46 PROCEEDINGS OF THE AMERICAN ACADEMY 



antimony, not with any expectation of coi-recting the atomic weight 

 of sulphur, but with a view possibly of verifying the higher value ob- 

 tained by Stas, and more especially of still further testing the accuracy 

 of our method of precipitating' sulphides ; as it was obvious that any • 

 hidden sources of error which could have impaired the accuracy of our 

 antimony determinations might be expected to reappear in the experi- 

 ments with silver, and then the well-established composition of sulphide 

 of silver would help us to detect them. 



We began these experiments by taking two adjacent portions of 

 the same piece of pure silver foil, and, having dissolved each in nitric 

 acid, we in the first place precipitated the silver fi'om both solutions, as 

 chloride, with the usual precautions. The argentic chloride from the 

 first solution was washed, collected, and weighed as before described. 

 That from the second solution, having been thoroughly washed by the 

 process of reverse filtering, was redissolved in the same vessel with 

 pure aqua-ammonia, and from this solution the silver was precipitated 

 as sulphide by adding the supersaturated solution of hydric sulphide, 

 using the same precautions with which we were familiar in the precipi- 

 tation of sulphide of antimony. The sulphide became granular on 

 boiling, and was readily washed and collected by the method of reverse 

 filtering. "We thus hoped to obtain a direct comparison of the atomic 

 weights of chlorine and sulphur, not depending on the absolute purity 

 of the metallic silver used, and moreover to obtain a confirmation or 

 otherwise of the general accuracy of our method of determining sul- 

 phide of antimony ; for it was obvious that the same causes of error 

 were likely to inhere in two such similar processes. Hence, although 

 such experiment would probably only confirm values already well 

 established, such a confirmation would give us 'confidence in the accu- 

 racy of our previous work. But although the mechanical details of 

 the process appeared perfect, and the results were not inconsistent 

 with the accepted values of the weights under discussion, yet they were 

 neither sufficiently sharp nor constant to answer the questions we had 

 proposed, owing probably to some slight solvent action of the ammo- 

 niacal menstruum on the precipitated sulphide of silver. We were 

 therefore led to modify the process by first preparing pure sulphide of 

 silver by the method we have described, and then determining the 

 relation of silver to sulphur by reducing weighed portions of this 

 sulphide in a stream of hydrogen gas. This result, compared with 

 the already well-known relation of silver to chlorine, — probably the 

 most accurately determined of all the atomic ratios, — would give us 

 the relation of sulphur to chlorine which we sought, and under essen- 



