with the value obtained by 

 Stas, whose later experiments with the chloride gave values ranging 

 from 39.142 to 39.15. 



Thus the evidence of this work affirms the trustworthiness of the 

 experiments with caesic chloride. 



V. Tiie Quantitative Decomposition of Nitrates. 



If a process could be devised which involved the use of a different set 

 of apparatus, and another series of salts, any constant error which might 

 possibly have crept into the previous method would probably not appear, 

 and the new results would furnish a severe test of the correctness of the 

 values found from the chlorine determinations. Hence a search was 

 instituted for an entirely new method, involving different manipulation. 



It is well known that when silicic oxide is heated with the nitrate of 

 such a metal as caesium, the salt breaks up according to the following 

 reaction : — 



2CsN0 3 + Si0 2 = Cs 2 (SiO a ) + N 2 5 



the nitrogen pentoxide decomposing and passing off, leaving the caesic 

 silicate. Therefore, from the loss in weight of a mixture containing a 

 known weight of caesic nitrate and an excess of silica one might compute 

 the atomic weight of caesium. This method, commonly used to determine 

 nitric acid, has never before served for this exact purpose. 



A series of preliminary experiments were of course necessary to deter- 

 mine the limit of accuracy attainable. 



The silicic dioxide was prepared from a very pure specimen of natural 

 sand. This material was washed several times with boiling water and 

 then digested with hot concentrated sulphuric acid for several days. 

 Subsequently it was again washed with water and covered with boiling 

 aqua regia. It was allowed to stand on the steam bath in this solution 

 for at least a week. Upon washing and drying the product, it was found 

 to be beautifully white, and when heated to a high temperature in the 

 blast lamp, it came almost at once to a constant weight. After powder- 

 ing very finely in an agate mortar and heating again it was ready for 

 use. Subsequently very pure artificial silica made from alkaline silicate 



