Petrogenesis and geodynamic significance of silicic volcanism in the western Trans-Mexican Volcanic Belt: role of gabbroic cumulates
- First Online:
- Cite this article as:
- Petrone, C.M., Orozco-Esquivel, T. & Ferrari, L. Contrib Mineral Petrol (2014) 167: 1006. doi:10.1007/s00410-014-1006-6
- 647 Views
In the western Trans-Mexican Volcanic Belt voluminous silicic volcanism has been associated with the rifting of the Jalisco block from mainland Mexico. Rhyolitic volcanism started at 7.5 Ma after a major pulse of basaltic volcanism aged 11–8.5 Ma associated with slab detachment. This was followed by a second period, between 4.9 and 2.9 Ma, associated with rhyolitic domes and ignimbrite coexisting with basaltic volcanism. The similarity in rare earth element contents between basalts and rhyolites excludes a simple liquid line of descent. The low Ba and Sr contents and the ferroan character of the rhyolites suggest extensive fractional crystallization. Late Miocene–early Pliocene rhyolite Sr isotope values are only slightly more radiogenic than the basalts, whereas Nd isotope ratios are indistinguishable. We successfully modelled the 7.5–3 Ma silicic magmatism as a result of partial melting of crustal gabbroic complexes that we infer to have formed in the mid-lower crust due to the high-density Fe-enriched composition of the late Miocene basaltic volcanism. Slab rollback since ~7.5 Ma favoured decompression melting and arrival of additional mafic magmas that intruded in the lower crust. These basalts heated and melted the gabbroic complexes forming the silicic magmas, which subsequently underwent assimilation and fractional crystallization processes. The first silicic pulse was emplaced during a period of low tectonic activity. Extensional faulting since the Pliocene favours the eruption of both silicic magma and lesser amount of mafic lavas.
KeywordsSilicic–bimodal volcanismFerroan magmaLow-Sr rhyolitesTholeiitic trendGabbroic cumulates meltingTrans-Mexican Volcanic Belt
The generation of silicic volcanism in arc environment is mainly attributed to two main processes or to a combination of them: differentiation of primary basaltic magmas through fractional crystallization in the crust (e.g., Gill 1981; Grove et al. 2003); partial melting of pre-existing crustal rocks (e.g., Ruiz et al. 1988; Bindeman et al. 2008; Bryan et al. 2008). Both of these two end members have been proposed for the genesis of rhyolites in the Taupo Volcanic Zone, one of the largest Quaternary rhyolitic provinces, and of the Sierra Madre Occidental of western Mexico, the largest Cenozoic silicic igneous province. In the case of Taupo, rhyolites have been related to differentiation from andesitic precursors (McCulloch et al. 1994; Price et al. 2005) or to the melting of greywacke and plutonic protoliths in the upper crust (Ewart and Stripp 1968; Charlier et al. 2005). For the Sierra Madre Occidental it has been suggested that the large volumes of rhyolites formed by fractional crystallization of basaltic or andesitic parental magmas with limited (20 %) crustal assimilation (AFC; Cameron et al. 1980; Lanphere et al. 1980; Wark 1991; Smith et al. 1996) or that crustal melting may have been the dominant process of rhyolite genesis (Albrecht and Goldstein 2000; Ruiz et al. 1988, 1990; Bryan et al. 2008; Bryan and Ferrari 2013).
A key problem with any fractional crystallization model for the origin of rhyolitic magmas is represented by the required volume of cumulates crystallised from the parental magma involved in the process, which is more than twice that of the produced evolved magma (Huppert and Sparks 1988; Annen et al. 2006). This process implies the formation of large volume of mafic cumulates, which is often not supported by geophysical or geological evidence (Annen et al. 2006). Density-driven sinking (Glazner 1994) and delamination (Jull and Keleman 2001) of the mafic cumulates have been suggested as a possibility to overcome the volume problem. However, crustal delamination can be invoked only in a number of locations and it may also induce crustal melting, as in the case of the central Andean silicic province (Kay and Kay 1993). Partial melting of pre-existing crustal rocks has also some limitations. Numerical models have shown that the heat budget necessary to produce large volumes of silicic magma via crustal melting alone is so high that allows melting only of the lowermost crust, or re-melting of young intrusive rocks (Annen et al. 2006; Brown 2007).
