Dating metasomatic events in the lithospheric mantle beneath the Calatrava volcanic field (central Spain)

We report the first attempt to date metasomatic events in peridotite xenoliths from the Subcontinental Lithospheric Mantle (SCLM) beneath the Cenozoic Calatrava volcanic field of central Spain. The most metasomatized xenoliths of the El Aprisco olivine melilitite maar were selected to perform a geochronological study on metasomatic apatite (U-Pb method) and amphibole (Ar-Ar), integrated with an enlarged chemical data set on these minerals. The metasomatic agents in studied samples are mainly carbonate-rich ultra-alkaline melts of prob-able asthenospheric derivation. Some samples have been overprinted by more than one metasomatic event. The geochronological data confirm three metasomatic events that occurred within the SCLM beneath central Spain in Cretaceous (118 Ma), Oligocene (29 Ma), and Miocene (16–4 Ma) times, much earlier than the host volcanic magmatism. To date, no magmatic events of those ages have been recorded in central Spain. However, a correlation with several cycles of sporadic intraplate magmatism of alkaline affinity in the Iberian microplate is suggested. This study illustrates that the SCLM preserves the memory of a complex history of melt and/or fluid percolation processes in a metasomatic record that is generally unrelated to shallower crustal magmatic events. Sm, Eu, Dy, Yb, and Ti (McKenzie and O’Nions, 1991). The compositional range of Italian carbonatites is from Stoppa and Sciazza (2013), whereas the composition of the Calatrava host volcanic rocks (melilitites and nephelinites) are from Cebriá and López-Ruiz (1995). Normalizing values are from McDonough and Sun (1995).


INTRODUCTION
Fluid or melt infiltration of the mantle in anorogenic settings may produce different styles of metasomatism, including modal, cryptic and stealth (e.g., Menzies et al., 1987;Ionov et al., 1997;O'Reilly and Griffin, 2013). The analyses of major, minor, and trace elements in metasomatic minerals (amphibole, clinopyroxene, apatite, phlogopite, carbonate) and interstitial glass, often associated with these minerals, is a powerful tool for unraveling the nature and composition of melts/fluids responsible for the different styles of metasomatism (e.g., Coltorti et al., 2007a;Raffone et al., 2009). Furthermore, several studies, based on the analysis of whole-rock or mineral separates, have attempted to date metasomatism of mantle rocks using an array of radiogenic isotopes (Rb-Sr, Sm-Nd, Pb-Pb, and Lu-Hf isochrons), but these have generally yielded spurious results (e.g., Nasir and Rollinson, 2009;Downes et al., 2015). Recently, the refinement of in situ analytical techniques for the Re-Os isotopic system has provided a more nuanced picture of the Re-Os system in basemetal sulfides and platinum-group elements from mantle-derived rocks, showing wide variability in the Os-isotope composition of these minerals (e.g., Alard et al., 2002;Ackerman et al., 2009). Such Os-isotope heterogeneity is reflected in complex distributions of Os-model ages recording the protracted passageway of melt/fluids through the mantle over a long period of time (e.g., González-Jiménez et al., 2013). Unfortunately, basemetal sulfides are highly sensitive to Re-Os isotopic isotopic exchange with Os-bearing fluid/melt(s), resulting in changes to the Re-Os isotopic system and mixed or meaningless model ages (Zangana, 1995). In this scenario, only when metasomatic minerals appear in amounts that are amenable to mineral separation or in situ techniques (zircon, apatite, amphibole, phlogopite) can dating success be achieved. Currently, the most widely used high-precision geochronometers are those employing the U-Pb method for accessory minerals and the 40 Ar/ 39 Ar method for K-rich minerals (Schmitz and Kuiper, 2013). The U-Pb method has been used in those rare mantle rocks that contain zircon (e.g., Zheng et al., 2006) or apatite (e.g., Morishita et al., 2008). The Ar-Ar dating method has been performed in the most common modal metasomatic minerals in mantle xenoliths: phlogopite (e.g., Wartho and Kelley, 2003;Hopp et al., 2008) and amphibole (e.g., Wartho and Kelley, 2003) and relies critically on the assumption of closed system behavior in the mantle environment.
Many peridotite xenoliths within volcanic rocks preserve a record of multi-stage metasomatism testifying to the percolation of magmatic melts and fluids within the lithospheric mantle. The potential correlation between events of metasomatism and events of magmatism recorded in the overlying crust is still unclear, suggesting that a significant portion of the magmatic budget that crosscuts a lithospheric section never reaches shallow crustal levels or the Earth's surface. One robust way to decipher the complexity of magmatic plumbing systems is to analyze and date the volcanic rocks (including their crystal cargo) and the metasomatic phases in peridotite xenoliths hosted by the volcanic rocks. The Calatrava volcanic field is an intraplate alkaline zone that has volcanic vents carrying mantle xenoliths of the Subcontinental Lithospheric Mantle (SCLM), which have recently been characterized (e.g., Bianchini et al., 2010;Villaseca et al., 2010;Lierenfeld and Mattsson, 2015). The involvement of different metasomatic agents in some of these peridotite xenolith suites (e.g., El Aprisco suite: Villaseca et al., 2010;González-Jiménez et al., 2014) provides an opportunity to precisely assess the chemical and temporal evolution of mantle metasomatic events, efforts not undertaken in previous studies of this volcanic field. Thus, the aims of this study are twofold: (1) to characterize and date metasomatic events in the lithospheric mantle beneath central Spain, and (2) to correlate these metasomatic events with the magmatism recorded in the Iberian Peninsula and its western Atlantic margin. Correlations of magmatic processes occurring at mantle depths with those exposed at the Earth's surface are anticipated to fill the gaps in knowledge of igneous processes occurring in central Iberia during Mesozoic and Cenozoic times. In order to achieve this goal we selected the most strongly metasomatized xenoliths of the El Aprisco olivine melilitite maar to perform a geochronological study on metasomatic apatite (U-Pb method) and amphibole (Ar-Ar dating), with an enlarged trace element and isotopic chemical data set for these minerals.

