6.1. Crustal Contamination
 Hildreth and Moorbath  argued that most of the chemical and isotopic variability of continental arc magmas develops at the crust-mantle boundary in zones of melting, assimilation, storage and homogenization (MASH), and that the processes that take place in the deep mantle are largely obscured. Magmas ascending from such zones can be later compositionally modified as they fractionate and assimilate upper crustal rocks, following Assimilation Fractional Crystallization (AFC) paths. Magmas formed in this way should be physical mixtures between mantle and lower crustal melts. Melting of the lower crust requires heat from the intruding magma, a process that induces its crystallization (i.e., differentiation), and generates a melt with higher silica than the original magma. Therefore arc signatures formed by a MASH process should be more pronounced in more evolved rocks and true basalts would rarely erupt. In contrast, magma formation by mantle metasomatism in subduction zones should initially create a basaltic melt in which the “arc signature” depends on the amount of fluxing from the slab, a process that can significantly modify the isotopic and trace element composition of the resulting melts, with only minor effects on the SiO2 and MgO contents. Therefore the chemical variations induced by subduction should be distinguishable from those associated with MASH processes or upper crustal contamination, because the physical processes of enrichment are not only different but should also bring contrasting chemical patterns, even though the participating components can have similar compositions.
 An accurate evaluation of magma genesis in continental arcs thus requires that the different components involved are known or can be reasonably inferred with the geologic information available. Unfortunately, the characteristics of continental basement rocks underneath the TMVB are largely unknown because most of them are covered by Mesozoic to Holocene sedimentary and volcanic sequences. Even so, different lines of evidence indicate that the volcanic arc is emplaced over large crustal masses or terranes, with contrasting ages and geologic characteristics [Sedlock et al., 1993; Ortega-Gutiérrez et al., 1994]. According to these studies, the basement rocks of eastern Mexico are largely Grenville age (∼1 Ga) gabbros, charnokites, anorthosites, and metapelites with metamorphic peaks in the granulite facies that belong to the so-called Oaxaquia microcontinent [Ortega-Gutiérrez et al., 1995]. Lower crustal xenoliths have been also recovered in several areas of northern Mexico and, although they are generally more mafic than the exposed terranes, their Nd model ages are generally consistent with the Grenville-age Oaxaquia microcontinent. Several studies have dealt with the geochemistry of the xenoliths and high-pressure terranes, and there is an extensive data base available in the literature [Ruiz et al., 1988a, 1988b; Roberts and Ruiz, 1989; Schaaf et al., 1994; Lawlor et al., 1999]. In general, Oaxaquia rocks show the typical characteristics of lower crustal materials: high Ba/Th and low Th/Nd ratios, negative Pb spikes, and strong positive Sr and Eu anomalies (Figure 6). Importantly, all lower crustal materials analyzed so far in Mexico have ‘enriched’ Sr and Nd isotopic compositions but unradiogenic Pb isotopes when compared to the Palma Sola rock suites (Figure 4). The closest outcrops of Grenville-age rocks are 220 km to the NW and 250 km to the south of Pico de Orizaba volcano, in Molango and Oaxaca, respectively, but Precambrian rocks have also been verified in exploratory wells along the northern Gulf coast. It is therefore usually assumed that the Oaxaquia microcontinent underlies most of eastern Mexico, even beyond the continental margin [Keppie and Ortega-Gutiérrez, 1995; Ortega-Gutiérrez et al., 1995].
Figure 6. Trace element patterns of possible crustal end-members and weighted bulk subducted sediment. Average Oaxaquia composition calculated from Lawlor et al. , weighted bulk sediment data (LaGatta et al., manuscript in preparation, 2003), and sample FO-99-1 from the local Paleozoic basement is the component used in the AFC modeling.
Download figure to PowerPoint
 In the Palma Sola area the oldest exposed rocks are Mesozoic to mid-Tertiary marine to continental sandstones, but their relative volume is small compared to the magmatic portion of the stratigraphic record. Exploratory wells have verified the existence of late Paleozoic plutonic and metasedimentary rocks underlying the Mesozoic sediments [Jacobo, 1985], and it has been proposed that rocks with similar age and lithologic characteristics extend to the northwest of the Cofre de Perote stratovolcano, in the so-called Teziutlan Massif [López-Infanzón and Torres-Vargas, 1984; López-Infanzón, 1991]. The structural base of the Teziutlan massif is made of a strongly tectonized metamorphic complex of mica-schists, plutons (granodiorites to granites) and metavolcanic rocks (andesites to dacites) that have been dated between 259 and 252 Ma [López-Infanzón and Torres-Vargas, 1984; López-Infanzón, 1991]. In order to better constrain the participation of the crust in the petrogenesis of the Palma Sola volcanic rocks, we analyzed three metamorphic samples from the structural base of the Teziutlan Massif (Tables 1, 2, and 3). The samples are metasedimentary mica-schists (sample FO-99-1) and slightly foliated and chloritized metavolcanics (samples FO-99-3 and FO-99-4). The trace element characteristics of the Paleozoic basement rocks show typical upper crustal systematics with high LILE/HFSE ratios and negative anomalies in Sr and Eu (Figure 6). Isotopic data for these rocks is also variable but they have high 87Sr/86Sr and 207Pb/204Pb, coupled with low 143Nd/144Nd, and similar 206Pb/204Pb when compared to the Palma Sola volcanics (Figure 4).
