Major Element Chemistry

8 Major element abundances profoundly influence physical properties such as magma viscosity and density, which in turn affect the rate at which magmas rise in the crust. The major element chemistry of parental magmas formed by partial melting is primarily controlled by phase equilibria. This principle is shown schematically in Figure 2 in which a magma of composition M, representing the minimum temperature in the system, is produced by melting of host rocks (e.g. X) with a wide range of bulk compositions. Once magma is produced, its major-element composition mostly controls what silicate minerals eventually grow from the magma, and so the chemical evolution of the cooling magma is constrained by phase equilibria. Experimental and observational evidence suggest that olivine, followed by clinopyroxene and plagioclase, are the earliest crystallizing phases from mafic magma. Fractional crystallization of ferromagnesian phases such as olivine and clinopyroxene leads to systematic depletion in FeO*, MgO, and enrichment in elements such as SiO2, Na2O and K2O. When plagioclase appears on the liquidus, the remaining liquid becomes depleted in CaO and Al2O3. Figure 2 shows the evolution of a typical basaltic liquid in a shallow magma chamber (i.e. low pressures). As long as the system remains closed, cooling of a basaltic liquid of composition X will initially produce forsterite (olivine), and then enstatite. Equilibrium crystallization or fractionation of these minerals drives the residual liquid composition away initially from olivine (X to T) until it reaches and then descends the olivineenstatite curve (T to P). Along pathway X to P, the magma becomes enriched in other components, including those of plagioclase (represented by anorthite). As path T to P is along a reaction curve, olivine reacts with the liquid and enstatite is precipitated. Once plagioclase begins to crystallize, the residual magma also becomes depleted in CaO and Al2O3. The overall result is that mafic liquids evolve towards more intermediate and felsic compositions.

Display large image of Figure 4

Display large image of Figure 5

9 The composition of the minerals also reflects fractionation. The FeO*/MgO ratio found in the olivines and pyroxenes increases with fractionation, as shown, for example, in the simple phase relationships in Figure 3. Olivine and clinopyroxene phenocrysts in arc basalts are more iron-rich than those that are in equilibrium with mantle rocks, suggesting that even the most primitive arc magmas are fractionated to some degree and that true primary magmas (i.e. melts directly derived from the mantle) are rare.

10 Mineral compositions are also susceptible to the tectonic environment in which they are formed. For example, Leterrier et al. (1982) showed that for clinopyroxenes of similar Ca content, those in arc-related mafic rocks have lower Ti + Cr than those in non-orogenic settings (Fig. 4).

11 As most volcanic rocks are dominated by a glassy or fine grained matrix, their minerals are poorly developed and so they are more effectively classified by their chemistry rather than their mineral content. The most familiar classification of lavas is based on silica content (Fig. 5). However, arc magmas are so chemically diverse, that to be able to understand the petrologic processes responsible, additional elements must be used. Magmas with similar silica content can be distinguished from one another by examining the concentration of other elements. For example, many basalts in subduction zone environments are characterized by high Al2O3 (17.0 – 21.0 wt %) and low MgO (< 6 wt %), compared with basalts in other tectonic settings, and are known as high-alumina basalt. Although their origin is unclear, their low MgO content suggests that they are fractionated (by removal of olivine and augite) from a more mafic parent. The high Al2O3 implies that plagioclase was not a liquidus phase.

12 Two suites, or collections of magmas, known as alkalic and subalkalic, can be distinguished on the basis of their alkali content (Na2O + K2O; MacDonald 1968; Cox et al. 1979; Wilson 1989; Rollinson 1993). For the same silica content, alkalic basalts have higher Na2O + K2O than subalkalic basalts, and alkalic rhyolites have higher Na2O + K2O than subalkalic rhyolites (Fig. 5). Subalkalic suites are further divided into tholeiitic or calcalkalic suites on the basis of their FeO* and MgO contents. In both suites, FeO* and MgO decreases with fractionation, which suggests the importance of ferromagnesian minerals, such as olivine and pyroxene in the evolution of both types of magma. However, for rocks of the same silica content, tholeiitic rocks have higher FeO*/MgO ratios than calc-alkalic rocks (Miyashiro 1974; Fig. 6).

Display large image of Figure 6

13 Even within an individual suite, there are further distinctions that can be made. For example, tholeiitic rocks that occur in arc environments can be distinguished from tholeiitic rocks in other settings by the concentrations of other elements, including their K2O content, which tends to be lower than in rift-related tholeiites (e.g. Gill 1981). These island-arc tholeiites (IAT) are also known as "low-K tholeiites" and typically have K2O < 0.8 wt%.

