LATER STAGES OF EVOLUTION OF IGNEOUS ROCKS 3 



in the light of the results hitherto obtained in the experimental 

 study of siHcate mixtures, has led the writer to believe that the 

 former process, assimilation, is comparatively unimportant quanti- 

 tatively, and that to understand the igneous rocks we must direct 

 our energies toward the elucidation of processes of differentiation. 

 Moreover, of the various processes of differentiation that have been 

 proposed, it is considered that differentiation, or fractionation, by 

 crystallization is vastly the most important. In the present paper 

 the facts which have led to these conclusions are stated and a 

 systematic petrogenic theory based on these conclusions is proposed. 



DIFFERENTIATION 



The term differentiation is applied to any process whereby a 

 magma, without foreign contamination, forms either a mass of rock 

 that has different compositions in different parts or separate masses 

 that differ from one another in composition. The subject has been 

 much discussed and apparently every conceivable process given a 

 place as a factor in producing the observed results. In the following 

 pages, discussion of these processes will be entered into only in so 

 far as is necessary in order to indicate the reasons for believing 

 that but one of these is of great importance, namely, crystalHzation 

 and the movement of crystals relative to the remaining liquid. 



Some investigators have considered that difference of composi- 

 tion in the magma itself , due to a temperature or a pressure gradient 

 or to gravity, is the important factor. These continuous variations 

 in the composition of the liquid, supported by Vogt, Iddings, and 

 others, which are due to a Diffusion oder Wanderung^ in a single 

 liquid phase,^ are to be carefully distinguished from a breaking up 

 of the magma into immiscible partial magmas, i.e., a separation into 

 two or more liquid phases, a process supported by Rosenbusch, 

 Backstrom, and others. 



Theoretically, continuous composition gradients should be estab- 

 lished if sufficient time were available. However, in an actual 

 case, there is reason to believe that crystallization begins long 



^ J. H. L. Vogt, "tjber anchi-monomineralisclie und anchi-eutektische Eruptiv- 

 gesteine," Vidensk. Selsk. Skr., I Math.-Naturv. Kl., 1908, No. 10, p. 6. 



2 When a phase shows a composition or concentration gradient it can be treated in 

 phase-rule considerations as divided into "regions"; see R. C. Tohnan, Jotir. Am. 

 Chem. Soc, XXXV (1913), 307- 



