Subduction-accretion complex with boninitic ophiolite slices and Triassic limestone seamounts: Ankara Mélange, central Anatolia Open Access

Subduction-accretion complexes constitute a major part of the active continental margin, forming belts hundreds of kilometres wide and more than 1000 km long, as for example in the present-day Makran (e.g. McCall & Kidd, 1982; Burg, 2018). Old subduction-accretion complexes, such as the Jurassic–Palaeogene Franciscan Complex of the western US (e.g. Ernst, 2011; Wakabayashi, 2015; Raymond, 2018) or the Cretaceous to Miocene Shimanto belt in Japan (e.g. Taira et al.1988), make up a significant part of the orogenic belts. They also provide important information on the ages of the former oceanic lithosphere, of oceanic seamounts and plateaus, and of subduction in the old convergent margins.

Subduction-accretion complexes consist of two contrasting rock associations: (a) tectonic slivers of the down-going oceanic crust, which constitute the ophiolitic part, and (b) clastic rocks derived from the erosion of the continental magmatic arc, which constitute the trench turbidites (e.g. Isozaki, 1997; Wakita & Metcalfe, 2005; Kusky et al.2013). In large subduction-accretion complexes, such as the Franciscan, Shimanto or Makran, sandstones form the bulk. In contrast, sandstones are rare in the ophiolitic mélanges in Anatolia, which represent Mesozoic Tethyan subduction-accretion complexes (e.g. Okay et al.1998, 2019; Parlak & Robertson, 2004; Plunder et al.2013; Pandeli et al.2018; Robertson et al. 2021). Here we describe a well-exposed segment of the ophiolitic mélange in the Beynam–Ankara region, depict its evolution and discuss the reason for the lack of the coarse-clastic component. In the Beynam area, the ophiolitic mélange is stratigraphically overlain by Upper Cretaceous fore-arc turbidites, which allows reconstruction of the geometry during the Late Cretaceous subduction-accretion.

The ophiolitic mélange in the Beynam–Ankara region consists of three distinct accretionary units (AUs) in the sense of Raymond et al. (2020), lithologically distinctive, mappable and fault-bounded rock bodies that were formed during subduction accretion and consist of lithologic or tectonostratigraphic units truncated at the AU margin by major faults.

Mesozoic subduction-accretion complexes, generally named as ophiolitic mélange, crop out over large areas along the İzmir–Ankara suture, that marks the boundary between the Pontides and the Anatolide–Tauride Block (Fig. 1). They represent remnants of a Mesozoic Tethyan ocean, which separated Pontides in the north, from the Anatolide–Tauride Block and the Kırşehir Massif in the south (e.g. Şengör & Yılmaz, 1981; Robertson et al.2009; Mueller et al. 2019; Okay et al.2020 a; van Hinsbergen et al.2020). The ophiolitic mélanges in central Anatolia constitute part of the Ankara Mélange, one of the earliest mélanges described (Bailey & McCallien, 1950, 1953). The Ankara Mélange is made up of two distinct subduction-accretion complexes: the Karakaya Complex in the west and the ophiolitic mélange in the east (Figs 1 and 2; e.g. Koçyiğit, 1991; Rojay, 2013). The Karakaya Complex forms a north–south-trending, 24 km wide belt in central Anatolia (Fig. 2). It consists of Permian and Carboniferous limestone blocks in a sheared Upper Triassic greywacke matrix (Upper Karakaya Complex) and a tectonically underlying sequence of metabasite, marble and phyllite (Lower Karakaya Complex). The Karakaya Complex is generally interpreted as a subduction-accretion unit, which was accreted to the southern margin of Laurasia during the Late Triassic (e.g. Robertson & Ustaömer, 2012; Okay & Nikishin, 2015). Subsequently the Karakaya Complex is unconformably overlain by Lower to Middle Jurassic sandstones and shales, which pass up into Upper Jurassic – Lower Cretaceous limestones, erosional remnants of which are preserved in a few places in the Ankara region (Figs 2 and 3; Bilgütay, 1960; Koçyiğit, 1989; Dönmez et al. 2008).

East of Ankara the Karakaya Complex is thrust eastwards over the ophiolitic mélange (Fig. 2; e.g. Rojay, 2013). The ophiolitic mélange consists of basalt, serpentinite, radiolarian chert and lesser amounts of pelagic shale and limestone. It forms a 13 km wide belt, and is thrust eastwards over the Upper Cretaceous turbidites, which are in turn thrust eastwards onto Palaeocene and Eocene clastic rocks (Fig. 2; Norman, 1972). Eocene sandstones and limestones also lie unconformably over the crystalline rocks of the Kırşehir Massif and constitute the oldest post-collisional marine sedimentary cover over the İzmir–Ankara suture (e.g. Gülyüz et al.2013; Okay et al.2020 b).

The Kırşehir Massif consists of gneiss, marble and amphibolite with Late Cretaceous metamorphic ages (91–83 Ma) and dismembered ophiolites, which were intruded by Late Cretaceous granites to gabbros (85–65 Ma; Whitney & Hamilton, 2004; Lefebvre et al.2011; van Hinsbergen et al.2016). It was a continental terrane, with affinities to the Anatolide–Tauride Block, which underwent metamorphism under an obducted Cretaceous ophiolite followed by the development of a magmatic arc during the Late Cretaceous. The collision of the Kırşehir arc with the Pontides occurred in the late Maastrichtian – Palaeocene and produced the arc-shaped geometry of the central Pontides (Fig. 1; Kaymakçı et al. 2009; Meijers et al.2010) and the present-day triangular shape of the Kırşehir Massif, which is a result of folding of the originally north–south-trending Kırşehir magmatic arc (Lefebvre et al.2013).

The methods employed during this study include geological mapping, biostratigraphy, geochemistry and zircon U–Pb geochronology. Geological mapping was done on 1:25 000 scale topographic maps. The locations of samples and observation points are given in UTM coordinates in Table S2 (in the Supplementary Material available online at https://doi.org/10.1017/S0016756822000504). Samples collected during geological mapping were studied for planktonic and benthic foraminifera and for radiolaria. Planktonic and benthic foraminifera and calpionellids from the Mesozoic rocks were identified in thin-sections based on Zaninetti (1976), Altıner & Zaninetti (1980), Zaninetti et al. (1982), Al-Shaibani et al. (1983), Altıner (1991), Altıner & Özkan (1991), Martini et al. (2004, 2009), Premoli Silva & Verga (2004) and Senowbari-Daryan & Link (2017). Larger benthic foraminifera (nummulitids, orthophragminids) from the Eocene sequence were studied both in thin-section and as loose individual foraminifera. The latter have been sectioned through the equatorial and axial planes for taxonomic studies.

