Retrograde eclogites (ranging from unaltered eclogite to retrograde blueschist and greenschist mantling the eclogite boulders) from Ring Mountain on the Tiburon Peninsula, near San Francisco, California, were examined for whole-rock major and trace elements to assess protolith compositions and the geochemical signature of fluids associated with retrogression. High field strength elements are highly correlated, indicating relatively immobile and conservative behavior during retrogression. These immobile elements were used to assess the relative losses or gains of other elements during retrogression. Rare earth elements and FeO content show only minimal open-system behavior. The rare earth abundance patterns, FeO contents, and Nb/Ta and Nb/La ratios show that the protoliths of these rocks were most likely normal to enriched mid-oceanic ridge basalts. Mass-balance considerations reveal two independent styles of metasomatic enrichment during retrogression. One style involves coupled enrichment in large ion lithophile elements (Cs, Rb, Ba, K, and Tl), likely caused by fluids from sediments or reaction with sediments. Another style involves coupled enrichment in Cr, Mg, Ni, and Pb, which may reflect overprinting by reaction of eclogite boulders with serpentinite, the latter of which are highly enriched in these elements. Pb is shown here and elsewhere to be nearly universally enriched in serpentinites and is likely to be selectively mobilized into eclogite-serpentinite reaction zones. Because all retrograde lithologies show reaction with serpentinites and sediments, exhumation of the eclogite must have been accompanied by chemical interaction with serpentinites along the entire retrograde path. The simplest interpretation is that the eclogites were transported within a deeply rooted serpentinite channel, presumably formed along the slab-mantle interface by infiltration of slab-derived fluids into the overlying mantle wedge. Physical models of channel flow show that the rapid exhumation rates required to preserve eclogites in a hydrous carrier matrix, such as serpentinite, are possible due to the buoyant and low-viscosity nature of serpentinite. However, the most rapid ascent rates occur during oblique subduction, suggesting that eclogite exhumation could be favored by, but not confined to, oblique subduction zones.


Subduction zones are places where crust is returned and recycled back into the mantle, but the presence of high-pressure metamorphic rocks, such as eclogites, in accretionary prisms indicates that deep crustal materials are, on occasion, exhumed to the surface in convergent margins (Platt, 1993). Geochronologic studies comparing ages of peak metamorphism and exhumation in different accretionary complexes indicate that the time between peak metamorphism (e.g., eclogite formation) and exhumation is <10–20 Ma, corresponding to minimum ascent rates of 2–5 km/Ma (Anczkiewicz et al., 2004; Baldwin et al., 2004; Cloos, 1985). Such rapid exhumation is required in order to avoid complete retrogression of high-grade metamorphic rocks during ascent. Such rapid return is remarkable given that the high densities of eclogites make them neutrally buoyant with respect to most lithologies in the subducted crust and the upper mantle.

The purpose of this paper is to use geochemistry to explore how eclogites return to the surface in ocean-continent subduction. In one class of models, slivers of eclogitized oceanic crust are ripped off during subduction and then transported to the surface by coupled erosion, extension, and return flow in the accretionary prism (Platt, 1986; Schemmann et al., 2008). Another class of models involves transport of eclogite fragments by diapirs or channels associated with low density and hence buoyant matrices, such as sediments (siliciclastics or carbonates) or serpentinites (Ernst, 1970; Ernst et al., 1997; Guillot et al., 2001; Moore, 1984; Schwartz et al., 2001). One of the difficulties in testing these hypotheses is that eclogite boulders often occur in highly tectonized and jumbled terranes, e.g., mélanges, so their original high-pressure structural relationships with surrounding rocks are poorly preserved. Here we investigate the geochemical signatures of variably retrograde eclogite lenses found in the Franciscan mélange, which is part of a Jurassic–Cretaceous accretionary complex that extends from California's central coast to coastal northern California (Platt, 1975). The purpose of examining variably retrograde eclogite lenses is to determine how the chemistry of eclogites changes during retrogression, providing insight into exhumation processes. We conducted a case study of Franciscan eclogites from the Tiburon Peninsula, just north of San Francisco, California. We show that the trace element geochemistry of the retrograde eclogites is consistent with transport from depth within a serpentinite matrix. We then use these observations to speculate on the nature of eclogite exhumation in ocean-continent subduction zones.

