Post-collisional Neogene magmatism of the Mediterranean Maghreb margin: a consequence of slab breakoff

C. R. Acad. Sci. Paris, Sciences de la Terre et des planètes / Earth and Planetary Sciences 331 (2000) 159–173 © 2000 Académie des sciences / Éditions scientifiques et médicales Elsevier SAS. Tous droits réservés S1251805000014063/REV

Geodynamics / Géodynamique

CONCISE REVIEW PAPER LE POINT SUR...

Post-collisional Neogene magmatism of the Mediterranean Maghreb margin: a consequence of slab breakoff René C. Maurya *, Serge Fourcadeb, Christian Coulonc, M’hammed El Azzouzia,d, Hervé Bellona, Alain Coutellea, Aziouz Ouabadib,e, Belkacem Semroude, M’hamed Megartsie, Joseph Cottena, Ouardia Belanteure, Amina Louni-Hacinie, Alain Piquéa, Ramon Capdevilab, Jean Hernandezf, Jean-Pierre Réhaulta a b c d e f

UMR 6538 Domaines océaniques, université de Bretagne occidentale, BP 809, 29285 Brest, France UPR 4661 Géosciences Rennes, université de Rennes-1, Campus de Beaulieu, 35042 Rennes cedex, France Laboratoire de pétrologie magmatique, université d’Aix-Marseille-3–Saint-Jérôme, 13397 Marseille cedex 20, France Faculté des sciences, université Mohammed-V, av. Ibn-Batouta, BP 1014, Rabat, Maroc Institut des sciences de la Terre, USTHB, BP 32, Bab Ezzouar, 16111 Alger, Algérie Institut de minéralogie et pétrographie, université de Lausanne, BSFH 2, 1015 Lausanne, Suisse

Received 25 April 2000; accepted 26 June 2000 Written on invitation of the Editorial Committee

Abstract – A 1 200 km-long linear magmatic belt extends along the Mediterranean coast of the Maghreb from Eastern Tunisia to Morocco. This belt is mainly composed of Langhian calc-alkaline metaluminous to peraluminous granitoids and associated andesites/ dacites in Central and Eastern Algeria. In Tunisia and Oranie/Western Morocco, calcalkaline activity started later (during the Serravallian) and was followed by the emplacement of alkali basalts and basanites since the Tortonian to the Pliocene and, in some places, the Pleistocene. Available data on the tectonic setting, petrology, age and geochemistry of this belt show that most of its striking features, e.g. (1) very low magma production rate, subduction-related geochemical imprint, extensive crustal contamination for the calc-alkaline magmatism and (2) progressive magmatic change from calc-alkaline to alkaline, are consistent with magma generation during a slab breakoff process as proposed by Carminati et al. in 1998. The magmatism associated with this breakoff started in Central Eastern Algeria at 16 Ma, then propagated eastwards and westwards. The upward flow of asthenospheric enriched plume-type mantle through the tear in the downgoing slab first triggered melting of the overlying lithospheric mantle which had been metasomatised during a previous subduction period. Heat supply from this uprising asthenosphere may have warmed up the continental crust and made its involvement in assimilation processes easier. As the asthenosphere ascended through the ‘window’ in the slab, partial melting occurred at the uprising boundary between asthenosphere and lithosphere, generating basalts with transitional characteristics between those of calc-alkaline and alkaline basalts. As the asthenospheric upwelling proceeded, partial melting then occurred in the sole asthenospheric mantle, producing alkali basalts. © 2000 Académie des sciences / Éditions scientifiques et médicales Elsevier SAS breakoff (slab) / magmatism / calc-alkaline / alkaline / Neogene / Maghreb

* Correspondence and reprints to: [email protected]

159

R.C. Maury et al. / C. R. Acad. Sci. Paris, Sciences de la Terre et des planètes / Earth and Planetary Sciences 331 (2000) 159–173

Résumé – Le magmatisme néogène post-collisionnel de la marge méditerranéenne du Maghreb : une conséquence de la rupture de lithosphère subduite. La marge septentrionale du Maghreb comporte une chaîne magmatique, qui s’étend sur 1 200 km des Mogods (Tunisie) à Ras Tarf (Maroc). En Algérie centrale et orientale, elle se compose principalement de granitoïdes métalumineux et peralumineux langhiens et d’andésites et dacites calco-alcalines associées. En Tunisie, en Oranie et au Maroc, l’activité calcoalcaline débute plus tard, au Serravallien. Elle est, par la suite, relayée par un magmatisme alcalin d’âge Tortonien à Pliocène et, localement, Pléistocène. La synthèse des données tectoniques, géochronologiques, pétrologiques et géochimiques disponibles nous permet de montrer qu’un processus de rupture de lithosphère plongeante proposé par Carminati et al. en 1998 rend compte de l’essentiel des particularités magmatiques de cette chaîne (comme le très faible taux de production magmatique, l’empreinte géochimique de subduction des roches calco-alcalines et la forte contamination crustale qu’elles ont subie, ainsi que la transition calco-alcalin–alcalin au cours du temps). Le magmatisme lié à cette délamination lithosphérique a débuté vers 16 Ma en Algérie centrale et orientale, puis s’est propagé vers l’est et vers l’ouest. Le flux ascendant de manteau asthénosphérique enrichi (de type panache) au travers de la déchirure lithosphérique a d’abord provoqué la fusion du manteau lithosphérique sus-jacent, métasomatisé lors d’une période antérieure de subduction, ce qui a entraîné la formation de magmas calcoalcalins. Ce flux asthénosphérique a également contribué à réchauffer la croûte continentale, facilitant ainsi son assimilation par les magmas d’origine mantellique. Par suite de la remontée progressive de l’asthénosphère dans la déchirure lithosphérique aux deux extrémités de la chaîne, la fusion partielle s’est produite à la limite lithosphère–asthénosphère, engendrant des magmas basaltiques, dont la signature géochimique est transitionnelle entre les types calco-alcalin et alcalin. Consécutivement à l’élargissement de la déchirure lithosphérique, la remontée asthénosphérique s’est poursuivie ; la fusion partielle s’est alors exercée dans le seul manteau asthénosphérique enrichi, produisant des liquides basaltiques alcalins. © 2000 Académie des sciences / Éditions scientifiques et médicales Elsevier SAS rupture (lithosphérique) / magmatisme / calco-alcalin / alcalin / Néogène / Maghreb

Version abrégée 1. Introduction En dépit des nombreux travaux régionaux consacrés, durant les trente dernières années, à l’étude de la chaîne magmatique néogène qui borde la marge méditerranéenne du Maghreb (MM), les modalités de son origine et de son évolution sont demeurées incertaines, voire controversées. Les premières études synthétiques [6–8, 36] ont mis en évidence : (1) la prédominance, en Algérie centrale et orientale, de magmas calco-alcalins d’âge Langhien et généralement riches en potassium, et (2) le déplacement, durant le Serravallien, de l’activité calco-alcaline vers l’est (Tunisie) et vers l’ouest (Oranie et Maroc septentrional). Les deux extrémités de la chaîne sont caractérisées par le développement d’un volcanisme alcalin du Tortonien au Plio-Pléistocène. L’ensemble des magmas s’est mis en place à un stade clairement post-collisionnel [44], recoupant les nappes des chaînes rifaine et tellienne, parfois même leur avant-pays africain. Bien que les signatures pétrologiques et géochimiques des magmas calco-alcalins de la MM traduisent clairement leur affinité avec les magmas des zones de subduction actuelle, ils s’en distinguent

160

par des volumes émis beaucoup plus faibles (2 000 km3 pour l’ensemble de la MM), un caractère globalement beaucoup plus potassique, et enfin, une très forte empreinte crustale. De telles caractéristiques sont difficilement explicables par un épisode de subduction péné-contemporain du magmatisme qui, d’ailleurs, serait à pendage sud, alors que l’enracinement des charriages miocènes au nord de la côte actuelle suggère un plongement vers le nord. D’autre part, la nature du magmatisme de la MM et son évolution au cours du temps sont compatibles avec un processus de rupture de lithosphère subduite [23, 81]. Il en est de même des particularités [16, 17] de l’évolution géologique de la MM au Langhien : magmatisme bimodal, développement d’hydrothermalisme et de minéralisations associées [56], exhumation d’unités métamorphiques profondes [60] dans un climat général d’extension [1, 43]. Le but de cet article est de proposer un modèle géodynamique rendant compte des caractéristiques pétrologiques, géochimiques et géochronologiques du magmatisme de la MM, en prenant en considération tant les structures tectoniques superficielles que les données de la tomographie sismique.

R.C. Maury et al. / C. R. Acad. Sci. Paris, Sciences de la Terre et des planètes / Earth and Planetary Sciences 331 (2000) 159–173

2. Particularités et sources des magmas de la marge méditerranéenne du Maghreb

transitionnelles d’Oranie est liée à la fusion d’un manteau situé à la limite lithosphère–asthénosphère [22].

