62 PROCEEDINGS OF THE AMERICAN ACADEMY 



the very constant per cent of chlorine, the assumption that any material 

 amount of impurity could be present appeared wholly untenable. Yet 

 we have seen how this assumption was forced back upon us by the 

 subsequent results of the investigation. 



Returning to the subject after our experiments with iodide of anti- 

 mony, we, for the first time, fully appreciated how very small an amount 

 of oxygen — the only real impurity present — was required to reduce 

 the per cent of chlorine in autimonious chloride from 47.02, the amount 

 corresponding to Sb= 120, to 46.61, which corresponds to Sb = 122 ; 

 for, while the effect is so differently produced, yet the result of the 

 action of the atmosphere on the chloride of antimony is wholly like 

 that of its action on the iodide. It ends in replacing a small amount 

 of chlorine by oxygen ; and although, in consequence of the smaller 

 atomic weight of chlorine, it requires in this last case a larger replace- 

 ment to produce a corresponding change of percentage composition, 

 yet still the amount required to make all the difference in question is 

 very small ; so that, when we come to sum up the supposed completed 

 results (as on page 42), it might easily be covered up by slight inaccu- 

 racies of the analytical work. An easy calculation will show that the 

 substitution of but -r^oViT ^^ <^"® P®'" <^®"*' ^^ oxygen for the equivalent 

 amount of chlorine would reduce the per cent of this last element in 

 the chloride from 47.020, corresponding to Sb = 120, to 46.608, which 

 corresponds to Sb= 122; and such a substitution would result from 

 the absorption of only 1^^ milligrammes of water by each gramme of 

 the chloride. The composition of the material would then be as 

 follows : — 



(IJomposition of Antimonious Chloride with tVj^ % of O when 

 Cl = 35.5 and Sb= 120. 



Chlorine 46.608 



Oxygen .146 



Antimony 53.246 



100.000 



Now it will be seen by referring to the tables, on pages 40 and 42, that 

 these percentages do not differ from the mean of the results of our pre- 

 vious analyses as much as these results differ among themselves ; and we 

 therefore determined to repeat these analyses, hoping that the experi- 

 ence we had acquired in both chlorine and antimony determinations 

 would now enable us to obtain results sufficiently sharp to show even 



