Geochemistry, Geophysics, Geosystems

Ultrahigh-pressure metabasaltic garnets as probes into deep subduction zone chemical cycling


  • Robert L. King,

    1. Department of Earth and Environmental Sciences, Lehigh University, Bethlehem, Pennsylvania, USA
    2. Pheasant Memorial Laboratory for Geochemistry and Cosmochemistry, Institute for Study of the Earth's Interior, Okayama University at Misasa, Tottori-ken, Japan
    Search for more papers by this author
  • Gray E. Bebout,

    1. Department of Earth and Environmental Sciences, Lehigh University, Bethlehem, Pennsylvania, USA
    2. Pheasant Memorial Laboratory for Geochemistry and Cosmochemistry, Institute for Study of the Earth's Interior, Okayama University at Misasa, Tottori-ken, Japan
    Search for more papers by this author
  • Katsura Kobayashi,

    1. Pheasant Memorial Laboratory for Geochemistry and Cosmochemistry, Institute for Study of the Earth's Interior, Okayama University at Misasa, Tottori-ken, Japan
    Search for more papers by this author
  • Eizo Nakamura,

    1. Pheasant Memorial Laboratory for Geochemistry and Cosmochemistry, Institute for Study of the Earth's Interior, Okayama University at Misasa, Tottori-ken, Japan
    Search for more papers by this author
  • Sebastiaan N. G. C. van der Klauw

    1. Jena, Germany
    Search for more papers by this author


[1] We demonstrate an approach to examining the metamorphic history of subducting oceanic crust that can complement records of subduction zone chemical cycling derived from studies of igneous rocks produced at volcanic arcs. By merging methods utilizing garnet zoning to establish prograde reaction histories with in situ high-resolution trace element geochemistry, and application to coesite-bearing mafic eclogites representing subduction to depths beneath arcs, we are able to directly identify geochemical manifestations of reactions contributing to element mobility in the subducting slab that are only inferred in studies of volcanic arcs or theoretical metamorphic models. Specifically, we identify a prograde metamorphic reaction, based solely on the zoning of geochemistry and mineral inclusions within garnet, and infer that these features are a record of the breakdown of coexisting clinozoisite + titanite and probable liberation of trace element–laden fluid from the rock during prograde metamorphism. We are then able to assign a specific depth interval for the reaction through calculation of the P-T dependence of the reaction for these eclogites and comparison with a published P-T trajectory. Because of the robust preservation of records of petrologic and geochemical processes by garnet, this methodology is particularly suited for study of ultrahigh-pressure (UHP) eclogites, in which severe retrograde alteration (generally related to exhumation) commonly obscures prograde history.

1. Introduction

[2] The distinctive geochemical composition of volcanic rocks produced at convergent margins (subduction zones) has led to a standard petrogenetic model for such “arc volcanics” that requires the addition, to the mantle, of a trace element-bearing hydrous fluid derived from metamorphic reactions within subducting materials. Experimentally based models of devolatilization within subducted oceanic crust and sediments indicate that both continuous and discontinuous reactions are important for mobilization of dominantly hydrous fluids, and that these reactions persist to 150–200 km within subduction zones [Schmidt and Poli, 1998, 2003]. However, a continuous mobilization of trace elements linked to this continuous dehydration is not realistic in that diffusional processes limit communication of trace elements hosted by solid phases not actively participating in devolatilizing reactions to mobilized fluid phases [Schmidt and Poli, 2003]. This reasoning led Schmidt and Poli [2003] to argue that trace element mobilization in fluids is probably controlled in pulses related to decomposition of host phases. Thus knowledge of the reaction histories of (generally accessory) host phases that partition trace elements in subducting oceanic crust and sediments [e.g., Sorensen and Grossman, 1993; Tribuzio et al., 1996; Sassi et al., 2000; Hermann, 2002; Spandler et al., 2003] is of critical importance in attempts to characterize mass transfer from the subducting slab to the volcanic arc. However, few connections have been made between the indirect geochemical data derived from study of igneous rocks formed along volcanic arcs and specific reactions involving trace element hosts.

[3] Ultrahigh-pressure (UHP) metamorphic suites, the direct products of the subduction of oceanic crust and sediment to sub-arc depths of approximately 70–120 km [Gill, 1981; Jarrard, 1986] and beyond, should provide a means of extracting information regarding the geochemical evolution of the subducting section. Unfortunately, however, UHP metamorphic rocks are notorious for their superposition of retrograde mineral assemblages (related to exhumation) over mineral assemblages and geochemistry reflecting prograde underthrusting history. Although traditional whole rock geochemical methods can be informative, their inability to discriminate prograde versus retrograde signals in bulk geochemistry ultimately results in a data set burdened by caveats, assumptions, and the purely unknown. In contrast, extraction of a prograde chemical record from “robust” phases such as garnet, tourmaline, and zircon could be productive because of the remarkable ability of these phases to preserve prograde metamorphic chemical and mineralogic records as growth zoning and mineral inclusions despite overprinting of the rock by later metamorphic history [e.g., Thompson et al., 1977; Kohn, 2003; Bebout and Nakamura, 2003]. Here we demonstrate the rich potential of UHP garnet as a recorder of prograde metamorphic reactions potentially mobilizing hydrous fluids within subducting oceanic crust. In this contribution, we recover such reaction histories through the combination of established geothermobarometric methods with trace element geochemical data obtained by secondary-ion mass spectrometry (SIMS; ion microprobe methods affording 10–15 μm spatial resolution) for UHP-metamorphosed metabasaltic rocks from the Lago di Cignana locality, western Italian Alps [van der Klauw et al., 1997; Reinecke, 1998] that are spectacular in their relatively low degrees of exhumation-related overprinting.