14 Although the generation of each suite of rocks is controversial, there are some general over-arching principles. Using the Japanese arc as an example, Kuno (1966) drew attention to the increasing K2O content in rocks that have similar SiO2, with increasing distance from the trench (and hence greater height above the subduction zone), the so-called K-h relationship (Dickinson 1975). Thus tholeiitic basalts, which typically occur nearer to the trench, have lower K2O contents than calc-alkalic basalts, whose K2O content increases farther away from the trench. Gill (1981) recognized volumetrically small, but petrologically significant, potassic-rich basaltic to intermediate lavas, known as shoshonites that occur far from the trench in continental arcs.

15 Some experimental data suggest that high PCO2 in the mantle may yield silica-undersaturated (i.e. olivinenepheline normative) alkalic basalts whereas high PH2O produces silica-saturated basalts (e.g. Mysen and Boettcher 1975, Boettcher 1977; Eggler 1978; Mysen 1988). As alkalies readily partition into melt, alkalic basalts are thought to represent small degrees of partial melting, possibly from a source mantle previously enriched in alkalies. However, the ascent of such small volumes of melt is mechanically problematic (see Murphy 2006) and so alkalic magmas in arc settings tend to be restricted to specialized settings, such as local rift zones within the arc or backarc region.

16 The major element composition of calc-alkalic and tholeiitic lavas is reflected in their mineral contents. Arc lavas, especially calc-alkalic lavas, contain abundant phenocrysts and microphenocrysts (up to 20%). A summary of the range of phenocryst content with magma type and composition is shown in (Table 1; after Wilson 1989). Plagioclase is the most abundant phenocryst, and occurs in both calcalkalic and tholeiitic suites in lavas that range from basaltic to rhyolitic in composition. Plagioclase is characterized by complex zoning patterns that reflect disequilibrium conditions (Shelley 1993; Stewart and Fowler 2001; Murphy 2006) For both tholeiitic and calcalkalic suites, clinopyroxene phenocrysts occur in mafic to intermediate lavas, olivine phenocrysts are typically restricted to the more mafic lavas and orthopyroxene, amphibole and quartz phenocrysts are more common in intermediate to felsic lavas. Many phenocrysts, like plagioclase, exhibit textural and chemical evidence of disequilibrium. Although magnetite is a common phenocryst in intermediate to felsic lavas of both magma series, it also occurs as a phenocryst in mafic lavas in calc-alkalic suites.

Table 1. Typical phenocryst content of island arc tholeiitic and calc-alkalic basalt (B), basaltic andesite (BA), andesite (A), dacite (D) and rhyolite (R) (after Wilson 1989). Solid lines, common; dashed lines, rare.

Display large image of Table 1

17 As the calc-alkalic suite is the most dominant suite in arc environments, especially so in mature arcs, the processes responsible for their limited increase in FeO*/MgO have received considerable attention. Phase-equilibria studies show that, like tholeiites, fractionation of olivine and pyroxene are important in the earliest stages of fractionation of calc-alkalic magmas. Miyashiro (1974) proposed that water derived from the subduction zone leads to oxidized melts, and the early precipitation of oxide minerals such as magnetite (Fe3O4). This interpretation is consistent with petrographic observations that magnetite is a common phenocryst in basaltic calc-alkalic basalts but is rare in tholeiitic basalts. Precipitation of magnetite would inhibit enrichment in FeO*/MgO with fractionation, leading to a calc-alkalic trend. Experimental evidence, however, yields conflicting results. Lee et al. (2005) found that the oxidation state of the magma in arc environments is similar to that of the mid-ocean ridge basalts (MORB), and that the degree of oxidation might be buffered by mantle mineral assemblages. Sisson and Grove (1993) show that low pressure (2 kb) fractionation of a H2O-rich (4-6 wt %) basalt can reproduce calc-alkalic trends by crystallization of olivine, Ca-rich plagioclase (An > 90) and either magnetite or spinel. Ringwood (1977) and Boettcher (1977) point out that H2O-rich magmas would increase the stability of hornblende, which would have a similar FeO*/MgO to the melt. Fractionation of hornblende would also inhibit the increase of FeO*/MgO during fractionation and so produce a calc-alkalic trend. The importance of amphibole as a phenocryst in intermediate rocks suggests it may play a role at this stage in the cooling of the magma.

