Major elements and compatible trace elements

Selected major element data are shown in SiO₂-variation diagrams in Fig. 4. The Malaga dykes display a limited compositional variation and cluster on most diagrams. They are moderately magnesian (MgO = 4·1-7·4) with 51·6-56·7% SiO₂ and have mg-numbers ranging from 46 to 64 [where mg-number = molecular Mg/(Mg + Fe_total)]. K₂O contents are <1% (Fig. 4d) and they record a tholeiitic crystallization history that involved moderate Fe enrichment (Torres-Roldán et al., 1986). Concentrations of compatible trace elements, such as Cr, Ni and Sc, are rather low, consistent with their generally evolved nature.

Figure 4. Major element variation diagrams highlighting the distinction between the lamproites and the Malaga dykes and calc-alkaline volcanic rocks which form a broadly contiguous array (see text for discussion). The data from this study (shown as filled symbols) are supplemented with published data (shown as open symbols) from the following sources: Malaga dykes (Torres-Roldán et al., 1986); calc-alkaline volcanics (Di Battistini et al., 1987); cordierite dacites (Munksgaard, 1984) taken as representative of crustal melts; Tallante alkali basalts (Ancochea & Nixon, 1987); lamproites (Nixon et al., 1984; Venturelli et al., 1984; Nelson et al., 1986).

The calc-alkaline volcanic rocks from the Cabo de Gata and Mazarrón-Cartagena regions form a broadly contiguous array in most major and trace element variation diagrams, extending from compositions similar to the more evolved Malaga dykes up to highly evolved rocks with SiO₂ contents >70% (Fig. 4). This is accompanied by systematic, broadly linear decreases in MgO, Al₂O₃, Fe₂O₃, TiO₂, CaO and Na₂O and increases in K₂O. On the K₂O-SiO₂ diagram (Fig. 4d), these rocks largely fall within the calc-alkaline to high-K calc-alkaline field. Compatible element concentrations are low to very low in all of these rocks. The cordierite-garnet dacites, taken as representative of local crustal melts, fall in the midst of the calc-alkaline volcanic array, having ~63% SiO₂ and 3·5% K₂O.

The Tallante alkali basalts are relatively evolved products with MgO ~4-5%, mg-number = 45-50 and low Cr and Ni contents. They have rather low SiO₂ and moderate K₂O ( Fig. 4). The hornblende-clinopyroxene cumulate has a complementary composition, with high MgO, Cr and Ni accompanied by low Al₂O₃ and K₂O. The spinel lherzolite xenolith compositions are dominated by their high modal olivine content (~80%, Ancochea & Nixon, 1987) and have high MgO and mg-number = 90, and low TiO₂ and CaO (Table 3).

Finally, the lamproites have high MgO, Ni and Cr with mg-numbers of 74-83, suggesting that they crystallized from primary mantle melts. They have relatively high SiO₂ but distinctly low Al₂O₃, Fe₂O₃ (total iron) and Na₂O for a given MgO or SiO₂ (Tables 2 and A1, Fig. 4). K₂O contents are high, and place the lamproites in the shoshonitic field in Fig. 3d. Thus, they are distinct from the other Neogene magmatic rocks in the Betics (Fig. 3), having extreme compositions, even for lamproites (e.g. Venturelli et al., 1984).

Incompatible trace elements

The general incompatible trace element characteristics are illustrated in primitive mantle normalized multi-element diagrams in Fig. 5 and in chondrite-normalized REE diagrams in Fig. 6. The Malaga dykes have low concentrations of most moderately incompatible trace elements and relatively flat REE patterns (Fig. 6a), similar to those of mid-ocean ridge basalts (MORB). However, they differ significantly from MORB in their enrichment of Rb, U, K, Pb and the light REE (LREE), and by having small negative Ti, Ta-Nb and Zr-Hf anomalies (Fig. 5a). Th and U show some scatter. They are characterized by sub-chondritic Nb/Ta ratios of 11-14, and one sample (M9) shows a small negative Eu anomaly (Fig. 6a).

Figure 5. Incompatible trace element diagrams for the four different rock suites in the Betic-Alboran Domain (see text for discussion). Normalized to primitive mantle values from Sun & McDonough, (1989).

Figure 6. Chondrite-normalized REE diagrams. The flat patterns for the Malaga dykes and the development of a negative Eu anomaly in the calc-alkaline volcanics should be noted. The lamproites are characterized by sigmoidal patterns with negative Eu anomalies. The Tallante alkali basalt lava is LREE enriched whereas the spinel lherzolite xenoliths have mildly LREE-depleted patterns.

The calc-alkaline volcanic rocks generally show a much greater degree of enrichment of the more incompatible trace elements, such as Rb, Th and U, and are markedly LREE enriched (Figs 5b and 6b). They have prominent negative Ba, Ta-Nb and Ti anomalies and a large positive Pb anomaly (Fig. 5b). In detail, the Cabo de Gata REE patterns are generally parallel to each other and strongly LREE enriched. However, they show relatively flat patterns from Dy to Lu and show variable development of a negative Eu anomaly (Fig. 6b). The REE pattern for sample B302 from Mazarrón is enriched in HREE relative to the Cabo de Gata volcanic rocks, whereas sample B309 shows a pattern more like the cordierite-garnet dacite (B321) than the other calc-alkaline volcanic rocks. Like the Malaga dykes, these rocks are also characterized by sub-chondritic Nb/Ta ratios of 9-13, excepting B302 (Nb/Ta = 16). The cordierite-garnet dacite has an incompatible trace element pattern which is broadly parallel with the calc-alkaline rocks in Fig. 5b but with higher overall concentrations; exceptions are formed by negative anomalies for Sr, P and Zr-Hf (see Fig. 5b). Its REE pattern is significantly more LREE enriched than the calc-alkaline volcanic rocks (excepting B309) but with similar heavy REE (HREE) concentrations (Fig. 6b).

