Komatiites

Introduction

Large igneous provinces (LIPs) and ocean islands basalts (OIB) are commonly hypothesized to be the products of thermal plumes originating at the core-mantle boundary. If so, then the size, strength and number of such mantle upwellings should primarily be a function of the temperature of the core (Davies, 1993). Thus one way to test the plume hypothesis is to look at the size and frequency of such proposed plumes throughout the history of the Earth. Numerical simulations of mantle convection and the cooling of the Earth typically show a rapid cooling in the first 1 billion years, followed by a slower, nearly linear cooling (Davies, 1993). Correspondingly, one would expect a steep decline in plume activity in the first 1 b.y. of Earth history. Unfortunately, the Archean and Hadean rock record is very scant. Furthermore, most modern proposed plume magmas erupt on oceanic crust, which only very rarely escapes destruction at convergent margins. It seems likely that such size-frequency information on mantle plumes simply has not been preserved in the rock record.

The strength (buoyancy flux) of the hypothetical plumes is another matter. Buoyancy flux is primarily a function of the temperature difference between the upwelling mantle and the ambient mantle. If one constrains the ambient mantle to be below (or at least near) its solidus (i.e. no magma ocean after ~ 4.4 Ga) then the primary influence on the strength of thermal plumes will be the temperature of the plume when it leaves the CMB. Studies (mostly of mid-ocean ridge volcanism) have shown that the temperature of a rising piece of mantle that undergoes anhydrous, adiabatic decompression melting can be correlated with the major element composition of the magmas it produces (Klein & Langmuir, 1987). Hotter mantle intersects its solidus at greater depths and melts over a larger pressure interval and to greater extents than cooler mantle. The melts produced will have higher MgO and FeO and lower Na₂O, Al₂O₃ and SiO₂ than melts of cooler mantle. So an alternative approach to quantifying the vigor of plume activity through time is to look at the secular evolution of the major element composition of magmas thought to be produced by plumes.

Like the cooling of the Earth, one would expect the most rapid changes and dramatic differences in plume magma chemistry to occur in the first billion years of Earth history. Indeed, studies of Archean volcanism (Arndt, 1976; Pyke et al., 1973; Viljoen & Viljoen, 1969a,b) have found high-MgO magmas that seem to fit the criteria for being the products of super-hot plumes (Herzberg, 1995). These magmas are called komatiites. Recently though, an old debate (Allegre, 1982) has been rekindled that argues these magmas are not produced by plumes, but rather by hydrous melting processes in the upper mantle (Parman & Grove, 2001).

Komatiites

Komatiites were first recognized in the late 1960s in the Barberton Mountainland greenstone belt in South Africa (Viljoen & Viljoen, 1969a,b). They have extremely high MgO contents; 18-30 wt.% compared to 10-15 wt.% for the most mafic mid-ocean ridge basalts (MORB) or ocean-island basalts (OIB, Figure 1). The MgO contents of magmas is proportional to their melting temperatures (higher MgO means hotter magmas) and the first experiments on komatiites (Green, 1975) were interpreted to imply melting conditions in excess of 1600°C (compared to 1250-1350°C for modern MORB, Figure 2; see also Temperature and Mantle temperature pages). Subsequent dating showed the Barberton komatiites to be 3.5 billion years old (Lopez-Martinez et al., 1992) and so the high temperatures inferred for the komatiite source region fit nicely with the concept of a hot early Earth.

Komatiites from the Superior province in Canada (the Munro komatiites) were the next to be well studied (Arndt, 1976; Pyke et al., 1973). These are younger (2.7 Ga) than the Barberton komatiites and have lower MgO contents (up to ~ 24 wt.%). This also fit well with the idea of a cooling Earth. At the time, there was still some debate about the tectonic setting of komatiites. A whole range of settings was considered, e.g., mid-ocean ridge, plume, giant impact and magma oceans.

Figure 1: SiO₂ and TiO₂ versus MgO in komatiites (fields with solid boundaries), basaltic komatiites (filled squares) and modern mafic magmas (fields with dashed boundaries). Solid line in top panel shows compositional effect that olivine fractionation would have on the most MgO-rich Barberton komatiite. The majority of the basaltic Barberton komatiites cannot be made by olivine fractionation from a komatiite parent. Boninites (modern subduction related magmas) show a large compositional overlap with Archean basaltic komatiites. Komatiites show a wide range of major and minor element composition. High-SiO₂ komatiites (dark fields) resemble modern boninite magmas that are produced by hydrous melting, while low-SiO₂ komatiites (light fields) resemble more closely modern basalts produced by anhydrous decompression melting. The low TiO₂ contents of komatiite magmas requires their source to have been depleted (i.e. to already have had a melt extracted). Komatiites: Barb. (Barberton, South Africa, 3.5 Ga,), C (Commondale, South Africa, 3.3 Ga), T (Tisdale, Canada, 2.7 Ga), B (Ball, Canada, 2.9 Ga), Munro (Munro, Canada, 2.7 Ga) and Gorg. (Gorgona, South America, 0.088 Ga). Modern magmas: OIB (ocean island basalt, GeoRoc online database), boninites (GeoRoc online database).

