Komatiites are rare ultramafic volcanic and subvolcanic rocks that occur, predominantly, in Archean and Paleoproterozoic greenstone belts. These rocks contain more than 18 wt% MgO (Le Bas 2000) and are distinguished from other magnesium-rich rocks, such as picrites and meimechites, by having spinifex texture (characterized by spectacular arrays of subparallel or randomly-oriented skeletal, platy and bladed olivine crystals set in a glassy groundmass; Fig. 1). Because all komatiites do not display spinifex texture, Arndt and Fowler (2004) defined komatiites as ultramafic rocks that either contain spinifex texture or are related to rocks that do (see Kerr and Arndt 2001). Komatiitic lavas are assumed to crystallize from magmas that have ultramafic compositions.

Komatiites were first identified by Viljoen and Viljoen (1969) in the Archean greenstone belt in the Barberton Mountainland of South Africa as a distinctive and "new" class of magnesium-rich (20–30 wt% MgO) volcanic rocks. Up until then, genuine ultramafic lavas were not known. Viljoen and Viljoen (1969) recognized that many high-Mg rocks in the Barberton greenstone belt were lava flows of significant lateral extent and thickness, with chilled or brecciated tops, amygdules, pillows and the distinctive quench textures that were subsequently named "spinifex" (after an Australian spiky grass - Triodia spinifex). The Viljoen brothers inferred that the rocks had characteristics of erupted liquids with distinctive chemical composition and named them after the Komati River flowing through the type locality. These 3.5 Ga flows are among the oldest known ultramafic rocks.

Similar ultramafic lavas were soon described from other Archean and Paleoproterozoic belts particularly from Canada, Africa, Australia and Finland. Most are Neoarchean to Paleoproterozoic (Fig. 2) but extend into the Phanerozoic, including Permian–Triassic komatiites from NW Vietnam (Glotov et al. 2001; Hanski et al. 2004) and Cretaceous (~ 90 Ma) komatiites from Gorgona Island off the coast of Columbia (Kerr et al. 1996; Kerr 2005; Brandon et al. 2003). However, komatiites have no known modern analogues. The predominance of komatiites in the Archean, their decreasing occurrence in the Proterozoic, and extreme rarity in the Phanerozoic have been interpreted to reflect secular cooling of the mantle (by up to 400EC; Nisbet et al. 1993). Although komatiites are considered to be important windows into Earth’s early mantle and therefore have major implications for constraining models of the thermal evolution of Earth, the temperatures, sources and depths of their formation remain controversial. In addition, komatiites are of economic interest as they are associated with significant magmatic nickel sulfide ore deposits in Australia and Canada.

Although, where present, they occur only in minor/subordinate amounts, komatiites are an integral part of subaqueous volcanic successions of Precambrian greenstone belts. De Wit and Ashwal (1997) estimated that komatiites constitute typically less than 10% of the total volume of volcanic rocks in most greenstone belts although in some areas they form up to 30% of the volcanic succession (Fig. 2). Komatiite units can be tens or even hundreds of metres thick and can be traced continuously for up to 20 km, indicating that komatiite eruptions were major events during the Precambrian.

Komatiites occur either as lava flows or subvolcanic bodies; rarely, they are also pyroclastic. Komatiite lavas range from thin (a few cm) to massive to thick (> 100 m) and are distinctly layered. Individual flows displaying well-developed layering (Arndt et al. 2004; Grove et al. 1997) show a textural division that includes an upper part (Zone A), characterized by spinifex-textured rock, and a lower part (Zone B) containing a high proportion of equant olivine crystals resembling various peridotite to dunite cumulates (Fig. 3). Both zones have been further subdivided into several subzones, although the individual subzones may not be continuous and may occasionally be absent. Many thin and even some thick komatiite flows are not layered and do not have spinifex zones. They may have a massive texture, and olivine phenocrysts are commonly concentrated in the centre of such flows. Dann (2001) reported massive komatiite sheets, 12 to 50 m thick, traceable for more than 11 km and compositionally and texturally homogeneous. Massive lavas may pass laterally into flows with well-developed spinifex textures. The variation in the textures and layering of komatiite units has been attributed to variations in flow volume, lava flow velocity, effusion rate, topography and cooling rate (Sylvester et al. 1997; Dann 2001; Dann and Grove 2007), all of which might also be related to distance from an eruptive site.