Case Study of Retrograde Franciscan Eclogites from the Tiburon Peninsula, California

The Franciscan mélange, associated with the accretionary wedge formed during Farallon plate subduction beneath North America, consists largely of graywacke, chert, and basaltic volcanic units disrupted by tectonic blocks and slivers of low-grade metamorphic rocks, such as greenschists and serpentinites, and medium- to high-grade metamorphic rocks, such as blueschists, amphibolites, and eclogites. Peak metamorphic ages for the high-grade rocks in the Franciscan mélange are between 149 and 169 Ma, based on Lu-Hf ages of eclogites and amphibolites (Anczkiewicz et al., 2004).

We collected samples from Ring Mountain, located on the Tiburon Peninsula in California (Catlos and Sorensen, 2003; Tsujimori et al., 2006a). Most of this region is made up of serpentinite and greenschist, within which are occasional large boulders and lenses of eclogite. We examined 11 samples, representing different parts of an eclogite boulder, in order to investigate protolith compositions and the effects of retrograde metasomatism on composition. We categorized these samples into three lithologies (Fig. 1), following descriptions given in previous studies (Saha et al., 2005; Tsujimori et al., 2006a). First are the eclogite cores of the high-grade lenses (Fig. 1A). These eclogites consist of ~40% garnet, ~60% omphacite pyroxene, and small amounts of chlorite alteration rims to pyroxenes (accessory phases include rutile and epidote). Eclogites overprinted by blueschists make up the second category and represent the first retrograde envelope surrounding the eclogite core (Fig. 1B). The blueschist-overprinted eclogites consist of <50% garnet, >30% chlorite, >50% glaucophane, and ~10% phengite; accessory minerals include titanite, apatite, Fe-oxides, and quartz. Garnets commonly show chlorite breakdown rims and all pyroxenes appear to have been converted to glaucophane or chlorite. The third category is made up of chloritized eclogite, representing the outermost retrograde rind (Fig. 1C). Pyroxene has been mostly transformed into glaucophane or chlorite with chlorite mode exceeding glaucophane due to replacement of glaucophane by chlorite. The mineralogy consists of <40% garnet, 40% chlorite, <20% glaucophane, and <1% phengite.

Based on detailed thermobarometric studies of relict prograde and retrograde mineralogies, Tsujimori et al. (2006a) showed that these eclogites followed a counterclockwise pressure-temperature (P-T) path and were exhumed along decreasing thermal gradients, perhaps due to progressive refrigeration of the mantle wedge by the subducting Farallon plate. The eclogites underwent peak metamorphism at ~2.2 GPa (70 km), while their blueschist retrogression occurred at ~1.1 GPa (35 km). Based on Lu-Hf for peak metamorphism and Ar-Ar ages of 147 Ma for blueschist retrogression (Anczkiewicz et al., 2004; Tsujimori et al., 2006a), Tsujimori (2006a) suggested that exhumation rates might have been as fast as ~5 km/Ma, similar to that inferred for exhumation rates in ultrahigh-pressure metamorphic terranes.

Previous Geochemical Studies on the Tiburon Eclogites

Trace element and isotopic studies of the Tiburon eclogites were conducted by Saha et al. (2005), who argued that the Tiburon eclogites and associated mafic metamorphic rocks had an island arc basalt protolith, based on slightly low Nd isotopic compositions, high Pb isotopic compositions, and in particular low Ce/Pb ratios. Low Ce/Pb ratios in basalts are generally thought to indicate preferential mobilization of Pb, presumably due to hydrous melting, and therefore are a signature of arc basalts. However, Pb is highly mobile during metamorphic reactions involving fluids, begging the question of whether the Pb contents (and Pb isotopic composition) of the eclogites truly reflect the protolith composition. In part, the goal of this study is to assess the role of open-system behavior during retrograde metamorphism on the geochemistry of the eclogites.