The Plume Model

It is not clear at what point plumes became the paradigm for the generation of komatiites. One important step was the discovery of Phanerozoic komatiites on the island of Gorgona (Aitken & Echeverria, 1984; Echeverria, 1980). These komatiites have even lower MgO and melting temperatures than the Munro komatiites, again fitting nicely into the cooling Earth model (Figure 2). Initially, the tectonic setting of Gorgona was unclear. The main issue was whether it is related to the orogenies on the mainland of the South American continent or to the Galapagos hotspot. Though I have not seen the resolution of this debate, more recent papers on Gorgona have assumed that it is related to a plume beneath the Galapagos, thus cementing the idea that komatiites are produced by plumes (Arndt et al., 1998; Echeverria & Aitken, 1986). Another benchmark in the study of komatiites was the publication of the book “Komatiites” (Arndt & Nisbet, 1982), which summarized much of what was then known about komatiites. The general consensus of the book is that komatiites are produced by plumes.

The basic concepts of the plume model are summarized in Figure 2 and follow fairly closely the models of adiabatic, decompression melting developed to explain melting at mid-ocean ridges (Klein & Langmuir, 1987). Barberton komatiites (3.5 Ga) had the hottest mantle sources. As the Barberton plume ascended, the mantle intersected the solidus at depths greater than 350 km (Walter, 1998). Melting occurs between the point of intersection with the solidus and the bottom of the lithosphere, perhaps 50-100 km thick in the Archean. Assuming ~1 percent melting per 3 km of ascent when cpx is present and half that after it has been all melted out, 60-70% melting is implied. By 2.7 Ga, the mantle has cooled considerably, and now the Munro plume adiabat crosses the solidus at much shallower depths, ~200 km, and produces 30-40% melting. Gorgona komatiites are cooler still and begin melting at ~150 km with melt extents of 20 – 30%. Most komatiites have been shown to have source regions that were depleted by a previous melt extraction event, and so the actual amount of melting that produced the komatiites is probably lower than the numbers given above.

Figure 2: Depths and temperatures of melting of komatiites in a plume setting inferred by Herzberg (1995). Black lines show the pressure-temperature path experienced by a body of mantle as it ascends adiabatically towards the surface. The gray lines are the solidus and liquidus of KLB-1, a fertile mantle composition. The solidus is the first point at which the mantle begins melting. The liquidus is the point at which the mantle is completely molten. Below the solidus, the slope of the lines reflects slight cooling due to decompression. The potential temperature is the temperature at which this path intersects the surface. The greater slope above the solidus reflects both the heat lost to produce melting and greater compressibility of melts relative to solid. In the plume model, the oldest komatiites (Barberton at 3.5 Ga) record the highest mantle potential temperatures, while late Archean (2.7 Ga) Munro komatiites and Cretaceous (8.8 Ma) Gorgona komatiites are formed in progressively cooler plumes. Note that if melting occurs by adiabatic decompression, then higher melting temperatures require that melting begins at greater depths. The range of melting temperatures for present day mid-ocean ridge basalts (MORB) is also shown.

The Subduction model

Along the way, as with all paradigms right or wrong, there were dissenting voices. Brooks & Hart (1972; 1974) first pointed out that the major-element chemistry of many of the komatiites and related magmas (komatiitic basalts) were more similar to modern mafic subduction magmas than to any magma thought to be produced by a modern plume (Gorgona had not been discovered at this point). Allegre et al. (1982) pointed out that the very high melting temperatures could be lowered if the melting incorporated significant amounts of H₂O. Perhaps the first strong evidence for a subduction origin for a suite of komatiites was presented by Wilson & Versfeld (1994) who studied the Nondweni komatiites. These komatiites have much higher SiO₂ than the Munro or Gorgona komatiites, and show some similarities to modern mafic subduction magmas (boninites, Figure 1). SiO₂ contents of magmas generally decrease as the pressure of melting increases, and so high SiO₂ is difficult to explain in a plume scenario. Boninites have similarly high SiO₂ at high MgO contents because they are high-degree melts produced at shallow depths. In an anhydrous, decompression melting regime, as plumes are thought to represent, high degrees of melting necessarily must begin at great depth, but in the case of boninites, large extents of melting can occur at shallow depth because the melting is caused by high H₂O contents (Crawford et al., 1989). While Wilson & Versfeld (1994) noted the similarities with boninites, they stopped short of arguing for a subduction origin.

