289 



with the exception of the alkaline ones, and the bromine 

 being supposed united with magnesium, and a deduction 

 made from the calcium for the small amount of sulphate of 

 lime present, it was easy, by giving to the manganese and 

 earthy metals their proper proportions of chlorine, to infer 

 the quantity of this principle united with the potassium and 

 sodium. Let this quantity, which was found greater than 

 what resulted from the direct determination of the chlorides 

 of these metals, be called w, and let m be to win the ratio of 

 the chloride of potassium to the chloride of sodium, as 

 already determined. Let x, also, be the true weight of the 

 chloride of potassium, y the true weight of the chloride of 

 sodium, a the ratio of the atomic weight of chloride of potas- 

 sium to chlorine, and b that of chloride of sodium to chlorine. 

 We will thus obviously have the two following equations : 



an -\-by — w ; and, nx =. my; 

 from which we readily deduce 



mw nw 



x = — -; and y = ■ — r . 



ma + no ' ma -f no 



From a fresh portion of the water the sulphuric acid was 

 got by nitrate of barytes, and the sulphate of barytes having 

 been separated, the chlorine and bromine were thrown down 

 in union with silver.* To infer, however, from this mixed 

 precipitate, the chlorine, it was necessary to estimate the 

 bromine present by a distinct process. 



With this view, a strong aqueous solution of chlorine 

 was mixed with a known weight of the water under analysis, 

 and the bromine liberated was removed by repeated wash- 

 ings with ether. From the etherial solution, the bromine mixed 

 with some chlorine was separated by barytic water, and the 

 ether being distilled off, the residue was evaporated to dry- 

 ness, and ignited, so as to reduce any bromate and chlorate 



* The excess of silver having been separated from the solution, by chloride of 

 sodium, the manganese was thrown down by the addition of bydrosulpbate of ammo- 

 nia. This metal was also estimated by the process recommended by Stromeyer* 



