1824.] Chrysoheryls from Haddani and Brazil. 431 



F. After the silica was separated from the liquid (E), the 

 alumina and oxide of iron were precipitated by means of a great 

 excess of subcarbonate of ammonia. After twenty-four hours, 

 the liquor was separated from the yellowish precipitate, and was 

 boiled, but no glucina was precipitated from it. The matter 

 precipitated by the subcarbonate of ammonia consisted of 3*68 

 grammes of alumina, or 73*60 per 100, and 0*19 gramme of per- 

 oxide of iron, which, on account of the colour of the mineral, 

 must be estimated as protoxide. The 0-19 gramme of peroxide 

 is equivalent to 0*169 of protoxide, or 3*38 per 100. 



The constituents of this chrysoberyl therefore are, 



Per 100 parts. 



A. Moisture 0*40 



C. Oxide of titanium 1*00 



D. Glucina 15*80 



E. Silica 4*00 



F. Alumina 73*60 



F. Protoxide of iron 3*38 



98*18 

 100*00 



Loss 1*82 



As the preceding results differed so essentially from the ana- 

 lyses of the chrysoberyl from Brazil by Klaproth and Arfwedson, 

 I determined to examine a specimen from that locality. 1*5 

 gramme was analyzed in the manner above-mentioned, and the 

 following results were obtained : — 



Per 100 parts. 



Water 0*666 



Oxide of titanium 2*666 



Glucina 16*000 



Silica 5*999 



Alumina 68*666 



Protoxide of iron 4*733 



98-730 

 100*000 



Loss 1*270 



In estimating these constituents according to the electro- 

 chemical theory, I believe that the oxide of titanium, notwith- 

 standing its important agency in the analytical experiments, 

 must he regarded as an accidental ingredient, as well as tin; 

 oxide of iron, which in some measure may have been derived 

 from the iron mortar. As the cymophane of Brazil appears to 

 be constituted more conformably to the hypothesis of chemical 

 proportions than that of Haddani, the following calculation may 



