The concept of silicate liquid immiscibility was invoked early in the history of petrology to explain certain pairs of compositionally divergent rocks, but. as a result of papers by Greig (Am. J. Sci.13, 1–44, 133–154) and Bowen (The Evolution of the Igneous Rocks), it fell into disfavor for many years. The discovery of immiscibility in geologically reasonable temperature ranges and compositions in experimental work on the system K₂O-FeO-Al₂O₃-SiO₂, and of evidence for immiscibility in a variety of lunar and terrestrial rocks, has reinstated the process.

Phase equilibria in the high-silica corner of the tetrahedron representing the system K₂O- FeO-Al₂O₃-SiO₂ are presented, in the form of constant FeO sections through the tetrahedron, at 10% increments. Those sections, showing the tentative relationships of the primary phase volumes, are based on 5631 quenching runs on 519 compositions, made in metallic iron containers in pure nitrogen. Thirteen crystalline compounds are involved, of which at least six show two or more crystal modifica-tions. Two separate phase volumes, in each of which two immiscible liquids, one iron-rich and the other iron-poor, are present at the liquidus. One of these volumes is entirely within the quaternary system, astride the 1:1 K₂O:Al₂O₃ plane. No quaternary compounds as such have been found, but evidence does point toward at least partial quaternary solid solution, with rapidly lowering liquidus temperatures, from K₂O·Al₂O₃· 2SiO₂ (‘potash nepheline’, kalsilite. kaliophilite) to the isostructural compound K₂O·FeO·3SiO₂, and from K₂O·Al₂O₃·4SiO₂ (leucite) to the isostructural compound K₂O·FeO·5SiO₂, Both of these series apparently involve substitution, in tetrahedral coordination. of a ferrous iron and a silicon ion for two aluminum ions. Some of the ‘impurities’ found in analyses of the natural phases may reflect these substitutions.

As a result of the geometry of the immiscibility volume located entirely within the quaternary system, compositions near it show a number of phase changes and large amounts of crystallization with small temperature changes, generally in the range 1100–1150 C. Similar low-temperature, high-alkali immiscibility was discovered in a few exploratory runs in the equivalent systems with Rb or Cs substituting for K. But not in those with Li or Na.

A review of the compositions and general behavior of systems involving immiscibility, both stable and metastable, and of the evidence for natural immiscibility. indicates that it may be a much more common feature than generally thought. Several examples of natural immiscibility are detailed; most yield a felsic. alkali-aluminosilicate melt and a mafic melt. from a wide variety of generally basaltic parental magmas, both under- and over saturated. Unfortunately, the best line of evidence for immiscibility in terrestrial rocks, a sharply defined meniscus between two compositionally disparate glasses, is by its very nature self-destructing, since it is effectively eliminated by either crystallization or gravitative separation and coalescence into separate magmas. Verification of operation of the exosolutionor ‘splitting’ process on a large scale will probably require careful study of isotopic and trace element partitioning in both laboratory and field.