Copper Systematics in Arc Magmas and Implications for Crust-Mantle Differentiation

See allHide authors and affiliations

Science  06 Apr 2012:
Vol. 336, Issue 6077, pp. 64-68
DOI: 10.1126/science.1217313


Arc magmas are important building blocks of the continental crust. Because many arc lavas are oxidized, continent formation is thought to be associated with oxidizing conditions. On the basis of copper’s (Cu’s) affinity for reduced sulfur phases, we tracked the redox state of arc magmas from mantle source to emplacement in the crust. Primary arc and mid-ocean ridge basalts have identical Cu contents, indicating that the redox states of primitive arc magmas are indistinguishable from that of mid-ocean ridge basalts. During magmatic differentiation, the Cu content of most arc magmas decreases markedly because of sulfide segregation. Because a similar depletion in Cu characterizes global continental crust, the formation of sulfide-bearing cumulates under reducing conditions may be a critical step in continent formation.

The composition and oxidation state of melts formed in subduction zones, collectively known as arc magmas, influence the formation and evolution of the continents, ore deposits, and possibly even the atmosphere (1). Arc magmas are oxidized relative to the average upper mantle; however, the means by which these lavas become oxidized is debated. The prevailing view is that arc magmas inherit their oxidized states from melting sub-arc mantle contaminated by subducted sediments and oceanic crust (15). Alternatively, primary arc magmas may be less oxidized than their more evolved counterparts if oxidation is caused by magmatic differentiation associated with crystallization and chemical interaction with preexisting crust (69).

To resolve this debate, the redox state of the mantle source regions of arc magmas must be determined. Unfortunately, primary arc magmas recording this signature are rare because by the time they have risen to the surface, they have already differentiated. One approach has been to investigate the redox state of melts trapped in phenocrysts (4), but melt inclusions rarely represent true primary magmas and are found only in extrusive rocks, which may not be representative of the entire arc crust. Another approach is to track the evolution of those metals, whose partitioning behaviors are sensitive to redox during mantle melting but largely redox-insensitive during early magmatic differentiation owing to their incompatibility in early fractionating minerals (6, 7, 10). The relative proportions of such metals then record the original redox conditions in the mantle but do not record redox evolution during differentiation.

One element that may provide insight into the entire redox history of magmas is copper (Cu). In the range of oxygen fugacities (fO2) typical of most igneous rocks, the valence state of Cu remains +1. However, its solid/melt partitioning behavior depends on the speciation of sulfur (S) (11, 12), which itself is redox-sensitive. At fO2 values of ΔFMQ = –1 to 0 (log10 unit deviations from the fayalite-magnetite-quartz buffer), typical of average uppermost mantle, the dominant oxidation state is S2–, which stabilizes sulfides. At high fO2, however, S6+ is stable as sulfate (SO42–) only. In silicate melts, complete conversion from sulfide to sulfate occurs within a narrow range between FMQ+1 and +2, resulting in a 10-fold increase in total S solubility (12). Thus, the redox state of magmas should be reflected in their S contents. However, determining pre-eruptive S contents is hampered by devolatilization of S at low pressures (13) as well as by post-eruptive hydrothermal alteration and weathering. Cu is strongly partitioned into sulfide minerals but is generally insensitive to such post-eruptive processes. Sulfide liquid/silicate melt partition coefficients (Dsf/melt) for Cu are between 600 and 1200 (1416). This makes Cu a potential tracer of pre-eruptive redox state of S, provided Cu does not partition into nonsulfide phases.

To confirm that Cu partitioning into nonsulfide minerals is negligible, we measured the distribution of Cu between olivine phenocrysts and host magma and between minerals in mantle peridotites and garnet pyroxenite cumulates from arc environments (17). Mineral/melt partition coefficients were calculated from measured olivine/melt (Dol/melt) and mineral/olivine partition coefficients (Fig. 1A, table S2, and fig. S1). Cu is found to be incompatible in olivine (Dol/melt = 0.05), orthopyroxene (0.035), clinopyroxene (0.04), amphibole (0.05), garnet (0.004), and spinel (0.2), indicating that in the presence of sulfide, the influence of these minerals on bulk partitioning of Cu is negligible (Fig. 1A). The bulk partitioning of Cu during peridotite melting Dper/melt therefore depends on how sulfide mode varies with melt extraction. The ubiquity of sulfide-bearing peridotite samples (18) indicates that the mantle is generally sulfide-saturated, and detailed studies of these peridotites suggest that typical fertile mantle contains ~200 parts per million (ppm) S in the form of monosulfide solid solution (1921). This corresponds to a sulfide mode of 0.06 weight percent (wt %) and a bulk Cu partition coefficient of ~0.5 at the onset of melting—that is, Cu is moderately incompatible, but in the absence of sulfide, Cu would be more incompatible (Fig. 1B). Thus, as melting progressively consumes sulfide, the bulk partition coefficient of Cu decreases (Fig. 2A), and when sulfide is finally exhausted, Cu becomes almost perfectly incompatible. The efficiency of S exhaustion—that is, the degree of melting needed to remove sulfide—depends on the solubility of S in the melt, which increases with decreasing pressure (P), increasing temperature (T), and increasing fO2 (supplementary text) (12, 22, 23).