 Variations of Sr, Nd, and Pb isotopic ratios with differentiation indices are shown in Figure 7. Isotope ratios and fractionation indices of the plateau basalts are uncorrelated. Primitive rocks with OIB and arc-like trace element patterns exist with almost the same MgO content (see Figure 3), a variation that few would attribute to crustal assimilation. Field evidence also indicates that the plateau basalts were most likely fed by dyke systems, since no central volcanic structures were associated with them. Also, at least one lava flow contains abundant spinel-lherzolite mantle xenoliths. These features indicate a rapid ascent to the surface and thus participation of the continental crust in their petrogenesis appears to be negligible. We thus conclude that the compositional variations of this suite are most likely mantle derived features.
Figure 7. Variations of Nd, Sr, and Pb isotopes with differentiation indices. Plateau basalts do not show a coherent correlation between isotopic enrichment and indices of fractionation. Cinder cones and plutons do show correlations between these parameters but in opposite directions. Evolved cinder cones appear to be consistent with the assimilation of an enriched upper crustal component, similar to the local Paleozoic basement. In contrast, the plutons require a depleted lower crustal component or the participation of melts coming from the subducted MORB (see text).
Download figure to PowerPoint
 Plutonic rocks and cinder cones do show correlations among the isotopic compositions and indices of differentiation, but they follow diverging trends (Figure 7). Sr and Nd isotopic composition of the cinder cones appear to be consistent with contamination by an isotopically enriched crustal component. The 206Pb/204Pb ratios in the cinder cones do not follow a simple correlation trend with MgO or SiO2 contents, but incipient correlations can be seen in the 207Pb/204Pb and 208Pb/204Pb ratios. The poor correlations observed in the 206Pb/204Pb ratios probably indicate the lack of a significant compositional contrast between the mantle wedge (modified by subduction or not) and the crustal contaminants.
 An Assimilation Fractional Crystallization (AFC) model [DePaolo, 1981] that uses the most depleted and primitive cinder cone lava flow as starting magma composition (sample CP-35), and the average composition of the Paleozoic crust as a contaminant, fails to explain the trend of enrichment in the Sr-Nd-Pb isotopic space (not shown). If an average lower crustal composition is used as a contaminant (i.e., Oaxaquia), the AFC model predicts a trend that strongly modifies the 206Pb/204Pb isotopic composition of the contaminated rocks (orange line in Figure 8a). This is because the Palma Sola cinder cones and Oaxaquia have very different Pb isotopic compositions. Instead, the 206Pb/204Pb ratios of the cinder cones remain fairly constant with differentiation, thus making the Oaxaquia lower crust an implausible contaminant. Using the most enriched Paleozoic crustal sample as a contaminant (FO-99-1), the resulting trend passes closer to most of the data in Sr-Nd-Pb isotopic space, and also can explain the variations in concentrations and ratios of most of the incompatible trace elements (Figures 8 and 10). Even though no unique crustal end-member analyzed so far can fully describe the compositional systematics of the cinder cones, and it is somehow unclear the extent of which sample FO-99-1 represents the basement lithology of the Palma Sola area, it seems nonetheless reasonable to conclude that at least the most evolved cinder cone lavas appear to be contaminated with an upper crustal component, similar in composition to the most enriched portions of the nearby Paleozoic basement.
Figure 8. Crustal contamination models (AFC). All models assume sample CP-35 (MgO = 8.4) as starting magma composition. Green lines consider sample FO-99-1 from the local Paleozoic basement as a contaminant. Orange lines use the average Oaxaquia crust as a contaminant. All models assume a r (Ma/Mc) value of 0.8. Bulk D values used are DSr = 0.307, DNd = 0.142, DPb = 0.128, DBa = 0.104, DLa = 0.047, compatible with crystallization of 35% Ol, 34% Cpx, 20% Opx, 10% Plag 1% Ilmenite. The enriched tip of the models represent a value of Ma/Mm = 1 (a 1:1 mixture of mantle and crustal materials) and 80% of magma remaining (F). All parameters are from DePaolo . (a) AFC models of Pb-Nd isotopic compositions. Contamination with a Precambrian lower crustal composition (Oaxaquia) strongly modifies the Pb isotopic composition of the resulting melts. (b) AFC modeling of Sr-Nd isotopic compositions; (c) 143Nd/144Nd versus Ba, (d) 143Nd/144Nd versus La.
