216 PROCEEDINGS OP THE AMERICAN ACADEMY. 



but the relative weights observed in the first two experiments were 

 of no value, because it was evident that a no inconsiderable amount of 

 Sb 2 3 was lost by sublimation. Since, however, the small sublimate 

 of oxide condensed in the glass combustion tube very much nearer the 

 nacelle than the very much larger sublimate of chloride, we varied the 

 apparatus in our third experiment so far as to place the nacelle in a 

 tube of the shape represented in the accompanying figure. 



t 7 



D 



" This tube was weighed with the nacelle, and was so selected that it 

 quite closely fitted the combustion tube within which it was placed for 

 heating, as shown in the figure by dotted lines. And it is evident 

 that, while with this arrangement the SbCl 3 would be swept by the 

 C0 2 gas into the colder portion of the combustion tube, the greater 

 part at least of the sublimed oxide would be retained in the small 

 tube, which was of course at each stage weighed with the nacelle as at 

 first. ... It was evident from this determination that the order of 

 the decomposition was precisely that indicated by our reactions." 



In these experiments with the double haloids very little of the Sb 2 3 

 was volatilized, but proceeding in the described manner the exact 

 amount of resulting Sb 2 3 could be found, from which was readily cal- 

 culated the per cent of oxygen. 



The determinations of water of crystallization were made in the 

 usual way by heating in an air bath a weighed quantity of salt to con- 

 stant weight at a temperature of 100-105°. This will be described 

 more fully later. 



Method of Complete Analysis. — Obviously the ignition of the salt in 

 a current of dry carbon-dioxide affords a means of determining the 

 total amount of each element. The salt, dried at 100°, is weighed and 

 placed in the porcelain boat with the covering as described. Upon 

 ignition the SbCl 3 volatilizes, some Sb 2 3 also volatilizes, but is con- 

 densed on the covering tube. There is left as a residue from ignition 

 a mixture of KC1 and Sb 2 3 (if any combined oxygen is present in 

 the salt). Hence we see that the data are sufficient to give the exact 

 weight of SbCl 3 , Sb 2 3 , and KC1, from which can be calculated the 

 total Sb, CI, K, and O. This method, after abundant proof of its cor- 

 rectness, obtained by comparing the percentages of Sb and CI result- 

 ing from ignition with the direct determination of the total Sb and CI, 

 was applied in all the later determinations for the estimation of potas- 