The first evidence for high H₂O contents came from Stone et al. (1997) who discovered hydrous amphibole in a komatiite. Most komatiites contain hydrous tremolite, but this is produced by the metamorphism that has affected all komatiites and is not indicative of the H₂O content of the magma. Stone et al. (1997) found pargasitic amphibole which is often an igneous mineral though it can also be produced by metamorphism. Hanski (1992) also found pargasitic amphibole in Fe-rich komatiites in Finland. Melt inclusions, now devitrified and possibly altered, have also been analyzed for H₂O contents (McDonough & Ireland, 1993; Shimizu et al., 2001) with up to 0.8 wt.% water being inferred.

I entered komatiite research as graduate student in Tim Grove’s lab at MIT. Maarten de Wit had shown Tim some work he was doing on the Barberton komatiites and Tim noted that the igneous clinopyroxene that was sometimes preserved in the rocks was fairly high in Ca (wollastonite component) given the low CaO contents of the magmas. Having worked extensively with hydrous, mafic magmas, Tim knew that one effect of H₂O was to produce high-Ca pyroxene.

So my task was set; run dry experiments and see if I could reproduce the high Ca clinopyroxene (augite) found in the Barberton komatiites. Needless to say, it didn’t work. Under dry conditions, the first pyroxene to appear was actually pigeonite (low Ca cpx), which is not surprising given the ultramafic composition of the komatiites (Parman et al., 1997). The augite that subsequently crystallized was significantly lower in CaO (Wo = 0.3 – 0.35) than found in the rocks (Wo = 0.4 – 0.43). This is true whether the augite is crystallized under isothermal or cooling conditions.

However, when ~ 6 wt.% H₂O was added to the melt, the pigeonite was suppressed and the first pyroxene to crystallize was augite (as is the case in the rocks). Even better, the compositions of the experimental augites were an excellent match for the natural augites (Figure 3). Correcting for olivine crystallization, the data suggest 3 – 4 wt.% H₂O in the Barberton komatiite melts.

Figure 3: Composition of augite in Barberton komatiites and in dry and wet experiments. Top: Augites produced in anhydrous experiments (labeled field) have distinctly lower wollastonite (Wo) contents than the cores of augites found in the komatiites (darker fields), while augite in experiments with 6 wt.% H₂O are a good match for the natural augite (labeled field). Points show the compositional zoning from core (high Mg#) to rim (low Mg#). Mg# = Mg/(Mg+Fe). Bottom: Augites in komatiites that grew near the chill margin (diamonds) and so grew under rapid cooling conditions incorporated high Al contents, while the augite that grew in the interior (circles; and cores in 4a) are close to the equilibrium values measured in experiments, indicating that kinetics have had little effect on the major-element composition of the augites.

At first, it was unclear whether the H₂O required the Barberton komatiites to be produced in a subduction zone, or whether a hydrous plume scenario was possible. Subsequently, we compared the major and trace element chemistry of komatiites and basaltic komatiites with modern mafic magmas (Parman et al., 2001; Parman et al., 2003). As had been found before (Cameron et al., 1979), the compositions of the basaltic komatiites is very similar to modern boninites (Figures 1 and 4b). Small differences are: slightly higher TiO₂ and FeO, smaller high-field-strength element (HFSE = Nb, Ta, Zr, Hf, Ti) anomalies, and less depletion in middle rare-earth elements (REE). Nevertheless, basaltic komatiites are much, much closer to boninites than to any OIB or supposedly plume related picrite (Figure 1). Many are indistinguishable from modern boninites.

Figure 4a: Trace-element composition of Barberton komatiites. Two samples from Barberton show HFSE depletions, while the third (B94-4) has the same unusual high Zr seen in boninite and basaltic komatiites (Figure 2). B94-4 also has a slightly U-shaped REE pattern.

Figure 4b: Comparison of the trace-element composition of Barberton basaltic komatiites with Phanerozoic boninites. Elements arranged in order of compatibility. Both boninites and basaltic komatiites show a wide range of LREE slopes, sometimes resulting in U-shaped patterns. HFSE are variable, sometimes showing large depletions and unusual Zr enrichments.