Samples were crushed and powdered in a Spex mill. One aliquot was sent to Washington State University, Pullman, for major element and select trace element analyses by X-ray fluorescence (XRF). Another aliquot (50 mg of powder) was dissolved for major and trace element analysis by solution inductively coupled plasma–mass spectrometry (ICP-MS) at Rice University, Houston. Samples were dissolved using 0.25 mL of a 1:1 mixture of concentrated HF and HClO3, heated at 140 °C, and followed by complete dry down at 170 °C. This procedure was repeated and the final dry down was taken up in 125 mL of 2 wt% HNO3 (one drop of 6 N HCl was added to keep Fe in solution). Procedural blanks and external standards (U.S. Geological Survey standards BHVO1 and BIR1) were processed through the same procedure simultaneously. A known quantity of indium was added to each solution to serve as an internal standard for instrumental drift during analysis. Solutions were introduced into a Thermo-Finnigan Element 2 ICP-MS by free-aspiration via an Elemental Scientific Teflon nebulizer (100 μL/min uptake) attached to a cyclonic spray chamber. Measurements were done in low and medium mass resolution modes, the former for trace elements and the latter for major elements and those trace elements in the mass range of 45–66 amu, where isobaric molecular interferences occur and can be resolved in medium mass resolution. Signal intensities were corrected for instrumental drift by monitoring In, and then converted to concentrations by calibration against BHVO1 and BIR1 basalt standards (using standard values from the Max Planck Institute GeoReM database; http://georem.mpch-mainz.gwdg.de/).


Geochemical data for the core to rim series (e.g., eclogite core to blueschist and greenschist mantles) are shown in Table 1 (data have not been normalized to 100%). Major element contents of the eclogite cores are broadly similar to basalts with the exception of the high Na contents, presumably reflecting a seawater-altered oceanic crust protolith (Fig. 2). Retrogression from eclogite to blueschist to greenschist is associated with decreasing SiO2 and Na2O but increasing Al2O3 content (Figs. 2B, 2D). MgO and K2O rise in the blueschists and then decrease in the greenschist. The high K2O contents are correlated with the appearance of phengite in the blueschists. Additional geochemical transitions as a function of Na content are shown in Figure 3. Figure 3A shows how Mg is enriched in the blueschists. Figure 3B shows that the retrograde lithologies, with the exception of one sample, are elevated in Cr relative to the eclogite core. Figure 3C shows higher Ni contents in blueschists, but similar Ni contents in greenschist and eclogite. Figure 3D shows that Li is strongly correlated with Na, indicating that retrogression of eclogite is accompanied by Li depletion.

To more meaningfully display all the geochemical data, however, relative enrichments or depletions were calculated by assuming that the high field strength elements (HFSEs; Ti, Zr, Nb, Ta) were immobile. This is a common assumption in metamorphic petrology as well as in soil geochemistry because such elements are much less soluble in aqueous fluids than monovalent or divalent cations (Ague, 2003; Ague and van Haren, 1996; Brimhall and Dietrich, 1987; Little and Lee, 2006; van der Straaten et al., 2008). Although perfect immobility is probably never attained, the high degree of correlation between Nb, Ti, and Ta (Fig. 4) suggests that the immobile approximation of these elements is probably reasonable for this case study (as the likelihood of each being equally mobile is small). We note that correlations of Zr with the elements shown in Figure 4 are poor, and hence they are not shown. This may be due partly to the fact that Zr measurements were done by XRF, whereas Ti, Nb, and Ta were done by solution ICP-MS, and therefore the latter set represent internally consistent data (Zr was not measured by solution ICP-MS due to our inability to dissolve all zircons). In any case, Nb/Ta and Nb/Ti ratios are similar to mid-oceanic ridge basalt (MORB) and ocean island basalts, a feature to which we will return.

Taking Nb, Ta, and Ti as immobile, mass loss or gain during retrogression relative to the original protolith mass, Mrock/Mo, can be estimated with the following equation (Brimhall and Dietrich, 1987):  
where Mrock is the mass of the retrograde lithology, Mo is the original mass of the protolith, Crocki is the concentration of an immobile element in the retrograde lithology, and Coi is the concentration of the same immobile element in the protolith. Thus, for an immobile element, an increase in its concentration requires mass loss from the system during retrogression. Conversely, a decrease in its concentration indicates mass gain. These mass conservation principles can be extended to all other elements to assess their mobility (relative to the immobile element i). The percentage increase or decrease in mass of a given element, forumla, during retrogression is given by (Brimhall and Dietrich, 1987; Little and Lee, 2006):  
where Crockj is the concentration of mobile element j in the retrograde lithology, Coj is the concentration of mobile element j in the protolith, moj is the mass of mobile element j in a given amount of protolith, and ΔCoj is the change in mass of element j during retrogression.

