190 



SCmNGE. 



[N. S. Vol. VII. No. 163. 



aay, with tlie expectation of finding tliat 

 tlie other chemist had made a mistake? 

 Judge of our surprise when we found that 

 our second analysis confirmed his figures 

 exactly. Our first and second analyses had 

 heen made by the same method, and by the 

 same operator, working on borings from the 

 same bottle, and the cause of the discrep- 

 ancy between the two was not, therefore, 

 at first sight apparent. On carefully ques- 

 tioning the operator, however, as to exactly 

 what he did at each step of the method, a 

 clew was obtained, which, when followed 

 out, cleared up the whole difiiculty and 

 ultimately led to a modification of the 

 method. The siHcon in these samples was 

 determined by what is known as Drown' s 

 method, which consists in dissolving the steel 

 in nitric acid, adding sulphuric, heating until 

 white fumes of the latter acid appear, to 

 render the silica insoluble, dilution with 

 water, filtration, washing and weighing. 

 The difference between our two analyses 

 consisted simply in this, that in the first 

 case, after the dilution with water, there 

 being considerable work in hand, the vessel 

 was allowed to stand overnight before fil- 

 tration, while in the second case filtt-ation 

 immediately followed dilution. Subsequent 

 work on this point showed that in this 

 method silica is not completely dehydrated 

 by heating in concentrated sulphuric acid in 

 presence of iron salts, but is apparently ren- 

 dered colloidal and sufficiently dehydrated, 

 so that if filtration follows soon after dilu- 

 tion fau'ly accurate results will be obtained. 

 On standing after dilution, however, this 

 colloidal, undehydrated silica, apparently 

 goes into solution again. Indeed, we were 

 able to get on this same sample, anywhere 

 from one-eighth up to the full amount of 

 sihcon present, by varying the time of 

 standing after dilution, the longest time 

 covered by our experiments being about 

 four days. 



Perhaps I may venture to give you still 



one more illustration of how, in the course 

 of analytical work, new and apparently 

 hitherto unnoticed reactions may be hit 

 upon and modifications of methods result. 

 Every chemist who has done much work in 

 determining phosphorus in iron or steel, by 

 the reduction of the molybdic acid of the 

 yellow ammonium phosphomolybdate and 

 subsequent titration of the reduced solution, 

 cannot fail to have been annoyed by the oc- 

 casional failure of duplicates to agree. Ap- 

 parently, in the two analj^ses everything 

 has been. done exactly aUke, and yet the 

 results do not agree. Every thoughtful 

 chemist cannot fail to have felt at such 

 times that somewhere in the method there 

 were conditions affecting the result that 

 were not fully controlled. During the last 

 six or eight months in our laboratory we 

 have apparently struck one of these hitherto 

 uncontrolled conditions, whose influence is 

 not large, and yet enough at times to cause 

 annoying discrepancies in duplicates, or be- 

 tween different chemists working on the 

 same sample. 



In order to make clear what follows, it 

 should be stated that in the ordinary work- 

 ing of this method the yellow precipitate, 

 after careful washing, is dissolved in am- 

 monia, and this solution is then treated with 

 sulphiiric acid largely in excess and diluted 

 to a definite volume, in which condition it 

 is passed through the reductor and subse- 

 quently titrated with standard potassium 

 manganate. The reductor in common use 

 consists, as is well known, of a tube of 

 heavy glass, about five-eighths of an inch 

 internal diameter, and about a foot long, 

 filled with powdered zinc, the top be- 

 ing fitted with a funnel, and the bottom 

 with a stopcock. Below the stopcock a 

 smaller tube carries the rubber cork by 

 means of which the reductor is fitted to the 

 flask which receives the reduced solution. 

 This smaller tube usually projects into the 

 flask an inch or two, and it is customarj' to 



