Geochemistry and U–Pb dating of felsic volcanic rocks in the Riotinto–Nerva unit, Iberian Pyrite Belt, Spain: crustal thinning, progressive crustal melting and massive sulphide genesis

Abstract: We present new geochemical, Sm–Nd and U–Pb data on the felsic volcanic rocks containing the world-class volcanic-hosted massive sulphide deposits in the Riotinto–Nerva unit, Iberian Pyrite Belt, Spain. Three new U–Pb ages from older plagioclase–quartz-phyric dacites and plagioclase-phyric rhyolites to youngest plagioclase–quartz-phyric rhyolites indicate a time span for felsic volcanism ranging from 351.5 ± 0.4 to 345.7 ± 0.6 Ma. The youngest felsic rocks exhibit lower &egr;Nd, as well as contrasting Ti, Sr, Zr, Hf and Eu/Eu* values. We interpret that in the Riotinto–Nerva unit crustal melting successively affected shallower, more evolved horizons in the crust. Progressive crustal melting is consistent with current interpretations of the Iberian Pyrite Belt in terms of a late Devonian–Early Carboniferous transtensional setting, coupled with underplating by basic magma. We suggest that low &egr;Nd, evolved crustal magmatism could be used as a proxy in studies of the genesis of these and possibly other Phanerozoic massive sulphide deposits. Supplementary material: Details of the analytical methods and whole-rock data for basalts are available at http://www.geolsoc.org.uk/SUP18452.

Volcanic-hosted massive sulphide (VHMS) deposits constitute today one of the major sources of base metals on Earth. Such deposits have formed at various times in a range of geological settings, most often related in time and space to felsic volcanic rocks within bimodal, mafic-felsic volcanic successions (Lentz 1998; Barrie & Hannington 1999;Allen et al. 2005). Among VHMS areas, the Iberian Pyrite Belt, in the SW Iberian Peninsula, is of great economic interest because it hosts a huge number of sulphide masses that are very diverse in tonnage and grade, including world-class deposits such as Riotinto or Neves Corvo. The Riotinto mining district alone is interpreted as the largest concentration of volcanic-hosted sulphides on Earth.
Given that VHMS deposits occur within volcano-sedimentary stratigraphic successions and are commonly coeval and coincident with volcanic rocks, several major research lines have been followed to assess the links between volcanism and VHMS deposition, including studies on volcanic architecture and geochemistry. These studies are particularly relevant to the Iberian Pyrite Belt, where relationships between volcanism and VHMS genesis are still a matter of debate (Leistel et al. 1998;Sáez et al. 1999;Tornos 2006, and references therein), among other reasons because detailed studies on volcanic rocks in the Iberian Pyrite Belt are still scarce, as are also isotopic and U-Pb dates.
Considering geochemical features of volcanic rocks related to VHMS deposits, it has been often argued that among the felsic rocks hosting these deposits in a given area, some geochemical types could be more prospective and therefore used as an exploration tool. For instance, it has been claimed that most 'fertile' host types are rhyolites characterized by relatively flat REE patterns, low Zr/Y and intermediate to high high field strength element (HFSE) contents (Lesher et al. 1986; Barrie et al. 1993;Lentz 1998;Hart et al. 2004;Piercey et al. 2008). However, the debate is complicated because it is agreed that VHMS deposits and their host felsic volcanic rocks have formed in a wide variety of tectonic settings, implying different petrogenetic models (Hart et al. 2004). Accordingly, the use of these geochemical features as a tool in VHMS exploration has been questioned (Mercier-Langevin et al. 2007;Gaboury & Pearson 2008).
The aim of this study is to interpret the geochemical and age relationships of the various types of felsic rocks occurring within the Riotinto-Nerva unit in the Spanish Iberian Pyrite Belt, which includes the Riotinto mining district, on the basis of new geochemical and Sm-Nd isotope data, as well as U-Pb dates for these rocks. We show that felsic volcanic rocks in the Riotinto-Nerva unit developed over a protracted time span and mostly predated VHMS deposition in the Riotinto mining district. We also interpret these data to place some constraints on the genesis of the VHMS and the regional crustal evolution. Finally, we discuss the potential use of isotopic and geochemical parameters as exploration tools in the Iberian Pyrite Belt and other Phanerozoic VHMS areas.

Geological setting
The Iberian Pyrite Belt is located in the central part of the South Portuguese Zone, the southernmost part of the Iberian Massif ( Fig. 1a). The Iberian Pyrite Belt consists of a threefold, late Devonian to Carboniferous stratigraphic sequence deformed during the Variscan orogeny, with plutonic rocks confined to the north. From base to top, the following stratigraphic units are distinguished (Schermerhorn 1971) (Fig. 1b): (1) the Phyllite-Quartzite Group; (2) a volcanic succession named the Volcano-Sedimentary Complex; (3) a turbidite sequence named the Culm Group.
