This paper represents the fruit of approximately 5 years of experimental studies at the Geophysical Laboratory. When Bowen and I were contemplating this study while I was still on the staff at the Naval Research Laboratory and he at the University of Chicago, the need for it was made apparent by some results on experimental studies on obsidian–H 2 O which I had carried out at the Geophysical Laboratory a few months before. The results were of great importance if equilibrium were attained, but of little significance if only metastability were involved (I had heated obsidian in the presence of water, and relatively large amounts, as much as 30 per cent, apparently went into the obsidian powder at approximately 500°C. and 30,000 psi.). As we had no knowledge of the effect of water on equilibrium relations in such compositions we decided to direct our efforts toward a study of the system NaAlSi 3 O 8 –KAlSi 3 O 8 –SiO 2 –H 2 O.). It was apparent that we would have to develop new experimental methods if we were to make progress in water-silicate mixtures. The hydrothermal quenching apparatus was therefore developed after many unsuccessful experiments with various types of pressure vessels then available. Once the hydrothermal quenching apparatus was perfected we were able to begin studies on the granite-water system. As the three bounding binary systems were unknown we had first to study these. The NaAlSi 3 O 8 –KAlSi 3 O 8 –H 2 O) system was the first, and results of this and various papers dealing with polymorphism in the feldspars, which came as by-products of these studies, have been published. . . .

This paper deals with the experimental determinations of phase-equilibrium relations in the system NaAlSi 3 O 8 (albite)–KAlSi 3 O 8 (orthoclase)–SiO 2 –H 2 O and with the application of these results to some petrologic problems. The laboratory experiments can be divided into two categories: (1) study of the liquidus, solidus, and subsolidus phase relations in synthetic mixtures, and (2) study of natural minerals and rocks. Phase relations in the binary systems SiO 2 –H 2 O, NaAlSi 3 O 8 –H 2 O, KAlSi 3 O 8 –H 2 O, the ternary systems NaAlSi 3 O 8 –KAlSi 3 O 8 –H 2 O, NaAlSi 3 O 8 –SiO 2 –H 2 O, KAlSi 2 O 8 –SiO 2 –H 2 O, and the quaternary system NaAlSi 3 O 8 –KAlSi 3 O 8 –SiO 2 –H 2 O were investigated with the aid of synthetic mixtures. Analyzed natural feldspars were used for subsolidus studies in the system NaAlSi 3 O 8 –KAlSi 3 O 8 , and the beginning of melting temperatures of natural granites were studied and compared with synthetic granites. Heating experiments with natural alkali feldspars demonstrated that sanidine cryptoperthites could be made homogeneous by heating at 700°C., whereas orthoclase cryptoperthites could not be homogenized by heating at any temperature below the solidus of the binary system (1060°C.) unless the heating was carried out for such a long time that the material inverted to sanidine. A solvus or miscibility gap was determined for each series, and both differed from the solvus for synthetic mixtures (high albite-high sanidine series). Only a portion of each of these solvuses is believed to represent stable equilibria. Phase-equilibrium relations in the system SiO 2 –H 2 O were considered in detail because the relations in this system are representative of those expected in all systems involving a rock-forming silicate and water. There are at least five invariant points in the binary system of which three involve liquid carrying 90 per cent silica. The effect of pressure on the high quartz–tridymite inversion was investigated; 1000 kg/cm 2 pressure raised the inversion temperature approximately 180°C. to 1050°C. Melting relations in the system NaAlSi 3 O 8 –H 2 O were determined by methods used throughout this investigation, and it was gratifying to find that the results were in good agreement with those obtained by Goranson (1938) using a different apparatus and method. The discovery of high-temperature albite was a by-product of this study. Experimental studies in the system NaAlSi 3 O 8 –KAlSi 3 O 8 –H 2 O demonstrated that the alkali feldspars form a complete series of solid solutions above 660°C., and below this temperature a solvus or miscibility gap is present. Homogeneous synthetic feldspars formed above 660°C. unmixed when held at lower temperatures, a confirmation of the accepted theory for the origin of most natural perthites. The composition of the two alkali feldspars in the synthetic mixtures can be determined with considerable accuracy by means of X-ray diffraction patterns. With certain limitations, the X ray can be used to determine the composition of the two phases of natural perthitic alkali feldspars. Unmixing of alkali feldspars in the presence of water vapor under pressure was so rapid that it is surprising that fine perthitic intergrowths and homogeneous feldspars are found in nature. They must have formed in a dry environment. Tridymite and albite are the stable crystalline phases at the liquidus below 300 kg/cm 2 pressure, whereas at higher pressures high quartz and albite are the stable phases. The change from tridymite to quartz is a consequence of the liquidus lowering by water dissolving in the melt together with the pressure raising of the quartz–tridymite inversion; at about 300 kg/cm 2 the pressure–temperature curve of the quartz–tridymite inversion intersects the liquidus. Phase studies in the quaternary system NaAlSi 3 O 8 –KAlSi 3 O 8 –SiO 2 –H 2 O provided quantitative data on the melting relations in these granitic compositions as well as information