18 Although there are some controversies, it is generally accepted that water plays a major role in producing calc-alkalic magmas, and so it is not surprising that these magmas are the dominant suite in many arcs. Several experimental studies (e.g. Sisson and Grove 1993; Grove et al. 2003) suggest that calc-alkalic magmas form by hydrous melting of the mantle, and then rise to the shallow crust where they undergo fractional crystallization under near-H2O saturated conditions. Sisson et al. (2005) point out that a significant quantity of rhyolitic liquid can be derived by differentiation or by low-degree partial re-melting of mantle-derived basalt.

19 Another alternative is that calc-alkaline trends and high volume of intermediate magmas are generated by mixing of mafic and felsic end-member magmas. This is supported by widespread petrographic evidence of textural disequilibrium in andesites (e.g. Eichelberger 1975; Eichelberger et al. 2000, 2006), and by geochemical and isotopic patterns consistent with mixing (Grove et al. 2005; see also the discussion below) and field observations (e.g. Wiebe et al. 2002).

20 Although several exceptions have been documented, arc magmas exhibit broad chemical trends in space and time during arc evolution. Tholeiitic magmas are common in the early stages of oceanic island-arc genesis and are closer to the trench. The calc-alkalic suite of magmas is the most common in mature oceanic arcs and generally occurs farther from the trench than arc tholeiites; this suite is especially common in continental arcs. There is also a tendency in both suites toward more silicic compositions with increasing arc maturity and crustal thickness, and many calc-alkalic suites range from basaltic to rhyolitic composition, and are typified by an abundance of intermediate (andesitic) rocks.

Table 2. Partition coefficients for various trace elements in rock-forming minerals (see Winter 2001, p. 157; Pearce and Norry 1979).

Display large image of Table 2

Trace Element and REE Geochemistry

21 Trace element, rare-earth-element (REE) and isotopic abundances reflect a variety of factors including the source of the magma, the percentage of the source rock that undergoes melting, the extent of crystal fractionation, and the contamination of either magma from adjacent rocks or from other magmas of other arc systems. In each case, trace element abundances can be modelled (see Appendix 1).

22 Trace element and REE abundances are sensitive to the appearance of minerals as liquidus phases. For example, Ni and Cr are heavily concentrated into early-crystallizing minerals such as olivine and pyroxene, and Y, Yb and Lu are concentrated in garnet. The fractionation of olivine and pyroxene depletes the remaining liquid in Ni and Cr, providing a sensitive indicator on the extent of their fractionation. Such elements are also known as compatible trace elements because they fit into the structure of early crystallizing minerals and so have a low concentration in the melt. Alternatively, trace elements such as Zr, Hf, Nb, Ta, Ti, and Th are known as incompatible trace elements. They strongly partition into magma, and so their abundances should increase systematically during crystal fractionation. Note that some elements behave incompatibly in some magmas but compatibly in others. For example, Ti is concentrated in horn-blende and both V and Ti are concentrated in magnetite. Nevertheless, these minerals can crystallize early or late in the evolution of magmas depending on PH2O and fO2 (e.g. Miyashiro 1974; Ringwood 1977). Most calc-alkalic suites record a decrease in Ni, Cr, Ti and V from basalt to andesite, consistent with experimental evidence for the fractionation of olivine and magnetite.

23 Generally, early crystallizing minerals such as olivine, clinopyroxene, plagioclase and orthopyroxene contain very low REE abundances (Table 2). Therefore, the REE are incompatible elements and partition into the melt causing the total REE to increase during fractionation from mafic to intermediate magmas. As the magma cools, REE elements eventually become concentrated in accessory phases such as zircon, rutile, ilmenite, titanite and monazite. In some highly siliceous magma, precipitation of accessory minerals is common (Miller and Mittlefehldt 1982) so that REE abundances can decrease during the later stages of magma evolution. Rare-earth-element abundances are most commonly displayed on diagrams in which the lanthanide elements are arranged from left to right with increasing atomic number (57-71, La to Lu on the periodic table). The lanthanides have similar chemical and physical properties (all but Eu are trivalent) and their ionic radius decreases with increasing atomic number (the so-called "lanthanide contraction"); Europium can be divalent or trivalent depending on the degree of oxidation of the magma. According to Goldschmidt’s rules, if ions have the same charge, the smaller ions are preferentially incorporated into growing crystals.

