GEOPHYSICAL LABORATORY. 97 



(d) All barium sulphate precipitates contain barium chloride. If the pre- 

 cipitation is made slowly (3 to 6 minutes for 2 grams BaS0 4 ), the amount 

 of this is only about 0.15 per cent in the unignited precipitate, and since all 

 but a trace of it is eliminated as hydrochloric acid during ignition it is not a 

 source of error. When a precipitate is very rapidly formed the amount of 

 the barium chloride is multiplied several times and the chlorine is no longer 

 entirely eliminated on ignition. The more rapid the precipitation, the finer 

 the precipitate, and, since the barium chloride retained increases with the 

 fineness, we regard it as probably held by adsorption. The other impurities 

 are diminished by rapid precipitation. 



(e) Sodium nitrate, and probably other nitrates also, are occluded by 

 barium sulphate, giving results which are therefore too high. 



An exact determination of sulphur, so far as we know, can only be done 

 by correction for the above-named errors. Nitrates are to be avoided ; also 

 chlorides and ammonium salts as far as possible. Corrections for occlusion 

 and solubility can be made directly, but for volatility, unless one has a 

 specially constructed platinum apparatus, a duplicate precipitation must be 

 made under identical conditions. 



The error which led us to make this investigation, and which must be 

 expected if these precautions are neglected, not infrequently reaches 2 per 

 cent. A carefully corrected determination of sulphur for the cases given 

 should be accurate to 0.1 to 0.2 per cent of the total sulphur. 



(9) The exact determination of sulphur in pyrite and marcasite. E. T. Allen and John 

 Johnston. Journ. Industrial and Eng. Chem., vol. 2, 1. 1910. 



It has been shown that when the sulphides of iron are ground for analysis 

 they suffer partial oxidation to sulphur dioxide and ferrous sulphate. If 

 they are finely pulverized this error is not negligible ; it may be reduced to 

 0.05 per cent by gently crushing to 20-mesh size, which is sufficiently fine for 

 the proposed method of analysis. Inhomogeneous material, such as would 

 be met with in commercial work, would have to be pulverized for the pur- 

 pose of accurate sampling. In that case the error could be determined by 

 washing a weighed portion of the powder with boiled water in an atmos- 

 phere of carbon dioxide and determining the iron in the washings ; the opera- 

 tion is rather exacting on account of the readiness with which the moist sul- 

 phide oxidizes. 



In the proposed method the sample is oxidized in sealed tubes, according 

 to Carius ; this avoids all possible loss of sulphur. Lunge's and even Frese- 

 nius's method gave in our hands lower results, a fact which we attributed to 

 a loss of sulphur during oxidation, occasioned by the use of coarse material, 

 or perhaps, in Fresenius's method, to the presence of iron in the barium 

 sulphate. 



The iron is removed by two precipitations with sodium carbonate. The 

 object of this is to avoid ammonium salts, which cause much larger losses 

 of sulphur. In precipitating the barium sulphate, the conditions described 

 in the previous article should be followed. The sum of the errors in the 

 method, apart from that involved in the grinding, should not be over 0.2 per 

 cent of the total sulphur. The extreme variation in our own analyses was 

 less than 0.15. Incidentally, it was learned that the precipitation of iron by 

 ammonia involves much larger errors than has been generally supposed, if, 

 as is customary, the ammonia is kept in glass bottles and the precipitation is 

 made in glass beakers. 



7 — YB 