2. Samples and Analytical Methods

[4] We present data for a mafic eclogite collected from the Lago di Cignana UHP metamorphic locality (sample CIG 91-1 of van der Klauw [1998]) [see also van der Klauw et al., 1997; Reinecke, 1998]. Mineral major element compositions and elemental X-ray maps of garnet porphyroblasts in this sample were obtained using the Lehigh University JEOL 733 and the Okayama University JEOL 8800 electron microprobes. Quantitative point analyses were made using wavelength-dispersive spectrometry, standard ZAF corrections, and beam conditions were a 15 keV accelerating voltage, 40 nA current on the Faraday cup, and counting times of 40 seconds on peak and 20 seconds each on two backgrounds. Well-characterized natural minerals were used as standards. Analytical errors (relative standard deviation, percent [RSD%]; 1σ) were 1–2% or better. Elemental X-ray maps within garnet were produced using a 15 kV, 1–1.5 μA beam, pixel sizes of 5 μm, and dwell times of 50 ms; mapping methodologies followed the techniques of Pyle and Spear [1999].

[5] Trace element compositions were measured using the Cameca ims5f secondary-ion mass spectrometer of the Pheasant Memorial Laboratory, Japan. Analytical methods were similar to those used by Nakamura and Kushiro [1998]. A primary O beam of 10 nA intensity was used to sputter samples; spot sizes were approximately 10 μm. Secondary ions were collected using a 4500 V accelerating voltage and measured by ion counting. A −45 eV energy offset was used to reduce mass interferences from polyatomic and oxide isobars; oxide isobars on Eu, Gd, and Dy were corrected as by Nakamura and Kushiro [1998]. Analytical errors (RSD%) were between 5–10% (1σ). Well-characterized natural minerals, glasses, and experimental products were used as standards.

3. Evidence for Reactions During Subduction

3.1. Petrography

[6] Primary petrographic evidence suggests that the Cignana eclogites preserve important geochemical evidence for devolatilization related to prograde subduction history. The metamorphic mineral assemblage representing equilibration at peak pressures and temperatures (2.7–2.9 GPa, ∼625°C [Reinecke, 1998]) is omphacite + garnet + rutile + zircon + phengite + apatite ± coesite ± clinozoisite ± glaucophane. Retrogradation in these eclogites is manifested as zoned amphiboles (barroisite-actinolite-hornblende rims on glaucophane and garnet), chlorite, albite, and a Na-biotite [van der Klauw, 1998; van der Klauw et al., 1997]; similar retrograde assemblages are reported by Bucher et al. [2003] for metabasites of the western Alps. Our exclusive use of high-resolution analytical techniques (SIMS methods) allows us to avoid and minimize any influence of retrogradation on peak mineral chemistries.

[7] We present data for two garnets ∼2 cm from each other in one mafic eclogite sample (CIG 91-1) showing relatively minor (∼15 vol.%) retrogradation and preserving coesite within the omphacite rock matrix (Figures 1a and 1b). Coesite, a high-pressure polymorph of common quartz, is a characteristic mineral of UHP metamorphism, providing a “smoking gun” indicator of subduction to depths of at least those corresponding to beneath arcs [e.g., Chopin, 1984]. The presence of coesite within matrix clinopyroxene (Figures 1a and 1b) provides a reliable indicator that mineral assemblages reflecting UHP conditions are preserved in the rock matrix, despite localized overprinting by exhumation-related assemblages. Occurrences of coesite in this sample as inclusions near rims in garnet porphyroblasts are indicative that at least some of the later garnet growth occurred within the coesite stability field [van der Klauw, 1998; van der Klauw et al., 1997] (see Figure 1c). Coesite is in general uncommon in the metabasaltic eclogites from Lago di Cignana in that they appear to be marginally SiO2-undersaturated, precluding the widespread stability of a free SiO2 phase such as quartz or coesite throughout the eclogites. However, coesite is abundant within more Si- and Mn-rich metasediments of the Lago di Cignana locality (as mineral inclusions in garnet, tourmaline, dolomite, and apatite [Reinecke, 1998]); these metasediments have been inferred to represent seafloor pelagic sediments formed concurrently with the eclogite basaltic protoliths within the Piemontese Ocean and have had identical metamorphic histories as the metabasalts [van der Klauw et al., 1997; Reinecke, 1998].

Figure 1.