The Phyllite-Quartzite Group (Fig. 1b) is a Late Devonian siliciclastic sequence, consisting of a monotonous succession of intercalated shale and sandstone. It is interpreted to represent deposition in an epicontinental sea (Leistel et al. 1998). Its upper contact with the Volcano-Sedimentary Complex is in many places tectonic, but in some areas is marked by the occurrence of deposits and structures indicating ground rupture and shaking, and therefore the onset of regional tectonic instability (Moreno et al. 1996).
The Volcano-Sedimentary Complex (Fig. 1b) consists of a bimodal mafic-felsic volcanic succession interbedded with clastic and chemical sedimentary rocks. It is between 0 and 1300 m thick (Schermerhorn 1971) and has been dated as Late Fammenian to Late Viséan (Oliveira 1990;Barrie et al. 2002;Dunning et al. 2002;Rosa et al. 2009). All of the massive sulphide deposits in the Iberian Pyrite Belt are hosted by the Volcano-Sedimentary Complex.
The Culm Group (Fig. 1b), the uppermost Iberian Pyrite Belt unit, consists of shale, litharenite and rare conglomerate with turbiditic features and has a maximum thickness of about 3000 m (Moreno 1993). The Culm Group ranges in age from Late Viséan to Middle-Late Pennsylvanian; it is interpreted to represent a synorogenic flysch related to the Variscan tectonic event and was partly sourced from the Volcano-Sedimentary Complex (Moreno 1993).
Volcanism within the Volcano-Sedimentary Complex is represented by a succession of bimodal volcanic rocks (Thiéblemont et al. 1998), with felsic rocks usually predominant. Magmatism was coeval with extensional tectonics. A mantle plume that developed under SW Iberia in the Early Carboniferous is the most plausible geodynamic scenario. The mantle plume would have been most active during the period 355-335 Ma, the age of most Variscan mafic rocks (Simancas et al. 2003).
After emplacement, the volcanic rocks of the Volcano-Sedimentary Complex have undergone several alteration processes, including hydration of volcanic glass and later hydrothermal alteration. The latter process comprises pervasive, regional alteration with low water-rock ratios, as well as the formation of local, intense chlorite-sericite haloes around ore deposits. Regional, very low-grade metamorphism also occurred (Silva et al. 1990;Fernández-Caliani et al. 1994). The effects of each of these alteration processes are difficult to distinguish and some of them are imprecisely defined. For instance, in some cases regional alteration and metamorphism have been referred to as hydrothermal metamorphism (Munhá & Kerrich 1980).
Plutonic rocks (Fig. 1b) comprise gabbros, diorites, tonalites and granites (Schutz et al. 1987;De la Rosa 1992). Although most of these rocks were emplaced after the main Variscan deformation phase affecting the Volcano-Sedimentary Complex (Simancas 1983;De la Rosa 1992;Quesada 1998), some plutonic rocks, close to the Riotinto mining district, probably intruded earlier (Stein et al. 1996). The geochemical similarities between plutonic and volcanic rocks (Schutz et al. 1987;Thiéblemont et al. 1998) as well as the available U/Pb evidence (Barrie et al. 2002;Dunning et al. 2002), also suggest that plutonic rocks were at least partly coeval with the Volcano-Sedimentary Complex.
VHMS deposits in the Iberian Pyrite Belt include not only those in the Riotinto mining district, but also a number of other supergiant deposits and more than 100 smaller VHMS bodies, some of them high-grade, occurring both in Spain and Portugal (Leistel et al. 1998). More than 400 small manganese deposits also occur in the Iberian Pyrite Belt. A number of VHMS deposits formed in Late Strunian (latest Famennian) time (González et al. 2002;Oliveira et al. 2003;González 2005;Rosa et al. 2009), whereas others are distinctly younger (Nesbitt et al. 1999;Barrie et al. 2002;Tornos 2006).
It is apparent that any potential interest of the study of volcanic rocks in the Volcano-Sedimentary Complex arises from their general relation in time and space with VHMS deposits. On the other hand, recent regional studies in the Spanish Iberian Pyrite Belt (ITGE 1999) have indicated a subdivision of this area into a number of units, each of them consisting of contrasting volcanic and stratigraphic successions, as well as VHMS deposits exhibiting different features. Therefore, detailed studies of volcanic rocks must be performed within each of these units. Accordingly, we have focused first on the Riotinto-Nerva unit (ITGE 1999), which contains a thick volcanic succession hosting the VHMS in the Riotinto mining district. Considering this unit, we have studied the Odiel river section in detail because volcanic rocks continuously crop out in a very well-exposed, relatively unaltered area, so that stratigraphic relationships between the various volcanic successions can be better assessed.
