Abstract

In one of the most studied Variscan exposures, the Órdenes allochthonous complex (NW Spain), the transition between medium-pressure (MP) and high-pressure (HP) units in the SW of the complex has been identified as an extensional shear zone: the Fornás detachment. Migmatitic paragneisses crop out discontinuously along that boundary, at the base of the MP ensemble (O Pino unit). The metamorphic reaction sequence, mass balance calculations, and phase diagram modeling investigated in these paragneisses are interpreted in terms of an approximately isobaric heating path (8 ± 0.8 kbar), from ∼650 °C to 740 °C, crossing into the melt- and K-feldspar–bearing stability fields. These anatectic conditions are evidenced by the presence of leucosomes through progressive muscovite and biotite melting reactions. Our results indicate that the heating path evidenced by the migmatitic paragneisses is directly related to the subtractive nature of the Fornás detachment, with heat transferred from the footwall to the hanging-wall unit.

INTRODUCTION

It is widely accepted that, depending on crustal variables such as composition and structure, different tectonic processes may produce pressure-temperature (P-T) conditions up to the granulite facies and associated heat and mass transport within the crust (e.g., Sandiford and Powell, 1986; Harley, 1989; Meissner, 1989; Selverstone and Chamberlain, 1990). In addition, depending on the crustal heterogeneity, mass and heat transport may be highly variable, for instance, where shear zones play a critical role in the bulk rheology and can act as barriers or channels for mass and heat transfer (e.g., Thompson and Connolly, 1992; Gottardi et al., 2013).

A high apparent thermal gradient at shallow crustal levels (from 50 to 150 °C/km; De Yoreo et al., 1991; Brown, 2007), as well as melt migration (e.g., Miyazaki 2004), or heat transfer by specific tectonic process, may trigger high-temperature (HT) metamorphism and partial melting of crustal rocks. Heat transfer will affect the metamorphic reaction sequence, the subsequent P-T paths, and the final interpretation of the geodynamic evolution. This is especially delicate for high-grade and anatectic metamorphic rocks, in which the homogenization of mineral grains (rapid rates of cation diffusion) may partly or totally erase the information required to track the P-T paths associated with the granulite-facies conditions.

In addition, isolated chemical analysis cannot define the sense of reaction progress along the P-T path (e.g., isobaric heating vs. cooling, or isothermal decompression vs. burial). Several studies have demonstrated that chemical (compositions and reaction balance) and petrological constraints (coherent P-T projections or pseudosections) must be complemented with textural analysis in metamorphic rocks to distinguish between alternative generic possibilities (e.g., Carmichael, 1969; Selverstone and Chamberlain, 1990; Hand et al., 1994; Kriegsman and Hensen, 1998; Cenki et al., 2002; Vernon, 2004; Álvarez-Valero and Kriegsman, 2007, 2010; Álvarez-Valero and Waters, 2010). Therefore, it is necessary to define reactants and products and successive mineral assemblages, as well as other fundamental parameters of the rock evolution, such as age patterns (trace-element distributions and isotopic compositions), in their properly defined microtextural setting (e.g., papers in Vance et al., 2003).

The direct comparison of mineral modes and composition with calculated phase relationships (through a P-T pseudosection) constrains P-T estimates, as well as P-T paths by the interpretation of the textural evolution of the rock. Hence, phase diagram modeling—not dependent on establishing original mineral compositions—represents an advantage over conventional thermobarometers in that it allows the observed assemblage to be quantitatively constrained for a specified rock (e.g., White et al., 2002; Powell and Holland, 2008) or microdomain composition (e.g., Álvarez-Valero and Kriegsman, 2007, 2010; Álvarez-Valero and Waters, 2010).

Concerning the possible tectonic process responsible for significant heat transfer, several contributions have related nearly isothermal decompression in the internal parts of an orogen with tectonic denudation (Thompson and England, 1984; Escuder Viruete et al., 1997; Gerbi et al., 2006). However, the outcome P-T paths are from units that were exhumed at the footwall of extensional detachments. Conversely, isobaric heating has been registered at the hanging wall of extensional detachments, but it is reported less frequently (De Jong, 1991; Escuder Viruete et al., 1994; Brown, 2009), whereas heating at constant depth has been generally interpreted as caused by magmatic intrusions (Abati et al., 2003; El-Shazly et al., 2011).

