Becker — Fractional Crystallization of Rocks. 259 



names the property of maximum fusibility in mixtures eutexia 

 and the bodies which exhibit this property he calls eutectic. 

 The phenomena are not always so simple as is supposed in the 

 illustration given above, especially when masses, as they 

 approach the temperature of solidification, divide into immis- 

 cible fractions. In such cases one has to do with two or more 

 eutectic mixtures. Supersaturation may also intervene to com- 

 plicate matters and change of pressure probably influences the 

 composition of the eutectics.* Thus it is at least conceivable 

 that very complicated cases should arise, while if the process 

 plays a part in lithogenesis the simplest case is probably the 

 most frequent. 



The fractional crystallization process depends essentially 

 upon convection currents. That it is not incompatible with 

 convection is clear, while convection is the mortal enemy of 

 any process of separation involving molecular flow. The only 

 function of diffusion in this case would be to preserve the 

 homogeneity of the residual fluid or mother liquor, so that the 

 eutectic state could not be attained by any sensible part of the 

 fluid until the whole mother liquor was reduced to this condition. 



The effect of the solidification of crusts on the walls of a 

 dike or laccolite is to liberate heat. This liberation does not 

 raise the temperature, for otherwise the crusts would remelt ; 

 but the liberated heat must be conducted through the walls 

 before the dike as a whole can congeal, and it therefore delays 

 the process of solidification, giving additional time for the evo- 

 lution of an eutectic magma. 



There appear to be some conditions under which eutectic 

 action could not be expected. Unless an intrusive rock pos- 

 sesses considerable mobility, chilling would proceed more rap- 

 idly than convection, and eutectic separation would be very 

 imperfect if not completely obscured. Viscosity of the mass 

 would also interfere seriously with the uniformity of composi- 

 tion of the mother liquor. If the mass cooled very slowly 

 indeed, this uniformity might be established even in a very 

 viscous mass ; but very slow cooling would also mean very 

 slight convection. In viscous lavas, therefore, fractional crys- 

 tallization is not very probable. There is seemingly no exact 

 way of defining the degree of viscosity compatible with frac- 

 tional crystallization, but enough mobility must certainly be 

 present to maintain uniformity in the melted mass when diffu- 

 sion and convection cooperate. I have shown that, in some 

 solutions at any rate (all for which I could find appropriate 

 experimental data) diffusivity is inversely as the square of vis- 

 cosity.f If any such law holds for magmas, a moderate amount 



* Ostwald, Allgem. Chemie, vol. i, 1891, p. 1027. 

 f This Journal, vol. iii, 1897, p. 284. 



