70 PROCEEDINGS OF THE AMERICAN ACADEMY 



the mean error being about -j-0.0013 gr. or -f-0.71 per cent.; while 

 in (54) to (58) the error varies from to -(-0.0012 gr. upon 0.1831 gr. 

 or 0.1875 gr. of phosphoric oxide, or from to -j-0.64 per cent., the 

 mean error being about -|-0.0004gr. or -4-0.22 per cent. ; and it is to 

 be remarked that in (54), (55) and (5G), in which dilute ammonia was 

 added after the non-ammoniacal magnesic chloride mixture, the mean 

 error — about -f-0.0002 gr. or -f-O.lO per cent. — is much smaller 

 than the mean error of (57) and (58) — about -4-0.0008 gr., or 

 -(-0.44 per cent. — in which strong ammonia was added after the non- 

 ammoniacal magnesic chloride mixture. The high results of experi- 

 ments (59) and (60), in which the non-ammoniacal magnesic chloride 

 mixture was added to the hot phosphate solution, are explained by 

 the fact that a part of the hydromagnesic phosphate first formed was 

 decomposed by the heat with the formation of trimagnesic phosphate, 

 and therefore not completely converted into ammonio-magnesic phos- 

 phate by the ammonia subsequently added. From experiments (61) 

 to (64) it will be seen that the presence of 3 gr. of sodic chloride in 

 the phosphate solution raises the figures of the analysis materially when 

 precipitation is effected by the ammoniacal magnesic chloride mixture, 

 even with small weights of phosphoric oxide, and that better results 

 are got in (65) and (66) by using the non-ammoniacal mixture. 



The use of an excessive quantity, too, of ammonic chloride tends 

 apparently to reduce the error. Thus, in experiments (48) to (53) 

 the mean error was about -j-0.0013, gr., or -4-0.71 per cent., while in 

 (44) and (45), in which free ammonia, but no ammonic chloride be- 

 yond the amount in the magnesic chloride mixture, was used, the 

 mean error is -{-0.0022 or -4-1.17 per cent., although less of the pre- 

 cipitant was used in the latter. 



The cause of all these differences in the accuracy of the determina- 

 tion of the phosphoric oxide of alkaline phosphates by precipitation 

 with magnesic chloride mixture seems to lie in the variations of the 

 rapidity with which the ammonio-magnesic phosphate is crystallized 

 from solution. The sudden addition of a large amount of the precip- 

 itant, or of an excess of the precipitant to a strongly ammoniacal 

 solution of the phosphate, or of strong ammonia to a solution of the 

 precipitate in acid, tends to hasten the deposition, and so to prevent 

 the complete exclusion of foreign material from the crystalline struc- 

 ture ; while in the precipitation of solutions containing large amounts 

 of ammonic chloride, or of weakly ammoniacal solutions, or of solu- 

 tions of the precipitate in acid, by the gradual addition of ammonia 

 the process of crystallization goes on more slowly and perfectly. 