GEOLOGICAL SETTING
The Calatrava Volcanic Field (CVF) is an anorogenic intracontinental zone formed in Neogene time within the circum-Mediterranean province. This volcanic field comprises more than 200 volcanic centers in an area of around 5500 km 2 (Ancochea, 1982;Fig. 1) located at the boundary of small Cenozoic sedimentary basins that are transgressive over the Variscan basement (Fig. 1). The distribution of volcanic vents broadly follows Variscan shear bands reactivated by the main compressional axes of the Betic collision (Cebriá et al., 2011). Their geodynamic setting is still controversial, with debate focused around three main models: (1) volcanism related to asthenospheric-mantle upwelling (hot-spot or diapiric instabilities) in a pre-rifting stage (Ancochea, 1982;López-Ruiz et al., 1993); (2) a megafault system affecting the western Mediterranean (López- Ruiz et al., 2002); (3) uplifting of the Variscan basement with lithospheric folding in the Calatrava region exploiting a reactivation of previous NW-SE-trending Variscan structures formed during the indentation of the Betic block onto their foreland (Cebriá et al., 2011;Granja et al., 2015). Despite significant thickening during the Variscan orogeny, the Calatrava area has a normal crust (around 32 km) but a thin lithosphere (around 85 km; Granja et al., 2015).
Volcanism in Calatrava took place in two different stages (Ancochea, 1982): (1) a minor ultrapotassic event around 8.7-6.4 Ma ago and (2) alkaline basalts, basanites, and olivine nephelinites and melilitites from 3.7 to 0.7 Ma. The marked isotopic (Sr-Nd-Pb) homogeneity of the primary magmas of the main volcanic event suggests that the Calatrava volcanic rocks are derived from partial melting of an enriched homogeneous asthenospheric source (Ancochea 1982;Cebriá and López-Ruiz, 1995). The isotopic data suggest a HIMU-FOZO-like reservoir (where HIMU is high μ [μ = 238 U/ 204 Pb], and FOZO is a focal zone that has an approximate isotopic composition of 87 Sr/ 86 Sr = 0.7025, ε Nd = +9, and 206 Pb/ 204 Pb = 19.5), similar to that defined for the European asthenospheric mantle (Wilson and Downes, 1991;Granet et al., 1995). An enriched asthenospheric reservoir is in agreement with the suggested presence of garnet and phlogopite within mantle sources based on the trace-element geochemistry of Calatrava volcanic rocks (e.g., Ancochea, 1982;López-Ruiz et al., 2002). A complex carbonatite-silicate association has recently been suggested from volcanic deposits in Calatrava (Bailey et al., 2005;Humphreys et al., 2010;Stoppa et al., 2012), but this is an open debate because some of the supposed mantle carbonates show isotopic evidence of sedimentary origins (Lustrino et al., 2016).

SAMPLES AND MINERALOGICAL FEATURES
Mantle xenoliths from the El Aprisco center (Universal Transverse Mercator Zone 30 S, 428483m E, 4299135m N), an olivine melilitite maar (Fig. 1), vary from lherzolite to wehrlite comprising four chemical groups on the basis of the degree and nature of the metasomatic agents (González-Jiménez et al., 2014). Group 1 and lherzolite 111658 of Group 2 are highly metasomatized peridotites affected by a carbonate-rich agent, whereas other peridotites show interaction with silica-undersaturated alkaline or subduction-related silicate melt/fluids (Groups 2, 3, and 4 of González-Jiménez et al., 2014). Lherzolites with the lowest incompatible trace element contents and having flat rare earth element (REE) patterns (Group 4) seem to be a slightly restitic peridotite source that is poorly metasomatized (González-Jiménez et al., 2014). The three xenoliths selected in this study for geochronological purposes are from Group 1 (two wehrlites: 111657 and the apatite-bearing 111659), and from Group 2 (lherzolite 111658 with accessory apatite).
Apatite only appears in two xenoliths: the 111659 wehrlite of the most metasomatized Group 1, and in the 111658 lherzolite of Group 2. It is always an accessory phase (< 0.3 modal %), being markedly rare in the Group 2 lherzolite, where it occurs as dispersed small grains (mostly < 100 µm) within mafic peridotite minerals that are in apparent microstructural equilibrium with the lherzolite mineral assemblage (Fig. 2D). In the wehrlite, apatite appears interstitially forming clustered polygonal dusty grains up to 1800 µm in diameter ( Fig. 2A and 2B). Apatite is unrelated to the metasomatic amphibole but locally coincides with the network of interconnected intergranular veins of zeolitized glass (with accessory calcite) that penetrate the wehrlite 111659 (Fig. 2B). Nevertheless, textural and compositional studies indicate that apatite did not contribute to the formation of interstitial glass in these xenoliths (González-Jiménez et al., 2014).
Amphibole in selected samples appears as a main mineral phase (> 5 modal%), with a medium to coarse grain size (González-Jiménez et al., 2014). It forms veins or small layers in these peridotites. Only in lherzolite from Group 2 does the amphibole shows a clear corroded aspect and spatial relation with interstitial glass veins and pockets (González-Jiménez et al., 2014). Glass in these xenoliths may have been formed by in situ melting reactions promoted by amphibole breakdown during volcanic entrainment (Villaseca et al., 2010) or, alternatively, by infiltrating mantle metasomatic melts reacting with previous amphibole shortly before the entrainment of the xenoliths in the host basalt (González-Jiménez et al., 2014). The interconnected network of altered glass (zeolitized) with occasional calcite that appears in Group 1 xenoliths is clearly another metasomatic stage unrelated to apatite or amphibole formation/breakdown.