Six samples of red radiolarian cherts were processed at the University of Lille with diluted hydrofluoric acid (HF 4 %) for at least three repetitive leachings, c. 24 hours each. Residues obtained from acid leaching were washed with the help of a 63 μm mesh sieve. Radiolarians were picked under a stereo-binocular microscope with an eyebrow, mounted on a thin brush stick. They were then mounted on SEM stubs and photographed with a ZEISS EVO scanning electron microscope.

Mineral separation for zircon dating was done in the Istanbul Technical University using classical techniques including crushing, sieving, magnetic and heavy liquid separation. The zircons were picked under a binocular microscope, mounted in epoxy and polished to nearly half-width of the grains. Internal structures of the mounted zircons were imaged by means of cathodoluminescence (CL) in the Hacettepe University (Ankara) by ZeissEvo-50SEM. The CL images of the analysed zircons are given in Figure S1 (in the Supplementary Material available online at https://doi.org/10.1017/S0016756822000504). Zircons were analysed using laser ablation – inductively coupled plasma – mass spectrometry (LA-ICP-MS) at the University of California, Santa Barbara. For details of the method employed, see Kylander-Clark et al. (2013) and Okay et al. (2020 a). Long-term reproducibility in secondary reference materials is <2 % and, as such, should be used when comparing ages obtained within this analytical session to ages elsewhere. The U–Pb analytical data are given in Table 1 and Table S1 (in the Supplementary Material available online at https://doi.org/10.1017/S0016756822000504).

Whole-rock analyses were performed at Bureau Veritas Mineral Laboratories in Vancouver, Canada. A 1–5 kg sample was first processed in a steel jaw crusher, and an aliquot of c. 30 g was powdered in an agate rind-disc mill. Rock powders were then dried at 105 °C for c. 24 hours, and were sent for analysis. Accuracy is reported to be better than 2 % for major- and better than 10 % for trace-element analysis. For details of the analysis procedure see https://commodities.bureauveritas.com/metals-minerals/exploration-and-mining/geoanalytical-services.

Brief petrographic descriptions of samples analysed for geochemistry and/or for geochronology are given in the Supplementary Material (available online at https://doi.org/10.1017/S0016756822000504), along with their UTM locations and representative microphotos.

The ophiolitic mélange in the Beynam area south of Ankara was previously briefly described by Sarıfakıoğlu et al. (2017), and a 1:100 000 scale geological quadrangle map including the Beynam area was published by Akyürek et al. (1997). In the Beynam region the ophiolitic mélange forms a sub-vertical tectonic sequence unconformably overlain by the Upper Cretaceous fore-arc turbidites (Figs 4 and 5a). At present the unconformity surface dips steeply north at 70–80°. When the unconformity surface is restored to its original Late Cretaceous horizontal position, the underlying ophiolitic mélange provides a well-exposed 6 km thick vertical section of the subduction-accretion complex. The section consists of three well-defined fault-bounded tectonic slices, called accretionary units (AUs), which have thickness of a few kilometres and lengths of over 8 km (Fig. 4). During the Late Cretaceous the tectonic stack consisted of the Kuyumcudağ AU at the base, overlain by Otlubel and Holos AUs. Below we describe the AUs from base upwards.

Kuyumcudağ AU consists of a serpentinite mélange. It dips at c. 70° to the NW, has a minimum structural thickness of c. 1 km (Figs 4 and 5a) and continues northwest under the Neogene sedimentary cover. It lies with a steep fault contact over the Otlubel AU.

About 70 % of the Kuyumcudağ AU consists of variably serpentinized, foliated or blocky harzburgites, and the rest consists of tectonic blocks of basalt, Jurassic red radiolarian chert and limestone (Fig. 5b, c). Foliation in the serpentinite dips steeply NW. Petrographically the basalt consists of plagioclase, augite and late chlorite and calcite. Sarıfakıoğlu et al. (2017) report Middle to Late Jurassic radiolaria ages from four chert samples, which are shown on the geological map in Figure 4. The sizes of the blocks range from a few metres to c. 50 m. The blocks tend to be aligned and define a steeply dipping tectonic fabric (Fig. 5c).

The sub-vertical Otlubel AU has a structural thickness of ∼1.8 km. It consists mainly of basalt and diabase with lesser amounts of gabbro, plagiogranite, conglomerate and pelagic limestone. The Otlubel AU shows a stratigraphy with thin lenses of gabbro in the north passing up to a thick layer of Jurassic basalt and diabase with small veins and patches of plagiogranite, overlain by debris flow conglomerates intercalated with basaltic flows, which pass up into Lower Cretaceous pelagic limestones (Figs 4 and 5a).

The gabbro occurs as thin (<30 m) lenses on the northern margin of the Otlubel AU. It consists of plagioclase (An₆₂) and hornblende with minor pyroxene, quartz and late epidote. Basalt and diabase are the dominant rock types in the Otlubel AU and make up more than 80 % of the bulk. They occur as lava flows showing agglomerate and pillow lava structures. Locally diabase is cut by isolated basaltic dykes, which, however, make up a few per cent of the Otlubel AU. The primary igneous mineral assemblage in the diabase and basalt is plagioclase and hornblende with minor augite and opaque minerals (cf. Supplementary Material available online at https://doi.org/10.1017/S0016756822000504); augite is preserved only in 4 out of 20 samples petrographically examined. Secondary minerals include chlorite, actinolite, epidote and calcite.

