234 
TilE VOYAGE OF H.M.S. CHALLENGER. 
single analyses. Results, 2*474, 2*338, 2*544, 2*492, 2*362, mean of the six results = 
•j-4 40; probable error of the single determination = ±0*053. My number is 2*405, and 
lb. probable error lies at about±0*03G. I very much wondered at the close agreement 
of S.hmelck’s value for potash with mine, because he used an analytical method of which 
I should never have thought that it could yield anything better than a rough approxima- 
tion. What he did was to first precipitate lime and magnesia by Classen’s method,* as 
oxalates, to filter, evaporate to dryness, expel the ammonia-salts by ignition, and then 
to convert the sulphates into chlorides by repeated ignition with sal-ammoniac. From 
the “alkaline chlorides” thus obtained, the potash was separated by Fresenius’s method 
as chloroplatinate and weighed as such. 
Schmelck himself says that his alkaline chlorides alwnys contained magnesia 
and sulphuric acid, so that the chloroplatinate was contaminated correspondingly ; but 
he states that duplicate determinations agreed with one another. Now the Finkener 
method, as applied by me to sea- water, as I state in the context (page 16), owes its 
exactitude to some extent to a compensation of errors. I therefore considered it quite 
possible that Schmelck’s method might perhaps be more exact than mine, and for the 
purpose of a preliminary inquiry into the matter, caused Mr. Barbour to apply it to 50 
e.c. of a sea-water of known composition. The result was not very encouraging ; the 
chloroplatinate of potash obtained was obviously impure, and its weight short of what it 
ought to have been. This in itself, of course, might have been owing to Mr. Barbour’s 
want of practice, but about one thing the experiment left no doubt in my mind, namely, 
that the Classen method, when applied to sea- water, fails to effect anything like a satis- 
factory elimination of the magnesia. I did not consider it necessary to inquire into the 
sal-ammoniac method for substituting chlorine for the S0 4 of the sulphates, because I 
knew it to be tedious and unsatisfactory. But all this I thought might be rectified by a 
new combination of methods ; and supposing the problem of eliminating the alkalies of a 
given sea-water in the form of pure chlorides to be solved, Fresenius’s process might 
possibly be the most exact method for the determination of the potash in these. In 
order to settle this question, I caused Mr. John M‘Arthur to carry out the following 
two test experiments. In both the general scheme followed was the same, namely, as 
follows : — A known weight of chloride of sodium, and a known weight of a standard 
solution of pure chloride of potassium, equivalent conjointly to the potassium and sodium 
in about 100 grins, of ocean-water, are dissolved in water, the solution is mixed with a 
little more chloride of platinum than is needed to convert both metals into chloro- 
platinates, and the mixture evaporated on a water-bath very nearly to dryness. The 
r* ndue, aft« r cooling, is digested (cold) in 30 c.c. of pure (non-methylated) alcohol 
<.f 80 per cent, by volume, the liquid decanted through a small filter, and the residue 
continued to be exhausted with the same alcohol, until the last runnings, when tested 
* Kruscniufl, ZciUchr./iir anal. Chcmic, 1879, p. 374. 