Dispositions of coesite in Cignana ultrahigh-pressure metabasalt sample CIG 91-1 of van der Klauw [1998]. (a) A cross-polarized light optical photomicrograph of coesite preserved within matrix omphacite; note 10 μm SIMS points in omphacite near coesite. (b) A Raman spectrum of characteristic coesite absorption bands for coesite grain shown in Figure 1a. (c) A photomicrograph of a coesite + quartz inclusion in another garnet porphyroblast in the same sample but not texturally suitable for the chemical traverses (i.e., this inclusion is not in garnet 3 or garnet 7, for which chemical data are shown in Figures 25). The presence of this coesite as an inclusion in garnet near their rims demonstrates at least some later-stage garnet growth in this rock occurred in the coesite stability field [van der Klauw et al., 1997]. Garnets in some metasedimentary samples from this locality similarly contain coesite inclusions only near their rims, indicative that the later-stage growth of these porphyroblasts occurred in the coesite stability field [Reinecke, 1998].

3.2. Inclusions in Garnet

[8] A core-to-rim progression in mineral inclusions in these garnets, from blueschist-facies to eclogite-facies mineral assemblages, provides a first-order suggestion that the garnets retain a geochemical record of the blueschist- to eclogite-facies transition and any related elemental redistribution (Figures 2 and 3). Within garnet cores, the inclusion assemblage glaucophane + paragonite ± titanite ± clinozoisite is present; all of these minerals are characteristic of blueschist-facies metabasalts. Garnet mantles preserve the higher-grade, apparently less hydrous, eclogite-facies assemblage omphacite + rutile + phengite ± clinozoisite and, rarely, coesite [van der Klauw et al., 1997] (Figure 1c). The mineral reactions representing the blueschist-to-eclogite facies transition are largely dehydration reactions, and this facies transition is thought to mobilize volatiles (and distinctive slab-derived trace element signatures) to volcanic arcs [e.g., Peacock, 1993; Schmidt and Poli, 1998, 2003; Forneris and Holloway, 2003].

Figure 2.

Elemental X-ray maps for Mn, Ca, and Y, and a line drawing of petrographic and chemical relationships within garnet 7 in sample CIG 91-1. Analytical points for SIMS data and major-element calibration of these maps are shown with petrographic relations. For all elemental maps, higher concentrations are represented by warmer colors, low concentrations are represented by cooler colors, and colors are ordered as in the electromagnetic spectrum. Mineral inclusions are denoted as colored shapes given in the legend. The petrographic summary does not include identification of every mineral inclusion to ensure clarity. SIMS points are shown as red for “normal” garnet and blue for garnet of the “reaction zone” illustrated in Figure 4. The quasi-circular pink zone indicates the zone of elevated trace element contents identified as the reaction zone and purposely excludes the high trace element concentration garnet core. Elemental maps of P concentrations (not shown) indicate P is uniformly low in all silicates and that garnets are not zoned in P content, both observations indicating that phosphates were not linked to trace element abundances. See Figure 4 for quantitative chemical data for garnet 7 and text for discussion.

Figure 3.

Elemental X-ray maps for Mn, Ca, and Y, and a line drawing of petrographic and chemical relationships within garnet 3 in sample CIG 91-1. Analytical points for SIMS data and major-element calibration of these maps are shown with petrographic relations. For all elemental maps, higher concentrations are represented by warmer colors, low concentrations are represented by cooler colors, and colors are ordered as in the electromagnetic spectrum. Mineral inclusions are denoted as colored shapes given in the legend. The petrographic summary does not include identification of every mineral inclusion to ensure clarity. SIMS points are shown as red for “normal” garnet and blue for garnet of the “reaction zone” illustrated in Figure 5. The quasi-circular pink zone indicates the zone of elevated trace element contents identified as the reaction zone and purposely excludes the high trace element concentration garnet core. Elemental maps of P concentrations (not shown) indicate P is uniformly low in all silicates and that garnets are not zoned in P content, both observations indicating that phosphates were not linked to trace element abundances. See Figure 5 for quantitative chemical data for garnet 3 and text for discussion.

3.3. Chemical Zoning of Garnet

[9] In addition to line drawings of petrographic relations within garnet, Figures 2 and 3 present elemental X-ray maps for Mn, Ca, and Y contents in garnet. Maps indicate generally declining Mn and Ca contents from garnet cores to rims. Mn “kick-ups” are not present along rims of these garnet grains, indicating little or no influence on rim chemistry due to retrogradation [Kohn and Spear, 2000]. Yttrium contents, which should mimic Mn concentrations [Otamendi et al., 2002], appear to deviate within garnet so that a median zone has Y concentrations anomalously higher than, or approximately as high as those of garnet cores. Aspects of elemental zoning are more fully discussed below with reference to quantitative point data, below.

