Geochemistry and origin of zircon in chlorite schists of the Ronda peridotites (Betic Cordilleras, southern Spain)

During the last years, most papers about the Ronda peridotites (southern Spain) have focused on geochemical, petrological, and structural topics on the peridotites, but issues tied to the serpentinization and rodingitization processes have passed almost unnoticed. The term rodingite, proposed by Bell et al. (1911), refers to calcium-enriched and SiO₂-undersatured metasomatic rocks rich in Ca-Al and Ca-Mg silicates (e.g., O’Hanley, 1996, and references therein). These rocks are interpreted as products of the metasomatism of silicate rocks in contact with or within serpentinized ultramafic rocks, as a result of the percolation of alkaline (pH ≈ 8–12) and Ca-rich solutions released during the serpentinization of the ultramafic rocks. Rodingites are commonly wrapped by chlorite shells, referred to as blackwalls or chlorite schists (Dubinska et al., 2004), formed from the alteration of the ultramafic rocks.

This work reports a detailed zircon geochemistry, petrographic, and mineral chemistry study of chlorite schists from the Ronda peridotites of the Betic Cordilleras, in southern Spain (Fig. 1A). Similar chlorite schists have been described in ultramafic massifs from the western Alps (Mével, 1984), the Bohemian Massif (Dubinska et al., 2004), the Armorican Massif (Paquette et al., 2017), the Oman ophiolite (Yoshitake et al., 2009), and in the serpentinite ridge that crops out along the passive continental margin of Iberia, at the Galicia Bank (Beslier et al., 1990; Schärer et al., 1995) and the Iberian Abyssal Plain (Cornen et al., 1999) of the Atlantic Ocean. In these regions, the chlorite schists have been interpreted as subproducts from gabbroic/doleritic (Schärer et al., 1995) or plagiogranitic protoliths (Cornen et al., 1999; Jöns et al., 2009) and as reactional blackwalls originating from rodingitization processes (Dubinska et al., 2004). In line with this last work, the Ronda chlorite schists have been also described as blackwall zones formed during the serpentinization of the ultramafic rocks by interactions between the ultramafic rocks and granite dikes (Esteban et al., 2007).

The Ronda chlorite schists are notable for their high zircon contents. In view of the zircon scarcity in orogenic peridotites, this could be a relevant feature for geochronological purposes. Indeed, regarding the Ronda massifs, no zircon has been reported in the peridotites, and only two zircon-bearing pyroxenite layers have been described until now (Sánchez-Rodríguez and Gebauer, 2000). Based on structural field data, the petrographic study of the chlorite schists, and the geochemistry of their zircon inclusions, this study sought to ascertain the origin of the zircon. To this end, we also aimed to constrain the temperature conditions prevailing during the zircon formation and evaluate possible sources of zirconium. Finally, we also discuss the possible influence of serpentinization of the Ronda peridotites and coeval rodingitization of the granite dikes in the development of the chlorite schists.

The Ronda peridotites are the largest massifs of orogenic lherzolites in the world. In decreasing size, the principal ultramafic bodies are the Sierra Bermeja, the Sierra Alpujata, and Carratraca massifs. They correspond to spinel and/or plagioclase peridotites and subordinate mafic layers (Obata, 1980). The mafic layers in the peridotites consist of strongly deformed websterites, orthopyroxenites, and clinopyroxenites. All three massifs form allochthonous nappes sandwiched into the middle continental crust of the metamorphic Internal zone of the Betic Cordilleras, in the western corner of the Alpine Mediterranean belt.

In the studied area, the peridotites display a foliation with a dominant NNE strike, parallel to the mafic layering, and are traversed by granite dikes (Figs. 1B and 2A) roughly perpendicular to the foliation (Cuevas et al., 2006). The peridotites can be weakly or strongly serpentinized around the dikes (Figs. 2A and 2B). These dikes have large textural and mineralogical variations of magmatic origin. These variations can be locally enhanced by the introduction of calcium-rich minerals, leading to amphibole formation (Fig. 2C) during later rodingitization events (Esteban et al., 2003). Many granite dikes show a vertical zonation, with minerals such as cordierite, biotite, and graphite concentrated (cordierite-bearing granite) at lower levels and tip zones enriched in quartz, oligoclase, and tourmaline (Fig. 2D). These dikes have been linked with a short-lived event of partial melting of the crustal rocks underlying the Ronda peridotites during its hot emplacement (Tubía, 1994; Tubía et al., 1997; Cuevas et al., 2006).