In Situ Major and Trace Element Mineral Chemistry
New data on 50 major (electron microprobe) and 36 trace element (LA-ICP-MS) spot analyses have been added to the chemical data set of González-Jiménez et al. (2014) to yield a comprehensive body of geochemical information. Major-element compositions of silicates and apatite were analyzed in polished thick sections using a Jeol JXA-8900 M electron microprobe with four wavelength dispersive spectrometers at the Centro de Microscopía Electrónica "Luis Bru" (Universidad Complutense de Madrid, Spain). Analytical conditions included an accelerating voltage of 15 kV and an electron beam current of 20 nA, with a beam diameter of 5 μm. Elements were counted for 10 s on the peak and 5 s on each background position. Corrections were made using an online ZAF method. Detection limits are 0.02 wt% for Al, Na, K, and P; 0.03 wt% for Ti, Fe, Mn, Mg, Ni, and Cr; and 0.04 wt% for Si. The CO 2 contents of apatite were analyzed after Cr coating of polished samples and standards, and using an LDE2 crystal with counting times of 20 s on peak. Smithsonian dolomite (NMNH 10057) was used as the standard. On the basis of counting statistics, the carbon detection limit was less than 1200 ppm. We have checked this method by analyzing Durango apatite and, as expected, found no detectable carbon in it. We are therefore confident that our approach yields true carbon concentrations.
Trace elements (including REE, Ba, Rb, Sr, Ga, Th, U, Nb, Ta, Pb, Zr, Hf, Y, Sc, V, Co, and Ni) were determined in situ on clinopyroxene, amphibole, and apatite in polished thick sections, by laser ablation (LA-ICP-MS). This was done at the Geochemical Analysis Unit, Australian Research Council (ARC) Centre of Excellence for Core to Crust Fluid Systems (CCFS)/ARC National Key Centre for Geochemical Evolution and Metallogeny of Continents (GEMOC), using an Agilent 7500CS ICP-MS coupled to a New Wave UP 213 laser source. The counting time for one analysis was typically 180 s (60 s on gas blank to establish background and 120 s for signal collection). The diameter of the laser beam was around 40 µm, the frequency 5 Hz, and laser energy 8 J/cm 2 . Ablation was performed in a He carrier gas (0.9-1.2 l/min). After exiting the sample cell, He was combined with Ar (0.9-1.2 l/min) gas in a 30 cm 3 mixing chamber prior to entering the ICP as described by Jackson et al. (2004). The data were processed using the GLITTER software (Griffin et al., 2008). The NIST 610 glass standard was used to calibrate relative element sensitivities for the analyses of the silicate minerals. Each analysis was normalized to Ca using concentrations determined by electron microprobe. Detection limits for each element are in the range of 0.01-0.06 ppm except for Sc (0.11 ppm). Precision and accuracy (both better than 10%) were assessed from repeated analyses of the BCR-2 standard. Representative mineral chemical data are presented in Table 1.

Amphibole Ar-Ar Dating
Amphibole was concentrated by mineral separation methods. Five samples were selected for Ar-Ar dating, two metasomatic amphiboles from peridotite xenoliths, and three megacrysts (amphibole and phlogopite are ~6 and 4.5 cm in length, respectively) from host volcanic rocks. The small size of the 111659 wehrlite xenolith precluded separation of a sufficient quantity of amphibole from this sample. In contrast, the two amphibole megacrysts from Calatrava volcanic rocks were split in two subsets (Table 2). Samples were crushed, sieved to select the 400-500 µm fractions, and washed in deionized water. High purity grains were handpicked under a binocular microscope and ultrasonically cleaned using ethanol. Samples were placed in copper or aluminum foil packets, together with neutron flux monitors, and stacked in quartz tubes. The relative positions of packets were precisely measured (±0.5mm) for later reconstruction of neutron flux gradients. The sample package was irradiated in the Oregon State University reactor, Cd-shielded facility. Fish Canyon sanidine (28.294 ± 0.036 [1σ) Ma, Renne et al., 2011) was used to monitor 39 Ar production and establish neutron flux values (J) for the samples. The neutron flux within a given disc is calculated by least-squares fitting of a surface to the J-monitors. Estimated errors in the neutron flux measurements are calculated from the residual deviation from the fitted surface. Gas was extracted from samples via step-heating using a mid-infrared CO 2  Note: Preferred ages are marked in bold. Values in italics (inverse isochron) are based in the youngest population. 40 Ar/ 36 Ar(i)-non-radiogenic intercept value; P-probability that the residuals can be explained by measurement errors alone; n-number of accepted steps (out of the total number of steps run). laser, with samples housed in a doubly pumped ZnS-window laser cell. Individual sample grains were loaded into a copper planchette containing 1x1 cm square wells. Liberated argon was then purified of active gases (e.g., CO 2 , H 2 O, H 2 , N 2 , CH 4 ) using three Zr-Al getters, one at 16 °C and two at 400 °C. Data were collected on a GVi instrument ARGUS 5-collector mass spectrometer using a variable sensitivity faraday collector array in static collection (non-peak hopping) mode (Sparks et al., 2008;Mark et al., 2009), at the NERC Argon Isotope Facility of the SUERC. Time-intensity data are regressed to t 0 with second-order polynomial fits to the data. Mass discrimination was monitored by comparison to running-average values of an air standard. The average total system blank for laser extractions, measured between each sample run, was 1 × 10 −15 mol for 40 Ar, 3 × 10 −17 mol for 39 Ar, 3 × 10 −18 mol for 36 Ar. All data are blank, with interference and mass discrimination corrected using the Masspec software package (authored by Al Deino, Berkeley Geochronology Center).
Plateau criteria included the following: the plateau consists of at least three contiguous steps, and the scatter between the ages of the steps is low, i.e., mean square of weighted deviates (MSWD) close to 1, and the fraction of 39 Ar released for these steps is ≥50%. Isochrons are calculated using only the plateau steps to confirm the composition of the trapped component. Generally, a plateau age is accepted if it is concordant at the 2σ level with the isochron age, has a trapped component indistinguishable from air (298.56 ± 0.31, 1σ) at the 2σ level, and meets the other criteria listed above. However, we note that the data for metasomatic amphiboles appear to be a mixture between excess and radiogenic Ar. Therefore, uncertainties on the trapped component compositions derived from the inverse isochron regressions are large and highly sensitive to individual data points. It is for this reason that we assume atmospheric trapped components in the amphiboles for purposes of plateau calculation. Summary data are presented in Table 2.

