LATER STAGES OF EVOLUTION OF IGNEOUS ROCKS 83 



more acid portion would commonly precede the basic portion. 

 But, on account of the fact that crystals accumulate as such in the 

 lower portions, these portions have Httle eruptibility and such 

 intrusions as may occur come from the still liquid upper portion 

 almost exclusively. Successive intrusions therefore commonly 

 exhibit the decreasing basicity which this Kquid portion experiences. 

 Some cases of reversal of the normal order may, perhaps, be reason- 

 ably attributed to the intrusion of Kquid from the middle portion 

 of the reservoir bearing a relatively small proportion of crystals 

 and following closely after the Kquid from the upper portion. 



It is not to be assumed from the foregoing discussion that the 

 Kquid portion itself from a lower layer would not have a different 

 composition from that in an upper layer. The coUection of the 

 early crystals in the lower layers causes the liquid in those parts to 

 foUow a different course from that followed in an upper layer, 

 even although the liquids are freely miscible to a homogeneous 

 liquid. It is the attempt of the liquid in different parts to maintain 

 equiKbrium with its immediate surroundings that determines the 

 different courses followed. The bearing of the existence of liquids 

 of somewhat different composition in contiguous layers on the 

 formation of rocks exhibiting primary banding has already been 

 discussed. 



CHILLED BORDER PHASES 



ChiUed border phases have already been referred to several 

 times, especially the formation of a basic border by restriction of 

 differentiation due to sudden chilling. Thus the formation of a 

 diorite border phase about a granite is explained as due to com- 

 pletion of crystallization near the border at stage 4 (see p. 75) and 

 continuance of crystallization and removal of crystals in the more 

 slowly cooled part removed from the border until stages 5 and 6 

 have been attained. The border phases produced by such chilKng 

 are, however, not necessarily basic. Thus in some cases the chilKng 

 might cause completion of crystallization toward the border at 

 stages 6-7 with the formation of a quartzose type, a granite, whereas 

 the more slowly cooled part might, by removal of crystals, con- 

 tinue its crystallization through stages 8 and 9 with the formation of 

 a nephelite-bearing type. It seems possible that the quartzose 



