82 



from a smaller one attached to it, according to the method 

 adopted by the author in the determination of the vapour 

 density of vanadium tetrachloride. The narrow neck of the 

 bulb was kept open during the experiment by inserting a 

 platinum wire, and after the sulphur or the mercury had 

 been boiling for some minutes the neck was sealed. 



The results of three experiments in sulphur vapour at 

 440° gave the density (H=l) as (1) 167-8, (2) 1097, (3) 

 168*8, Two determinations in mercury vapour at 350° 

 gave (1) 1907, (2) 191-2.* The fact of the alteration of the 

 vapour density from 190 at 350° (closely approaching the 

 normal density 198-5) to 167 at 440° shows pretty clearly 

 that the anomalous vapour density is to be ascribed rather 

 to dissociation than explained by Persoz's suggestion of an 

 error in the atomic weight; and this conclusion is fully 

 borne out by further experiments detailed in the sequel. 



The residual chloride from the bulb possesses the same 

 properties and composition as the original substance, there 

 is no trace of free chlorine found in the cold bulb, nor does 

 the colour of the vapour of the hexachloride change when 

 it is strongly heated. 



On heating the residue with water, a difference between 

 its behaviour and that of the original hexachloride can how- 

 ever be detected, as the residue yielded an oxide which was 

 perfectly yellow, but had a greenish colour, showing the 

 existence of traces of oxides lower than WO3, although pre- 

 sent in too small quantity to affect the analysis. 



In order to ascertain whether the gaseous hexachloride 



is decomposed at high temperatures, a portion of the ])ure 



chloride was distilled upwards in a current of dry carbonic 



acid for several hours. A continuous liberation of clilorine 



was clearly shown to occur, for, on passing the exit carbonic 



acid through a solution of potassium iodide considerable 



* Eieth has lately determined the vapour density of " Wolfram Chlorid," 

 showing that its molecule contains 187 instead of 184 of metal, but there is 

 nothing to show whether the substance thus examined was the hexa- or the 

 penta-chloride. 