In Situ Apatite U-Pb Dating
U-Pb dating of apatite is difficult due to its generally low U contents and incorporation of common lead during formation (e.g., Thomson et al., 2012). In addition, LA-ICP-MS U-Pb dating of apatite also presents laser-induced U-Th-Pb fractionation and yields high age uncertainties in young Cenozoic crystals (Chew et al., 2011). Nevertheless, the new wellcharacterized matrix-matched reference apatites and the usually high U contents of metasomatic mantle apatite combine to make in situ apatite U-Pb dating a promising chronometer of metasomatic events in peridotite xenoliths. Indeed, studied metasomatic apatite has > 38 ppm U (Table 1), making it suitable for U-Pb dating.
Thick (85 µm) polished sections were mounted close to a 2.5-cm diameter epoxy disk containing four apatite reference material (Kovdor, Slyudyanka, Mount McClure, and Emerald Lake) and two zircon standards (GJ and 91500) to control instrument stability. Analyses were performed in situ using a 213 nm Nd:YAG laser microprobe attached to an Agilent 7500 quadrupole ICP-MS system at the GEMOC laboratories. Samples were ablated using a 55 µm (40 µm in the small apatite from sample 111658) laser beam that was rastered over the sample surface to create a 200 µm line to minimize laser-induced Pb/U fractionation. The laser energy was set at 6.5J/cm 2 , with a 5Hz repetition rate. He carrier gas transports the ablated sample from the laser-ablation cell via a mixing chamber to the ICP-MS torch with the same gas flow conditions (0.9-1.2 L/min) as used for the ICPMS trace-element work.
Slyudyanka carbonatite apatite was used as a calibration standard due to its chemical homogeneity. Together with every six analyses of unknowns, well-characterized zircons (91500 and Mud Tank), and apatites have been analyzed as a quality control to monitor variability in operation conditions. The list of "check up" apatites includes Kovdor carbonatite (377.5 ± 3.5 Ma, Amelin and Zaitsev, 2002), Mt McClure syenite (523.5 ± 1.5 Ma, Schoene and Bowring, 2006), and Emerald Lake granite (92.2 ± 0.9 Ma, Coulson et al., 2002). Their analyzed weighted-mean 207 Pbcorrected ages agree within the published analytical results (367 ± 13, 538 ± 8 Ma, 99 ± 9 Ma, respectively) (GSA Data Repository Figure DR1 1 ), mostly falling within the uncertainty limits on the assumed ages of the reference material.
Isotopic ratios and uncertainties on individual analyses are presented with 2σ errors in Table 3. Tera-Wasserburg concordia and weighted average ages were calculated using Isoplot Ex 3.0 (Ludwig, 2003).

In Situ Apatite Sr-Nd Isotopic Ratios
Only large apatite crystals of wehrlite 111659 were amenable to in situ Sr-Nd isotopic analyses. Eleven analyses close to previous U-Pb spots were ablated during Sr and Nd isotopic measurements at the Geochemical Analysis Unit at CCFS/GEMOC with a Wave/Merchantek UP 213 laser microprobe with a modified ablation cell coupled with a Nu Plasma Multicollector ICP-MS. During the ablation runs, ion beams were collected using a mix of Faraday cups and ion counters. The laser was fired at a frequency of 5 Hz (90%-95% power and fluency of 5-10 J/cm −2 ) and a spot size of 65-100 µm (40-55 µm for some apatite spots). During ablation runs, the Sr and Nd analyses were monitored by repeated measurements of the Ice River perovskite standard (with 87 Sr/ 86 Sr = 0.70293 ± 2 and 143 Nd/ 144 Nd = 0.512615 ± 9, Yang et al., 2009) and the Batbjerg clinopyroxene standard ( 87 Sr/ 86 Sr = 0.70444 ± 9, Neumann, 2004). The IR perovskite analyses yielded 87 Sr/ 86 Sr = 0.70287 ± 6 (1SE) and 143 Nd/ 144 Nd = 0.512608 ± 24 (1SE), whereas the Batbjerg clinopyroxene gave 87 Sr/ 86 Sr = 0.70453 ± 5 (1 standard error [SE]), which are within the uncertainty of the accepted values.
Isotopic ratios and uncertainties on individual analyses are presented with 2σ errors in Table 4.

MINERAL CHEMISTRY
Most of the trace-element signatures of wehrlites from Group 1 of El Aprisco, such as their convex upward LREE enriched REE patterns combined with prominent positive Nb-Ta and small negative Hf and Ti anomalies ( fig. 4 of González-Jiménez et al., 2014), are inherited from the trace element composition of the abundant metasomatic amphibole (14-18 vol% in modal amount). Primary clinopyroxene in wehrlites does not present positive Nb-Ta anomalies in its multitrace element pattern and has lower LILE-Ti contents than associated amphibole (Table 1). Nevertheless, they share a similar REE pattern and small negative Pb-Sr-Hf-Ti anomalies in multi-trace element patterns suggesting an approach to chemical equilibrium shared between the two minerals, as departures from a D amp/cpx of 1 occur in elements (Nb, Ta, Ti, Ba), which usually show higher partition coefficients in peridotite amphibole (e.g., Chazot et al., 1996a;Raffone et al., 2009).