For our case study, we assume the eclogite core represents the protolith and we adopt Ti as our immobile element i. Unfortunately, however, the Ti contents in our two eclogite core samples (UH1 and UH8) differ considerably (Fig. 4A). The Ti and Nb contents of UH8 are the only elemental concentrations that differ substantially between UH1 and UH8. In TiO2 versus Na2O plots (not shown), UH8 appears to be an outlier, so we attribute the low Ti contents of UH8 to nugget effects associated with dissolution of heterogeneous samples wherein Ti is housed primarily in accessory oxide phases. We thus take the Ti content of UH1 as an estimate of the protolith's Ti content. Figure 5 shows percent mass loss or gain of individual elements relative to the average of the two eclogite samples (by definition, Ti mass change is zero). The red horizontal bar qualitatively defines the regions of 30% mass loss or gain for reference. We note that the HFSEs all plot close to zero (Fig. 6), suggesting that our choice of Ti protolith concentration is reasonable. If UH8's Ti value was used instead, all the HFSEs would have been shifted considerably from the zero line, inconsistent with their presumed immobile behavior.

The most obvious features in Figure 5 are as follows. K, Cs, Rb, Ba, Tl, and Cu show the most pronounced enrichments (>300%), particularly in the blueschists. Figures 6A and 6B show that the enrichments in Ba, Rb, and K (as well as Cs and Tl; not shown) are correlated. Sr also shows substantial enrichments, but the most pronounced enrichments are in the greenschists, and hence Sr enrichments are decoupled from Ba, Rb, and K. Li and Na show depletions (Fig. 5) and therefore are anticorrelated to enrichments in Ba, Rb, K, Cs, and Tl. Cr shows relative enrichments of 200%–800% (Figs. 5 and 6C–6F), but such enrichments show no correlation with enrichments in large ion lithophile elements (LILEs) like Ba (Fig. 6C). Cr enrichment instead appears to be correlated with smaller enrichments in Mg and Ni (Figs. 6D, 6E); this correlation holds even if the most Cr- and Mg-enriched sample is ignored. The covariation diagrams in Figure 6 have additional significance in that any correlations are essentially independent from choice of normalizing element (e.g., the immobile element of choice). We note that choosing a different immobile element could cause those elements that show moderate (<50%) enrichment or depletion factors, such as Mg and Ni, to shift about the zero line. However, the fact that the large Cr mass changes correlate with the smaller Mg and Ni mass changes (Fig. 6) suggests that Mg and Ni have undergone relative enrichments. What is surprising is that Cr enrichment is strongly correlated with Pb enrichment, even though these two elements in general do not behave geochemically similarly.

Rare earth element (REE) mobility in Figure 5 appears to be largely limited as nearly all samples cluster around zero for the heavy REEs. One blueschist sample (UH10) appears to show substantial relative REE enrichment, but this is certainly due to its anomalously low Ti content, because the absolute REE contents are only slightly higher than the other blueschists. In any case, the REE systematics in the form of primitive mantle normalized values (Fig. 7A) show that all of the lithologies in this study have REE patterns similar to those of MORBs, albeit with slight enrichments in the light REEs. These slight enrichments in the light REEs are most pronounced in the retrograde lithologies, suggesting that there has been some infiltration of light REE components.

Ce/Pb is often used as a provenance indicator in basaltic magmas because Ce and Pb are thought to have identical geochemical behaviors during anhydrous mantle melting (Hofmann et al., 1986). During hydrous melting, however, it is thought that Pb is more mobile, resulting in fractionation of Ce/Pb ratios and generating the low Ce/Pb ratios characteristic of arc basalts. Figure 8 shows that decreasing Ce/Pb ratio is associated with increasing Pb contents with little to no changes in Ce content. The lowest Ce/Pb ratios are associated with the blueschists, consistent with the Pb enrichment inferred from simple mass-balance approaches. In all likelihood, even the eclogites have also undergone considerable Pb enrichment.