24 The concentration of each REE is divided by the concentration of that element in chondrite, which is thought to reflect the composition of the early Earth. This results in smooth patterns that facilitate comparisons among samples and between two suites. The resultant chondrite-normalized rare-earth-element plot (note that the y axis is dimensionless) allows an analysis of the enrichment (or depletion) in each lanthanide element relative to the same standard reference, in this case, that of the early Earth (Fig. 7). Analysis of these plots focuses on two different aspects. First, as REE elements are equally incompatible during fractionation of the early crystallizing phases, the slope on the REE diagram (as represented by normalized La/Lu) is virtually unaffected by fractionation and so represents their relative abundances in the source rock. An overall envelope around the RE E data from a given suite may therefore yield valuable information of the chemistry of that source. Second, total RE E (ZREE) should increase until the latest stages of fractionation (Fig. 7a), and suites, derived from the same source and related by fractionation, should yield parallel profiles. For crystal fractionation to be a viable model, samples that have higher ZRE E should have higher abundances of incompatible elements and so the RE E profiles are generally checked against abundances of elements such as Zr and Nb.

Display large image of Figure 7

25 Island arc tholeiites exhibit relatively flat RE E profiles, with minor depletion in light rare-earth-elements LREE). These profiles are broadly similar to MOR B (although MOR B has more pronounced LRE E depletion) suggesting derivation from a similar source, i.e. the depleted mantle. In contrast, calc-alkalic basalts show moderate degrees of light rare-earth enrichment, although they are not as enriched as alkalic or plume-related basalts, suggesting derivation from a mantle source enriched in these elements relative to the source for MORB or IAT.

26 Tholeiitic and calc-alkalic lavas are both characterized by flat heavy RE E (HREE) profiles. These profiles imply a source in the shallow (spinel lherzolite) mantle (i.e. < 50 km depth) rather than a deeper (garnet lherzolite) mantle. Garnet preferentially incorporates HRE E over LREE, so that a steepened RE E slope (i.e. higher La/Lu) will result if either fractional crystallization of garnet occurs during cooling, or if garnet remains in the residuum during partial melting (Fig. 7b). The effect of garnet on RE E profiles contrasts with other silicates, the fractionation of which increases ZRE E but does not change the slope of RE E profiles. In addition, the flat HRE E patterns are also a strong argument that the magmas are not directly derived from melting of the subducted slab, because oceanic lithosphere subducted to 110 km depth would be metamorphosed to eclogite (a garnetomphacite bearing metamorphic rock).

27 Plagioclase incorporates more Eu than the other REE because, unlike other lanthanides, which are predominantly trivalent, Eu is predominantly divalent (Eu2+/Eu3+ depends on fO2) and so can be incorporated into the Ca2+ site in the plagioclase structure. If fractionation of plagioclase occurs during cooling, the increase in Eu concentration lags behind the increases in other REE, resulting in a negative "kink" in chondrite-normalized REE patterns (Fig. 7b), known as a "negative europium anomaly". The extent of the anomaly increases with fractionation, and so is very pronounced in felsic rocks as can be seen by comparing normalized Eu abundance with a hypothetical Eu value obtained by extrapolating a hypothetical line from neigh-bouring elements (dashed line, Fig. 7b). The presence of a negative Eu anomaly in many tholeiitic and calc-alkalic volcanic arc suites is consistent with plagioclase fractionation, which is typical of processes in crustal magma chambers and matches field and petrographic evidence of the abundance of plagioclase phenocrysts.

28 When trace elements and REE elements are considered together, it is evident that arc systems are complex and magma compositions represent interplay of a large number of factors, including magma mixing, crustal contamination, crustal assimilation, and fluid transport. These complexities are revealed if ratios between incompatible trace elements in basaltic rocks are studied. If arc liquids evolve by simple fractionation, ratios such as La/Nb and Zr/Nb remain constant because La, Zr and Nb are (approximately) equally incompatible elements. Geo-chemical studies from most arc magmas show that this is rarely the case, and while they do not negate the possibility that fractional crystallization played an important role, they also indicate that other processes are responsible for these variations, such as source heterogeneities, magma mixing or crustal assimilation.