Metric-sized layers and pods of zircon-bearing chlorite schists, parallel to granite dikes (Figs. 1B and 2B) and perpendicular to the mafic layering, are also present in the western corner of the Sierra Bermeja massif (Fig. 1A). Their structural relationships preclude an origin from older mafic layers and link the chlorite schist to granite dikes (Esteban et al., 2007). An age of 19.2 ± 1.1 Ma and temperatures between 289 °C and 322 °C have been proposed for the formation of the chlorite schist (Esteban et al., 2007). The present study presents new data on the trace-element contents of the zircon crystals in the chlorite schists that help to constrain the origin of the unusually high zircon content of these rocks.

The textural identification was based on petrographical study of more than 100 thin sections from 12 samples of chlorite schists and partially serpentinized peridotites. In addition, mineral composition was analyzed with a Cameca SX100 electron microprobe at the Oviedo University (Spain). Measurements were taken at an accelerating voltage of 15 kV, 10 s counting time, and 15 nA current. Total iron is given as FeO. The obtained mineral compositions are provided in GSA Data Repository Tables DR1–DR6.¹

According to the petrographic features, three main stages of mineral growth can be established: (1) a prechloritization event of rodingitization, (2) the chloritization sensu stricto, and (3) a postchloritization stage of late rodingitization. The dominant mineral resulting after this sequential process is chlorite, which has a large variety of textures. In addition, there are noteworthy amounts of idiomorphic zircon in all the samples.

The minerals corresponding to initial stages of the rodingitization (prechloritization: zircon, monazite, a first generation of apatite [Ap₁], sphene, and amphibole) are scarce and poorly preserved due to the great extent of the subsequent chloritization. Ca-rich amphiboles replacing older mylonitic peridotite textures are occasionally preserved in partially chloritized peridotites (Fig. 3A). Fresh amphiboles with random orientations are idiomorphic and poikilitic and contain inclusions of zircon (Fig. 3B). Sphene, which occurs as xenomorphic and corroded crystals (Fig. 3C), is often partially transformed into prehnite and thomsonite. The presence of xenomorphic crystals of apatite (Fig. 3D, left) is a distinctive feature of the prechloritization stage. Apatite appears as stubby, poikilitic, and corroded crystals that display a highly visible cleavage enhanced by the preferred arrangement of tiny inclusions, including zircon (Fig. 3D). From an analytical point of view, the apatite can be classified as F-rich, with up to 3% F and 2.5% Cl contents (Data Repository Table DR2).

The analyzed amphibole of the rodingite (sample tb-732) is classified as Ca-rich amphibole (Leake et al., 1997, 2003) with end members corresponding to tremolite and magnesio-hornblende (Data Repository Table DR1). These compositions contrast with those of amphiboles from late amphibole veins (samples tb-759 and tb-757), which are of magnesio-hornblende and edenite composition (Data Repository Table DR1), or from nephrite jades formed by submicroscopic to fine-grained aggregates of tremolite in partially rodingitized granite dikes (Tubía et al., 2009).

The extensive chloritization event was described by Esteban et al. (2007), who identified ilmenite and two different types of textural chlorites: (1) large radial clusters of needle-like chlorite crystals of different sizes, and (2) mesh and bastite pseudomorphic textures of chlorite. All these textures point to the peridotites being the chlorite schists precursors. The finding of the Ca-rich amphiboles described above replacing previous mylonitic peridotites supports this hypothesis.

The postchloritization mineral association includes (Data Repository Tables DR2–DR6): (1) prehnite (Fig. 3E) and a second generation of apatite (Fig. 3D, right), both concentrated in late veins and amygdules; (2) anorthite, thomsonite, hydrogarnet, and clinochlore, which replace prehnite; and (3) epidote, sphene, and aerinite superimposed onto previous textures. Prehnite occurs in late veins and amygdules crosscutting or replacing older chlorite. Apatite appears along late veins and occurs as subhedral crystals and shows common triple points (Fig. 3D, right). Thomsonite is intergrown with anorthite and replaces prehnite (Fig. 3F) or sphene (Fig. 3G). Sphene forms idiomorphic crystals in optical continuity with ilmenite (Fig. 3H, left). Aerinite occurs as idiomorphic prismatic crystals superimposed on chlorite aggregates (Fig. 3H, right). Finally, epidote is found filling in late voids.