Komatiites are associated with komatiitic basalts (12–18 wt% MgO; Arndt et al. 1997), which have similar flow facies and textures as komatiites, except that the dominant phenocrysts are pyroxenes. Their differentiated flows contain upper pyroxene spinifex zones and lower cumulate zones.

Komatiite lavas are highly fluid because of their low viscosity and high liquidus temperature, so they would be expected to form rather thin flows. The thick komatiitic flows were probably ponded (lava lakes and rivers; Hill et al. 1995) or were thickened by flow inflation (flow ballooning) due to repeated flow pulses, like pahoehoe basalts of Kilauea, Hawaii (Dann 2001; Dann and Grove 2007). Komatiite lavas are inferred to have the ability to flow over considerable distances from eruption sites (Dann 2001; Sproule et al. 2002). Once a solid crust formed around a flow, it would create an efficient insulating surface that would facilitate the flow of hot komatiitic lava.

Some komatiitic rocks are intrusive; they usually form high-level dikes and sills and even layered differentiated bodies. These rocks have textures and compositions similar to those of their volcanic equivalents and might be the hypabyssal parts of the komatiitic volcanic suites (Arndt et al. 2004). Small amounts (typically <1%) of komatiitic volcaniclastic rocks have been documented in some greenstone belts (Sylvester et al. 1997).

A variety of geological environments has been proposed for the emplacement of komatiites including mid-ocean ridges, plumes, oceanic plateaus, giant meteorite impacts and magma oceans (Grove and Parman 2004). There has been a suggestion that some komatiites are allochthonous, i.e. that they have been tectonically transported and interleaved with rocks of different environments (Sylvester et al. 1997). Most komatiites are associated with Archean and Paleoproterozoic greenstone belts (Arndt et al. 1997) where komatiites occur mainly in lithological assemblages that are considered to be remnants of ocean plateaus or arcs (Condie 2001). Some komatiites, including those from the quartzite–komatiite association that is widespread in the western Superior Province, overlie older granitic basement and are inferred to be autochthonous (Thurston and Chivers 1990; Bickle et al. 1994; Sylvester et al. 1997; Shimizu et al. 2004).

In Canada, some of the most prominent komatiitic occurrences are in the Abitibi belt, which contains several komatiitic successions that were emplaced over ca. 20 Ma (2724–2703 Ma). Komatiitic rocks of the Abitibi belt occur mainly in the Timmins– Kirkland Lake–Lake Abitibi region of Ontario and Québec.

Komatiites and related rocks have been affected to variable degrees by metamorphism, hydrothermal and seafloor alteration, and deformation, which have, at least in part, obliterated the original textures and primary mineralogy. Hence, komatiites generally contain metamorphic minerals in place of their primary assemblages, relics of which may nonetheless be fortuitously preserved in some instances. Low-grade metamorphism of komatiites produced mineral assemblages dominated by serpentine–antigorite, chlorite, talc, tremolite, magnesite–dolomite and magnetite. At higher metamorphic grades, metamorphosed komatiites contain anthophyllite, enstatite, olivine and diopside.

The primary mineralogy of komatiites is simple: phenocrysts of olivine and spinel/chromite (+/pyroxene) are enclosed in a groundmass composed of glass, calcic clinopyroxene and minor orthopyroxene. In the spinifex-textured komatiites (Fig. 1), elongate and skeletal olivine (or pyroxene in komatiitic basalts) blades ranging in length from mm to tens of cm are set in a fine-grained matrix that originally contained a large proportion of volcanic glass ranging in composition from komatiite to komatiitic basalt (Donaldson 1982). Olivine is highly magnesian and shows normal zoning from cores of Fo95-90 to rims of Fo92-84. Their contents of Ni and Cr are high (up to > 4000 ppm and >2000 ppm, respectively; Donaldson 1982).