This petrological study aims to (1) assess the tectonothermal feedback between a particular tectonic process, the development of an extensional detachment, and localized partial melting at its hanging wall, (2) put constraints on the conditions of early Variscan extensional collapse, and (3) compare the derived P-T path with other paths published in the literature.

GEOLOGICAL SETTING

The allochthonous complexes in NW Iberia consist of a pile of exotic units characterized by distinct tectonothermal histories and lithological associations, the mutual tectonic contacts of which are either thrust faults or extensional detachments (Martínez Catalán et al., 2002, 2009). In the Órdenes complex (Fig. 1), three types of units have been recognized from bottom to top at the nappe stack and grouped into the Lower, Middle, and Upper Allochthons. The Upper Allochthon consists of units that are derived from an ensialic magmatic arc developed in the northern margin of the Gondwana supercontinent during the Late Cambrian and earliest Ordovician (Fig. 2). In the Middle Devonian, and associated with early Variscan plate convergence, the arc was incorporated into an accretionary wedge by partial subduction and internal imbrication. Its lower section underwent a high-P and high-T (HP/HT) episode, while its upper section was thickened under medium-pressure (MP) conditions (Gómez Barreiro et al., 2006, 2007; Fernández-Suárez et al., 2007).

The Middle Allochthon consists of ophiolites and units of oceanic affinity related to the former Rheic Ocean. These units where partially subducted and imbricated in the deep levels of the accretionary wedge, becoming a significant part of it (Fig. 2).

The Lower Allochthon represents the thinned passive margin of northern Gondwana, which was subducted under the wedge upon ocean closure. Underthrusting and imbrication of the Middle and Lower Allochthons provoked the extension and thinning of the overlying Upper Allochthon, where a first generation of extensional detachments formed during the Late Devonian and earliest Carboniferous. The most important are the Fornás and Corredoiras detachments, which put the HP/HT units (at least 15 kbar and >800 °C; Díaz García et al., 1999; Arenas and Martínez Catalán, 2002) and the MP units into contact, with the MP units at the hanging wall (Figs. 1B and 1C). Once subduction of the Lower Allochthon became locked, plate convergence continued during the early and middle Carboniferous under a collisional regime, during which the Upper Allochthon underwent a renewed episode of shortening, with recumbent folding and thrusting toward the Gondwanan foreland, and finally, renewed extension during orogenic collapse in late Variscan times (Díaz-García et al., 1999; Gómez Barreiro et al., 2006, 2007).

In the Upper Allochthon (Fig. 2), intrusive rocks related to the ensialic magmatic arc and to coeval crustal extension have been dated at 520–490 Ma (Peucat et al., 1990; Ordóñez-Casado, 1998; Abati et al., 1999, 2007; Fernández-Suárez et al., 2007; Castiñeiras et al., 2010). The age of HP/HT metamorphism in the Upper Allochthon is constrained between ca. 405 Ma and 390 Ma in eclogites, mafic granulites, and quartzo-feldspathic gneisses (Schäfer et al., 1993; Santos Zalduegui et al., 1996; Ordóñez-Casado et al., 2001; Fernández-Suárez et al., 2002, 2007). Ophiolites were imbricated between 390 and 375 Ma (Peucat et al., 1990; Dallmeyer et al., 1991, 1997), and continental subduction of the Lower Allochthon occurred at 375–365 Ma (Santos Zalduegui et al., 1995; Rodríguez et al., 2003; Abati et al., 2010). The collapse of the accretionary wedge (Fig. 2) took place during the imbrication of the Middle and Lower Allochthon, that is, between 390 and 365 Ma. Recumbent folding and thrusting during the collisional stage operated in the internal zones at ca. 365–330 Ma (Dallmeyer et al., 1997; Martínez Catalán et al., 2009). This was followed by extension and collapse until 315 Ma and renewed shortening and folding related to transcurrence and oroclinal bending until ca. 305 Ma (Aerden, 2004; Martínez Catalán, 2011, 2012).

