A JXote on Thermal Diffusion. 251 



process would take place more slowly than with the previous 

 one. In the next experiment, therefore, eighteen hours were 

 allowed, and the results were 



,~. J Hot bulb (220° C.) 63-2 per cent. H 2 : 36-8 per cent. SO s 

 [0 ' \ Cold bulb (10° C.) 59-7 percent. H 2 : 403 percent. SO s 



Some experiments w T ere also undertaken on a mixture of 

 hydrogen and hydrochloric acid gas, but they were not con- 

 tinued since in every case there were indications ihat the acid 

 had attacked the alkali of the glass. 



The effect of thermal diffusion was verified qualitatively 

 also in another way. The hot and cold bulbs were con- 

 nected to the two sides of a tube (A) divided in the middle 

 by a porous diaphragm. After the bulbs had been main- 

 tained at their lespective temperatures sufficiently long for 

 thermal diffusion to act, the, stopcock between them was 

 closed, and the gases were allowed to cool. The side taps 

 connecting with A, which had till then been closed, were 

 opened so as to admit the mixtures in the hot and cold bulbs 

 to their respective portions of the tube A. The pressure 

 was adjusted so as to be initially the same on the two sides 

 of the diaphmgm. The difference of composition of the 

 mixtures on the two sides quickly manifested itself by the 

 production of an inequality of pressure between them, owing 

 to the more rapid diffusion and equalization of density of 

 the lighter constituent of the mixture. A paraffin pressure- 

 gauge showed differences of level amounting to 10 mm. or 

 less, on the two sides, the gases used being hydrogen and 

 carbon dioxide, and hydrogen and sulphur dioxide. In each- 

 case the sign of the difference indicated an initial excess of 

 the lighter gas on the side of the hot bulb. 



We will now briefly consider the theoretical difference of 

 composition to be expected in the cases (1), (2), (3), and (5); 

 in case (4) the steady state would appear not to have been 

 attained, so that we may neglect the figures for that, 

 experiment. We will write 



^l = V( v i + *>«)> ^2 =v 2 /(v 1 + v 2 ), 



The equation (a) may be conveniently re-written as follows : — 



~dx "dx d# 



It is easy to calculate' a first approximation to k f on the 