Les principales caractéristiques géochronologiques (40K–40Ar), pétrologiques et géochimiques (rapports isotopiques 87Sr/86Sr et 143Nd/144Nd) des magmas de la MM sont présentées sur la figure 1. Outre les faits majeurs exposés précédemment, on note l’existence, dans cette chaîne, de roches basaltiques, ou plus rarement andésitiques, transitionnelles entre les types magmatiques calco-alcalin et alcalin. Ces laves transitionnelles existent aux deux extrémités de la chaîne volcanique, en Tunisie (secteur 3 de la figure 1, d’après les données de Halloul [40]) et surtout en Oranie (secteur 12 [22, 52]) et au Maroc (secteurs 14 et 15 [31]). Elles sont, pour l’essentiel, contemporaines, entre 10 et 7 Ma, de la fin de l’épisode calco-alcalin. Leur mise en place est suivie, dans les mêmes régions ou à proximité, d’un volcanisme alcalin typique (basaltes alcalins et basanites). Les spectres d’éléments incompatibles (figure 2) témoignent de l’étonnante diversité des roches de la MM. Les magmas calco-alcalins (La/Nb > 2) varient depuis des types faiblement potassiques (secteur 10 C) jusqu’à des shoshonites et des roches très potassiques (secteur 11 C) : tous présentent les anomalies négatives en Nb et, à un degré moindre, en Ti, typiques des magmas d’arc. Les basaltes alcalins (secteurs 12 A et 13 A) sont, en revanche, caractérisés par des anomalies positives en Nb, similaires à celles des basaltes d’îles océaniques (OIB) de type HIMU (La/Nb < 1). Les laves transitionnelles ont, quant à elles, des rapports La/Nb compris entre 1 et 2. Les types pétrographiques, les rapports isotopiques 87 Sr/86Sr et 143Nd/144Nd et les âges 40K–40Ar des roches magmatiques de la MM sont corrélés (figure 3). On note la grande hétérogénéité du groupe calco-alcalin, due pour partie à une très forte contamination crustale. Les basaltes alcalins présentent des rapports isotopiques de type OIB attribuables à l’influence d’un panache asthénosphérique. La signature géochimique des laves transitionnelles est intermédiaire entre celles des deux pôles précédents. Ces données permettent d’identifier trois types majeurs de sources pour les magmas de la MM : un manteau asthénosphérique enrichi, de type panache, similaire à la source principale du volcanisme alcalin ouest-européen [47] ; un manteau lithosphérique, caractérisé par une empreinte géochimique de subduction, et enfin la croûte continentale, inférieure et supérieure, vraisemblablement très hétérogène sous la MM. Des modèles géochimiques basés sur les éléments majeurs et en traces ainsi que sur les isotopes de Sr, Nd et O permettent d’aboutir aux conclusions suivantes : (1) les roches calco-alcalines (granitoïdes et volcanites), bien que très fortement contaminées par la croûte de la MM, dérivent de la fusion du manteau lithosphérique modifié par la subduction [35] ; (2) l’origine des laves

3. Le modèle de rupture de lithosphère subduite Les caractéristiques des magmas de la MM sont explicables dans le cadre d’un modèle de détachement de lithosphère subduite [16, 17, 23, 68, 81] dont les modalités sont schématisées sur la figure 4. Le processus de rupture lithosphérique s’inscrit dans le cadre de l’histoire complexe de la marge africaine, bordée au nord par les blocs continentaux du Rif et des Kabylies et les sillons de flyschs alpins (stade 1). Ce système entre en compression nord–sud à partir du Cénomanien. La suturation du sillon marginal africain intervient précocement à la fin de l’Éocène par le développement d’une subduction à pendage nord (stade 3). Nous admettons, en tenant compte des données tomographiques sur l’Algérie orientale, que la partie profonde de la plaque subduite, ancrée dans le manteau, s’est désolidarisée de sa partie superficielle, qui a poursuivi sa migration vers le nord (stade 4). Les effets magmatiques du détachement ne sont apparus que tardivement, à 16 Ma (Langhien : stade 5), à la faveur des distensions qui ont suivi la mise en place des nappes rifaines et telliennes, en face du « poinçon maghrébin », qui s’est développé au niveau de l’Algérie centrale et orientale [67]. La fusion du manteau lithosphérique, hydraté et enrichi en éléments incompatibles lors de la période de subduction précédente, est déclenchée par l’anomalie thermique due à la remontée de manteau asthénosphérique, au niveau de la déchirure lithosphérique qui apparaît sous l’Algérie centrale et orientale. Cet épisode de fusion produit des magmas calco-alcalins basiques, qui se contaminent très fortement en traversant la croûte continentale africaine (elle-même affectée par l’anomalie thermique qui ne provoque cependant que très localement son anatexie) et donnent naissance aux granitoïdes et aux laves associées d’âge Langhien. Dans un stade ultérieur, au Serravallien (stade 6), la rupture lithosphérique se propage latéralement vers les ailes du dispositif, et atteint la Tunisie d’une part, l’Oranie et le Maroc septentrional d’autre part. Dans ces secteurs, la fusion du manteau lithosphérique modifié par la subduction antérieure se produit sous l’effet de la remontée de l’asthénosphère. Cet épisode donne naissance à des associations calco-alcalines plus récentes que celles d’Algérie centrale et orientale. Au Tortonien, entre 10 et 7 Ma, le magmatisme cesse dans la partie centrale du dispositif (stade 7), alors que l’élargissement de la déchirure lithosphérique aux deux extrémités de la MM a pour conséquence la remontée de la limite lithosphère–asthénosphère. La fusion partielle se produit alors : (1) soit dans cette zone frontière, où elle engendre des basaltes transitionnels (dont la source comporte 10 à 35 % de composant asthénosphérique en Oranie), (2) soit au sein de l’asthénos-

161

R.C. Maury et al. / C. R. Acad. Sci. Paris, Sciences de la Terre et des planètes / Earth and Planetary Sciences 331 (2000) 159–173

phère, avec la formation des premiers magmas alcalins (Mogods, secteur 2 et Guilliz, secteur 15). Enfin, les dernières manifestations de la contribution du manteau lithosphérique modifié par la subduction se traduisent, au Messinien, par l’épanchement de laves shoshonitiques au Maroc dans le massif du Gourougou. Le volcanisme pliocène de la MM, typiquement alcalin,

dérive de la fusion du manteau asthénosphérique, les liquides basaltiques interagissant localement avec le manteau subcontinental du Maroc [31]. Au Pléistocène, le volcanisme alcalin cesse dans la MM et se déplace vers le sud-ouest (Moyen Atlas), au niveau du linéament transmarocain, qui a pu canaliser les remontées asthénosphériques.

1. Introduction

(16–15 Ma) towards Tunisia (14–8 Ma) and Morocco (12–5 Ma); (2) younger alkaline volcanism which started at ca. 8 Ma in Tunisia and Morocco [7] and was active up to the Pleistocene in Oranie, Algeria [52] and Morocco [31]. A major difficulty in the study of post-collision magmatism is to understand which tectonic process(es) can lead to partial melting of very variegated mantle and crustal sources in an often restricted time span and in a limited area. For instance, the occurrence of Neogene ‘orogenic’ calc-alkaline and alkali basaltic magmas in the MM has been documented since several decades (see [8, 36] and references therein) but no satisfactory global tectonic interpretation was proposed then. A long-standing problem with the post-collision calcalkaline associations in the MM has been to reconcile their subduction-related geochemical signature with the apparent lack of a contemporaneous subduction zone beneath the Maghreb margin and the coeval opening of oceanic domains in the Western Mediterranean. Many authors considered that these calc-alkaline associations were subduction-related, but outlined several weaknesses of this hypothesis (e.g. [9]). They noted, for instance, that the MM belt differs from the majority of active volcanic arcs by its rather low magma production and by their K-rich character with respect to average calc-alkaline andesite. Moreover, the timing of the magmatic events [6, 7] compared to that of the tectonic phases did not support the subduction model. Hernandez and Lepvrier [44] were the first petrologists to state clearly that the magmatism of the Algiers area was postcollisional, although Glangeaud [37] already demonstrated it post-dated the major Alpine compressive phase. Hernandez et al. [43] and de Larouzière et al. [26] proposed that shear heating at the level of the Betic transcurrent shear zone may have favoured the melting of an heterogeneous lithospheric mantle and the overlying continental crust; however this explanation can hardly be generalised to the whole MM belt. During the 80s, an increasing number of authors (e.g. [69, 70]) favoured the hypothesis of a Cenozoic northward-dipping subduction previously proposed by Wezel [84], Dercourt [28], Boccaletti and Guazzone [14], Auzende et al., [4], Bellon [6], Biju-Duval et al., [13], Durand-Delga and Fontboté [30] and Cohen [18]. This last author even envisioned that this subduction regime ended through a Langhian–Serravallian breakoff of the subducting oceanic slab (his figure 5, p. 292).