RESEARCH
As in wehrlites, the slightly high LILE, Nb, Ta, and Ti contents of the amphibole from lherzolite 111658 (Group 2) also indicate chemical equilibrium with respect to the composition of the associated clinopyroxene, and are evidence of the metasomatism undergone by the studied peridotite xenoliths (see also González-Jiménez et al., 2014).
Petrographical features of apatite in studied xenoliths are similar to the dispersed grains of apatite-A described by O'Reilly and Griffin (2000), i.e., dusty crystals not associated with modal amphibole. The relatively high MgO (up to 1.17 wt%), FeO (0.54 wt%) and Na 2 O (0.80 wt%) contents are also in agreement with values found in apatite-A of mantle xenoliths. Moreover, apatite of El Aprisco xenoliths enlarges the compositional field of metasomatic A-type apatite in peridotites showing higher halogen, REE, and U contents (Fig. 3). The high Sr (6700-11050 ppm), Th (140-950 ppm), U (38-220 ppm), Ce (2400-5600 ppm), and low Y (80-200 ppm) contents, combined with some aspects of their REE patterns: (1) almost no Eu/Eu* anomaly (0.87-0.91) and (2) high (La/Lu) N or (Ce/Yb) N ratios, resembling those of apatite from carbonatites (Belousova et al., 2002), especially the more enriched apatite from the 111658 lherzolite (Table 1). The most relevant difference of apatite from carbonatites is in the halogen contents expressed by the high Cl/F ratio of apatite from El Aprisco xenoliths (1-13), which is clearly at odds with that of carbonatite apatite (typically below 0.05, Chakhmouradian et al., 2017). Nevertheless, the halogen contents of apatite from El Aprisco xenoliths are common in metasomatic apatite-A in peridotites (Fig. 3) (O'Reilly and Griffin, 2000). Likewise, the carbonate content in studied apatites (0.8-2.0 wt%, Table 1) is in the range of chlorapatite-A (O'Reilly and Griffin, 2000).
The in situ Sr-Nd isotopic data obtained for apatite in the 111659 wehrlite yielded a signature similar to that of the Calatrava volcanic rocks, showing a high averaged εNd and a low 87 Sr/ 86 Sr ratio (FOZO or HIMU composition), although Nd isotopic heterogeneity is significant in apatite crystals (Fig. 4).

Ar-Ar Dating
Ar-Ar data on the mafic megacrysts from the basaltic lavas give ages in the range of other Calatrava volcanic rocks. Amphibole from the El Aprisco volcano (megacryst 114404, Table 2) does not yield a plateau age but a clustering of final steps at around 2.8 ± 0.1 Ma can be taken as the best estimate of the melilitite emplacement age (Data Repository Fig.  DR2). Two replicates of the amphibole 111653 megacryst from the Cerro Pelado scoria cone yielded plateau ages of 1.59 ± 0.07 and 1.80 ± 0.06 Ma, whereas the phlogopite 114403 megacryst from the same volcano (Table 2) yielded a good plateau age of 2.23 ± 0.04 Ma (Data Repository Fig. DR3). The phlogopite being slightly older than the amphibole seems to be of more geological significance due to the absence of excess argon in this sample.
Ar-Ar analyses were conducted for metasomatic amphiboles in two peridotite xenoliths from the El Aprisco maar (Table 2)  9 VILLASECA ET AL. | Dating metasomatic events in Calatrava RESEARCH plateau ages because they are affected by excess argon. Inverse isochron plots for these samples show trapped component compositions different from those of atmospheric argon. The amphibole of the lherzolite 111658 sample (with accessory apatite) yielded an integrated age of 10.2 ± 0.9 Ma, whereas amphibole from wehrlite 111657 yielded an integrated age of 21.2 ± 0.2 Ma. When only considering final steps (after release of most of the excess argon), ages of 3.9-15.7 Ma, respectively, are obtained (Data Repository Fig. DR2; Table 2). These poorly constrained ages, based on a very restricted number of steps, can be regarded merely as an indication of amphibole formation during metasomatic events before entrainment of the xenolith by the volcanic magma.