Geochemical Constraints on Protolith

As shown in the Results section and in Figures 4 and 5, the HFSEs (Nb, Ta, Zr, Ti) and the REEs seem to have undergone less disturbance during retrogression. This is evidenced by the constant Nb/Ta and Nb/Ti ratios and the clustering of REE mass changes around zero in Figure 5 for the entire suite of lithologies. The low mobility of these elements allows us to use their relative proportions as geochemical fingerprints of possible protoliths. For example, Nb/La and Nb/Ta ratios cluster around 1 and 17, respectively. These values coincide with values of the model bulk silicate Earth (e.g., the Earth's primitive mantle; McDonough and Sun, 1995), MORBs, and ocean island basalts (hotspot magmas), but contrast strongly with the low Nb/La and Nb/Ta ratios seen in island arc basalts (Figs. 4B and 7B) (Hofmann, 1988; Hofmann, 1997; Kelemen et al., 1990). The absolute abundances of the heavy REEs are also very similar to those of MORBs, unlike arcs, which have slightly lower heavy REE abundances, as can be seen in Figure 7A (Sun and McDonough, 1989). The REEs normalized to model bulk silicate Earth (McDonough and Sun, 1995), and plotted in order of increasing atomic weight (and increasing compatibility in the solid), exhibit flat to slightly negatively sloped abundance patterns. Such patterns are typical of enriched MORBs (e.g., seamounts) or ocean island basalts (Fig. 3D) (Hofmann, 1997; Kelemen et al., 2003; Sun and McDonough, 1989), but not island arcs.

The FeO contents (all Fe taken as Fe2+) of the eclogites are ~10 wt%, which is higher than that seen in island arc basalts (~8 wt%) and on the high end of MORBs. High Fe contents could be related to high-temperature and high-pressure melting of the mantle if the eclogite protolith was once a basaltic liquid. Assuming that the eclogite protoliths represent fractionated basalts, we can estimate the primary parental magma of the protolith by correcting for crystal fractionation (assumed to be dominated by oli vine) until the hypothetical magma is in equilibrium with a mantle source typical of the upper mantle [molar Mg/(Mg + Fe) ~0.89–0.9; Fig. 9]. Mg-Fe exchange between olivine and liquid can then be used to estimate temperature (Courtier et al., 2007; Putirka et al., 2007; Roeder and Emslie, 1970), yielding temperatures of ~1500 °C (Fig. 9). These temperatures, however, are likely to be upper bounds because the addition of Fe-rich cumulate crystals during igneous differentiation, Fe addition during hydrothermal alteration, and preferential mass loss during metamorphism could result in an elevation of Fe content. We thus conclude that our calculated temperatures are similar to oceanic plateaus, hotspots, seamounts (>1400 °C), and MORBs (1300–1400 °C) (Courtier et al., 2007; Herzberg et al., 2007; Putirka et al., 2007).

Based on the above trace element and major element compositions, we estimate that the protoliths of the Ring Mountain eclogites were MORBs or enriched MORBs and thus represent oceanic crust or seamounts. Previous studies argued that the protoliths were island arc basalts on the basis of Pb isotopic compositions and low Ce/Pb ratios (Saha et al., 2005), but as we discussed above, low Ce/Pb ratios are a feature of preferential Pb addition during retrogression.

Geochemical Signatures of Retrogression and the Role of Serpentinites

There appear to be at least two types of metasomatic enrichments associated with retrogression. As outlined in the Results section, one type of enrichment is characterized by the addition of LILEs such as K, Rb, Cs, Ba, and Tl. Similar metasomatic enrichments are found in eclogites and blueschists from Catalina Island, California, and were used to argue that these mafic metamorphic rocks were “bathed” in LILE-enriched fluids, since these elements are thought to be fluid mobile (Catlos and Sorensen, 2003; Giaramita and Sorensen, 1994; Sorensen and Grossman, 1989). These elements ultimately must derive from fluids released from a source enriched in these elements. Metamorphosed sediments, such as pelites, are probably the most likely source for these elements (Bebout, 2007; Bebout and Barton, 2002; Catlos and Sorensen, 2003). In contrast, Na and Li are rather depleted in the blueschists and greenschists compared to the eclogite protolith. This seems counterintuitive because their geochemical behaviors should be similar to the other LILEs; in addition, Na is a major constituent in glaucophane. However, such depletions have been shown in similarly retrograde eclogites from Tian Shan, China (van der Straaten et al., 2008).