Post-collision magmatism is characterised by the exceptional petrologic and geochemical diversity of its products. These usually range from calc-alkaline (low-K to high-K and often shoshonitic) to alkaline (alkali basaltic to basanitic and even ultrapotassic) and are often bimodal, i.e. mafic and acidic [21, 59]. They are thought to derive from various sources, including: • (1) the deep asthenospheric mantle, either subcontinental or suboceanic [80]; • (2) the subcontinental or suboceanic lithospheric mantle, the latter being either depleted (MORB-type source) or metasomatised during previous subduction events by hydrous fluids carrying a sedimentary component [64]; • (3) the upper and/or lower continental crust, which can be involved through anatectic melting [88], and/or crustal contamination of mantle-derived melts [63] coupled with fractional crystallization (AFC) [27], or more complex processes, (e.g. melting, assimilation, storage and homogeneization, MASH [46]); • (4) subducted slivers of oceanic crust, which can melt to generate adakitic magmas [58]. The contributions of these various sources and processes to collision zone magmatism usually change through time. A temporal transition is commonly observed from ‘orogenic’ calc-alkaline magmas variably contaminated by continental crust towards continental intraplate alkaline or ultrapotassic associations, e.g. in Western Turkey and Anatolia [63], Tibet [21, 59] or the Permian magmatism of Corsica [15] and Sardinia [19]. Such a magmatic evolution occurred during the Neogene in several areas of the Western Mediterranean Basin: Sardinia [20]; Southern Spain [88]; CentralSouthern Italy [64]; North Africa [31, 32, 52]. In North Africa, Miocene to Quaternary magmatic centres are scattered along the Mediterranean coasts of Morocco, Algeria, and Tunisia (the Maghreb). They form a 1 200 km-long but narrow (ca. 50 km-wide) magmatic belt running parallel to the coast from La Galite Island and the Mogods massif in Tunisia to Ras Tarf in Morocco (figure 1). This volcanic chain is referred to hereafter as the Maghreb Margin (MM) belt. It includes two main types of magmatism: (1) orogenic calcalkaline to shoshonitic volcanic and plutonic associations, the ages of which decrease from Eastern Algeria

162

R.C. Maury et al. / C. R. Acad. Sci. Paris, Sciences de la Terre et des planètes / Earth and Planetary Sciences 331 (2000) 159–173

Figure 1. Summary of the magmatic and chronological features of the MM. The range of available K–Ar ages and Sr–Nd ratios (not calculated back to emplacement ages) are given for each area together with petrologic indications for the dominant rock type based on Peccerillo and Taylor’s [65] classification of calc-alkaline series. LK: low-K; MK: medium-K; HK: high-K; SH: shoshonitic; KR: K-rich alkaline. B: basalts; Gb: gabbros; A: andesites; D: dacites and rhyodacites; G: granodiorites, granites and monzogranites (and their microgranular varieties). ALB: alkali basalts; BAS: basanites; TALB: transitional alkali basalts; TSHA: transitional shoshonitic andesites (shoshonites) [2, 42, 48, 50, 57, 66, 72, 79]. Figure 1. Résumé des caractéristiques des roches magmatiques de la MM. Pour chaque massif ou zone sont indiquées les fourchettes d’âge K–Ar et des rapports isotopiques mesurés de Sr et Nd, ainsi que des indications sur le type de roche dominant, défini suivant la classification de Peccerillo et Taylor [65].

This hypothesis received a strong support from tomographic studies by Spakman [75], showing the presence beneath the Western Mediterranean of a large body of oceanic lithosphere dipping northwards. These results have been confirmed and extended by more recent data [16, 17, 25, 76, 86] and are in agreement with tectonic/ geodynamic models of the western Mediterranean [38, 53, 68].

In the meantime, Davies and von Blanckenburg [23, 81] integrated the complex features of post-collision magmatism within a general model of collision-related processes involving slab detachment (or breakoff, or failure). In their model, as the subducted oceanic plate breaks off, the underlying asthenosphere rises into the lithosphere break and impinges at the base of the thickened lithosphere of the overlying plate, resulting: (i) first

163

R.C. Maury et al. / C. R. Acad. Sci. Paris, Sciences de la Terre et des planètes / Earth and Planetary Sciences 331 (2000) 159–173

Figure 2. Primitive mantle-normalised multi-element patterns of representative MM magmatic rocks. Numbers as in figure 1, letters denote the type. C: calcalkaline; T: transitional; A: alkaline. Normalisation values from [77]. Figure 2. Spectres multiélémentaires normalisés au manteau primitif [77] de roches magmatiques de la MM. Les numéros sont identiques à ceux de la figure 1. C : calco-alcalin ; T : transitionnel ; A : alcalin.

in a heat supply which can induce melting of this lithospheric mantle metasomatised during the previous subduction period; the ascending mafic magmas can interact with the continental crust and cause its melting, (ii) in lithospheric thinning through thermal erosion. Finally, in the latter stages, melting of the uprising asthenosphere becomes prominent, generating mafic magmas similar to plume-related basalts. Obviously such a scenario might account for the sequence of magmatic events observed in the MM. Carminati et al. [16] showed that the Early Langhian (16–15 Ma) geological features of the MM fit the five major characteristics of a slab breakoff process [23]: (i) ‘bimodal magmatism with basaltic mantle melts and granitoids formed by lower crustal melting, with a mantle parentage’ — a rather good summary of the magmatic features described below; (ii) rapid uplift and exhumation of high pressure metamorphic rocks (e.g. in the Edough massif [60]); (iii) regional metamorphism, hydrothermalism and mineralization [55, 56] reflecting increasing heat and fluid flows, in response to tearing and detachment of the slab [24], (iv) extensional tectonics [1] and finally (v) clastic sedimentation in intramontane basins (e.g., in Grande Kabylie, the Miocene TiziOuzou, Dellys and Thenia basins [54]). The purpose of this paper is to test this model using field, petrological, geochemical and geochronological data collected on the magmatic rocks of the MM, in order to assess the temporal evolution of the mantle and crustal sources involved in the genesis of these magmas.

2. Neogene magmatism from the MM: distribution, age and petrology 2.1. Petrologic types

The locations, ages and ranges of Sr and Nd isotopic ratios of MM magmatic centres are shown in figure 1

164

(see [67] for additional informations, e.g. La/Nb ratios and CIPW normative parameters). Three types of magmatic associations have been distinguished, as previously proposed by Louni-Hacini et al. [52], Piqué et al. [67], El Azzouzi et al. [31] and Coulon et al. [22]; related representative incompatible multi-element plots are shown in figure 2. Calc-alkaline (C) magmas (lato sensu) show the usual chemical characteristics of ‘orogenic’ magmas, i.e. porphyritic character (15–30 % phenocrysts, typically plagioclase + orthopyroxene + clinopyroxene + Fe–Ti oxides ± amphibole ± biotite), negative anomalies in High Field Strength Elements (HFSE) reflected by La/Nb ratios higher than 2, and a ‘crustal/sedimentary’ isotopic Sr–Nd imprint. They range in composition from low-K (LK) to shoshonitic (SH) and potassium-rich (KR) through medium-K (MK) and high-K (HK) calc-alkaline compositions. They are mostly represented by andesites (A), dacites and rhyodacites (D), granodiorites, granites and monzogranites (G), with less common basalts (B) and gabbros (Gb). Acidic types are either metaluminous (areas 8 to 16) or strongly peraluminous and, in this latter case, generally cordierite-bearing (areas 1 to 7). The alkali basaltic (A) association is mostly represented by alkali basalts (ALB) and basanites (BAS) which show strong mineralogical and chemical affinities with Ocean Island Basalts (OIB) and intraplate alkali basalts and basanites from the Central-Western European Neogene Province [49, 83]. These basalts are silicaundersaturated (they commonly contain modal analcite ± nepheline) and exhibit positive HFSE anomalies (e.g. La/Nb ratio lower than unity) like those often observed in HIMU-type OIB basalts. Their isotopic signature is typical of an enriched mantle source. The third group is mainly represented by basalts which display mineralogical and chemical features transitional (T) between those of the two former types. Their mineralogy and their slightly silica-undersaturated major

R.C. Maury et al. / C. R. Acad. Sci. Paris, Sciences de la Terre et des planètes / Earth and Planetary Sciences 331 (2000) 159–173

element compositions are generally close to those of alkali basalts (TALB), but they differ from the latter by their weak negative Nb anomalies (La/Nb ratios ranging from 1 to 2) and their isotopic signatures which are more radiogenic in Sr and less radiogenic in Nd. These isotopic signatures suggest a small but significant crustal/sedimentary imprint. With respect to the calcalkaline group, the transitional association displays less marked enrichments in Rb, Ba, Th, K and is devoid of negative Ti anomalies (figure 2). The occurrence of these transitional basalts is well documented for Oranie and Guilliz. Limited chemical data on Tunisian basalts [40] suggest they are also present in the Nefza and possibly the Mogods areas. In the Gourougou volcano (Morocco), however, lavas with transitional La/Nb ratios and Sr–Nd isotopic signatures are petrographically closer to shoshonites (TSHA). The chemistry of the calcic clinopyroxene phenocrysts reflects the differences between the three groups. In the Ti vs. Ca + Na diagram (not shown) of Leterrier et al. [51], calcic clinopyroxenes of the calc-alkaline and alkaline lavas plot into two distinct compositional domains, while those from the transitional basalts straddle the boundary between them.