Recent models envisage a mixed scenario in which arc-related intermediate to silicic magmas are generated through both fractional crystallization of H2O-rich parental basalt and partial melting of surrounding crustal rocks favoured by heat and H2O transfer from the cooling basalts (Annen et al. 2006). One critical point with this and similar models is linked to the generation of intermediate compositions (i.e., andesites) in close genetic relationship with silicic arc magmas. In cases where intermediate compositions are lacking and a pure bimodal (basalt-rhyolite) association is present, all proposed genetic models are questioned and no simple solution to the genesis of silicic and bimodal magmatism in subduction-zone settings can be put forward.
On the basis of geochemical evidence Nelson and Hegre (1990) proposed that the peralkaline rhyolites of the mid-Pleistocene Las Navajas volcano might be the results of fractional crystallization of large amounts of alkaline mafic lavas widely distributed in the area. By contrast, in the only isotopic study published so far on the rhyolites, Mahood and Halliday (1988) suggested that the peralkaline rhyolites of the La Primavera Caldera might be related to partial melting of Mesozoic or Tertiary mafic intrusive rocks or lower-crustal metamorphic equivalents. Partial melting of Tertiary ignimbrites or their plutonic equivalent has been also proposed for the thick succession of Pliocene rhyolites by Frey et al. (2007). Indeed, they suggested that the low Sr abundances of the Pliocene rhyolites required a source with moderate Sr that would leave a feldspar-rich residue on partial melting. Whereas this model is consistent with the trace element provided by these authors, their main supporting argument was the occurrence of one Late Oligocene sanidine crystal in one sample, which can be hardly considered representative for an estimated volume of ~500 km3 of rhyolites.
In this paper we present new mineralogical, geochemical and isotopic data for the silicic magmatism that affected the western part of the Trans-Mexican Volcanic Belt starting at 7.5 Ma following an unusual pulse of mafic volcanism at 11–8.5 Ma and preceding the establishment of more typical arc-magmatism since ~3 Ma. Using our new data and those from the literature we propose a new petrogenetic model for the genesis of the late Miocene–early Pliocene silicic magmatism, which, for the first time, successfully reproduce both trace elements and isotope data. Our petrogenetic model is also consistent with the geologic and geodynamic evolution of the western part of the arc. New data on late Pliocene–Quaternary rocks will be also presented in this paper, but we will use these data only for comparison as the petrogenesis of these rocks has been discussed in previous works (e.g., Petrone et al. 2003, 2006; Petrone and Ferrari 2008; Petrone 2010).
The geology and petrogenesis of the WTMVB as well as the geodynamic evolution of the region has been described in Ferrari et al. (2000, 2003, 2012), Gomez-Tuena et al. (2007) and Petrone et al. (2003, 2006). A brief summary of the geological evolution of the region is given here.
The WTMVB is characterized by a system of WNW-ESE trending extensional faults collectively forming the Tepic-Zacoalco rift (TZR), which have been active at least since the Pliocene and have controlled the emplacement of mafic volcanism (Ferrari and Rosas 2000; Ferrari et al. 2012) (Fig. 1). Volcanism shows a marked across-arc difference with mostly mafic lavas in the volcanic front and a bimodal volcanism in the rear part of the arc. The volcanic front has migrated toward the trench since the end of Miocene but volcanism remained active along the TZR making the arc wider. Volcanic activity began in the rear part of the arc with a major pulse of mafic volcanism (~1,800 km3) at ~11–8.5 Ma (Fig. 1). This was followed between 7.5 and 5 Ma by the emplacement of silicic domes and minor pyroclastic flows (~370 km3, Ferrari et al. 2001) to the north of Guadalajara (Rossotti et al. 2002), in the area with the largest volume of mafic lavas of the previous episode. Since the beginning of the Pliocene, minor amount of high-Ti basalts erupted at several places along the TZR followed shortly after (4.9–2.9 Ma) by large amount of rhyolitic lavas and ash flow tuffs (~500 km3, Ferrari et al. 2001) distributed in a 30 km wide belt along most of the arc. A new episode of mostly effusive rhyolitic volcanism (~430 km3, Ferrari et al. 2001) occurred in the Pleistocene, mainly between Tequila and Guadalajara (Fig. 1). Peralkaline rhyolites of La Primavera caldera and the Las Navajas volcano represent the last episodes of rhyolitic volcanism and are located at the southeastern and northwestern end of the western arc, respectively. Stratovolcanoes of intermediate composition and truly alkaline basalts (Na-rich) (Petrone et al. 2003) erupted only in the last 1 Ma along the main extensional fault systems of the TZR.