U-Pb Dating
Twenty-seven spot analyses were performed on apatite crystals (wehrlite sample 111659), and 24 of these analyses were used for age calculation employing the Tera-Wasserburg concordia approach, which failed to yield a reliable lower intercept age (33 ± 34 Ma; MSWD = 1.2).
However, a weighted average of 207 Pb-corrected ages approach, which used all 27 analyses, gave the age of 29.0 ± 6.3 Ma with a good MSWD value of 0.22 (Fig. 5). This age was considered to be the most reliable estimate for the timing of metasomatic events forming these apatites.
Only 12 analyses were possible for the lherzolite 111658 due to apatite scarcity and small grain sizes. The Tera-Wasserburg concordia lower intercept age of 121 ± 34 Ma (MSWD = 2.7) and weighted average 207 Pb-corrected age of 118.1 ± 9.4 Ma (MSWD = 0.29) (Fig. 5) are values suggesting an old Cretaceous metasomatic event, much older than the one that formed the associated amphibole (~3.9 Ma).
More information on the chemical characteristics of the metasomatic agents can be derived by calculating the composition of melts in equilibrium with metasomatic phases: clinopyroxene, amphibole, and apatite. We have used clinopyroxene and amphibole K D 's (partition coefficient for mineral-melt equilibria) of basaltic systems (Hart and Dunn, 1993;Zack and Brumm, 1998;La Tourrette et al., 1995;McKenzie and O'Nions, 1991). The calculated melts show high LREE, Th-U, and negative anomalies of Zr, Hf, and Ti characteristics of carbonate-rich magmas (Figs. 8 and 9). Melts in equilibrium with metasomatized clinopyroxene and amphibole from wehrlites (Group 1) and from apatite-poor lherzolite 111658 (Group 2) also yielded a significant Nb-Ta depletion, typical of carbonatites, and very different in normalized trace-element patterns from silicate melts from the CVF (Fig. 9).
Calculated melt composition in equilibrium with apatite using carbonatite systems (Dawson and Hinton, 2003;Hammouda et al., 2010) also yielded a trace element pattern similar to that from clinopyroxene. Thus, calculated melts show consistent features of LREE-enriched melts with marked troughs in Nb-Ta, Zr-Hf, and Ti values, common in carbonate-rich liquids (Fig. 8). When using apatite partition coefficients for basaltic systems (Paster et al., 1974;Prowatke and Klemme, 2006), calculated melts show smaller positive Th-U and negative Ta-Nb, Hf-Zr, and Ti anomalies, although they still look like carbonatite melts.
The isotopic Sr-Nd signature of the analyzed metasomatic apatite is similar to alkaline magmas produced in the Atlantic region and in the European circum-Mediterranean area (e.g., Wagner et al., 2003;Lustrino and Wilson, 2007) (Fig. 4). However, significant Nd-isotope heterogeneity in apatite grains (wehrlite 111659), which is up to 12 epsilon units (Table 4), occurring even at the grain scale, suggests an important chemical modification of the old metasomatic apatite. This Nd-isotope heterogeneity contrasts with the homogeneity of major and trace element compositions and the Sr-isotope signatures of apatite crystals. The lack of isotopic equilibrium between minerals of mantle xenoliths is not uncommon (e.g., Menzies et al., 1987;Raffone et al., 2009). The presence of a network of infiltrating (zeolitized) glass veins close to studied apatite crystals (Fig. 2B) suggests a significant percolative metasomatic agent affecting apatite in later stages. The vein network probably results from alkaline silica-undersaturated melt/fluids generated during Neogene magmatic events, as Nd isotopic heterogeneity in the studied apatite (from MORB to EM-I) encompasses values found in the Iberian Cenozoic volcanism, for example, Olot (NE Spain), and the ultrapotassic leucitites of the Calatrava volcanic field (Fig. 4). However, the question of why Nd-isotope ratios are modified in the absence of any other appreciable chemical modification in the apatite crystal awaits further investigation.

Correlation of Metasomatic Ages with Volcanic Events
In situ U-Pb dating of apatite, as reported herein, is obtained using the inherently high spatial resolution and high-precision LA-ICP-MS technique. Due to the limited number of apatite grains available in some samples, a larger data set would be desirable in future studies to better constrain ages obtained during this research. On the other hand, Ar-Ar geochronology on amphibole had its own challenges due to the presence of excess argon and the potential for open versus closed system behavior (Kelley and Wartho 2000), which resulted in complicated age data being obtained. Assuming closed system behavior for argon in the lithospheric source region, the maximum Ar ages reported herein indicate that

12
VILLASECA ET AL. | Dating metasomatic events in Calatrava RESEARCH metasomatic interactions in the studied samples cannot be attributed to a single event, given that ages recorded from apatite are much older than those obtained from the amphibole samples (Table 5). For these reasons, we believe our geochronological data to be sufficiently robust to initiate discussion of the correlation between metasomatic events undergone by Calatrava peridotites and coeval magmatism recorded in the Iberian microplate, as well as its implications for geodynamic models pertaining to the circum-Mediterranean province. The Calatrava volcanic field was formed during two separate stages of magmatism. The oldest one was a minor ultrapotassic event at around 8.7-6.4 Ma, whereas the youngest was the main Na-rich alkaline event occurring from 4.7 to 0.7 Ma (e.g., Ancochea, 1982). Our Ar-Ar geochronological data on mafic megacrysts from basaltic magmas give ages in the range of other Calatrava volcanic rocks: 2.8 ± 0.1 Ma for amphibole from the El Aprisco volcano, and 2.2 ± 0.1-1.8 ± 0.1 Ma for phlogopite and amphibole, respectively, from the nearby Cerro Pelado volcano (Table  2; Data Repository Fig. DR3). These ages are within the timeframe of 4.7 to around 0.7 Ma reported for similar alkaline volcanic rocks of the Calatrava field in previous studies (Ancochea, 1982;Herrero-Hernández et al., 2015). However, some previous K-Ar dating of mafic megacrysts of the olivine nephelinite from the Cerro Pelado volcano yielded ages of 4.6 ± 0.7 (amphibole) to 3.4 ± 0.4 Ma (phlogopite) (Ancochea, 1982), a slightly older range than the dating obtained in this study.
Most dated metasomatic minerals have yielded ages greater than the known period of volcanic activity in the Calatrava field (Table 5). Nevertheless, interstitial peridotite amphiboles have estimated maximum Ar-Ar ages of 3.9 (lherzolite Group 2) to 15.7 Ma (wehrlite Group 1), almost overlapping the time of alkaline eruptions in the Calatrava volcanism.
In contrast, U-Pb ages obtained for metasomatic apatites indicate the existence of much older metasomatic events in the mantle beneath central Spain. The measured ages range from Aptian to Rupelian, indicating a previously unknown occurrence of a Cretaceous to Oligocene alkaline magmatism in central Spain. These new geochronological data provide evidence that apatite-forming metasomatic events occurred significantly earlier than the age range of the Calatrava volcanism and also record older stages, as compared to the metasomatic events that produced amphibole in the same peridotite xenolith (Table 5). Therefore, at least three metasomatic events have been identified within the SCLM beneath central Spain in Cretaceous (118 Ma), Oligocene (29 Ma) and Miocene (16-4 Ma) times, but the vestiges of these events in the Iberian microplate are very scarce.