[10] Quantitative major and trace element zoning of garnets (Figures 4 and 5, Tables 14) indicates that growth zoning of garnet is preserved; however, trace element compositions deviate from ideal Rayleigh fractionation/distillation behavior [cf. Otamendi et al., 2002]. In addition to measured compositions, Figures 4 and 5 present a model for Rayleigh distillation of Y by garnet growth following the equations of Hollister [1966] and Otamendi et al. [2002] and all “heavy” rare earth elements (HREE) should follow a similar law. We emphasize that this model only accounts for sequestration of Y and HREE by garnet, and is not truly valid in this system where other minerals will compete for these elements (i.e., titanite, clinozoisite). However, the model provides a useful reference frame for comparison of measured compositions to expectations for normal garnet growth zoning. Positive deviations of Y and HREE contents from the ideal Rayleigh model suggest disturbance of the system by a reaction among REE-partitioning phases. Variable Ca contents within the deviated trace element zones, rather than a smooth decline in concentration, suggest buffering of Ca in garnet in connection with the reaction affecting trace element compositions. Numerous lines of evidence support this interpretation of a mineral reaction affecting garnet compositions: (1) deviated zones within garnet directly coincide with the aforementioned change in the mineral inclusion assemblage in garnets, suggesting a causal relation; (2) concentrations of numerous HREEs are higher within the deviated zones than within garnet cores (Figure 4) or apparently buffered at levels similar to the core following an initial drop in concentration (Figure 5); both relationships indicate liberation of Y and HREEs from a phase that would also sequester these elements; (3) the high field strength element (HFSE) Hf displays a similar positive deviation that is well correlated with other elements in this zone, suggesting the decomposition of a phase that partitions HFSE as well as HREEs; (4) relative normalized concentrations of HREE internal (i.e., “coreward”) of the deviated zone differ from those external to the zone (i.e., “rimward”; Figure 4), strongly suggesting the overall REE composition of the rock reservoir available to the growing garnet had changed and indicating a shift in the bulk distribution coefficients buffered by the mineral assemblage in the system. We have observed such an anomalous chemical zone in garnet in all Cignana metabasaltic samples we have analyzed. Furthermore, identical results have been independently documented in the Cignana garnets by a recent combined Lu-Hf/Sm-Nd geochronologic study that quantified the duration of garnet growth for these eclogites (∼12Ma [see Lapen et al., 2003, Figure 2]), suggesting it is a pervasive feature of the metamorphic history for the locality and not an aberration or analytical artifact.

Figure 4.

(a) Major element and (b) trace element (NMORB-normalized [Sun and McDonough, 1989]) zoning profiles in garnet 7 in sample CIG 91-1. The mole fraction (X) of Mn in the garnet decays from high values in the core to lower values toward the garnet rim, as expected for growth zoning produced by Rayleigh distillation. Within the anomalous zone (gray shaded area), XCa is variable, while elevated Y and HREE contents deviate from an ideal Rayleigh distillation law (modeled for Y, dashed line), suggesting a reaction involving a REE-bearing calcic phase (see text). The coupling of Hf with the HREE within this zone supports titanite as a phase involved in this reaction. Distinct concentrations of HREE relative to one another to either side of the suspected reaction, as well as higher concentrations of Gd rimward of the reaction zone, indicate distinct REE partitioning behavior within the rock prior to and following the reaction, further supporting a mineralogic reaction among phases that strongly partition REE.

Figure 5.

(a) Major element and (b) trace element (NMORB-normalized [Sun and McDonough, 1989]) zoning profiles in garnet 3 in sample CIG 91-1. Similar to Figure 4, XMn supports retention of growth zoning by this garnet. In contrast with the relationships in Figure 4, this garnet preserves the reaction zone (gray area) as a much broader zone at approximately equal REE concentrations as present in Figure 4; however, Hf concentrations are approximately one order of magnitude higher within this garnet compared to Figure 4. Dashed line is a pure Rayleigh distillation model for Y during garnet growth.

Table 1. Major-Element Data for Garnet 7: Electron Microprobe Analysesa
Element OxideTraverse
CorePoint 2Point 3Point 4Point 5Reaction ZonePoint 8Point 9
Point 1Point 6Point 7
  • a

    Notes: Element oxides reported as weight percent; oxygen calculated using stoichiometry.

Element OxideTraverse
Point 10Point 11Point 12Point 13Point 14Point 15Point 16Point 17Rim
Point 18
Table 2. Major-Element Data for Garnet 3: Electron Microprobe Analysesa
Element OxideTraverse
CorePoint 2Point 3Point 4Reaction ZonePoint 10Rim
Point 1Point 5Point 6Point 7Point 8Point 9Point 11
  • a

    Notes: Element Oxides reported as weight percent; oxygen calculated using stoichiometry.

Table 3. Trace Element Data for Garnet 7: Ion Microprobe Analysesa
CorePoint 2Point 3Point 4Point 5Reaction ZonePoint 8Point 9
Point 1Point 6Point 7
  • a

    Notes: Element concentrations reported as parts per million (ppm); b.d. indicates element was below detection limits.

Point 10Point 11Point 12Point 13Point 14Point 15Point 16Point 17Rim
Point 18
Table 4. Trace Element Data for Garnet 3: Ion Microprobe Analysesa
CorePoint 2Point 3Point 4Reaction ZoneRim
Point 1Point 5Point 6Point 7Point 8Point 9Point 10Point 11
  • a

    Notes: Element concentrations reported as parts per million (ppm); b.d. indicates element was below detection limits.