Komatiites are higher in MgO than boninites, and one cannot make a direct comparison. However, there are similarities. High SiO₂ for given MgO, very low TiO₂ (Figure 1), variable LREE (Figure 4a) and in particular an interesting decoupling between Zr-Hf and Ti. Usually these HFSE behave similarly, but in some komatiites (and many modern boninites), Zr and Hf are enriched relative to the REE (e.g., sample B94-4 in Figure 4b) while Ti is depleted. How this is produced in boninites, let alone the komatiites, is not clear, but it is a very uncommon feature of other types of magmas and has been used as diagnostic feature of boninites (Hickey & Frey, 1982).

Figure 5: Subduction model for the origin of komatiites in subduction zones, modeled after melting process inferred for boninites (Stern & Bloomer, 1992). 1. Boninite-like melting event produces a range of high-MgO melts: komatiite, komatiitic basaltic magmas and low-Ti tholeiites (dark shading) and a corresponding heavily depleted mantle residue (stippled region). 2. Continued subduction of the lithosphere cools the mantle and establishes a mature subduction. Subsequent lower temperature hydrous magmas (calc-alkaline andesites) are emplaced on top of early, ultra-mafic crust. The cold, buoyant residual mantle remains beneath the fore-arc. 3. Komatiitic crust is obducted onto continents during continent collision at the end of subduction. The depleted mantle is thickened and incorporated into the continental lithospheric mantle.

The chemical similarities between boninites, basaltic komatiites and komatiites has led us to the conclusion that komatiites were produced by similar melting processes that produce modern boninites, with the primary difference being that the mantle was about 100°C hotter in the Archean (Parman et al., 2001). This estimate is based on high-pressure, H₂O-undersaturated melting experiments (Parman & Grove, 2004). Of course, the origin of boninites themselves is a matter of some debate. For simplicity, we have focused on the model that seems most plausible to us (Stern & Bloomer, 1992) in which boninites are formed at the initiation of subduction (Figure 5). Other models suggest subduction of mid-ocean ridges or the influence of plumes leads to the production of boninites (Crawford et al., 1989) and it is seems likely that boninites are formed by various arc-related melting processes. But in all cases, the inferred melting conditions (shallow, hydrous melting) are the same and do not change the ultimate conclusion that production of komatiites by hydrous melting would require a slightly (~ 100°C) hotter upper mantle in the Archean.

Subsequent to the publication of our model, Wilson discovered another high-SiO₂ komatiite location – Commondale (Wilson, 2003; Wilson et al., 2003). These have over 50 wt.% SiO₂ and trace element contents that suggest a very depleted mantle source. They are the most boninitic of all komatiites in major element composition and source depletion. Unlike most boninites, their LREE and LILE have not been enriched by the hydrous fluid that is hypothesized to have instigated the melting event. On the other hand, there are a number of boninites that are not LILE enriched (Figure 4b). Wilson and colleagues conclude that both of these characteristics are difficult to reconcile with a plume origin, and require some type of hydrous melting process, probably subduction.

Final Thoughts

A trend that has become clear in the last few years is that advocates of a subduction origin for komatiites focus on the high-SiO₂ komatiites e.g., Commondale, Nondweni and Barberton, while plume advocates focus on low SiO₂, OIB-like komatiites e.g., Munro and Gorgona. This may suggest that komatiites were produced in more than one tectonic setting, though personally I find this unappealing. The similarities in composition and geology of komatiites seem to me to be much greater than their differences. To reconcile the two opposing hypotheses, some researchers have proposed hybrid models in which komatiites are produced by hydrous melting in Archean plumes (Asahara & Ohtani, 2001). The water is envisioned to be primordial H₂O inherited from the accretion of the planet.

Whatever one's view on komatiites, I think it is fair to say that their origins are the subject of active debate. Furthermore, there are some komatiites, Commondale in particular, that are very difficult to fit into a plume scenario because their high SiO₂ suggests relatively shallow melting. If it is eventually shown that all (or most) komatiites were produced by subduction-related melting processes, this would call into question the presence of thermal mantle plumes in the modern world. Komatiites, as the product of super-hot Archean plumes, are an important stone in the foundation of plume theory in general. If komatiites are not produced in plumes, then there is no evidence for great plume activity in the early Earth, even though this is a fundamental prediction of mantle convection models that incorporate thermal plumes. Then, views of thermal plumes in the modern mantle are also called into question or, at least, need to be modified substantially. Conversely, if modern OIB e.g., at Hawaii are indeed the product of upper mantle melting, and not of thermal plumes from the CMB, this also calls into question the origin of komatiites in thermal plumes.