There is a clear temporal pattern for the western migration of calc-alkaline magmatism from the Algiers area (area 10) towards Oranie and Morocco where it starts at 13–12 Ma and ends at 7 or even 5 Ma. It is worth noting that, west of Cherchell (area 11), this calc-alkaline activity is purely extrusive (lavas, pyroclastic flows and associated domes). Typical alkali basaltic volcanism is only documented at the two extremities of the MM (areas 2 and 12–15, figure 1). In a given area, this volcanism always postdates the calc-alkaline magmatism. It emplaced lava flows and maar-type hydromagmatic breccias, which either lay over the African basement (South of the Tell and Rif nappes; areas 13, 15), crosscut the nappes (areas 12, 14) or are deformed with them in Tunisia (area 2). The alkaline basaltic volcanism of the western MM is also part of a volcanic lineament which fits with the SW–NE Transmoroccan fault system [67] and which also includes the comparatively larger Moyen Atlas volcanic chain [31]. Transitional volcanic activity also occurs at the two extremities of the MM, where it is mostly (areas 12, 14) or partly (area 3) concomitant with the calc-alkaline events. 2.3. Sr–Nd isotopic features

2.2. Spatial and temporal patterns

Neogene calc-alkaline activity started at 16 Ma in Central and Eastern Algeria (areas 4 to 10 in figure 1). Reworked rhyolitic tuffs and ash layers interbedded within faunistically-dated Uppermost Aquitanian to Lower Burdigalian marls in Grande Kabylie [71] have given a somewhat older 40K–40Ar age of 19.1 ± 1 Ma. However, given the alteration of these ashes, the comparison with any of the younger magmatic rocks from the MM is not reliable. The calc-alkaline acidic magmas emplaced between 16 and 15 Ma usually occur as shallow intrusives, often cordierite-bearing, which either crosscut the basement of the Tell chain (Grande Kabylie, Kabylie de Collo, Edough) or their external zones, i.e. the Tell nappes (areas 5, 8, 9). They often display textures transitional between granular and microgranular or even microlitic, and are commonly associated with rhyolitic dikes or even pyroclastic flow deposits (area 10). Associated mafic magmas, either gabbros or basaltic flows, hydromagmatic breccias and intrusions are volumetrically very minor. In their main occurrences, i.e. Dellys and Cap Djinet (area 10), low-K submarine flows and hyaloclastites are intercalated within Late Burdigalian molasses and marls; this suggests that the emplacement of the mafic magmas may pre-date that of the volumetrically dominant Langhian acidic types. Magmatic calc-alkaline activity stops in Eastern Algeria at ca. 15 Ma, and migrates with time towards both east and west. In Tunisia, it ends only at 10 Ma (area 1) and 8 Ma (area 3) with the emplacement of cordieritebearing dacitic domes and microdioritic intrusions.

Patterns of isotopic variations are illustrated in figure 3. The Filfila granite, not shown in these diagrams, differs from all the other MM magmas by very high and variable 87Sr/86Sr ratios (> 0.730; figure 1). This feature, together with its strongly fractionated Heavy Rare Earth Element (HREE) pattern and trace element heterogeneity, is consistent with an origin through crustal anatexis; it constitutes the sole example of this process in the MM although its very evolved character (this massif corresponds to a Li-mica and topaz-bearing granitic cupola) obscures the primary chemical characteristics [34, 35, 61]. The three magmatic groups display contrasted isotopic signatures in the age vs. 87Sr/86Sr (figure 3a) and the 143 Nd/144Nd vs. 87Sr/86Sr (figure 3b) plots. Alkali basalts exhibit the lowest 87Sr/86Sr ratios (0.703 2–0.704 9) and the highest but quite variable 143Nd/144Nd ratios (0.512 5–0.512 9) which, calculated back to emplacement ages, correspond to positive eNd values (+2.1 to +5.5). These features are grossly similar to those of the Tamazert lamprophyres [11], and to those of the subcontinental mantle rocks exposed in the Ronda and Beni Bousera massifs [49]. Calc-alkaline magmas display an obvious crustal/ sedimentary imprint, with highly variable radiogenic 87 Sr/86Sr ratios (0.705 8–0.722 0) and low 143Nd/144Nd ratios (0.512 1–0.512 5). The latter, calculated back to emplacement ages, correspond to negative εNd values (–1.8 to –10.5). The Sr isotopic ratios are rather variable from one centre to another and also within a given plutonic massif (e.g. areas 6 and 7). They also vary with petrographic type: for instance, the cordierite-bearing

165

R.C. Maury et al. / C. R. Acad. Sci. Paris, Sciences de la Terre et des planètes / Earth and Planetary Sciences 331 (2000) 159–173

Figure 3. Correlations between Sr isotopic ratios and ages (a) and between Sr and Nd isotopic ratios (b) for the MM magmatic rocks. Numbers refer to the areas and letters to the types as in figures 1 and 2. 3Ca: cordierite-free rocks from Nefza; 3Cb: cordieritebearing rocks from Nefza. Figure 3. Corrélations entre rapports isotopiques du Sr et âges (a) et entre rapports isotopiques du Sr et du Nd (b) pour les roches magmatiques de la MM. Conventions identiques à celles de la figure 2.

lavas from the Nefza (3Cb) are much more radiogenic than the cordierite-free ones (3Ca). These features suggest that the contribution of the crustal/sedimentary component was highly variable both in time and space. Transitional lavas are characterised by Sr and Nd ratios in-between those of the two former groups (87Sr/ 86 Sr =0.704 4 – 0.707 8 and slightly negative or positive εNd(T) values of –4.4 to +1.3), indicative of a variable but significant crustal/sedimentary imprint. Whatever their geographic location, they form a rather coherent group (figure 3). They were mostly emplaced between 10 and 7 Ma, i.e. at the end of the calc-alkaline cycle and prior to the emplacement of alkali basalts (with the exception of Guilliz). Thus, it appears that the key period for geochemical changes (implying the contribution of new sources) was also 10–7 Ma, with a sudden drop in the 87Sr/86Sr ratio which is correlated with a drop in the corresponding La/Nb ratio [22, 67].

166

3. A model for the geochemical evolution of the MM 3.1. Parental calc-alkaline magmas and crustal contamination

The highly variable crustal imprint of the calc-alkaline group may reflect a variety of magmatic processes including: (i) anatectic melting of metasediments from the MM continental crust, (ii) contamination and/or AFC of mantle-derived mafic magmas by the upper or lower MM continental crust, (iii) melting of mantle previously metasomatised by fluids or magmas carrying a crustal imprint that was inherited from either continent-derived detrital or pelagic subducted sediments, or finally (iv) any combination of the former processes, e.g. mixing of crust- and mantle-derived magmas.

R.C. Maury et al. / C. R. Acad. Sci. Paris, Sciences de la Terre et des planètes / Earth and Planetary Sciences 331 (2000) 159–173

The mafic lavas from Algiers area and the gabbros from the Cap de Fer [33] carry a typical ‘orogenic’ major, trace element and isotopic signature. Some of them are K-poor but relatively rich in Mg and compatible transition elements (e.g. sample DLDR from Dellys: MgO = 9%; Cr = 214 ppm; Co = 41 ppm; Ni = 94 ppm [5]), and therefore they can be classified as low-K calcalkaline basalts (= island arc tholeiites ; see pattern 10 C in figure 2) similar to those typical of intraoceanic active island arcs. Their most likely source is a depleted mantle previously metasomatised, in a subduction setting, by hydrous fluids carrying a crustal/sedimentary component [3, 78]. The relatively low δ18O (+7 to +7.8 per mil) of liquids in equilibrium with clinopyroxenes from the Oranie andesites or from the Cap de Fer gabbros [22] are readily explained by minor contamination processes affecting mafic magmas derived from the melting of depleted mantle plus a few per cent sediment. The only anatectic acidic bodies identified in the MM correspond to the cordierite-bearing rhyodacites from the Nefza (volumetrically minor) [40] and to the small cordierite-bearing Filfila granitic massif. In most of the other intermediate to acidic MM magmas, the crustal imprint is so strong that their initial mantle signature is overprinted. Their ultimate derivation from a mantle source can nevertheless be invoked in these cases, taking into account the relatively mafic character of the dioritic/andesitic metaluminous rocks (e.g. in area 9; [73]). An additional and substantial process of crustal contamination or AFC is suggested by their variable (usually increasing with SiO2) LILE/HFSE and 87Sr/86Sr ratios, as well as by the common occurrence of partly assimilated crustal xenoliths. Although the crustal imprint is overwhelming [61] in the peraluminous cordierite-bearing granitoids from Eastern Algeria (areas 4, 6, 7), their mineralogical, geochemical (major and trace elements) and isotopic (O, H, Sr, Nd) study led Fourcade et al. [34, 35] to conclude in favor of their derivation through melting of an ‘orogenic’ mantle, followed by massive mixing with crustal melts and assimilation of metapelitic country-rocks. A similar interpretation has been proposed by Benito et al. [10] for the high-K calc-alkaline rocks from SE Spain. Quantitative modelling of these contamination processes is difficult given the scarcity of mafic end-members and also the probably highly variable composition of the crustal components. For instance, the O–Sr isotopic features of Bejaïa–Amizour granitoids are best explained by contamination of mafic mantle-derived magmas by lower continental crust [22]. The most enigmatic feature of the MM calc-alkaline magmas is their large variability in potassium contents (and in the most incompatible trace elements, e.g. Rb, Ba and Th, figure 2). This is partly due to hydrothermal alteration effects which have been considerably underestimated in previous studies, especially in the plutonic rocks [35]. However, there is no easy explanation to

account for the bulk range of K2O contents, which often change abruptly in space and time at equivalent differentiation levels [41]. Hildebrand and Bowring [45] have proposed that slab detachment leads to large quantities of crustal material being recycled into the mantle. Such direct deep crustal recycling should increase more efficiently the incompatible element content of the lithospheric mantle than fluid-induced metasomatism which operates beneath most arcs. 3.2. Sources of alkali basalts

