LATER STAGES OF EVOLUTION OF IGNEOUS ROCKS 7 



gas bubbles at a later stage. The persistence of this layering, 

 or, better, continuous composition gradient, is due only to the 

 inability of diffusion to overcome it in such viscous melts in any 

 reasonable time. The phenomenon was noted long ago by Sorby. 

 In describing the melting of Mount Sorrell "syenite," a hornblende 

 granite, he says, "The hornblende melted more easily than the 

 quartz and feldspar, and a portion of those minerals has risen 

 upward."^ 



The phenomena which take place in the glass-maker's pot are 

 essentially of the same general nature; that is, the difficulty con- 

 nected with obtaining a homogeneous melt in glass practice is due 

 to initial inhomogeneity and not to any tendency of the liquid, once 

 homogeneous, to become inhomogeneous. The truth of this state- 

 ment becomes obvious when it is realized that the glass-maker, in 

 order to reduce inhomogeneity, resorts to long-continued heating, 

 thereby affording diffusion sufficient time to accompHsh the desired 

 result. If there were any tendency to become inhomogeneous, con- 

 tinued heating would give precisely the opposite effect. 



SEPARATION OF DISTINCT PHASES 



The possibiHties of composition differences that have been con- 

 sidered are those which may conceivably exist in a magma still 

 entirely fluid and consisting of but one phase. They are not 

 adequate to explain the differences actually observed in nature in 

 the crystallized product of the magma. As soon, however, as new 

 phases begin to separate from the magma, the possibiHty of obtain- 

 ing composition differences in different parts is enormously in- 

 creased, if there is any mechanism whereby a relative concentration 

 of the new phase or phases in certain parts may be accomplished. 

 The new phases might be either liquid or crystalline and the factors 

 which might result in their concentration in certain parts are diffu- 

 sion, convection, gravity, and earth movements. Consideration 

 will first be given to the question whether the separation of Hquid 

 phases, i.e., immiscibility in the Kquid state, is a factor in differ- 

 entiation. 



^ Proc. Geol. and Polytech. Soc. W. Yorkshire, IV (1863), 302. 