The mineral composition of five selected samples (three late Miocene rhyolites; two early Pliocene rhyolites) was determined at The Natural History Museum (NHM) of London using a CAMECA SX100 electron microprobe equipped with five WDS spectrometers and one EDS spectrometer. The electron microprobe was operated at 20 keV and 20 nA current with 10 μm beam diameter for feldspar and amphibole analysis and variable peak counting time: 10 s for Na and K, 30 s for F, Cl, Sr and Ba and 20 s for all the other elements. Calibration was performed using international USGS and internal standards. Matrix effects were corrected using the Cameca X-PHI protocol. Data figures given in Online Resources 1–3 are in accordance with analytical precision.
Geochemical and isotope data were obtained in the course of several years and some of data used in this study have been already published (Petrone et al. 2003, 2006; Petrone 2010). Details of the analytical techniques are thus given only for new and unpublished data. Whole rock major elements and some trace elements have been analysed by X-ray flourescence (XRF) and traditional wet chemistry for Na2O and MgO at the Dipartimento di Scienze della Terra, Università degli Studi di Firenze. In this same laboratory FeO was determined by titration and the loss on ignition (LOI) via gravimetry. A set of samples (12 samples) was analysed for trace elements in the Actlab commercial laboratory in Canada. The complete set of trace elements was determined at the Centro de Geociencias (CGEO), UNAM, Queretaro, Mexico (17 samples) using a Thermo Series XII instrument. Samples were prepared in a clean lab according to the procedure given by Mori et al. (2007). Calibration and data reduction were based on digestion of five international rock standards (BHVO-2, AGV-2, BCR-2, JR-1 and JB-2), repeated in-house rocks standard (MAR, ZZ) (Gomez-Tuena et al. 2003) and blanks that followed the same chemical procedure as the samples.
Sr and Nd isotope ratios were measured at the Isotope Geochemistry Laboratory of the Dipartimento di Scienze della Terra, Università degli Studi di Firenze using a Finnigan TRITON© thermo-ionisation mass-spectrometer equipped with nine Faraday cups. Chemical analytical procedures and instrument conditions are the same as those given in Avanzinelli et al. (2005). All Sr isotope ratios are corrected for mass fractionation to 86Sr/87Sr = 0.1194, whereas Nd isotopic ratios are corrected for mass fractionation to 146Nd/144Nd = 0.7129. 87Sr/86Sr value for SRM987 measured during the course of these analyses was 0.710262 ± 13 (n = 15), with current lab average 0.710266 ± 13 (n = 36) and compared to the reported value of 0.71025. 143Nd/144Nd values for La Jolla standard measured over the course of this work were 0.511843 ± 14 (n = 7), with current average 0.511844 ± 7 (n = 10). Sr analytical blank measured over the course of this work was 108 pg. The complete set of new geochemical and isotope analysis is given in Online Resource 4 and the location of the analysed samples is given in Online Resource 5.
Petrography and mineral chemistry
Late Miocene basalts are characterized by sub-aphyric to porphyritic-seriate textures. The main mineral phase is generally olivine followed by plagioclase and minor clinopyroxene. The groundmass ranges from ophitic to microcrystalline with the same mineral phases found as phenocrysts along with opaque minerals and minor amount of glass. Secondary calcite and chlorite has been found in some samples.
Late Miocene rhyolites range from aphyric to porphyritic textures with phenocrysts of sanidine, quartz, opaque minerals and minor fayalite in a vitrophyric groundmass that frequently shows devitrification textures. Femic minerals are overall very scarce although rare microphenocrysts of amphibole are also present in a limited number of samples. Phenocrysts of sanidine and quartz are euhedral, whereas fayalite shows resorption textures. Zircon and apatite are found as accessory phases, and secondary clay minerals are found in some samples.
Early Pliocene sub-alkaline and transitional basalts are characterized by porphyritic textures with variable degree of porphyricity from low (<5 vol%) up to medium (~20–30 vol%) with phenocrysts of plagioclase and olivine and microphenocrysts of plagioclase + olivine + opaque and minor pyroxene. Groundmass is made up by the above mineral phases and range from micro- to crypto-crystalline. Rare megacrysts of plagioclase have been found in some samples. They are mostly euhedral with no evident compositional zoning. The main difference between sub-alkaline and transitional basalt is the lack of pyroxene in the transitional basalts.