The geochemical signature of the MM alkali basalts and basanites (relatively unradiogenic Sr and positive εNd, without any noticeable crustal imprint) strongly supports an origin from an enriched asthenospheric mantle. Although no isotopic Pb compositions are presently available, their very low La/Nb ratios (0.5 to 0.8) suggest they are HIMU-type OIB basalts [82]. Their asthenospheric source is probably similar to that documented by the Sr, Nd and Pb isotopic study of Cenozoic volcanic rocks from the Western Mediterranean and Central European Volcanic Provinces [47]. Using tomographic and geochemical results, these authors have concluded to the existence of a large sheet-like region of upwelling sublithospheric mantle domain extending beneath Western Europe, the Western Mediterranean and the Eastern Atlantic margins. This asthenospheric reservoir is characterised by 87Sr/86Sr = 0.703 0 to 143 0.703 4; Nd/144Nd = 0.512 82 to 0.512 94 (εNd = +3.5 to +5.9); 206Pb/204Pb = 19.9 to 20.1; 207Pb/ 204 Pb = 15.62 to 15.68; 208Pb/204Pb = 39.60 to 39.90. The hypothesis of an asthenospheric upwelling beneath western North Africa is corroborated by the existence of a NNE-trending volcanic alkaline belt running from the Cape Verde islands to Central Europe. This volcanic belt is interpreted as originated from a NNE-directed sublithospheric plume, channeled and disrupted as a consequence of the asymmetric opening of the Central Atlantic. The composition of this asthenospheric source is close to the HIMU-type mantle [62, 83, 85]. However, this hypothesis does not fully account for the unusually wide range of εNd found in the MM alkali basalts and basanites. Isotopic studies of the Ronda and Beni Bousera ultramafic massifs [49] conclude that they represent a highly heterogeneous subcontinental lithospheric mantle (exhumated at ca. 22 Ma), the isotopic imprint of which being characterised by rather constant Sr but highly variable Nd. Therefore, the presence of such an imprint in the MM alkali lavas may result from the chemical interaction of asthenospheric magmas en route to the surface with an heterogeneous lithospheric mantle [31]. A similar model was proposed by BernardGriffiths et al. [12] for the Ormonde-Sierra de Monchique alkaline magmatism. 3.3. Origin of transitional lavas

Several processes may account for the intermediate chemical features of this group: (i) magma mixing

167

R.C. Maury et al. / C. R. Acad. Sci. Paris, Sciences de la Terre et des planètes / Earth and Planetary Sciences 331 (2000) 159–173

between calc-alkaline and alkaline liquids, (ii) contamination of ascending asthenospheric liquids by the continental crust or the lithospheric mantle, (iii) melting of a lithospheric ‘orogenic’ mantle modified through injection of asthenosphere-derived alkali basaltic liquids, and finally (iv) melting of a transitional zone at the boundary between an asthenospheric mantle and a lithospheric subduction-modified mantle. Hypothesis (ii) has been favoured for Tunisia [40] and Morocco [31] but, in both cases, the number of transitional samples was small and the lack of detailed isotopic studies prevented any quantitative modelling. By contrast, in Oranie, 8 samples of transitional basalts have been dated [52] and studied for Sr and Nd isotopes [22]. The combined modelling of the La/Nb ratios and Sr or Nd isotopic compositions allows to discard hypotheses (i), (ii) and (iii). It shows that the isotopic characteristics of the transitional lavas fit with an origin through partial melting of a mantle zone encompassing the frontier between asthenosphere and lithosphere (hypothesis iv), with a 10–35 % contribution of the asthenospheric reservoir. The temporal magmatic evolution is consistent with this hypothesis [22], because the less refractory reservoir is the lithospheric mantle (owing to its hydrated subduction-related component). In the context of an asthenospheric thermal pulse, this reservoir would be the first to melt, and the melting zone might then develop into the transition zone between lithosphere and asthenosphere. Melting of the sole ‘dry’ asthenospheric reservoir would require higher temperature and/or lower pressure conditions; these could be realised later, in the event of an asthenospheric upwelling, with the formation of alkaline basalts. This scenario implies the progressive development of a thermal anomaly at the lithosphere–asthenosphere boundary, the origin of which is discussed below.

4. Tectonic setting of the MM: the slab breakoff model 4.1. Temporal contributions of magmatic sources

From 16 to 10 Ma (Langhian and Serravallian), calcalkaline magmas resulted from the melting of a subduction-modified ‘orogenic’ lithospheric mantle followed by extensive contamination of these melts by the lower and uppercontinental crust (AFC, MASH). The calc-alkaline activity started at the Burdigalian–Langhian boundary with the eruption of mafic low-K lavas (arc tholeiites) in the Algiers area and extended during the Langhian (16–14.5 Ma) to Central-Eastern Algeria areas 10 to 4) where it mostly induced the emplacement of granitoids and acidic lavas displaying a strong crustal imprint. Then, the calc-alkaline magmatism stopped in this region but propagated eastwards to Tunisia, again with a strong crustal signature (cordierite-bearing rhyodacites) and westwards to Western Algeria and Morocco

168

during the Serravallian and the Early Tortonian. The magmas emplaced during this 16–10 Ma period were globally much more potassic than the average andesite from active arcs. A major change occurred during the Tortonian (10–7 Ma): magmatic activity was then restricted to the two extremities of the MM. Only minor amounts of calc-alkaline rocks were emplaced in areas 3 and 12. Transitional magmas (areas 3, 12, 14) resulted (at least in the case of Oranie) from the melting of a zone encompassing the asthenosphere–lithosphere boundary. The contribution of the deep OIB-type asthenospheric mantle increased during this period, whereas that of the subduction-modified lithospheric mantle vanished progressively through time. The first typical alkali basalts originating from melting of the sole asthenospheric source were emplaced at the end of this period in Tunisia (Mogods) and Morocco (Guilliz). Finally, the Messinian, Pliocene and Early Pleistocene were characterised by the nearly exclusive contribution of the deep OIB-type mantle, with additional interactions of the corresponding melts with subcontinental mantle. The magmatic activity stopped in Tunisia at the Miocene–Pliocene limit, but continued in Morocco along a SW–NE lineament oblique to the MM. The very last contributions of the subduction-modified mantle are evidenced by the emplacement of shoshonites in the Gourougou at 5.4 Ma and transitional lavas in the Guilliz at 2.2 Ma. 4.2. Magmatic arguments for a slab detachment model

Any proposed geodynamic model must take into account the fact that the MM magmatic chain differs from typical subduction-related arcs not only by its unusually K-rich character and strong crustal signature, but also by the very small amounts of magma emplaced during its whole history. A very rough estimation of the volumes of the presently exposed magmatic units of the MM is ca. 2 000 km3 for the calc-alkaline magmas and less than 200 km3 for the transitional and alkali basaltic lavas. For comparison, the volumes of Plio-Quaternary calc-alkaline magmatic rocks in Java Island (Sunda arc, Indonesia), the length of which is similar to that of the MM, can be estimated to a minimum of 50 000 km3 [74]. Obviously, a mechanism more complex and much less efficient from a magmatic point of view than oceanic subduction is needed to explain both the small magma production and the involvement of contrasted chemical reservoirs (e.g., the temporally decreasing contribution of a subduction-modified lithospheric mantle on behalf of an OIB-type asthenospheric source). The main effect of the detachment of a lithospheric slab is that the asthenosphere underlying the downgoing plate flows up into the widening gap above the sinking slab [23, 39, 81]. Input of heat from the upwelling asthenosphere could induce partial melting of the overlying lithospheric mantle previously metasomatised dur-

R.C. Maury et al. / C. R. Acad. Sci. Paris, Sciences de la Terre et des planètes / Earth and Planetary Sciences 331 (2000) 159–173

ing subduction [23, 87]. This enriched ‘orogenic’ mantle displaying a typical subduction-related geochemical imprint (and easily melted because it still contains metasomatic water-bearing minerals) is the most likely source of the calc-alkaline magmatic rocks of the MM, and its signature is prominent in the earlyemplaced mafic rocks (areas 10 and 4). Then, the heat supply from the uprising asthenosphere can affect the continental crust and yields local crustal melting (area 5); more generally, it facilitates the involvement of a crustal component within mantle-derived melts through extensive contamination, AFC and/or MASH processes (areas 4, 6, 7, 8, 9, 10). Since the heat source, due to the asthenospheric uprise, follows the tear of the breakoff, the resulting magmatism is expected to form a narrow and elongated belt like the MM. 4.3. Present-day mantle structures

First tomographic data led to propose the existence of a 100–500 km-deep northward-dipping subducted slab of oceanic lithosphere beneath the Southwestern Mediterranean [75]. However, later results were ambiguous regarding the existence of such a subducted lithosphere below the MM due to poor spatial resolution [76]. Considerable refinements of the method led to a more complex picture described by Carminati et al. [16]. These authors published three north–south tomographic crosssections of the MM (their figure 7, p.658) at the levels of Oran (area 12), Filfila (area 5) and East of Tunis, respectively. The latter, located at the level of the Pantelleria Trough, is too far away from areas 1 and 2 to be considered in the present discussion. The Eastern Algeria profile shows a large positive speed anomaly dipping quite regularly northwards at depths from ca. 200 km below the MM down to 500 km southeast of Sardinia. This anomaly corresponds to the tomographic signature of a subducted oceanic lithosphere. The lack of any negative anomaly which could be explained by the presence of a hot mantle beneath the MM at this level is in agreement with the absence of Plio-Quaternary magmatic rocks in Eastern Algeria. The Oranie profile shows a strikingly different situation. A strong positive speed anomaly is observed at depths between 200 and 700 km beneath the Betic chain and the Alboran basin: it has been interpreted as a detached and vertically dipping slab of oceanic lithosphere [16]. In turn, a clear negative anomaly occurs at shallow depths (< 300 km) below the volcanic region of Oranie (area 12), a feature which obviously fits with the occurrence of Late Pliocene–Quaternary (3.9 to 0.8 Ma-old) alkali basalts in this sector. 4.4. Cenozoic tectono-magmatic evolution of the MM

