How far has geochemistry led geology?


Granite pmg ss 2006
Thin section of a typical granite: clear white and grey grains are quarts (silica); striped black and white is feldspar; coloured minerals are micas (credit: Wikipedia)

In the Solar System the Earth is unique in having a surface split into two distinct categories according to their relative elevation; one covered by water, the other not. More than 60% of its surface – the ocean basins – falls between 2 to 11 km below sea level with a mean around 4 to 5 km deep. A bit less than 40% – land and the continental shelves – stands higher than 1 km below sea level up to almost 9 km above, with a mean around 1 km high. Between 1 and 2 km below sea level is represented by only around 3 % of the surface area. This combined hypsography and wetness is reckoned to have had a massive bearing on the course of climate and biological evolution, as far as allowing our own emergence. The Earth’s bimodal elevation stems from the near-surface rock beneath each division having different densities: continental crust is less dense than its oceanic counterpart, and there is very little crustal rock with an intermediate density. Gravitational equilibrium ensures that continents rise higher than oceans. That continents were underpinned mainly by rocks of granitic composition and density, roughly speaking, was well known by geologists at the close of the 19th century. What lay beneath the oceans didn’t fully emerge until after the advent of plate tectonics and the notion of simple basaltic magmas pouring out as plates became detached.

In 1915 Canadian geologist Norman Levi Bowen resolved previously acquired knowledge of the field relations, mineralogy and, to a much lesser extent, the chemistry of igneous rocks, predominantly those on the continents in a theory to account for the origin of continents. This involved a process of distillation or fractionation in which the high-temperature crystallisation of mafic (magnesium- and iron-rich) minerals from basaltic magma left a residual melt with lower Mg and Fe, higher amounts of alkalis and alkaline earth elements and especially enriched in SiO2 (silica). A basalt with ~50% silica could give rise to rocks of roughly granitic composition (~60% SiO2) – the ‘light’ rocks that buoy-up the continental surface – through Bowen’s hypothetical fractional crystallisation. Later authors in the 1930s, including Bowen’s teacher Reginald Aldworth Daly, came up with the idea that granites may form by basalt magma digesting older SiO2-rich rocks or by partially melting older crustal rocks as suggested by British geologist Herbert Harold Read. But, of course, this merely shifted the formation of silica-rich crust further back in time

A great deal of field, microscope and, more recently, geochemical lab time has been spent since on to-ing and fro-ing between these hypotheses, as well as on the petrology of basaltic magmas since the arrival of plate theory and the discovery of the predominance of basalt beneath ocean floors. By the 1990s one of the main flaws seen in Bowen’s hypothesis was removed, seemingly at a stroke. Surely, if a basalt magma split into a dense Fe- Mg-rich cumulate in the lower crust and a less dense, SiO2-rich residual magma in the upper continental crust the bulk density of that crust ought to remain the same as the original basalt. But if the dense part somehow fell back into the mantle what remained would be more able to float proud. Although a neat idea, outside of proxy indications that such delamination had taken place, it could not be proved.

Since the 1960s geochemical analysis has became steadily easier, quicker and cheaper, using predominantly X-ray fluorescence and mass-spectrometric techniques. So geochemical data steadily caught up with traditional analysis of thin sections of rock using petrological microscopes. Beginning in the late 1960s igneous geochemistry became almost a cottage industry and millions of rocks have been analysed. Recently, about 850 thousand multi-element analyses of igneous rocks have been archived with US NSF funding in the EarthChem library. A group from the US universities of Princeton, California – Los Angeles and Wisconsin – Madison extracted 123 thousand plutonic and 172 thousand volcanic igneous rocks of continental affinities from EarthChem to ‘sledgehammer’ the issue of continent formation into a unified theory (Keller, C.B. et al. 2015. Volcanic-plutonic parity and the differentiation of the continental crust. Nature, v. 523, p. 301-307).