Download figure to PowerPoint
 Plutonic rocks follow an opposite correlation trend to the cinder cones, with the most evolved plutonic sample (PS-99-38) being the most isotopically depleted (Figure 7). This particular rock sample also has highly fractionated Gd/Yb ratios, low HREE concentrations and positive Sr anomalies (Figure 3a). If the geochemical characteristics of the plutons are controlled by contamination, MASH processes, or crustal melting, then the composition of the crust involved has to be isotopically depleted in Sr, Nd and Pb, and it must have retained residual garnet, otherwise the highly fractionated Gd/Yb ratios and positive Sr spikes would not be preserved. In other words, an overthickened and isotopically depleted lower crust should be invoked. Clearly, this crustal component has to be different from the local Paleozoic rocks that apparently contaminated the Quaternary cinder cones. Lower crustal xenoliths and high pressure Precambrian terranes in Mexico (i.e., Oaxaquia), although very depleted in Pb isotopes, are all significantly more enriched in their Sr and Nd isotopic composition than the plutons making them again unlikely candidates (Figures 4 and 8). On the other hand, the contemporary crustal thickness in the Palma Sola area reaches only 20 km [Molina-Garza and Urrutia-Fucugauchi, 1993], a low pressure for garnet to be stable during melting of a mafic lower crust [Rapp and Watson, 1995]. With these considerations, the geochemical characteristics of the plutons can be equally explained by: (1) assimilation and/or melting of a garnet-bearing and isotopically depleted lower continental crust that was delaminated and recycled into the mantle shortly after the plutons were emplaced or (2) melting of the subducted oceanic crust and the formation of adakite magmas [Kay, 1978; Defant and Drummond, 1990] metasomatizing the subarc mantle wedge. Given that there is little geological evidence to support the existence of a thick, garnet bearing lower crust in this area, and the delamination scenario is ad hoc with no supporting evidence, the second possibility will be further explored below.
6.2. Mantle Wedge
 The composition of the mantle wedge beneath the eastern TMVB can be investigated by comparing the isotopic compositions, and the concentrations and ratios of the HFSE and the HREE in volcanic rocks with known petrologic reservoirs. Plateau basalts from the Palma Sola area show significantly higher Nb concentrations and Gd/Yb ratios than the Pacific MORB, with values that overlap OIB reservoirs (Figure 9a). Nb/Ta ratios of the Palma Sola volcanic rocks (∼18 on average) also tend to be higher than the average EPR MORB (∼15.5 on average), overlapping only with the most enriched portions of the EPR (Figure 9b). Data arrays involving Pb isotopes of the plateau basalts do not point toward the MORB field, but rather they terminate (or begin) at high 206Pb/204Pb, 207Pb/204Pb, and 208Pb/204Pb in samples with the smallest subduction signatures (Figures 4 and 5). The most primitive plutons and cinder cones mostly share these elevated values in their Pb isotopes, indicating that this mantle end-member is contributing to some extent to all rock suites. All these characteristics suggest that the source region of the eastern TMVB is more likely related to an enriched OIB-like mantle wedge that has been compositionally modified by various subduction components.
Figure 9. Mantle source. (a) Comparison of trace element patterns: MORB, E-MORB and OIB [Sun and McDonough, 1989], average alkaline basalt from the Mexican basin and range province (Ventura, San Luis Potosí) [Lassiter and Luhr, 2001] and sample PS-99-25. (b) Nb/Ta versus Gd/Yb ratios of the Palma Sola volcanics and EPR MORB between 5°–15° N (PETDB, 2002, http://petdb.ldeo.columbia.edu/). Plateau basalts have higher Gd/Yb ratios at similar values of Nb/Ta than the rest of the magmatic suites. This is indicative of melting of a deeper region of the mantle, where garnet is more abundant. Higher Nb/Ta and Gd/Yb ratios of the most isotopically depleted pluton (sample PS-99-38) could be an indication of melts coming from the subducted MORB, in which garnet and rutile are residual phases. (c) Gd/Yb ratios do not show a coherent correlation with subduction signatures (i.e., Ba/Nb) indicating that Gd/Yb ratios are probably not a proxy for extent of melting, but provide some insight into the relative depth from which these magmas were extracted.