Early Pliocene rhyolites show low to medium porphyritic textures with plagioclase + quartz and absence or very minor amounts of K-feldspar. Femic minerals are represented by clinopyroxene and a slightly pleochroic pale yellow to pale green ferrosilitic orthopyroxene. Microphenocrysts of opaque minerals are common. Apatite is a common accessory phase. Groundmasses are generally micro- or crypto-crystalline with abundant glass and frequent devitrified textures.
Late Pliocene–Quaternary mafic to intermediate sub-alkaline rocks have seriate porphyritic textures with phenocrysts of plagioclase, orthopyroxene, olivine and clinopyroxene in a hypocrystalline groundmass. Opaque minerals are found only as micro-phenocrysts. Plio–Quaternary transitional mafic rocks show the same seriate porphyritic textures as the sub-alkaline rocks even though with a lower porphyritic index (5–10 vs. 20 vol%, Petrone et al. 2003) and rare orthopyroxene. The Na-alkaline Plio–Quaternary mafic rocks are characterized by pheno- and micro-phenocrysts of olivine and plagioclase in a hypocrystalline groundmass.
The studied late Pliocene–Quaternary rhyolites are porphyritic with glassy groundmass. In few cases microcrystalline groundmass is present. The main mineral phases are represented by plagioclase and quartz in the majority of the cases, with some rhyolites characterised by the presence of sanidine instead of plagioclase. Femic phases are not abundant and are mainly represented by opaque minerals and by ferrosilite orthopyroxene slightly pleochroic form pale yellow to pale green. Opaque minerals are commonly found as microphenocrysts. Zircon is a relatively common accessory mineral phase. Devitrified textures are frequently observed in the glassy groundmass, which often is altered to clay minerals.
Mineral composition has been determined in three late Miocene rhyolites and two early Pliocene rhyolites. There are some differences observed with time. Indeed, clinopyroxene is only found as microphenocrysts and groundmass microlite in the early Pliocene rhyolite (SPC 166). Fe–Mg minerals are augite in composition En39Fe20Wo41 (En = enstatite; Fe = ferrosilite; Wo = wollastonite) with limited or absent zoning. Orthopyroxene is more common than clinopyroxene and has a composition falling within the enstatite-ferrosilite boundary with Mg♯ [Mg/(Mg + Fe2++Fe3++Mn)] of 0.47–0.68. Few microphenocryst of amphiboles have been found in late Miocene rhyolite. Amphibole has a composition intermediate between the tremolite and edenite fields [AlIV of 0.5–0.7 a.p.f.u. and (Na + K) of 0.8–1.2 a.p.f.u.] and it is relatively F-rich (0.7–1.3 wt%). Zoning is absent. Analysis of pyroxene and amphibole are reported in Online Resource 1.
Late Miocene rhyolites are characterised by the presence of K-feldspar with no plagioclase. K-feldspar has a limited compositional range from Na-sanidine (Or32–47) to few microphenocryst of anorthoclase (Or12–19) (Fig. 2b). On the contrary, early Pliocene rhyolites are characterized by the presence of plagioclase with intermediate composition (An29–50) and by the absence of K-feldspar. Zoning is not particularly evident and when present it is normal zoning (Online Resource 3).
Pre-eruptive temperature and oxygen fugacity
Pre-eruptive temperature and oxygen fugacity (fO2) have been evaluated using the CORBA program based on the Fe–Ti two-oxide geothermometry and oxybarometry of Ghiorso and Evans (2008). The equilibrium between coexisting ilmenite and magnetite was tested on the basis of Mg/Mn partitioning following the equilibrium calibration of Bacon and Hirschmann (1988). Three ilmenite-magnetite pairs from sample GDL 212 and two pairs from samples GDL 211, both late Miocene rhyolites, passed the equilibrium test. Nevertheless, the CORBA algorithm gave results only for the three Fe–Ti oxide pairs in sample GDL 212. Estimated temperatures are as follows: 687, 837 and 856 °C. The range is quite large, but according to Ghiorso and Evans (2008) the uncertainty on the T can be as high as 100 °C. The estimated fO2 for the GDL 212 rhyolite are as follows: −16.61, −13.09 and −12.72 log10fO2. These data translate respectively into −1.02, −1.36 and −1.37 log bar units below the Nickel–Nickel Oxide buffer (ΔNNO).