In a nutshell, the authors compared the two divisions in this vast data bank; the superficial volcanic with the deep-crustal plutonic kinds of continental igneous rock. The gist of their approach is a means of comparative igneous geochemistry with an even longer pedigree, which was devised in 1909 by British geologist Alfred Harker. The Harker Diagram plots all other elements against the proportionally most variable major component of igneous rocks, SiO2. If the dominant process involved mixing of basalt magma with or partial melting of older silica-rich rocks such simple plots should approximate straight lines. It turns out – and this is not news to most igneous geochemists with far smaller data sets – that the plots deviate considerably from straight lines. So it seems that old Bowen was right all along, the differing deviations from linearity stemming from subtleties in the process of initial melting of mantle to form basalt and then its fractionation at crustal depths. Keller and colleagues found an unexpected similarity between the plutonic rocks of subduction-related volcanic arcs and those in zones of continental rifting. Both record the influence of water in the process, which lowers the crystallisation temperature of granitic magma so that it freezes before the bulk can migrate to the surface and extrude as lava. Previously. rift-related magmas had been thought to be drier than those formed in arcs so that silica-rich magma should tend to be extruded.

But there is a snag, the EarthChem archive hosts only data from igneous rocks formed in the Phanerozoic, most being less than 100 Ma old. It has long been known that continental crust had formed as far back as 4 billion years ago, and many geologists believe that most of the continental crust was in place by the end of the Precambrian about half a billion years ago. Some even reckon that igneous process may have been fundamentally different before 3 billion years ago(see: Dhuime, B., Wuestefeld, A. & Hawkesworth, C. J. 2015. Emergence of modern continental crust about 3 billion years ago.  Nature Geoscience, v. 8, p.552–555). So big-science data mining may flatter to deceive and leave some novel questions unanswered .


4 thoughts on “How far has geochemistry led geology?

  1. Senior author Brenhin Keller sent this comment by e-mail. (Better to use the “Leave a comment” option in the item concerned, as this gets direct to me)
    Steve Drury

    A coauthor recently pointed me in the direction of this Earth-Pages article ( on our paper in Nature last week. I liked the article (especially the historical context!) but just wanted to make one small clarification: there is actually quite a bit of Pre-Phanerozoic data in EarthChem, just not in the tectonic environments (active arcs and rifts) that we filtered for in Figures 1-3 (and in extended data Figures 3 and 5-7). The trends in Figures 5 and 6, however, are based on samples from all tectonic settings, including pre-Phanerozoic shields and cratons. The samples used to produce the trends in Figure 6 are sampled evenly from all exposed igneous crust, yet the curvature of differentiation trends (and subsequent conclusions about fractional crystallization) persists.

    I think we may not have made this distinction very clear – since we were more concerned with ensuring that our samples filtered as “arc” and “rift” were in fact from arc and rift tectonic settings and accurately represented melt production and differentiation in these environments. In any case, we are well aware that the processes of continental crust formation may have been different in the past (we have an older paper on that here and two more in the works, also using EarthChem data).

    There wasn’t any contact information for the author Steve Drury on the Earth-Pages website, but the contact page led us to you – perhaps you can pass this on?

    Best regards,

    Thanks Brenhin,


    1. Hi Brenhin

      I appreciate that Earth Chem contains geochem data from many Precambrian sources, but meta-igneous rocks (eg gneisses, amphibolites etc) are tagged as metamorphic rather than volcanic or plutonic. My guess is that a large proportion of rocks from the known crystalline crust of whatever age do not fall under the volcanic-plutonic hat in EarthChem and are certainly not tagged according to tectonic setting of formation. Mind you, I was only a geochemist in the days when one plotted data by hand (1967 to about 1985). (Joking!) I moved into other fields thereafter. I wrote the piece on finding how much influence old Harker, Bowen and Daly still have from your nice Nature paper.


      Steve Drury


  2. Quite right – only relatively unmetamorphosed rocks are included here. We’re working on it though!


  3. We are often too self-esteem and dont realize that we have very incomplete data because we barely scratched the surface of the crust. We have no direct data from deep crust parts and we have very imprecise and indirect estimation about the geochemistry of the Earth 2-3 billions y. ago. The surface and ocean geochemistry was very different for sure – just the switch from anoxic to highly oxidizing environment was probably very dramatic! We often tend to make huge theories based on vast maps, cratons, and thousands of bulk rock analyses. And we often forget that too many of these are poorly sampled, poorly analysed and that the sampling was often highly selective. Also trace elements and accessory minerals (often the most important to understanding the geochemistry!) are way too often ignored. Despite these facts, building big rock and geochemical databases has potential to bring more understanding of rock forming processes. Petrology will probably more and more tend towards regional petrology – connected with (plate)tectonics and big geological events; the second branch will go more towards chemistry and physics and study the rock forming processes – as the mineralogy already did. Both ways require lot of gechemical data.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s