PRISMATIC STRUCTURE IN IGNEOUS ROCKS 223 



appears when the wax is bent. Dauzere pointed out the strong 

 probability that certain symmetrical columns in Auvergne have 

 been due to convection in the basalt in which they are formed. 1 



In a horizontal sheet of molten rock which has come to rest after 

 extrusion or intrusion it is obvious that we have some at least of 

 the conditions necessary for the formation of convection cells. If 

 the cells succeed in leaving any permanent record of themselves 

 when the sheet solidifies, then subsequent contraction may bring 

 out the structure by cracking the rock along the boundaries of the 



cells 



In general there are two ways in which the convection cells 

 might impress themselves on the crystallized rock. In the first 

 case the axes of the liquid convection cells and of the solid prisms 

 are coincident. Benard found that a finely powdered substance 

 which is heavier than the liquid tends to gather on the bottom of 

 the vessel in little heaps situated on the axes of the convection 

 cells giving an appearance from above of uniformly spaced round 

 spots A floating substance, on the other hand, gathers along the 

 boundaries of the cells at the surface. A substance in suspension 

 gathers within the interior portion of the closed curves of Fig. 2, 

 so that the liquid shows transparent both on the axes and along the 

 boundaries of the cells. In a mixture, therefore, in which different 

 crystalline phases are separating at different temperatures, a cer- 

 tain amount of segregation is to be expected, and the solid prisms 

 will coincide with the convection cells. 



In a substance which crystallizes as a unit, on the other hand, 

 whether it be a pure substance or a considerably undercooled mix- 

 ture prisms may be formed without segregation. Benard observed 

 that in spermaceti the crystallization began at the corners of the 

 cells In pure stearin Dauzere found that crystallization beginning 

 at centers on the cell boundaries extended uniformly in all directions 

 until the growing cylindrical groups intersected to form prisms It 

 is evident that in this case the prisms will not coincide with the 

 convection cells, but will nevertheless be symmetrical and regularly 

 spaced. 



1 C. Dauzere, Assn. franc, av. sci., 1908, pp. 436-38; also Longchambon, Bull. 

 Soc. Geol. Fr., XIII (1913), 33-38. 



