294 Shepherd, Ran~kin, Wright — Binary Systems of 



As has often been said, the geologist is dependent on the 

 petrographic study of rocks which have passed through many 

 changes. A study of relatively simple two- and three-com- 

 ponent systems shows the futility of attempting to deduce the 

 past history of any polycomponent system from its final condi- 

 tion. Petrographers have already observed the order of 

 crystallization for certain minerals, but the exact correlation of 

 this order with the composition of the rock remains to be 

 determined. The phase rule shows that the first mineral to 

 separate from a freezing solution depends primarily on the 

 initial composition and only indirectly on the melting point of 

 the mineral. Furthermore, it is always possible for a mineral 

 to be redissolved after it has begun to separate and sub- 

 sequently to reappear at a lower temperature. If the eutectic 

 relation was the only one encountered in the study of minerals, 

 the classification of rocks, as well as the interpretation of the 

 occurrence of the femic and salic rocks, would be relatively 

 simple. But studies of simple systems leave no room for doubt 

 that the phenomena in a great rock magma are exceedingly 

 complex, and we have no choice but to begin with the labora- 

 tory study (quantitative measurement) of relatively simple 

 mixtures of pure minerals. We are entering upon a new 

 science in which it is absolutely essential that we proceed in an 

 orderly way through simple relations and conditions to those 

 which are more intricate. Experiments on the rocks, if under- 

 taken first, would be as futile as to design a power plant with- 

 out a first acquaintance with the physical properties of 

 materials. 



Take another obvious instance. It is of course desirable to 

 know the molecular weight of the various minerals, whether 

 or not they dissociate in solution, and the effect of these 

 phenomena on their inter-reactions. At the present time such 

 questions are practically unanswerable. The molecular weight 

 of salts in aqueous solution is determined by the lowering of 

 the freezing point, but even at this low temperature and with 

 the most delicate apparatus, an error of five per cent is regarded 

 as doing very well. jSTot only must the temperatures be 

 measured to the nearest hundredth of a degree, but the 

 formulae have thus far been found to apply only to " infinitely " 

 dilute solutions, — never over one per cent. Silicate melts are 

 too viscous to be stirred, and freezing occurs in a region of 

 variable temperature distribution and always over an interval 

 of a whole degree or more. Furthermore, we have as yet 

 no means of knowing that the fundamental assumptions 

 underlying the Van't Hoff-Raoult relation hold true for 

 silicates. We ought, therefore, to hesitate before forcing the 