Arc Magmas on Discrimination Diagrams

29 Since the early 1970s, a plethora of geochemical discrimination diagrams have been published utilizing the varying behaviours of compatible and incompatible trace elements to distinguish magmas by type (alkalic, tholeiitic, calc-alkalic) and by tectonic setting (arc, within plate, etc), and also to constrain the chemistry of the magma source. In this approach, incompatible elements are subdivided on the basis of their ionic potential (charge/ionic radius) into high-field-strength elements and large ion lithophile (LIL) elements. High-field-strength elements are relatively small incompatible elements that have a high charge density (e.g. Ti4+, Zr5+) and high ionic potential; other HFS elements include Th, U, Hf, Nb, Y, Ta and REE. They are typically insoluble in H2O-rich solutions. Large ion lithophile elements, however, have low charge and low ionic potential and are soluble in water-rich solutions (e.g. K, Rb, Cs, Ba, Sr). As H2O plays a key role in the genesis of arc magmas, the contrasting behaviour of LIL and HFS is crucial to understanding their petro-genesis and many arc magmas have high LIL/HFS elemental ratios (Pearce 1996).

30 Although the foundation for many discrimination diagrams (e.g. Figs 8-10) is basically empirical (they are based on data from volcanic rocks in known tectonic settings), and the reasons for these variations are not fully understood, they do provide some general constraints for the origin of arc basalts.

Some Limitations to the use of Discrimination Diagrams

40 While these diagrams have been a valuable tool in fingerprinting magma type and tectonic settings, there are some limitations and inconsistencies in their application. First, these diagrams assume that the abundances of these elements in a given rock sample are essentially unchanged since the time of crystallization. This assumption generally holds because HFS elements are typically insoluble in water-rich solutions and are most concentrated in stable accessory phases such as zircon, ilmenite and titanite. However, dissolution of these phases can occur under certain conditions (e.g. rocks subjected to hydrothermal alteration or carbonatization (Hynes 1980)) and samples from rocks so affected yield spurious results.

41 Second, there are some inconsistencies among the diagrams themselves. For example, on the Zr/Y vs Zr plot, within plate basalts (e.g. continental rift-related basalts) are classified by Zr/Y > 4. If one plots this line on the Zr-Ti-Y discrimination diagram, it is evident that some rocks with Zr/Y > 4 plot in the IAT, CAB or MORB fields (Fig. 8b, c). It is important to remember that these diagrams are projections that look at a very restricted compositional plane and so a variety of diagrams must be employed to be sure of consistency in interpretation.

Display large image of Figure 10

Display large image of Figure 11

Spider Diagrams

44 Geochemical data are also presented on normalized multi-element diagrams, in which the abundances of a range of elements are compared with a reference source, known as spider diagrams (Figs. 14 a, b). Many spidergrams compare the trace element content of samples with those of a number of diverse reservoirs that could represent potential sources, such as MORB, depleted mantle (DM, the mantle from which basalt has been extracted), and enriched mantle (EM, mantle that has been extensively metasomatized). The elements are ordered with increasing compatibility from left to right. Spider diagrams are analyzed in a similar fashion to chondrite-normalized diagrams. The data are normalized relative to a candidate reservoir. If basaltic suites are being studied, typically they are either compared with MORB or with a mantle reservoir (e.g. DM or EM). An envelope around the data is compared with that of the candidate mantle reservoir and yields valuable evidence about the source of the magma. Variations within the data set provide information about the extent of fractionation. Fractionated magmas derived from the mantle reservoir, for example, ideally should yield a set of parallel profiles within the envelope.

Display large image of Figure 13

Display large image of Figure 14

45 Mafic to intermediate arc rocks typically exhibit jagged MORB-normalized patterns, and most notably, a negative Nb anomaly relative to Th or La (Fig. 14a, e.g. see Sun and McDonough 1989). In contrast, mafic rocks from non-arc settings are characterized by relatively smooth MORB-normalized patterns (especially ocean-island basalts) and some have positive Nb anomalies. Jagged patterns are also evident in mantle normalized plots for mafic and felsic rocks generated in arc environments (Fig. 14b). In essence, the jagged patterns show the same LIL element enrichment relative to HFS elements (see Nb, Ti, Zr, and Ta), which is evident in several discrimination diagrams. Although these same traits can be recognized on standard discrimination diagrams, the advantage of these plots is that trends can be shown and compared for many elements on the same diagram. Although mafic magmas in both arc and non-arc settings are derived from the mantle, these diagrams also identify the distinctive high LIL/HF S ratio of arc magmas. When the metasomatized mantle wedge melts, LIL elements strongly partition into the melt, whereas HFS elements preferentially remain in stable accessory minerals. Melts derived from this metasomatized mantle wedge are therefore enriched in LIL relative to HFS elements (e.g. Miškovic and Francis 2006). Tatsumi and Kogiso (2003) have used this approach to propose that the dehydration of subducted sediments and oceanic crust released water and LIL elements into the mantle wedge. This hypothesis is supported by boron and beryllium studies. Boron and 10Be are enriched in sediments, but escape very easily upon entering the subduction zone. Many modern arc volcanic rocks have elevated concentrations in these elements, which can only be derived from subduction of recent sediments (e.g. Morris et al. 1990; Morris and Ryan 2004).