Our research is centered around the Fornás detachment, an early Variscan extensional shear zone with a top-to-the-NNW sense of motion, which separates metapelites and paragneisses of the O Pino unit in the hanging wall from underlying HP/HT, granulite-facies metabasites of the Fornás unit in the footwall (Fig. 1D). In addition, the detachment is subtractive, as there is a jump in metamorphic conditions across it, with the footwall having registered the higher-P and -T conditions. To the east of the detachment, the O Pino metapelites reflect upper-amphibolite-facies transitional to granulite-facies conditions (9–11 kbar and 700–800 °C, which were constrained by conventional geothermobarometry that did not use the whole bulk composition of the entire melt-bearing mineral assemblage; Castiñeiras, 2005). However, there are discontinuous outcrops at the base of the unit, spatially related to the Fornás detachment, which record a heating at relatively low pressure and the injection of granitic leucosomes (asterisks on Fig. 1D). This event occurred after the metamorphic peak of the unit, as supported by textural and field evidence (Gómez Barreiro, 2007). Here, we focus on some of those rocks that appear as a small (700 × 300 m) extensional horse, the Donas slice (Fig. 1D), where migmatitic paragneisses have been preserved. We will focus on that slice because (1) it registered a relatively small motion along the detachment, and (2) the subtraction between it and the footwall seems to be less than in the rest of the extensional fault.

The Donas slice records the higher-T conditions that affected the hanging wall of the detachment. It underwent migmatization during extension related with underthrusting of the Middle and Lower Allochthons beneath the Upper Allochthon accretionary wedge, in the early stages of Variscan convergence (Fig. 2). Later contractional structures related to the Variscan collision include recumbent folds and thrust faults with east vergence, and late Variscan extensional detachments of the Bembibre–Pico Sacro system (Figs. 1B, 1C, and 1D; Gómez Barreiro et al., 2010). Actually, the slice occurs at the hinge zone of a small recumbent anticline, close to the Pico Sacro detachment. However, these structures developed under lower metamorphic conditions than the episode of migmatization (lower amphibolite facies; Gómez Barreiro, 2007; Gómez Barreiro et al., 2007), and samples where the previous mineral assemblages have been preserved are common in the Donas slice.

Sample Description

The pelitic and semipelitic paragneisses show leucosomes (0.5–20 cm wide) defining the main foliation that appears transposed by minor asymmetric folds. They are medium- to coarse-grained rocks with granolepidoblastic texture, and biotite and quartz-feldspar segregations defining the planar fabric. At the thin-section scale, the textural microdomains show quartz-rich layers where evidence of dynamic recrystallization and crystal plasticity is evident in both quartz and plagioclase. Melanocratic layers mostly include euhedral garnet, aluminosilicate (kyanite and fibrolite), and biotite (Figs. 3A, 3B, and 3C).

A first garnet generation typically displays equidimensional shape and slightly resorbed rims, and is up to 1 mm in diameter. This garnet locally encloses biotite, quartz, rutile, and kyanite. A local second overgrowing garnet occurs as aggregates of smaller euhedral crystals and also includes biotite (e.g., Figs. 3D and 3E).

Biotite is a major component of the matrix, in contact with garnet, or locally included in garnet as well (Fig. 3E), kyanite, and plagioclase. In places, skeletal biotite also occurs in patches with muscovite and surrounding plagioclase crystals (Fig. 3F). We observed two chemical and texturally different generations of plagioclase. Plagioclase 1 is a minor component of the matrix, in contact with or enclosing kyanite, biotite, garnet, and muscovite, whereas plagioclase 2 rims euhedral small garnets with local boundaries of fibrolite, muscovite, and skeletal biotite (Figs. 3A, 3C, 3D, and 3F).

METHODOLOGY

We sampled, studied their respective thin sections, and chemically analyzed five representative migmatitic paragneisses of the Donas slice. We focused on key microtextural patches, which, integrated with the field studies of Castiñeiras (2005), constrain the overall P-T path of the Donas migmatites. Ours approach combines petrography, phase diagram modeling, and geochemistry in the NCKFMASHT (Na2O-CaO-K2O-FeO-MgO-Al2O3-SiO2-H2O-TiO2) system.

Microtextures, Mineral Chemistry, and Reaction Balancing

Our method is similar to that applied to migmatites and granulite-facies crustal xenoliths by Kriegsman and Hensen (1998), Cenki et al. (2002), and Álvarez-Valero and Kriegsman (2010), among others. These authors showed the importance of reaction balancing in correctly identifying reactants and products, especially for the case in which one of the reactions has been fully consumed.

In summary, the methodological sequence followed is: (1) detailed textural analysis (Fig. 3); (2) definition of a simplified chemical system for each reaction texture to see whether or not the number of phases fits to the phase rule (Table 1); (3) proper reaction balancing, including the calculation of volume ratios of product phases (Table 1); and (4) quantification using phase diagram modeling (Fig. 4).