Significance of Magmatic (Metasomatic Imprint) Ages
During the Late Mesozoic to Cenozoic times, Iberia registered several cycles of sporadic intraplate magmatism of alkaline affinity (e.g., Mata et al., 2015).
A first cycle of basic mildly alkaline magmas at 148-140 Ma appears to have been associated with peripheral rift basins in central western Portugal, suggesting a passive-type margin scenario (Mata et al., 2015). Their trace-element compositions combined with Sr-Nd isotopic signatures, close to CHUR values, suggest derivation from a metasomatized lithospheric source (Mata et al., 2015). A unique alkaline basaltic outcrop at the Languedoc volcanic field (SE France) yielded 161 Ma (Dautria et al., 2010), but slightly older Middle Jurassic alkaline volcanism (~176-165 Ma) has been described in the Iberian Range, also considered to be related to passive margin settings (Lago et al., 2004;Cortés and Gómez, 2016).
The third anorogenic magmatic cycle in Iberia is defined as being from 11 Ma to the present time (Ancochea, 2004), although orogenic calc-alkaline magmatism is slightly older in SE Spain (up to 18.5 Ma, Duggen et al., 2004). The third episode includes the studied Calatrava volcanic field with magmatic activity from 7 to 1 Ma. This magmatism is well recorded in continental Western Europe and in the Mediterranean area, and is characterized by Na-rich alkaline magmas, with signatures typical of OIB and a significant HIMU-FOZO component (e.g., Lustrino and Wilson, 2007;Faccena et al., 2010).
When summarizing age data from specific microplates (such as the Iberian plate), magmatic activity seems to be marked by the occurrence of magmatic episodes separated by periods of quiescence (Fig. 11). In contrast, when considering all of Western Europe-North Africa and Central Atlantic seamounts and islands, anorogenic magmatism seems to be more continuous, although scarce (Fig. 10). For instance, the apparent absence of magmatism in Iberia for more than 60 Ma, i.e., from 76 Ma for Portuguese and Catalonian alkaline intrusions (Grange et al., 2010;Ubide, 2013;Ubide et al., 2014) to Neogene intraplate volcanism, may simply reflect incomplete sampling. Indeed, the age reported herein of 16-29 Ma for the carbonate-rich alkaline magmatism that metasomatized the SCLM below Calatrava (forming wehrlites) fills this temporal gap and is close to the age of carbonatite-silicate magmatism in northern Morocco, at Azrou-Timahdite (35 Ma, Raffone et al., 2009) and at Tamazert (44-35 Ma, Marks et al., 2009;Bouabdellah et al., 2010;Fig. 11). The Late Cretaceous age found in metasomatic apatite from the lherzolite 111658 sample not only implies that this sporadic second magmatic cycle occurs in many parts of the Iberian peninsula, having first been reported from central Spain, but also suggests that a significant part of deep mantle-derived magmatic activity leaves no trace in overlying crustal levels. Percolating melts and fluids within the lithospheric mantle have little chance to progress upwards and reach the upper crust or even the surface to form volcanic rocks. Indeed, it is suggested that less than 10% of the budget of a magmatic cycle yields volcanic deposits, in contrast to the dominant plutonic counterpart trapped at greater depths (e.g., Perfit and Davidson, 2000). Recently, estimates of plutonic-to-volcanic ratios have increased to values up to 65:1 in subduction environments with high magmatic fluxes (de Silva and Kay, 2018), indicating that most melts of magmatic plumbing systems never reach the surface.