Volcanic rocks of the Riotinto-Nerva unit: the Odiel river area In the Riotinto-Nerva unit, three main stratigraphic successions are distinguished in the Volcano-Sedimentary Complex from base to top ( Fig. 2): (1) a lower mafic-siliciclastic succession of Strunian age (Rodríguez et al. 2002), containing a number of dark shale horizons and basaltic rocks, both volcanic and subvolcanic (García-Palomero 1980;Boulter 1993aBoulter ,b, 1996Tornos & Almodóvar 2003;Boulter et al. 2004;Mellado et al. 2006); (2) a volcanic felsic succession in which volcaniclastic rocks alternate with rhyolitic flows and sills (Boulter 1993a;Pascual et al. 2000;Valenzuela et al. 2002;Boulter et al. 2004); (3) an upper sedimentary succession. The occurrence of rhyolite fragments within this last unit, as well as the chloritic alteration haloes within the felsic succession, indicate that the VHMS deposits in the Riotinto-Nerva unit postdate most of the Volcano-Sedimentary Complex sequence, at least in the Riotinto mining district (García-Palomero 1980).
To assess the role and evolution of magmatism in the Riotinto-Nerva unit in detail, we have carried out a study in the Odiel river section, located about 10 km west of Riotinto (Fig.  2). This section constitutes a monoclinal sequence dipping to the north, separated from the Riotinto antiform by a narrow synform. Major basic and rhyolitic successions are traceable continuously from the Odiel area ( Fig. 3a) to the Riotinto mining district (Fig.  3b) and eastwards, at least as far as to the Jarama river section (Boulter 1993a) (Fig. 3c), with local variations that are to be expected in volcanic successions. In the Odiel river section, the Volcano-Sedimentary Complex crops out continuously with a minimum thickness of 1000 m. Most of the igneous rocks are only weakly deformed and, although regionally altered, are not affected by sericitic or chloritic alteration haloes related to VHMS deposits.
Studies in the Odiel river area (Fig. 3a) have revealed a number of volcanic and subvolcanic facies, indicating a complex scenario in which bimodal magmatism took place, including basic and felsic rocks. Field evidence indicates successive emplacement of basaltic rocks, including sills, pillow lavas and basaltic breccias, followed by felsic volcanic rocks belonging to two groups that change in composition with time. These two groups contain both volcaniclastic and coherent rocks, and are characterized by contrasting petrographic and chemical features. Accordingly, they have been respectively named plagioclasephyric rhyolite and plagioclase-quartz-phyric rhyolite volcanic successions (Fig. 3a). Apart from basaltic and rhyolitic volcanic successions, an intrusive sill, composed of plagioclase-quartzphyric dacites, occurs at the base of the Volcano-Sedimentary Complex in the Odiel area.
Because of the close spatial and temporal relation with VHMS deposits throughout the Riotinto-Nerva unit, we have focused our study on felsic rocks. Among these, the plagioclase-phyric and plagioclase-quartz-phyric rhyolite volcanic successions can be traced throughout this unit, whereas the basal porphyritic dacite rocks, representing an intrusive body, crop out only in the Odiel area (Fig. 3a).
The plagioclase-phyric rhyolite volcanic succession overlies a black shale horizon that separates it from the lower basic succession. The bottom of the plagioclase-phyric rhyolite succession is formed by a pumice breccia, followed by a number of horizons of graded, monomictic breccias. The top of the succession consists of a lava flow that is up to 60 m thick and is continuous over more than 3 km. All of these rocks are classified together because they have similar chemical features, as described below. Apart from the basal pumice breccia, which is aphyric, the fragments of the volcaniclastic rocks and the lava flow also have similar petrographic features, showing porphyritic textures only with sparse plagioclase crystals (Fig. 4a).
The plagioclase-quartz-phyric rhyolite volcanic succession is separated from the plagioclase-phyric rhyolite by a continuous chert horizon. It consists of four sills, continuous over distances up to 2 km, with thicknesses ranging from 20 to 100 m, interbedded with breccias, all of them sharing petrographic and chemical features that distinctly contrast with the plagioclasephyric rhyolite volcanic succession. Plagioclase-quartz-phyric rhyolite rocks, either coherent or volcaniclastic, have a porphyritic texture with plagioclase and quartz phenocrysts embedded in a fine-grained equigranular groundmass. Quartz-plagioclase micrographic aggregates up to 3 mm in diameter, occurring only in the plagioclase-quartz-phyric rhyolite, are very common (Fig. 4b). (c) Jarama river (Boulter 1993a). Symbols mark the stratigraphic position of samples used in whole-rock chemical analyses and isotopic study.
The plagioclase-quartz-phyric dacite consists of a massive, tabular coherent body. It has contacts at its top and bottom with the basaltic pillow lava and breccia and the Phyllite-Quartzite Group, respectively (Fig. 3a). It has a porphyritic texture, consisting of up to 25% phenocrysts of albitized plagioclase and embayed quartz in a microcrystalline groundmass (Fig. 4c). We suggest a subvolcanic origin in view of its high phenocryst content, which strongly contrasts with felsic rocks in regional volcanic successions. Although located at the base of the Volcano-Sedimentary Complex, most of the contacts at the bottom and top are tectonic, so that field relationships with host sediments cannot be clearly interpreted. Nevertheless, U-Pb dating confirms that plagioclase-quartz-phyric dacite is older than the rest of the felsic rock groups, as shown below.