The analyzed zircon crystals (Fig. 4) were separated by conventional mineral separation techniques (crushing, sieving, magnetic separator, and 3.3 g/cm³ density methyl iodide) from three chlorite schists (tb-733, tb-734, and tb-735). Most zircon crystals were idiomorphic and displayed subequant habits with a mean axis ranging between 80 and 230 µm. Prismatic zircon crystals (∼200 µm long and 100 µm wide), while less common, were also present. The shape and internal structure of the zircon crystals were best recognized in backscattered electron (BSE) images (Fig. 4), obtained using a scanning electron microscope (JEOL 6400-JSM) at the University of the Basque Country (Spain). The images of single crystals revealed idiomorphic zircon crystals with sector zoning dominant with regard to weak oscillatory zoning (Fig. 4). No inherited cores were detected.

The chlorite schist zircons often contained inclusions with ovoid, hourglass (Figs. 5A–5C), droplet (Figs. 5D–5F), and irregular shapes. Tiny gas bubbles (Figs. 5G–5I) and decrepitation processes (Fig. 5E) were also identified. Their size ranged from a few to a few tens of micrometers. These inclusions were frequently formed by two contrasted glasses, identified by scanning electron microprobe (SEM) as Gl₁ and Gl₂ (Fig. 6), which included small idiomorphic white mica (Fig. 6), apatite, quartz, feldspar, rutile, phlogopite, and thorite (a common mineral in felsic igneous rocks and pegmatites). The qualitative energy-dispersive X-ray spectroscopy (EDX) spot analyses obtained by SEM of these glasses yielded mostly strong silicic and peraluminous ([SiO₂] > 70%; aluminum saturation index [ASI] > 1) compositions. All these features suggest that these inclusions can be interpreted as silicate melt inclusions (e.g., Thomas et al., 2003).

Cathodoluminescence images (Fig. 7) of zircons from four granite dike samples (5-R-SB, tb-15–17, tb-06–838, and tb-06–842) were also taken at the A.P. Karpinsky Russian Geological Research Institute (VSEGEI), Saint Petersburg (Russia), in order to compare the structural/morphological features of the chlorite schist zircons with those of the dikes. In contrast with the chlorite schist zircons, granite zircons showed inherited oscillatory zoning in cores and peripheral growth rims indicating multiple growth stages.

Trace and rare earth elements (REEs; Data Repository Table DR7) were analyzed by laser ablation–inductively coupled plasma–mass spectrometry (LA-ICP-MS) at the University of the Basque Country (IBERCRON, SGIker), using a 213 nm New Wave Nd:YAG laser with 3.65 J/cm² energy density at a repetition rate of 10 Hz and coupled to a Thermo iCap quadrupole ICP-MS. Acquisition time was 70 s (25 s of background and 45 s of analytical signal). Analytical spot size was 40 µm in diameter, and, in most cases, the zircon crystals were completely pierced through. External calibration was performed with the National Institute of Standards and Technology (NIST) SRM 612 standard (Jochum et al., 2011), and an internal standard was the stoichiometry-calculated zircon (Zr = 49.76%). The data reduction was carried out using Iolite 3.32 software (Paton et al., 2011; Paul et al., 2012) by the laboratory staff.

After the observation of backscattered images of the zircon crystals from samples tb-733, tb-734, and tb-735, more than 80 crystals were selected for LA-ICP-MS analyses. Measurements tried to avoid chemical heterogeneities resulting from the tiny melt inclusions observed in a few zircon crystals. Several analyses were rejected due to the presence of inclusions or short acquisition signal time. The selected 69 zircon data set that was ultimately selected is presented in Data Repository Table DR7.

