OF ARTS AND SCIENCES. 45 



minations, Nos. 14 to 17, in the table on page 40. The material for 

 these analyses was first purified by repeated distillations as before, first 

 over powdered antimony and then over zinc. Through the melted 

 chloride was then passed, for a short time, a current of dry chlorine 

 gas, in order to make sure that no such thing as a subchloride of anti- 

 mony, if it can exist, or particles of metallic antimony, could be present. 

 The chloride was then again redistilled several times, and this product 

 purified by ten or twelve repeated crystallizations from a solution in 

 disulphide of carbon, the material being protected all the time as far as 

 possible from contact with moist air. The few grammes of pure 

 chloride thus obtained from more than a kilogramme of so-called pure 

 commercial chloride of antimony were submitted to analysis, and, as will 

 be noticed, the results were in complete accordance with those we had 

 obtained before. Lastly, we found nearly the same result, with only 

 the little loss that was to be expected, when the antimony was removed 

 from the solution before precipitating the chlorine. 



Now, it is evident that, if the sulphide of antimony we weighed is pure, 

 we are forced, even by these last analyses, to the conclusion that the 

 atomic weight of antimony must be very nearly at least 120, if that of 

 sulphur is 32, although the singular discrepancy which our results pre- 

 sented served at the time to I'ender the problem exceedingly puzzling. 

 The facts indeed seemed to indicate that, while antimony combined with 

 clilorine in the proportion of 122 to 3 times 35.5, it combined with sulphur 

 in the proportion of 120 to | times 32, or 122 to f times 32.53 ; in other 

 words, that the relation of the atomic weights of chlorine and sulphur 

 was not as accepted 35.5 to 32, but 35.5 to 32.53, And, although, after 

 the investigations of Dumas, Stas, Marignac, and others, it was clearly 

 out of the question that these values should be in error to the extent 

 indicated, yet, as we have seen, Stas had found for the atomic weight of 

 sulphur 32.074, and the results of our synthesis of sulphide of anti- 

 mony calculated on this basis would give for the atomic weight of anti- 

 mony 120.28. Moreover, it appeared that when the analyses of 

 antimonious chloride made by Dumas were recalculated with Stas's 

 values of the atomic weights of chlorine and silver (01^35.457, 

 Ag = 107.93), they gave for the same atomic weight the number 

 121.95. This reduced the difference between the two determinations 

 to 1.67, and it did not seem impossible that the whole discrepancy 

 might result from the accumulation of a number of similar small errors. 

 "We were thus led to undertake a new comparison of the atomic weights 

 of chlorine and sulphur, based on the precipitation of sulphide of silver, 

 by the same process we had employed in precipitating sulphide of 