Although the above field and petrographic descriptions are based on the Odiel section, felsic sequences in other areas in the Riotinto-Nerva unit can be correlated with this section. For instance, pumice breccias equivalent to the plagioclase-phyric rhyolite in the Odiel section also occur in the Riotinto mining district at the same stratigraphic position (Mellado et al. 2006;Fig. 3). Correlation of the plagioclase-quartz-phyric rhyolite away from the Odiel section is supported as well by petrographic, chemical and U-Pb dating.

Whole-rock major-and trace-element data
A threefold classification of the felsic rocks in the Riotinto-Nerva unit emerges on the basis of the above stratigraphic and petrographic descriptions: the successive plagioclase-quartz-phyric dacite, plagioclase-phyric rhyolite and plagioclase-quartzphyric rhyolite groups differ not only in stratigraphic position, but also in petrography. To assess if these differences also imply successive volcanic cycles, also contrasting in geochemistry, we have obtained major-and trace-element chemical analyses from 17 samples of the three volcanic groups, collected in the Odiel river section and the Riotinto mining district. We consider this number of samples sufficient, in view of the internal petrographic homogeneity of each of the rock groups under study.
Whole-rock analyses were performed at SGS Laboratories in Canada by a combination of analytical techniques, including X-ray fluorescence spectrometry (XRF) and inductively coupled plasma mass spectrometry (ICP-MS). Data are given in Table 1.

Chemical effects of alteration
To provide quantitative estimates of the intensity of alteration, the AI-CCPI (Ishikawa alteration index v. chlorite-carbonatepyrite index) bivariate plot of Large et al. (2001) has been used ( Fig. 5). Most of the felsic rocks analysed plot within the leastaltered box. The four exceptions shown correspond to three regionally altered plagioclase-quartz-phyric rhyolites, one from the Odiel river and two from the Riotinto area, plus a sample from the stockwork zone in the San Dionisio orebody in Riotinto, which plots farther outside the least-altered box towards the K-feldspar + sericite end member. In all samples, however, the large scatter of most mobile elements (Na 2 O, K 2 O, CaO, Rb, Ba) when plotted against TiO 2 is consistent with their mobility during hydrothermal alteration (Lentz 1999;Large et al. 2001). In contrast, Al 2 O 3 , P 2 O 5 , HFSE (Zr, Y, Hf, REE), Sc and V all show linear trends with TiO 2 (Fig. 5), attesting to their relative immobility (Piercey et al. 2001;Dusel-Bacon et al. 2004;Gifkins et al. 2005). Alteration precludes the use of total alkalis-silica (TAS) for rock classification.

Immobile element geochemistry
Variation diagrams of immobile elements v. TiO 2 confirm the existence of the three types of felsic rocks previously distin-  guished by field and petrographic study. Plagioclase-phyric and plagioclase-quartz-phyric rhyolites are characterized by their chemical homogeneity, whereas the plagioclase-quartz-phyric dacite samples tend to show well-defined linear variation trends with regard to TiO 2 (Fig. 6). The Zr/TiO 2 v. Nb/Y discrimination diagram of Winchester & Floyd (1977) (Fig. 7) shows that the Odiel river felsic rocks have a subalkaline character. The plagioclase-quartz-phyric dacite samples plot within the dacite and rhyodacite fields, whereas the plagioclase-phyric and plagioclase-quartz-phyric rhyolites straddle the dacite-rhyodacite and rhyolite border. On primitive mantle-normalized spider plots, the felsic rocks have downward-sloping profiles with light rare earth element (LREE) enrichment and distinctive negative Nb, Sr, Eu and Ti anomalies (Fig. 8). Plagioclase-quartz-phyric dacites are dacites characterized by moderate Zr/Y and intermediate abundances of HFSE (Y, Zr, Hf). Plagioclase-phyric rhyolites show similar, moderate Zr/Y but slightly higher abundances of HFSE and heavy REE (HREE) than plagioclase-quartz-phyric dacite. In contrast, plagioclase-quartz-phyric rhyolites are characterized by low Zr/Y and HFSE abundances and more pronounced Sr and Ti anomalies.
To compare the dacites and rhyolites in the Riotinto mining district with those occurring in other VHMS areas, we have plotted them on the classification proposed by Hart et al. (2004). The felsic rocks of the Riotinto mining district show chemical features similar to FII dacite-rhyolite, ascribed by Lesher et al. (1986) and Hart et al. (2004) to a calc-alkaline affinity (Fig. 9).