Petrologists who initially studied spinifex texture noticed that the skeletal olivine phenocrysts in komatiites (Fig. 1) resemble quench crystals (formed at very rapid cooling rates) in experimental melts (Donaldson 1982). Solidification under the conditions of supercooling (low nucleation rates and rapid crystal growth rates) produces a few large skeletal or dendritic crystals (Fig. 1). However, this model cannot explain the occurrence of spinifex texture within thick komatiite flows well below the upper chilled crust. Large skeletal crystals may have crystallized at depths, many metres below the surfaces of the flows (Fig. 3) where cooling rates must have been low. To overcome this problem, Viljoen and Viljoen (1969), among others, suggested that komatiites, most of which were related to submarine eruptions, were rapidly cooled by sea water. However, present day submarine basalts have only narrow rims that show features of rapid cooling and neither Archean basalts coeval with komatiites nor modern submarine basalts have spinifex textures. The origin of spinifex textures thus presents a dilemma: whereas the shape of crystals is suggestive of fast cooling near the flow margins, the blades formed deep within the flows.

As the spinifex texture is confined to MgO-rich basaltic to ultramafic rocks, a partial explanation is that spinifex texture is related to the temperature difference between the liquidus and solidus, which is very large for komatiites and Mg-rich rocks (400–500EC; Faure et al. 2006) compared with typical basaltic rocks (<100EC). However, in detail, the origin of the spinifex texture remains problematic. Because olivine crystals are thermally anisotropic, Shore and Fowler (1999) suggested that heat transfer increases the cooling rate in front of the crystal tips leading to the formation of platy crystals. Alternatively, a recent experimental study of Faure et al. (2006) inferred that spinifex texture is a result of slow cooling of ultramafic magma in a thermal gradient within a layer that separates magma from the solid outer crust. Another explanation has been put forward by Grove et al. (1997), who proposed that elevated water content in komatiitic magmas would lead to the rapid growth of large crystals, and accompanying degassing of hydrous komatiites would generate a strongly supercooled liquid. However, this hypothesis has been challenged (Arndt et al. 1998).

Virtually all komatiites and associated rocks have been metamorphosed and chemically altered. Many elements used in discussion of modern volcanic rocks such as Na, K, Rb, Sr and Ba as well as H2O and CO2 have commonly been redistributed. Thus geochemical/ petrological investigations of komatiites have been based upon elements that are considered to be less mobile, including Al, Ti, high-field-strength elements (HFSE) and rare-earth elements (REE).

Another potential complication is crustal contamination. Komatiitic magmas are prone to contamination because of their high temperatures. Crustal contamination of komatiite enriches light REE (LREE) and Th relative to heavy REE (HREE) and HFSE, particularly Nb and Ta (Jochum et al. 1991). Because of a significant compositional contrast between continental crust and komatiites, element ratios such as Nb/Th, Nb/U, Th/La and Nb/LREE are sensitive indicators of contamination (Jochum et al. 1991). Similarly, radiogenic isotopes, particularly Nd, provide useful constraints on contamination by older continental crust. The geochemical data suggest that crustal contamination did not play a major role in the genesis of most komatiites (e.g. Sproule et al. 2002).

Komatiites have a chemical composition similar to peridotite or dunite, with high MgO, Ni and Cr but low contents of SiO2, TiO2 (<1 wt%), K2O (< 0.5 wt%), Na2O and incompatible trace elements. Arndt et al. (1997, 2004) inferred that original komatiitic liquids (i.e. lavas without olivine phenocrysts) had about 28–30 wt % MgO. This is close to the composition of aphyric komatiites and chilled margins of komatiitic flows. Olivine that crystallized from a parental magma of 28–30 wt% MgO would have a composition of ~ Fo94, similar to olivines in the lava flows (Arndt et al. 2004; Lesher and Arndt 1995).