4. Reconstruction and Implications of the Reaction History

[11] Garnet in mafic eclogites forms through a variety of both continuous reactions (at the expense of Fe-Mg-Al phases such as amphibole and chlorite) and discontinuous reactions (through the decomposition of Ca-Al phases such as lawsonite, titanite, and clinozoisite). We infer that the net effect of garnet production through both types of reactions allows for relatively continuous growth of garnet along a prograde subduction path. The features outlined above are most easily explained as representing the geochemical record of a discontinuous reaction that is probably common to eclogites approaching or experiencing UHP conditions, the breakdown of coexisting clinozoisite and titanite defined by the equilibrium of Manning and Bohlen [1991]:

equation image

The occurrence of this reaction explains inclusion relations of Ti- and Ca-bearing minerals within garnet and most easily resolves all aspects of the observed chemical zonation within garnet. The production of additional pure grossular (the Ca-Al-garnet end-member) by this reaction provides a mechanism for variations in Ca measured for garnets while other chemical end-members were unaffected.

[12] Differences between the measured trace element zoning profiles in Figures 4 and 5 at first appear to pose a significant dilemma in our interpretation of a specific reaction creating trace element anomalies in garnet. Potentially, such differences could arise from a sectioning bias inherited from preparation of the sample; however, we discount this possibility in that the similar overall pattern of HREE zoning within garnet (i.e., high-concentration core/subsequent decline/anomalous zone/final decline) suggests sections sample core compositions and differences in the width of the anomalous zone could not be produced by such a bias. Therefore we believe that our interpretation for a single genetic mechanism for the zoning of the two garnets is valid and that differences between the garnet profiles are an artifact of heterogeneous modal abundances for titanite within the system and local-scale buffering of HREEs and Hf during the reaction. We base much of our reasoning on the observation that across the profile of garnet 7, where the anomalous zone is present as a sharp peak in elemental abundances, Hf concentrations are consistently about one order of magnitude lower than Hf concentrations within garnet 3, where the anomalous zone is diffuse. Prior to the reaction, suitable mineral hosts for Hf in blueschists or eclogites are limited to titanite and zircon; given the well-known refractory nature of zircon, we assume that, once incorporated into zircon, Hf is essentially removed from the active metamorphic chemical reservoir. This leaves titanite as the most important host for Hf. We next assume that Hf partitioning among any and all titanite crystals within the system prior to the reaction is relatively constant. If these assumptions are valid, the lower Hf within the anomalous zone of garnet 7 (Figure 4) as compared to that of garnet 3 (Figure 5) is most simply explained as a reflection of the amount of modal titanite locally available to feed garnet growth. This interpretation is consistent with models for diffusionally limited growth of garnet during metamorphic reactions [Carlson, 1989, 1991, 2002]; this theory suggests that growth of garnet porphyroblasts during metamorphic reactions is limited by intergranular diffusion of cations compatible in garnet from reactants. Consequently, as garnet porphyroblasts grow, each creates a chemical potential well about itself, within which it can “feed” on the cations available from garnet-forming reactions [Carlson, 1989, 1991, 2002]. Therefore modal proportions of titanite were probably higher locally near the site of garnet 3 than near garnet 7 and this governed the redistribution of compatible trace elements such as the HREE in garnet.

[13] Given this theory, the sharp peak in HREE and Hf abundances in garnet 7 (Figure 4) reflects consumption of only a low modal amount of titanite within the feeding zone for this garnet grain. A greater inferred modal proportion of titanite within the diffusional domain of garnet 3 resulted in a broader zone recorded within garnet for the same reaction as larger local amounts of nutrients were available for garnet growth, resulting in a smoothed trace element profile (Figure 5) reminiscent of a buffered zone. Furthermore, the concentrations of HREEs across the breadth of the anomalous zone in garnet 3 (∼8–10x NMORB for ∼500 μm; Figure 5), are only modestly lower than the HREE concentrations within the comparatively narrow zone of garnet 7 (∼10–12x NMORB for ∼100 μm; Figure 4), which implies the absolute amount of HREEs integrated within the diffuse anomalous zone of garnet 3 is much higher. This further implicates a higher modal proportion of titanite available to garnet 3 yet reveals essentially constant (i.e., approximately equilibrium) partitioning of HREEs between phases during the reaction. The persistence of clinozoisite as a matrix mineral indicates that excess clinozoisite was available and would not limit the reaction. The opposite paragenesis, titanite + rutile (rather than clinozoisite + rutile), was documented for UHP eclogites by Sassi et al. [2000], and this probably reflects differences in bulk composition for the UHP eclogites of the Dabie-shan and Lago di Cignana. Therefore the anomalous trace element zones in garnet represent a transient stage of elemental partitioning between all product phases of the reaction as the rock adjusted to the removal of titanite.

