4 JOURNAL OF GEOLOGY— SUPPLEMENT 



before even very small effects due to these causes can be obtained. 

 With crystallization, possibiHties of differentiation are introduced 

 which completely overshadow any of the possible effects of diffusion 

 in a liquid phase. 



THE SORET ACTION 



The kind of composition gradient most commonly considered 

 is that due to a temperature gradient, commonly known as the Soret 

 action. It is generally assumed that a concentration takes place 

 toward the colder parts of those substances whose crystallization is 

 imminent. This assumption is designed to explain the observed 

 fact that there is often a richness in the minerals of early crystalliza- 

 tion toward the boundaries of igneous-rock bodies. But the Soret 

 action takes place entirely in the liquid and is wholly independent 

 of the solubility of the various substances in the liquid. There is 

 therefore no reason to believe that the substances concentrated in 

 the cold part would be those that are about to separate out. If a 

 tube filled with a very dilute solution of common salt is warmed at 

 one end, there will result a concentration of salt toward the cold 

 end. If the whole tube is now gradually cooled, this same differ- 

 ence of temperature being maintained, a time will come when ice, 

 not salt, crystallizes at the cool end. The substance concentrated 

 at the cool end as a result of the Soret action was salt, the substance 

 first crystallized there was ice. The Soret action has produced an 

 effect in this case precisely the opposite of that which is commonly 

 supposed to result from it. It is clearly unsafe to assume that 

 the substances concentrated toward the cooling boundary would 

 be those of early crystallization. 



There is also a difficulty in obtaining a sufficient quantitative 

 effect from the Soret action. Harker has demonstrated this truth 

 by imagining certain temperature differences and calculating the 

 possible composition differences that could be produced,"^ assuming 

 that the concentration varies as the absolute temperature. This 

 calculation shows that for a temperature difference of ioo° at 1200° 

 the relative concentrations would be as 1.07:1.^ Even this esti- 



^ The Natural History of Igneous Rocks, p. 316. 



2 The Soret action has recently been reinvestigated for salt solutions and it has 

 been found, as might be expected, that the simple relation deduced from osmotic- 

 pressure laws does not hold (August Eilert, Zs.f. anorg. Chemie, LXXXVIII [1914], i). 