In contrast to the other three suites of rocks, the Tallante alkali basalts have reasonably smooth, convex-up incompatible trace element patterns, typical of alkali basalts (Fig. 5c). Departures from the smooth patterns involve a negative K anomaly and a positive anomaly at Pb. The REE patterns are LREE enriched with relatively flat HREE and no Eu anomaly (Fig. 6c). The hornblende-clinopyroxene cumulate has an incompatible element pattern which broadly mirrors that of its host lava, albeit at lower concentrations. Its REE pattern is also similar to that of the host lava but has higher middle REE (MREE) concentrations. The spinel lherzolite xenoliths have very low incompatible trace element concentrations and relatively flat patterns except for negative anomalies at Ba and K (Fig. 5c). Their REE patterns are essentially flat with rather high concentrations (1-2 times chondrite) and, again, no Eu anomaly.

The overall patterns of the lamproites are broadly similar to those of the calc-alkaline rocks in terms of having negative Ba, Ta-Nb and Ti anomalies. However, they are distinguished by far more extreme enrichment of the more incompatible elements and a strikingly large positive Pb anomaly (Fig. 5d). Additionally, the lamproites possess negative Sr and P anomalies, and K is significantly depleted relative to Th and U despite the high overall K abundances. The REE patterns for the lamproites are distinct from those of all the other Neogene Betic volcanic rocks, and the LREE abundances reach up to 400-500 times chondrite. These patterns are strongly sigmoidal, with flat LREE and HREE, steep MREE, and have negative Eu anomalies (Fig. 6d).

Radiogenic isotopes

The four rock suites are well distinguished by their ¹⁴³Nd/¹⁴⁴Nd_i-⁸⁷Sr/⁸⁶Sr_i systematics (Fig. 7a). The Malaga dykes have relatively restricted ¹⁴³Nd/¹⁴⁴Nd_i (0·5129-0·5128); however, their ⁸⁷Sr/⁸⁶Sr_i ratios range from 0·705 to 0·709 and the highest ⁸⁷Sr/⁸⁶Sr_i ratios may reflect alteration, as these occur in samples M9 and M2, which also have elevated loss on ignition (Table 1). In contrast, the calc-alkaline rocks are displaced to much lower ¹⁴³Nd/¹⁴⁴Nd_i and higher ⁸⁷Sr/⁸⁶Sr_i. The cordierite-garnet dacite, taken as representative of local crustal rocks, has the highest ⁸⁷Sr/⁸⁶Sr_i ratio. The Tallante alkali basalt and cumulate have ¹⁴³Nd/¹⁴⁴Nd_i ~ 0·5126 and ⁸⁷Sr/⁸⁶Sr_i = 0·705-0·706. The lamproites have extreme isotope ratios with significantly lower ¹⁴³Nd/¹⁴⁴Nd_i and higher ⁸⁷Sr/⁸⁶Sr_i than either the crustal melt or the calc-alkaline rocks (e.g. Nelson et al., 1986).

Figure 7. Nd-Sr and Pb-Pb isotope diagrams (see text for discussion). Symbols and data sources as in Fig. 4; northern hemisphere reference line (NHRL) from Hart, (1984).

The Betic dykes and volcanic rocks have elevated ²⁰⁷Pb/²⁰⁴Pb and ²⁰⁸Pb/²⁰⁴Pb at a given ²⁰⁶Pb/²⁰⁴Pb ratio, and all lie well above the northern hemisphere reference line (NHRL) of Hart, (1984), with the exception of the Tallante lava, which lies close to the NHRL. In the ²⁰⁷Pb/²⁰⁴Pb-²⁰⁶Pb/²⁰⁴Pb diagram (Fig. 7b), the Malaga dykes form a linear array roughly parallel to, and above, the NHRL. The lamproites crudely form an extension of this array to higher ²⁰⁶Pb/²⁰⁴Pb. The calc-alkaline volcanic rocks lie below this array, with lower ²⁰⁷Pb/²⁰⁴Pb and a range of ²⁰⁶Pb/²⁰⁴Pb ratios that encompasses the range shown by the lamproites. The cordierite-garnet dacite lies at the low ²⁰⁶Pb/²⁰⁴Pb end of this array and thus at lower ²⁰⁷Pb/²⁰⁴Pb than the Malaga dykes and lamproites. Finally, and in contrast to the other rocks, the Tallante lava has much lower ²⁰⁷Pb/²⁰⁴Pb and lies on the NHRL at the high ²⁰⁶Pb/²⁰⁴Pb end of the spectrum shown by the Betic rocks. Interestingly, the Tallante cumulate has higher ²⁰⁷Pb/²⁰⁴Pb and lower ²⁰⁶Pb/²⁰⁴Pb, and lies near the field of the calc-alkaline volcanic rocks. Similar relationships are observed in the ²⁰⁸Pb/²⁰⁴Pb-²⁰⁶Pb/²⁰⁴Pb diagram (Fig. 7c).

Malaga dykes

The Malaga dykes are mantle-derived tholeiites which record a crystallization history that involved moderate Fe enrichment (Torres-Roldán et al., 1986). Although crustal contamination has clearly occurred (see below), the relatively low Fe and high Si in the least contaminated, highest-MgO dykes suggest that partial melting took place at relatively shallow levels (e.g. Klein & Langmuir, 1987). Consistent with this, they also possess relatively flat REE patterns and have HREE concentrations that are 10-15 times chondritic values, indicating that partial melting did not take place in the presence of residual garnet. Figure 8a is a plot of Tb/Yb-La/Yb in which curves for partial melting of a primitive mantle composition within shallow spinel facies and deeper garnet facies peridotite are shown. The Malaga dykes have not undergone significant REE fractionation and lie close to the primitive mantle composition, despite the majority of them being relatively evolved diorites. Figure 8b is a plot of Nb/Zr vs Nb, which should discriminate between the effects of partial melting and fractionation, assuming that both elements are incompatible with respect to the source mineralogy. In this diagram, the Malaga dykes largely fall around a mantle partial melting curve indicating degrees of partial melting generally >5% and more typically around 10-15%, which is consistent with their tholeiitic character and REE inversion modelling (D. P. McKenzie, personal communication, 1998). Given these degrees of partial melting, their low Tb/Yb indicates that partial melting took place in the spinel facies (Fig. 8a) and thus at depths shallower than 60-70 km. The sub-chondritic Nb/Ta ratios may indicate that the mantle source had undergone an episode of prior melt depletion (Plank & White, 1995).