The diabase is locally cut by plagiogranite veins, which form an irregular network and are generally tens of centimetres wide (Fig. 6a). The plagiogranite consists mainly of quartz and plagioclase with minor hornblende and opaque; secondary minerals include actinolite, chlorite and epidote. It makes up less than 0.1 % of the Otlubel AU. Out of three plagiogranite samples processed, only one (sample 12526) yielded four zircons, all of which contained appreciable common Pb (Table 1). Though a 204-based common-Pb correction proves challenging (low 204 counts, Hg interference etc.), it nevertheless yields equivalent Late Jurassic (Oxfordian) dates of 161.2 ± 3.2 Ma for two of the four grains. The other two zircons yield Late Carboniferous ages. Late Jurassic (150 ± 4 Ma) Ar–Ar hornblende ages from ophiolitic gabbros are also reported by Çelik et al. (2013) from further north in the Eldivan region (Fig. 2). Plagiogranites from the Eldivan region produced Early Jurassic (c. 180 Ma) zircon U–Pb ages (Dilek & Thy, 2006; Sarıfakıoğlu et al.2017). Early Jurassic (174–183 Ma) plagiogranite and diabase bodies are also described from the ophiolitic mélanges and from the largely intact ophiolite bodies along the İzmir–Ankara suture (Robertson et al. 2013; Topuz et al.2013; Uysal et al.2013; Sarıfakıoğlu et al.2017; Balcı & Sayit, 2020; Okay et al. 2020 a).

The basalt and diabase are overlain by conglomerates with clasts of pelagic limestone and basalt in a volcanogenic argillaceous matrix (Fig. 6b). Although the actual contact is not well exposed, the absence of serpentinite slivers or foliation along the contact suggests that it is stratigraphic. The conglomerate beds are up to 15 m thick and are separated by basaltic flows (Fig. 6c). Poor sorting and angular clasts indicate that the conglomerates represent debris flows. The conglomerates are overlain by pelagic limestones, which crop out as a continuous 150 m thick horizon along the southern margin of the Otlubel AU (Figs 4 and 6d). The limestones are medium-bedded, light-grey, beige, radiolaria-bearing micrites with thin shale interbeds. Most limestone samples contain just calcified radiolaria tests; however, two samples comprise Calpionella alpina, C. grandalpina, C. elliptica, Tintinnopsella carpathica, Crassicollaria parvula and Remaniella ferasini (Fig. 7, photos 44–62) besides abundant radiolaria, which indicate an earliest Cretaceous (Berriasian) age. Sarıfakıoğlu et al. (2017) also report Early Cretaceous ages from the limestone horizon, including an Aptian–Albian radiolaria age. The age of the limestone is compatible with the Late Jurassic age of the underlying oceanic crust. The Lower Cretaceous limestones are tectonically underlain by a continuous serpentinite slice, 20 m to 300 m in thickness, which dips at high angles to the north (70–80°) and marks the boundary between the Otlubel and Holos AUs (Figs 4 and 6d). North of the village of Karaali the serpentinite slice merges to a larger area of serpentinite, locally with blocks of basalt and chert (Fig. 4).

The Holos AU forms a sub-vertical tectonic slice between the Otlubel AU and the fore-arc turbidites, the Haymana Formation (Figs 4 and 5a). The Holos AU has a thickness of 2.1 km, and is made up predominantly of basalt (>65 %) with lesser amounts of Jurassic radiolarian chert, Triassic limestone and serpentinite. It differs from the Otlubel AU by the lack of any recognizable stratigraphy and by the presence of Triassic limestones, radiolarian cherts and serpentinite. Basalts in the Holos AU occur both as porphyritic pillow basalt with large white plagioclase phenocrysts set on a matrix of pinkish Ti-augite, plagioclase and opaque (Fig. 8f) and aphyric alkali basalts with a similar mineral assemblage. Some of the basalt samples contain analcime or kaersutite, attesting to their strongly alkaline character (cf. Supplementary Material available online at https://doi.org/10.1017/S0016756822000504). Kaersutite forms phenocrystals, up to 5 mm long, in a matrix of feldspar and opaque. Analcime occurs as idioblastic microphenocrysts in a matrix of augite, feldspar and opaque. Analcime was described previously from the basalts of the ophiolitic mélange in the Ankara Kalecik region (Çapan & Buket, 1975) and has been shown to be secondary after leucite (Varol, 2013). All the basalts show alteration involving formation of chlorite, calcite and epidote; the amygdales in the basalts are commonly filled by calcite. Serpentinite makes up less than 5 % of the Holos AU, occurs as steeply dipping narrow slivers, a few to tens of metres wide and up to a few hundred metres long, and marks tectonic contacts.

A characteristic feature of the Holos AU is the presence of shallow marine Upper Triassic limestone blocks embedded in a volcanoclastic matrix (Fig. 8). This Hörç Limestone makes up c. 10 % of the Holos AU. Akyürek et al. (1997) have mapped these limestones as Lower Jurassic; Sarıfakıoğlu et al. (2017) mentions the presence of Triassic limestone blocks in the Beynam area without giving any details. The Hörç Limestone is white, pale-grey, massive and contains corals, algae, bivalves, belemnites and gastropods (Fig. 8f). There are also very rare pink, micritic limestone blocks of probable Late Triassic age with thin-shelled bivalves and ammonites, which may represent forereef facies. The size of the limestone clasts ranges from a few mm to 50 m (Fig. 8). They occur in debris flows locally with blocks of basalt (Fig. 8d). The debris flow conglomerates form two distinct horizons, which can be followed for more than 2 km along strike (Fig. 4). Large blocks are worked in quarries, which allow three-dimensional observation of the blocks (Fig. 8b, c). More than 15 blocks were sampled for palaeontological study. They contain an Upper Triassic (upper Norian – Rhaetian) foraminiferal fauna including Galeanella laticarinata, Galeanella? minuta, Siculocosta floriformis?, Miliolipora cuvillieri, Orthotrinacria sp. Decapoalina schaeferae, Ophthalmidium leischneri, O. maximum, Triadodiscus eomesozoicus, Aulotortus communis, A. tumidus, A. planidiscoides, A. ex gr. sinuosus, Parvalamella praegaschei, P. friedli, Triasina hantkeni, Auloconus permodiscoides, Trocholina ultraspirata, Semiinvoluta sp., Foliotortus spinosus, Duotaxis birmanica, D. metula, ‘Tetrataxis’ humilis, ‘Tetrataxis’ inflata, Trochammina jaunensis, Reophax tauricus, Gandinella falsofriedli, Glomospirella amplificata, Endoteba controversa, Dentalina vadaszi, Austrocolomia canaliculata, Polarisella spp., Globochaete sp., Thaumatoporella parvovesiculifera, Tubiphytes obscurus and Baccanella floriformis (Fig. 7, photos 1–42 and Fig. 9).