[14] Partitioning and buffering of trace elements likely would have operated among all phases participating in the reaction. For discussion of REE systematics quartz/coesite can be ignored, and rutile has been shown to not partition REEs within eclogites [Zack et al., 2002], although it is the dominant new host for titanite-derived Nb and Ta. Accordingly, buffering of REEs would only be a function of garnet-clinozoisite-water partitioning. HREEs are probably preferred in garnet relative to either of the other phases, while LREEs are not compatible in garnet. LREE contents for clinozoisite have not been systematically investigated, but preliminary data [King et al., 2003a] do not support elevated concentrations in LREE for matrix clinozoisite compared to inclusions, as would be required if LREE back-reacted into clinozoisite via a hypothetical net-transfer mechanism [cf. Kohn and Spear, 2000]. This suggests some appreciable amount of LREEs would have been present in the aqueous phase produced by the reaction. Similarly, as clinozoisite is a major host for Sr and also that Sr is notably soluble in aqueous fluids but not in any solid product of the reaction, the fluid phase probably contained a Sr-LREE-enriched signature.

[15] The extent to which such trace element laden fluids are effectively mobilized out of eclogites is a subject of some debate. Several studies endeavoring to examine fluid-related elemental mobility during eclogite-facies metamorphism have found evidence for chemical depletion assigned to fluid mobility [e.g., Becker et al., 1999, 2000], while others have indicated fluids are unable to escape, fluids and/or elemental depletions are locally retained by or redistributed within eclogites, or that no evidence exists for fluid/elemental mobility of any extent [e.g., Philippot and Selverstone, 1991; Chalot-Prat et al., 2003; Spandler et al., 2003]. This has led many authors to conclude that the transport of fluids equilibrated with eclogite-facies metamorphism to arc sources is highly unlikely. From our previous research bearing on lower-grade lithologies, we have observed that fluid flow and metasomatic alteration are strongly controlled by high-permeability structures within subduction complexes, such as fractures and mélange zones [Bebout, 1991; Bebout and Barton, 1993, 2002; King et al., 2003b]. Within the Alpine orogenic chain, several studies of high-pressure meta-ophiolitic suites have found limited evidence for alteration or elemental mobility within coherent blocks, and have speculated that high-permeability pathways such as mélange or fault zones probably existed during subduction, but were destroyed by exhumation processes [e.g., Barnicoat and Cartwright, 1995; Miller et al., 2001]. Our working hypothesis is that fluid flow is focused along mélange zones [Bebout and Barton, 1993, 2002] and the mineralogy of these zones (dominantly talc, chlorite and amphibole) dictates that they are fundamentally weak structures within the subducting section and are probably preferentially reactivated as shear zones during exhumation of high-pressure metamorphic terranes. As such, they are probably under-represented in the natural rock record due to intense deformation or misidentified when encountered due to pervasive later alteration. Arguments regarding limited mobility of fluids in eclogites could therefore be an artifact of the natural bias of the rock record. We believe that the evidence presented in this paper for fluids produced by a specific mineral reaction, whereby this fluid phase is able to acquire a trace element composition compatible with enrichments observed within modern volcanic arcs, certainly demonstrates the potential for selective elemental loss in fluids. Given the volume of the subducting section and experimental constraints that imply essentially continuous dehydration beneath the mantle wedge [Schmidt and Poli, 1998, 2003], sufficient integrated volumes of hydrous fluids are probably produced to ensure development of mélange zones as pathways for fluid mobility.

[16] Our interpretation of a specific mineral reaction to produce the anomalous trace element zonation for these garnets implies that this discontinuous reaction may represent one (of several) significant trace element “pulses” in the overall continuous dehydration cycle occurring during subduction as postulated by Schmidt and Poli [1998, 2003]. Furthermore, our interpretation of controls on preservation of the reaction based on modal abundances of minerals supports arguments made by Schmidt and Poli [2003] for “smearing out” of reactions due to variations in modal abundances and mineral activities resulting from solid solution. In that elevated concentrations of Sr and LREEs in arc volcanic rocks have been repeatedly interpreted as being the result of fluid-mediated additions from the subducting slab to the sub-arc mantle [Gill, 1981; Elliott et al., 1997], we propose that this reaction may be a primary candidate for the source of trace element enrichments observed for some arc volcanics. Ultimately, an analysis of such a scenario would incorporate the P-T dependency of candidate metamorphic reactions that may form a trace element pulse (such as equation (1); see calculation below) with models of material transfer in subduction zones [e.g., Kincaid and Sacks, 1997; Conder et al., 2002] to test whether trace element enrichment for known volcanic centers can be linked to hypothetical mass transfer vectors ending in the region of a proposed reaction; such a test is a logical extension of arguments made by Kelemen et al. [2003]. Regardless, the trace element partitioning behavior of eclogites following this reaction will be significantly distinct from that prior to the breakdown of coexisting clinozoisite and titanite, suggesting that this reaction may have a fundamental control on trace element compositions of fluids buffered by eclogites at higher pressures.