Figure 8. Partial melting models for the Betic-Alboran Domain rocks. (a) Tb/Yb vs La/Yb diagram with partial melting curves (percent melting indicated) for both spinel and garnet facies upper mantle. The shaded region encompasses the field of likely upper-mantle melts. The Malaga dykes and less crustally contaminated calc-alkaline volcanic rocks lie within this field close to the spinel facies curve, consistent with their derivation by partial melting of the asthenosphere at shallow levels (<70 km). In contrast, the lamproites and alkali basalts have much higher La/Yb, requiring an LREE-enriched mantle source inferred to lie within the sub-continental mantle lithosphere. (b) Nb/Zr vs Nb diagram to separate the effects of partial melting and fractionation-assimilation. The Malaga dykes and calc-alkaline volcanic rocks broadly straddle an asthenospheric partial melting vector (amounts of partial melting indicated) but lie increasingly to the left of this line at higher Nb/Zr, consistent with the combined effects of fractionation and crustal assimilation. As in (a), the lamproites fall well off the melting curve and require an enriched mantle source. Modelling assumed batch partial melting of a MORB mantle peridotite source with distribution coefficients taken from the compilation by Halliday et al., (1995). Symbols and data sources as in Fig. 4.

The Malaga dykes have small negative Ta-Nb, Ti and Zr-Hf anomalies, and positive Rb, U, K and Pb anomalies (Fig. 5a). Additionally, their ⁸⁷Sr/⁸⁶Sr_i and ¹⁴³Nd/¹⁴⁴Nd_i isotope ratios are higher and lower, respectively, than those of the contemporary depleted mantle (⁸⁷Sr/⁸⁶Sr ~0·7029 and ¹⁴³Nd/¹⁴⁴Nd ~0·5132) and they are displaced well above the NHRL with elevated ²⁰⁸Pb/²⁰⁴Pb and ²⁰⁷Pb/²⁰⁴Pb. Thus, these rocks clearly do not represent unmodified melts from the depleted upper mantle. Elevated ⁸⁷Sr/⁸⁶Sr_i and negative Ta-Nb anomalies suggest mixing with a contaminant bearing a crustal signature. In a plot of Zr/Nb vs Ti/Nb (Fig. 9a) the Malaga dykes form a scattered array between MORB-like compositions and those of the lamproites and crustal melts. Consistent with the inferences from the isotopic data, this suggests that the dykes could have been affected either by crustal contamination or by mixing with lamproitic partial melts from the lithospheric mantle as the magmas rose towards the surface (e.g. Ellam & Cox, 1991). The lamproites are characterized by high K₂O and have low Al₂O₃ compared with the crustal melts (Fig. 4), and so K₂O/Al₂O₃ should readily distinguish between these two possibilities. Figure 9b shows that the Malaga dykes form the low-SiO₂ end of a tight array with the calc-alkaline volcanic rocks which ends in the field of the crustal melts. By contrast, the majority of the lamproites have much higher K₂O/Al₂O₃ and lower SiO₂. It therefore seems clear that crustal, rather than lithospheric mantle, contamination is responsible for the trend towards elevated ⁸⁷Sr/⁸⁶Sr_i. Although there are insufficient data for a rigorous analysis, Fig. 10 shows that the compositional array of the Malaga dykes can be approximated by calculated MORB-crust mixing curves. Approximately 5-15% crustal contamination is indicated, though, as discussed above, the higher ⁸⁷Sr/⁸⁶Sr_i in samples M2 and M9 may reflect alteration. It is also likely that crustal contamination was coupled with increasing degrees of fractional crystallization (e.g. DePaolo, 1981), as SiO₂ increases and Eu/Eu* decreases with increasing ⁸⁷Sr/⁸⁶Sr_i (Fig. 10a and b). Notwithstanding the limitations of the data, the critical observation from Fig. 10 is that the negative Nb anomalies (as measured by Ce/Nb) and elevated ⁸⁷Sr/⁸⁶Sr_i ratios of the Malaga dykes can be explained by crustal contamination (Fig. 10c). The data arrays project back to compositions typical of MORB and there is no requirement for a unique source composition such as subduction-modified mantle wedge or enriched lithospheric mantle.

Figure 9. (a) Zr/Nb vs Ti/Nb diagram showing that the Malaga dykes scatter around a composition broadly similar to MORB whereas the lamproites form a steep positive array. The calc-alkaline volcanic rocks lie between the Malaga dyke suite and the lamproite-crustal melt fields, implying a mixing relationship. (b) K₂O/Al₂O₃ vs SiO₂ to distinguish between a crustal or lamproitic contaminant for the Malaga dyke-calc-alkaline volcanic suites. The Malaga dyke-calc-alkaline volcanic rocks form a tight array extending towards the crustal melts whereas the lamproites have higher K₂O/Al₂O₃ and generally lower SiO₂. Symbols and data sources as in Fig. 4.