Whole rock geochemistry and isotope data
The compositional diversity of mafic to intermediate rocks in the Tepic-Zacoalco rift has been described in several works (e.g., Ferrari et al. 2000; Petrone et al. 2003; Petrone and Ferrari 2008; Petrone 2010); the chemical characteristics of mafic rocks of different ages are exemplified with the data presented in Fig. 6a, c and e. The late Miocene and early Pliocene rocks have incompatible element patterns with positive spikes at Ba and Pb, and variably developed negative Nb–Ta anomalies, from almost absent in rocks of transitional character (early Pliocene), to clearly developed in sub-alkaline rocks. Transitional and sub-alkaline rocks of late Pliocene to Quaternary age display similar features, whereas the Na-alkaline rocks of this age show the typical patterns of ocean island basalt (OIB)-like intraplate basalts lacking Ba, Nb–Ta, Pb, and Sr anomalies.
In some cases rhyolites show devitrification textures, as observed above, due to post-magmatic alteration, which is also accompanied by element mobilization. In particular, the large-ion lithophile elements (LILE) such as Rb, Sr and Ba are the more prone to mobilization, whereas the high-field strength elements (HFSE: Ti, P, Th, Zr, Nb, Hf and Y) and the rare earth elements (REE) are usually immobile. One common effect of post-magmatic alteration is the enrichment in Rb and depletion in Sr (Riley et al. 2001). Thus the strong depletion in Sr coupled with Rb enrichment shown by our silicic rocks (Fig. 6) might be the result of post-magmatic alteration. The immobile elements (Nb, Ta, Th, Hf and Y) have a similar behaviour of Rb but correlate negatively with Sr (not shown). This argues against post-emplacements mobilization of mobile elements Rb and Sr. Thus, the very low Sr contents of late Miocene and early Pliocene rhyolites are the result of pre-eruptive processes (e.g., crystal fractionation) and cannot be explained by post-emplacement alteration.
Origin of the rhyolites
The similarity in REE contents between late Miocene and early Pliocene basalts and rhyolites excludes simple fractional crystallization process from basalt to rhyolite (Fig. 5c, d). All REE display a similar trend when plotted against a differentiation index despite their different compatibility/incompatibility in rhyolites and basalts and among rhyolites. In fact the light REE (LREE, e.g., La, Ce) are compatible (total partition coefficient D = 1.1–1.3, mineral/melt partition coefficients from GERM database: http://earthref.org/index.html), whereas Sm and Tb are incompatible (D = 0.7–0.5) when considering the actual mineralogy of rhyolite made up of feldspar + quartz + opaque + fayalite ± pyroxene and trace amount of apatite and zircon. Even if we take into account the possible presence of low amounts of accessory phases, such as chevinkite and astrophyllite which will make LREE, Zr and Nb more compatible (White 2003; Troll et al. 2003), it is difficult to explain the similarity in all REE contents between basalt and rhyolites, considering that HREE budget should not be controlled by these REE–Ti rich minerals (Troll et al. 2003). Only Eu behaves differently and clearly reflects feldspar crystallization (i.e., Eu negative anomaly, Fig. 6). Nevertheless, the very low Sr and Ba contents of the rhyolites suggest extensive fractional crystallization processes but not necessarily from a mafic parent. Indeed, the variation trend of Sr (and Ba, not shown) versus SiO2 (Fig. 5a) clearly does not point to late Miocene and early Pliocene basalts suggesting that they do not represent the composition of rhyolite’s parental magmas. In addition, the production of large volume of rhyolites from fractional crystallization of basalt (>95 % of basalt fractionation, Mahood and Halliday 1988) seems rather improbable because it will require even larger volume of ultramafic and mafic cumulates. Large volume of high-silica rhyolites with low Sr and Ba contents can be produced via extensive fractional crystallization of hybrid intermediate rocks (Mahood and Halliday 1988) or partial melting of older silicic lavas and ignimbrites, as suggested by Frey et al. (2007) for the Pliocene silicic flare-up of western Mexico.