The Cretaceous African margin was bordered by continental blocks (the future Rif and Kabylies) and deep, at least partly oceanic, basins [29] in which Alpine flyschs were deposited (stage 1, figure 4). During the Late Cre-

taceous, the whole system was affected by a north– south convergence (stage 2, figure 4) the rate of which increased during the Eocene. This convergence was accompanied by a north-dipping subduction which resulted in the metasomatism of the lithospheric mantle underlying the Rif and the Kabylies. It ended by the collision of the African margin with these massifs and the emersion of the future MM during the Late Eocene (stage 3). It was followed by the closure of the flysch basins and the related emplacement of the Rif and Tell nappes during the Miocene. This collision period was followed by post-collisional distension in the MM between 28 and 11 Ma [1]. There is no obvious geological indication of slab breakoff until the Langhian (16 Ma). We postulate that the lower part of the African plate, anchored in the mantle as suggested by tomographic data, was teared apart from its upper part which was still moving northwards with respect to Europe during the Early Miocene (stage 4, figure 4). The only explanation for the temporal migration of the calc-alkaline activity towards east and west is that the slab detachment occurred first below Eastern and Central Algeria (stage 5, figure 4) and then propagated laterally towards Tunisia and Oranie/ Morocco during the Serravallian (stage 6, figure 4). This hypothesis is in agreement with the model of a progressively developing Algerian tectonic indenter [67]. From this indenter, the western and eastern edges are assumed to have experienced lateral escape along, respectively, NE–SW and NW–SE transcurrent faults. Then magmatic activity stopped in the central part of the MM (stage 7, figure 4). We assume that the deep detached slab sank at the two edges of the MM, and there the rising asthenosphere started to melt when it reached shallow depths. In Oranie (area 12), the Tortonian (10–7 Ma) transitional lavas document a process during which the melting zone straddled the uprising lithosphere–asthenosphere limit [22]. The most easily melted source was still the subduction-modified hydrated lithospheric mantle, but the asthenospheric contribution was already significant (up to 35 %). Finally, the subduction-modified mantle imprint vanished during the Messinian (area 14), owing to the progressive thinning of the lithosphere through thermal erosion. During the calk-alkaline and transitional magmatic activities, the regional shortening was oriented NNE–SSW; then, it became north–south during the Messinian, activating transcurrent megashear zones and causing crustal stretching [67]. The alkaline volcanism which is observed at the two extremities of the MM is clearly associated with the second tectonic period, which marked the latest stages of the nearly north–south convergence between Africa and Europe. PlioQuaternary alkali basalts and basanites of the MM (areas 2, 12, 13, 14, 15) originated from the melting of the deep asthenospheric source. The geochemical characteristics of these basalts suggest that this source was inherited from the OIB-type asthenospheric mantle

169

R.C. Maury et al. / C. R. Acad. Sci. Paris, Sciences de la Terre et des planètes / Earth and Planetary Sciences 331 (2000) 159–173

Figure 4. Cartoons depicting the tectono-magmatic evolution of the central part of the MM (Algiers area). 1: calc-alkaline magmatism; 2: deformed crustal materials (enhanced scale); 3: European continental crust; 4: Kabylian continental crust; 5: African continental crust; 6: metasomatised lithospheric mantle; 7: lithospheric mantle; 8: asthenosphere; 9: asthenospheric flow. See text for explanations. Figure 4. Coupes très schématiques montrant l’évolution tectonomagmatique de la partie centrale de la MM (Algérois). 1 : magmatisme calcoalcalin ; 2 : matériaux crustaux déformés (échelle dilatée) ; 3 : croûte continentale européenne ; 4 : croûte continentale kabyle ; 5 : croûte continentale africaine ; 6 : manteau lithosphérique métasomatisé ; 7 : manteau lithosphérique ; 8 : asthénosphère ; 9 : flux asthénosphérique. Explications dans le texte.

plume extending from the Eastern Atlantic to Northern Africa, which is supposed to have been dismembered and channeled during the opening of the Central Atlantic [62].

5. Conclusions Geological and geophysical features of the MM, as well as the observed magmatic changes through time, lead us to propose lithospheric failure (slab breakoff) as a clue to account for the recent (Neogene to PlioQuaternary) geodynamic evolution of this continental

170

margin. The slab detachment model explains most of the unusual features of the Neogene magmatism of the Mediterranean Maghreb Margin, which have been considered as rather enigmatic during the last 30 years, despite a number of regional studies. 1. The MM magmatism post-dated the end of a suspected north-dipping subduction period and the collision between the Kabylies and Africa by at least 20 Ma. Such a time span fits with a slab detachment process. Magmatic activity started when the thermal anomaly due to uprise of the asthenosphere through the tear in the slab reached the overlying subduction-modified and hydrated lithospheric mantle, triggering its melting. The

R.C. Maury et al. / C. R. Acad. Sci. Paris, Sciences de la Terre et des planètes / Earth and Planetary Sciences 331 (2000) 159–173

fact that the MM magmatic chain is remarkably linear overmore than 1 000 km is in agreement with such a process. 2. The Langhian to Serravallian rocks of this postcollisional MM magmatism display the typical petrological and geochemical characteristics of subductionrelated calc-alkaline magmas. Strongly peraluminous granitoids are also generated at that period but their origin (mostly crustal) is linked to the thermal anomalies associated with this calc-alkaline magmatism. This reflects the geochemical ‘subduction imprint’ acquired by the lithospheric mantle overlying the tear during a previous (Upper Cretaceous–Eocene) subduction event. However, it is worth noting that the volumes of these calc-alkaline rocks are very small compared to most active magmatic arcs (ca 2 km3 per km of chain, i.e. 25 times less than in the Sunda arc). Compared to active oceanic subduction, during which water and incompatible elements are continuously supplied from the subducting slab to the arc mantle wedge, slab detachment is a ‘passive’ (and thus less productive) process from a magmatic point of view, because the previously subduction-modified lithospheric mantle is no more ‘fed’ through sinking and dehydration of the slab. 3. The observed migration of the MM magmatism eastwards and westwards from Central/Eastern Algeria might reflect the lateral propagation of the slab tear. 4. The K-rich calc-alkaline magmas of the MM display a strong crustal imprint. In the slab breakoff model, the thermal anomaly due to the uprising

asthenosphere favours crustal contribution through anatexis, as well as AFC and other contamination-related processes affecting mantle-derived melts. Very high K contents might reflect direct incorporation of crustal slivers into the mantle. 5. Calc-alkaline magmatic activity stopped during the Serravallian in Central and Eastern Algeria. In Tunisia and Oranie/Morocco, it ended later and was accompanied by the eruption of transitional lavas, followed by Plio-Quaternary alkali basalts and basanites. Tomographic data suggest that these features reflect the presence of a hot asthenospheric mantle anomaly beneath Tunisia and Oranie/Morocco. The indenter model provides an explanation for the ascent of alkaline melts along, respectively, NW–SE and NE–SW transcurrent faults in these regions. 6. In Oranie, the magmatic change from calcalkaline to alkaline lavas (through transitional basalts) has been ascribed to the uprise of the lithosphere–asthenosphere boundary, as a consequence of the flowing up of the asthenosphere through the widening tear of a sinking slab. In this area, the transitional basalts originated from a melting zone straddling the lithosphereasthenosphere limit. 7. In Morocco, the youngest alkaline lavas were emplaced within the African foreland and not within the Rif/Tell domains. This may be an effect of the channeling of the asthenospheric upwelling through a preexisting weakness zone of the African lithosphere (the Transmoroccan fault system).

Acknowledgements. Most of the data used in this paper have been obtained within the Algerian-French cooperative program 91MDU175 Magmatisme néogène de la marge nord-algérienne and within the Moroccan-French cooperative program Action intégrée 98-162 Les océans actuels et anciens au Maroc. We are grateful to J. Bernard-Griffiths for his contribution to the isotopic study of volcanic rocks from Oranie and Morocco. We thank L.-E. Ricou for his critical review of the manuscript.

References [1] Aïte M.O., Gélard J.-P., Distension néogène post-collisionnelle sur le transect de Grande Kabylie (Algérie), Bull. Soc. géol. France 168 (4) (1997) 423–436. [2] Andries D., Bellon H., Ages isotopiques 40K–40Ar du volcanisme alcalin néogène d’Oujda (Maroc oriental) et implications tectoniques, Sci. Géol. Mém. Strasbourg 84 (1989) 107–116. [3] Arculus R.J., Aspects of magma genesis in arcs, Lithos 33 (1994) 189–208. [4] Auzende J.-M., Bonnin J., Olivet J.-L., The origin of the Western Mediterranean basin, J. Geol. Soc. London 129 (1973) 607–620. [5] Belanteur O., Bellon H., Maury R.C., Ouabadi A., Coutelle A., Semroud B., Megartsi M., Fourcade S., Le magmatisme miocène de l’Est de l’Algérois : géologie, géochimie et géochronologie 40K–40Ar, C. R. Acad. Sci. Paris, série IIa 321 (1995) 489–496. [6] Bellon H., Séries magmatiques néogènes et quaternaires du pourtour méditerranéen occidental, comparées dans leur cadre géochronométriques. Implications géodynamiques, thèse d’État, université Paris-Sud–Orsay, 1976, 367 p. [7] Bellon H., Chronologie radiométrique K–Ar des manifestations magmatiques autour de la Méditerranée occidentale entre 33 et 1 Ma, in: Wezel F.C. (Ed.), Sedimentary basins of Mediterranean margins, Tecnoprint, Bologne, 1981, pp. 341–360.