Figure 10. MORB-crust mixing curves compared with the Malaga dyke-calc-alkaline volcanic array in diagrams of (a) SiO₂, (b) Eu/Eu* and (c) Ce/Nb vs ⁸⁷Sr/⁸⁶Sr_i, suggesting the Malaga dykes have been contaminated by 5-20% of a crustal component whereas the crustal contribution to the calc-alkaline volcanic rocks ranges from 20 to 100%. It should be noted that the elevated ⁸⁷Sr/⁸⁶Sr_i for two of the Malaga dykes may in part reflect alteration; our choice of ⁸⁷Sr/⁸⁶Sr_i for these diagrams was motivated by the availability of additional literature data for the lamproites and crustal melts. Symbols and data sources as in Fig. 4.

Cabo de Gata and Mazarrón-Cartagena calc-alkaline volcanic rocks

Although apparently erupted significantly later, the calc-alkaline volcanic rocks appear to form a continuum with the Malaga dyke arrays. They exhibit a large compositional range with MgO, Al₂O₃ and Fe₂O₃ decreasing, and K₂O increasing, as SiO₂ increases (Fig. 4), implying an important role for crystal fractionation in their petrogenesis (e.g. Di Battistini et al., 1987). Compared with the Malaga dykes they are much more enriched in the more incompatible trace elements whereas their REE fractionation is similar, also precluding partial melting in the presence of residual garnet (see Fig. 8a). In Fig. 8b the calc-alkaline volcanic rocks spread out from the theoretical partial melting curve, and the Malaga dyke array, to higher Nb, consistent with increases in Nb content induced by fractionation ± contamination.

The calc-alkaline volcanic rocks have larger negative Ba, Ta-Nb and Ti anomalies than the Malaga dykes, similar to those exhibited by the cordierite-garnet dacites. These dacites have [delta]¹⁸O values ranging from +12·2 to +15·6%° and formed by partial melting of semi-pelitic source rocks (Munksgaard, 1984). The low Sr and negative Eu, P and Zr-Hf anomalies exhibited by these rocks probably reflect residual plagioclase, clinopyroxene and accessory phases (zircon and apatite) during partial melting. Compared with the Malaga dykes, the calc-alkaline volcanic rocks are displaced to higher ⁸⁷Sr/⁸⁶Sr_i and lower ¹⁴³Nd/¹⁴⁴Nd_i, ²⁰⁸Pb/²⁰⁴Pb and ²⁰⁷Pb/²⁰⁴Pb. In the ⁸⁷Sr/⁸⁶Sr_i-¹⁴³Nd/¹⁴⁴Nd_i diagram (Fig. 7a) they form a broad negative array which extends to much lower ⁸⁷Sr/⁸⁶Sr_i and higher ¹⁴³Nd/¹⁴⁴Nd_i than the cordierite-garnet dacite crustal melt, suggesting that these rocks do not simply reflect crustal melts. None the less, the cordierite-garnet dacite lies at one end of the Sr-Nd (but not Pb-Pb; see discussion below) isotope array formed by the calc-alkaline rocks, and the calc-alkaline volcanic rocks also lie at the end of the Malaga dyke array close to the crustal melts and lamproites in Fig. 9a. The implication is that they represent more contaminated and fractionated equivalents of the Malaga magmas.

In Fig. 10, the calc-alkaline volcanic rocks scatter around the same MORB-crust mixing curves as the Malaga dykes, though the implied amounts of crustal contamination are significantly higher (20-60%). It should be noted that, although the lamproites also lie at the end of these arrays, Fig. 9b rules out the lamproites as an end member for these rocks. Thus, partial melting of the crust has played a major role in the magmatism in the Betic-Alboran Domain, and the coupled increases in SiO₂ and ⁸⁷Sr/⁸⁶Sr combined with decreasing Eu/Eu* (Fig. 10a and b) provide good evidence that crustal contamination was coupled with crystal fractionation. As discussed above for the Malaga dykes, Fig. 10c suggests that the negative Ta-Nb anomaly in these rocks could be accounted for by crustal contamination of MORB-like magmas. A similar mantle source to that of the Malaga dykes is suggested by the similarly sub-chondritic Nb/Ta ratios of the calc-alkaline rocks. The implied amounts of crustal contamination can explain also the high ²⁰⁸Pb/²⁰⁴Pb (Fig. 11) and ²⁰⁷Pb/²⁰⁴Pb (not shown) ratios of these rocks. However, the same is not true for the ²⁰⁶Pb/²⁰⁴Pb variations, as the cordierite dacite lies between the Malaga dykes and calc-alkaline volcanic rocks in Fig. 7b and c. Unfortunately, with only a single Pb isotope analysis of the local crustal rocks, a more detailed assessment of the significance of this discrepancy must await more detailed work. The large positive Pb spike, which characterizes all of the rocks in Fig. 5, suggests that some process is dominating the Pb budget and by inference also controlling the Pb isotope signature.

Figure 11. Ce/Pb vs ²⁰⁸Pb/²⁰⁴Pb, illustrating that contamination of asthenospheric magmas by as little as 5% of a crustal component dominates the Pb isotope budget of the resultant lavas. Symbols and data sources as in Fig. 4.

Tallante alkali basalt and xenoliths

Some time after emplacement of the Malaga dykes and the onset of calc-alkaline volcanic activity, alkali basalts were erupted around Tallante. The presence of the cumulate nodules in these lavas provides evidence for clinopyroxene-hornblende fractionation, and the relatively smooth incompatible trace element patterns of the Tallante alkali basalts and their lack of depletion in Ba or Ta-Nb relative to Th and La indicate that they have not significantly interacted with the continental crust. Thus, although data on the alkali basalts remain scarce, these lavas and their inclusions may provide important insights into processes occurring in the mantle beneath the Betic Cordillera.