Fig. 1

(A) Mineral/melt partition coefficients Dmin/melt for clinopyroxene (cpx), orthopyroxene (opx), olivine (ol), and garnet (gt) plotted in order of decreasing relative abundances in MORBs. Elements with D > 1 and D < 1 are compatible and incompatible in the mineral, the latter preferentially partitioned into the melt. D values for Cu are from this study; others are taken from the literature (table S4). Cu does not fit into silicate minerals but is highly compatible in sulfide phases; value of ~800 represents partitioning of Cu between sulfide liquid and silicate liquid. (B) Bulk partition coefficients for peridotite (62% ol, 20% opx, and 17% cpx), clinopyroxenite, and garnet pyroxenite (30% gt and 70% cpx) containing sulfide. Sulfide mode (0.06 wt %) in peridotite is equivalent to 200 ppm S assumed for primitive mantle (21). Sulfide mode in pyroxenites (0.36 wt %) is defined by the solubility of sulfide in basaltic magmas undergoing fractional crystallization. (C) Primitive-mantle (19) normalized elemental abundances in basalts, based on globally averaged MORBs (for these elements, island arc basalts are similar) (39), and bulk, upper, and lower continental crust (BCC, UCC, and LCC) (28). Cu and Sc become compatible during continental crust evolution but are incompatible during mantle melting to form basalts. (D) Predicted residual melt compositions resulting from 50% fractional crystallization of basalt by pyroxenite and garnet pyroxenite cumulates, showing coupling of Cu and Sc depletion.

Fig. 2

Mantle melting models. (A) Variation of aggregated melt and residual mantle composition with degree of melting (F) at 2 GPa, 1350°C and fO2 at FMQ+0. Initial S of 200 ppm and sulfide mode of 0.06 wt % are assumed. Melting depletes S (and sulfide) in the peridotite residue. Cu is strongly compatible in sulfide and thus moderately incompatible during early melting of peridotite, when only a small decrease in bulk Cu content in the residue occurs. After sulfide is exhausted (F = 0.2), Cu becomes highly incompatible, and its behavior is defined solely by silicate-melt partitioning. Cu content in aggregate liquid is low during initial melting and reaches a maximum at F = 0.2 when sulfide is exhausted. At F > 0.2, Cu decreases because of dilution. Cu content of instantaneous liquids is shown for reference. (B) Primary melt Cu content as a function of F and contoured against log fO2 deviations from FMQ buffer (ΔFMQ); same P-T conditions as in (A). White squares represent MORBs from the East Pacific Rise, Indian Ridge, and Mid-Atlantic Ridge; black circles represent arc basalts from the Aleutians, Marianas, Mexico, Cascades, and Andes based on extrapolation of Cu-Mg/(Mg+Fe) trends to Mg/(Mg+Fe) = 0.72. Gray field is possible range of F in arc basalts (40). (C) Cu content of aggregate liquids versus fO2s for different F. Gray field refers to primitive MORBs and arc Cu from (B) and Fig. 3.

The above concepts can be quantified by modeling fractional melting of a fertile mantle composition [30 ppm Cu and 200 ppm S (19)] to generate a series of peridotite residues and instantaneous fractional melts, which are then pooled (Fig. 2). Melts are initially enriched in Cu because Dper/melt < 1, but as melting degree F increases, sulfide mode decreases, causing Dper/melt to decrease even further so that Cu is more efficiently transferred into the melt. A maximum Cu content in the melt is reached at F ~ 0.2, when all sulfide is exhausted and the system becomes sulfide-undersaturated. At higher fO2, sulfide is exhausted at lower F, resulting in more efficient transfer of Cu into the melt (Fig. 2A, inset). For example, maximum Cu contents in melts exceed 100 ppm for FMQ+1 but are below 100 ppm for fO2 values less than FMQ (Fig. 2A, inset). As shown in Fig. 2B, primitive (MgO > 8 wt %) mid-ocean ridge basalts (MORBs) have 60 to 70 ppm Cu and therefore cannot be explained by melting at fO2 > FMQ (Fig. 2B). This is consistent with the generally accepted fO2 in MORBs (4). More surprising is that the Cu contents of primitive arc magmas fall between 50 and 90 ppm (Figs. 2B and 3), which is indistinguishable from MORBs. This overlap suggests that the fO2 and Cu content of sub-arc mantle must be similar to that of MORB mantle; if arc mantle had fO2s > FMQ+1 or Cu > 40 ppm (the latter requiring enrichment by Cu-rich fluids, which in turn require high fO2), magmatic Cu contents >200 ppm would be predicted, which is not observed. Our results are consistent with the study of Jenner et al. (8) on an individual volcanic center in the Marianas arc and require that sub-arc mantle is not highly oxidized (≤FMQ+1).