Thinly bedded red radiolarian cherts occur as 5–20 m thick discontinuous horizons interbedded with red pelagic limestone and basalt. The radiolarian cherts occur throughout the Holos AU, and Jurassic radiolarian ages are determined from both its NW and SE margins. Red pelagic limestones did not produce any age diagnostic fossils, whereas several samples of radiolarian cherts contain a moderately well-preserved radiolarian fauna (Fig. 10) of Middle to Late Jurassic (Bathonian – Oxfordian) age. More particularly, sample 14584 yielded an age diagnostic species, Kilinora (?) oblongula (Fig. 10i), known to occur only in Unitary Association Zones (UAZ) 6 – 8 of the biozonation of Baumgartner et al. (1995); therefore sample 14584 may be correlated with the middle Bathonian to early Oxfordian time interval. The same age is deduced for sample 15127, as the co-occurrence of species Archaeohagiastrum munitum and Paronaella mulleri (Fig. 10m and n) allows its correlation with UAZ 6–8 of the biozonation of Baumgartner et al. (1995). Finally, sample 14585 may be assigned with certainty to the latest Bajocian/early Bathonian to late Oxfordian/early Kimmeridgian interval based on the presence of species Palinandromeda podbielensis, which is known from UAZ 5– 9 of Baumgartner et al. (1995). Furthermore, the presence of a poorly preserved specimen, which has been tentatively identified as Guexella nudata, suggests that the age of this sample may possibly be restricted to the latest Bajocian/early Bathonian to early Oxfordian interval (UAZ 5–8).

The ophiolitic mélange in the Beynam area does not show any regional metamorphism, and foliation is locally developed along shear zones and close to the contacts of the AUs, and is invariably steeply dipping (Fig. 8g). The steep fabric of the ophiolitic mélange is especially well marked in the Kuyumcudağ and Holos AUs. In the Kuyumcudağ AU, intense shearing is defined by sub-vertical foliation in the serpentinite; the blocks in the serpentinite also show a steep alignment. Shear zones are much rarer in the Otlubel AU (<3 %), which appears to have largely preserved the oceanic crustal stratigraphy.

The Holos AU has a fabric characterized by the alignment of debris flow conglomerates, pelagic limestone, chert and serpentinite bodies (Figs 4, 5 and 8g). The repetition of the Triassic debris flow horizons and Jurassic cherts and limestones within the Holos AU (Fig. 4) indicates that it consists of an imbricated thrust stack. However, in the absence of a few metres thick serpentinite slivers, individual faults in the Holos AU are difficult to map. The thrust imbrication must have occurred during the subduction before the deposition of the Upper Cretaceous fore-arc turbidites. Shear zones, as opposed to faults, are rare in the Holos AU (<10 % of the bulk, Fig. 8) and occur mainly close to its NW contact.

Over 40 magmatic rock samples from the Beynam area were petrographically studied. The samples show low-temperature (sub-greenschist facies) alteration involving formation of chlorite, calcite, epidote and pumpellyite. Among least altered samples fourteen were analysed for major and trace elements: six from Holos AU, seven from Otlubel AU and one from Kuyumcudağ AU. The locations of the samples are shown on the geological map in Figure 4, and the geochemical data are given in Table 2. Loss-on-ignition values range from 1.80 to 9.40 wt %, indicating the variably altered nature of the samples. The analysed samples are mostly basalt, andesite and diabase, with one gabbro and one plagiogranite.

The geochemical data reveal distinct differences in the geochemistry of the magmatic rocks from the Holos and Otlubel AUs. All the mafic rocks from the Holos AU are alkaline (Fig. 11a, b), as also shown by the presence of analcime, kaersutite and T-augite in the mineral assemblage of some samples. On a volatile-free basis, the mafic rocks in the Holos AU have lower SiO₂ (46–54 wt %) and MgO (2.9–7.5 %) and higher Al₂O₃ (13.7–21.2 %) and TiO₂ (1.0–3.6 %) than those from the Otlubel AU (SiO₂ 53–59 %, MgO 6.1–12.3 %, Al₂O₃ 12.4–17.2 %, TiO₂ 0.2–0.4 %). The rare earth element (REE) and trace element patterns of the Holos AU are characterized by a strongly fractionated shape typical of ocean island basalts (OIB; Fig. 11c, d). On the multi-element variation plots normalized to primitive mantle, they do not show any negative Nb and Ta anomaly in line with their anorogenic nature.

In the Zr/Ti versus Nb/Y plot of Pearce (1996) the mafic rocks from the Holos AU lie in the alkali basalt, and those from the Otlubel AU in the basalt fields (Fig. 11b). However, the volcanic rocks from the Otlubel AU have high SiO₂ contents (SiO₂ 53–59 %), which place them in the basaltic andesite and andesite fields in the total alkali – SiO₂ plot (not shown). They also have higher Mg numbers of 61–75 (molecular MgO*100/(MgO + FeO_total) compared to the mafic rocks from the Holos AU (45–55), and lower concentrations of P₂O₅ (0.01–0.003 wt %), Zr (7–15 ppm), Nb (1–2 ppm), Y (6–12 ppm), Th (0.3–0.6 ppm) and Hf (0.3–0.5 ppm) than the alkali basalts from the Holos AU (P₂O₅ 0.22–1.04 wt %, Zr 98–429 ppm, Nb 16–85 ppm, Y 17–37 ppm, Th 2–35 ppm and Hf 3–11 ppm; Table 2). The magmatic rocks from the Otlubel AU show less enrichment in rare earths than average enriched and normal mid-ocean ridge basalt (E-MORB and N-MORB; Fig. 11c) and have a typical flattened concave-upward pattern (Fig. 11c), which is observed in boninites (e.g. Crawford & Cameron, 1985; Pearce et al.1992). They also show minor depletion in Nb and Ta on multi-element variation plots normalized to primitive mantle, suggesting subduction influence. The mafic rocks from the Otlubel AU with their elevated SiO₂ (53–59 %) and low TiO₂ (0.2–0.4 %) contents, high Mg numbers (61–75) and REE pattern can be classified as boninites and plot in the boninite field of Pearce & Reagan (2019) (Fig. 11f).