Partial melting models show that the observed LREE concentrations (200 times chondrite) and La/Yb ratios of 33, exhibited by the Tallante lava, cannot be produced from a primitive mantle source, even by very small degrees (<1%) of melting (Fig. 8a). The high La/Yb relative to Tb/Yb for this lava suggests an origin involving partial melting of an LREE-enriched mantle source. The elevated HREE concentrations and flat HREE patterns suggest that garnet was not a major residual phase, which would be consistent with the presence of spinel lherzolite (and absence of garnet lherzolite) xenoliths and implies a relatively shallow mantle source region. The absence of residual garnet, the need for an LREE-enriched source, and the elevated ⁸⁷Sr/⁸⁶Sr_i and low ¹⁴³Nd/¹⁴⁴Nd_i isotope ratios of the lava (Fig. 7a), rule out the asthenosphere as the source of these lavas. Moreover, the lack of negative Ta-Nb or Ti anomalies (Fig. 5c), which characterize the local crust, indicates that the elevated ⁸⁷Sr/⁸⁶Sr_i and low ¹⁴³Nd/¹⁴⁴Nd_i isotope ratios of the Tallante alkali basalts are unlikely to have resulted from crustal contamination.

Although the incompatible trace element patterns for the Tallante alkali basalts (and cumulate) are relatively smooth, they are both characterized by negative K anomalies and positive Pb anomalies, which are likely to reflect real source features. Venturelli et al., (1984) noted that phlogopite in these lavas has a higher Ni/MgO ratio than coexisting olivine, an observation they attributed to disequilibrium partial melting of a phlogopite-bearing source. The notion of a relatively shallow, enriched mantle source region for the Tallante alkali basalts is also supported by the observation of veining and metasomatism in some of the spinel lherzolite xenoliths and their relatively low equilibration temperatures of ~850°C (Ancochea & Nixon, 1987), which would correspond to depths of ~70 km for a typical continental geotherm and are consistent with the absence of a garnet signature. The Cr, Ni, CaO, Al₂O₃ and MgO contents of the three xenoliths analysed here (Table 3) are typical of mantle residues (rather than cumulates). However, these xenoliths have relatively flat REE patterns (Fig. 6c) and incompatible trace element concentrations are atypical of those expected in a mantle residue. Indeed, their negative Ba and K anomalies may reflect enrichment by partial melts that were themselves in equilibrium with residual amphibole or phlogopite. This sort of enrichment, either by fluids or small degree partial melts from the convecting upper mantle, may be characteristic of the lower mechanical boundary layer of the lithosphere, i.e. that part of the lower lithospheric mantle which does not convect and so can preserve trace element anomalies and develop resultant isotope anomalies over geological time-scales (e.g. McKenzie, 1989). If so, the Nd-depleted mantle model age for the Tallante alkali basalts suggests this occurred around 700 Ma. However, this constraint is significantly relaxed if enrichment occurred by addition of small partial melts from a mantle plume, in which case the age of enrichment could be significantly younger than 700 Ma.

Lamproites

The origins of the Spanish lamproites have been discussed in some detail previously (Nixon et al., 1984; Venturelli et al., 1984, , 1988; Nelson et al., 1986). The mg-numbers of the lamproites (74-83) and their olivine phenocrysts (Fo_90-94, Venturelli et al., 1984, , 1988) suggest that these rocks represent relatively unmodified primary mantle melts. Nevertheless, they are also characterized by having high SiO₂ but distinctly low Al₂O₃, Fe₂O₃ (total iron) and Na₂O relative to melts known to have come directly from the depleted upper mantle. Such features suggest derivation from peridotite that had been depleted during a previous melt extraction event (e.g. Turner & Hawkesworth, 1995). Similarly, Venturelli et al., (1984) observed that the high Ni/MgO ratios of both the whole rocks and their phlogopite phenocrysts probably requires their source was melt depleted. As melt depletion can result in a mantle source with sub-chondritic Nb/Ta it is interesting that the lamproites have chondritic Nb/Ta ratios of 16-18; however, this ratio may have been increased by subsequent source enrichment events (see below) and/or by small degrees of partial melting.

Another striking feature of the lamproites is their extreme degree of enrichment in incompatible trace elements. The concentrations of these elements are significantly higher than in the local crustal melts, which in conjunction with the primary nature of these rocks makes models in which their trace element or isotope signatures reflect solely the effects of crustal contamination untenable. Moreover, Fig. 8a and b shows that the LREE-HREE fractionation and incompatible trace element concentrations in these rocks cannot be produced from a primitive mantle source even by unrealistically low degrees (<1%) of partial melting (e.g. Venturelli et al., 1984). This difficulty is further compounded by the major element evidence for prior melt depletion of the source, which would have effectively stripped it of all its incompatible elements. Thus, the data require at least a two-stage source history in which addition of incompatible trace elements occurred in an enrichment event subsequent to melt depletion and the high K₂O/Na₂O ratios and the sigmoidal REE patterns may require an even more complex, two- or three-stage, source history (e.g. Venturelli et al., 1984). Such mantle source histories of melt extraction followed by enrichment events are a characteristic of old continental mantle lithosphere (e.g. Hawkesworth et al., 1990; Turner & Hawkesworth, 1995).