The relative large range in Nd isotope ratios (Fig. 7) is consistent with the fact that most of the lavas come from different and discrete crustal reservoirs and not from a single magma chamber. Sr and Nd isotope composition of the less radiogenic late Miocene rhyolites is similar to that of late Miocene basalts possibly pointing to a source of the rhyolites with Sr and Nd isotopic composition similar to that of the basalt. Nevertheless, part of the western Mexico crust has also low radiogenic isotope composition, not dissimilar to that of the Miocene basalts and rhyolites. In particular, the available data for the closest lower crust granulites (Schaaf et al. 1994) indicate variable Sr and Nd isotope compositions, which in part might be compatible with the low radiogenic rhyolites. On the other hand, the Sr and Nd isotope compositional range of rhyolites differs from the very low εNd of the Punta Mita and Puerto Vallarta batholiths (Schaaf et al. 2003) exposed to the south of the WTMVB but resembles that of the granitic basement located just to the east of Colima rift (the Jilotan and Manzanillo batholiths). The more radiogenic late Miocene rhyolites are not dissimilar from the isotopic composition of the granitic basement of Jalisco Block, and the positive correlation between 87Sr/86Sr and 1/Sr (Fig. 7), implies contamination en-route to the surface by crustal lithologies with radiogenic Sr isotope compositions, such as the granitic basement.
All the above, while excluding an origin of the rhyolites through fractional crystallization of the associated basalts, points to a role of crustal components, although none of the available crustal compositions match exactly the Nd isotopic composition of WTMVB rhyolites. Apart from differentiation of primary magmas, common models for the genesis of rhyolites in arc setting envisage the formation of evolved melts by partial melting of older crustal rocks driven by heat and H2O released by cooling of basaltic magmas, eventually with additional evolved magmas originating through assimilation of crustal rocks by mantle-derived magmas (Annen et al. 2006). Crustal assimilation, mixing and hybridization of different melts are common processes (DePaolo et al. 1992; Hildreth and Moorbath 1988).
Large volume of high-silica rhyolites are commonly considered to be anorogenic or related to extensional settings (Christiansen et al. 1983, 1986; Orozco-Esquivel et al. 2002; Karacik et al. 2013) and have been related to melting of lower granulitic crust. Some of these high-silica rhyolites are characterized by low Sr, Ba and Eu, high F, Rb, U and Th and flat REE patterns (La/YbN = 1–3, Christiansen et al. 1983). The presence of topaz is also remarkable and hence they are known as topaz rhyolites (Christiansen et al. 1983). Nevertheless, as pointed out by Orozco-Esquivel et al. (2002) and Christiansen et al. (2007) not all the rhyolites with geochemical characteristics similar to the topaz rhyolites contain topaz, but they can be still the result of low degree of partial melting of lower crust granulites. Both late Miocene and early Pliocene rhyolites show trace elements patterns which might resemble those of the topaz rhyolites from the Upper Sequence of Mesa Central (Mexico, Orozco-Esquivel et al. 2002) with low concentration of Sr and Ba, high Rb contents and flat M- HREE pattern (Fig. 6d, f). Thus, a similar origin via partial melting of lower granulite crustal composition cannot be excluded despite the lack of topaz.
A new petrogenetic model for the production of silicic volcanism
If late Miocene–early Pliocene rhyolites are not generated via partial melting of lower crust granulite, as discussed in the previous section, a different origin needs to be explored. One possibility is that they are the results of partial melting or re-melting of older silicic rocks. As proposed by Frey et al. (2007), the Pliocene silicic flare-up in western Mexico could have been produced by re-melting or partially melting older silicic lavas or their plutonic equivalents related to the Sierra Madre Occidental. In their view, the preferential partitioning of Sr into feldspar relative to rhyolite liquids will generate rhyolite liquids with low Sr contents. Their main argument is the presence of one 26 Ma sanidine crystal in a rhyolitic pyroclastic flow NE of Tepetiltic which they consider as derived from a Sierra Madre Occidental protolith and preserved during partial melting. Our main concern with this model is that Sr and Nd isotope composition of both Sierra Madre Occidental ignimbrites and crust (Fig. 7) is too radiogenic with respect to the Sr and Nd isotope composition of late Miocene and early Pliocene rhyolites. Thus, melting of Sierra Madre Occidental lithologies with negative εNd and high 87Sr/86Sr (>0.705: Albrecht and Goldstein 2000) cannot generate silicic liquids with positive εNd and lower 87Sr/86Sr (0.7038–0.706). If the rhyolites are the result of melting of any older silicic rocks, the protolith is not represented by the Sierra Madre Occidental or by any older silicic volcanic rocks in the area.
