ORTGTN TN IGNEOUS HOOKS. 
91 
but little tendency for the alumina to unite with magnesia to foiiu 
double silicates when the magma is a magnesium silicate. 
In the following table these same phenomena have been sunnnai'- 
ized in a way that j)erhaps makes the various results stand out more 
prondnently. 
depuration of alumina from molten magmas.*^ 
Magma silicates. 
Dissolved oxides. 
Minerals formed. 
rAl,0,:R=l:l 
Corundum. 
+ excess SiO.^ 
Sillimanite 
Ca-Na-K (silicates) 
R=other bases. 
AlA-MgO-FeO 
4- excess Al^Oj 
Spinel. 
Corundum. 
+ excess SiOj 
/AI2O, 
Cordierite. 
>Corundum. 
Spinel. 
Corundum. 
Do. 
No excess MgO 
AlA 
-f- excess MgO 
-f- excess AI2O3 
AloOq-CroO, 
Do, 
(Mg,Fe)Si04 
No excess MgO 
Chromite. 
AiPa-Cr.Og 
J Corundum, 
tpicotite. 
rMitchellite. 
IChromite. 
Plagioclase feldspar. 
Corundum. 
+ excess MgO 
CaO-Na^O-Al^Og-SiO, 
.+ excess AI2O3 
It is very evident that there are a series of magmas that contain 
an excess of alumina, just as there are magmas containing an excesi- 
3f silica, and that the alumina separates out as corundum according 
o the same laws that govern the separation of silica as quartz in 
granitic rocks. 
In the case of the amphibolites and pyroxenites the general laws 
yiven above wall apply to the separation of the alumina as corundum 
Prom these rocks when they were in a molten condition. As they are 
Por the most part in narrow dike-like formations instead of lenticular 
nasses, the corundum would be generally distributed through them 
^ather than concentrated along their borders. 
"For comparison see on p. 94 the origin of corundum in plumasite. 