The second metasomatic overprint is characterized by enrichments in Cr, Mg, Ni, and Pb, and appears to be independent of the LILE enrichments. Enrichments in Cr, Mg, and Ni have been seen in retrograde eclogites from Tian Shan, China (van der Straaten et al., 2008). Enrichments in Cr and Mg have also been seen in the envelopes of eclogites and amphibolites in contact with serpentinized ultramafics in the Catalina mélange complex in California (Bebout and Barton, 2002). Van der Straaten et al. (2008) suggested that these enrichments in Cr, Mg, and Ni were due to reaction of the eclogites with a serpentinite matrix during retrogression because these elements are enriched in the ultramafic protoliths of serpentinites.

The origin of Pb is less clear. Pb enrichments could derive from pelitic-sourced fluids, but the correlation of Pb and Cr enrichment (Fig. 6F) suggests that the Pb enrichment is coupled with Cr, Mg, and Ni enrichments. If the latter are associated with serpentinite-derived fluids, the implication is that Pb may also be related to serpentinites. Although we have not studied the serpentinites in Ring Mountain because they are scattered throughout the mélange and may not be directly related to the eclogite boulders, Pb enrichment (and low Ce/Pb) appears to be a universal feature of serpentinized peridotites in ocean basins and arc environments. For example, extremely high Pb anomalies (as much as 40 times that of primitive mantle, which approaches the levels seen in continental crust) have been found in serpentinized ophiolites (Li and Lee, 2006) and abyssal peridotites (Kelemen et al., 2007; Niu, 2004) (Fig. 3A). The origin of high Pb in serpentinites is not known, but Pb is likely introduced from seawater during the serpentinization process because Pb enrichments are much less common in unserpentinized mantle xenoliths (Lee, 2005). It has been suggested that the high Pb content in serpentinites may be a global feature, perhaps even resolving the Pb mass balance for the silicate Earth (Godard et al., 2005; Kelemen et al., 2007). It is thus reasonable to speculate that the Pb enrichment in the retrograde eclogites may have derived from serpentinites during reactions between eclogite and surrounding serpentinite. We note, however, that the Tian Shan retrograde eclogite series show no coupled Pb enrichments with Cr, and instead show depletions in Pb (van der Straaten et al., 2008). Assuming that reaction with serpentinite is involved in the retrogression of both the Tian Shan and Ring Mountain eclogites, these differences in Pb behavior require explanation. One speculation is that Pb mobility may be related to the different P-T paths undergone by these lithologies in the two regions. In the Tian Shan case, inferred P-T paths are clockwise, with the retrograde path showing increasing thermal gradients during exhumation (van der Straaten et al., 2008). In contrast, the Ring Mountain case shows a counterclockwise path wherein the retrograde segment follows a decreasing thermal gradient during exhumation, a feature thought to be related to refrigeration by the downgoing Farallon plate (Tsujimori et al., 2006a). Thus, one possibility is that Pb was completely mobilized out of the system in the Tian Shan case due to the higher temperatures of retrograde metamorphism (~600 °C at 1.5 GPa) compared to that seen in the Ring Mountain case (~400 °C at 1.5 GPa).

Eclogite Exhumation via Serpentinite Channels

Our geochemical observations indicate that eclogite exhumation and retrogression were accompanied by reaction with pelitic sediments and serpentinites. The fact that the trace element signatures of retrograde blueschist envelopes and chloritized eclogites all indicate reaction with serpentinite requires that eclogite and serpentinite were juxtaposed and reacted along the entire retrograde path from high to low pressures. This suggests that the eclogites were carried up in a long serpentinite channel extending from peak metamorphic pressures to the surface. This does not rule out the possibility that the last stages of exhumation were accommodated by crustal level extension along the lines of metamorphic core complexes.

