The compositional ranges of regional hotspot-related volcanic products have been attributed to mixing of three or more end-members in addition to a depleted MORB source (White et al., 1993; Kurz and Geist, 1999; Harpp and White, 2001; Blichert-Toft and White, 2001; Harpp et al., 2005). One end-member (most prevalent in lavas from the Wolf-Darwin Lineament) accounts exclusively for elevated Pb isotope systematics but a small amount of total variation (3.3%) (Harpp and White, 2001; Harpp et al., 2005). As this data set does not yet exist for our samples, the remaining regional geochemical end-members reduce to (1) depleted, MORB source; (2) enriched, plume source; and (3) elevated incompatible trace element (ITE), Sr radiogenic (Bow, 1978; Bow and Geist, 1992; Geist, 1992; Graham et al., 1993; Harpp and White, 2001; Blichert-Toft and White, 2001). The locus of this last potential end-member ("FLO" of Harpp and White, 2001) is thought to be beneath Isla Floreana in the southern Galápagos archipelago, but it may not represent a signal of the plume (Kurz and Geist, 1999).
Neodymium isotopes are used here to help constrain mixing relationships between the aforementioned geochemical mantle end-members and as a prerequisite for the identification of reasonable parental trace element composition that will, in turn, be used to model petrogenesis. Figure F10 plots Sr-Nd isotopic values of samples from Subunits 4A and 4B from Leg 205 and fields of the EPR, CNS, and the Galápagos Islands, along with tholeiitic seamounts north of the Cocos Ridge axis proximal to the locations of Legs 170 and 205. The restricted range in Nd isotopes (0.512937–0.512974, N = 25; one value extends to 0.513020) indicates that Subunits 4A and 4B sample an approximately uniform 143Nd/144Nd source despite the wide regional isotopic variation. This mantle source is more isotopically enriched (less radiogenic Nd) than that of EPR MORB and is likely a mixture of more and less enriched mantles, similar to the source for the Galápagos Islands, regional seamounts, and the Cocos and Carnegie Ridges (White et al., 1993; Hoernle et al., 2000, 2002; Harpp and White, 2001; Harpp et al., 2005). The overlapping fields suggest widespread mixing, as has been argued previously for the CNS (White et al., 1993; Hoernle et al., 2000). An extended discussion on the relationship between the enriched signatures of Subunits 4A and 4B with those of seamounts on the flanks of the EPR existing far from plume influence (Nui and Batiza, 1997; Nui et al., 2002) is beyond the scope of this paper. However, a link between Subunits 4A and 4B and the Galápagos hotspot is suggested from geochemical similarities and the proximity of the hotspot and the paleolocation of emplacement. Though not conclusive in resolving the source of enrichment, we note that the 143Nd/144Nd ratios of Subunits 4A and 4B (only one value above 0.513000) are more enriched than the vast majority of EPR flank seamount lavas (143Nd/144Nd = 0.512956–0.513183) (Nui et al., 2002).
Assuming that elevated (higher 87Sr/86Sr) ratios seen in Leg 205 samples is due to seawater interaction, and therefore not primary, Nd isotopes are used to calculate mixing proportions between depleted MORB mantle and the enriched Galápagos end-member. The majority of Leg 205 samples are derived from a mantle source that is 50%–70% enriched; one sample requires ~30% enrichment to explain elevated 144Nd/144Nd = 0.513020. If the mantle source is isotopically homogeneous as argued, then variations observed in minor and fluid-immobile incompatible trace elements are most likely due to differences in degrees of partial melting and/or fractional crystallization. The premise that elevated Sr isotope values are due to seawater interaction rather than admixture of the FLO end-member is supported by several independent lines of evidence. Most Sr isotope excursions are correlated with increased ratios of fluid-mobile:less-fluid-mobile trace elements and measurements of increased fracture density (Pfender and Villinger, this volume; Chavagnac et al., in prep), suggesting a secondary origin. The trend of Leg 205 Sr-Nd isotopes is primarily horizontal toward elevated Sr isotopic values with very limited variation in Nd isotopic values. This is in contrast with trends predicted by modeling mixing of the FLO end-member with a uniform mantle source, itself composed of depleted MORB and enriched plume components. Furthermore, 87Sr/86Sr values of several samples extend beyond the range of values expected from a three-component mixture (Fig. F10). We conclude that compositions of Leg 205 samples are dominated by binary mixing of enriched and depleted sources; this question will be revisited with trace elements.