[17] Application of relevant phase equilibria to the Cignana eclogites allows for relatively precise identification of the depth interval in which the inferred elemental mobility occurred. Figure 6 presents metamorphic facies and the P-T path of Reinecke [1998] for the Lago di Cignana locality as well as geothermobarometric constraints from this study (see Table 5 for mineral compositions). The equilibrium position of equation (1) [Manning and Bohlen, 1991] is plotted for pure phases and the range for the logarithm of the equilibrium constant for equation (1) (log K1) of −2.0 to −2.5. Several petrographic factors have dictated that we are unable to provide a unique barometric solution to the position of equation (1). Compositions for clinozoisite and titanite must be derived from inclusions in garnet, and while each are present as inclusions in garnet, both phases are not present within the same garnet grain. This undoubtedly reflects small-scale heterogeneity in mineralogy within the eclogites, discussed in detail above. Due to the heterogeneity of mineral inclusions preserved in garnet, significant doubt exists as to whether measured compositions used in the calculations represent a true equilibrium pair. In our calculations we used clinozoisite-titanite pairs that coexist with nearly identical garnet compositions. For products of equation (1), we assumed a unity activity for rutile, the activity of grossular was derived using the activity model of Berman [1990] and a range in aSiO2 of 0.8–1.0 was used to evaluate scenarios of slight SiO2 undersaturation. For the aH2O, we used a range below unity (0.3 ≤ aH2O ≤ 0.6), as mixed-volatile species appear to predominate in subduction zones where continental material has been deeply subducted, such as is the case for the western Alps [cf. Manning, 2004]. Despite the collective uncertainties present in equation (1) for the Cignana eclogites, these activities result in an overlap of the published P-T path [Reinecke, 1998] and log K1 (Figure 6) that agrees well with petrographic evidence for breakdown of coexisting titanite + clinozoisite at or shortly following the transition from the blueschist facies.

Figure 6.

Phase equilibria constraints on the reaction history. At left, pressure-temperature phase diagram illustrating the P-T path for Lago di Cignana (green line with arrows [Reinecke, 1998]) with petrologic constraints on the proposed reaction history for Cignana metabasaltic eclogites. Bold solid lines are metamorphic facies; AM, amphibolite; BS, blueschist; EA, epidote amphibolite; GS, greenschist; HGR, high-T granulite; LGR, low-T granulite. Bold dashed lines are subdivisions of the eclogite facies: AEC, amphibole eclogite; DEC, dry eclogite; LEC, lawsonite eclogite; ZEC, zoisite eclogite. Fine solid lines are common metamorphic reactions; mineral abbreviations are as follows: Arg, aragonite; Cal, calcite; Coe, coesite; Dia, diamond; Gr, graphite; Hab, high albite; Jd, jadeite; Lab, low albite; Qtz, quartz. Phase diagram after Oh and Liou [1998] and Okamoto and Maruyama [1999]. Bold blue line is the equilibrium position of equation (1) for pure phases [Manning and Bohlen, 1991]; fine blue lines are for calculated values of log K1; see text for assumptions regarding calculation of equation (1). Red lines are calculated positions of the garnet-clinopyroxene Fe-Mg exchange thermometer using the calibrations of Powell [1985] and Krogh [1988], labeled P and K, respectively. At right, schematic illustration depicting the petrologic progression of mineral growth within metabasalts and inferred fluid mobility during subduction for Lago di Cignana.

Table 5. Major-Element Data for Other Minerals: Electron Microprobe Analysesa
Element OxideGarnet-Cpx ThermometrybInclusions in Garnetc
  • a

    Notes: Reported as weight percent; oxygen calculated using stoichiometry.

  • b

    Thermometry compositions are average compositions for garnet rims (n = 6) and matrix clinopyroxene (n = 5) retaining well-preserved equilibrium textures.

  • c

    Inclusion compositions are average data; titanite n = 3, clinozoisite n = 2.


[18] Figure 6 also presents peak metamorphic temperatures calculated using calibrations of the garnet-clinopyroxene Fe-Mg exchange thermometer by Powell [1985] and Krogh [1988]. We chose these two calibrations in that they resulted in the greatest difference in temperature (∼80°C) of many calibrations surveyed for use in this study. Garnet rim compositions were used with average matrix clinopyroxene compositions; corrections for Fe3+ were made using stoichiometry. Our results are consistent with the published P-T path [Reinecke, 1998], although extending to slightly higher temperatures. The consistency of the thermometry results and the presence of coesite with matrix clinopyroxene further support our interpretation that reliable peak metamorphic chemical compositions are preserved in these samples.

