THE PROBLEM OF THE ANORTHOSITES 93 



sisting of two different rocks. In the same way differentiation 

 may be induced by the floating of the lighter crystals, as well during 

 the crystallization of a liquid magma as in the process of anatexis. 

 An old experiment of Sorby^ is very instructive. When he fused 

 a hornblende granite from Mount Sorrel he observed that the 

 hornblende melted first and some of the feldspar and quartz floated. 



The formation in this way of a small anorthositic or pyroxenitic 

 facies in a gabbro body is easily conceivable. On the other hand, 

 the formation of a large anorthosite body requires rather that the 

 sinking of pyroxene crystals should be accompanied by their 

 resolution in the deeper parts, until the magma is finally almost 

 completely separated into anorthosite and pyroxenite with gabbro 

 between them. 



2. According to Bowen the magma does not contain definite 

 minerals as these are dissociated. The first crystalHzed minerals, 

 being only temporary products of this dissociation, are again dis- 

 solved and disappear if the period of crystallization is sufficiently 

 quiet and long. So, for instance, in a magma corresponding to a 

 certain mixture of diopside and labradorite, ohvine is formed at the 

 beginning of crystallization, and in this way a certain quantity of 

 free silica is produced. In the later stages of crystallization the 

 oHvine is again dissolved and, combining with the silica, crystallizes 

 at last as diopside. But if the olivine, formed at the beginning, is 

 eliminated by sinking, the magma might fully differentiate into an 

 olivine rock, gabbro, syenite, or even granite. This conception is 

 based on Bowen's experiments: chilling an enstatite melt demon- 

 strated that at the beginning of crystallization oHvine was formed; 

 when crystallization was slowly brought to an end the oHvine dis- 

 appeared and the whole melt crystallized as clinoenstatite. It is 

 quite certain (see Bowen's photographs) that olivine is really formed 

 in these conditions from an enstatite melt and that it tends to con- 

 centrate in the lower parts of the crucible. But, in conceding this, 

 we are nevertheless very far from concluding that this process is 

 realized on a large scale in the earth's crust, and that a basaltic 

 (gabbroid) magma will split into dunite, pyroxenite, gabbro, syenite, 

 and granite. In order to support such a conclusion, we have to 



^ H. Sorby, Proc. Geol. a. Polytechn. Soc. W. Yorkshire, Vol. IV (1863), p. 302. 