Metamorphic Modeling Procedure

We focus on microtextural analysis combined with thermodynamic modeling to constrain P-T conditions and regional evolution. Pressure and temperature were constrained from the best fit between observed and calculated mineral data by utilizing Perple_X (Connolly, 2005; including the thermodynamic database of Holland and Powell [1998], with updates).

The local bulk composition for modeling the migmatitic paragneisses was determined by combining phase compositions and modal proportions derived from image analysis of photomicrographs, supported by visual estimation (e.g., Álvarez-Valero and Waters, 2010). Whole-rock bulk analyses for metamorphic rocks, especially those that underwent partial melting conditions, have limited use for phase diagram modeling because each microdomain may represent a different “equilibrium” stage (different equilibrium volumes sensu stricto; Stüwe, 1997) within the same rock (or thin section). Hence, we investigated the evolution of our paragneisses using discrete modal bulk compositions, calculated from local mineral modes and composition, to avoid the construction of unreliable pseudosections (Fig. 4).

Mineral compositional zoning, e.g., in garnet, is minor, and its relict cores are similar in composition to equilibrated rims. The main variation in garnet is in the MnO content, varying over ∼2 wt%, largely compensated by FeO. A mean composition, weighted toward that of the rims, was used for the bulk composition calculation. In view of the minor compositional zoning and the small volume contribution from fractionated grain cores, we did not adjust the estimated bulk compositions for this effect.

REACTION BALANCING RESULTS

Reaction balancing allows estimation of the effects of changing a compositional variable using a simple spreadsheet. A balance can only be achieved in a pseudo-univariant system, as one can only solve for n variables by having n independent equations. Being interested only in ratios and not absolute quantities, one reaction coefficient is set to 1, reducing it to n – 1 equations for n variables (Thompson, 1982). In order to balance reactions with variance >1, solid solution phases can be split into end members (see the Ab-An example of the divariant reaction 2). After the calculations, the phases can be reconstituted from the end members (for more balancing examples, see also Álvarez-Valero and Kriegsman, 2010). Melt composition for balancing is similar to the electron microprobe (EMP)–analyzed glasses (after melting a pelite) of Acosta-Vigil et al. (2007) and Álvarez-Valero et al. (2007).

Melt-Producing Reactions

The Donas paragneisses show melanosomes with the assemblage Grt + Bt + Ky/Sil + Pl ± Ms and leucosomes containing Qtz + Pl + Kfs ± Grt ± Sil ± Ms (mineral abbreviations after Kretz, 1983; M—melt). The presence of Grt in leucosomes suggests a vapor-absent biotite dehydration-melting reaction (e.g., Patiño Douce and Harris, 1998; Clemens and Droop, 1998; Spear and Daniel, 2001; Kriegsman, 2001; Kriegsman and Álvarez-Valero, 2010). We focus on three main reactions using real chemical phase compositions, which define all the microstructures observed in the paragneisses. All three are melt-producing reactions and are responsible for the progressive consumption of biotite and production of garnet. For the earliest one, in the KFMASH system and its subsystems KFASH and KMASH, the degenerate reaction with the water-absent continuous balance 
graphic
represents the minimum melt reaction for pelites, and little or no melting (i.e., ± M) is expected to occur at this stage (Fig. 3B; Table 1A; see also Spear et al., 1999). In a five-component chemical system K(FM)ASH (reducing Fe and Mg to one component), six phases or phase components are needed to balance a reaction. Alternatively, adding plagioclase, the balance is in the NCK(FM)ASH system. Two chemically different plagioclases are observed: Plagioclase 1 (Pl1), associated with fibrolite and skeletal biotite in the matrix, ranges from An17 to An28, whereas plagioclase 2 (Pl2) , within the leucosomes and next to Bt + Sil recrystallization boundaries, ranges from An67 to An75 and locally hosts euhedral garnets smaller than those within the melanosomes (Fig. 3; Table 1B). There are no systematic differences between core and outer rim of individual plagioclase crystals. Kinetic factors may control the shape of garnet crystals in the leucosomes, which are smaller and more euhedral than those formed within the melanosomes (e.g., Waters, 2001; Figs. 3A and 3C).
Replacing plagioclase by the two end members (i.e., Ab- and An-rich components) gives the right number of phase components (Table 1B). Thereby, we used a seven-component chemical system and eight phases or phase components to balance the reaction. After balancing, Pl compositions can be reconstituted from the two end-member components. This balance matches the observed Grt/Pl2 volume ratio (∼0.24). This leads to a balance of the form 
graphic
At progressively higher T, the P-T path crosses into the Sil and Kfs stability fields (Ms-absent), where Grt1 (product of reactions 1 and 2) reacts with the relic matrix phases Bt, Sil, and Qtz to form a new garnet generation (Grt2 or garnet overgrowth). Thereby, partial resorption of Grt1 is related to growth of Grt2, whereas partial resorption of Grt2 is interpreted in terms of a back reaction with crystallizing melt: 
graphic