Fig. 3

Covariation of Cu and MgO in (A) MORB, (B and C) island arcs, and (D to F) continental arcs. Horizontal band represents average range of MORB Cu contents. Data sources are from GEOROC and RidgePetDB (supplementary text). Curves in (A) to (F) represent differentiation along fO2 evolution paths from FMQ = 0 to the value denoted. Curves are defined by a tholeiitic fractionation series at 0.1 GPa involving olivine and plagioclase in (A) to (C) and by a calc-alkaline differentiation series at 0.8 GPa involving olivine and clinopyroxene in (D) to (F) (supplementary text). Initial starting composition is that of a primitive basalt generated at FMQ+0. Tick marks on right side of each panel show Cu in primary magma as a function of fO2 (Fig. 2, B and C) for an average F = 0.1. Star indicates average composition of the BCC (28).

Only during magmatic differentiation does the partitioning of Cu diverge (Fig. 3). In most arcs, Cu becomes compatible during differentiation, as evidenced by monotonic decrease in Cu with decreasing MgO, an index that correlates with progressive fractional crystallization. However, in some arc magmas (mostly in island arcs) Cu initially behaves incompatibly and then increases with decreasing MgO until a maximum is reached, after which Cu becomes compatible and plummets as MgO decreases further. A quantitative understanding of these differentiation trends can be achieved by modeling the behavior of S during fractional crystallization at various fO2 (Figs. 3 and 4). In these calculations, early crystallization sequences specific for island arcs and continental arcs were adopted (only magmas with MgO > 4 wt % were considered) in order to match major element systematics (supplementary text). Cu depletion in the magma requires sulfide segregation (probably as sulfide droplets included in fractionating silicate minerals), whereas Cu enrichment requires suppression of sulfide crystallization, which can be achieved, as mentioned above, through decompression and fO2 increase. The divergence of Cu in arc magmas suggests different fO2 trends during differentiation, but the fact that all arc magmatic suites eventually evolve toward highly depleted Cu contents implies that all melts ultimately reach sulfide saturation. This limits fO2s during early magmatic differentiation to less than FMQ+1.3.

Fig. 4

Plutonic rocks and cumulates from Cretaceous Cordilleran batholiths in western North America. (A) Cu versus MgO in plutonic rocks from the Peninsular Ranges batholith in southern California (27). Curves represent evolution of mantle-derived basaltic magma during fractional crystallization of olivine-pyroxene cumulates (calc-alkaline trend) in the crust for different fO2 values (Fig. 3, caption). (B) Cu versus MgO in garnet pyroxenite cumulate xenoliths from the Sierra Nevada batholith, one segment of the Cordilleran batholith. Curves represent compositions of instantaneous cumulates during fractional crystallization. These are the complements of the residual melt curves in (A). (C) Photomicrographs of sulfide (sf; pyrrhotite) inclusion within clinopyroxene (cpx) in the Sierran garnet pyroxenite cumulates. (Top) Reflected light. (Bottom) Unpolarized transmitted light. Star indicates BCC (28).

The above explanation for the depletion of Cu in evolved arc magmas predicts the formation of Cu-rich cumulates in the lower crust or lithospheric mantle. To test this prediction, we examined pyroxenite xenoliths hosted in Late Miocene basaltic volcanoes in eastern California. These pyroxenites are cumulates associated with the formation of the Cretaceous Sierra Nevada continental arc in western North America (2427). The Cu content of the pyroxenites is highly scattered, which is expected for cumulates, particularly in continental arcs, where magma intrusion and crystallization occur through multiple events. The majority of pyroxenites is enriched in Cu compared with the primitive mantle, many containing more Cu (<400 ppm) than that of primitive basalts (70 ppm) (Fig. 4B). Both the low and high Cu abundances are consistent with our model calculations of cumulate compositions. Samples with the highest Cu contain Cu-rich sulfides (pyrrhotite and chalcopyrite) included within magmatic clinopyroxenes, which is evidence for sulfide+pyroxene co-precipitation (Fig. 4C and table S1). On the basis of the Cu content of the pyrrhotites (1 to 5 wt %) (table S3), we infer a sulfide fraction of ~1 wt % and a whole-rock S content of ~3000 ppm for the most Cu-rich pyroxenites.

