64 Mr. W. Makower on the Molecular 



lower surface of the plug, preventing diffusion from proceed- 

 ing as rapidly as theory demands. When a light gas such 

 as hydrogen is used, the air which enters the diffusion-vessel 

 through the plug would tend to fall through the hydrogen, 

 and the error due to the formation of a layer of air is greatly 

 reduced if not entirely obviated. That the failure of the 

 ordinary law of diffusion was not due to any peculiarity of 

 the plugs themselves was proved by testing them in the 

 manner described by Graham (Transactions Roy. Soc. Edin- 

 burgh, vol. xii. 1834). The rubber connecting the diffusion- 

 vessel with the adjustable mercury reservoir was detached, 

 and the open end of the vessel placed under water contained 

 in a tall cylinder ; the diffusion-vessel was then filled with 

 hydrogen and allowed to stand until all the hydrogen had 

 been displaced by air, water being added from time to time 

 to the cylinder to keep the level inside the diffusion-vessel 

 equal to the level outside. The volume at the end of the 

 experiment was to the original volume of hydrogen as the 

 square root of the density of hydrogen is to the square root 

 of the density of air, within the limits of experimental error. 

 It was of course impossible to adopt this method for the 

 emanation, since it depends on measurements of the volumes 

 of gases escaping and entering through the plugs. 



A series of experiments was therefore next made to find 

 out empirically how the rate of diffusion of a gas through 

 the plugs used depended on its density, and for the purpose 

 oxygen, carbon dioxide, and sulphur dioxide were used. 

 Unfortunately, before taking these observations the porosity 

 of plug I. was found to have changed *, so that the results 

 given below cannot be compared with those obtained above 

 with the same plug. Plug II. was removed from the second 

 diffusion-vessel and replaced by a similar but somewhat 

 thicker one (plug III.). Some observations were, however, 

 first taken with plug II., and these are quoted below. 



5. Final Experiments. 



Preparation and Analysis of (rases. — Oxygen. This gas 

 was prepared by heating pure dry potassium permanganate 

 contained in a horizontal glass tube closed at one end and 

 drawm off at the other, which was connected to a mercury 

 manometer by a T-piece, the other limb of which was 



* Owing' to an accident the stopcock of the diffusion-vessel was 

 broken and a new one attached. It is possible that during the operation 

 of making the glass joint near the plug, the change mentioned may have 

 occurred. 