REE abundances of the analyzed zircons were normalized to C1-chondrite of McDonough and Sun (1995) and plotted on logarithmic graphs (Figs. 8A–8C). All the chondrite patterns were quite similar, with strong depletion in light REE (LREE) content, a very steep positive slope from La to Lu [(Lu/La)_N > 5000], large Ce positive anomaly (Ce* = 11–278, mean: 110), and smaller negative Eu anomaly (Eu* = 0.24–0.35, mean of 0.31), suggesting a common origin for all the zircons. Regarding the LREE and heavy (H) REE patterns, they displayed steep slopes from La to Sm, (Sm/La)_N ratios ranging from 30 to 2247 (mean 907), and a slight slope increasing from Gd to Lu, with (Lu/Gd)_N values from 18 to 45. Measured Th/U ratios from the samples (Data Repository Table DR7) were quite similar and displayed an average value of 0.15 (Fig. 8D).

The crystallization temperature is a useful parameter to differentiate between hydrothermal and magmatic zircon, as the latter give apparently higher temperatures, usually over 600–650 °C. The application of the zircon-saturation geothermometer (e.g., Watson and Harrison, 1983; Boehnke et al., 2013), addressed mainly to the petrogenesis of peraluminous to alkaline granite rocks, could be considered as a powerful tool with which to decipher crystallization temperature. However, its applicability is linked to those cases when the bulk-rock composition is equal or quite similar to the parental melt. In the case of the chlorite schists, where the chlorite formation is linked to metasomatism between the serpentinized peridotites and granite protoliths, the composition obviously does not correspond to the parental magma, and, therefore, we avoided using the zircon-saturation geothermometer. Temperature was calculated according to the calibration of Watson et al. (2006), and the activities of SiO₂ and TiO₂ were considered to be 1 due to the presence of quartz and rutile inclusions within zircon. The Ti-in content in all the zircons analyzed was quite uniform. The measured values ranged from 3.21 to 5.32 ppm, with a mean value of 3.9 ppm. The application of the Watson et al. (2006) geothermometer yielded mean temperatures of 672 ± 6 °C, 659 ± 7 °C, and 658 ± 4 °C for samples tb-733, tb-734, and tb-735, respectively (Data Repository Table DR7). A pressure correction for the Ti-in-zircon thermometer also has to be taken into account, because such temperature values are estimated for a pressure of 10 kbar, but temperature drops 50 °C at pressure conditions of 1 kbar (Watson et al., 2006). Therefore, the aforementioned values have to be reduced under lower-pressure conditions. In order to correct for the pressure effect on the Ti-in geothermometer, we used a previous pressure calculation for a garnet- and cordierite-bearing granite dike, where a pressure of 4.2 kbar and a temperature of around 680 °C were proposed for the dike intrusion (Tubía et al., 1997). Therefore, assuming a linear dependence between temperature and pressure in the Ti-in zircon geothermometer, the temperature would decrease around 20 °C. Therefore, a temperature range between 630 °C and 655 °C can be considered as a good approximation for zircon crystallization. These temperatures are significantly higher than those (289–322 °C) reported previously for the chlorite formation (Esteban et al., 2007), but they are consistent with the thermal conditions proposed for the emplacement of the granite dikes in the Ronda peridotites (Esteban, 2005; Cuevas et al., 2006).

For decades, zircon has been considered to be a rare or absent phase in ultramafic rocks due to the low Zr content and Si activity of these rocks (e.g., Palme and O’Neill, 2003; Hermann et al., 2006; Zheng, 2012). However, zircon is being increasingly reported within mantle-derived xenoliths (Liati and Gebauer, 2002; Liati et al., 2004; Zheng et al., 2006a, 2006b; Li et al., 2016), mafic granulites (Zhang et al., 2018), and peridotites (Grieco et al., 2001; Katayama et al., 2003; Zhang et al., 2011; Zheng et al., 2014), or even within rodingites and their blackwalls (Dubinska et al., 2004; Li et al., 2010; Fukuyama et al., 2014; Zhang et al., 2014). With the exception of rodingites, zircon coexists in all other types of rocks with mineral assemblages equilibrated at high mantle temperatures, which is consistent with experimental results that point to the stability of zircon in dunite at 1400–1550 °C (Anfilogov et al., 2015). In these cases, the formation of the zircon is linked to melt-peridotite metasomatism due to the migration of fluids/melts during the upwelling of the asthenosphere or downgoing of the lithosphere. Therefore, the occurrence of zircon in these types of rocks provides evidence of crustal metasomatism/refertilization of the ultramafic rocks at high pressure-temperature (P-T) conditions. In contrast, in the rodingites, the formation of zircon is attributed to crustal metasomatism at lower P-T conditions.