Both compositional Grt1 and Grt2 profiles are slightly bell-shaped with sharp rims of resorption, especially for MnO (Table 1C). The reaction balance calculation is in the K(FM)ASHMn system that includes a six-component chemical system (reducing Fe and Mg to one component), and seven phases or phase components (i.e., Bt, Sil, Grt1, Grt2, Qtz, Kfs, and M). Manganese is added as a chemical component to investigate its exchange between the two Grt generations. This balance predicts the correct Grt2/Kfs volume ratio (∼5).

Melt-Consuming Reactions

There are several types of textural evidence that support the occurrence of a subsequent cooling (and possibly decompressive) episode in the evolution of the migmatites (Fig. 4): (1) resorbed Grt2 that produces Bt + Sil around the leucosomes or within the matrix (Fig. 3E), as well as local growth of ilmenite crystals possibly after rutile; and (2) some Kfs grains in the leucosomes that have reacted to late muscovite, which occurs both cutting across the fabric and within leucosome cores showing aggregates of Bt and Sil (Figs. 3A, 3C, and 3F). The water needed is likely to have been derived from the crystallizing melt.

MODELING RESULTS

Phase diagram modeling reveals that the migmatites record isobaric heating from ∼650 °C to 740 °C (±20 °C) at 8 ± 0.8 (2σ) kbar. At this P, the solidus lies at ∼660 °C. Given that staurolite occurs in the O Pino unit above the Donas slice (Van Zuuren, 1969; Castiñeiras, 2005; Gómez Barreiro, 2007), it was included in the models to find the lower-T limit of the Donas metamorphic evolution.

Subsequently, microtextures and reaction balancing helped to constrain the isobaric heating path (estimates in the NCKFMASHT system through the best fit between observed and modeled modes) until it crossed into the melt- and Kfs-bearing stability fields (Fig. 4), and the transformation of Ky into Sil. The chemically distinct generations of garnet and plagioclase are consistent with the heating path and the observations of both garnet overgrowth (Fig. 3D) and large patches of plagioclase and quartz that host the small euhedral garnets (Fig. 3C). Later, during the cooling and exhumation episode, the newly grown garnet (+ rutile) reacted with the melt and formed a retrograde biotite + sillimanite + muscovite (+ ilmenite) assemblage wrapping around the leucosomes or growing within the matrix (Fig. 3E). Evidence for a slight decompression on the retrograde path comes from the growth of sillimanite + ilmenite and the lack of retrograde kyanite (Fig. 3).

DISCUSSION AND CONCLUDING REMARKS

Microtextural observations combined with mass balance calculations track the metamorphic reaction sequence of the Donas metapelitic migmatites, documenting an isobaric heating episode within the tectono-metamorphic history of the O Pino unit in NW Iberia. Our data show that coeval thinning did not occur on the hanging wall to the detachment (isobaric heating path in Figs. 4A and 5), and therefore the partial melting at the base of O Pino unit was not induced by decompression in the overlying upper crust. Heat transfer from the hotter footwall seems a better option. In fact, partial melting was a generalized phenomenon at the footwall to the detachment, in the HP/HT units. The granulitized metagabbroic rocks of the Fornás unit were migmatized to some extent, and particularly the Fornás amphibolites are often banded rocks with abundant amphibole-plagioclase veins strongly deformed under high-T conditions (Van Zuuren, 1969; Gómez Barreiro, 2007).

