CRYSTALLIZATION OF IGNEOUS ROCKS. 
99 
would not obtain if the amount of the silica and alkalies 
passed a certain limit. In the latter case it should happen 
that the alkali-aluminous compounds and free silica crys¬ 
tallize before the ferro-magnesian compounds. 
It is to be remarked that in a cooling solution of various 
compounds, when a point is reached where certain of the 
compounds are in excess of what the solution is then capable 
of holding dissolved, this excess is crystallized out and the 
relative proportions of the various compounds is changed. 
In a subsequent phase of the cooling another compound may 
reach the point of excess and it may start to crystallize. In 
this manner a number of crystallizations may take place 
beside one another, though they started to crystallize at suc¬ 
cessive periods of the cooling. 
In the case of magmas highly charged with silica and the 
alkalies it is probable that the early separations would con¬ 
sist of these compounds, and that the ferro-magnesian com¬ 
pounds would crystallize later and would finish their sepa¬ 
ration before the alkali compounds and the free silica did. 
(2.) The exact chemical nature of compounds when in so¬ 
lution is as yet little known; their mutual relations when 
in a mixed solution is still less understood. The influence 
of the temperature and pressure upon their crystallization 
from mixed solutions has not been determined for a sufficient 
number of substances to lead to general conclusions; but 
enough has been observed to indicate that the resulting crys¬ 
tallizations vary within certain limits according to the 
physical conditions under which they separate from mixed 
solutions. 
Hence it should happen that chemically similar magmas 
that cool under similar physical conditions should crystallize 
alike, and that differences in the nature of the crystallization 
of such magmas should indicate differences in the physical 
conditions attending their cooling. Further, that chemically 
different magmas which cool under similar physical condi¬ 
tions should differ in the nature of their crystallization ac¬ 
cording to their chemical variations; and, finally, that if one 
