FACTORS IN COLUMNAR STRUCTURE IN LAVAS 459 



to contraction on cooling is able to overcome this power of extension, 

 the rock cracks. 



Let us consider in the first place a lava mass cooling from one sur- 

 face only, the top of the flow (Fig. i). It is evident that the surface 

 will cool very rapidly owing to the scoriaceous nature of the surface, 

 the convection currents in the air above the heated rock, the vol- 

 canic rain (if near a vent) , and the collection of water on the warped 

 surface. Thus a solid crust will form over the molten lava and 

 this liquid will cool very slowly 

 because rock is a very poor con- 

 ductor of heat. While cooling and 

 shrinking in the liquid state, it 

 will adjust itself to the tension by ^ ^ . , . , 



riG. I. — Section showing now 



a movement of the Uquid, but columns are formed in the brittle 



when the lava solidifies from b to solid lava, when tension due to 



C (Fig. l) the rock is subjected contraction overcomes the expansive 



. power of the lava. Columns are not 



to tension m h o r 1 z o n t a 1 and ^^^^^^ „^^^ ^^^ ^^^^^^^ 

 vertical directions. The vertical 



tension is relieved by movement of the liquid beneath c and by the 

 shortening of the column be. Vertical tension at this stage does 

 not cause joints in the rock but the horizontal tension can only 

 be relieved by cracking. Mallet' puts the temperature of cracking 

 between 315° C. and 500° C, and points out that the temperature 

 is lower when the rate of cooling is very slow. He also shows that 

 if the rock were perfectly homogeneous the surface would be 

 covered with cracks separating the surface into equal areas, i.e., 

 triangles, squares, or hexagons.^ The smallest amount of work is 

 done by separating the surface into hexagons rather than into 

 squares or triangles. These cracks would extend down into the hot 

 sohd rock only a short distance, i.e., to that point where the tension 

 was equal to the extensibihty. As the rock cooled, the cracks 

 would extend downward and at the same time the deeper Hquid 

 lava would be solidifying and shrinking from two causes: cooling 

 and crystallization. The slower the cooHng the more complete the 

 crystallization and therefore the greater the contraction. We 



' R. Mallet, Phil. Mag., L (1875), p. 130. 

 ""Op. cit.,pp- 125-30. 