The generality of the above-described behavior of Cu and S can be assessed by examining the bulk Cu systematics of the time-integrated average global continental crust. Globally averaged continental crust has ~30 ppm Cu (28), more than a factor of two lower than primary basaltic magmas in mid-ocean ridge and arc environments (Figs. 1C, 2, and 3A). Plotting the relative abundances of transition metals in order of compatibility during mantle melting shows that the continental crust is marked by a strong negative Cu anomaly relative to possible parental magmas, such as primitive island arc basalts or MORBs (Fig. 1C). This Cu depletion appears to be associated with depletions in scandium (Sc), chromium (Cr), and nickel (Ni). Although Ni and Cr depletion may be attributed to olivine and spinel fractionation, respectively, only clinopyroxene (and possibly amphibole) removal can simultaneously deplete Sc, Cr, and Ni (Fig. 1D). Coupled Cu and Sc depletions in global continental crust thus suggest that sulfide-pyroxenite fractionation is an important step in the evolution and formation of continents as well as arc magmas. Mass balance indicates that 70 to 80% and 20 to 30% of the original Cu and S in parental arc basalts are sequestered in such cumulates (supplementary text and fig. S3).

Our findings thus suggest that oxidizing conditions are not required to generate arc magmas or continents in general. One important implication is that the mantle source regions of arc magmas are not anomalously enriched in Cu. This may seem surprising because there is a well-known association of Cu-porphyry ore deposits with arcs (5, 29). Cu-porphyries are dominantly associated with continental arcs and trace-element signatures [yttrium (Y) depletions] indicative of residual garnet (30), leading to the view that porphyry Cu deposits derive from mantle sources enriched in Cu via highly oxidized melting of subducted oceanic crust enriched in Cu and S during hydrothermal alteration beneath the sea floor (5, 31). If, however, the mantle source is not enriched in Cu as suggested here, alternative explanations are required. Pyroxenite cumulates in the deep roots of arcs may be the only enriched source of Cu, raising the question of whether high-degree remelting of such pyroxenites can liberate, by sulfide exhaustion, the Cu necessary for forming ore deposits. Such conditions should occur when magmatic or tectonic thickening displaces cumulates to greater depths, where they are more prone to heating by the surrounding asthenospheric mantle and the passage of hot magmas. Melting at these pressures would generate peritectic garnet (32), resulting in magmas depleted in garnet- and clinopyroxene-compatible elements, such as Y and Sc, respectively. Although the formation of Cu-porphyry deposits still requires unusual concentration processes at shallow levels (33), remelting of pyroxenite cumulates in thickened arc roots may be a necessary step. Such a process may explain why porphyry Cu deposits are mainly found in thick, mature continental arcs and less in island arcs or incipient continental arcs (34, 35).

What is the eventual fate of Cu-rich pyroxenite cumulates beneath arcs? Pyroxenites, especially if they have undergone metamorphic transformation into garnet pyroxenites, are denser than underlying peridotitic mantle and therefore thought to eventually founder en masse into the convecting mantle (36). Indeed, foundering of mafic lower crust is thought to be an important step in driving the bulk composition of continental crust toward silicic compositions (37). Foundering of deep arc roots may thus result in permanent removal of Cu and other sulfide-compatible metals, such as lead (Pb), from continents. This implies that formation of Cu-porphyries may be possible only in a narrow window within the history of a continental arc (fig. S4) (38). Last, because Pb is the daughter product of uranium, which itself is not partitioned into sulfide-bearing pyroxenites, foundering of these pyroxenites through Earth’s history could have generated an unradiogenic Pb isotope reservoir deep in the mantle, leaving behind a continental crust and upper mantle characterized by radiogenic Pb isotopes.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S4

Tables S1 to S4

References (4167)

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online
  2. Acknowledgments: This work was funded by an NSF grant to C.-T.A.L. Discussions with F. Albarede, J. Blichert-Toft, S. Huang, D. Anderson, R, Rudnick, W. McDonough, and R. Arevalo are appreciated. D.J.’s participation in this project was made possible by F. Steinkamp’s Scientific Research and Design course at William P. Clements High School in Sugar Land, TX. All data are available in the supplementary materials. C.-T.A.L. planned the research and wrote the manuscript. E.J.C. collected the whole-rock data for Sierran garnet pyroxenites. R.B. collected the olivine phenocryst-groundmass trace-element data. C.-T.A.L. and P.L. collected the mineral/mineral trace-element partition coefficients in peridotites, pyroxenites, and sulfides. D.J. developed the ideas behind Fig. 1C. All authors participated in interpretation of the data and modeling.
View Abstract

Navigate This Article