HYDRATES AND CARBONATES OF THE ALKALI-METALS, ETC. 453 



sometimes at 200°/' by igniting a known weight with powdered quartz and determining 

 the loss of weight. He found, in two analyses — 



Minimum. Maximum. Mean. 



6-986 7-021 7-008 



III. Diehl, in 1857, analysed the carbonate, after drying it at 130°, by decomposing 

 it with dilute sulphuric acid and ascertaining the loss of weight. Found in four 

 analyses — 



Minimum. Maximum. Mean. 



7-013 7-037 7-027 



IV. Teoost, in 1857,* analysed the chloride by weighing the chlorine as AgCl. Two 

 analyses gave — 



Minimum. Maximum. Mean. 



6-944 7-002 6-966 



V. Stas analysed the chloride three times by titration with silver, and found — 



Minimum. Maximum. Mean. 



7-023 7-028 7-026 



VI. Stas converted the chloride into the nitrate and determined the ratio of 

 LiCl : LiN0 3 ; he found in three analyses — 



Minimum. Maximum. Mean. 



7-026 7-035 7-030 



In contrasting these several results with my own, Li = 6*892, I naturally begin with 

 the results of Diehl's and Troost's analyses of the carbonate, and note that they agree 

 with one another but differ from me, Diehl by about -f 0*135 units =19 times my 

 probable error, and Troost by +0*115 = 16 times my probable error. But Diehl 

 determined his carbonic acid by means of a method of insufficient precision, and Troost 

 dried his carbonate " sometimes in vacuo, sometimes at 200°," and yet produces only two 

 analyses ! Besides, it is not proved that carbonate of lithia loses all its water at ordinary 

 temperatures in vacuo, nor is it proved that it retains all its carbonic acid in air at 200°. 

 Supposing the negatives to hold, his value for Li must be too high. Hence his results, 

 no more than Diehl's, can be said to be incompatible with mine. 



In now passing to Mallet, I must admit that his three analyses of the chloride agree 

 fairly well, and yet the mean of the two " means " 6*942, is greater than my result by 0*05 

 units = to seven times my probable error. But, at Mallet's time Dumas's excellent 

 method for at the same time dehydrating and deoxygenating a metallic chloride (by heating 

 it in HC1 gas) was not invented yet; his chloride was bound to contain oxychloride, 

 and his analysis to give too high a value for Li. 



But, unfortunately for me, Stas's result is still higher than Mallet's, and his chloride 

 cannot be presumed to have been contaminated with oxychloride, because, if his silver- 



* See footnote, page 452. 

 VOL. XXXV. PART H. (NO. 12). 4 F 