So far, only three zircon-bearing ultramafic samples have been reported in the Ronda peridotite massifs. Sánchez-Rodríguez and Gebauer (2000) found zircon in a corundum-bearing pyroxenite and in a pyroxenite with graphite pseudomorphs after diamond, and Esteban et al. (2007) described zircon-rich chloritites associated with the serpentinization of the peridotites. Assuming pressures of 30 kbar in the diamond stability field, temperatures above 1400 °C have been suggested for the origin of the graphite-bearing pyroxenites, which were subsequently equilibrated at lower-temperature (1100–1200 °C) and lower-pressure (20–25 kbar) conditions (Davies et al., 1993). Also, following a nearly adiabatic path from 15 to 10 kbar, the corundum-bearing pyroxenite evolved from early metamorphic conditions at T ≥ 900 °C to T ∼ 800–900 °C (Morishita et al., 2001). By contrast, our results yielded much lower temperatures, between 630 °C and 655 °C for the crystallization of zircon in the chlorite schists (see above). In line with the previously cited bibliographic cases, this thermal gap reflects different origins of zircon from the Ronda ultramafic rocks: magmatic and/or metamorphic at mantle conditions for the pyroxenite layers (Sánchez-Rodríguez and Gebauer, 2000) and metasomatic for the chlorite schists (Esteban et al., 2007).

Since the first description of the so-called “recrystallization front” by van der Wal and Vissers (1993), the Ronda peridotites have become a good target to study those processes related to the refertilization of subcontinental mantle by the upwelling of asthenosphere (Tubía et al., 2004). This front was described as a thermal boundary (“thermal-induced recrystallization front”) or a partial narrow melting domain caused by the thinning and coeval upwelling of the asthenosphere, leading to annealing and refertilization of the former lithosphere. This process can be considered as one of the necessary factors in the development of zircon within ultramafic rocks (e.g., Griffin et al., 2004; Zheng et al., 2006a, 2006b). However, the refertilization of the lithosphere can be ruled out for the studied samples because (1) the main structures related to the emplacement of the Ronda peridotites over the crust postdate the development of the “recrystallization front” (van der Wal, 1993; van der Wal and Vissers, 1993), (2) the zircon-bearing chlorite schists are always structurally linked to granite dikes (Esteban et al., 2007; Tubía et al., 2009), as they are parallel, and (3) the dike intrusion and the emplacement of the Ronda peridotites into the crust occurred during the Aquitanian–Burdigalian (Priem et al., 1979; Monié et al., 1994; Sánchez-Rodríguez and Gebauer, 2000; Sosson et al., 1998; Esteban et al., 2011), which agrees with chlorite schist zircon formation ages (Esteban et al., 2007). Looking at other possibilities, Dubinska et al. (2004) proposed that in the Sudetic ophiolite (Poland), the presence of fluid inclusions and layer silicates of chemical composition close to Mg-chlorite inclusions within zircon are some of the main markers of the concomitant zircon and blackwall formation at low-temperature conditions (≈270–300 °C). However, such an interpretation is not valid here, since fluid inclusions trapped during zircon growth have not been observed. In addition, in the samples tested in this work, there is a systematic presence of zircon crystals with straight edges and idiomorphic shape that lack internal zoned patterns. These microstructural features indicate that the zircon crystals of the chlorite schists probably grew freely in a fluid/melt-rich environment. Although the morphology of the zircons alone must not be considered as a diagnostic feature to identify their origin (Schaltegger, 2007), the lack of dipyramidal zircon with {111} faces could suggest a magmatic origin for prismatic zircons with well-developed {100} prisms (Yang et al., 2014). The observation of melt inclusions (Figs. 5 and 6) also precludes low thermal hydrothermalism (250–400 °C) as the formation mechanism of the zircons.