Many komatiites show significant compositional variations, even within a single lava flow (Fig. 4). Most of these variations can be accounted for by fractionation of the first crystallizing mineral, olivine, involving either the gain or the loss of olivine phenocrysts. In differentiated flows, olivine cumulates at the base of the flows have 30–45 wt% MgO (Figs. 4, 5) whereas the spinifex-textured komatiites of the upper zone (residual liquids) have lower MgO (Fig. 4). This decrease of MgO in the differentiated komatiite flows from the basal cumulates to the spinifex zone is accompanied by an increase in Al2O3, TiO2, CaO and incompatible trace elements because these components are excluded from olivine (Fig. 5). Crystallization of olivine does not modify the element ratios as these elements will be enriched in the residual liquid to the same degree. The chondrite- and mantle-normalized trace-element patterns of the related spinifex and cumulate rocks are parallel although the former have higher absolute concentrations (Figs. 6, 7). Pyroxene spinifex-textured komatiitic basalts have lower MgO and Ni and frequently compositionally grade into komatiites.

According to their chemical composition, komatiites are usually subdivided into two major types (Nesbitt et al. 1979; Arndt et al. 1997): aluminum (Al)-undepleted (or Munrotype, named after their type locality in the Abitibi greenstone belt) and aluminum-depleted (or Barberton-type). Al-undepleted komatiites (AUK) are characterized by ratios of Al2O3/TiO2 ~ 20 and CaO/Al2O3 ~ 1, values that are similar to those of chondritic meteorites and primitive mantle. Their REE patterns (Fig. 6) are typically slightly depleted in LREE and have a flat HREE segment with (Gd/Yb)n ~ 1 (nchondrite-normalized). They resemble the patterns of recent N-type midocean ridge basalts (MORB) although the absolute REE abundances in komatiites are significantly lower than in MORB (Fig. 6). The mantle-normalized patterns of AUK are also characterized by near chondritic Ti/Zr (~110) and depletion of Th, Nb and LREE (Fig. 7). Al-undepleted komatiites are the most widespread komatiitic lavas and occur predominantly in 2.7 Ga Archean greenstone belts including the Abitibi (Ontario-Québec), Belingwe (Zimbabwe) and Norseman-Wiluna (Australia) but are rare in pre-3.0 Ga greenstone belts. They include the 2718–2710 Ma Kidd– Munro assemblage (eastern Ontario), which contains the ~1000 m-thick Munro komatiite flows. Some AUKs are also of Proterozoic and Phanerozoic ages such as those from Gorgona Island (Kerr et al. 1996).

Al-depleted komatiites (ADK) are lower in Al2O3 and have low Al2O3/TiO2 (typically <12; Fig. 8) but high CaO/Al2O3 (~2–2.5). Their REE patterns (Fig. 6) have fractionated HREE with (Gd/Yb)n >1.3. Al-depleted komatiites have higher contents of strongly incompatible trace elements (Th, LREE) than the AUK but their mantle-normalized patterns typically show small negative Zr and Hf anomalies (Fig. 7). Al-depleted komatiites are the most abundant komatiitic lavas in the 3.5–3.0 Ga Barberton greenstone belt of South Africa and the ~3.5 Ga greenstone belts of the Pilbara craton (Australia) but in other greenstone belts, particularly those of post-3.0 Ga age, they are rare. An association of both AUK and ADK is uncommon but has been documented in the Abitibi greenstone belt (e.g. Dostal and Mueller 1997; Fan and Kerrich 1997) among other places. Komatiitic basalts include both Al-undepleted and Aldepleted types and their characteristics match those of komatiites. In addition to the two komatiite groups, several other komatiite subtypes have been recognized in particular regions but they are only subordinate (e.g. Sproule et al. 2002).