Download figure to PowerPoint
 Cinder cones and most of the plutons have similar Nb/Ta ratios to the plateau basalts but lower Gd/Yb ratios (Figure 9b). While elevated Nb/Ta ratios appears to be an ubiquitous mantle feature shared by all rock suites, variations toward higher Gd/Yb ratios in the plateau basalts provide some insight into the mantle melting process. Higher Gd/Yb ratios in the plateau basalts could be controlled by lower extents of melting or by a relatively larger amount of modal garnet in the source at higher pressures [Green and Ringwood, 1970]. Even though the Gd/Yb ratios alone cannot distinguish between these two process, the relative concentrations of major elements like SiO2 and Fe2O3 at constant MgO contents are sensitive to variations in pressure and water contents in the sub-arc mantle source [Gaetani and Grove, 1998]. Cinder cones and plutons have higher SiO2 and lower Fe2O3 (higher SiO2/Fe2O3 + MgO ratios) than the plateau basalts at similar MgO contents (Figure 2). Experimental studies have shown that the effect of adding water to mantle peridotite at fixed pressures is to increase the silica activity of the melt, so that water-rich mantle melts should display higher SiO2 and lower Fe2O3 contents that those formed by anhydrous melting [Gaetani and Grove, 1998]. On the other hand, these experiments showed that the same composition is obtained by decreasing pressure at constant water contents. It has been also shown that the amount of fluid fluxing into the wedge should be positively correlated with the extent of mantle melting [Stolper and Newman, 1994; Reiners et al., 2000]. Following this rationale, the geochemical features of the cinder cones and most of the plutons (with lower Gd/Yb ratios) could be consistent with H2O-rich magmas generated by larger extents of partial melting than the plateau basalts at similar pressures. Nonetheless, Gd/Yb ratios do not correlate with other indicators of fluid contributions (i.e., Ba/Nb ratios) within and among rock suites (Figure 9c). Cinder cones and plateau basalts with similar Ba/Nb ratios have different Gd/Yb ratios, while most of the plutons have higher Ba/Nb ratios at similar Gd/Yb values to the cinder cones. These systematics thus imply that the relative concentrations of SiO2 and Fe2O3, combined with Gd/Yb ratios, are not proxies for extent of melting or water contents in the source but provide an insight on the relative depth from which these magmas were extracted. We thus conclude that the geochemical characteristics of the plateau basalts are consistent with melting of a relatively deeper region in the mantle wedge where garnet was more abundant. The significant overlap in isotopic composition and Nb/Ta ratios of the most primitive cones and plutons with the plateau basalts indicates that the subarc mantle column has a similar bulk chemical composition, but that changes in mantle mineralogy at different pressures have a strong effect on the REE patterns of the resulting partial melts. We further suggest that the mantle wedge below the Palma Sola region most likely melted by a combination of processes. The existence of interstratified plateau lavas with and without subduction signatures indicates that slab fluxing into the mantle is a heterogeneous process, and that at least a small part of this wedge melted by decompression and not by fluxing.
 An exception to this general rule is sample PS-99-38, the isotopically most depleted pluton. This rock has the highest Gd/Yb and Nb/Ta ratios of the whole data set. The partition coefficients of Nb and Ta in mantle mineralogies are so similar that the ratios between these two elements should not be significantly affected by melting or subsequent fractional crystallization. The only known mineral phase that can selectively fractionate Ta from Nb is rutile in the presence of a silicate melt [Green, 1995]. Experimental studies have shown that rutile is not a likely residual phase in the source of basaltic melts, as TiO2 contents in basalts are too low, but rutile appears to be stable during partial melting of hydrous basalts and eclogites [Green and Pearson, 1986; Ryerson and Watson, 1987]. Therefore the highly fractionated Gd/Yb and Nb/Ta ratios of this sample are strong indications that both garnet and rutile were probably residual phases during partial melting of an eclogite or a garnet amphibolite source.
6.3. Subduction Components
 While most researchers agree that the source region of convergent margin volcanism is the mantle wedge, and that element fluxing from the slab is the main mechanism triggering partial melting, the specific transport process of the subduction components remains controversial. Are elements transported by hydrous fluids or by water-rich silicate melts? And if a distinction can be drawn between these two processes, which reactions are involved and under which conditions? Coupling the geological information and geochemistry of arc lavas with experimental constrains provide some insights into these questions.
 In this section we discuss the different contributions from the subduction environment to the Palma Sola volcanics. Although the following figures will show data from the Quaternary cinder cones, the interpretations and discussions regarding the subduction systematics are based on the petrologic characteristics of the plutonic rocks and the plateau basalts because, as mentioned before, there is strong evidence that the most enriched cinder cones are contaminated with the upper continental crust.
 Trace element ratios between Th, Pb, and Nd are instructive because they allow us to recognize the contributions from the different subduction components. Th and Nd are considered to be fluid immobile, whereas Pb is soluble in aqueous fluids [Brenan et al., 1995c]. Therefore the Th and Nd budgets of the magmas should be mainly controlled by their abundances in the mantle wedge plus the subducted bulk sediments or their partial melts [Plank and Langmuir, 1993; Elliott et al., 1997; Class et al., 2000]. The Pb contributions, on the other hand, can be either controlled by fluid additions and/or the additions from the subducted sediments. The melt/solid partition coefficients between Pb and Nd are very similar and thus fractionation between these two elements is negligible during mantle melting. Additionally, the use of these element ratios allows us to construct mixing plots with the isotopic ratios of Nd and Pb where the correlations are linear.
