120 THE ATOMIC WEIGHTS. 



In the second series of experiments a small excess of silver was added 

 as before, and the jn'ecipitate of silver chloride was removed by filtra- 

 tion. The filtrate and wash waters were concentrated to small bulk 

 whereupon a trace of silver chloride was obtained and taken into account. 

 The excess of silver remaining was then thrown down as silver bromide, 

 and from the weight of tlie latter the silver was calculated, and sub- 

 tracted from the original amount. 



6.59993 

 5-55229 

 4.06380 



.0054 



The third series involved mixing solutions of barium chloride and 

 silver in as nearly as possible equivalent amounts, and then determining 

 the actual quantities of silver and chlorine left unprecipitated. The 

 filtrate and wash waters were divided into two ]iortions, one-half being 

 evaporated with hydrobromic acid and the other with silver nitrate. 

 The small amounts of silver bromide and chloride thus obtained were 

 determined by reduction and the use of Volhard's method : 



Ag. BaCI,. 



4.4355 4.2815 



2.7440 2.6488 



6.1S65 5-9712 



34023 3-2841 



Mean, 96.526, ± .0035 



Two final experiments were carried out by Stas' metliod, somewhat as 

 in the first series, with variations and greater refinement in the observa- 

 tion of the end point. The results were as follows : 



Ag. 

 6.7342 

 10.6023 



Mean, 96.524, -±z .0007 



A careful study of Richards' paper will show that, altbough the last 

 two experiments are probably the best, they are not entitled to such pre- 

 ])onderance of weight as the " probable error" here computed would 

 give them. I therefore treat Richards' work as I have already done tliat 

 of Marignac and Dumas, regarding all of his series as one, which gives for 

 the value of the ratio 96.520, rt .0025. This combines with the previous 

 series thus : 



