388 Of the Propagation of Heat 



winter, the water was frozen, and formed a cake of ice 

 at its surface about an inch thick. Letting the dish re- 

 main in its place, in order that the surface of the ice 

 might remain perfectly horizontal (which was necessary 

 to the complete success of the experiment, as will pres- 

 ently be seen), I entered the room with a chafing-dish 

 filled with live coals, in the midst of which was my iron 

 bullet, perfectly red-hot ; and taking out the bullet from 

 among those burning coals, I held it over the center of 

 this horizontal sheet of iee, and distant from it about 

 y^Q- of an inch. 



The ice directly under the red-hot bullet was soon 

 thawed, but the depth to which it was thawed was very 

 inconsiderable ; the water, however, extended itself 

 slowly from the center towards the circumference, and at 

 length a circular spot 2 or 3 inches in diameter in the 

 center of the surface of the ice was covered with it, 

 though but to a very inconsiderable depth. 



This little spreading sea appeared to prey on the wall 

 of ice by which it was surrounded on every side. 



The particles of water in contact with this wall being 

 rendered specifically lighter on becoming ice-cold, they 

 move upwards, and, making way for other warmer parti- 

 cles to advance from below, cause currents in opposite 

 directions to set between the center (where the hot iron 

 remains) and the circumference. As a current at the 

 temperature of 41° must necessarily set downwards at 

 the middle of the circle, this current, striking against the 

 middle of the excavation formed in the ice, ought to 

 deepen it gradually in that part, though but slowly, — 

 and this is what was actually found to be the case ; for 

 the bottom of this excavation was not perfectly flat, but 

 was deeper at and near its center than at its sides. 