In the Th/Yb versus Nb/Yb plot (Pearce, 2008), the Holos samples lie on the MORB–OIB array and close to the OIB field, whereas the Otlubel samples lie close to E-MORB but show a slight Th enrichment, which indicates a subduction influence (Fig. 11e). Only one basalt sample was analysed from the Kuyumcudağ AU, and it is geochemically similar to the alkali basalts from the Holos AU.

The plagiogranite sample is characterized by low K₂O (0.41 wt %) and high Na₂O contents (3.91 wt %; Table 2) and is chemically similar to oceanic plagiogranites (e.g. Coleman & Peterman, 1975). Its trace and rare earth element concentrations show a similar pattern to the mafic rocks of the Otlubel AU and are distinct from those of the Holos AU (Fig. 11c, d, e). This supports the view that the magmatic rocks of the Otlubel AU form a comagmatic sequence distinct from those of the Holos AU.

There are a large number of studies on the geochemistry of basaltic rocks from the ophiolitic mélanges in the Ankara region (e.g. Çapan & Floyd, 1989; Floyd, 1993; Tankut et al.1998; Rojay et al.2001, 2004; Gökten & Floyd, 2007; Göncüoğlu et al.2010; Dangerfield et al.2011; Sarıfakıoğlu et al.2014, 2017; Bortolotti et al.2018). These studies have shown that the dominant rock in the ophiolitic mélange is an ocean island type alkali basalt, similar to those from the Holos AU. This is also the case, for example, in the Japanese subduction-accretion complexes (e.g. Isozaki et al. 1990), and is explained by preferential accretion of topographic highs at the subduction zones. In the Ankara ophiolitic mélanges there are also smaller amounts of E-MORB and N-MORB type tholeiitic basalts locally with subduction influence (e.g. Rojay et al. 2004; Nairn et al. 2013; Bortolotti et al. 2018). In most cases there is no clear spatial or tectonic separation of different basalt types in the ophiolitic mélange. Dangerfield et al. (2011) notes the presence of three types of basalt in an area as small as 20 km². However, in the Beynam case, different types of basalts clearly belong to different subduction-accretion units. The Holos AU has alkali basalts, whereas Otlubel AU contains boninites, which appear to be rare in the Anatolian ophiolitic mélanges. Sarıfakıoğlu et al. (2017) describe boninites from the ophiolitic mélange in the Eldivan region, 110 km NE of Beynam (Fig. 2), and Robertson et al. (2013) from the southern margin of the easternmost Pontides.

The Holos AU is stratigraphically overlain by a thick sequence of Upper Cretaceous turbidites of the Haymana Formation. Upper Cretaceous turbidites have a wide distribution in central Anatolia and have been interpreted as a fore-arc sequence (Fig. 2; Görür et al.1984; Koçyiğit, 1991; Nairn et al.2013; Gülyüz et al. 2019). The Haymana Formation lies mostly on continental crustal sequences of the Sakarya Zone, including over the Upper Cretaceous limestones and olistostromes (Fig. 3; Ünalan et al.1976; Okay & Altıner, 2016; Okay et al.2019). Although some authors indicate a stratigraphic contact between the ophiolitic mélange and the Upper Cretaceous turbidites (e.g. Norman, 1972; Ünalan, 1981; Akyürek et al.1997; Nairn et al. 2013), there is no detailed description of such a contact. The Beynam region is one of the few places in the Ankara region where the stratigraphic contact between the ophiolitic mélange and the fore-arc sequence is well exposed. Two detailed sections were measured across this contact (Fig. 12).

At the base of section-1, there are sheared basalt, radiolarian chert and pelagic shale. This is overlain by a coherent stratigraphic sequence of red radiolarian cherts intercalated with shale and basalt and rare pelagic limestone, 100 m in thickness (Fig. 12). Two samples of radiolarian cherts contain Middle to Late Jurassic (Bathonian–Oxfordian) radiolaria including Kilinora (?) oblongula, Palinandromeda podbielensis and ? Guexella nudata (Fig. 10). The Haymana Formation starts above this Jurassic oceanic crustal slice with a 12 m thick, massive conglomerate (Figs 11 and 12a, b). The conglomerate contains rounded, 1–5 cm large clasts of black marble, basalt, neritic limestone, serpentinite and diabase in a sandy matrix. A sample from a limestone clast contains Jurassic incertae sedis Crescentiella morronensis. The conglomerate is overlain by violet, red and green silty clayey marls (Figs 12 and 13c), which contain early to middle Campanian (∼83–79 Ma) planktonic foraminifera Globotruncanita elevata, Globotruncana linneiana, G. bulloides, G. arca, Planoheterohelix globulosa and Muricohedbergella spp. (Fig. 7, photos 54–67). The marls are overlain by the typical thickly bedded, brown volcanogenic sandstones and shales of the Haymana Formation (Figs 12 and 14d).

At the base of section-2 there are also sheared, red radiolarian chert, red shale, basalt, red pelagic shale and rare serpentinite. Here, the Haymana Formation starts with medium, thickly bedded to massive, locally laminated volcanogenic sandstones, c. 20 m thick (Figs 12 and 13d), which lie unconformably over the ophiolitic mélange. The sandstones are overlain by violet, red mudstone and red radiolarian cherts, which are followed by a second horizon of 45 m thick volcanogenic sandstone with mudstone and siltstone layers. Samples from the mudstones contain Muricohedbergella planispira, Macroglobigerinolloides sp. and radiolaria, which indicate a post-Aptian age. The poorly sorted sandstones are rich in large feldspar and quartz grains and lithic volcanic clasts (see Supplementary Material available online at https://doi.org/10.1017/S0016756822000504). A sandstone sample (11169) was analysed for detrital zircons. The zircon U–Pb age spectrum is shown in Figure 14b, and the isotopic data and the CL images of the zircons are given in Table S1 and Figure S1, respectively (see Supplementary Material). The Upper Cretaceous detrital zircons are predominantly euhedral and show igneous zoning (Fig. S1), and their Th/U ratios are above 0.2 (Table S1), which indicate an igneous origin (see Supplementary Material). Out of 118 zircons analysed, 47 were concordant at 90–100 %, and 30 of those are Santonian to early Campanian (81–87 Ma). The oldest ages (87 ± 3 Ma and 86 ± 3 Ma) are represented by single zircons each, whereas there are five zircons with 85 ± 3 Ma ages. The youngest detrital zircons are 81 ± 3 Ma (three zircons) and provide an early Campanian lower age limit for the Haymana Formation.

