CRYSTALLIZATION-DIFFERENTIATION IN MAGMAS 413 



crystal settling? The answer suggested by experience is that it 

 would act as a deterrent. Sedimentation in nature is always 

 delayed by convection, and in the laboratory when one wishes to 

 make separations by means of heavy solutions one always obtains a 

 more ready separation by jacketing the heavy solution in order to 

 prevent even the very moderate convection that results from the 

 slight draughts in the laboratory. 



Perhaps, however, the more important reason for advancing 

 convection was not its greater efficiency in bringing about accumu- 

 lation of crystals as compared with the "far-fetched" process of 

 settling of crystals. Possibly the particular reason was the b5,nding 

 of the gabbro. In order to explain the banding it is necessary, 

 however, to assume a simplicity of convection that experience of 

 that phenomenon does not wari-ant. It is known, for example, 

 that a mass of liquid experiencing convection may divide up 

 into columnar cells of hexagonal section, each of which has its 

 own system of convection characterized principally by vertical 

 currents. While it need not be urged that convection in a large 

 mass of magma would be exactly of this nature, it seems inevit- 

 able that a very large number of currents would be set up 

 whose motion would be principally vertical, and that any banding 

 of the rock that might result from convection would be principally 

 vertical. There seems little reason to believe in a great current 

 sweeping over the floor and depositing its load in bands parallel to 

 the floor in the manner Pirsson^ and Grout have suggested. How- 

 ever, assuming that such a current did exist, it is noteworthy that in 

 order to explain banding it is necessary to make the further assump- 

 tion of a rhythm in crystallization, such, that the crystals brought 

 down by convection vary in their nature as time goes on and may 

 even alternate. And convection is supposed to have a remarkable 

 advantage over crystal settling in producing alternation of layers- 

 as a result of this rhythmic crystallization. But while I do not 

 urge rhythmic crystallization I would point out that the crystals 

 brought to the bottom by settling would show precisely the same 

 alternation as those brought there by convection if this rhythmic 

 crystallization can be assumed. 



^ The Igneous Rocks of Highwood Mountains, U.S. Geol. Survey, Bulletin 237. 