 In the Th/Nd and Pb/Nd correlation diagram (Figure 10a) all samples follow a positive correlation trend with plutonic rocks (and cinder cones) having a higher Pb/Nd at similar Th/Nd values than the plateau basalts. In general, plutonic rocks have significantly higher LILE/HFSE ratios than the plateau basalts. If the geochemistry of these sequences is not controlled by the continental crust, then we can safely assume that the plutonic rocks and the plateau basalts have been affected by two different subduction components. The compositional difference between these components is confirmed in a plot of Nd/Pb versus Pb isotopes (Figure 10b). While the plateau basalts form a tight negative correlation between the OIB-like mantle source and the bulk subducted sediments, the plutons clearly need an additional component with low Nd/Pb and unradiogenic Pb. And again, an AFC process with the local Paleozoic crust can explain the composition of the most enriched cinder cones in this diagram.
Figure 10. Subduction components. (a) Plutons have higher Pd/Nd at similar values of Th/Nd than the plateau basalts, suggesting that the two magmatic suites were affected by two different subduction components. (b) In a simple mixing diagram of Nd/Pb versus 207Pb/204Pb, the plateau basalts are bracketed by the OIB-like mantle wedge and the subducted sediments. In contrast, the plutons require an additional component with unradiogenic Pb. This component is apparently consistent with melts coming from the subducted oceanic crust (see text). Geochemical data from late Miocene Baja California adakites [Aguillón-Robles et al., 2001] also show low Nd/Pb values and plot as an extension of the Palma Sola plutons. (c) Th/Nd versus 143Nd/144Nd variation for the Palma Sola volcanics. Because Th is relatively more incompatible than Nd during partial melting of the subducted sediment [Johnson and Plank, 1999], the negative correlation array formed by the plutons and most of the plateau basalts is interpreted as additions of sediment melts into the OIB-like mantle wedge. An AFC model with the same parameters as in Figure 8 and DTh = 0.034, can explain the different trend followed by the cinder cones. (d) The sediment component also has lower Ba/Nb (and Pb/Nd) than the bulk subducted sediment. This indicates that the sediment was significantly depleted in fluid mobile elements prior to melting.
Download figure to PowerPoint
 Th/Nd ratios of plutonic rocks and most of the plateau basalts display a negative correlation with Nd isotopes (Figure 10c). The correlation trends to a higher Th/Nd ratio than the bulk sediment, a characteristic that has been previously documented as evidence that the sediment addition is in the form of a partial melt [Elliott et al., 1997; Class et al., 2000; Hochstaedter et al., 2001]. Indeed, the recently determined partition coefficients from Johnson and Plank  indicate that, during partial melting of the sediment, Th is more incompatible than Nd. These results provide strong evidence that the plutonic rocks and plateau basalts have been affected by different proportions of a sediment partial melt. The Th/Nd ratios of the cinder cones are lower than those of the plateau basalts and plutons, and do not follow a simple correlation trend. A process of assimilation with an enriched crustal component will create a trend with lower Th/Nd and can explain reasonably well the geochemical variations of the most enriched cinder cones.
 Although it seems that both sequences, plateau basalts and plutonic rocks, received contributions from a sediment melt, plutonic rocks require a component with higher LILE/HFSE ratios, lower Nd/Pb ratios and unradiogenic Pb isotopes (Figure 10b). At first sight, these characteristics might indicate the participation of a second component rich in fluid mobile elements but isotopically depleted, a component usually associated with a fluid derived from the dehydration of the subducted oceanic crust [Miller et al., 1994]. In detail, however, the injection of isotopically depleted fluids alone does not explain the strong garnet signature and the high Nb/Ta ratio of the most depleted plutonic sample (PS-99-38) because fluids do not modify these parameters. In addition, the sole injection of MORB fluids into the mantle wedge would create a mixing array connecting the composition of the OIB mantle and MORB in the Pb isotopes correlation diagrams (Figure 4). Partial melting of subducted oceanic crust could explain the high Nb/Ta and Gd/Yb ratios and the depleted isotopic compositions but, as with the fluids, the injection of pure slab melts into the wedge would create a mixing array connecting MORB and the OIB mantle in Pb isotopic space. Therefore a pure MORB melt component injected into the mantle does not fully explain the geochemical data and a more complicated scenario should be envisaged.