Finally, felsic rocks display similar REE patterns (Fig. 10), with (La/Sm) N . 1, negative Eu anomalies and (Tb/Yb) N < 1. However, plagioclase-quartz-phyric rhyolites, as well as the rhyolitic rocks hosting the Riotinto VHMS, show significantly more pronounced, negative Eu anomalies (Fig. 10). In all, the REE patterns are closely similar to those of felsic volcanic rocks in other areas in the Iberian Pyrite Belt (Thiéblemont et al. 1998;Rosa et al. 2006;Barrett et al. 2008).

U-Pb geochronology
We include data for all of the felsic rocks we have distinguished in the Odiel river sequence: ISOD-22 (plagioclase-phyric rhyolite), ISOD-23 (plagioclase-quartz-phyric rhyolite), and ISOD-24 (plagioclase-quartz-phyric dacite). To assess the timing of the Odiel volcanism with regard to the VHMS deposits in the Riotinto mining district, we have also analysed two samples of rhyolites from the San Dionisio orebody (Fig. 3b) in the Riotinto mining district, an intensely altered rhyolite in the footwall of the VHMS (sample ISOD-26) and a regionally altered rhyolite (sample ISOD-27). These latter rocks are petrographically equivalent to the plagioclase-quartz-phyric rhyolite in the Odiel river section.
Zircons from these felsic rocks samples were analysed by high-precision isotope dilution thermal ionization mass spectrometry (ID-TIMS) at the Jack Satterly Geochronology Laboratory (Department of Geology, University of Toronto, Canada). U-Pb data for all samples are presented in Table 2 and the results are displayed graphically in the associated U-Pb concordia plots (Fig. 11). In these, associated ages and errors were calculated using the IsoPlot/Ex (v. 3.00) program of Ludwig (2003).
In some cases, the U-Pb magmatic age constraint is provided not by a cluster of data, but by one or two youngest identified single grain fractions among a population showing variable inheritance; for these samples, the ages are strictly interpreted to represent maximum age limits and the actual time of magmatism may be slightly younger. U-Pb dating from the Odiel river section shows that ISOD-24 (plagioclase-quartz-phyric dacite) is significantly older (351.5 AE 0.4 Ma) than associated ISOD-22 (plagioclase-phyric rhyolite) (349.4 AE 0.6 Ma), and represents the earliest felsic magmatic episode in the area ( Fig. 11a and b). Consistent with their relative stratigraphic position, the age of ISOD-23 (plagioclase-quartz-phyric rhyolite) postdates that of ISOD-22 at 345.7 AE 0.6 Ma (Fig. 11c). The plagioclase-quartzphyric rhyolite samples hosting the San Dionisio orebody have an age of 348.4 AE 0.7 Ma (ISOD-26), based on a single young analysis (Z3) (Fig. 11d), and 349.5 AE 0.4 Ma (ISOD-27) (Fig.  11e); this latter age is close to that obtained by Barrie et al. (2002).

Sm-Nd isotopic data
To evaluate the petrological significance of the differences in age and geochemical features found among plagioclase-quartzphyric dacite, plagioclase-phyric rhyolite and plagioclase-quartzphyric rhyolite felsic rocks, we have obtained Sm-Nd isotope data from the same five samples as used for U-Pb dating. Sm-Nd data are resistant to disturbances caused by low-grade metamorphism and alteration processes by aqueous solutions, and therefore offer a powerful tracer of the potential sources of meta-igneous rocks. Specifically, the 143 Nd/ 144 Nd isotopic ratio corrected for in situ radioactive decay of 147 Sm since the time of igneous emplacement, usually expressed by using the å-notation, reflects the time-integrated value of the Sm/Nd ratio; that is, the degree of fractionation of the LREE in the source reservoir of the sample analysed.
Sm and Nd concentrations were measured by ID-TIMS at the Université Blaise-Pascal (Clermont-Ferrand, France), using the analytical methods of Pin et al. (1990). å Nd(350) values were calculated relative to a chondritic uniform reservoir (CHUR, equivalent to Bulk Earth) with the following present-day characteristics: 143 Nd/ 144 Nd ¼ 0.512638, 147 Sm/ 144 Nd ¼ 0.1966 (Jacobsen & Wasserburg 1980). T DM neodymium model ages were calculated relative to a model depleted mantle described by å Nd (T) ¼ 0.25T 2 À 3T + 8.5, where T is age in Ga (DePaolo 1981). Data are given in Table 3.
The compositional changes between the groups described above are also reflected in tracer Nd isotopic data. Plagioclasequartz-phyric dacite and plagioclase-phyric rhyolite show similar initial å Nd values, slightly positive and with little or no mutual contrast (+0.3, +0.4), whereas the plagioclase-quartz-phyric rhyolite and the samples hosting the VHMS in Riotinto (ISOD-26 and ISOD-27) show distinctly negative initial å Nd values of À1.7 to À2.8.