The Campanian volcanogenic sandstones are stratigraphically overlain by two conglomerate beds, 5 to 15 m in thickness, separated by volcanogenic sandstones (Fig. 12). The conglomerates consist of rounded, 1–5 cm large clasts of black marble, basalt, limestone, serpentinite and diabase in a sandy matrix. Limestone clasts from the conglomerate contain benthic foraminifera Crescentiella morronensis, Mohlerina basiliensis and Charentia sp. (Fig. 7, photos 69–71) characteristic of the Upper Jurassic – Lower Cretaceous. In terms of facies, lithology and fauna, the limestone clasts in the conglomerate are similar to those of the Bilecik Group of the Sakarya Zone (Fig. 3; Altıner et al.1991). The conglomerates are in turn overlain by purple, green silty marls with a middle Campanian to middle Maastrichtian foraminiferal fauna including Globotruncana linneiana, Gl. ventricosa and G. bulloides, which pass up into volcanogenic sandstone and shale, which constitute the typical lithology of the Haymana Formation (Ünalan et al.1976; Okay & Altıner, 2016). Sandstones from the Haymana Formation are generally poorly sorted greywackes rich in volcanic and plutonic clasts and their petrography indicates a magmatic arc provenance (Çetin et al. 1986).

In the Beynam region the Haymana Formation has a stratigraphic thickness in excess of 1 km. The predominant facies is an intercalation of volcanogenic sandstone and shale (Fig. 14d). The sandstones show typical turbidite features including graded bedding, parallel and convolute lamination and slumps. There is also a 40 m thick channelized debris flow horizon NE of the village of Günalan with poorly sorted clasts of basalt (Fig. 4). Within the sandstone–shale sequence, there are rare (<3 %) marl and calc-arenite beds; several samples from such beds contain planktonic foraminifera of middle Campanian to middle Maastrichtian age range including Globotruncana linneiana, G. lapparenti, G. bulloides, G. arca, Muricohedbergella monmouthensis and Heterohelix globulosa (Fig. 7, photos 58, 67) and also transported benthic foraminifera Orbitoides sp. and Lepidorbitoides sp.

Palaeontological and isotopic data show that: (a) the base of the fore-arc sequence (the Haymana Formation) is early to middle Campanian in age (c. 81 Ma), which provides an upper age limit for the tectonic assembly of the AUs in the Beynam area, and (b) the fore-arc basin was sourced from Santonian to early Campanian (87–81 Ma) magmatic arc and from the ophiolitic mélange. However, there is no indication that the ophiolitic mélange was subaerially exposed prior to the deposition of the Haymana Formation. The prominent conglomerate levels in the base of the Haymana Formation represent debris flows, and the absence of volcanogenic sandstones in section-1 is most probably due to scouring by the massive debris flow.

The Haymana Formation is overlain with an angular unconformity by an Eocene shallow marine to continental sequence of sandstone, limestone and gypsum (Fig. 4). The Eocene sequence starts with sandy, pebbly limestones very rich in large benthic foraminifera. Individual foraminifera and rock samples, collected from the base of the Eocene section, yielded a shallow marine benthic foraminera fauna including Nummulites ex. gr. perforatus, Nummulites spp., Asterocyclina alticostata, Asterocyclina sp., Discocyclina sp., Orbitoclypeus sp., Asterigerina sp. and Victoriella sp. (Fig. S2 in the Supplementary Material available online at https://doi.org/10.1017/S0016756822000504). The fauna indicates a Middle Eocene age (middle–late Lutetian to early Bartonian). Considering that there are no marine Bartonian deposits in central and northern Anatolia (Okay et al.2020 b), the basal Eocene transgression can be dated to middle–late Lutetian (c. 45 Ma). Middle Eocene sandstones lie unconformably over the older units both in the Sakarya Zone and in the Kırşehir Massif (e.g. Gülyüz et al.2013; Nairn et al. 2013; Okay et al.2020 b) and represent the oldest post-collisional cover.

Geological mapping has shown that the ophiolitic mélange in the Beynam region consists of three AUs, distinguished by lithology, structure, age and geochemistry. The Kuyumcudağ AU is the only one which can be defined strictly as a tectonic mélange (e.g. Raymond, 1984; Festa et al.2020), in that it has a well-defined matrix of sheared serpentinite with blocks of basalt, Jurassic chert and limestone. The Otlubel AU is a semi-intact piece of Late Jurassic oceanic crust with boninite chemistry. The Holos AU is an imbricate tectonic stack of alkali basalt, chert and limestone. Accretionary units with similarly distinct lithological, geochemical, structural and temporal features are also are described from the Franciscan Complex (e.g. Wakabayashi, 2015; Raymond et al.2020) and from Japan (e.g. Isozaki et al. 1990: Isozaki, 1997).

While it is easy to explain the presence of alkali basalts in the Holos AU, as having formed on an oceanic seamount, the presence of boninites in the Otlubel AU poses a problem. Boninites are typically found in supra-subduction ophiolites, are commonly associated with subduction initiation and are located in the upper plate in a subduction system (e.g. Crawford et al.1989; Pearce & Reagan, 2019). The Late Jurassic Otlubel AU is located in a subduction-accretion complex sandwiched between two other AUs, and thus was clearly part of the lower plate. However, it could have formed above a mid-oceanic subduction zone, and later been incorporated in the subduction-accretion complex (Fig. 15), as discussed in Section 9.5 further below.

A number of observations indicate that the Triassic limestones in the Holos AU were deposited on an oceanic seamount, as also suggested by Sarıfakıoğlu et al. (2017). These observations include: (a) Lack of continent-derived siliciclastic detritus in the Holos AU, showing that the Triassic limestone blocks could not have been derived from a continent. (b) In the debris flows Triassic limestone and basalt blocks are closely intermixed (Fig. 8d), suggesting that the Hörç limestone was deposited on oceanic basalt. (c) The geochemistry of the basalts in the Holos AU suggests an oceanic island setting. (d) The Hörç limestone is predominantly shallow marine. (e) The lithology and fauna of the Triassic limestones in the Holos AU are unique and cannot be correlated with any of the Triassic limestone sequences in the Pontides.