Regional geochemical end-members have characteristic trace element abundances. Figure F11A plots data from regional basalts and illustrates end-member mixing relationships in terms of ratios of chondrite-normalized (La/Sm)N vs. Hf/Ta, forming the hyperbolic "mantle mixing" trend. Samples from Subunits 4A and 4B are well separated; ratios of additional immobile incompatible elements (e.g., Nb/Y and Zr/Y) give similar results. Generally, Subunit 4A has (La/Sm)N >1.75 and Hf/Ta <3.5, and samples from Subunit 4B have (La/Sm)N <1.5 and Hf/Ta >3.5, with the exception of one sample (Hf/Ta ~2.5). Leg 170 and 205 samples cluster toward the enriched end-member and lie within a field outlined by the Galápagos Islands. Samples from the "main series" from Isla Floreana of the Galápagos Islands trend toward anomalously high (La/Sm)N, the locus of proposed ITE-rich FLO end-member. Melt modeling, presented in detail in the following section, shows that increasingly small degrees of partial melting of a hybrid mixture of enriched and depleted sources would approach the high (La/Sm)N, low Hf/Ta of the ITE-rich Isla Floreana samples, but melt processes alone are unable to explain the distinctive isotope systematics of this end-member (namely, high 206,208Pb/204Pb, high 87Sr/86Sr, low 143Nd/144Nd, and relatively high 176Hf/177Hf) (cf. Kurz and Geist, 1999). Variable decoupling between isotopes and ITE ratios during igneous processing of Galápagos hotspot products is discussed in Harpp et al, (2005). Without more definitive Pb and Hf isotope data, we tentatively conclude that the FLO end-member is not a significant component in the source of Leg 170 and 205 samples and have instead focused, therefore, on two-component mixing between enriched and depleted end-members. In sum, both Nd isotopes and ratios of ITEs indicate that source mixing is an important process in both Leg 170 and 205 samples. Overlap of some samples from the Galápagos Islands with a few of the most enriched CNS and EPR basalts, observed with both isotope (Fig. F10) and trace element systematics (Fig. F11A), suggests regional influence of the plume and smearing of the source signals (cf. White et al., 1993; Harpp and White, 2001; Blichert-Toft and White, 2001, and references therein).
In the binary Sr-Nd isotope mixing model presented above, we assumed that the enriched and depleted mantle end-members mix as solid materials. We have also examined the possibility that one or both of the parental sources of Subunits 4A and 4B magmas were melts themselves, in which case the resulting mixing proportions and trace element compositions will differ. To test these possibilities we also modeled (1) an enriched melt mixing with solid depleted ambient mantle and (2) an enriched melt mixing with a depleted melt. Because an enriched melt will have comparatively large abundance of trace elements, admixture of this end-member is highly restricted because the resultant hybrid magma must ultimately reproduce observed Nd isotope values. In this case, Leg 205 samples would represent a mixture of ~5% enriched melt with 95% depleted ambient mantle. In the second case of an enriched melt mixing with a depleted melt, resultant mixing proportions change only modestly compared to solid-source mixing if extents of melting are less than a few percent for both enriched and depleted sources. In this scenario, Leg 205 samples would represent 30%–50% enrichment over ambient depleted mantle, as approximated by average oceanic island basalt (OIB) and MORB of Sun and McDonough (1989). In both cases, subsequent modeling of reasonable degrees of melting and crystallization cannot reproduce the observed range in ITE abundances and ratios of Subunits 4A and 4B. Therefore, we consider these scenarios unlikely compared to solid source mixing, which is discussed in the next section.
Given the approximately uniform Nd isotope source of Subunits 4A and 4B, variable degrees of partial melting and fractional crystallization are potential causes of their observed geochemical variation. Melt generation models using ratios of incompatible trace elements are examined in conjunction with previously established end-member mixing proportions to deconvolve the roles of source mixing and subsequent igneous processing. In the following sections, we evaluate these combined functions in the generation of the distinct trace element compositions of Subunits 4A and 4B. Figure F11B superimposes models of binary mixing, partial (batch) melting of ~60% enriched source, and subsequent fractional crystallization (see "Fractional Crystallization and Crystal Accumulation") with data from Subunits 4A and 4B as well as regional basalts; Plank and Langmuir (1992) demonstrate that batch melting closely reproduces a diverse range of complex melting mechanisms and approximate the net effects of the melting process remarkably well.