According to the literature, magmatic and hydrothermal zircons can display distinct REE and trace-elements patterns that can be used as discrimination criteria. It is worth noting that hydrothermal zircons commonly display REE distribution patterns characterized by flat or slight slopes from La to Sm and steep ones from Sm to Lu, with small Ce anomalies and high REE contents, but in a few cases, they rather show “magmatic-like” REE patterns (Pettke et al., 2005; Pelleter et al., 2007; Guo et al., 2011; Fukuyama et al., 2014). In fact, the REE distribution in zircon, apart from their origin, can be a function of the crystallization sequence of accessory minerals (apatite, xenotime, monazite, allanite, ilmenite), the temporal changes in the magma composition, complex substitution mechanisms, or even accidental samples of REE-bearing micro-inclusions, among other things (e.g., Chapman et al., 2016; Zhong et al., 2018). The geochemistry of zircons from the Ronda chlorite schists has been tested against the composition of zircons from hydrothermal environments elsewhere (Hoskin, 2005; Zhu et al., 2016; Hu et al., 2017; Li et al., 2010; Pettke et al., 2005; Pelleter et al., 2007; Guo et al., 2011; Fukuyama et al., 2014), which have been classified according to their “magmatic” (Fig. 9A) or even “hydrothermal-like” (Fig. 9B) REE patterns. The studied zircon REE patterns plot within the “magmatic-like” patterns defined by the hydrothermal group of zircons of Figure 7B and support the observation that the REE patterns of the zircons alone are not indicative of their magmatic or hydrothermal origin. In this regard, the work of Pettke et al. (2005) in the Mole Granite could shed light on this apparent contradiction, as high-temperature hydrothermal zircons have similar REE patterns to late magmatic ones.

Bilogarithmic plots of U/Ce ratio versus Th concentration (e.g., Castiñeiras et al., 2011) or the Ta/Nb and HfO₂/Y₂O₃ ratios (Pettke et al., 2005) can help to decipher the nature of the zircons. In the first plot, there is a main group concentrated in the igneous field (Fig. 8E). The Ta/Nb and HfO₂/Y₂O₃ ratios are commonly used to discriminate the nature of early magmatic zircons from late magmatic or hydrothermal zircons, as early magmatic zircons show lower Ta/Nb and HfO₂/Y₂O₃ ratios (usually <1). In the studied zircons, the low mean value (≈1) of the Ta/Nb ratio (Data Repository Table DR7) prevents its use as a possible discrimination parameter, in contrast with the high HfO₂/Y₂O₃ values (from 1.45 to 7.86), which are consistent with a late magmatic origin of the zircon. Bilogarithmic plots of Hoskin (2005) also agree with zircons being of magmatic origin.

Figures 2A and 2B represent good field examples, showing the close structural relationships between the granite dike swarm and the layers of chlorite schists. It is worth nothing that field orientations of the chlorite schists are parallel to the granite dikes and perpendicular to the mafic layers, which precludes an origin from the older mafic layers (Esteban et al., 2007). Similar U-Pb sensitive high-resolution ion microprobe ages, within their error limits, have been published for granite dikes and chlorite schist zircons. The granite dike intrusion has yielded ages of 22 ± 4 (2σ) Ma (Priem et al., 1979), 18.8 ± 6 (2σ) Ma (Monié et al., 1994), 18.8 ± 4.9 (1σ) Ma (Sánchez-Rodríguez and Gebauer, 2000), 18.9 ± 0.8 (2σ) Ma (Sosson et al., 1998), and 22.6 ± 1.8 (2σ) and 21.5 ± 3.8 (2σ) Ma (Esteban et al., 2011), while the zircon in the chlorite schists (Esteban et al., 2007) has been dated at 21.8 ± 0.5 Ma (95% confidence limit). Such similarity therefore supports chlorite schist zircons being related to granite dikes and growing more or less freely in a melt-rich environment.

The presence of zircon as inclusions in amphibole (Fig. 3B) and the transformation of amphibole to chlorite suggest that the crystallization of zircon preceded the pervasive growth of chlorite. Chlorite formation has been dated at 19.2 ± 1.1 Ma by means of zircon fission tracks (Esteban et al., 2007). However, such a time sequence does not enable us to assess the source of the high-temperature, zirconium- (Zr) rich percolating melts that promoted the crystallization of zircon.

