When phosphoric acid solutions which contain no iron, lime, alumina, or manganese, are precipitated in presence of ammonium citrate the results obtained vary markedly with the quantity of magnesia mixture employed. Grupe and Tollens[48] were the first to point out that a portion of the phosphoric acid might remain in solution, but that the precipitate might contain a sufficient excess of magnesia to compensate for the loss. It has been further shown by Glaser[49] that a portion of the phosphoric acid may be lost by volatilization in the citrate method. When the ignition is carried on in a crucible where the cover is coated with magnesia to intercept the volatilized acid, a considerable quantity of it can be recovered by the molybdic method. Where too little magnesia mixture is employed, therefore, two sources of loss are to be guarded against; _viz._, a part of the phosphoric acid may remain in solution and another part be volatilized on ignition. The explanation of the volatilization is as follows: In the presence of ammonium citrate, magnesium chlorid may be partly converted into magnesium citrate and ammonium chlorid. There may be a time, therefore, in the precipitation with not too great excess of magnesia mixture, when proportionally there is little magnesium chlorid and much ammonium chlorid present. The formation of a salt represented by the formula Mg(NH₄)₄(PO₄)₂ may take place which, upon ignition, breaks up into Mg(PO₃)₂ and finally passes into Mg₂P₂O₇ with loss of P₂O₅. This theoretical condition has but little weight, however, practically in the analysis of fertilizers, since in these cases a large quantity of lime is always present. But even in these cases traces of volatile P₂O₅ may be discovered. Wells[50] has shown that the citrate method gives good results in certain conditions but that this accuracy is reached by a fortunate compensation of errors. The ammonium magnesium salt does not precipitate all the phosphoric acid in this process, but contains enough impurities to make up for this loss. Johnson[51] in conjunction with Osborne has shown that the results by the citrate method practiced in accordance with the details laid down by Vögel, are too low, but that this difficulty could be overcome by using more and stronger magnesia mixture and a larger quantity of strong ammonia solution. The citrate method was found to give unsatisfactory results when iron and alumina were present in any considerable quantity. In the examination of the final ignited precipitate, which should be pure magnesium pyrophosphate, it was found to consist of only 94.98 to 97.83 per cent of that salt. The chief impurity found was calcium oxid, the percentage of which varied from 2.05 to 3.95 in six cases. There was also a considerable percentage of loss due, probably, to magnesia and pyrophosphoric acid.