18 KEPORT — 1877. 



From these experiments it appears that the employment of the processes 

 of Fresenius and Frank leads to results sensibly above the truth if a largo 

 excess of platinum he employed. The fact that in all the experiments the 

 error is in the same direction, indicates that it is not due to defective mani- 

 pulation. When only a slight excess of platinum is employed in the above 

 methods the results are decidedly better, but present greater differences among 

 themselves, as if some other disturbing cause came into operation. This is 

 notably the case with Frank's method, the error in only six experiments 

 varying from "01 per cent, to 1'07. 



The residts by Tatlock's method distinctly indicate a tendency to loss ; but 

 this is chiefly noticeable in the cases in which the proportion of sodium 

 chloride was very high (50 per cent.). In fact four experiments with a 

 mixture similar to that which usually occurs in practice (i. e. V2 per cent. 

 KC1 and 18 JSTaCl) gave results showing an error in excess of the truth 

 varying from -09 to -20 per cent. The thirteen determinations by Tatlock's 

 method show a maximum error of — '40 per cent. 



In this experiment the quantity of material employed was measured in 

 the pipette, and for several reasons this plan was found lese trustworthy 

 than the weighing of the solution used. 



With the view of ascertaining the cause of the loss observed in some cases 

 by Tatlock's method in presence of a large proportion of sodium chloride, an 

 experiment was made by treating a mixture of 30 milligrammes of KC1 and 

 •7 gramme of pure jN T aCl with 30 c. c. of the platinum solution (the usual 

 quantity), and estimating the potassium in the usual way. 



By employing a small quantity of KC1 it was thought that other errors of 

 manipulation would be avoided, and that the experiment would be practically 

 to ascertain the extent to which chloroplatinate of potassium was soluble in 

 a solution of platinic chloride containing much chloride of sodium (or, in 

 other words, in a solution of sodium chloroplatinate). The weight of potas- 

 sium chloroplatinate which should have been yielded by the above quantity 

 of KG is "0982 gramme, whereas the weight actually obtained was only 

 •0915 gramme. Hence there was a loss of "0067 gramme. In another 

 experiment in which only -35 gramme of Nad were used, the quantities of 

 KG and platinum solution remaining as before, a loss of -0042 gramme of 

 chloroplatinate was observed. In this last experiment the potassium chloride 

 corresponding to the chloroplatinate obtained was only 95-7 per cent, of the 

 quantity added, while in the previous experiment it amounted to only 93'2 

 per cent. From these residts, and those recorded in the Tables, it appears 

 that the percentage error is larger the greater the proportion of sodium salts 

 present, a fact which appears to point to the solubility of the precipitate in 

 solution of sodium chloroplatinate as the origin of the loss. Thus in the 

 experiments in which pure chloride of potassium was employed, and in those 

 in which the amount of sodium chloride was small, the variation from the 

 truth was exceedingly slight, but the errors became greater with the amount 

 of sodium chloride present. In experiments 42 to 50 the amounts of chloride 

 of sodium and platinum solution employed were the same as in the test ex- 

 periment, in which a deficiency of '0042 gramme of precipitate was observed. 

 If we assume that this loss is the weight of K.,PtCl c dissolved by the use 

 of -35 gramme of NaCl and 30 c. c. of platinum solution, then a correction 

 of -0042 gramme ought to be applied to each of the results of experiments 42 

 to 50. This correction of -0042 gramme in the weight of the precipitate 

 corresponds to "37 per cent, of KG. The mean of the nine experiments 

 above referred to is 99*58 per cent, of KG, which, with the correction -37, 

 amounts to 99-95 per cent. 