46 The origin of the jagged mantle-normalized patterns in felsic magmas could arise in several ways. As fractionation generates a series of parallel trends, these patterns could be derived from fractionation of a more mafic parent. Alternatively, these features can be inherited from mafic magma derived from metasomatized mantle, and the differentiated products from such magma reproduce the jagged pattern. Supporting evidence for this hypothesis is gained from the analyses of xenoliths (retrieved from kimberlite bodies) that are derived from the mantle beneath some continental arcs. These data indicate that the mantle source may have suffered extensive and long-term metasomatism. Alternatively, as continental crust is primarily formed by earlier arc magmatism, they may reflect partial melting of continental crust that was mostly generated by earlier arc activity. In this case, the jagged pattern is inherited from the crustal source, rather than a reflection of coeval subduction.

47 Intermediate to felsic magmas in arc settings are also depleted in HFS elements relative to felsic magmas from within-plate settings. However, there is considerable debate about the origin of intermediate to felsic magmas in arc settings, and their origin may vary from one arc to another (Gill 1981; Carmichael 2002). As these magmas tend to be most dominant in mature arcs with thick continental crust, such as the Andes, much discussion focuses on whether the thickened crust allows more scope for fractionation, or whether the crust supplies a direct chemical contribution (see Carmichael 2002, and references therein). In some instances it is argued that this signature is inherited from basalts by fractional crystallization. In others, the felsic composition is attributed to partial melting of the crust or contamination of mafic to intermediate lavas by the crust. Since this crust is depleted in HFS elements, it is difficult to distinguish between these hypotheses using trace-element geochemistry alone. Given that the average continental crust is very similar in composition to calc-alkaline andesites (e.g. Taylor 1995), resolution of the problem has fundamental implications for understanding the evolution of the crust. The following section discusses how radiogenic isotopes, particularly Sm-Nd isotopic studies, can help distinguish among these various possibilities.

ISOTOPIC CHARACTERISTICS

48 In addition to their use in dating rocks and minerals, certain types of isotopes can be used as tracers to yield information about the original source material from which a volcanic rock was derived. Several isotopic methods are used, including Rb-Sr, and Sm-Nd and U-Pb. Rubidium-Sr and Sm-Nd are commonly considered together.

OTHER ARC ROCKS

73 Not all magmas generated in arc environments fit neatly into calc-alkalic, tholeiitic or alkalic classifications. Shoshonites, boninites and adakites are volumetrically small, but are nonetheless petrologically significant igneous rocks that form in specialized arc environments. Their occurrence reveals much about the thermal structure of the arc and the potential effects of some complications induced by local tectonic activity.

74 Shoshonite is an alkali-rich (Na2O+K2O > 6.5 wt %) trachyandesite that occurs in continental rift, collisional and arc settings (Rogers and Setterfield 1994). Shoshonite occurs in association with calc–alkaline volcanism in intra-arc rift settings, in intraoceanic island arcs and backarc basins, (Sun and Stern 2001; Adams et al. 2005) and continental arcs (Conrey et al. 1997). In oceanic arcs, shoshonite magma development is related to a change in subduction processes or arcrifting (Rogers and Setterfield 1994). In continental arcs, shoshonite generally occurs a significant distance from the trench where the subducted slab is at a great depth and is commonly associated with rifting of the arc or backarc (Gill 1981; Baluyev et al. 1988). Shoshonites in all environments are markedly enriched in LIL (100-1000× chondrite), including LREE. The high Ce/Yb is attributed to LREE enrichment, and the role of residual garnet in the source. The subduction component in shoshonites is also indicated by pronounced negative Nb, Ti and Ta anomalies (e.g. Turner et al. 1996).

