202 G. HARKEK. 



the percentages for fluorine by methods (1) and (3), the silica and 

 alumina were also estimated, so that with the exception of water r 

 the whole of the constituents were now obtained. 



It has been found by Penfield and Minor (Am. J. Sc. 94) and 

 Jannasch and Locke (J. Chem. Soc. 94) that all topazes contain 

 more or less water of constitution. The former found 2*45 and 

 the latter 2*69 per cent, in Brazilian topaz. The water takes the 

 place of the fluorine. Penfield and Minor decomposed the topaz 

 with sodium carbonate and the water free from acid was weighed 

 in sulphuric acid or calcium chloride tubes. Jannasch and Locke 

 decomposed with litharge in a bulb tube and collected the water 

 in a calcium chloride tube, a stream of dry air being passed 

 through the apparatus. This last method was found more con- 

 venient as a much lower temperature could be used to decompose 

 the topaz ; both methods however gave the same results. In the 

 experiments the topaz was first heated to bright redness over a 

 strong bunsen to drive out any contained water ; it was then 

 decomposed by litharge and its so-called water of constitution 

 determined. Blank tests on the litharge shewed that the calcium 

 chloride tube gained two to three deci-milligrams and in calculat- 

 ing the percentages this was deducted. In the experiments 

 about -5 gram, of topaz was mixed with 2 - 2*5 grams, of previously 

 heated litharge, the same amount of litharge being used in the 

 blank tests. 



Water of Constitution. 



Locality. Amount taken. Water found. Percentage. 



D ., (-6324 grams. -0134 2-12 )„,„ 



Brazil ... j. 5080 * . 0107 2-11 ] 2U 



XT T7 1 J ( ' 5038 ,0050 '" I 1-07 



New England | . 57g() . 0066 1<u J 1 07 



m a J ,490 ° * 0038 * 78 \ -75 



Mudgee ... |. 495S . 0038 . 73 J 75 



Results of analyses. — Appended is a table giving the complete 

 analyses of these three varieties of topaz. In the latter it will be 

 seen that the silica and alumina obtained by Wohler's method are 

 always higher than by Berzelius', and this is apparently due to 



