at its Maximum Density. 271 



have a tendency to descend, and consequently to leave 

 the surface of the metal, as soon as they begin to acquire 

 heat ; but experiment showed that, instead of descend- 

 ing, they were actually pushed upwards : this proves 

 that they were heated so rapidly that, before they had 

 time to leave the surface of the metal and to escape 

 from its calorific influence, they had acquired a temper- 

 ature so elevated that their density, after having passed 

 rapidly the point of its maximum, became even less than 

 it was at the temperature of melting ice. But after 

 some moments, the metallic body having cooled some- 

 what, and the communication of heat to the particles of 

 water taking place more slowly, these particles, having 

 become more dense on account of a slight increase 

 of temperature, had time to escape before becoming 

 warmer, and at that time the descending current sud- 

 denly began. 



This fact interests me the more, as it may serve in 

 some sort to explain a phenomenon which I observed in 

 an experiment made eight years ago, an account of 

 which I gave in my Essay on the Propagation of Heat 

 in Fluids.* 



The phenomenon to which I have alluded was this : 

 Having poured some mercury into a small cylindrical 

 glass vessel 2 inches in diameter and j| inches deep, 

 until this fluid filled the vessel to the height of an inch, 



1 poured on to the mercury twice as much water (that is, 



2 inches), and, plunging the vessel up to the level of the 

 upper surface of the mercury into a freezing mixture of 

 pounded ice and sea-salt, the temperature of the air 

 being 60 F., I allowed the whole to cool quietly, in 

 order to see in what part of the water the ice would first 



* See Vol. I. p. 357. 