The Nd isotopic data on relatively fresh komatiites as well as their fresh pyroxenes from the Abitibi belt yielded GNd values of 2.8 to 3.8 (e.g. Lahaye et al. 1995). Some komatiites have slightly higher GNd values (e.g. Lesher and Arndt 1995).

Komatiites and komatiitic basalts locally host magmatic Ni-sulfide (Ni-Cu-PGE) mineralization. Famous deposits include those in the Archean Yilgarn and Pilbara cratons of Western Australia, the Abitibi greenstone belt, Zimbabwe craton, the Proterozoic (1.8 Ga) Cape Smith (Ungava) belt of Québec and the Proterozoic Thompson belt of Manitoba. Principal ore minerals are pyrrhotite, pentlandite, pyrite and chalcopyrite. The sulfides formed as immiscible liquids and thus they are of primary magmatic origin. The production of large amounts of immiscible sulfide droplets that settle in the host magma to form an orebody requires the attainment of sulfide saturation in the komatiitic melts, mainly through the assimilation of sulfur from sulfurrich country rocks. Nickel, copper and platinoid elements as well as iron would then be scavenged by the sulfur to form sulfide droplets.

Naldrett (2004) subdivided these komatiite-hosted deposits into three groups. The first type includes ore bodies that are generally small (1 to 5 x 10⁶ tonnes) but high grade (1.5–3.5 wt% Ni) and occur where massive sulfides are concentrated at the base of the host ultramafic flows in zones up to about 50 m thick. Examples include the 2.7 Ga deposits of the Kambalda district (with 0.8–1.4 wt% Cu), which lie in the Yilgarn craton, and the Langmuir deposit of the Abitibi greenstone belt, about 40 km southeast of Timmins, Ontario.

The second deposit type encompasses large (100 to 250 x 10⁶ tonnes) but low grade (~0.6 wt% Ni) deposits of disseminated sulfides in olivine-rich cumulate dikes/sills that feed komatiitic flows. Examples of these deposits are the Six Mile and Mt. Keith deposits near Yakabindi, Western Australia and the Dumont deposit of Québec (~ 60 km NE of Rouyn). The third type includes deposits related to Proterozoic komatiitic basalts that occur in rifted continental margin environments. They host deposits of the Thompson (Manitoba) area and the Cape Smith fold belt (Ungava Peninsula, Québec).

Komatiites provide a window into the composition and thermal dynamics of the Archean mantle. They are usually considered to be the product of mantle plumes generated at depth, possibly near the core–mantle boundary (Campbell et al. 1989; Arndt et al. 1998; Sproule et al. 2002). The chemical compositions of komatiites and related Mg-rich rocks appear to vary with time. For example, with respect to the MgO contents of common ultramafic magmas, komatiites characteristic of the Archean have >18 wt% MgO, komatiitic basalts dominant in the Proterozoic have 12–18 wt% MgO and picrites in the Phanerozoic have 10–13% wt% MgO. Because the MgO contents of dry magmas are related to their melting temperature, the trend is consistent with secular cooling of the Earth’s interior. However, recently, some researchers (Parman et al. 1997; Grove and Parman 2004; Wilson et al. 2003) have challenged the plume model and attendant volatile-free mantle melting, and postulated that komatiites may be produced by hydrous melting at shallow mantle depth in a subduction (arc) environment. In this scenario, temperatures would be considerably lower than those required by the plume model. Since many current models of chemical differentiation and thermal evolution of the Earth are based on mantle plume-generated komatiites, the subduction model, if correct, would significantly change current views of the evolution of Archean Earth. Although recent data imply that many komatiites are derived from a dry mantle source, there are still some rocks that could be subduction-related, such as high-SiO2 komatiites from the ~3.33 Ga Commondale greenstone belt in South Africa (Wilson et al. 2003), and some komatiitic basalts that resemble boninites. More studies of the trace element and isotopic characteristics of komatiites and their minerals as well as experimental studies, particularly under wet melting conditions, are needed to constrain their origin. Obviously, there is much left to learn about these fascinating rocks.