75 Boninites (named after the Bonin Islands, south of Japan, where they are most abundant, Cameron et al. 1979, 1983) are high MgO (> 6 wt% MgO) andesites that predominantly occur in the early stages of oceanic arc development in fore-arc regions, an unusual locality for magma generation. Compared to typical andesites, they have higher Cr, Ni and LIL elements but lower Ti and Nb (Arndt 2003). They are generally thought to represent melting of a hydrated and refractory Mg-rich mantle wedge that was contaminated with LIL elements transported from the subduction zone. It is clear that to generate boninitic magma, a steep geothermal gradient is needed to melt depleted peridotite. Such environments can occur in some extensional environments within the arc, either at shallow levels in the fore-arc, or where a backarc spreading centre propagates into an active arc. Published models for the origin of boninites include 1) the initiation of subduction (Stern and Bloomer 1992), 2) intra-arc rifting (Crawford et al. 1981; 3) ridge subduction (Cameron 1989), 4) the effect of seafloor spreading in a forearc basin (Bedard et al. 1998) or 5) where an active arc undergoes backarc spreading (e.g. Falloon and Crawford 1991; Monzier et al. 1993).

76 Adakites (Kay 1978; Defant and Drummond 1990; Defant and Kepezhinskas 2001; Grove et al. 2005), named after Adak Island in the Aleutians, are also high-Mg andesites, but unlike boninites, they are very LREE enriched (normalized La/Yb > 40), contain high Sr (> 400 ppm), high Sr/Y values, low initial 87Sr/86Sr and low ratios of radiogenic to non-radiogenic Pb. Kay (1978) attributed adakites to the interaction of a LIL element-rich hydrous melt from the subducted oceanic crust with overlying mantle, and then eruption without interaction with the island arc crust. The question then becomes: under what conditions does the slab become hot enough to melt? Models favouring the subduction of young crust (e.g. Green and Harry 1999) are supported by experiments where adakites are produced by reaction of slab melts with peridotite in the mantle wedge (e.g. Rapp et al. 1999). The thermal structure of subduction zones (e.g. Peacock et al. 1994) predicts that only very young oceanic crust (<5 Ma) can melt. This prediction is at odds with geologic data (e.g. Defant and Drummond 1990) which indicate that slab melting occurs in crust as old as 20 Ma. This discrepancy is attributed to a higher contribution from shear heating in subduction zones than is accounted for in the Peacock et al. model (Green and Harry 1999). Dickinson and Snyder (1979) and Thorkelson (1996) show that in a ridge-trench collisional environment, a "window" in the slab forms between the downgoing plates that is filled with upwelling asthenosphere. Such an environment can induce slab melting along the thinned margins of the window (Johnston and Thorkelson 1997; Thorkelson and Breitsprecher 2005). Defant and Kepezhinskas (2001) propose that slab melting and adakite production can also occur near tears in the subducted slab, which also facilitates asthenospheric upwelling. It is clear that slab melting may occur in several environments, but the common denominator appears to be anomalous local supply of heat.

77 Adakites have recently been recognized in many other localities in the modern and ancient record. Indeed, some authors use them as an analogy for generating Archean magmas, in which it is viewed that the more elevated geothermal gradient might result in more common melting of the slab (Martin et al. 2005). Archean complexes are dominated by trondhjemite, tonalite and granodiorite (the so called TTG suite, Martin 1987). During the Archean, it is generally believed that the geothermal gradient was higher implying a hotter upper mantle than today (e.g. de Wit and Hynes 1995) which could promote slab melting (e.g. Martin 1987; Defant and Drummond 1990). Experiments suggest that slabs under these conditions can melt, and if that melt interacts with the mantle, it can produce intermediate compositions similar to those of Archean complexes (Rapp et al. 1999; Tatsumi 2000).

DISCUSSION

78 There are several possible sources for arc magmas in the vicinity of the subduction zone, including the subducted oceanic crust and sediments, as well as the mantle wedge and crust above the subduction zone.

79 To a greater or lesser extent, a chemical signal from all these sources can be found in arc magmas. However, when the thermal conditions in the vicinity of the subduction zone are considered in conjunction with phase equilibria, it is clear that arc magmas are most commonly generated in the mantle wedge and crust above the subduction zone. The thermal regime at subduction zones is profoundly influenced by the rate of descent of cold oceanic lithosphere, which is generally too fast to allow significant transfer of heat from the surrounding mantle. As a result, the isotherms are deflected downwards, and this cooling leads to a low geothermal gradient (10-15°C per km) and low heat flow in the fore-arc region above the subducted slab. As it descends, the slab is slowly warmed up, so that sediments, oceanic crust and lithosphere undergo prograde blueschist-eclogite facies metamorphism resulting in dehydration of the slab.