These observations and arguments are thus consistent with eclogite exhumation via serpentinite-bearing channels (Ernst, 1970; Ernst et al., 1997; Guillot et al., 2001; Moore, 1984; Schwartz et al., 2001). These serpentinite channels could represent original serpentinites within the subducting oceanic lithosphere itself, that is, just beneath the oceanic crust. Alternatively, the serpentinite channel could consist of secondary serpentinites, formed in the mantle wedge above the slab-mantle interface. Secondary serpentinite in the corner of the mantle wedge above the subducting slab are thought to form as water in pores or bound in hydrous minerals in subducted sediments and altered oceanic crust is released into the corner of the overlying mantle wedge. Serpentinized mantle wedges are the best explanations for low seismic velocity corners in the mantle wedge observed in some active subduction zones (Bostock and Van Decar, 1994).

If a serpentinite channel does exist, it could represent a convenient exhumation pathway for high-grade metamorphic rocks. Serpentinite has a much lower density than the mantle (2700 kg/m3 compared to 3300 kg/m3) and a very low viscosity (Hilairet et al., 2007). The former provides a strong buoyancy force generating a pressure gradient that can drive flow of the serpentinite to the surface, whereas the low viscosity reduces the viscous resisting forces (Ernst, 1970; Ernst et al., 1997; Guillot et al., 2001). Coupled with erosion of the serpentinite at the surface (e.g., via serpentinite mud volcanoes; Fryer et al., 2000), considerable upward transport along this serpentine channel may occur. As these serpentinites rise, they may disrupt eclogitized oceanic crust or seamounts, transporting such fragments to the surface. Of interest are the ascent rates of these serpentinite channels. If ascent rates are too low, the eclogites would probably undergo complete retrograde metamorphism. Retrogression kinetics would be further enhanced by the fact that the eclogites are surrounded by a hydrous matrix of serpentinites or sediments. Thus, the question is whether ascent rates within serpentinite channels are ever fast enough to match those constrained by the geochronologic studies discussed above and to prevent complete retrogression.

Ascent Rates in Serpentinite Channels and the Role of Oblique Convergence

In order to assess the physical plausibility of the serpentinite channel hypothesis, we now explore simple quantitative models of eclogite exhumation in serpentinite channels. A number of simple models of serpentinite channel flow have been presented (Guillot et al., 2001; Schwartz et al., 2001). It is important that there is a trade-off between serpentinite layer thickness and ascent rate, because the thinner the layer thickness, the greater the viscous resisting forces and the slower the ascent rate. Thus, large serpentinite layer thicknesses are needed for serpentinite ascent rates to exceed the downward dragging force associated with the subducting slab (layer thicknesses must be >10 km for a ~10 cm/yr plate velocity; Schwartz et al., 2001). However, in the case of the Franciscan eclogites at Ring Mountain, it has been suggested that the angle of subduction was initially low, possibly making it difficult to accommodate thick serpentinite channels. If the serpentinite channel hypothesis is correct, as is implied from our observations, additional physics must be explored.

In this regard, it is worth exploring the possible relationship between exhumation and obliquity of subduction. It has been suggested that exhumation of high-grade metamorphic rocks is globally related to oblique convergence (Tsujimori et al., 2006b). Eclogites in Central America, the Caribbean islands, and northern Venezuela and Ecuador all occur in serpentinite mélanges and were formed along oblique subduction margins between the Caribbean plate and North and South America (Avé Lallemant and Guth, 1990; Giaramita and Sorensen, 1994; Harlow et al., 2004). Serpentinite-hosted eclogites in Papua New Guinea also crop out along highly oblique convergence zones between the Pacific and Australian plates (Baldwin et al., 2004). As for the obliquity of Farallon–North American convergence during eclogite exhumation, there is still considerable debate. During the Jurassic and Early Cretaceous, when many of the Franciscan eclogites were formed and exhumed, subduction is thought to have been oblique (Wakabayashi, 1992). However, it has also been suggested that effusive continental arc magmatism during the mid-Cretaceous required more normal convergence (Ernst et al., 2008).