 The plutons form a positive correlation in the 206Pb/204Pb-207Pb/204Pb diagram (Figure 4b), bracketed by the MORB and the sediments, indicating that the component we are looking for is probably a mixture between these two end-members. Figure 10b also shows that the isotopically depleted component required to explain the geochemistry of the plutons has a very low Nd/Pb ratio. The Nd/Pb ratio of average EPR MORB is very high (∼19), thus implying that during slab melting Pb must be preferentially partitioned into the liquid phase while Nd should behave relatively more compatible. Recently published partitioning data for garnet (DNd/DPb ≈ 2.12, on average), and clinopyroxene (DNd/DPb ≈ 24, on average) with hydrous tonalitic melts [Barth et al., 2002], combined with partition coefficients for amphibole and andesitic melts (DNd/DPb ≈ 5.16, on average) [Brenan et al., 1995b] indicate that Pb should indeed be preferentially partitioned into the liquid phase during hydrous melting of an eclogite or a garnet amphibolite source. Although very few data exist in the literature for Nd and Pb concentrations and isotopic ratios in adakites, it is notable that recently published geochemical data from late Miocene Baja California adakites [Aguillón-Robles et al., 2001] also show low Nd/Pb ratios, and plot mostly as an extension of the Palma Sola plutons (Figure 10b). The Baja adakites, believed to have formed by melting of a subducted East-Pacific Rise ridge, provide additional evidence supporting the interpretation that the Palma Sola plutons were influenced by a slab-melt subduction component. Assimilation of a small proportion of sediments into the slab derived melts could have played a role in decreasing the Nd/Pb ratio of the mixture, because sediments have very low Nd/Pb values. Since the Pb isotopic composition of the most depleted plutonic sample is more radiogenic than MORB, it seems likely that some sediments were in fact assimilated by the slab melts forming hybrid magmas. Pockets of melts coming from the sediments and oceanic crust could also interact with the mantle wedge in a complex way, making it extremely difficult to recognize the exact proportions or characteristics of the initial components. Even though such a scenario is difficult to visualize and even harder to quantify, it is probably the simplest way to explain the overall geochemical features of the Palma Sola plutons.
 Plateau basalts geochemistry appears to be related to a much simpler scenario in which sediment melts alone can explain the isotopic and trace element variations. Even so, simple mixing lines connecting the most depleted sample (the OIB-like component) and the bulk sediment (or its melt) does not reproduce the plateau basalt data because the Ba/Nb and Pb/Nd ratios of the sediment component are too high (Figures 10a and 10d). Considerably lower Ba/Nb and Pb/Nd ratios are necessary to reproduce the mixing trends. This would be possible if some of the Ba and Pb were extracted from the sediment prior to melting, while Th and Nd remained immobile. This depletion of the fluid mobile elements could be an indication that the sediment involved was significantly dehydrated before entering the melting regime of the Palma Sola volcanics.
6.4. Quantitative Modeling of Subduction Components
 The discussion thus far suggests the mantle wedge below the eastern TMVB is similar to an enriched OIB-like source, and that this source can melt at different depths generating magmas with distinct major and trace element characteristics. The geochemical signatures of the plutonic rocks also record the participation of at least two different slab components (MORB melts and sediment melts), while the geochemistry of the plateau basalts indicate the participation of a dehydrated sediment melt only. The chemical characteristics of the cinder cones, in contrast, appear to be largely influenced by assimilation of the upper continental crust.
 In order to quantify the different subduction fluxes we assume that our most isotopically depleted plateau basalt sample, that shows no subduction signature (see sample PS-99-25 in Figure 3b), represents a pristine mantle melt that records the composition of the mantle source prior being modified by the subduction agents. The trace element pattern and isotopic composition of sample PS-99-25 are similar to alkaline rocks sampled in the Mexican Basin and Range province, where modifications derived from the subduction environment are negligible [Luhr et al., 1989b; Pier et al., 1989]. This composition is also similar to the average OIB of Sun and McDonough  (Figure 9a). These data and comparisons provide additional support to our assumption that the source region of sample PS-99-25 was not significantly modified by subduction. We then reconstructed the mantle composition by assuming that sample PS-99-25 is a mantle melt formed by 3% batch melting of a source made of 54% olivine, 19% Cpx, 24% Opx and 3% garnet (Table 4). Mantle mineralogy was determined using the BATCH program and the experimental constraints described by Longhi  for low degree mantle melts. The choice of 3% melting was based on comparisons of incompatible elements concentrations with the alkaline basalts form Ventura and Chichinautzin volcanic fields, assumed to represent about 1–5% melting of a relatively enriched mantle source [Luhr et al., 1989b; Wallace and Carmichael, 1999]. The model peridotite composition is similar to those used in previous studies where enriched mantle domains are considered [Borg et al., 1997; Reiners et al., 2000]. Even though this mantle reconstruction is not strictly correct, as sample PS-99-25 has most likely undergone a small amount of fractionation (MgO = 7.3 wt%), we use this composition because it is convenient for the purpose of estimating subduction fluxes in samples that have suffered similar levels of fractionation (i.e., sample PS-99-5c with MgO = 7.8 wt%, see Figure 3b). Accounting for fractionation would yield a source that is roughly parallel in shape but slightly more depleted in trace element concentrations than the one used. For this reason we emphasize that the calculated contributions from the slab are maxima.