Seamounts and islands consisting of shallow marine limestone on a basaltic substratum are common in the present Pacific Ocean (Nunn et al.2016). Accreted seamounts are common in the western Pacific convergent margins (e.g. Dilek & Ogawa, 2021), and shallow-marine limestone blocks including Late Triassic ones are described from subduction-accretion complexes in Japan (e.g. Isozaki et al.1990; Chablais et al.2010). The collapse of the limestone carapace occurs during the attempted subduction of the seamount (e.g. Sano & Kanmera, 1991). The Hörç Limestone must also have been deposited on a Triassic oceanic seamount in the Tethyan ocean, and then been incorporated into an accretion complex during the subduction (Fig. 15a, b).

Continent-derived sediments are absent in the ophiolitic mélange in the Beynam region, therefore all rocks in the ophiolitic mélange must have been formed or deposited on oceanic crust. Palaeontological and isotopic data from the ophiolitic mélange in the Beynam region indicate the presence of Triassic and Late Jurassic oceanic crust. The oceanic seamount recorded in the Holos AU must have been constructed on a pre-Late Triassic oceanic crust. Further support for the presence of a Triassic oceanic crust in the İzmir–Ankara ocean is found in the ophiolitic mélanges further west, in the Mihaliçcık region, which include Upper Triassic radiolarian cherts (Tekin et al.2002), and further east in the Late Triassic plagiogranites (c. 225 Ma; Çelik et al.2018). The age of the plagiogranite and the ages of sedimentary rocks in the Beynam region, including radiolarian cherts, provide solid data for the Jurassic oceanic crust. Similar Jurassic ages are reported from other parts of the ophiolitic mélange in the Ankara region (Dilek & Thy, 2006; Çelik et al.2011; Sarıfakıoğlu et al.2017; Bortolotti et al.2018) and further east in the Eastern Pontides (Topuz et al.2013; Robertson et al.2013) and Lesser Caucasus (Danelian et al. 2010, 2016; Rolland et al. 2010). The youngest rocks in the Beynam area are Lower Cretaceous pelagic limestones.

Published ages of radiolaria and foraminifera from the oceanic sedimentary rocks in the ophiolitic mélanges in central Anatolia range from Late Triassic (late Norian) to early Late Cretaceous (Cenomanian, c. 100 Ma; Fig. 3). There are no confirmed post-early Late Cretaceous palaeontological ages from the ophiolitic mélange, which suggests that the subducting oceanic lithosphere was Triassic to early Late Cretaceous in age. The radiolarian chert sequences described from the ophiolitic mélanges also have short age ranges (Fig. 3), which implies a short period between the generation of oceanic lithosphere and its subduction.

The Late Cretaceous subduction of the İzmir–Ankara ocean was followed by the Palaeocene hard collision between the Sakarya Zone and the Kırşehir Massif (e.g. Kaymakçı et al.2009; Meijers et al.2010; Lefebvre et al.2013). The architecture of the ophiolitic mélange in the Ankara region is therefore the combined result of subduction and collision. The preservation of the Upper Cretaceous fore-arc turbidites, the Haymana Formation, in the Beynam region allows a separation between these two types of structures. The Haymana Formation is characterized by sub-vertical dips but does not show any small-scale folding or disruption of bedding in the Beynam area, where it is overlain unconformably by Middle Eocene sedimentary rocks. This shows that the steep tectonic fabric in the ophiolitic mélange and in the Haymana Formation is a result of continental collision. The Palaeocene collision in central Anatolia was not intense since the ongoing convergence between Eurasia and the Arabian Platform was partly taken up by subduction along the Eastern Mediterranean and the Bitlis–Zagros ocean (Fig. 1b). In the Haymana region, which was located further away from the trench than the Beynam area (Fig. 2), the marine sedimentation continues without a break from the Late Cretaceous into the Middle Eocene (e.g. Ünalan et al.1976; Kaymakçı et al. 2009; Nairn et al. 2013).

Restoration of the bedding in the Haymana Formation to the horizontal results in a sub-horizontal fabric in the ophiolitic mélange similar to that observed in frontally accreted subduction-accretion complexes. Frontal accretion is also compatible with the lack of high-pressure metamorphism in the ophiolitic mélange in the Ankara region. The ophiolitic mélanges in the Tavşanlı Zone further west show an incipient HP metamorphism with the development of lawsonite, aragonite, sodic amphibole and sodic pyroxene (Topuz et al.2006; Plunder et al.2013), which are not found in central Anatolia.

Pontides have been an active margin at least since the Triassic. However, subduction accretion and arc magmatism have been episodic, with major phases during the Late Triassic, Middle Jurassic (175–155 Ma) and Late Cretaceous (e.g. Fig. 3; Okay & Nikishin, 2015; Akdoğan et al. 2018, 2019; Okay et al.2019). The Late Jurassic – Early Cretaceous (157–135 Ma), on the other hand, was a period of carbonate deposition in the Pontides, with little evidence for arc magmatism or subduction accretion. However, data from the Beynam region suggest that subduction was initiated during the Late Jurassic (c. 161 Ma), leading to the formation of a supra-subduction oceanic lithosphere, as represented by the boninitic Otlubel AU (Fig. 15a). The surface of the newly created oceanic crust must have been above the carbonate-compensation depth as shown by the deposition of the pelagic carbonates, which also characterize the Upper Jurassic – Lower Cretaceous stratigraphy of the Sakarya Zone (Fig. 3; Altıner, 1991). This suggests that pelagic carbonate deposition extended from the continental margin to the supra-subduction oceanic crust (Fig. 15a). Robertson et al. (2021) describe a similar situation from the ophiolitic mélange in the Aladag region in the Taurides. The interaction between the pelagic carbonates and Late Jurassic volcanism may have formed the debris-flow conglomerates observed in the Otlubel AU. The accretion of the Holos AU most probably occurred during the Late Jurassic – Early Cretaceous in this intra-oceanic subduction zone, which involved collision and partial accretion of the Hörç seamount (Fig. 15a).

The incorporation of the Late Jurassic supra-subduction oceanic crust in the subduction-accretion complex requires that in the early Late Cretaceous the subduction zone jumped inboard to the ocean–continent boundary, resulting in an Andean-type subduction zone (Fig. 15b; Dangerfield et al. 2011). This might have been initiated through soft collision with the Kırşehir Massif (Fig. 15c). This also led to the development of a continental magmatic arc and of a fore-arc basin (Fig. 15c). This Palaeo-Galatian arc, first suggested by Koçyiğit (1991), is mainly represented by the Late Cretaceous (80–73 Ma) I-type Beypazarı Granite with a geochemistry compatible with a magmatic arc setting (Öztürk et al.2012) and a few outcrops of Upper Cretaceous volcanic rocks (Fig. 2; Koçyiğit et al.2003; Okay et al.2019); most of it is concealed under the Neogene sedimentary cover (Fig. 2).