The purpose here is not to resolve whether oblique convergence was involved in any of the above-mentioned regions where eclogite exhumation took place. Instead, we simply explore the consequences of oblique convergence in eclogite exhumation. We first assume that the eclogite bodies are small enough that they are passively carried by the serpentinite matrix. Hence, we consider only the flow of the serpentinite. The buoyancy of the serpentinite generates a pressure gradient that drives flow along a narrow channel. This pressure gradient is given by Δρgsinθ, where Δρ is the positive (upward) density contrast between serpentinite and mantle, g is gravitational acceleration, and θ is the dip angle of subduction (Figs. 10A, 10B). Relative to the channel, the average upward velocity of the serpentinite can be approximated by Poiseulle channel flow, and is given by:  
where H is the full width of the channel and η is the serpentinite viscosity. The average net velocity upward, Vnet, however, is the difference between the upward channel velocity and the downward-sinking velocity of the plate (e.g., a Couette component). The downward-sinking velocity of the plate is given by Vs = VPcosβ, where VP is the relative plate velocity and β is the angle of oblique convergence (0° reflects pure normal convergence and 90° represents pure strike-slip motion).

In Figures 10C and 10D, we consider one set of variables to illustrate how the time, t, needed to return eclogites from their depths of formation (Zserpentinite ~50 km), i.e., t = Zserp/(Vnetsinθ), depends on dip angle and obliqueness. We assume serpentinite to be weak (η = 1018 Pa s), a density contrast Δρ of 500 kg/m3, a full channel width of ~5 km, and a relative plate velocity VP of 10 cm/yr. It can be seen that return times are shortest for high dip angles because the pressure gradient is greatest and the trench-normal distance along the slab to Zserp is shortest. However, high dip angles are uncommon in subduction zones (though possibly common in continent-continent collisions). We thus focus our attention on lower dips, where the model is most relevant. For example, the global average dip of subducting slabs is ~23°. At these angles, the pressure gradient driving flow is lower, so the effect of oblique convergence becomes more important. For a given dip, exhumation is most rapid (on the order of several kilometers per million years) during oblique convergence (Fig. 10D) because the downward subduction component is reduced and hence competes less with the upward channel flow of serpentinite. In the extreme case of nearly pure strike-slip motion (β = 90°), the relevance of the model breaks down because there is no subduction component to generate eclogites. Our model shows that for a given combination of plate velocity and slab dip, rapid eclogite exhumation similar to that inferred from the geochronologic studies discussed in the Introduction appears to be favored during oblique subduction. Many of the parameters in this simple model can be varied considerably, but the general physics and trends presented here will not change.

Our simple calculations shown here indicate that eclogite exhumation can occur by transport in buoyant serpentinite channels. Although not a requisite for exhumation, oblique convergence can increase ascent rates, thereby decreasing the elapsed time between peak metamorphism and exhumation. We suggest that in oblique convergence zones, eclogite ascent rates are fast enough to prevent complete retrogression of high-grade metamorphic rocks.


The Ring Mountain eclogites show petrographic evidence on their rinds for retrograde metamorphism into blueschists and greenschists. Based on major element systematics, REE abundance patterns, and Nb/Ta and Nb/La ratios, the protoliths of these mafic lithologies are consistent with tholeiitic basalts (normal MORB to enriched MORB). Retrograde lithologies show two independent styles of meta-somatic enrichments, one represented by addition of LILEs (Cs, Rb, K, Ba, and Tl) and another represented by correlated enrichments of Cr, Mg, Ni, and Pb. The former are likely related to reaction with metamorphosed pelitic sediments or with fluids derived from them; the latter are best explained by reaction with serpentinites. Because all of the retrograde lithologies show the serpentinite overprint, the eclogites must have undergone reaction with serpentinite along the entire retrograde path. These obser vations are thus consistent with eclogite exhumation along serpentinite channels extending down to the depths of the eclogites, perhaps formed just above the slab-mantle wedge interface by the infiltration of dehydrating fluids from the subducting slab. Simple models show that the low viscosity and low density of serpentinite are sufficient to drive the upward flow of serpentinite. However, the rapid ascent rates needed to return eclogites to the surface without complete retrograde metamorphism are most easily attained in oblique subduction zone settings.

Grants from the National Science Foundation and the Packard Foundation supported this research. This work constitutes part of Horodyskyj's undergraduate research projects at Rice University. The comments of two anonymous reviewers, and discussions with A. Basu, S. Sorensen, J. Wakabayashi, and J. Platt are appreciated. We especially thank W.G. Ernst and H. Avé Lallemant for the generous donation of their time and discussions. We also thank R. Halama for bringing to our attention the work on retrograde eclogites in the Tian Shan.