Discussion
Source of the felsic rocks in the Riotinto-Nerva unit: progressive melting and crustal thinning The above data confirm that felsic rocks in the Riotinto-Nerva unit formed during a protracted interval of magmatism, lasting at least 6 Ma, as indicated by previous U-Pb dating (Barrie et al. 2002). Probably, this time interval was significantly longer  because in the Odiel river area, as well as in the whole unit, thick pillow lava packages constitute the lower part of the volcanic succession. These pillow lavas are interbedded with black shales, which have been dated here as Strunian (c. 360 Ma, Rodríguez et al. 2002), as well as in other areas in the Iberian Pyrite Belt (González et al. 2002). This implies that igneous activity probably lasted a minimum of 15 Ma.
The Odiel succession indicates a scenario dominated by bimodal magmatism with few or no intermediate rocks, in which compositional variation trends of felsic and basic rocks point to an independent genesis and evolution of basic and felsic magmas, as previously recognized in the Volcano-Sedimentary Complex (Munhá 1983;Mitjavila et al. 1997;Thiéblemont et al. 1998). In the Odiel section, this is shown by the TiO 2 v. Zr plot (Fig. 12), among others. Therefore, fractional crystallization of basic magmas cannot account for the genesis of felsic magmas, even considering the most primitive ones. Similarly, mixing of felsic and basic magmas or crustal assimilation by basic magmas must have had, at most, a minor role in the genesis of felsic magmas in the Volcano-Sedimentary Complex.
Assuming the above constraints, variations in å Nd found in felsic rocks (Fig. 13) must indicate source variation with time from dacitic to rhyolitic compositions. Accordingly, plagioclasequartz-phyric dacite and plagioclase-phyric rhyolite could have been generated by partial melting of depleted upper mantle and/ or juvenile mafic crust, whereas plagioclase-quartz-phyric rhyolite magmas, both in the Odiel and in Riotinto areas, would have formed by partial melting involving more evolved crustal segments. In the rocks under study, a crustal magma source is more consistent with the negative Nb anomalies shown by all the felsic rocks (Fig. 8).
Partial melting of mafic sources has been proposed to form felsic rocks similar to plagioclase-quartz-phyric dacite and plagioclase-phyric rhyolite in other bimodal volcanic successions in VHMS areas in a wide range of P-T conditions and tectonic environments (Lentz & Goodfellow 1992;Hart et al. 2004). In our case, the similar å Nd signature shown by the plagioclasequartz-phyric dacite and plagioclase-phyric rhyolite indicates no mutual compositional contrast in the source areas. Also, Tb/Yb values close to unity shown by both rock groups indicate that melting did not occur in equilibrium with residual garnet. Accordingly, we suggest that plagioclase-quartz-phyric dacite and plagioclase-phyric rhyolite melts were generated from chemically similar sources, at pressures equivalent to 7.5 kbar or less. Considering the experimental data (Beard & Lofgren 1991;Wyllie & Wolf 1993;Wolf & Wyllie 1995), compositional differences between plagioclase-quartz-phyric dacite and plagioclase-phyric rhyolite can be explained by two-stage partial melting of a similar mafic source, plagioclase-phyric rhyolite having been generated from a more intensely dehydrated crustal segment, at distinctly lower P and higher T. The time span between plagioclase-quartz-phyric dacite and plagioclase-phyric rhyolite may account for high-T crustal dehydration before  plagioclase-phyric rhyolite genesis, involving formation of large amounts of clinopyroxene in the source area by amphibole breakdown.
The above interpretation may account for most of the geochemical differences between plagioclase-quartz-phyric dacite and plagioclase-phyric rhyolite (Fig. 8). Plagioclase-phyric rhyolite melts, formed from a source equilibrated with clinopyroxene and larger plagioclase amounts, are accordingly characterized by lower Sr, higher REE abundance and Eu/Eu* ratios (Fig. 10). On the other hand, the higher-T, lower melting rate of such a dehydrated source accounts for the lower Ti and higher Zr and Hf contents of plagioclase-phyric rhyolite melts (Fig. 6). Regarding the lower Ti content in the plagioclase-phyric rhyolites, this is more probably related to low melting rate because it cannot be explained by significant differences in magnetite or ilmenite content in the source rocks (Beard & Lofgren 1991). Higher Zr and Hf are to be related to the lower mineral-melt partition coefficient of these elements in pyroxenes with regard to amphiboles (Deering et al. 2008). Moreover, these higher HFSE elements would be more soluble in these high-T rhyolites because of their less polymerized structure (Hart et al. 2004).
This suggested explanation of the contrasting Ti and HFSE distribution in plagioclase-quartz-phyric dacite and plagioclasephyric rhyolite is based on considering Zr and Hf as incompatible elements, implying that zircon was scarce or absent during melting events. This interpretation is consistent with the low å Nd ratios of plagioclase-quartz-phyric dacite and plagioclase-phyric rhyolite, indicating mafic sources in which zircon abundance is unlikely. Also, if the higher Zr and Hf abundance in plagioclasephyric rhyolite was caused by highly efficient melting of refractory minerals, such as zircon, this should have been coupled with higher melting rates of mafic minerals, resulting in a positive Ti-Zr correlation between plagioclase-quartz-phyric dacite and plagioclase-phyric rhyolite, as described in other VHMS areas (Piercey et al. 2008).
