"Inside the globe [there] exist mysterious forces, whose effects become apparent on the surface. Eruptions of vapors, glowing lava and new volcanic rocks…"
Alexander von Humboldt
At the end of the 19th century and after the victory of "Plutonism" in the great Granite War, geologists accepted the idea that igneous rocks originate from deep inside earth. However the great variability of volcanic and plutonic rocks, from dark basalt to light-colored granite, was difficult to explain, as Earth´s interior was assumed to be relatively uniform (based on the idea that earth formed by condensation of primordial matter, it was imagined like a succession of concentric shells).
Some geologists, the "Granitizers" after a term coined by French geologist P.T. Virlet d´Aoust, proposed that igneous rocks formed by melting of pre-existing rocks. This model could explain the distribution of certain magmatic rocks (granite was found only on continents, as it forms by melting other crustal rocks) however it couldn´t explain the sharp limits between volcanic conduits often seen in the field. Also the energy required to melt the "source-rocks" was very high and the necessary chemical diffusion of elements between old and new rocks very low.
According to the "Magmatists", from the term "magma, coined by French naturalist de Dolomieu in 1794, granitic and other plutonic rocks formed indeed by cooling of different melts. German chemist Robert Wilhelm E. Bunsen (1811-1899) proposed in 1851 a simplified model based on this premise and his geologic observations in Iceland ("Über die Processe der vulkanischen Gesteinsbildung Islands"). He imagined that there are only two magma-types on Earth - the basic "pyroxenite" and the acidic "trachyte", formed in the different shells of earth´s crust. By mixing these two main magma-types all other volcanic rocks could be generated.
A discovery in andesites and dacites of the San Juan Mountains (Colorado) seemed to confirm this model. Here single large feldspar-crystals differ significantly in their chemical composition from the surrounding matrix with small feldspar-crystals. For Bunsen the larger feldspar-crystals formed independently and were later mixed into the melted fine-grained rock.
In 1844 geologist Charles Darwin proposed a petrological model needing only one type of magma:
"Much of the difficulty which geologists have experienced, when they have compared the composition of volcanic with plutonic formations, will, I think, be removed, if we may believe, that most plutonic masses have been, to a certain extent, drained of those comparatively weighty and easily liquefied elements, which compose the trappean and basaltic series of rocks."
Darwin, based only on observations in the field, imagined a mineralogical differentiation of magma by segregation of minerals. Grains of minerals form in the cooling magma, the heavy minerals sink to the ground of the magma chamber, the residual melt is now impoverished in heavy elements and can form another types of igneous rocks.
Fig.2. The "evolutionary tree" of igneous rocks - with the most basic (found in earth´s mantle - like peridotite) on the base and the more derived (like granite) on top. From KRÜGER, K. (1954): "Das Reich der Gesteine", image in public domain.
But only with the experiments by Canadian Norman Levi Bowen (1887-1956) geologists started to realize the mechanisms that generate igneous minerals and rocks.
N.L. Bowen was born in Kingston (Ontario), but after university left for the Massachusetts Institute of Technology, where he meet petrologist Arthur L. Day (1871-1957), director of the Geophysical Laboratory of the Carnegie Institution, newly established at Washington in 1905. Day was especially interested to understand how melts form from a mixture of different minerals. He proposed to Bowen to study the properties of the plagioclase-feldspar, a common and important component in igneous rocks, composed of a mixture of the two minerals Anorthite and Albite. Thanks to new electric ovens, high-temperature resistant probes and new analysis-methods Bowen finally published in 1913 one of the most famous diagrams in petrology - the phase diagram of plagioclase:
Fig.3. Phase diagram of plagioclase feldspar from Bowen´s "The Melting Phenomena of the Plagioclase Feldspars" (1913) This diagram shows how crystals form from a cooling melt, note how the composition from the crystals (lower line) differs from the composition of the residual melt (upper line). The crystals continue to react with the cooling melt, following the respective line. However if the crystals are removed from the system the starting point of the system will change. Cooling this melt will now produce another variety of feldspar with the chemical composition of the residual melt. This mechanism is known (and feared) by geology students as "fractional crystallization" (image used under fair use for educational purpose).
Bowen was convinced that alone fractional crystallization could explain the observed variety of igneous rocks. However there was one great weakness with this model - fractional crystallization is very inefficient; you need a volume of source-magma nine times greater than the observed granitic rocks, also fractional crystallization should occur all over the planet, but granitic rocks are concentrated in continental crust.
Something was missing…
HÖLDER, H. (1989): Kurze Geschichte der Geologie und Paläontologie - Ein Lesebuch. Springer Verlag, Heidelberg: 243
YOUNG, D.A. (2002): Norman Levi Bowen (1187-1956) and igneous rock diversity. From OLDROYD, D.R. (ed.); The Earth Inside and Out: Some Major Contributions to Geology in the Twentieth Century. Geological Society, London, Special Publications Nr. 192: 99-111