FREEZING OF WATER. 37 



greater than might have been produced from the fame 

 water in the other proceffes in a reafonable time ; and in 

 them the production of air had no hmit. 



Difappointed in my expectations of getting by this 

 means ice perfedly free from air, (which when a large 

 quantity of water freezes very flowly it is eafy to do, 

 the air contained in it retiring from that which is frozen 

 to that which remains fluid) 1 diffolved ice that was per- 

 fectly tranfparent, and therefore free from air, in veffels 

 containing mercury, and expofed it to froft a fecond 

 time. But I always found that when the whole of it 

 was frozen, though the extreme parts were tranf- 

 parent, and therefore free from air, the central parts 

 were opaque, and when dilTolved yielded air. And 

 though I repeated this procefs ten or a dozen times with 

 the fame water, always letting out the air that was pro- 

 cured by freezing prefently after it was extricated under 

 mercury, and before it could have been reabforbed, yet on 

 expofmg it to another freezing, I never failed to get more 

 air ; and the harder the froft was the more air I procured. 



As there is an evaporation from ice, no lefs than from 

 water, the interftices formed by the cryftallization of the 

 water when it is converted into ice will foon be filled 

 with vapour ; and this vapour, like that which is form- 

 ed by heat, becomes, 1 fuppofe, the bafis of a quantity 

 of air. Since, however, ice that is the moft ti'anfparent 

 fwims in water, this alfo muft have interftices ; but they 

 contain no air ; being fuch as exift in the moft folid 

 bodies, in which (gold itfelf not excepted) the com- 

 ponent particles are not in perfeCt contact ; fmce they are 

 reduced into lefs dimenfions by cold. 



As the veffels I made ufe of in thefe experiments were 

 either cylindrical jars, or conical wine glaifes, and con- 

 fequently the bubbles of air procured by freezing were 

 expofed to a confiderable furface of water, and would in 



time 



