in Fluids. 285 



is changed to -ice; but when, in cooling, its temperature 

 has reached to the 4<Dth degree of Fahrenheit's scale, or 

 eight degrees above freezing, it ceases to be farther con- 

 densed ; and on being cooled still farther, it actually ex- 

 pands, and continues to expand as it goes on. to lose more 

 of its Heat, till at last it freezes ; and at the moment 

 when it becomes solid, and even after it has become solid, 

 it expands still more on growing colder. This fact, which 

 is noticed by M.. de Luc in his excellent treatise on the 

 modifications of the atmosphere, has since been farther in- 

 vestigated and put beyond all doubt by Sir Charles Blag- 

 den. See Philosophical Transactions, Vol. LXXVIII. 



Now, as water in contact with melting ice is always at 

 the temperature of 32, it is evident that water at that 

 temperature must be specifically lighter than water which 

 is eight degrees warmer, or at the temperature of 40; 

 consequently, if two parcels of water at these two tem- 

 peratures be contained in the same vessel, that which is 

 the coldest and lightest must necessarily give place to 

 that which is warmer and heavier, and currents of the 

 warmer water will descend in that which is colder. 



In the two last experiments, as the circular tin plate 

 which covered the surface of the ice served to confine 

 the thin sheet of water which was between the plate and 

 the ice, as this water could not rise upwards, being hin- 

 dered by the plate, and as it had no tendency to descend, 

 it is probable that it remained in its place ; and as it was 

 ice-cold, it was not capable of melting the ice on which 

 it reposed. 



But as the tin plate had a circular hole in its center, 

 the surface of the ice in that part was of course naked, 

 and, the ice-cold water in contact with it being displaced 

 by the warmer and heavier water from above, an excava- 