We consider three alternative origins for these zircons: early igneous, late igneous, and hydrothermal. According to the obtained data, identified silicate melt inclusions contravene the hydrothermal origin for the zircons. An early igneous origin directly coupled with the intrusion of the granite dikes could be considered from the structural relationships between the dikes and the chlorite schists and their similar ages and orientations. However, the striking microstructural differences between the zircons extracted from the granite dikes (Fig. 7) and the chlorite schists (Fig. 4) also rule out this interpretation. From their low temperatures, <800 °C (Dickey and Obata, 1974; Cuevas et al., 2006), the Ronda granite dikes should be included in the cold granite field of Miller et al. (2003), which means low solubility of zircon and a broad preservation of inherited zircon xenocrysts. The granite dikes meet these conditions, as they frequently contain zircon crystals with tiny idiomorphic rims surrounding large xenomorphic cores (Fig. 7). Instead, with regard to the zircon from the chlorite schists, their idiomorphic shape and lack of inherited cores indicate that they cannot be interpreted as xenocrysts dragged from the neighboring granite dikes. Therefore, an alternative origin should be taken into account for these zircons, as their chemical features (common magmatic REE pattern), crystallization temperature, melt inclusions, and lack of fluid inclusions suggest that they could have come from late igneous melts sourced from the dikes and enriched in Zr. At first sight, this interpretation would seem to be in conflict with the aforementioned issue of the low solubility of zircon in the dikes. Nevertheless, this apparent contradiction is solved when the marked mineralogical variability of the dikes is taken into account. In this regard, a factor of particular relevance is the existence of apatite and even of fluorite in some granite dikes (Priem et al., 1979), since the F contents play a significant role in the solubility of zircon in igneous melts. According to the experimental work of Keppler (1993) on haplogranitic melts at temperatures of 800 °C and pressures of 2 kbar, the presence of F favors the enrichment of Zr in residual melts, because the solubility of zircon increases by 1–2 orders of magnitude when 2 wt% F is present in the melt. The frequent observation of granite dikes showing highly differentiated apices (Fig. 2D) and the finding of fluorapatite coexisting with zircon in the chlorite schists (Fig. 3D) are consistent with the existence of late igneous melts enriched in F and Zr that percolated from the granite dikes to the host ultramafic rocks. This interpretation is concordant with previous results that ascribed the influx of boron and halogens into Ronda serpentinites to fluids linked to the granite dikes (Pereira et al., 2003).

Zirconium has been considered as an immobile element during the last decades; however, its mobility is already proven in magmatic and hydrothermal systems, even in very low-grade metasedimentary rocks and in soils at different pH conditions (Rubin et al., 1989, 1993; Kerrich and King, 1993; Dempster et al., 2004; Fraser et al., 2004; Rasmussen, 2005). It is widely accepted that acidic fluid activity (Ayers and Watson, 1991; Schmidt et al., 2006) and the presence of F and Cl (e.g., Gieré, 1986; Rubin et al., 1989, 1993) can be some of the main factors responsible for Zr transfer in magmatic systems, since late and evolved magmatic fluids generally become more acidic. The presence of F- and Cl-rich apatites (see “Chlorite schists” section) and Ca-rich amphiboles in the chlorite schists points to the interaction between Ca-rich fluids, released during the beginning of the alteration of the peridotites, and highly evolved F-Cl–rich melts leading to Zr mobilization. In the studied area, such highly evolved melts frequently crystallize in the apical zone of the zoned dikes as tourmaline-rich leucogranite (Fig. 2D) and react with partially serpentinized peridotites, leading to the formation of rodingites. We propose that a late pH increase in the system, induced by the interaction of those fluids (8 ≤ pH ≤ 12) liberated during the first stages of serpentinization (O’Hanley, 1996) with highly evolved F-Cl–rich fluids, could promote (1) the crystallization of new idiomorphic zircon and its concentration toward the apical zone of the dikes and (2) the subsequent rodingitization of dikes and peridotites, leading to amphibole formation (Fig. 2C). At high pH conditions, Ti has been considered as a mobile element in the rodingites of the Bohemian massif (Dubinska, 1997; Dubinska et al., 2004). This interpretation can be extrapolated to the Ronda peridotite chlorite schists, because the ilmenite is one of the main phases of the chlorite schists, and a temperature around 300–350 °C for the formation of the chlorite (Esteban et al., 2007) is coherent with serpentinization temperatures (O’Hanley, 1996). In addition, the age of chloritization (19.2 ± 1.1 Ma) by means of zircon fission tracks (Esteban et al., 2007) postdates the dike intrusion (22.6–21.5 Ma). Therefore, we propose that the formation of zircon and amphibole arose from the interplay between high-temperature melts saturated in F, Cl, and Zr issuing from the granite dikes and Ca-rich fluids released during the beginning of the serpentinization. Finally, a late hydrothermal fluid, highly alkaline and rich in Ca(Mg), incoming from the widespread serpentinization, would have promoted the subsequent chloritization.

