OM THE ESTIMATION OF POTASH AN I) PHOSPHORIC ACID. 29 



so closely, even after considerable practice. A difference of 0-1 cub. centim. 

 in the volume of the standard alkali employed corresponds in the last three 

 experiments to about -2.3 per cent, of the chloride of potassium taken, and 

 In the first six experiments to '5 per cent, of the sample. As the quantity of 

 potassium chloride worked on cannot be conveniently increased beyond the 

 weights used in experiments 10, 11, and 12, it is evident that the process 

 is not susceptible of great accuracy even if no other disturbing influence 

 existed. 



The fact that the volumetric method gives results below the truth is pro- 

 bably due to the difficulty of decomposing tho last traces of silicofluoride by 

 alkali, without introducing an excess of the latter. The trace of free 

 alkali which suffices to change the tint of the litmus to blue seems incapable 

 of reacting on the silicofluoride. An attempt was made to overcome this 

 difficulty by adding a sensible excess of alkali, boiling well, and titrating back 

 with standard acid ; but the result was not satisfactory, the end of the 

 reaction being very obscure. 



In practice it would be preferable to sot the standard alkali by its action 

 on moist silicofluoride prepared from a known quantity of potassium chloride, 

 rather than to trust to its theoretical neutralizing effect. 



As the drying and weighing of the silicofluoride requires but little more 

 time than the titration with alkali, and gives better results, the gravimetric 

 estimation is to be preferred. Although the process is not to be compared in 

 accuracy to the precipitation and weighing of potassium as potassium 

 chloroplatinate, it might no doubt be advantageously employed in particular 

 cases. 



The next experiments were made on a mixture of 75 per cent, of chloride 

 of potassium with 25 per cent, of chloride of sodium. 



-p, . . Wt. of KC1 sola. KaC'l C. c. nurmal _ j-,-,, „ , KC1 found for 



1 ' taken. taken. alkali used. ' 100 parts taken. 



13 .. ..11-7505 -375 42-95 S 1-6019 136-3 



14 ....11-7530 -375 42-95 S 1-6019 136-3 



These results show that 96-9 per cent, of the total amount of alkali metal 

 present was precipitated as silicofluoride. In two other experiments of 

 equal weights of potassium and sodium chlorides, 222 and 218 parts of Iv( 1 

 were found for 100 parts taken. The former number represents a precipi- 

 tation of 96-8 per cent, of the sum of the alkali metals present. 



Two experiments were next made on mixtures of potassium and sodium 

 chlorides by precipitating the solution with hydrofluosilicic acid as before, 

 but using a smaller proportion of spirit. One third of the bulk of solution 

 and wash water consisted of rectified spirit, instead of one half, as in all 

 previous experiments. The weights of the precipitates corresponded respec- 

 tively to 109 and 188 parts of potassium chloride for 100 parts taken. 



It is evident from these experiments that the process is quite worthless 

 for the separation of potassium from sodium, and consequently that the 

 number of cases in which it can he advantageousl}' employed is greatly 

 limited. Althoivgh this result was anticipated from the known properties of 

 sodium silicofluoride, it was thought desirable to establish the fact by direct 

 experiment. 



Since the above experiments were completed the original paper of Stolba 

 has been consulted *. The author recommends the suspension of the preci- 

 pitated silicofluoride in a much larger quantity of water than was employed 



* ZeiUch. lur anal. Chem. iii. p. 298. 