The plagioclase-quartz-phyric rhyolite rocks, representing the latest stage of felsic magmatism in the Riotinto-Nerva unit, display markedly lower, negative initial å Nd values, corresponding to more evolved crustal sections. They additionally differ from plagioclase-phyric rhyolite in a number of other chemical features, including lower Ti, Sr, Zr and Hf contents and distinctly lower Eu/Eu* values. Both Sr contents and Eu/Eu* ratios may be explained by melting of a plagioclase-richer source, whereas low Ti could be interpreted as a result of a more leucocratic nature of the protolith. Low Zr and Hf contents indicate that these elements did not behave as trace elements during melting, but as essential structural components (sensu Hanson 1989), probably owing to the occurrence of zircon grains in the source rocks and low melting degree of zircon and/or other refractory accessories. Both HFSE and å Nd data suggest the participation of shallower crustal source rocks, not necessarily mafic, in the genesis of plagioclase-quartz-phyric rhyolite, again suggesting that melting gradually affected horizons of an increasingly attenuated crust. To discriminate between possible types of such source areas (e.g. more 'crustal' mafic sources (Hart et al. 2004) or a metasedimentary source contribution (Reid 1983)) is precluded by difficulties in mass-balance calculations, owing to rock alteration resulting in high SiO 2 , Na 2 O and K 2 O mobility.
Considering U-Pb ages together with the additional constraint of Nd isotopic signatures (Fig. 13), the data indicate that the generation of magmas from a shallower crust was not strictly synchronous along the Riotinto-Nerva unit, beginning slightly earlier in the Riotinto area than in the Odiel river area. However, emplacement of felsic rocks in Riotinto probably continued later, as shown by the fact that the U-Pb age of the plutonic rocks in the Riotinto area (Campofrio biotite tonalite, Barrie et al. 2002) is equivalent to that of the plagioclase-quartz-phyric rhyolite in the Odiel section.
We therefore suggest that the bimodal igneous activity reflects a transient thermal pulse, in a locally transtensional tectonic regime. This thermal pulse would be related to the emplacement of a large basic sill identified in seismic reflection profiles  Pb* is total amount (in picograms) of radiogenic Pb. Pb C is total measured common Pb (in picograms) assuming the isotopic composition of laboratory blank: 206/204, 18.221; 207/204, 15.612; 208/204, 39.360 (errors of 2% (Simancas et al. 2003). This scenario implies a within-plate setting, in a (probably oblique) extensional setting, also to account for the deposition of voluminous siliciclastic material during and after igneous activity. This basin was rather shortlived and later evolved into a fold-and-thrust belt (Silva et al. 1990;Oliveira & Quesada 1998).

Felsic magmatism and genesis of VHMS deposits
Given the age and negative initial å Nd values found both in footwall and regional rhyolitic rocks in the Riotinto mining district area, it is apparent that the VHMS deposits therein postdate the emplacement of the upper plagioclase-quartz-phyric rhyolite volcanic succession in the Riotinto-Nerva unit. Therefore, these orebodies postdate a thick volcanic succession, which lasted at least 15 Ma as indicated by combined palynological and U-Pb evidence (Barrie et al. 2002;Mellado et al. 2006). This volcanic succession formed in a transtensional environment, following a scenario characterized by mantle upwelling and crustal thinning. Such scenarios are not rare in VHMS areas, because elevated geothermal gradients would have had the elevated heat flow needed to drive robust hydrothermal systems at sufficiently permeable levels (Cathles 1991;Cathles et al. 1997;, in particular in subaqueous volcanic environments in which both crustal thinning and enhanced structural permeability occur (Sillitoe 1982;Lentz 1998;Hart et al. 2004). It has been accordingly suggested that the occurrence of rocks reflecting melting of shallower areas of this crust, having formed at a higher temperature, would indicate the most favourable conditions for VHMS formation. If so, selected geochemical features of the felsic rocks could be used as a proxy in VHMS exploration. This idea has been previously developed in other VHMS areas by a number of researchers (Mortensen & Godwin 1982;Lesher et al. 1986;Lentz 1998;Piercey et al. 2001Piercey et al. , 2008Dusel-Bacon et al. 2004;Hart et al. 2004), in most instances stressing the role of HFSE and REE in discrimination between 'fertile' and barren rhyolites, particularly in the case of those  mining districts in which mineralization is episodic throughout an interval of several million years, including the Iberian Pyrite Belt.