The data for Subunits 4A and 4B plot along or near the modeled partial melting curves and at higher (La/Sm)N and lower Hf/Ta of the modeled mantle (PLUME-DMM) mixing trend. Both garnet and garnet-free (spinel) source mineralogies were considered, which may affect fractionation within the REEs and Hf from Ta during partial melting because of their contrasting incompatibilities in the presence of garnet. Additional melt modeling (not shown) using the REE concentrations of Subunits 4A and 4B indicates that a small, early formed proportion of the total melt occurring in the presence of garnet can transmit the characteristic slope of the HREE to the subsequent melts. Although published melting models for the vast majority of Galápagos Island seamounts and the Cocos and Carnegie Ridges (Harpp and White, 2001; Harpp et al., 2005) indicate the majority of melting in the presence of garnet, the gentle slope of normalized HREE patterns for both Subunits 4A and 4B suggests that the bulk of melting did not occur in the garnet stability field (i.e., >60 km) (Fig. F9).
Combined results suggest that the majority of Subunit 4A and 4B samples are adequately explained by 2%–7% partial melting of a hybrid source containing 50%–70% of the enriched end-member (Fig. F11B; see caption for model details). Furthermore, Subunit 4A is distinguished from Subunit 4B by lower degrees of partial melting; a difference of at least 2% melting separates Subunit 4A from Subunit 4B (Fig. F11B). Offset linear trends of V vs. TiO2 (Fig. F7) and variable ITE ratios (such as Hf/Ta) (Fig. F8) at similar Mg# are also consistent with variations in degree of partial melting. More sophisticated models of polybaric partial melting that consider the compositional dependence of partition coefficients have yet to be applied, but it appears that Subunits 4A and 4B are not associated with extensive melting at lithospheric levels deeper than typical MORB despite geochemical evidence for hotspot overprinting. These conditions place limitations on melt generation in proximity to the plume (see "Models of Sill Emplacement"). In the next section, we discuss the causes of variable trace element abundances within the igneous units.
Petrologic relations and geochemical variation within Subunits 4A and 4B indicate that they are likely to have experienced a significant degree of fractional crystallization. Previously, we have shown that the isotopic and ITE ratio compositions of the majority of Leg 170 and 205 samples can be modeled with a restricted range of mantle mixing and melting. These geochemical parameters were chosen, in part, based on their relative insensitivity to the degree of shallow fractional crystallization. Models indicate that reasonable degrees of fractional crystallization of basaltic liquids are not responsible for the observed differences in ITE ratios between Subunits 4A and 4B (Fig. F11B, inset). For example, 50% modal fractional crystallization from a basaltic melt results in an increase of La/Sm of ~3% in the remaining melt, whereas the minimum difference between Subunits 4A and 4B is ~20%. However, a two-fold range in the abundances of the REE with similarly shaped patterns is evidence of additional variation with Subunit 4B, which we relate to the effects of fractional crystallization and crystal accumulation.