In more detail, the relatively high concentrations and flat pattern of the HREE suggests that these melts did not form in the presence of residual garnet, although their source region was probably enriched by a component that had seen residual garnet (e.g. Nixon et al., 1984). Additionally, some of the lamproites host spinel lherzolite xenoliths bearing evidence for veining and metasomatism (Ancochea & Nixon, 1987). Both pieces of evidence are consistent with a relatively shallow, lithospheric origin for these rocks. Phlogopite is a liquidus phase in these rocks and the high K₂O contents of the lamproites mean that these melts would have been increasingly saturated in phlogopite at higher pressures, though the presence of phlogopite and absence of garnet constrain the depth of the source region to <2·5 GPa (Edgar et al., 1976; Mengel & Green, 1989). Thus, given the primary nature of the lamproites, we suggest that their source region contained phlogopite (and probably also apatite; Venturelli et al., 1988). Furthermore, the low concentrations of K relative to Th or La in Fig. 5d, and evidence that K has been buffered relative to other incompatible trace element ratios or concentrations (Fig. 12a and b), indicate that phlogopite was a residual phase during partial melting. Venturelli et al., (1988) have suggested that partial melting may have occurred at relatively shallow pressures at a phlogopite + olivine + orthopyroxene + liquid peritectic. Even more intriguingly, Nb shows an almost identical behaviour to K, with Nb/Zr decreasing as Ce/Y increases (Fig. 12c) and Nb concentrations showing little or no correlated increase with increasing Zr, even though Nb concentrations span a considerable range (Fig. 12d). Thus, it would appear that Nb concentrations have also been controlled by a residual source phase. The most obvious candidates would be a residual Ti-rich phase such as rutile or ilmenite, which might be stabilized at elevated oxygen fugacities within the lithospheric mantle. However, Fig. 13 shows that Nb concentrations increase as the size of the negative Ti anomaly in Fig. 5d increases. Therefore this phase is unlikely to be a residual Ti-rich phase. The similarity in the behaviour of K and Nb (Fig. 12) suggests that the same phase may be responsible for the buffering of concentrations of both elements, and recent data from subduction metasomatized mantle xenoliths show that phlogopite can contain appreciable concentrations of Nb and Ta (Ionov & Hofmann, 1995).

Figure 12. Evidence for the role of minor residual phases during partial melting within the lithospheric mantle to produce the lamproites. The lamproites display a negative trend in K/Ce vs Ce/Y in (a) and K₂O remains relatively constant over a large range of Zr in (b) suggesting that K was compatible (in phlogopite) during partial melting. In contrast, the dykes and calc-alkaline volcanic rocks define a steep positive trend in (b) indicating similar incompatibility of K₂O and Zr. Similarly in (c) and (d), Nb/Zr forms a negative array against Ce/Y for the lamproites and Nb is relatively constant over a large range of Zr compared with the steep trend shown by the dykes and calc-alkaline volcanic rocks. This implies that Nb was compatible during partial melting, possibly in residual phlogopite (see text for discussion). Symbols and data sources as in Fig. 4.

Figure 13. Nb vs Ti/Ti* (where Ti* is the expected Ti concentration based on an interpolation between Zr and Y in Fig. 5d), showing that Nb concentrations increase as the size of the negative Ti anomaly increases; this implies that Nb is not being controlled by a residual Ti phase (e.g. rutile, ilmenite). Symbols and data sources as in Fig. 4.

The enrichment event(s) that affected the lithospheric mantle source region of the lamproites resulted in the addition of incompatible elements, including volatiles conducive to phlogopite stabilization. If this enrichment resulted from the addition of small partial melts from the depleted upper mantle (e.g. McKenzie, 1989), then the elevated ⁸⁷Sr/⁸⁶Sr_i and low ¹⁴³Nd/¹⁴⁴Nd_i of the lamproites would suggest that this occurred in the Proterozoic, as the Nd-depleted mantle model ages for the lamproites are 1·6-1·9 Ga. However, the Spanish lithosphere may not be old enough for this isotope ingrowth model (Nelson et al., 1986), and such a model does not easily explain the negative Eu anomalies that characterize these rocks. Such features are unlikely to have developed during partial melting. Negative Eu anomalies can be produced by plagioclase, as either a residual or a fractionating phase. However, plagioclase is not generally stable at mantle depths and is not a liquidus phase in the lamproites, which approximate primary melts and thus cannot have undergone significant crystal fractionation. The alternative is that the enrichment process involved the addition of ancient incompatible element enriched material, which itself already possessed these trace element and isotopic characteristics. Nelson et al., (1986) have noted that the elevated ²⁰⁸Pb/²⁰⁴Pb and ²⁰⁷Pb/²⁰⁴Pb of the lamproites is most similar to modern oceanic pelagic sediments and would be consistent also with their high ⁸⁷Sr/⁸⁶Sr_i and low ¹⁴³Nd/¹⁴⁴Nd_i. Furthermore, negative Ta-Nb, Ti and Eu anomalies are features of crustal sediments and therefore it is argued that the mantle source region of the lamproites was enriched by the addition of small amounts (~2-3% on the basis of the Ce/Pb and Eu/Eu* ratios) of a sediment component, most probably partial melts of sediments introduced during an ancient subduction event (e.g. Hergt et al., 1989).

Direct evidence for fluxing of the mantle beneath the Betic-Alboran Domain with fluids containing crustal components comes from the geochemistry of the peridotites and pyroxenites of the Ronda (Reisberg & Zindler, 1986) and Beni Bousera peridotites (Pearson et al., 1991, , 1993; Kumar et al., 1996). The pyroxenites at Beni Bousera have anomalously high oxygen isotope compositions (Pearson et al., 1991), which, together with their radiogenic isotope systematics that overlap those of the lamproites and Tallante alkali basalt, indicate a sediment-derived component in their source (Pearson et al., 1993). Model Nd and Os ages are Proterozoic (Pearson et al., 1993; Kumar et al., 1996) and hence may support ancient enrichment as a means of generating the alkali basalt and even lamproite isotopic systematics, although some of the pyroxenites carrying this `enriched' signature may well be younger (Pearson et al., 1993).