on fractional and equilibrium crystallization. At constant pressure, the system is characterized by a minimum melting temperature on the boundary between quartz and feldspar solid solutions. Liquids throughout the system move toward this minimum on crystallization, and if fractionation is pronounced most liquids will reach the minimum. A plot of the normative albite, orthoclase, and quartz in all analyzed granites and rhyolites from Washington’s Tables (1917) demonstrates that the minimum at low water-vapor pressure, corresponding to a water content of 1–2 per cent, falls at the composition of the average granite and rhyolite. It is suggested that this demonstrates that crystal–liquid equilibria control granite compositions; therefore granites not formed at magmatic temperatures will be rare and will not have compositions related to the minimum. The liquids can originate by fractional crystallization of more basic liquids ( i.e., basalts) or by fractional melting of appropriate sedimentary and metamorphic rocks. The beginning of melting of two granites—The Westerly, Rhode Island, and the Quincy, Massachusetts—one a normal calc-alkaline granite, and the other an alkaline granite, was determined at a series of water-vapor pressures; a PT curve for the beginning of melting of the two granites corresponds within the experimental error to the beginning of melting at the isobaric minimum in the quaternary system. Evidence is presented to show that a continuous gradation from magma to hydrothermal solution will obtain in hydrous granitic compositions if the alkali to alumina ratio is such that crystallization results in concentration of alkali silicates in the residual liquids. The vapor in equilibrium with hydrous granitic liquids can remove the silica, feldspars, and quartz from the liquid phase by vapor transport or by diffusion through the vapor, and in long runs these materials were transferred to the cooler part of the pressure vessel. CaO, MgO, FeO, and P 2 O 5 were concentrated by this process, and in one experiment with the Westerly granite the vapor removed essentially all the feldspar and quartz, leaving a residue of garnet, pyroxene, and apatite. This tendency for the oxides abundant in the basic and ultrabasic rocks to be relatively insoluble in the vapor suggests that such a mechanism may produce the basic zones commonly found at granite contacts. The amphiboles anthophyllite, grunerite, and riebeckite appeared to be unstable in the presence of water vapor under pressure, and the absence of amphiboles in the granite pegmatites and their almost universal presence in the perthite-quartz granites indicate that the pegmatites were produced in a water-rich environment while the perthite-quartz granites crystallized in a water-deficient environment. The rapakivi granite problem has been reviewed in the light of the experimental results, and it is pointed out that normal crystallization of magmas containing somewhat more potassium than the average granite can produce the rapakivi texture, providing water is concentrated during crystallization and the liquidus is depressed below the feldspar miscibility gap. The Tertiary granites of Skye are normal granites, chemically, and to some extent texturally; mineralogically, however, they are similar to rhyolitic rocks. The quartz and feldspars resemble in many respects the corresponding phenocrysts of extrusive rocks. It is suggested that these young granites represent quenched granites which, as a consequence, have some properties of both granites and rhyolites. Melting is expected in the earth’s crust at depths of 12–21 km in geosynclinal areas where the initial gradient is on the order of 30°C./km. Complete melting will take place if 9–10 per cent water is available. If the water content is 2 per cent, melting will still begin at the same depth; complete melting will not take place until some greater depth has been reached. This range of melting will produce a zone in the earth’s crust which may range in thickness from a few to 20 km with the amount of liquid increasing downward. It is proposed that this zone of melting, where temperatures are high enough to melt granite completely and more basic compositions at least partially, may offer a mechanism for producing large batholithic masses of granite. A classification of salic rocks based on the nature of the alkali feldspar is proposed. The classification has two major divisions: (1) subsolidus, and (2) hypersolvus, depending on the whereabouts of the soda feldspar. In the hypersolvus rocks all the soda feldspar is or was in solid solution in the potash feldspar whereas in the sub-solvus rocks the plagioclase is present as discrete grains. The two major divisions are further subdivided according to the nature of the alkali feldspar modification. The suggestion that most granites finished crystallization with a single alkali feldspar precipitating has been questioned by some petrologists because rhyolites commonly carry phenocrysts of plagioclase and sanidine feldspar. A study of the feldspars of extrusive rocks indicates that the plagioclase phenocrysts may react with the liquid during crystallization leaving a single alkali feldspar. If fractionation takes place, the tendency to complete crystallization with only a single feldspar crystallizing from the liquid is greatly enhanced. The proposition that two-feldspar granites may have gone through a one-feldspar stage has been examined in the light of experimental studies, and it was concluded that the required adjustments in mineral composition and texture are reasonable.