[19] Finally, REE compositions for inclusion and matrix clinopyroxene, when evaluated in light of garnet REE zoning, demonstrate the potential utility of mineral inclusions in garnet in identifying trace element and isotopic fractionation resulting from subduction zone metamorphism (see data for sample CIG 91-1 in Figure 7a, Table 6). The progressive depletion of HREE in garnet from core to rim reflects depletion of the REE budget available to all minerals in the rock due to Rayleigh fractionation of HREE by garnet [e.g., Otamendi et al., 2002], in some cases with superimposed concentration anomalies related to reactions among key mineral hosts (see Figures 4 and 5). Lowered normalized HREE compositions for matrix clinopyroxene, as compared with HREE concentrations in clinopyroxene inclusions, (Figure 7a) most likely reflect chemical equilibration with this progressively depleting reservoir that evolved along the prograde subduction path. Therefore combined spatial relationships and evolving REE compositions of clinopyroxene inclusions in garnet may be robust recorders of petrologic evolution and processes potentially controlling the geochemical evolution of fluid-mobile light isotopic tracers of elements such as lithium within eclogites. Zack et al. [2003] found that Li is primarily hosted by clinopyroxene in eclogites and hypothesized that extremely fractionated “light” values were produced during subduction zone metamorphism. SIMS analytical results for Li concentrations in Cignana eclogite minerals (Figure 7b, Table 7) support conclusions that clinopyroxene is the dominant Li host in eclogites and that Li is strongly partitioned into clinopyroxene relative to garnet, in agreement with experimental results [Brenan et al., 1998]. Although Li concentrations between inclusion and matrix clinopyroxene do not appear to vary significantly in our limited data set (Figure 7b), any isotopic fractionation of Li isotopes produced by fluid mobility during eclogite-facies metamorphism may be correlated with changes in the REE composition of clinopyroxene, providing a powerful connection between prograde metamorphism and Li isotopic fractionation. In that these inclusions can be tied to the metamorphic P-T history recorded by garnet, detailed records of temperature-dependent fractionation of lithium isotopes during subduction, if it occurs, should be recoverable. Similar arguments apply to phengite inclusions, which strongly partition boron, and the boron isotopic system [cf. Bebout et al., 2004].

Figure 7.

(a) NMORB-normalized [Sun and McDonough, 1989] REE variation diagram illustrating the progressive depletion of HREE as recorded by garnet and clinopyroxene (CPX) in sample CIG 91-1 of van der Klauw [1998]. Spatial relationships and chemical records of mineral growth such as these (see text) indicate detailed histories of chemical and isotopic fractionation may be present and assessed using SIMS methods. (b) Comparison of Li contents for matrix clinopyroxene, garnet, and clinopyroxene inclusions in garnet.

Table 6. Trace Element Data for Clinopyroxene: Ion Microprobe Analysesa
ElementMatrix ClinopyroxenesInclusions in Garnet
Matrix 1Matrix 2Matrix 3Matrix 4Matrix 5Matrix 6Incl. 1Incl. 2
  • a

    Notes: Element concentrations reported as parts per million (ppm).

Table 7. Lithium Concentration Data: Ion Microprobe Analysesa
Inclusion CPXGarnetMatrix CPX
  • a

    Notes: Lithium concentrations reported as parts per million (ppm).


5. Prospects and Conclusions

[20] In summary, this study provides direct analytical evidence from samples that have experienced subduction zone metamorphism to sub-arc depths for a specific metamorphic reaction that could promote fluid mobility with a predicted trace element signature likely to occur within modern subduction zones. Complementary to our findings, a recent geochemical study of volcanic rocks from northeastern Japan has concluded that anomalous across-arc elemental and isotopic variations expressed within this arc are most likely due to the pronounced effect of lawsonite decomposition mobilizing a specific trace element signature to this arc [Moriguti et al., 2004]. This new insight from arc volcanics for recognition of a specific metamorphic reaction indicates that an explicit reconciliation between metamorphic reactions within the slab and geochemical enrichments for arc volcanic centers is a realistic goal of future studies.

[21] As demonstrated here, in order to retrieve metamorphic records of subduction in exhumed and overprinted tocks, identification of phases that preserve prograde histories is critical. Other recent work on the B isotope compositions of tourmaline produced during UHP metamorphism and exhumation in metasedimentary rocks at Lago di Cignana [Bebout and Nakamura, 2003] indicates that careful examination of phases that are relatively “robust” to chemical changes during retrograde metamorphic history (such as garnet and tourmaline) can afford reconstruction of reaction history and geochemical evolution in subducting materials. Reinecke [1998] reported that, for the metasedimentary rocks of the Lago di Cignana locality, only garnet, tourmaline, apatite, and dolomite preserve chemical information regarding prograde history, whereas all other phases in these samples represent extensive re-equilibration during exhumation. Similar studies of other UHP terranes likewise document the utility of zircon as a robust phase that encapsulates UHP histories [e.g., Katayama et al., 2002]. Petrologic evidence such as this argues strongly against the utility of whole rock geochemical data in reconstructing geochemical records of subduction. Future research bearing on subduction zone geochemistry, based on studies of metamorphic suites, should thus emphasize elemental and isotopic microanalysis of robust phases and other phases preserved within them utilizing SIMS and/or LA-ICPMS methods [e.g., Usui et al., 2003; Spandler et al., 2003; Bebout and Nakamura, 2003]. Careful placement of these geochemical observations into the context of P-T reconstructions can lead to greater understanding of material recycling through subduction zones and quantification of the fundamental metamorphic controls on arc magmatism.


[22] We thank M. Walter for kindly providing the laser Raman spectrum as well as R. Tanaka and T. Moriyama for assistance with the ISEI electron probe. Valuable discussions with M. Kohn, C. Parkinson, T. Usui, and members of the PML improved interpretations of our data. W. Carlson, associate editor T. Elliott, and two anonymous reviews greatly improved the manuscript. This study was supported by grants from the U.S. National Science Foundation (EAR-0079331, to Bebout) and from the 21st Century of Excellence Program, Japanese Ministry of Education, Science, Sports, and Culture (to Nakamura).