[8] Bellon H., Brousse R., Le magmatisme périméditerranéen occidental. Essai de synthèse, Bull. Soc. géol. France XIX (7) (1977) 469–480. [9] Bellon H., Letouzey J., Volcanism related to plate tectonics in the western and eastern Mediterranean, in: XXV Congr. Assoc. CIESM, Split, Technip (Ed.), Paris, 1977, pp. 165–184. [10] Benito R., Lopez-Ruiz J., Cebrià J.M., Hertogen J., Doblas M., Oyarzun R., Demaiffe D., Sr and O isotopic constraints on source and crustal contamination in the high-K calc-alkaline and shoshonitic Neogene volcanic rocks of SE Spain, Lithos 46 (1999) 773–802. [11] Bernard-Griffiths J., Fourcade S., Dupuy C., Isotopic study (Sr, Nd, O and C) of lamprophyres and associated dykes from Tamazert (Morocco): crustal contamination processes and source characteristics, Earth Planet. Sci. Lett. 103 (1991) 190–199. [12] Bernard-Griffiths J., Gruau G., Cornen G., Azambre B., Macé J., Continental lithospheric contribution to alkaline magmatism: isotopic (Nd, Sr, Pb) and geochemical (REE) evidence from Serra de Monchique and Mount Ormonde Complexes, J. Petrol. 38 (1997) 115–132. [13] Biju-Duval B., Letouzey J., Montadert C., Structure and evolution of the Mediterranean basins, in: Hsü K.J., Montadert L. et al., Init. Rep. DSDP, 42, US Gov. Print. Off., Washington DC, 1978, pp.951–984. [14] Boccaletti M., Guazzone G., Gliarchi appenninici, il Mare Ligure ed il Tirreno nel quadro della tettonica dei bacini marginali retroarco, Mem. Soc. Geol. Italiana 11 (1972) 201–216.

171

R.C. Maury et al. / C. R. Acad. Sci. Paris, Sciences de la Terre et des planètes / Earth and Planetary Sciences 331 (2000) 159–173 [15] Cabanis B., Cochemé J.-J., Vellutini P.J., Joron J.-L., Treuil M., Post-collisional Permian volcanism in north-western Corsica: an assessment based on mineralogy and trace-element geochemistry, J. Volcanol. Geoth. Res. 44 (1990) 51–67. [16] Carminati E., Wortel M.J.R., Spakman W., Sabadini R., The role of slab detachment processes in the opening of the western-central Mediterranean basins: some geological and geophysical evidence, Earth Planet. Sci. Lett. 160 (1998) 651–665. [17] Carminati E., Wortel M.J.R., Meijer P.T., Sabadini R., The twostage opening of the western-central Mediterranean basins: a forward modeling test to a new evolutionary model, Earth Planet. Sci. Lett. 160 (1998) 667–679. [18] Cohen C.R., Plate tectonic model for the Oligo-Miocene evolution of the Western Mediterranean, Tectonophysics 68 (1980) 283–311. [19] Cortesogno L., Cassinis G., Dallagiovanna G., Gaggero L., Oggiano G., Ronchi A., Seno S., Vanossi M., The Variscan postcollisional volcanism in Late Carboniferous–Permian sequences of Ligurian Alps, Southern Alps and Sardinia (Italy): a synthesis, Lithos 45 (1998) 305–328. [20] Coulon C., Le volcanisme calco-alcalin cénozoïque de Sardaigne (Italie). Pétrographie, géochimie et genèse des laves andésitiques et des ignimbrites. Signification géodynamique, thèse d’État, université de Marseille, 1977, 385 p. [21] Coulon C., Maluski H., Bollinger C., Wang, S., Mesozoic and Cenozoic volcanic rocks from central and southern Tibet: 39Ar/40Ar dating, petrological characteristics and geodynamic significance, Earth Planet. Sci. Lett. 79 (1986) 281–302. [22] Coulon C., Megartsi M., Fourcade S., Maury R.C., Bellon H., Louni-Hacini A., Cotten J., Hermitte D., The transition from calcalkaline to alkaline volcanism during the Neogene in Oranie (Algeria): magmatic expression of a slab breakoff, Tectonophysics (submitted). [23] Davies J.H., von Blanckenburg F., Slab breakoff: a model of lithosphere detachment and its test in the magmatism and deformation of collisional orogens, Earth Planet. Sci. Lett. 129 (1995) 85–102. [24] De Boorder H., Spakman W., White S.H., Wortel M.J.R., Late Cenozoic mineralization, orogenic collapse and slab detachment in the European Alpine Belt, Earth Planet. Sci. Lett. 164 (1998) 569–575. [25] De Jonge M.R., Wortel M.J.R., Spakman W., Regional scale tectonic evolution and the seismic velocity structure of the lithosphere and upper mantle: the Mediterranean region, J. Geophys. Res. 99 (1994) 12091–12108. [26] De Larouzière F.D., Bolze J., Bordet P., Hernandez J., Montenat C., Ott d’Estevou P., The Betic segment of the lithospheric TransAlboran shear zone during the Late Miocene, Tectonophysics 152 (1988) 41–52. [27] De Paolo D.J., Trace element and isotopic effects of combined wall rock assimilation and fractional crystallization, Earth Planet. Sci. Lett. 53 (1981) 189–202. [28] Dercourt J., L’expansion océanique actuelle et fossile, ses implications géotectoniques, Bull. Soc. géol. France XII (7) (1970) 261–317. [29] Dercourt J., Ricou L.-E., Vrielynck B. (Eds.), Atlas Tethys Palaeoenvironmental maps, CCGM, Paris, 1993. [30] Durand-Delga M., Fontboté J.-M., Le cadre structural de la Méditerranée occidentale, 26th Int. Geol. Congress, Paris, Geology of the Alpine chains born of the Tethys, Mém. BRGM, 115, 1980, pp. 67–85. [31] El Azzouzi M., Bernard-Griffiths J., Bellon H., Maury R.C., Piqué A., Fourcade S., Cotten J., Hernandez J., Évolution des sources du volcanisme marocain au cours du Néogène, C. R. Acad. Sci. Paris, série IIa 329 (1999) 95–102. [32] El Bakkali S., Gourgaud A., Bourdier J.-L., Bellon H., Gundogdu N., Post-collision Neogene volcanism of the Eastern Rif (Morocco): magmatic evolution through time, Lithos 45 (1998) 523–543. [33] Fougnot J., Le magmatisme miocène du littoral nordconstantinois, thèse, INPL Nancy, 1990, 358 p. [34] Fourcade S., Capdevila R., Ouabadi A., Martineau F., Abstracts IVth Int. Conf. on Granites, Clermont-Ferrand, Coeval calc-alkaline metaluminous and peraluminous cordierite-bearing granitoids of Miocene age in Northern Algeria: sources and geodynamic significance, 1999, p. 146. [35] Fourcade S., Capdevila R., Ouabadi A., Martineau F., The origin and geodynamic significance of the Alpine cordierite-bearing granitoids

172

of Northern Algeria. A combined petrological, mineralogical, geochemical and isotopic (O, H, Sr, Nd) study, Lithos (submitted). [36] Girod M., Girod N., Contribution de la pétrologie à la connaissance de l’évolution de la Méditerranée occidentale depuis l’Oligocène, Bull. Soc. géol. France XIX (7) (1977) 481–488. [37] Glangeaud L., Sur les différents modes de gisement des roches intrusives tertiaires du littoral algérien, de Ténès à Djidjelli. Leurs relations avec la tectonique de l’Atlas, Bull. Soc. géol. France IV (5) (1934) 237–249. [38] Gueguen E., Doglioni C., Fernandez M., On the post-25 Ma geodynamic evolution of the western Mediterranean, Tectonophysics 298 (1998) 259–269. [39] Gvirtzman Z., Nur A., Plate detachment, asthenosphere upwelling, and topography across subduction zones, Geology 27 (1999) 563–566. [40] Halloul N., Géologie, pétrologie et géochimie du bimagmatisme néogène de la Tunisie septentrionale (Nefza et Mogod). Implications pétrogénétiques et interprétation géodynamique, thèse, université Blaise-Pascal, Clermont-Ferrand, 1989, 203 p. [41] Hernandez J., Pétrologie du massif volcanique du Guilliz (Maroc oriental); cristallisation fractionnée, mélanges de magmas et transferts de fluides dans une série shoshonitique, J. Afr. Earth Sci. 5 (4) (1986) 381–389. [42] Hernandez J., Bellon H., Chronologie K–Ar du volcanisme miocène du Rif oriental (Maroc): implications tectoniques et magmatologiques, Rev. Géol. Dyn. Géogr. Phys. 26 (1985) 85–94. [43] Hernandez J., de Larouzière F.D., Bolze J., Bordet P., Le magmatisme néogène bético-rifain et le couloir de décrochement transAlboran, Bull. Soc. géol. France (8) III (1987) 257–267. [44] Hernandez J., Lepvrier C., Le volcanisme calco-alcalin miocène de la région d’Alger (Algérie) : pétrologie et signification géodynamique, Bull. Soc. géol. France XXI (7) (1979) 73–86. [45] Hildebrand R.S., Bowring S.A., Crustal recycling by slab failure, Geology 27 (1999) 11–14. [46] Hildreth W., Moorbath S., Crustal contributions to arc magmatism in the Andes of Central Chile, Contrib. Mineral. Petrol. 98 (1988) 455–489. [47] Hoernle K., Zhang Y.-S., Graham D., Seismic and geochemical evidence for large-scale mantle upwelling beneath the eastern Atlantic and western and central Europe, Nature 374 (1995) 34–39. [48] Juteau M., Michard A., Albarède F., The Pb–Sr–Nd isotope geochemistry of some recent circum-Mediterranean granites, Contrib. Mineral. Petrol. 92 (1986) 331–340. [49] Kumar N., Reisberg L., Zindler A., A major and trace element and strontium, neodymium, and osmium isotopic study of a thick pyroxenite layer from the Beni Bousera ultramafic complex of northern Morocco, Geochim. Cosmochim. Acta 60 (8) (1996) 1429–1444. [50] Lepvrier C., Velde D., À propos des intrusions tertiaires de la marge nord-africaine entre Cherchel et Ténès (Algérie), Bull. Soc. géol. France 18 (7) (1976) 991–998. [51] Leterrier J., Maury R.C., Thonon P., Girard M., Marchal M., Clinopyroxene composition as a method of identification of the magmatic affinities of paleo-volcanic series, Earth Planet. Sci. Lett. 59 (1982) 139–154. [52] Louni-Hacini A., Bellon H., Maury R.C., Megartsi M., Coulon C., Semroud B., Cotten J., Coutelle A., Datation 40K–40Ar de la transition du volcanisme calco-alcalin en Oranie au Miocène supérieur, C. R. Acad. Sci. Paris, série IIa 321 (1995) 975–982. [53] Maillard A., Mauffret A., Structure et volcanisme de la fosse de Valence (Méditerranée nord-occidentale), Bull. Soc. géol. France 164 (3) (1993) 365–383. [54] Magné J., Raymond D., Le Néogène post-nappes de la région de Dellys-Tizi Ouzou (Algérie) ; un enregistreur de l’évolution dynamique du NW de la Grande Kabylie après le Burdigalien, Bull. Soc. géol. France XVI (7) (1974) 47–52. [55] Marignac C., Les minéralisations filoniennes d’Aïn Barbar (Algérie). Un exemple d’hydrothermalisme lié à l’activité géothermique alpine en Afrique du Nord, thèse d’État, INPL, Nancy, 1985, 1 163 p. [56] Marignac C., Zimmermann J.-L., Âges K–Ar de l’événement hydrothermal et des intrusions associées dans le district minéralisé miocène d’Aïn Barbar (Est Constantinois, Algérie), Miner. Deposita 18 (1983) 457–467.