The similarity in εNd isotope compositions between basalt and rhyolites strongly argues for a “mantle-origin” of the rhyolites although, as pointed out in the previous section, a simple liquid line of descent from basalt to rhyolite is not evident. Nevertheless, the low concentration of rhyolite compatible elements (i.e., Sr and Ba) might be the results of fractional crystallization processes, possibly from an unerupted mafic to intermediate composition or from hybrid intermediate magmas generated by crustal assimilation and crystal fractionation (AFC) of basalt. Mahood and Halliday (1988) suggested that the Quaternary peralkaline rhyolite of La Primavera were generated by partial melting of monzonitic rocks under relatively dry conditions. This mechanism will be able to generated large volume of silicic magmas with peralkaline character and low Sr and Ba contents as consequence of the alkali nature of the source, large fraction of refractory plagioclase and extensive fractional crystallisation.
Late Miocene and early Pliocene silicic rocks are ferroan (Fig. 4), K2O-rich, Sr- and Al-poor close to the composition of the Snake Rive Plain ferroan rhyolites as reported by Christiansen and McCurry (2008). WTMVB rhyolites are also reduced (low fO2, below NNO −1.02 to −1.37) with ΔNNO values in the range of the reduced fayalite-bearing Yellowstone silicic ash flows (Carmichael 1991), which are characterized by an average value of NNO of −1.62. These are the characteristic features of high-K fayalite rhyolite, typically found in bimodal association, as described by Frost and Frost (1997). As shown by Christiansen and McCurry (2008), reduced and dry physical conditions lead to early precipitation of plagioclase and retard the crystallization of Fe–Ti oxides leading to high-density ferroan magmas and their subsequent mid-crustal level entrapment (Christiansen and McCurry 2008) and favouring the formation of intrusive complexes. According to these authors, parental magmas of ferroan rhyolites of the Snake River Plain are derived by small degree of melting of the associated intrusive complex. We envisage that a similar process can be invoked for the production of the 7.5–3 Ma silicic magmatism. Indeed, given that the intensive parameters for the rhyolites are similar to those of the Snake River Plain magmatism, it is possibly that the intrusive complex, which in turn originated the parental magmas of WTMVB rhyolites, formed as results of crystal fractionation in similar reduced and dry conditions leading to the formation of unerupted high-density ferroan mafic magmas.
Late Miocene and early Pliocene WTMVB mafic rocks are tholeiitic to ferroan and so characterised by moderate to extensive Fe-enrichment (Fig. 4a). Thus, they can have followed the tholeiitic differentiation trend, leading to silica-rich ferroan compositions, under dry and reduced conditions (Frost and Frost 1997; Christiansen and McCurry 2008). These intensive parameters are commonly associated with anorogenic settings, but they can be also found in arc settings characterised by water-poor conditions. In this case, arc magmas will be forced to follow the tholeiitic trend instead of the more common calc-alkaline one (Tatsumi and Suzuki 2009). In a recent review on the TMVB, Ferrari et al. (2012) pointed out that the 11–8.5 Ma pulse of mafic magmatism (our late Miocene basalts) is associated with decompression melting favoured by slab detachment triggered by the cessation of the subduction west of Baja California. Dry and hot conditions are expected in the region since the Late Miocene as a consequence of the influx of hot asthenosphere into the mantle wedge from the Gulf of California area to the northwest (Ferrari et al. 2001, 2012; Ferrari 2004).
The oxygen fugacities of rhyolite reported in this work (ΔNNO −1.02 to −1.37) indicate that magma evolved under reduced conditions and that they crystallized at oxygen fugacities below the fayalite-magnetite-quartz buffer. The obtained values are within the ΔNNO range of fayalite-bearing silicic rocks (−0.9 to −2) reported by Carmichael (1991), and are significantly lower than values for late Pliocene to Quaternary alkali basalts, calc-alkaline basalts, basaltic andesite and hornblende andesite from the WTMVB (−0.5 to +3.5) (Carmichael 1991; Righter and Carmichael 1992; Righter et al. 1995; Righter and Rosas-Elguera 2001). Considering that oxygen fugacities of magmas reflect those of their sources as long as oxygen exchange can be precluded (Carmichael 1991), and that the most probable effect of the assimilation of crustal material would be the increase of fO2, the low oxygen fugacity values of the rhyolitic magmas would indicate that they originated from a source or parental magma at least at similar reduced conditions.