Summary

Magmatism in the Betic-Alboran Domain was largely diffuse with the only sense of temporal and spatial migration being outward directed, as suggested by the location of the lamproites lying to the north of the calc-alkaline rocks (Fig. 2). Magmatism appears to have commenced with shallow, decompression melting within the asthenosphere following removal of lithospheric mantle to produce a swarm of tholeiitic dykes. Subsequently, protracted eruption of calc-alkaline magmas appears to reflect increasing degrees of crustal contamination of similar, asthenosphere-derived magmas. Although it may not be possible to rule out a continental lithospheric mantle origin for the Malaga dykes and/or the calc-alkaline rocks, the geochemical data at hand do not demand such a source. One piece of evidence that may favour such an origin is the sub-chondritic Nb/Ta ratios that characterize both the dykes and calc-alkaline rocks and may reflect a melt-depleted lithospheric source. The calc-alkaline rocks were succeeded by crustal partial melts. Finally, alkali basalts and then lamproites formed by partial melting of enriched lithospheric mantle, some 4 my after the onset of calc-alkaline magmatism.

Retreating subduction zone

The retreating subduction zone model (e.g. Lonergan & White, 1996) predicts a westward migrating volcanic arc showing a characteristic subduction-related geochemical signature. One of the great difficulties in interpreting the geochemistry of subduction-related rocks is distinguishing between the effects of sediment addition to the mantle source and those of shallow level, crustal contamination (e.g. Davidson, 1987). In the present case, although the main phase of magmatism in the Alboran Sea resembles that in island arcs, the development of the subduction signature appears to result from combined assimilation and fractional crystallization. In other words, features such as the negative Nb anomalies are not primary features that demand that the mantle source had been modified by subduction (see Fig. 10c). Furthermore, the geometry and timing of magmatism in no way suggests an island arc migrating behind a retreating subduction zone. For these reasons we feel that some form of lithospheric removal of lithosphere provides the best explanation for the syn- to post-extensional volcanism in the Alboran region, and this mechanism is most easily reconciled with the seismological evidence for detached bodies of cold mantle at depth (Grimison & Chen, 1986; Blanco & Spakman, 1993; Seber et al., 1996).

Slab detachment

Slab detachment (Davies & von Blanckenburg, 1995) predicts a narrow, linear zone of magmatism and uplift that propagates along strike and is of limited extent. These magmas should be derived from the overlying lithosphere that was metasomatized during the preceding period of subduction, unless detachment occurs at a depth shallower than 50 km, in which case decompression melting of the asthenosphere is predicted (Davies & von Blanckenburg, 1995). This model seems inappropriate to the Betic-Alboran Domain on geometrical grounds. The pattern of magmatism and exhumation in this region does not define a linear zone and the magmatism is of considerable extent if the volcanic thicknesses in the Cabo de Gata region (~1000 m) are extrapolated beneath the Alboran Sea (Fig. 1). Also, the lithospheric mantle-derived magmas appear to be derived from ancient enriched lithosphere (recall Nd model ages ~1·6-1·9 Ga), rather than one metasomatized during recent subduction. Thus, we do not find evidence to support the slab detachment model.

Delamination

Delamination of the lithospheric mantle (Bird, 1979) will bring hot asthenosphere into contact with the Moho. This should promote massive crustal melting and predicts a progressive migration of the resultant volcanism in the direction of delamination propagation. In the Betic-Alboran Domain, the only sense of migration to the magmatism is diffusely outwards from the Alboran Sea region (Fig. 2). Moreover, although unambiguous crustal melts do exist (e.g. the cordierite dacites), these are of small volume compared with the calc-alkaline volcanic rocks, which have a clear mantle component (see Fig. 10). Therefore, if massive crustal melting should be the hallmark of delamination, we find little evidence for it in the Betic-Alboran Domain.

Convective removal

Convective thinning of the lithospheric mantle (Houseman et al., 1981) may produce an outward pattern of propagation of lithospheric melts rich in large ion lithophile and volatile elements (Turner et al., 1992; Platt & England, 1994), depending on how much lithosphere has been removed. If a combination of the amount of thinning and any resultant extension raises the asthenosphere to within 50 km of the surface then decompression melting of the asthenosphere is predicted, with increasing amounts of crustal contamination of the asthenospheric melts as conductive and magmatic heating induces crustal melting.

It may not be possible to conclusively distinguish between the different mechanical and geometrical models for lithosphere removal using the geochemistry (but see discussion above), but for two reasons we will couch the remaining discussion in terms of the concept of convective thinning of the lithosphere, as originally proposed by Houseman et al., (1981), and recently elaborated by Houseman & Molnar, (1997). The first is that the pattern of both extension and volcanism in the Alboran region is diffuse, lacking any obvious linear trend. The only suggestion of a temporal migration of activity is outwards: extension and magmatism started in early Miocene time in the Alboran Basin itself, whereas local extensional basins and volcanic centres developed within the peripheral Betic and Rif thrust belts to the north and south of the Alboran Domain in late Miocene and Pliocene time (Figs 1 and 2). This pattern is most easily reconciled with convective removal beneath the Alboran Domain, propagating outwards, possibly in response to the continued convergence between Africa and Iberia.

The second reason for favouring convective thinning is that the mechanism has a sound physical basis (Houseman et al., 1981; Houseman & Molnar, 1997); the process is supported by observational evidence from the ocean basins (Parsons & McKenzie, 1978); and the thermal, mechanical, and magmatic consequences have been discussed in some detail (England & Houseman, 1989; Sandiford & Powell, 1990; Turner et al., 1992; Zhou & Sandiford, 1992; Platt & England, 1994).

The magmatic record of the Betic-Alboran orogenic domain and its surroundings are shown in a series of N-S sections in Fig. 14. Convergent thickening in the Alboran Domain continued until ~40 Ma, resulting in a doubling of crustal thickness from ~30 to 60 km. Modelling of the P-T-time path experienced by rocks recovered on ODP Leg 161 from beneath the Alboran Sea (e.g. Platt et al., 1998) suggests that extension started at ~30-27 Ma and was closely linked to intrusion of the Malaga dykes. These signify the onset of significant degrees (~10-15%) of decompression melting within the asthenosphere. Assuming upper mantle with a normal potential temperature of ~1280°C, partial melting will commence only if the asthenosphere is brought to within ~50 km of the surface (McKenzie & Bickle, 1988), consistent with the geochemical evidence that the Malaga dykes represent partial melting in the absence of residual garnet. Given the inferred crustal thicknesses, this requires that the lithospheric thickness was reduced to ~50 km by some combination of convective removal and stretching (Fig. 14b).