Neither the origin of the zircon in the chlorite schists, nor its temperature has been clarified previously (Esteban et al., 2007). Two options were hypothesized by these authors for their formation: (1) late high-temperature metamorphic fluids or (2) late high-temperature igneous melts, in both cases, emanating from the crustally derived granite dikes. However, the presented REE patterns and the identification for the first time of silicate melt inclusions in zircons agree with zircons being of late magmatic origin. This origin agrees again with the temperatures obtained for zircon crystallization (630–655 °C), and it fits with the expected temperatures (lower than 720 °C) for the intrusion of the granite dikes of the Ronda peridotites (Cuevas et al., 2006). The identification of such (1) high-temperature melts, saturated in F and Cl, which are usually characterized as efficient melts for Zr transportation under low-pH conditions, and (2) late hydrothermal alkaline fluids incoming from the serpentinization, which locally could increase the pH of the high-temperature melts, could lead to zircon precipitation.

Therefore, according to the presented data, the formation of the chlorite schists in the Ronda peridotites can be divided into the following stages (Fig. 10):

• (1) intrusion of the granite dikes within partially serpentinized peridotites at temperatures of ∼720 °C;

• (2) magmatic differentiation leading to highly differentiated melts saturated in Zr, Cl, and F and the beginning of zircon/F-rich apatite (Ap₁) crystallization at temperatures around 630–655 °C as the rodingitization and serpentinization progressed (Fig. 10A) due to the basification of the system;

• (3) advance of the rodingitization, leading to Ca-rich mineral precipitation (sphene and tremolitic-hornblende; Fig. 10B) from serpentinized peridotites and granite dikes;

• (4) beginning of chloritization (289–322 °C), leading to chlorite and ilmenite formation from previously rodingitized and serpentinized peridotites (Fig. 10C); and

• (5) overprint of late rodingitization minerals as prehnite, thomsonite, anorthite, clinochlore, hydrogarnet, epidote, sphene, apatite (Ap₂), and aerinite (Fig. 10D).

• (1) The presence of silicate melt-inclusions confirms the igneous nature of chlorite schist zircons in the Ronda peridotites.

• (2) Ti-in-zircon geothermometry yields temperatures of 630–655 °C for zircon crystallization.

• (3) The identification F- and Cl-rich apatite agrees with the presence of efficient melts for Zr transportation under low-pH conditions, whereas their late interaction with hydrothermal alkaline fluids, incoming from the serpentinization, led to zircon precipitation.

• (4) The serpentinization and concomitant rodingitization can be considered as determinant factors for zircon concentration in the chlorite schists derived from the ultrabasic rocks.

We acknowledge the work of three reviewers and are particularly grateful to two of them because their thorough and constructive comments greatly contributed to the improvement of the work. Special thanks go to Francisco Velasco Roldán and Jose Ignacio Gil Ibarguchi for their comments, discussions, and ideas on the genesis of zircon that improved a previous version of the manuscript. The editorial work of Damian Nance is also appreciated. This work was supported by grants GIU17/033 of the “Grupos de Investigación” of the University of the Basque Country and CGL2017–82976 (Ministerio de Economía, Industria y Competitividad/Agencia Estatal de Investigación/Fondo Europeo de Desarrollo Regional, European Union).