In the preceding discussion, we have shown that the majority of samples are modeled as a partial melts of a mixture of enriched and depleted mantle sources. These melts produce magmas with elevated REE abundances, adequately matching compositions observed in the majority of samples from Subunits 4A and 4B. However, Subunit 4B samples with the lowest REE abundance cannot be reconciled without additional consideration. We suggest that the compositional differences within Subunit 4B can be related by crystal accumulation. As envisioned, ITE-rich melts are effectively "diluted" by the addition of variable amounts of early-formed fractionates that are relatively ITE poor (Fig. F12). Whole-rock geochemical data are used to place plausible constraints on this process. We are able to reproduce the entire REE abundance range of Subunit 4B by adding increments of an instantaneous fractionating phase assemblage to the Subunit 4B sample with the highest REE abundance. The modeled fractionating phase assemblages (plagioclase = 0.57; clinopyroxene = 0.25; olivine = 0.09; orthopyroxene = 0.08; magnetite = 0.01) are consistent with approximate modal abundances within Leg 170 and 205 samples. The La/Sm ratio of the modeled bulk partition coefficient is 1.3, indicating that little LREE fractionation would occur during this relatively small interval of crystallization (Mg# = 0.63–0.51). This is consistent with the observed (La/Sm)N ratios of Subunit 4B samples (1.24–1.51). Samples with the lowest REE abundance occur within the interval 460–509 mbsf and exhibit the highest whole-rock Mg# and most positive Eu anomalies (up to Eu/Eu* = 1.15), which may be consistent with these samples having a higher proportion of earlier formed minerals. However, there are no reports of a distinct change in the crystallinity or modal mineralogy from onboard visual core descriptions of this interval (Morris, Villinger, Klaus, et al., 2003), suggesting that the differences in the bulk compositions may be due to variations in mineral chemistry rather than mode. Without geochemical analyses of individual phenocrysts we are unable to more quantitatively model this process.
Geochemistry and tectonics together provide constraints on mechanisms for sill emplacement and its regional impact. Trace element and Nd isotope systematics indicate that both subunits cored during Legs 170 and 205 share geochemical similarities with other volcanic products of the Galápagos hotspot. At the time of intrusion, after 15.6–18.2 Ma, the paleolocation of Subunit 4A was quite distant from both the EPR (~400 km) and the CNS (~140 km) (Fig. F1). Distal off-axis emplacement is unlikely because morphological, petrographic, and geochemical studies of the EPR show that the vast majority of off-axis volcanism occurs within ~10 km of the axis (Wilson, Teagle, Acton, et al., 2003; Reynolds and Langmuir, 2000; Sims et al., 2003), although some melt production extends >100 km from the fast-spreading EPR axis at 17°–19°S (Scheirer et al., 1998). Although precise ages are yet unknown, existing petrographic and geochemical similarities imply a link between Subunits 4A and 4B, which are unlikely to represent true EPR basement. Similarly, at its nearest point to the Galápagos hotspot at 14–12 Ma, the paleolocation of Legs 170 and 205 was still several hundred kilometers to the northwest (see Meschede and Barckhausen, 2000, 2001; Harpp et al., 2005) (Fig. F1). If Subunits 4A and 4B are related to the Galapagos hotspot, then it is likely that they are far-field manifestations of plume-ridge interaction.
A possible explanation for the age-location relationship of the igneous complex is distal emplacement of a melt originating near the hotspot. As envisioned, preexisting lithospheric weakness resulting from the initial fracture of the Farallon plate and/or prior ridge jumps on the CNS may have permitted enhanced plume influence at greater distances. A similar scenario may explain a series of alkalic and tholeiitic OIB-like ~14-Ma seamounts that are clustered north of the Cocos Ridge near an extinct spreading center of the CNS (Barckhausen et al., 2001), and this age is consistent with an intrusion age after 15.6–18.2 Ma for Subunit 4A. Geochemical constraints based on mixing and melting models discussed above restrict this style of (enriched) melt-induced mixing with ambient MORB-source mantle and argue instead for solid source mixing prior to melting.
Alternatively, the generation and emplacement of the Leg 170 and 205 igneous complex may involve melting of upper mantle that was previously enriched. This enrichment could exist as Galápagos plume material entrained within the ambient depleted mantle. An age of 8–10 Ma brackets a high-amplitude seismic reflector associated with widespread deformation of the Cocos plate called the sill intrusion zone, which includes the locations of Leg 170 and 205 (Silver et al., 2004) (Fig. F2). Although controversy surrounds the cause of the deformation, this age may be coincident with the Cocos Ridge collision with the MAT (Abratis and Woerner, 2001, but cf. Graefe et al., 2002), leading to the development of a stress regime that may have fostered renewed decompression melting on a local to regional scale. Such a scenario is consistent with a recently published model by Harpp et al. (2005) regarding the tectonic-petrogenetic formation of the Cocos and Carnegie Ridges and associated seamounts. This model suggests that postabandonment alkalic and tholeiitic volcanism immediately followed and was probably caused by either a ridge jump or rift failure and not the direct activity of the Galápagos hotspot.