158 PROCEEDINGS OF THE AMERICAN ACADEMY. 



different analyses of the same sample, that in the present state of our 

 knowledge it does not seem justifiable to reject the analyses of Samples 

 N and 0. This conclusion is supported by the fact that the water 

 determinations failed to show a difference between these samples. 

 The results, however, indicate that the average ratio 1.34562 (77.300 

 per cent of silver) may be very slightly too low, owing to the presence 

 of disilver hydrogen phosphate. The ratio 1.34562, assuming the 

 atomic weight of silver to be 107-88, and assuming that silver bromide 

 contains 57.4453 per cent of silver, leads to an atomic weight of 

 31.043 for phosphorus, whereas the ratio 1.34564 derived from Sam- 

 ples N and P gives the value 31.037. The rounded-off value, 31.04, 

 may be considered to be essentially free from error from this source. 



We are greatly indebted to the Carnegie Institution of Washington 

 for generous pecuniary assistance in pursuing this investigation ; also 

 to the Cyrus M. Warren Fund for Research in Harvard University for 

 many pieces of platinum apparatus. 



Summary. 



1. A careful study has been made of the conditions necessary for 

 the preparation of pure trisilver phosphate. 



2. It is found that silver phosphate can be almost completely dried 

 without fusion by heating in a current of dry air. 



3. The density of silver phosphate is found to be 6.37. 



4. It is found that silver phosphate does not adsorb a significant 

 amount of air. 



5. Nine analyses, made with four different samples, show that one 

 gram of silver phosphate yields 1.34562 grams of silver bromide, whence 

 the per cent of silver in silver phosphate is 77.300. 



Therefore, 



If Ag= 107.88 P = 31.04 



If Ag = 107.87 P = 31.03 



If Ag= 107.86 P = 31.02 



Cambridge, Mass., November 12, 1909. 