However, the migmatization products of the Fornás unit seem to have remained in situ, with the exception of small leucosomes close to and partially incorporated in the detachment shear zone (Gómez Barreiro, 2007). Since no coeval large bodies of Late Devonian intrusive rocks have been found in this part of NW Iberia, heat transfer may be related to the higher T of the footwall rather than to magmatic intrusion. The existence of a localized thermal event linked to the Fornás detachment suggests that this structure locally controlled the flow of heat. Although we suggest that heat transfer from the footwall to the detachment induced synkinematic melting at the base of its hanging wall, limited melt migration from the footwall unit into the shear zone may have reinforced the local character of the isobaric event.

The proposed structural context of the Donas slice is sketched in Figures 1 and 2. An extensional detachment would have put the middle-crustal O Pino unit, essentially metapelitic, into contact with the lower-crustal Fornás unit, a granulitized and amphibolitized metagabbro. Heat transferred from the O Pino unit would have induced migmatization under upper-amphibolite-facies to transitional granulite-facies conditions at the base of the hanging wall to the detachment. The crustal segment above it was not thinning at the time, but the footwall unit was probably thinned, as it was under generalized partial melting conditions and consequently weak from a rheological point of view. Published time constraints on the tectonic evolution of the NW Iberian allochthonous complexes suggest that the whole crust and lithosphere were being thickened at the same time. Extension and thinning of the Upper Allochthon were coeval with underthrusting and imbrication of the Middle Allochthon and continental subduction of the Lower Allochthon (e.g., Martínez Catalán, 2011, 2012).

The P-T paths for the Fornás and O Pino units, deduced by Castiñeiras (2005), Gómez Barreiro (2007), and Gómez Barreiro et al. (2007), are shown in Figure 5. The first part of the Fornás trajectory, characterized by pressurization due to building of the initial accretionary wedge, is unconstrained. However, data from similar HP/HT units indicate at least 15 kbar and >800 °C (Díaz García et al., 1999; Arenas and Martínez Catalán, 2002). In the Fornás unit, the path is known for the exhumation following the early Variscan HP/HT event (path 1; Gómez Barreiro, 2007). Regarding the O Pino unit, the path corresponds to the deep levels of the unit, just above the early Variscan Fornás detachment (path 2) and the one obtained in the present study (path 3). A Variscan clockwise trajectory (path 4) is also included, being common for both Fornás and the O Pino units. This late path represents the crustal thickening by recumbent folding and thrusting followed by late Variscan exhumation (Fig. 2). Paths A and B (Fig. 5) are previously published trajectories for the Fornás and O Pino units, respectively (Gómez Barreiro et al., 2007), that are reinterpreted according to our new results as paths 2 and 3, continuing into path 4, which represents the Variscan collisional stage (Gómez Barreiro et al., 2007).

The complexity of the trajectories correlates with that of the structural evolution, which is explained in Figure 2. The main difference between the old and new paths for the Fornás unit (paths 1 and A) occurs at the part related to the early Variscan exhumation: The Fornás detachment may have developed after the maximum crustal thickness was reached by the MP units of the Upper Allochthon, or nearly so, and not after a significant decompression and limited retrogression, as suggested by path A of Figure 5. Once put into contact by the Fornás detachment, the MP and HP/HT units followed a common P-T trajectory characterized by cooling and minor decompression. This path was probably driven by overlying extensional shear zones such as the Ponte Carreira detachment (Fig. 1), and by a swarm of extensional shear zones affecting the Fornás unit and developed at relatively low pressures (below 4 kbar in their final stages; Gómez Barreiro, 2007).

We conclude that the main outcome of this paper, i.e., the localized heat transfer across an extensional detachment, has fundamental importance in metamorphic geology because the small set of samples that follow heating paths in high-T terranes may reflect local rather than regional conditions and are not necessarily related to the intrusion of igneous bodies.

This contribution has been funded by the research project CGL2011-22728 of the Spanish Ministry of Science and Innovation, as part of the National Program of Projects in Fundamental Research, in the frame of the VI National Plan of Scientific Research, Development and Technologic Innovation 2008–2011. Álvarez-Valero and Gómez Barreiro appreciate financial support by the Spanish Ministry of Science and Innovation through the “Ramón y Cajal” research program. We deeply appreciate the constructive comments of Brendan Murphy, Leo Kriegsman, and Arlo Weil that helped to improve the manuscript. Alampi acknowledges the funded stay at the University of Salamanca in 2012 by the MISTI program (Massachusetts Institute of Technology [MIT] International Science and Technology Initiatives), as well as the Department of Earth, Atmospheric, and Planetary Science, MIT, for providing conference travel funding.