The application of the above geochemical models to the felsic rocks in the Riotinto-Nerva unit, however, does not allow discrimination between 'fertile' and barren rhyolites. For instance, using the geochemical parameters proposed by Hart et al. (2004) leads to a same classification of the plagioclase-quartzphyric dacite, plagioclase-phyric rhyolite and plagioclase-quartzphyric rhyolite as FII rhyolites (Fig. 9), with no discrimination between VHMS-related and barren rocks. Moreover, all the studied rocks are classified as having a low VHMS productivity, including the plagioclase-quartz-phyric rhyolite directly related to the world-class Riotinto VHMS deposit. Similar discrepancies have been also described in other VHMS areas worldwide (Gaboury & Pearson 2008;Piercey et al. 2008). The use of Zr and Hf abundances, specifically suggested for Phanerozoic VHMS areas (Piercey et al. 2008), also leads to inconsistent results in this case, as the Zr-and Hf-richer, plagioclase-phyric rhyolites are unrelated to mineralization.
Alternatively, we consider that å Nd study, coupled with U-Pb dating, is a reliable geochemical tool in assessing progressive crustal melting, hence favouring optimal conditions for sustained hydrothermal activity; younger, lower-å Nd rhyolites are a proxy in VHMS exploration in the Riotinto-Nerva unit, as possibly are low-å Nd felsic rocks throughout the Iberian Pyrite Belt. Other geochemical parameters, including lower Ti, Sr, Zr and Hf contents as well as more pronounced Eu anomalies, could also contribute to discriminate 'fertile' rhyolites. These latter features, however, should be used together with å Nd determination and applied only to the Riotinto-Nerva unit because, as previously suggested by Gaboury & Pearson (2008), both the geochemical features and the evolution of the crust are probably variable in any large VHMS province, and even more so if different VHMS provinces are compared.
The above suggestions imply that å Nd and geochemical parameters can contribute to provide constraints in VHMS exploration. This does not mean, however, that any of these parameters alone could account for VHMS genesis in the Iberian Pyrite Belt, as ore genesis, including heat and metal source, must be controlled by a number of other factors. In fact, VHMS deposits in the Iberian Pyrite Belt cannot be grouped in a single category. In contrast, they form an exceptional number of supergiant deposits together with more than 100 smaller masses. Moreover, both supergiant and smaller deposits formed at distinctly different times (Barrie et al. 2002), were hosted by diverse rock types and strongly differ in ore grade (Leistel et al. 1998). Such diversity within a single province implies that several factors, each having different influence in each case, are to be invoked in VHMS genesis in the Iberian Pyrite Belt. Among these, extensional regime, felsic magmatism, and anoxic bottom waters must account for the genesis of ore deposits. Beyond these genetic differences, however, it is apparent that most of the footwall rocks underlying VHMS deposits in the Iberian Pyrite Belt consist of felsic volcanic rocks, a feature that the Iberian Pyrite Belt orebodies share with most VHMS deposits worldwide (Allen et al. 2005).
Taking these constraints in account, we interpret that the lowå Nd rhyolites in the Riotinto-Nerva unit are an indirect prerequisite to VHMS genesis because they marked a more advanced stage in basinal evolution, at which sustained hydrothermal activity could have been more effective in producing VHMS deposits. In contrast, less evolved, high-å Nd rhyolites could not be associated with VHMS deposits because at earlier stages of crustal melting the basin had not yet attained the adequate conditions for ore deposition.

Conclusions
We present new geochemical, Sm-Nd and U-Pb data on the felsic volcanic rocks in a stratigraphic section at the Odiel river, belonging to the Riotinto-Nerva unit, in the Iberian Pyrite Belt, Spain. U-Pb dating has permitted three successive stages of felsic magmatism to be distinguished between 351.5 AE 0.4 and 345.7 AE 0.6 Ma, whereas å Nd variation from +0.3 to À2.8 indicates a progressively higher contribution of an evolved crust to volcanism. The evolution of felsic volcanism with time in the  Riotinto-Nerva unit is also marked by significant differences in petrographic and geochemical features, the younger rhyolitic rocks being characterized by lower Ti, Sr, Zr and Hf and a more pronounced, negative Eu/Eu* ratio. Geological evidence, å Nd data and U-Pb ages indicate that regional and footwall rocks hosting the world-class VHMS deposits in the Riotinto mining district correspond to the latest felsic volcanic stage in the Riotinto-Nerva unit.
We interpret that these successive stages of felsic magmatism were related to melting of progressively shallower crustal levels in a transtensional geodynamic environment, coeval with underplating by huge amount of basic rocks. This interpretation, involving crustal thinning, is consistent with regional tectonic interpretations, as well as with geophysical evidence (Simancas et al. 2003).
The link between VHMS deposits and younger, lower å Nd rhyolites suggests that crustal thinning was a prerequisite in VHMS formation, as previously interpreted in other VHMS areas (Hart et al. 2004). However, geochemical criteria used previously as indicators of high-T magmatism in an attenuated crust (Hart et al. 2004;Piercey et al. 2008) do not fit the geochemical features of the studied felsic rocks. Alternatively, we suggest that lower å Nd values discriminate shallow melting in the Iberian Pyrite Belt, indicating optimal conditions of hydrothermal circulation and ore deposition and thus constituting a useful proxy in mineral exploration.