The low fO2 calculated for the rhyolites points to the existence of reduced conditions during the late Miocene allowing the high-density Fe-enriched tholeiitic magmas to pond at mid-crustal level forming intrusive gabbroic complexes. On the basis of major and trace elements, Mori et al. (2009) suggested that the parental tholeiitic basaltic magmas originating the thick late Miocene basaltic succession in the Los Altos de Jalisco, east of Guadalajara, experienced high-pressure fractional crystallisation at the base of the continental crust. Given that the late Miocene basaltic episode is a major widespread magmatic pulse that affected the entire TMVB belt from west to east (Ferrari 2004) and given that the thickness of the crust is not significantly different from Los Altos de Jalisco and west of Guadalajara (Ferrari et al. 2012), it is highly probable that the parental tholeiitic basaltic magmas in the WTMVB experienced the same high pressure fractionation at the base of the continental crust, favouring the formation of intrusive gabbroic complexes. Experimental data on arc magma composition at 1.2 GPa (~40 km) indicate that H2O contents in the liquid change the crystallisation sequence (Muntener et al. 2001). Low H2O contents stabilise plagioclase earlier than garnet and amphibole while the liquid remain quartz saturated. Late Miocene basalts have low La/Yb and Gd/Yb ratios (2.36–9.47 and 0.58–1.60 respectively) which suggest that the liquid was not in equilibrium with garnet, whereas the negative Eu/Eu* anomaly (0.14–1.04) points to the involvement of plagioclase.
Fractionation of plagioclase can also contribute to an increase of magma density (Sparks and Huppert 1984). Constraints on the density of the unerupted mafic magmas can be obtained from the composition of the less evolved (SiO2 < 50 %) ferroan late Miocene basaltic rocks from the WTMVB. Magma densities between 2.85 and 2.87 g/cm3 were estimated with the method of Bottinga and Weill (1970) for pressures corresponding to the base of the ca. 40 km thick crust in the study area (see Ferrari et al. 2012; Fig. 4), and thermodynamic properties from Lange and Carmichael (1990), Lange (1997), and Ochs and Lange (1997). The calculated densities represent minimum values for the unerupted magmas because magmas emplaced on the surface may have underwent further differentiation during ascent through the crust. The composition and density of the crustal column in the WTMVB is poorly constrained, but estimates of 2.7–2.9 g/cm3 for the lower crust and of 2.65–2.70 g/cm3 for the upper crust have been reported for neighbouring areas (Ortega-Gutiérrez et al. 2008). These data indicate that the mafic magmas might have been underplated at the base of the crust or stalled in the crust because of their density contrast and crystallised to form gabbroic complexes. Continuous basalt intrusion can supply the heat and additional enthalpy able to favour partial melting of these gabbroic complexes, thus forming silicic magmas, which in turn can be subsequently modified by AFC differentiation processes at shallow level in the crust or en-route to the surface. Thus, late Miocene and early Pliocene rhyolites might be the result of partial melting of this newly formed crustal gabbroic complexes crystallized from the unerupted high-density portion of the late Miocene basaltic magmas.
In this scenario the rhyolites would be the results of a two-step process: (1) melting of gabbro cumulates formed during the previous basaltic magmatic episode and ponded in the lower-middle crust due to density contrast, and thus forming new mantle-derived crust; (2) variable degree of AFC processes en-route to the surface which subsequently modified the gabbro-derived melts.
Geodynamic constraint and final remarks
The work was partly supported by a bilateral Grant CONACyT-SRE (Mexico)—MAE (Italia). The authors thank S. Tommasini, L. Francalanci and all the staff of the Isotope Lab of the University of Florence for allowing access to their isotope lab; O. Perez-Arvizu, CGEO-UNAM for ICP-MS trace element analyses; John Spratt at NHM for allowing access to the EPMA lab. The authors would like to thank T. Grove for handling the manuscript, G. Mahood and an anonymous reviewer for improving the quality and focus of the manuscript.