Table 4. End-Member Compositions and Partition Coefficients Used in Models
| ||Mantle Wedgea||Dmantle/meltb||Bulk Sedc||Deh Sedd||Dsed/melte||Sed Meltf||Slabg||Dslab/melth|
|Ta|| || || || || || ||0.4||0.3676|
|87Sr/86Sr||0.703096|| ||0.70846||0.70846|| ||0.70846||0.70275|| |
|206Pb/204Pb||18.963|| ||18.687||18.687|| ||18.687||18.303|| |
|207Pb/204Pb||15.592|| ||15.601||15.601|| ||15.601||15.483|| |
|208Pb/204Pb||38.534|| ||38.423||38.423|| ||38.423||37.749|| |
|143Nd/144Nd||0.512941|| ||0.51252||0.51252|| ||0.51252||0.51314|| |
 The composition of the sediment melt (Table 4) is calculated using the weighted bulk sediment composition from the DSDP site 487 (LaGatta et al., manuscript in preparation, 2003) and assuming 5% batch melting using the partition coefficients of Johnson and Plank  (except Nb, see below). The composition of the slab melts (Table 4) were calculated using average and best fit mineral/melt partition coefficients of andesitic and dacitic liquids [Rollinson, 1993, and references therein; Green, 1995; van Westrenen et al., 2001; Barth et al., 2002; Geochemical Earth Reference Model (GERM) database, 2001, http://www.earthref.org/] and assuming that the metamorphosed subducted MORB is made of 35% Cpx, 30% Gt, 34.8% Amphibole and 0.2% Rutile. The bulk chemical composition of the subducted slab is an average East Pacific Rise MORB between 5°–15° North (Petrological Database of the Ocean Floor (PETDB), 2002, http://petdb.ldeo.columbia.edu/), with slightly higher concentrations of fluid mobile elements to account for hydrothermal alteration (Table 4).
 Since the isotopic and trace element compositional variations of the plateau basalts appear to follow simple mixing trends between the mantle source and a sediment melt, we concentrate on their characteristics before moving on to the more complex petrology of the plutonic rocks. In this case, the geochemical modifications imparted by the subduction environment are modeled by mixing different proportions of sediment melts to the reconstructed mantle source. In reality, however, sediment melts may react with the overlying peridotites, as such melts are most likely saturated with the SiO2 component [Kelemen, 1998; Rapp et al., 1999; Tatsumi, 2001]. Given that mass additions from the sediments to the mantle wedge are usually minor (<5%) [Kay et al., 1978; Bourdon et al., 2000; Class et al., 2000], and that the differences in SiO2 and MgO contents between slab-modified and unmodified samples to be modeled here are small (see samples PS-99-25 and PS-99-5c in Table 2), we assume that the mantle mineral paragenesis did not change significantly during melt/rock interaction, and thus the sediment contributions can be modeled as a simple mixing process.
 Figure 11 shows the geochemical modifications imparted to the mantle wedge by adding sediment melts in different proportions. In the Pb-Nd isotopic space (Figure 11a), if the bulk subducted sediment Pb/Nd ratio is considered (∼1.1), the resulting model predicts slightly lower 206Pb/204Pb values for a given 143Nd/144Nd ratio (orange lines). A similar phenomenon is observed in the Sr-Nd isotopic space (Figure 11b), where the use of the bulk sediment Sr/Nd ratio (∼6.1) predicts a trend with higher 87Sr/86Sr at a given 143Nd/144Nd value, with a slope that misses most of the data. These features validate the previous observation that a significant proportion of the fluid mobile elements were probably lost from the sediment prior to melting. Lower Pb/Nd (∼0.53) and Sr/Nd (∼4) ratios are required to accurately reproduce the Pb-Nd-Sr isotopic mixing relationships (pink lines in Figure 11). The data thus indicate that the sediment component needed to reproduce the Palma Sola plateau basalts has been significantly dehydrated with nearly 35% of the Sr and about 52% of the Pb lost in a fluid phase prior melting (Table 4). With these considerations, the isotopic data of the most enriched plateau basalts can be closely reproduced by adding less than 4% of a dehydrated sediment melt to the OIB-like source (Figure 11). Figure 12a shows that the trace element pattern of one of the most enriched plateau basalts (PS-99-5c, also see Figure 3b) can also be reproduced by adding ∼3% dehydrated sediment melt, and assuming about 3% batch melting of the metasomatized OIB-like mantle. The partition coefficients of Johnson and Plank  for the sediment melts thus appear to be reasonable for most of the trace elements, with the exception of Nb in which a significantly higher D (∼20) must be assumed. This further confirms the idea that rutile is a residual phase during partial melting of the subducted sediment, so that Nb, Ta and TiO2 should not be significantly transferred to the wedge by the sediment melt input.
