184 JV. L. Bowen — Crystallization — 



perature gradient such as cannot reasonably be supposed to 

 exist in any considerable magma body except, perhaps, quite 

 close to its margin. The thermal expansion of water is very 

 considerable. Between 4° C. and 100° C. it expands more 

 than 4 per cent of its volume. Diabase glass (undercooled 

 liquid) expands only a little more, about 6 per cent, in a range 

 of more than 1100°, its kilogram volume being 362 at 25° and 

 384 at 1150°.* The viscosity of water must also be considered 

 very much less than that of magmas. Each individual factor 

 affecting convection is decidedly in favor of the familiar case 

 of a beaker of water placed over a flame. It is not to be 

 expected that the vigor of convection in a body of magma 

 would even remotely approach that established in the beaker. 



Vigorous convection, by causing a general stirring of the 

 liquid, might seriously interfere with the sinking of crystals. 

 It has been shown above that convection of such a character is 

 scarcely to be expected in natural magmas. In spite of this 

 expectation it was considered desirable to conduct the experi- 

 mental attempts to obtain the sinking of crystals under condi 

 tions which favored the existence of convection currents. With 

 this end in view the temperature distribution within the fur- 

 nace was carefully explored with two thermoelements free 

 from contamination. They agreed in placing the maximum 

 temperature a little below the middle of the vertical cylindri- 

 cal tube of the furnace as might be expected. From this 

 point the temperature falls off both upward and downward at 

 first slowly and then more rapidly. By holding the crucible 

 above the position of maximum temperature, in the region 

 where the temperature increases downward, the bottom of the 

 crucible could be maintained at a slightly higher temperature 

 than the top, a condition favoring the establishment of convec- 

 tion currents. It was under such conditions that the results 

 already described were obtained. The crucible was held in a 

 position where the temperature gradient had been found to be 

 rather steep, about 4° per cm. The gradient cannot be sup- 

 posed as great in the contents of the crucible itself but it must 

 have been considerable. In spite of such currents as may have 

 been set up the definite results already described were obtained. 

 In a magma body a gradient of 1° per cm. is probably extreme 

 except in the region near the margin for a brief period prior 

 to crystallization in that region. A gradient of 1° in several 

 meters is probably more common in the main mass of a body 

 of magma. The tendency towards convection must be small, 

 and would probably not interfere with the sinking of crystals 

 any more effectively than it did in the crucible containing the 

 artificial melt. 



* Day, Sosman and Hostetter : Determination of Mineral and Rock Densi- 

 ties at High Temperatures, this Journal, (4), xxxvii, 27 and 35, 1914. 