The time of initiation of the Andean-type subduction zone can be estimated from the timing of continental arc magmatism, the best record of which is found in the fore-arc basin sediments (e.g. Paterson & Ducea, 2015; Sharman et al.2015), represented in this case by the Haymana Formation. In the Beynam region, detrital zircons from the basal sandstone beds of the Haymana Formation are predominantly Santonian to Campanian (87–81 Ma; Fig. 14b), suggesting c. 87 Ma for the start of arc magmatism. Further east, in the Haymana and Alcı areas, the detrital zircons from the Haymana Formation are younger mostly Campanian to early Maastrichtian (80–71 Ma; Fig. 14c; Okay et al.2019). This indicates that Late Cretaceous arc magmatism ranged between 87 Ma and 71 Ma (Santonian – early Maastrichtian). The time between the initiation of subduction and inception of arc magmatism depends on the slab dip angle and convergence rate, and ranges from a few million to 10 million years (e.g. Stern, 2002). The well-dated Coniacian (89–87 Ma) olistostromes in the Sakarya Zone (Fig. 2) contain clasts of ophiolitic mélange, which must have been derived from an accretionary complex (Okay et al.2019). These constrain the inception of Andean-type subduction to Cenomanian–Turonian (c. 97–90 Ma; Fig. 15c). Hence, the incorporation of the Otlubel AU in the ophiolitic mélange must have occurred during the Late Cretaceous. The Late Cretaceous subduction zone in the Ankara region had a NNE trend, which was also suggested for the western boundary of the Kırşehir Massif (Lefebvre et al. 2013; van Hinsbergen et al.2016; Maffione et al.2017).

All large subduction-accretion complexes, such as the Franciscan Complex in California or the Makran Complex in Iran, are dominated by greywacke-type sandstones derived from the magmatic arc and deposited in the trench and fore-arc. In contrast, no sandstones are recognized in the ophiolitic mélange in the Beynam region; in other parts of Anatolia, sandstones generally make up a few per cent of the ophiolitic mélange (e.g. Plunder et al.2013; Okay et al.2020 a). Detrital zircons from the rare sandstone bodies in the ophiolitic mélange west of Ankara do not contain any Jurassic or Cretaceous zircons (Okay et al.2020 a). The absence of sandstones in the Anatolian ophiolitic mélanges can be ascribed to three factors: (a) The main phase of accretion took place during the Late Jurassic – Early Cretaceous in an intra-oceanic subduction zone away from continental influence. (b) The Late Jurassic – Early Cretaceous was a period of marine limestone deposition on the continent, on the Sakarya Zone (Figs 3 and 15b; Altıner, 1991). (c) By the early Late Cretaceous, when the subduction zone jumped inboard, most of the Tethyan oceanic lithosphere was already subducted, and minor accretion took place during the Late Cretaceous (Fig. 15d).

1. The ophiolitic mélange in the Beynam (Ankara) region consists of distinct subduction-accretion units distinguished by lithology, age, structure and geochemistry, which were tectonically assembled mainly during the Late Jurassic and Early Cretaceous oceanic subduction.

2. In the Beynam (Ankara) region three subduction-accretion units (AU) are defined (Fig. 4). The structurally lowest one is a serpentinite mélange; the intermediate one, the Otlubel AU, is a semi-intact Late Jurassic oceanic crust with a boninite geochemistry; and the topmost Holos AU consists of OIB-type alkali basalts with Upper Triassic seamount-derived shallow marine limestones and Middle to Upper Jurassic radiolarian cherts and pelagic limestones.

3. The Late Jurassic oceanic crust was formed above a subduction zone, and then was incorporated into an accretionary complex, when the subduction jumped inboard in the early Late Cretaceous creating an Andean-type convergent margin (Fig. 15).

4. The ophiolitic mélange is stratigraphically overlain by a fore-arc turbidite sequence (the Haymana Formation), the base of which is dated to early to middle Campanian (c. 81 Ma). This provides an upper age limit for the tectonic assembly of the ophiolitic mélange in the Beynam area. The detrital zircons from basal sandstones in the fore-arc sequence are dominated by Santonian and Campanian zircons (87–81 Ma) derived from the magmatic arc.

5. Unlike most other subduction-accretion complexes, the ophiolitic mélange consists solely of oceanic crustal units and is devoid of land-derived coarse-clastic rocks. This is due to accretion occurring in an intra-oceanic subduction zone away from continental influence. During the Late Jurassic and Early Cretaceous period, the upper plate, the Sakarya Zone, was also a site of marine carbonate deposition (Fig. 3) and hence was not a source of clastic sediment.

6. During the Late Cretaceous the ophiolitic mélange, the fore-arc sequence (the Haymana Formation) and the Palaeo-Galatian arc formed a N–S-trending Andean-type convergent margin.

7. Ophiolitic mélanges are generally defined as monolithic units; however, geological mapping in the Beynam area has shown that at least locally they can be divided into distinct mappable tectonic units defined by lithology, age and geochemistry.

8. Ophiolitic mélanges in Anatolia are generally considered as of Late Cretaceous age; however, data from the Ankara region show that they formed mostly during the Late Jurassic and Early Cretaceous.

To view supplementary material for this article, please visit https://doi.org/10.1017/S0016756822000504

This study was supported by İTÜ-BAP project (AIO and EÖ, project no. 41644) and by TÜBA for AIO. T.D. thanks the Région Hauts-de-France, the Ministère de l’Enseignement Supérieur et de la Recherche (CPER Climibio) and the European Fund for Regional Economic Development for their financial support. We thank Bora Rojay for discussions on the geology of the Ankara region. Ezgi Sağlam, Sinan Yılmazer and Turgut Duzman are thanked for help with rock preparation and mineral separation and Ali Osman Yücel and Sylvie Régnier for help with palaeontological preparations. We also thank Alastair Robertson and an anonymous reviewer for detailed and constructive reviews, which improved the text.