R.C. Maury et al. / C. R. Acad. Sci. Paris, Sciences de la Terre et des planètes / Earth and Planetary Sciences 331 (2000) 159–173 [57] Mauduit F., Le volcanisme néogène de la Tunisie continentale, thèse de 3e cycle, université Paris-Sud–Orsay, 1978, 157 p. [58] Maury R.C., Sajona F.G., Pubellier M., Bellon H., Defant M., Fusion de la croûte océanique dans les zones de subduction/collision récentes : l’exemple de Mindanao, Philippines, Bull. Soc. géol. France 167 (1999) 579–595. [59] Miller C., Schuster R., Klotzli U., Frank W., Purtscheller F., Postcollisional potassic and ultrapotassic magmatism in SW Tibet: geochemical and Sr–Nd–Pb–O isotopic constraints for mantle source characteristics and petrogenesis, J. Petrol. 40 (19??) 1399–1424. [60] Monié P., Montigny R., Maluski H., Âge burdigalien de la tectonique ductile extensive dans le massif de l’Edough (Kabylies, Algérie). Données radiométriques 39Ar–40Ar, Bull. Soc. géol. France 5 (1992) 571–584. [61] Ouabadi, A., Pétrologie, géochimie et origine des granitoïdes peralumineux à cordiérite (Cap Bougaroun, Beni-Touffout et Filfila), Algérie nord-orientale, thèse, université Rennes-I, 1994, 257 p. [62] Oyarzun R., Doblas M., Lopez-Ruiz J., Cebria J.M., Opening of the Central Atlantic and asymmetric mantle upwelling phenomena: implications for long-lived magmatism in western North Africa and Europe, Geology 25 (1997) 727–730. [63] Pearce J.A., Bender J.F., De Long S.E., Kidd W.S.F., Low P.J., Güner Y., Saroglu F., Yilmaz Y., Moorbath S., Mitchell J.G., Genesis of collision volcanism in Eastern Anatolia, Turkey, J. Volcanol. Geotherm. Res. 44 (1990) 189–229. [64] Peccerillo A., Multiple mantle metasomatism in centralsouthern Italy: geochemical effects, timing and geodynamic implications, Geology 27 (1999) 315–318. [65] Peccerillo A., Taylor S.R., Geochemistry of Eocene calc-alkaline volcanic rocks from the Kastamonu area, Northern Turkey, Contrib. Mineral. Petrol. 58 (1976) 63–81. [66] Penven M.-J., Zimmermann J.-L., Mise en évidence, par la méthode potassium–argon, d’un âge Langhien pour le plutonisme calco-alcalin de la Kabylie de Collo (Algérie), C. R. Acad. Sci. Paris, série II 303 (1986) 403–406. [67] Piqué A., Aït Brahim L., El Azzouzi M., Maury R.C., Bellon H., Semroud B., Laville E., Le poinçon maghrébin : contraintes structurales et géochimiques, C. R. Acad. Sci. Paris, série IIa 326 (1998) 575–581. [68] Platt J.P., Soto J.-I., Whitehouse M.J., Hurford A.J., Kelley S.P., Thermal evolution, rate of exhumation, and tectonic significance of metamorphic rocks from the floor of the Alboran extensional basin, western Mediterranean, Tectonics 17 (1998) 671–689. [69] Réhault J.-P., Boillot G., Mauffret A., The Western Mediterranean basin geological evolution, Mar. Geol. 55 (1984) 447–477. [70] Ricou L.-E., Dercourt J., Geyssant J., Grandjacquet C., Lepvrier C., Biju-Duval B., Geological constraints on the Alpine evolution of the Mediterranean Tethys, Tectonophysics 123 (1986) 83–122. [71] Rivière M., Bouillin J.-P., Courtois C., Gélard J.-P., Raoult J.F., Étude minéralogique et géochimique des tuffites découvertes dans l’Oligo-Miocène Kabyle (Grande Kabylie, Algérie). Comparaison avec les tuffites de la région de Malaga (Espagne), Bull. Soc. géol. France 5 (1977) 1171–1177.

[72] Robin C., Étude géodynamique du massif volcanique du Cap Cavallo (El Aouana), Algérie, thèse de 3e cycle, université Paris-6, 1970, 152 p. [73] Semroud B., Maury R.C., Ouabadi A., Cotten J., Fourcade S., Fabriès J., Gravelle M., Géochimie des granitoïdes miocènes de BejaiaAmizour (Algérie du Nord), C. R. Acad. Sci. Paris, série II 319 (1994) 95–102. [74] Soeria-Atmadja R., Maury R.C., Bellon H., Pringgoprawiro H., Polvé M., Priadi B., Tertiary magmatic belts in Java, J. Southeast Asia Earth Sci. 9 (1994) 13–27. [75] Spakman W., Subduction beneath Eurasia in connection with the Mesozoic Tethys, Geologie en Mijnbouw 65 (1986) 145–153. [76] Spakman W., van der Lee S., van der Hilst R., Travel-time tomography of the European-Mediterranean mantle down to 1 400 km, Phys. Earth Planet. Inter. 79 (1993) 3–74. [77] Sun S.S., Mc Donough W.F., Chemical and isotopic systematics of oceanic basalts: implications for mantle composition and processes, in: Saunders A.D., Norry M.J. (Eds.), Magmatism in ocean basins, Geol. Soc. London Spec. Publ., 42, 1989, pp. 313–345. [78] Tatsumi Y., Kogiso T., Trace element transport during dehydration processes in the subducted oceanic crust. 2. Origin of chemical and physical characteristics in arc magmatism, Earth Planet. Sci. Lett. 148 (1997) 207–221. [79] Vass D., Bagdasarjan G.P., Bajanik S., Contributions to the Neogene radiometric scale from northern Africa, Geol. Mag. 111 (1974) 149–155. [80] Vollmer R., On the origin of the Italian potassic magmas. 1. A discussion contribution, Chem. Geol. 74 (1989) 229–239. [81] von Blanckenburg F., Davies J.H., Slab breakoff: a model for syncollisional magmatism and tectonics in the Alps, Tectonics 14 (1995) 120–131. [82] Weaver B.L., Trace element evidence for the origin of oceanisland basalts, Geology 19 (1991) 123–126. [83] Wedepohl K.H., Baumann A., Central European Cenozoic plume volcanism with OIB characteristics and indications of a lower mantle source, Contrib. Mineral. Petrol. 136 (1999) 225–239. [84] Wezel F.C., Interpretazione dinamica delle “eugeosinclinale meso-mediterranea”, Rivista Miner. Sicil. 187-198 (1970) 124–126. [85] Wilson M., Downes H., Tertiary–Quaternary extension-related alkaline magmatism in western and central Europe, J. Petrol. 32 (1991) 811–849. [86] Wortel M.J.R., Spakman W., Structure and dynamics of subducted lithosphere in the Mediterranean region, Proc. Kon. Ned. Acad. Wetensch. 95 (1992) 325–347. [87] Zeck H.P., Betic-Rif orogeny: subduction of Mesozoic Tethys lithosphere under eastward drifting Iberia, slab detachment shortly before 22 Ma, and subsequent uplift and extensional tectonics, Tectonophysics 254 (1996) 1–16. [88] Zeck H.P., Kristensen A.B., Williams I.S., Post-collisional volcanism in a sinking slab setting. Crustal anatectic origin of pyroxeneandesite magma, Caldear Volcanic Group, Neogene Alboran volcanic province, southeastern Spain, Lithos 45 (1998) 499–522.

173