Figure 14. Tectonic and magmatic evolution of the Betic-Alboran orogenic domain and its surroundings, shown in a series of N-S sections along the present-day 2°W meridian. Sections are true-scale, and the motions (after Platt & Vissers, 1989) are shown relative to the Iberian plate. (a) Section for ~27 Ma, after roughly homogeneous thickening of the Alboran crust and lithosphere. MBL, mechanical boundary layer; TBL, thermal boundary layer. The thermal boundary layer consists of lower lithospheric mantle with a low enough viscosity to be involved intermittently in convection. The position of the base of the MBL is speculative: we have taken into account the effects of radioactive heating of the overlying thickened crust (Platt et al., 1998) and of total lithospheric thickness (Houseman & Molnar, 1997). (b) At ~27 Ma convective removal of lower lithospheric mantle results in decompression melting within the underlying asthenosphere and surface uplift. The resulting increase in the gravitational potential energy leads to extensional collapse of the Alboran Domain. (c) By 19 Ma the margins of the extending Alboran Domain have been thrust onto the adjacent Iberian and African continental margins. It should be noted that the main direction of extension was E-W, so that cross-sectional area is not conserved in this figure. Melting within the asthenosphere continues, and by 15 Ma these magmas have mixed with crustal meltsforming near the Moho to create the main phase of calc-alkaline volcanism. Continuing convergence of the Iberian and African plates causes further lithospheric thickening beneath the margins of the Alboran domain. (d) As Africa-Iberia convergence continues, laterally propagating convective removal of lithosphere causes lithospheric melting and alkaline to lamproitic volcanism on the margins of the orogen. PTB, peripheral thrust belts.

Removal of the dense lithospheric root beneath the orogen would result in major increases in elevation (~5 km) accompanied by increased gravitational potential energy and outwardly directed buoyancy forces (e.g. England & Houseman, 1989; Sandiford & Powell, 1990). A second response will be a rapid rise in temperature caused by placing hot asthenosphere close to the Moho (Fig. 14b). These predictions are substantiated by the P-T history of the metamorphic rocks recovered from beneath the Alboran Sea during ODP Leg 161. These record a rapid episode of decompression that was accompanied by heating of the order of 100°C at ~20-18 Ma (Soto & Platt, 1999; Soto et al., 1999). There are two consequences. The first is that the strength of the lithosphere will be greatly reduced such that the potential energy resulting from the elevation increases will lead to extensional collapse of the orogen (e.g. England & Houseman, 1989; Zhou & Sandiford, 1992) as illustrated in Fig. 14b and c. The second consequence is the onset of partial melting within the lower parts of the crust (Fig. 14c). The Malaga dykes provide evidence for interaction with the crust. The temporal evolution to calc-alkaline volcanism in which the crustal contribution is 20-50% and the subsequent eruption of cordierite-garnet dacites, which may represent 100% crustal melts, corroborates these predictions. In this context it should be recalled that there is evidence that much of the Alboran Sea is underlain by equivalents of the Cabo de Gata volcanic rocks (Comas et al., 1999) and so the volume of these magmas may well be considerable (though poorly constrained, we estimate a minimum volume of 40 000 km³).

Removal of lithospheric mantle results in juxtaposition of hot asthenosphere against previously cooler lithospheric mantle (Fig. 14b). As noted by Turner et al., (1992, , 1993, 1996a) and Platt & England, (1994), any portions of the lithospheric mantle previously enriched by volatiles will have a lower solidus, relative to anhydrous asthenospheric peridotite, and will be likely to undergo partial melting upon heating by the newly juxtaposed asthenosphere. The likely rates of conductive heating suggest that this will occur on time-scales of 0-10 my (Turner et al., 1996b) depending on the location of volatile-enriched portions in the lithospheric mantle relative to the amount removed. The lamproites clearly provide evidence for such magmatism. Moreover, the temporal progression from the Tallante alkali basalts, interpreted as being derived from the lower mechanical boundary layer of the lithospheric mantle, to the lamproites that were derived from shallower lithospheric mantle, which had undergone a previous melt extraction event, is consistent with progressive heating of the lithosphere from below. However, the emplacement of the alkali basalts and the lamproites post-dates the onset of asthenospheric and crustal-derived magmatism. It may be that older, lithosphere-derived rocks lie buried beneath the Alboran Sea calc-alkaline volcanic rocks, although this does not resolve the temporal evolution observed inland. Another possibility is suggested by the spatial restriction of the lamproites to regions inland of the calc-alkaline volcanic rocks (Figs 2 and 14d). The greater crustal thickness and elevation of the emergent part of the orogen suggests that it may be underlain by thicker lithosphere than the Alboran Sea, as a result of the emplacement of the northern part of the Alboran Domain onto the Iberian continental margin during the Miocene. If this is correct, it implies that limited convective removal of Iberian lithospheric mantle may have occurred during the late Miocene, leading to partial melting within the lower part of the remaining lithospheric mantle and generation of the lamproites after the onset of calc-alkaline magmatism within the Alboran Domain. This is supported by the coincidence of the lamproite emplacement with a phase of uplift of this area which elevated marine Messinian to as much as 700 m above sea level. Thus, it would appear that early Miocene extension in the Alboran Sea region caused thrusting around its periphery, followed by outwardly propagating removal of lithospheric mantle. Finally, present-day lithospheric thicknesses beneath the Alboran Sea range from 40 to 90 km (Polyak et al., 1996), suggesting that the lithosphere has rethickened by thermal accretion.