Figure 11. Isotopic modeling of subduction components. (a) Pb-Nd, (b) Sr-Nd, and (c) Sr-Pb isotopic variations. The yellow lines in (a) and (b) describe the isotopic modifications imparted to the mantle wedge by adding different proportions of non-dehydrated sediment melts (Pb/Nd = ∼1.1 and Sr/Nd = ∼6). These models largely fail to reproduce the data of the plateau basalts (see inset in Figure 11a). The pink lines consider a model in which the sediment has lost a significant proportion of the fluid mobile elements before melting (Pb/Nd = 0.53 and Sr/Nd = 4). This model accurately describes the isotopic variations of the plateau basalts by adding less than 4% of a dehydrated sediment melt (see inset in Figure 11a). The blue lines are simple mixing models of the altered oceanic crust (AOC) and sediments sampled in DSDP site 487 [Verma, 1999b; LaGatta et al., manuscript in preparation, 2003]. These models fail to reproduce the isotopic composition of the most depleted plutonic sample, suggesting that this AOC is probably not representative of the bulk oceanic crust being subducted (see text). The green lines are simple mixing models of an average EPR MORB, with a slightly higher Sr contents and 87Sr/86Sr ratio (∼0.70275), and the dehydrated sediments. Using these end-members, the most isotopically depleted plutonic sample can be reproduced by a 20:80 sediment:MORB mixture.
Download figure to PowerPoint
Figure 12. Trace element modeling. (a) The trace element pattern of sample PS-99-5c (see also Figure 3) can be closely modeled by melting an OIB mantle source metasomatized with ∼3% dehydrated sediment-melt, and assuming ∼3% batch melting. Sediment melt partition coefficients are from Johnson and Plank , except DNb = 20. Mantle/melt partition coefficients in Table 4. The unusual flat pattern in the normalized Rb-Ba-Th-U is an indication of the dehydrated sediment due to a significant Ba loss in a fluid phase prior to melting. (b) The trace element pattern of sample PS-99-38 can be reproduced by melting a 20:80 sediment:MORB mixture, using partition coefficients of andesitic to dacitic melts (Table 4). Slab melt residual mineralogy is assumed to be 0.35 Cpx, 0.3 Gt, 0.348 Amph, 0.002 Rut. High Nb/Ta = 22, La/Yb = 46 and Sr/Y = 53 ratios for this sample are all consistent with slab melting.
Download figure to PowerPoint
 As mentioned before, the geochemistry of the plutonic rocks appear to be related to a complex mixture of slab and sediment melts metasomatizing an OIB mantle wedge. Thus modeling the proportions of the components involved in such a scenario is not a simple task. There are also many uncertainties regarding the composition of the altered oceanic crust that is being subducted. Verma [1999b] reported trace element data and isotopic compositions of altered oceanic crust (AOC) from the Cocos plate collected at DSDP site 487, and a larger trace element data set was obtained by LaGatta et al. (manuscript in preparation, 2003) on the same samples. Figure 11 shows that a mixing line connecting this AOC composition and the dehydrated sediments fails to reproduce the isotopic data of the most depleted plutonic samples because the Sr/Nd and Sr/Pb ratios of the AOC are lower than those required to satisfy the mixing relationships (blue lines in Figure 11). This AOC composition is highly depleted in REE when compared to NMORB and even more so against average EPR MORB. It is also notable that the 143Nd/144Nd ratio of this AOC is higher than most of the fresh EPR MORB samples (Figure 11). These features suggest that the samples used to define AOC, collected in the upper couple of meters of the oceanic crust, are probably not representative of the bulk oceanic crust being subducted. We thus consider it more reliable to use the extensive EPR MORB database available in the literature as a MORB end-member (PETDB, 2002, http://petdb.ldeo.columbia.edu/). If the average composition of EPR MORB with a slightly higher Sr content and 86Sr/87Sr ratio is considered as a slab end-member (Table 4), the mixing line passes close to the most depleted plutonic sample with a mixture made of 20% sediment and 80% MORB (green lines in Figure 11). A batch melting model of that mixture, that uses andesitic to dacitic partition coefficients (Table 4), can reproduce the general trace element pattern of sample PS-99-38, the most depleted pluton (Figure 12b). Strong garnet signatures, high Nb/Ta ratios, positive Sr anomalies (thus high Sr/Y ratios), and the most depleted isotopic composition compared to the other plutonic rocks all suggest that this particular sample could be related to a mixture between subducted MORB melts and sediment melts [Defant and Drummond, 1990]. The rest of the plutons could then be associated with mantle melts metasomatized with both sediment melts and hydrous slab-melts in various proportions.