12 J. A1TKEN ON THE NUMBER OF DUST PARTICLES IN THE ATMOSPHERE. 



lime after time the air would be perfectly free from all traces of condensation when 

 expansion was made, then apparently without reason it would break down, a dense 

 shower of rain would fall, and then the next time perhaps none, but only to be followed 

 by failure before long. 



As we shall see, there were many causes for the failure of the filtered air to keep free 

 from condensation when expansion was made. These we shall consider separately. This 

 failure, of the air to keep clear, at first seemed unaccountable with so large a factor of 

 safety as was given by 12 inches of cotton wool, which is four times what was found to 

 be required for perfect filtration under what appeared to be the same conditions. The 

 first thing that was suspected was some fault in the joints of the tubes where the metal 

 or glass joined the india-rubber. When any new failure occurred this was always the 

 first thing suspected. Time after time were the joints taken down, and remade with a 

 solution of india-rubber, and the stopcocks cleaned, greased, and tested ; but in almost 

 no instance were these found to be at fault, as india-rubber solution makes a perfect joint 

 so long as it has not been severely strained. 



One explanation of the failure seemed to be that after all the filtering might not be 

 perfect, that the filter only kept back the larger particles while it passed the extremely 

 small ones. Sir William Thomson has concluded from certain phenomena of capillarity 

 that the vapour pressure at a concave surface is less than that at a flat one ; and Professor 

 Clerk Maxwell has extended this conclusion to convex surfaces, and has shown that the 

 vapour pressure at a convex surface will be greater than that at a flat one, and that the 

 smaller the body, that is the quicker its curvature, the higher will the vapour pressure be 

 at its surface. From this we see that air which is saturated at a flat surface is not 

 saturated at a convex one, and also that the smaller the body the higher will the super- 

 saturation require to be to produce condensation on it. In our ignorance, it seemed from 

 this just possible that the degree of supersaturation produced by the amount of expansion 

 used in the experiments might just be about enough to cause condensation on these 

 extremely small particles, so that in one test the degree might be exceeded, while in the 

 next it might not be reached. 



These considerations pointed to the necessity of studying this point separately. 

 Given a quantity of ordinary air, we have, in a previous communication, shown reasons 

 for supposing that condensation begins before supersaturation is reached, owing to an 

 affinity between the material of the dust particles and water vapour. When supersaturation 

 is made in this air, these particles, owing to their affinity for water vapour, are the first 

 to become visible by the vapour condensing on them. Suppose these are allowed to settle, 

 then, when the next expansion and supersaturation is made, the probability is that the 

 largest particles will take the next burden of vapour, and that the smaller ones will be 

 left to be taken down by a subsequent supersaturation, till at last the smallest particles 

 only are left to be brought down in a similar manner. This is theoretically what we 

 might expect to happen, because the smaller the particle the higher will the vapour 

 pressure be at its surface, and the higher therefore the supersaturation must be before 