Geodynamic Implications of Cretaceous-Cenozoic Magmatism in Central Spain
In Iberia, the breakup of Pangea is manifested mainly by dyke intrusions (e.g., Messejana-Plasencia dyke, Dunn et al., 1998;Cebriá et al., 2003) forming part of a tholeiitic magmatic cycle at 202-198 Ma, which is coeval with the beginning of the Central Atlantic opening as a part of the Central Magmatic Province (e.g., Marzoli et al., 1999). Subsequently, the Iberian microplate was subjected to a complex evolution mostly constrained by the relative motion of the surrounding European and African lithospheric plates and the progressive opening of the Atlantic Ocean.
During the Mesozoic, two alkaline magmatic cycles in the Iberian microplate generated minor volumes of mildly alkaline volcanic rocks at the Jurassic-Cretaceous transition into more abundant Late Cretaceous alkaline massifs (Mata et al., 2015) (Fig. 11). The geological setting for these magmatic events is described as an extensional regime linked to the opening of the North Atlantic Ocean evolving from rift-related small intracontinental basins toward major involvement of mantle thermal anomalies (plumes?) related to the motion of the Iberian plate. The opening of the Bay of Biscay and allied rotation of Iberia with respect to Europe led to the emplacement of more silica-undersaturated rocks toward the end of this cycle (e.g., Grange et al., 2010;Ubide et al., 2014).
Data from the three magmatic alkaline cycles in Iberia and adjacent areas suggest a progression over time of several features that are significant for the evolving geodynamical setting of the Western Europe: (i) magmatism becomes more silica-undersatured with time, (ii) there is a broad geographical evolution toward (south)-eastern magma emplacement, and (iii) magma flux from deeper mantle levels increases (from lithospheredominant sources during Mesozoic to major involvement of asthenospherederived magmas in the Cenozoic).
The Jurassic-Cretaceous Lusitanian volcanic rocks of the western Iberian margin are moderately alkaline to slightly subalkaline (tholeiitic) basic magmas of lithospheric origin (Mata et al., 2015). During the Late Cretaceous, the magmatism evolves toward more alkaline affinity either in the Pyrenees or in SW Portugal, showing in both areas an evolution to more silica-undersaturated magmas with time (lamprophyres of 79 Ma in the Catalonian Coastal Ranges; Monchique suite of 69 Ma, in SW Portugal). Markedly less alkaline rocks of Iberia have older Late Cretaceous ages: 105-85 Ma in the Basque-Cantabrian basin and the Pyrenees (Ubide et al., 2014), 88-76 Ma in SW Portugal (Grange et al., 2010), and 97-87 Ma in Galicia (Ancochea et al., 1992). The generation of alkaline magmas in the Iberian microplate with HIMU-FOZO signatures occurred during this ultra-alkaline and silica-undersaturated Campanian-Maastrichian sub-episode (Ubide, 2013;Grange et al., 2010). This mantle composition coupled with ultra-alkaline silica-undersaturated magmas is typical of the Cenozoic alkaline magmatism of Iberia in their volcanic fields: e.g., Calatrava (López Ruiz et al., 2002) and Olot (Cebriá et al., 2000). Notably, enriched lithospheric components such as CHUR (Tallante, Beccaluva et al., 2004) or EM1 (leucititic magmas of Calatrava, López Ruiz et al., 2002) are also found in coeval volcanic products from these areas.
A geographical shift toward the SE over time is possible to envisage not only in a broad comparison among the locations of the three alkaline magmatic cycles (Fig. 10) but also when considering its evolution in the different Iberian sectors. During the Late Cretaceous, the described geochemical evolution goes along with a displacement of the magmatism toward eastern (Pyrenees) and southeastern (Portugal) locations. Moreover, the easternmost alkaline volcanic fields in Iberia (Olot, Cofrentes, Tallante) start their activity during the Late Cenozoic (last 10 Ma, López Ruiz et al., 2002). Thus, a migration of magmatism from the Atlantic toward the Mediterranean Iberian margin is evident.
The involvement of the ubiquitous FOZO component in the Iberian OIB-type basalts from Late Cretaceous times introduces important questions of its appearance in intracontinental settings with variable lithospheric sections. First, there is the potential involvement of sub-lithospheric deep mantle sources (the so-called European asthenospheric reservoir, Downes et al., 2003), and second, the possibility of focused upwelling of mantle material by plume-like structures. The scarce, discontinuous, and low volume magmatism that characterizes these alkaline cycles in Iberia and related areas are arguments against plume magmatism (e.g., Mata et al., 2015). This magmatism may have been triggered by mantle flow instabilities from transform faults (as the Tore-Madeira ridge) or small hotspots aligned parallel to the original western limit of the continent-ocean transition zone, during Atlantic opening (as Canary or Cape Verde volcanic lineaments). The onset of the Africa-Eurasia collision, mostly from 45 Ma onwards (e.g., Carminati et al., 2012), produced a more complex scenario for the Iberian plate. Geophysical models of the western Mediterranean indicate a marked lithospheric removal or small-scale convective heterogeneities in the upper mantle related to complex oceanic subduction trajectories (e.g., Faccena et al., 2010;Thurner et al., 2014). Crustal subduction events have occurred during the opening of the Alboran basin (SE Spain, Carminati et al., 2012, and references therein) and others that operate within a mosaic of microplates in the Mediterranean region (e.g., Faccena and Becker, 2010), generating complex small-scale convection associated with retreating slabs (Faccena et al., 2010;Thurner et al., 2014). The delamination of continental margin lithosphere when oceanic crust is consumed by subduction and slab rollback could initiate lithosphere detachment and permit asthenosphere upwelling (e.g., Thurner et al., 2014). This mechanism, combined with accommodation along transform faults, could explain the generation of anorogenic magmas during the Neogene (Melchiorre et al., 2017). A similar scenario is also envisaged in central Iberia where the existence of large Variscan shear bands weakened the crust and facilitated the ascent of underplated material stacked below the Moho boundary during the generation of the CVF (Granja et al., 2015).

CONCLUSION
To our knowledge, this is the first study dating mantle metasomatic minerals using a combination of different methodologies (in situ U-Pb on apatite coupled with Ar-Ar on amphibole separates) on a peridotite suite from a Cenozoic volcanic center within the circum-Mediterranean province. The geochronological data obtained imply that metasomatic interactions in the studied samples cannot be attributed to a single event, but rather successive metasomatic events recording subsequent percolation of ultra-alkaline melts within the subcontinental lithospheric mantle of central Spain over the last 120 Ma. This sporadic percolation of carbonatite or carbonated silicate melts from late Cretaceous to Neogene times occurred in at least three successive stages (Fig. 11). Carbonatites of Eocene to Miocene ages are abundant in the Atlantic islands (Cape Verde, Canary) and NW Africa (e.g., Madeira et al., 2005;Muñoz et al., 2005;Bouabdellah et al., 2010), and also in Cenozoic volcanic fields of Western Europe (Rhine, French Massif Central, Italy) (e.g., Stoppa and Sciazza, 2013;Wang et al., 2014). Correlating the Cretaceous carbonatite signature found in the apatite crystals from peridotite xenolith 111658 of the El Aprisco maar (Table 5) is, however, more difficult as no carbonatitic magmatism of similar age has been recorded in the circum-Mediterranean realm except in northwestern Africa where some few intrusive complexes have been described (e.g., Montero et al., 2016, and references therein). This issue merits further investigation.
Ages recorded from apatite are much older than those from associated amphibole in the studied samples (Table 5)

RESEARCH
apatite are especially significant when this accessory mineral appears disseminated in the peridotite matrix and is not related to melt pockets or glass veins, as in the studied samples (Fig. 2), suggesting that apatite is not involved in partial melting processes (see also Chazot et al., 1996b). The resistance to melting of this mantle mineral can be used to study old metasomatic events and agents. In our present study, mineral chemistry results indicate carbonated ultra-alkaline signatures for the successive metasomatic imprints (Table 5). This research highlights the possibility of a great age span between metasomatic overprints in the same section of the mantle due to the preservation of relictic metasomatic minerals.