80 Although magma is not produced, the subducted oceanic litho-sphere does play a vital role in the origin of arc magmas because the meta-morphic fluids are released and migrate upward to invade and chemically modify the overlying mantle wedge. Of all the magmas formed in a subduction zone environment, IATs show the least contamination from subduction zone processes, an interpretation consistent with their occurrence in immature arcs. In more mature arcs, the effect of the fluids is more profound. The hydrated mantle melts at a significantly lower temperature than dry mantle and the magmas produced are oxidized and hydrated, inducing early crystallization of minerals such as magnetite and hornblende, which inhibit the rise in Fe/Mg ratio and so produce a calc-alkalic trend.

81 These oxidizing fluids tend to be enriched in water-soluble LIL elements. They also may stabilize HFS-bearing accessory minerals so that magmas derived from the mantle wedge tend to be enriched in LIL but depleted in HFS elements, patterns that are clearly evident in MORB-normalized or mantle-normalized spider-gram plots in which HFS depletion is identified by negative Nb and Ti anomalies.

82 Except in rare cases, the subducted oceanic lithosphere does not attain temperatures that are sufficient for it to melt, so that the slab itself cannot be the main source of arc magmas. Relatively flat HREE profiles for most arc magmas are consistent with a derivation from the spinel lherzolite portion (< 50 km depth) of the mantle. For similar reasons, the asthenospheric mantle below the slab does not warm up appreciably and so cannot be an important source of magma. Adakites, however, which do show significant HREE depletion, may well be representative of slab melts, but they are rare, and their formation requires anomalous local supplies of heat within the subduction regime.

83 The Sr, Nd and Pb isotopic signatures of continental arc magmas commonly have a crustal component. In some arcs, these signatures are imposed by crustal contamination as the magma rises toward the surface. However, other arcs have this signature, even if they are capped by oceanic, rather than continental crust. The only plausible sedimentary source is from the subduction zone itself, supporting the notion that sediments are dragged down the subduction zone, and LIL elements derived from them are introduced into the mantle wedge. This hypothesis is supported by elevated concentrations of boron and beryllium in arc volcanic rocks, which can only be derived from subduction of recent sediments (e.g. Morris et al. 1990).

84 In mature subduction zones, original chemistry of the mantle wedge is progressively blurred by metasomatic fluids from the subducting slab. Depending on previous history, the mantle wedge may have a quite variable initial chemistry, which can influence the compositions of early arc magmas.

85 As arc magmas rise toward the surface, their compositions are strongly modified by interaction with their surroundings, especially during final cooling. Intermediate to silicic arc magmas can form in a variety of ways, including fractional crystallization of a more mafic parent, mixing between mafic and felsic magmas, partial melting of the crust, and assimilation and digestion of the crust by more mafic magma. Where felsic magmas are voluminous relative to intermediate or mafic melts, they are unlikely to be generated by fractional crystallization because there is not enough parent mafic magma to produce the felsic differentiates. Andesites formed either by mixing of basaltic and rhyolitic magmas, or by contamination of a more mafic parent by the continental crust commonly display textural evidence of disequilibrium (Eichelberger 1975) or field evidence of magma mixing or host rock assimilation (Wiebe et al. 2002). All these processes seem to occur in arc magmas to some extent although their relative importance may vary from one suite to another, or even within a given suite.

86 The extent of crustal involvement in the composition of intermediate to felsic magmas can be assessed using isotopes. Samarium-Nd isotopes are particularly robust because they are normally not affected by secondary processes. Magmas related by fractional crystallization should yield similar Sm-Nd isotopic signatures. Those formed by anatexis of older crust should have depleted mantle model ages that are significantly older than their crystallization age and those that reflect contamination or assimilation should plot on calculated mixing curves and should have a range in TDM ages between the end member components (DePaolo 1981a, b; Arndt and Goldstein 1987).

87 Given that 1) intermediate rocks are the dominant product of arc magmatism, and 2) they approximate the average composition of the continental crust, identifying the dominant process that forms intermediate magmas is fundamental to the understanding of crustal evolution.

88 Although hotly debated (e.g. Hamilton 1998; Stern 2005), many geologists contend that arcs have existed at least since the end of the Archean. Much of the petrological evidence is derived from geochemical and isotopic studies, which indicate that Late Archean volcanic and plutonic rocks share many similar geochemical traits with their modern counterparts (see Dostal and Mueller 1997; Corcoran and Dostal 2001; Canil 2004; Cousens et al. 2004; Dostal et al., 2004). If so, then processes like those occurring in modern arcs may well have had an influence on the evolution of continental crust for at least the last 2.5 billion years.

Websites: