Age and mantle sources of Quaternary basalts associated with “leaky” transform faults of the migrating Anatolia-Arabia-Africa triple junction

Near the Gulf of İskenderun in the eastern Mediterranean Sea, the Africa, Anatolia, and Arabia tectonic plates form an unstable triple junction (referred to here as the A³ triple junction) along left-lateral transform faults (Fig. 1; Dewey et al., 1986). This instability is associated with significant crustal deformation reflected by broad regions of high seismicity (Ergin et al., 2004), apparent crustal thinning (Abgarmi et al., 2017), and the development of transtensional basins (e.g., Cilicia-Adana-İskenderun and Samandağ-Amik-Karasu-Narlı) and transpressional ranges (e.g., Amanos Mountains) under conditions of complex strain partitioning (Fig. 2; Tatar et al., 2004; Aksu et al., 2005; Seyrek et al., 2007). Within this dynamic setting, small volumes of Quaternary mafic lava occur locally along transform fault segments and/or between rotating crustal blocks within fault zones (Fig. 2; e.g., Tatar et al., 2004; Kavak et al., 2009). Because these near-primary basalts are restricted spatially to transform faults of the developing A³ triple junction, this magmatism provides a rare opportunity to study melt generation within an exposed and actively developing plate boundary. More generally, the unusual setting of basaltic magmatism along transform faults within continental crust is significant because it provides context for their petrogenesis. Although a rare occurrence within continental crust, such small magmatic centers are known to occur in association with transform faults and fracture zones within oceanic crust (e.g., Hékinian et al., 1995; Perfit et al., 1996), with the potential for large-scale fluid and mineral systems forming along transform faults associated with triple junctions (e.g., Fletcher et al., 2020). Moreover, understanding the regional tectonic implications of triple junctions and their migration (e.g., Furlong and Schwartz, 2004), including the role of triple junction evolution in the development of escape tectonics, constitutes an important aspect of global tectonics that is not well understood.

Modern-day geophysical methods provide insight into physical heterogeneities within Earth’s crust and mantle at depths where magmas are sourced, and this insight is particularly useful for characterizing the genesis of recently erupted near-primary melts (e.g., Plank and Forsyth, 2016; Reid et al., 2019). The integration of geochemical and geophysical data sets furthers our understanding of the chemical, thermal, and physical evolution of tectonically active regions. New data from the U.S. National Science Foundation (NSF) Continental Dynamics–Central Anatolian Tectonics (CD-CAT) seismic deployment (e.g., Abgarmi et al., 2017; Delph et al., 2017; Portner et al., 2018) vastly increased seismic coverage in south-central Anatolia, thus providing unprecedented imaging of physical contrasts in the crust and upper mantle. Geophysical images of the upper mantle beneath southern Anatolia near the A³ triple junction highlight inferred regions of subducted Neotethyan oceanic lithosphere and the associated African continental margin (Biryol et al., 2011; Abgarmi et al., 2017; Delph et al., 2017; Portner et al., 2018). Lateral changes in Anatolian upper-mantle velocity structure appear to be related to the regional distribution of Quaternary magmatism, such that magmatism near the triple junction is generally restricted to areas underlain by thinner lithosphere (Delph et al., 2017). However, understanding the ways in which these lithospheric-scale contrasts in seismic properties may correlate with the geochemistry of Quaternary volcanic rocks requires a more careful analysis of magma source characteristics.

In this study, we present rock compositions, high-precision ⁴⁰Ar/³⁹Ar geochronology, and radiogenic isotope geochemistry (Sr, Nd, Hf, Pb) data of Quaternary basalts from two volcanic fields proximal to transform fault segments of the A³ triple junction alongside a compilation of published results, and we compare the integrated data set with seismic images of the uppermost mantle. We demonstrate that the geochemical and geophysical data support magma generation at two distinct depth intervals, consistent with melting of both lithospheric and asthenospheric mantle sources. When placed within the tectonic context of the region, the ascent of these basaltic magmas is likely facilitated by “leaky” transform faults. These results have implications for the timing and position of the developing triple junction, evolution of plate-boundary faults, and tectonic escape of the Anatolia microplate.

The tectonic forces controlling movements between the tectonic plates of Anatolia, Africa, and Arabia in the eastern Mediterranean have led to the development of a tectonically complex and unstable triple junction (e.g., Şengör et al., 1985; Perinçek and Çemen, 1990; Rojay et al., 2001; Reilinger et al., 2006). Modern geodesy and radar measurements near the triple junction show relative plate displacements up to 20 mm yr^–1 (Reilinger et al., 2006; Mahmoud et al., 2013; Cavalié and Jónsson, 2014; Simão et al., 2016). The A³ triple junction is currently centered near the Amik subbasin within the Karasu rift (Fig. 2; Dewey et al., 1986; Over et al., 2004; Yuce et al., 2014), although other triple junction locations have been proposed owing to the tectonic complexity and distributed deformation within the region (e.g., Karig and Kozlu, 1990; Tarı et al., 2014). The three plates are separated by two major transform fault systems: the left-lateral NE-SW–trending East Anatolian fault zone and the left-lateral N-S–trending Dead Sea fault zone (Fig. 2; Dewey and Şengör, 1979; Rojay et al., 2001; Westaway, 2004; Tatar et al., 2004; Duman and Emre, 2013). Here, both transtension and transpression are recorded in a network of smaller faults subparallel to the plate boundaries (e.g., Kavak et al., 2009). Focal mechanisms between fault zones reflect their complex interplay, showing normal, thrust, and mixed compressional and extensional oblique-slip motion over a large range of depths (Ergin et al., 2004). Within the field area, transtension across a wide region surrounding the A³ triple junction has resulted in the development of major basins, including the Cilicia-Adana-İskenderun Basins and the Karasu Basin, with its Samandağ, Amik, and Narlı subbasins, located northwest and southeast of the East Anatolian fault zone, respectively (Fig. 2).

The Dead Sea fault zone and the East Anatolian fault zone are neotectonic structures that crosscut an earlier orogenic belt of Late Cretaceous ophiolites, including the Kızıldağ (Amanos) and Baer Bassit ophiolites (Fig. 2). These ophiolites represent the former collision zone (Bitlis suture) between the Nubia (Africa-Arabia) and Eurasia (Anatolia) plates and remnants of the intervening Neotethys oceanic basement. This ophiolite mélange was itself overthrusted during the late Miocene by Jurassic–Cretaceous platform carbonates, equivalent to those exposed in the Kyrenia mountain range in Cyprus, forming an imbricated stack of ophiolite and carbonate rocks (e.g., Şengör and Yılmaz, 1981). Near the Gulf of İskenderun, the transition from a largely compressional tectonic system to one dominated by transtension and transpression developed when rifting in the Red Sea began in the late Miocene, leading to formation of the translational Dead Sea–East Anatolian fault zone system, and finally to tectonic extrusion of the Anatolian plate along the East Anatolian fault zone (e.g., Şengör et al., 1985; Hempton, 1987).

The East Anatolian fault zone has been active since the Pliocene (e.g., McKenzie, 1976; Şengör et al., 1985; Muehlberger and Gordon, 1987), with present-day average left-lateral slip estimated by global positioning system (GPS) data of ∼10 mm yr^–1 (Reilinger et al., 2006). Near the town of Çelikhan, the East Anatolian fault zone splits into two main segments (a northern minor strand and a southern main strand) largely separated by the Amanos Mountains (Fig. 2; Duman and Emre, 2013). The Amanos Mountains, an ∼2300-m-elevation range, represent an uplifted block of Precambrian-cored continental crust within a major zone of transpression, and, despite its position between the two major strands of the East Anatolian fault zone, it contains few mapped Quaternary faults. This apparent tectonic rigidity is consistent with its relatively recent and coherent uplift since 4 Ma (Seyrek at al., 2008), and the stark relief of the Amanos Mountains relative to the low-relief Arabian margin suggests a strong asymmetry in structural deformation and differential crustal uplift (e.g., Rojay et al., 2001; Reilinger et al., 2006).

The main southern strand of the East Anatolian fault zone extends ∼580 km from the so-called Karlıova triple junction in the east, where it joins the North Anatolian fault, to the Amik Basin in the west (Rojay et al., 2001; Westaway, 2004; Kavak et al., 2009; Duman and Emre, 2013). The Amik Basin (Fig. 2) approximately coincides with the northern termination of the Dead Sea fault zone, which accommodates ∼5 mm yr^–1 of left-lateral motion between the Africa and Arabia plates (Reilinger et al., 2006; Mahmoud et al., 2013), and its continuation as the main southern strand of the East Anatolian fault zone (e.g., Arpat and Şaroğlu, 1972; Rojay et al., 2001; Westaway, 2004; Kavak et al., 2009; Duman and Emre, 2013). Where the East Anatolian fault zone transits the Karasu valley (also referred to as rift, trough, basin, or graben), it is characterized by a diffuse network of individual fault segments comprising several restraining and releasing bends (Duman and Emre, 2013). The Karasu valley is bordered on its western margin by the Amanos Mountains and the east-dipping Karasu (Amanos) fault zone, and on its eastern margin by the west-dipping Yesemek (East Hatay) fault zone (Fig. 2; Rojay et al., 2001; Seyrek et al., 2008; Duman and Emre, 2013). Farther to the southeast, the Aaferin fault represents a major fault splay of the Dead Sea fault zone that roughly parallels the Karasu valley before linking with the main strand of the East Anatolian fault zone near the town of Kahramanmaraş (Fig. 2).

The northern segment of the East Anatolian fault zone diverges from its main segment near the village of Çelikhan, continuing ∼350 km toward the Gulf of İskenderun. Similar to the main southern strand of the East Anatolian fault zone, the northern strand consists of several individual fault segments connected by releasing and restraining bends and ultimately projecting offshore on strike toward the Kyrenia-Misis fault zone (Fig. 2; Duman and Emre, 2013). Both the northern and southern strands of the East Anatolian fault zone project offshore into the Mediterranean, likely connecting to major fault networks associated with the Cyprean trench (e.g., Duman and Emre, 2013), where strike-slip deformation linked to the interaction of the three plates gives way to compressional deformation related to the northward movement of the Africa plate (e.g., Imprescia et al., 2012). The northern strand of the East Anatolian fault zone becomes diffuse as it approaches the Gulf of İskenderun; deformation is distributed across smaller faults within a broad zone of transtension. Along the northern margin of the gulf, two subparallel fault segments, the Karataş and Yumurtalık faults (Fig. 2), both predominantly sinistral strike-slip faults with some reversed throw, can be considered as examples of small discontinuous faults that are part of a much broader network of transtensional structures. The Karataş and Yumurtalık faults were previously referred to as the Karataş-Osmaniye fault (e.g., McKenzie, 1976), and they have often been considered to be the boundary between Anatolian and African lithosphere (e.g., Parlak et al., 1997; Gürsoy et al., 2003; Westaway, 2003, 2004; Seyrek et al., 2008; Kavak et al., 2009).

There are two small Quaternary volcanic fields (here referred to as the Toprakkale and Karasu volcanic fields) locally occurring along and/or near both the northern and southern fault segments of the East Anatolian fault zone (Figs. 2 and 3); they consist of volcanic cones, lava flows, and ash-fall deposits (e.g., Parlak et al., 1997, 1998; Yurtmen et al., 2000; Rojay et al., 2001; Alıcı et al., 2001; Kavak et al., 2009). Quaternary volcanic rocks along the East Anatolian fault zone are neither voluminous nor widespread and are absent west of the İskenderun Basin and similarly absent northeast of Fevzipaşa along the main strand of the East Anatolian fault zone (Fig. 3).

The Toprakkale volcanic field refers to all of the Quaternary volcanic centers and lava flows of the Ceyhan-Osmaniye-Erzin-Toprakkale region, including the basaltic ash-fall deposits along the northern shore of the Gulf of İskenderun near İncirli (Fig. 3; e.g., Polat et al., 1997; Arger et al., 2000; Yurtmen et al., 2000; Bağcı et al., 2011; Italiano et al., 2017; Oyan, 2018). Volcanism in the Toprakkale volcanic field is spatially associated with the roughly parallel NE-SW–trending Karataş and Yumurtalık faults and a small stepover near the Delihalil volcano represented by the Toprakkale fault (Duman and Emre, 2013).

Published ⁴⁰Ar/³⁹Ar and K-Ar ages between 2.3 and 0.1 Ma for basalts in the Toprakkale volcanic field (Arger et al., 2000; Oyan, 2018) are consistent with paleomagnetic studies of volcanic rocks along the northern strand of the East Anatolian fault zone that have both normal and reversed polarities (Gürsoy et al., 2003). The Toprakkale volcanic rocks are mafic, with high combined alkalis (Na₂O + K₂O), and they plot as basanites and mostly alkaline basalts (Fig. 4). Relative to primitive mid-ocean-ridge basalts (MORB), they are enriched in light rare earth elements (LREEs), large ion lithophile elements (LILEs), and high field strength elements (HFSEs; e.g., Polat et al., 1997; Arger et al., 2000; Yurtmen et al., 2000; Bağcı et al., 2011). Values of ⁸⁷Sr/⁸⁶Sr and ¹⁴³Nd/¹⁴⁴Nd are consistent with derivation from an ocean-island basalt (OIB)–like asthenospheric mantle source together with some component of partial melting of lithospheric mantle. Trace-element differences between the basanites and alkaline basalts have been argued to indicate differential degrees of lithospheric mantle partial melting (1.5%–4.6% vs. 9.2%; Bağcı et al., 2011; Oyan, 2018).

The Karasu volcanic field includes all localized Quaternary lava flows and vents in the Amik Basin, a subbasin within the Karasu rift that is exposed over a distance of ∼60 km between Kırıkhan and Fevzipaşa (Figs. 2 and 3; Rojay et al., 2001). Northeast of Fevzipaşa, a tectonic saddle separates the Amik and Narli subbasins, which contain either Quaternary (Amik) or Miocene (Narlı) volcanic rocks (Rojay et al., 2001). Borehole data within the Karasu valley are consistent with sedimentation and volcanism occurring throughout much of the Quaternary, as the Karasu valley continues to act as a zone of active infilling, with sag structures (subbasins) consistent with active transpression and intermittent basaltic volcanism (Rojay et al., 2001).

When plotted on a conventional total alkali versus SiO₂ rock discrimination plot (Fig. 4), the volcanic rocks from the Karasu volcanic field straddle the boundary between the alkaline and subalkaline fields (Irvine and Baragar, 1971), which represents the low-pressure thermal divide defined by olivine-clinopyroxene-plagioclase stability. Fractional crystallization of any of these phases would drive the residual magma away from the divide, indicating that the Karasu basalts show evidence for both feldspathoid-normative and quartz-normative compositions. In notable contrast with the Toprakkale volcanic field, no basanites have been described from the Karasu volcanic field, but both alkali olivine basalts and quartz- or olivine-tholeiites have been described (Çapan et al., 1987; Alıcı et al., 2001). Any petrogenetic signature is poorly resolved in the geochemical and isotopic data, with, for example, ⁸⁷Sr/⁸⁶Sr and ¹⁴³Nd/¹⁴⁴Nd values showing considerable overlap. The wide range in ⁸⁷Sr/⁸⁶Sr (0.703353–0.705490) and trace-element compositions could indicate crustal contamination in many of the samples from the Karasu valley (Alıcı et al., 2001).

Eight samples of fresh volcanic rock were collected from the volcanic fields, five from the Toprakkale volcanic field and three from the Karasu volcanic field, in July 2012 (Fig. 3). All of the samples were either directly or proximally associated with fault segments of the East Anatolian fault zone. Within the Toprakkale volcanic field, samples of dark basaltic rock were collected from outcrops including near the summit of the Quaternary Delihalil volcano (R12OS-1; Fig. 5A), along the flank of Delihalil (R12OS-3; Fig. 5B), and outside the town of Toprakkale (R12OS-4; Fig. 5C). Additional samples were collected from layered volcanic flow and ash deposits that occur along the northern margin of the Gulf of İskenderun, in the vicinity of the Yumurtalık fault (R12OS-5, R12OS-6; Fig. 5D). Within the Karasu volcanic field, two samples were collected near the village of Hassa, along the Karasu (Amanos) fault segment of the southern strand of the East Anatolian fault zone (Fig. 3). Sample R12HA-1 (Fig. 5E) was collected from a fresh lava flow exposed near a flowing creek bed, and sample R12HA-02 (Fig. 5F) was collected from stratigraphically overlying fresh lava. Sample R12Ha-03 was collected near the village of Fevzipaşa. Samples R12HA-1 and R12HA-2 were collected in an area where several small volcanic centers (Büyük, Güvenç, Hassa, Bosalan) occur, possibly within local dilational zones (“spenochasms”; Tatar et al., 2004) formed in response to differential stress along the East Anatolian fault zone. These sample sites are within the much larger regional K-Ar geochronology study by Rojay et al. (2001) and were collected to try to gain better precision on the ages from this area. Specific sampling procedures corresponding to different analytical methods are described below.

Surficial contamination was removed from rock samples by extensive rinsing in water, followed by a sequence of 5–10 min of ultrasonication that progressed from dilute (∼2%) hydrogen peroxide to dilute (∼0.01 N) hydrochloric acid, with intermediate and final steps in distilled water. Major- and trace-element concentrations (Table 1) were determined on rock powders by quantitative X-ray fluorescence spectrometry, inductively coupled plasma–mass spectrometry, and inductively coupled plasma–atomic emission mass spectrometry. Samples for this investigation were analyzed by SGS Minerals in Lakefield, Ontario, Canada.

The ⁴⁰Ar/³⁹Ar analyses were performed at the U.S. Geological Survey (USGS) in Denver, Colorado. Fresh rock fragments ∼3 mm³ in size were prepared by crushing, washing in deionized water, and handpicking. Together with the neutron fluence monitor Fish Canyon sanidine, samples were loaded into precise positions within 18 mm Al disks, stacked, wrapped in Al foil, and encapsulated under vacuum in a quartz tube. The quartz tube was sealed into an Al canister and rotated at 1 rpm during neutron irradiation for 11 MWh in the central thimble position of the USGS TRIGA reactor (Dalrymple et al., 1981). Following irradiation, the samples and fluence monitors were loaded with tweezers into a stainless-steel sample holder and then placed into a laser chamber with an externally pumped ZnSe window, which was attached to a custom-built ultrahigh-vacuum extraction line. The volume of the mostly stainless-steel vacuum extraction line, including a cryogenic trap operated at −130 °C and two SAES™ GP50 getters (one at room temperature and one operated at 2 A [ampheres]), is estimated at ∼450 cm³. A combination of turbo molecular pumps and ion pumps maintained steady pressures within the extraction line of <1.33 × 10⁻⁷ Pa. Samples were incrementally heated in steps of 90 s by controlled power output of a 50 W CO₂ laser equipped with a beam-homogenizing lens, resulting in uniform energy over the entire sample surface. The reported incremental heating data represent results from individual mineral grains. During laser heating, any sample gas released was exposed to the cryogenic trap and was further purified for an additional 120 s by exposure to both the cryogenic trap and the SAES getters. The sample gas was expanded into a Thermo Scientific ARGUSVI™ mass spectrometer, and argon isotopes were analyzed simultaneously using four Faraday detectors (⁴⁰Ar, ³⁹Ar, ³⁸Ar, ³⁷Ar) and one ion counter (³⁶Ar). Following data acquisition of 10 min, time zero intercepts were fit to the data (using parabolic and/or linear best fits) and corrected for backgrounds, detector intercalibrations, and nucleogenic interferences. The mass spectrometer computer program written by A. Deino of the Berkeley Geochronology Center was used for data acquisition, age calculations, and plotting. All ⁴⁰Ar/³⁹Ar ages reported in Table 2 are referenced to an age of 28.201 ± 0.046 Ma for the Fish Canyon sanidine (Kuiper et al., 2008), the decay constants of Min et al. (2000), and an atmospheric ⁴⁰Ar/³⁶Ar ratio of 298.56 ± 0.31 (Lee et al., 2006). Laser fusion of >10 individual Fish Canyon Tuff sanidine crystals at each closely monitored position within the irradiation package resulted in neutron flux ratios reproducible to ∼0.25% (2σ). Isotopic production ratios were determined from irradiated CaF₂ and KCl salts, and the following values were measured for this study: (³⁶Ar/³⁷Ar)_Ca = (2.4 ± 0.05) × 10⁻⁴; (³⁹Ar/³⁷Ar)_Ca = (6.59 ± 0.10) × 10⁻⁴; and (³⁸Ar/³⁹Ar)_K = (1.29 ± 0.03) × 10⁻². Cadmium shielding during irradiation prevented any measurable (⁴⁰Ar/³⁹Ar)_K. The ⁴⁰Ar/³⁹Ar plateau ages (and uncertainties) are considered to be the best estimate of the cooling age of the minerals and were calculated from samples if three or more consecutive heating steps released >50% of the total ³⁹Ar and also had statistically (2σ) indistinguishable ⁴⁰Ar/³⁹Ar ages.

The analytical techniques used for Pb-Sr-Nd isotopic systematics on whole-rock powders at the USGS were similar to those reported by Tatsumoto and Unruh (1976) and more recently by Premo and Loucks (2000) and Premo and Taylor (2010). The whole-rock powders were dissolved in 7 mL perfluoroalkoxy Teflon^TM vials with ultrapure concentrated HF + HNO₃ while heating to ∼150 °C on a hot plate for at least 5 d. Lead was extracted from the dissolved effluent using AG1-X8 anion-exchange resin in Teflon columns and a very dilute HBr medium. The Pb laboratory contamination (blank) varied between 60 and 300 pg during the analytical session. Strontium and rare earth elements (REEs) were separated from the rest of the sample using a 30-mL-volume cation-exchange column (Birck and Allégre, 1978), while Sm was separated from Nd using the α-isobutyric method of Lugmair et al. (1975). Sr and Nd blanks were less than 300 pg each.

Pb, Sr, and Nd isotopic measurements were conducted using a Triton™ (Thermo-Fisher Scientific) nine-collector mass spectrometer, which was operated with a 10 kV acceleration voltage and 10¹¹ Ω resistors for the Faraday cups. Pb residues were redissolved in H₃PO₄ and loaded onto single Re filaments. The measured Pb isotopic ratios were corrected using the algorithms and programming of Ludwig (1980, 1985) for laboratory blanks with measured compositions of ²⁰⁶Pb/²⁰⁴Pb = 19.34 ± 0.53, ²⁰⁷Pb/²⁰⁴Pb = 15.53 ± 0.08, and ²⁰⁸Pb/²⁰⁴Pb = 38.11 ± 0.20, and instrumental mass fractionation (0.08% ± 0.03% per a.m.u.) as determined from multiple runs of the NIST Pb standards SRM-981 and SRM-982.

Between 500 and 1000 ng aliquots of Sr and Nd were loaded separately on outgassed Re filaments with a Ta₂O₅ activator. Within-run 1 standard error (SE) values for conventionally fractionation-corrected isotope ratios using accepted ⁸⁸Sr/⁸⁶Sr = 8.375209 typically were ∼0.0004% for ⁸⁷Sr/⁸⁶Sr and ∼0.002% for ⁸⁴Sr/⁸⁶Sr in unspiked samples. Unspiked NIST SrCO₃ standard SRM-987 was run after each five unknowns, and its average value was used to determine a normalization factor (F) for measured ⁸⁷Sr/⁸⁶Sr values as F = (⁸⁷Sr/⁸⁶Sr_Measured)/0.710248. The long-term weighted average for the NIST-corrected ⁸⁷Sr/⁸⁶Sr value in the USGS carbonate standard EN-1 (giant clam shell) was 0.709176 ± 0.000002 (n = 31, mean square of weighted deviates [MSWD] = 0.82, probability of fit = 0.74). Measured ¹⁴³Nd/¹⁴⁴Nd data were normalized to ¹⁴⁶Nd/¹⁴⁴Nd = 0.7219 and monitored for instrumental bias using the JNd-1 Nd standard (Tanaka et al., 2000). A mean ¹⁴³Nd/¹⁴⁴Nd value of 0.512087 ± 0.00006 (2SE%) was obtained during the same runs, and the average of two JNdi-1 analyses was 0.512085 ± 0.0006 (2SE%).

Hf and duplicate Pb isotope analyses were performed at the Ecole Normale Supérieure in Lyon (ENSL), France, on the same whole-rock powders as those used for USGS chemical and isotopic analyses. In this case, an ∼600 mg aliquot of whole-rock powder was leached in high-purity 6 M HCl in stages alternating between heating to 125 °C and ultrasonication. The total duration of leaching was 50 min. The leachate was decanted, and the residue was washed repeatedly in ultrapure water before digestion in double-distilled concentrated HNO₃ and HF in proportions of 1:3. After evaporation, Hf was leached from the digested residue using high-purity concentrated HF and separated first from the remaining sample matrix and then from Ti using a two-stage chromatographic procedure (Blichert-Toft et al., 1997). The Ca-Mg fluoride precipitate remaining after the HF-leaching step was dissolved in double-distilled 6 M HCl, evaporated to dryness, and then converted to bromides. Pb was separated from this solution on a 500 µL AG1-X8 anion-exchange column using dilute HBr to elute the sample matrix and 6 M HCl to collect the Pb.

Isotopic analyses were performed at ENSL on a Nu Plasma 500 HR multiple-collector–inductively coupled plasma–mass spectrometer. Groups of two to four Hf isotope analyses were bracketed by analyses of the JMC-475 Hf standard. Results were normalized for instrumental mass fractionation using an exponential law relative to ¹⁷⁹Hf/¹⁷⁷Hf = 0.7325. The JMC-475 Hf standard run in alternation in the session with the unknowns yielded ¹⁷⁶Hf/¹⁷⁷Hf = 0.282158 ± 0.000004 (2 standard deviation [sd]; n = 11). Pb was analyzed using Tl doping (5 ppb) and sample-standard bracketing (White et al., 2000), wherein the NIST 981 Pb standard was run systematically after every second sample. Analyses were normalized to the values of Eisele et al. (2003). External 2σ reproducibility of ²⁰⁶Pb/²⁰⁴Pb, ²⁰⁷Pb/²⁰⁴Pb, and ²⁰⁸Pb/²⁰⁴Pb was ±100–200 ppm (or 0.01%–0.02%), while that of ²⁰⁷Pb/²⁰⁶Pb and ²⁰⁸Pb/²⁰⁶Pb was ±50 ppm (or 0.005%). Internal run errors on all Pb isotope ratios were better than the external reproducibility. Hf and Pb blanks were <20 pg. The Sr, Nd, Hf, and Pb isotope data are listed in Table 3.

Petrographic inspection of the samples revealed fresh basalts with phenocrysts of olivine and clinopyroxene in a matrix of plagioclase and iron-titanium oxides, and some samples preserved fresh glass and vesicles (Fig. 6). Major- and trace-element compositions of samples determined from this study from both the Toprakkale and Karasu volcanic fields support previous analyses (e.g., Polat et al., 1997; Yurtmen et al., 2000; Alıcı et al., 2001; Nikogosian et al., 2018; Oyan, 2018), in that many samples were alkaline (Fig. 4). CIPW normative mineralogy calculations indicated that the basanites are nepheline-normative. There is a clear divergence from mostly subalkaline rocks in the Karasu volcanic field to mostly alkaline rocks in the Toprakkale volcanic field, and basanite is only reported from the Toprakkale volcanic field (this and previous studies). Relative to primitive mantle, all samples from this and other studies have compositions that are enriched in LILEs and LREEs (Fig. 7), and that share more similarities to OIB. Plots of REE relative to chondrites normalized to trivalent cations, rather than normalized abundances, result in smooth curves for which their slope and curvature provide a new means of basalt discrimination (O’Neill, 2016). A plot of REE data following this approach from both the Toprakkale and Karasu volcanic fields (Fig. 8) shows that the Toprakkale and Karasu samples are largely distinct from MORB and plot in regions consistent with garnet-bearing OIB (O’Neill, 2016). Moreover, the data for these two volcanic fields plot as largely separate fields from each other, with limited overlap, consistent with the separate fields defined by their alkaline versus subalkaline compositions (Fig. 4).

The ⁴⁰Ar/³⁹Ar geochronological data from this study constrain alkalic basalt volcanism in the Toprakkale and Karasu volcanic fields to between 2.63 ± 0.17 Ma and 46 ± 13 ka (Fig. 9). The Delihalil volcano is the most prominent edifice in the Toprakkale volcanic field, with ⁴⁰Ar/³⁹Ar plateau ages (2σ errors throughout) of 46 ± 13 ka (R12-OS1) and 50 ± 20 ka (R12-OS3) determined from basanites collected near and below the summit, respectively. The young Delihalil volcanic edifice erupted onto alkaline basalt lavas dated at 750 ± 90 ka (R12-OS4), with similar ages of 810 ± 60 ka (R12-OS5) and 780 ± 60 ka (R12-OS6) determined from samples collected along the northern margin of the Gulf of İskenderun; all three of these samples have identical geochemical and petrographic characteristics, despite being collected 30 km apart.

In the Karasu volcanic field, the basalt flow (R12HA-01) exposed near the village of Hassa produced a ⁴⁰Ar/³⁹Ar plateau age of 2.63 ± 0.17 Ma, and stratigraphically overlying basalt (R12HA-02) produced the youngest basalt dated from the Karasu volcanic field with a ⁴⁰Ar/³⁹Ar plateau age of 52 ± 16 ka. The lava flow toward the northern end of the Karasu volcanic field near the village of Fevzipaşa recorded a ⁴⁰Ar/³⁹Ar plateau age of 230 ± 30 ka (R12HA-03).

The two basanites analyzed from the Toprakkale volcanic field have the following isotopic ratios: ⁸⁷Sr/⁸⁶Sr of 0.703070 and 0.703136; ¹⁴³Nd/¹⁴⁴Nd of 0.512931 and 0.512893 (ε_Nd = +5.8 to +5.0); ¹⁷⁶Hf/¹⁷⁷Hf of 0.283019 and 0.282995 (ε_Hf = +8.7 to +7.9); ²⁰⁶Pb/²⁰⁴Pb of 19.087 and 19.155; and ²⁰⁸Pb/²⁰⁴Pb of 38.861 and 38.915, where the Pb isotope values are averages of those obtained at the USGS and ENSL. The ¹⁴³Nd/¹⁴⁴Nd and ⁸⁷Sr/⁸⁶Sr values for Toprakkale basanites (Parlak et al., 2000; Bağcı et al., 2011; this study) are among the highest and lowest values, respectively, when compared to Miocene mafic Anatolian lavas (cf. compilations by McNab et al., 2018; Uslular and Gençalioğlu-Kuşcu, 2019a; Fig. 10 herein). They also overlap with those of the most depleted mantle-like basalts in northern and central Arabia (Bertrand et al., 2003; Ma et al., 2011), albeit displaced to slightly lower ⁸⁷Sr/⁸⁶Sr. Plots of ¹⁷⁶Hf/¹⁷⁷Hf and ¹⁴³Nd/¹⁴⁴Nd lie within the distribution of oceanic mantle compositions defined by MORB and OIB (deviation from the mantle array, Δε_Hf, of –2; Fig. 10), and they resemble those of late Miocene basalts of western Anatolia (Aldanmaz et al., 2015). Pb isotope compositions lie close to, but above, the Northern Hemisphere reference line (NHRL; Hart, 1984) in ²⁰⁷Pb/²⁰⁴Pb-²⁰⁶Pb/²⁰⁴Pb and ²⁰⁸Pb/²⁰⁴Pb -²⁰⁶Pb/²⁰⁴Pb space (Δ7/4 deviations of ∼5 and Δ8/4 of ∼15; Fig. 10), similar to those of Miocene basalts from the northern and central Arabia plate as well as to those of late Miocene and Quaternary basalts in western Anatolia.

Quaternary alkaline Toprakkale basalts have somewhat higher ¹⁷⁶Hf/¹⁷⁷Hf ratios (0.282966–0.283019; ε_Hf = +6.9 to +8.7) than alkaline Karasu basalts (0.282837–0.282965; ε_Hf = +2.3 to +6.8) and somewhat higher ¹⁴³Nd/¹⁴⁴Nd ratios (0.512781–0.512866 and ε_Nd = +2.8 to +4.5, versus 0.512648–0.512806 and ε_Nd = +0.2 to +3.3). Plots of ¹⁷⁶Hf/¹⁷⁷Hf and ¹⁴³Nd/¹⁴⁴Nd values lie within the oceanic mantle array (Δε_Hf of ∼–2.1 to +0.4; Fig. 10), in this respect resembling late Miocene basalts of western Anatolia (Aldanmaz et al., 2015), in contrast to contemporaneous near-primary basalts in central Anatolia, which have similar Nd isotope ratios that are displaced to ε_Hf values ∼6 units higher (Reid et al., 2017). The ²⁰⁶Pb/²⁰⁴Pb values for alkaline Toprakkale basalts are relatively invariant (19.025 ± 0.001), whereas those for Karasu basalts vary from 18.800 to 19.324; ²⁰⁸Pb/²⁰⁴Pb ratios overlap but are displaced to lower values in Toprakkale basalts (38.978–39.103) relative to Karasu basalts (39.101–39.219). More generally, Pb isotope ratios extend to values well above the NHRL in ²⁰⁷Pb/²⁰⁴Pb-²⁰⁶Pb/²⁰⁴Pb and ²⁰⁸Pb/²⁰⁴Pb -²⁰⁶Pb/²⁰⁴Pb space (Δ7/4 up to 17 and Δ8/4 up to 75; Fig. 10). Although there is some overlap in radiogenic isotope ratios, the combined data highlight differences between the two Quaternary volcanic fields, with the Toprakkale basanites displaying the most restricted and depleted mantle-like compositions.

The Hf and Pb isotope compositions obtained for Toprakkale basanites, considered together with Sr and Nd isotope data, provide important new constraints on their mantle source(s). New and published Sr isotope data for Toprakkale basanites exhibit minor variability (average ⁸⁷Sr/⁸⁶Sr = 0.70315 ± 0.00007 [1 sd]; n = 14) despite approximately twofold variations in La/Nb, La/Yb, and Ba/Nb. Helium isotope values averaging R/R_a = 7.6 ± 0.3 obtained for five Toprakkale basanites are indistinguishable from those of MORB mantle (Italiano et al., 2017). More generally, the isotopic features of the Toprakkale basanites, and specifically the Hf and Pb isotopic data reported here, lie within mantle arrays defined by the isotopic characteristics of oceanic basalts, and close to those with a common mantle component (“C”; Hanan and Graham, 1996). As noted previously by several authors (Polat et al., 1997; Alıcı et al., 2001; Yurtmen et al., 2002; Oyan, 2018; Nikogosian et al., 2018), the incompatible element characteristics of the basanites resemble those of OIB (Fig. 10). Thus, both isotopic and incompatible element characteristics of the Toprakkale basanites indicate that they are OIB-type, asthenosphere-derived melts largely unmodified by crustal contributions.

Depths of peridotite-derived melt equilibration can be estimated using olivine-orthopyroxene-melt barometry (Lee et al., 2009; Plank and Forsyth, 2016). Relevant conditions for melt equilibration are assumed to be: (1) primary melts in equilibrium with an Fo₈₉ source (a fairly conservative estimate judging by Arabian mantle xenoliths; Stern and Johnson, 2010); (2) minimum water contents given by H₂O (ppm) = 200•Ce (ppm), a relation typical for oceanic basalts (Reid et al., 2017); and (3) Fe²⁺/ΣFe = 0.18 and 0.16 for the basanites and basalts, respectively, extrapolated based on estimated water contents (Kelley and Cottrell, 2009). Barometer uncertainty is ±8 km (Plank and Forsyth, 2016). Basanites were filtered for Mg# >55 and CaO/Al₂O₃ >0.6, since those with lower values are more likely to have experienced clinopyroxene fractionation (Italiano et al., 2017).

Basanites could have been derived by partial melting of garnet pyroxenite (Keshav and Gudfinnsson, 2004), hornblende peridotite (Pilet et al., 2008), or, in this study area specifically, phlogopite-bearing garnet peridotite (Oyan, 2018). The latter source was identified for Toprakkale basalts based on the apparent buffering of alkalis and alkaline earth elements relative to LREEs during melting. Pressure estimates for basanite melt equilibration with peridotite derived from new and published chemical analyses average 2.7 ± 0.2 GPa (n = 31; data in Supplemental Table S1¹), corresponding to a mean depth estimate of 91 ± 3 km, assuming an average crustal thickness of 30 km as indicated by seismic data (Abgarmi et al., 2017). Inferred water contents are 1.0–1.6 wt% but could be higher if the source is more hydrous than typical suboceanic mantle. Increasing water contents by a factor of two, for example, reduces the depth estimate by only 2 km. Allowing for an uncertainty in Fe²⁺/ΣFe of ±0.05 (likely a maximum) would increase the depth uncertainty by ±4 km. The ∼90 km depth estimate for melt equilibration is consistent with REE and HFSE evidence for Toprakkale basanite generation in the garnet stability field (Fig. 8; Oyan, 2018; Nikogosian et al., 2018). The corresponding equilibration pressures are within 0.3 GPa of values reported previously (Çoban, 2007; Reid et al., 2017; Nikogosian et al., 2018), albeit for a larger number of samples and with differing assumptions about primary melt compositions. These pressure estimates are considerably lower than the value of 3.7 GPa estimated by Oyan (2018) using the barometer formulation of Lee et al. (2009). Much of this difference is probably attributable to different barometer formulations.

The isotopic and enriched trace-element compositions of basalts from the Toprakkale and Karasu volcanic fields have been variously ascribed to subduction-modified sources (Yurtmen et al., 2000; Italiano et al., 2017), mantle signatures overprinted by crustal contributions (Alıcı et al., 2001; Oyan, 2018), and mantle source heterogeneity (Nikogosian et al., 2018). The elevated ²⁰⁷Pb/²⁰⁴Pb and ²⁰⁸Pb/²⁰⁴Pb values of the basalts could, by analogy with Miocene central Anatolian basalts, reflect metasomatism of the lithosphere by a subduction-related component. However, the basalts are largely characterized by positive HFSE anomalies (e.g., La/Nb_N = 0.6–0.9; Fig. 7), in contrast to the pronounced negative anomalies in relatively primitive subduction-influenced Miocene basalts from central Anatolia (La/Nb_N mostly >2; Reid et al., 2017; Uslular and Gençalioğlu-Kuşcu, 2019b). The oceanic mantle array–like Hf-Nd isotopic characteristics of Toprakkale and Karasu basalts also contrast with Hf isotope signatures that are displaced from the oceanic mantle array in central Anatolian basalts (Reid et al., 2017). Instead, incompatible element patterns for the A³ triple junction basalts are broadly similar to, albeit at lower abundances than, the basanites, as permitted if the basalts were generated by a greater degree of melting (Bağcı et al., 2011; Oyan, 2018).

The variations in new Hf, Δ7/4, and Δ8/4 isotope data along with new and previously determined Sr and Nd isotope ratios additionally require either source heterogeneity and/or contributions from more than one source. Paired isotope ratios from both volcanic fields define broadly linear arrays that trend from the basanites to the most evolved basalt (Fig. 10), despite occurring in two spatially distinct areas. The more radiogenic isotopic compositions of the alkaline Toprakkale basalts relative to the basanites could be the result of 2%–5% crustal assimilation (Oyan, 2018), with olivine-normative basalts displaced even farther in Nd and Hf isotope space. The somewhat different isotopic trends defined for the two volcanic fields would require different crustal end members if they resulted from crustal assimilation.

In detail, interelement ratios for the basalts (compiled data) and olivine-hosted melt inclusions (Nikogosian et al., 2018) in the Toprakkale volcanic field are more MORB-like than the basanites, an observation that is not easily reconciled with crustal assimilation. This has led to the hypothesis that at least some of the trace-element heterogeneity exhibited by the basalts is derived from a more MORB-like mantle source than that of the basanites (Nikogosian et al., 2018), although the more radiogenic Sr, Nd, and Hf isotope signatures of the basalts do not support this interpretation.

Pressure estimates obtained for the Toprakkale and Karasu basalts, assuming melt equilibration with peridotite, average 1.9 ± 0.1 GPa (n = 26) and 1.8 ± 0.3 GPa (n = 15), respectively (data filtered as described above). These pressures correspond to average depth estimates of 64 ± 4 km and 63 ± 8 km, or ∼30 km shallower than those estimated for the basanites. Such pressures are similar to those of 1.8–2.3 GPa and 2.05–2.17 GPa obtained in previous investigations (Reid et al., 2017; McNab et al., 2018; Nikogosian et al., 2018) and imply depths in the spinel peridotite stability field, contributions from which have been identified by other authors (Italiano et al., 2017; Oyan, 2018). Melting in the presence of garnet is also apparent from REE and HFSE systematics (Bağcı et al., 2011; Nikogosian et al., 2018) and was likely inherited as melt upwelled from depth; the “garnet-signatures” would not be erased by equilibration or mixing with more shallowly derived melts. Inferred water contents are from 0.6 to 1.4 wt%.

Data from this and previous studies are consistent with a model in which both the Toprakkale and Karasu volcanic fields were derived by partial melting of an oceanic basalt–like asthenospheric mantle source (basanites) with possible contributions from lithospheric mantle ± crustal source (basalts). However, observed geochemical and isotopic variations between the volcanic fields may also be impacted by the regional tectonic complexities of an evolving triple junction, perhaps explaining why alkaline basalts occur in both volcanic fields while basanites have only been identified in the Toprakkale volcanic field. Moreover, the melting depths inferred for the Toprakkale basanites are not only the deepest from either volcanic field, but they are also restricted to an area of transtension between the major strands of the East Anatolian fault zone. The basanites are the most depleted mantle-like rocks in terms of Sr, Nd, and Hf isotopic data, with low overall variability compared to the less mafic rocks from both the Karasu and Toprakkale volcanic fields. The compositions of the basanites support relatively deep melting (mostly 80–105 km, averaging ∼90 km), in contrast to shallower melting inferred for the Toprakkale alkaline basalts (mostly 60–70 km; Polat et al., 1997; Yurtmen et al., 2000; Bağcı et al., 2011; Oyan, 2018).

Significant variations in crustal and upper-mantle seismic structure are found near the A³ triple junction. Using P-wave receiver functions, Abgarmi et al. (2017) imaged a significant decrease in crustal thickness at the transition from the central Taurus Mountains (>40 km) to the Adana Basin (∼30 km). This thinning is likely due to the extensional component of the dominantly transtensional A³ triple junction. By combining Rayleigh wave dispersion measurements with the receiver function measurements of Abgarmi et al. (2017) and jointly inverting to solve for the three-dimensional shear-wave velocity structure of central Anatolia, Delph et al. (2017) observed a widespread and vertically extensive fast shear-wave velocity anomaly beneath the central Taurus Mountains and western Adana Basin, which they interpreted as the underthrusted Cyprean slab. These fast velocities abruptly terminate near the central Adana Basin, transitioning to very slow upper-mantle shear velocities beneath the Iskenderun Basin and areas farther east (<4.2 km/s at >60 km; Fig. 11). These observations are consistent with thin crust (∼30 km) and thin lithospheric mantle (∼30 km) underlain by a hot and shallow melt-bearing asthenosphere beneath the A³ triple junction and extending beneath the Arabia plate.

Excluding the basanites, melt equilibration depths for Quaternary basaltic magmatism in the Toprakkale and Karasu volcanic fields agree with the seismically imaged lithosphere-asthenosphere boundary as defined by a sharp transition at ∼60 km to very slow upper-mantle shear velocities (Fig. 11). A general (mostly within 5 km) co-occurrence between melt equilibration and lithosphere-asthenosphere boundary depth from seismic imaging has also been observed elsewhere in Turkey (Reid et al., 2017) and in the western United States (Reid et al., 2012; Plank and Forsyth, 2016). The correlation can be explained by partial melts being able to segregate from upwelling mantle as the relatively rigid lithospheric lid is approached. Chemical and isotopic heterogeneity exhibited by these basalts would therefore reflect hybridization of asthenosphere-derived melts with lithosphere that is being thermally and chemically eroded, except where arising from crustal contamination.

The relatively thin lithosphere (∼50–60 km) below the İskenderun Basin extends farther to the east beneath the Arabian foreland (Angus et al., 2006; Delph et al., 2017) and is much thinner than the Arabian Shield farther to the south, which has a lithospheric thickness of >100 km (Hansen et al., 2007). The comparatively thin Arabian foreland lithosphere may represent its original thickness as a continental passive margin prior to collision. Another possibility is that thinning of the Arabian lithospheric mantle is occurring through an asthenospheric mantle erosion process, which could be linked to crustal tectonics and magmatism along the East Anatolian fault zone. For example, northward mantle flow beneath the Arabia plate (Becker and Faccenna, 2011) coupled with the abrupt decrease in lithospheric thickness from >100 km to ∼60 km could lead to shear-induced, edge-driven convective erosion of the lithospheric mantle (e.g., Kaislaniemi and van Hunen, 2014). The convective removal of lithospheric mantle due to significant lithosphere-asthenosphere boundary topography has also been invoked in Australia as well as near the edges of other subduction zones (Davies and Rawlinson, 2014; Levander et al., 2014). Such edge-driven convection may provide a mechanism for adiabatic decompression melting of mantle that has entrained blocks of detached lithosphere, explaining the occurrence of localized small-volume basalts along the high-angle plate-boundary faults (East Anatolian fault zone) and also their geochemical and isotopic variations. Similarly, the Quaternary to recent mafic magmatism in the Karacadağ volcanic field in the Arabian foreland farther east (e.g., Lustrino et al., 2010; Ekici et al., 2014; Ekici and Macpherson, 2019) may also have resulted from such a process, supporting regional-scale mantle lithospheric erosion of the Arabian foreland, extending into the A³ triple junction. Indeed, magmatism at the Karacadağ volcanic field extends back to ca. 17 Ma (Ekici and Macpherson, 2019), showing a minor southeastward migration through time along a major regional fault (Keskin et al., 2012), consistent with the convective removal of the lithospheric mantle.

The considerably deeper melt equilibration depths obtained here and in previous investigations of the Toprakkale basanites (∼90 km; Reid et al., 2017; Oyan, 2018; Nikogosian et al., 2018) appear to require a different process of melt generation. The basanites could be melts of an oceanic basalt–like sublithospheric mantle source, as supported by the Pb and Hf isotope data reported here and recognized previously; they appear to define one end member for the basalts (Fig. 10). Basanite melts could simply have separated from mantle upwelling along an adiabat at depths greater than the lithosphere-asthenosphere boundary, as observed in the western Grand Canyon region of the western United States and proposed for the Sivas region of central Anatolia (Plank and Forsyth, 2016; Reid et al., 2019).

A signature feature of the basanites is their noticeable depletions of alkalis and Pb relative to similarly incompatible elements, a signature that is also observed in other Toprakkale basanites attributed to HIMU-OIB mantle sources (where HIMU stands for high µ = ²³⁸U/²⁰⁴Pb; Fig. 7; Bağcı et al., 2011; Oyan, 2018; Nikogosian et al., 2018). More subtle deviations from HIMU are also present in some samples of this study with elevated Sr and P, and sometimes lower Zr and Hf relative to nominally similar incompatible elements. Basanites from oceanic islands that exhibit similar geochemical characteristics (e.g., St. Helena, Cook-Austral islands) are inferred to have been derived from upwelling mantle plumes. The lack of high (>20) ²⁰⁶Pb/²⁰⁴Pb (or HIMU) in the Toprakkale basanites would reflect insufficient time for radiogenic ²⁰⁶Pb ingrowth after recycling of oceanic crust into the plume source. Alternative origins have been proposed for HIMU signatures in the Toprakkale volcanic field and other localities in the Anatolia-Arabia region via melting of hydrous metasomatic veins impregnated within, or at the base of, the lithosphere or by reactive porous flow of melts migrating through the lithosphere (Bağcı et al., 2011; Aldanmaz et al., 2006; Oyan, 2018).

In this study, the seismologically imaged depth of the lithosphere-asthenosphere boundary directly below the Toprakkale volcanic field is similar to melt equilibration depths for the basalts (∼60 km). We therefore considered three alternatives to explain why melt extraction of the basanites may have taken place at greater depths than the lithosphere-asthenosphere boundary: (1) Geochemical signatures of the nearby Cyprean slab lithospheric mantle (Fig. 11) were incorporated near the source; (2) the basanite melts were generated by heating and partial melting of suspended blocks within the asthenospheric mantle that delaminated/dripped from the base of the lithospheric mantle; or (3) the pressure estimates for basanite melt equilibration are invalid. Regarding the last hypothesis, melt equilibration depths could be in error if the source lithology lacked orthopyroxene and/or olivine (e.g., if it were a pyroxenite) or if there were considerable pyroxene fractionation after melt extraction. Previous investigations (e.g., Oyan, 2018) and our data filtering suggest that neither scenario applies here.

Basanites are only observed in the Toprakkale volcanic field, which is separated from the Karasu volcanic by the East Anatolian fault zone and the Amanos mountains (Fig. 2). Assuming that the melt generation depths of ∼90 km are reliable, the Toprakkale basanites originated from ∼30 km below the underlying seismically imaged lithosphere-asthenosphere boundary (Fig. 11). Irregular plate geometries at depth could induce small-scale convection cells, perhaps leading to localized heating and partial melting of the neighboring Cyprean slab lithospheric mantle. Because the Cyprean slab at depth occurs below the western Adana Basin (Delph et al., 2017), just west of the A³ triple junction, melt generation and migration near the adjacent Cyprean slab edge may follow permeable lithospheric-scale pathways associated with this plate boundary as has been proposed for oceanic transform faults (Hékinian et al., 1995). Although there is no a priori reason that such a source would impart HIMU-like trace-element signatures on the basanites, we cannot discount a hypothesis wherein the basanites are partial melts derived from the Cyprean slab lithospheric mantle at depths of ∼90 km. Another mechanism for explaining the deeper melting depths of the basanites is that they represent melts sourced from lithospheric blocks within the asthenosphere, for example, remnants of the Bitlis slab or blocks detached via convective removal from the base of the Anatolian lithosphere. This latter hypothesis is consistent with the abrupt contrast in seismically imaged lithospheric thicknesses and the spatial distribution of Quaternary basalts farther east within the Arabian foreland (e.g., Chorowicz et al., 2005; Stern and Johnson, 2019), and it may be the process that is largely responsible for the shallow lithosphere-asthenosphere boundary beneath eastern Anatolia. In either case, the basanites appear to include some contribution of partial melting of lithospheric mantle at ∼90 km depth that was either “attached” to the Cyprean slab, or in the form of “detached” blocks foundered within the asthenospheric mantle.

The combined K-Ar and ⁴⁰Ar/³⁹Ar ages indicate that the initial stages of volcanism began in the Toprakkale and Karasu volcanic fields with first appearances of alkaline volcanism at 2.25 ± 1.56 Ma (2 sd; Arger et al., 2000) and 2.63 ± 0.17 Ma (this study), respectively. Basaltic volcanism in the Karasu valley began nearly 1 m.y. earlier than previously known (1.57 ± 0.08 Ma) from the study by Rojay et al. (2001). The new ⁴⁰Ar/³⁹Ar data from this study support a semicontinuous 2.5 m.y. history of volcanism within the Karasu valley, consistent with borehole data showing intercalated basalt and fluvial-lacustrine sedimentation with significantly more sedimentation occurring along the western margin of the valley (Rojay et al., 2001). The 46 ± 13 ka basanite lava erupted from the Delihalil volcano, and the 50 ± 20 ka alkaline basalt lava flow in the Karasu valley also demonstrate that volcanism remains intermittently active in both the Toprakkale and Karasu volcanic fields. The initial period of volcanism in both volcanic fields was limited, as only one sample from each area records evidence of volcanism prior to 2 Ma, providing the earliest evidence of magmas erupted along developing fault boundaries (Fig. 12). However, between 2 and 1 Ma, more widespread basaltic volcanism was localized along the Karasu fault zone and along the Yumurtalık fault zone in the Karasu and Toprakkale volcanic fields, respectively. Over the last 1 m.y., volcanism appears to have intensified and produced most of the exposed volume of volcanic rocks within both volcanic fields. These post–1 Ma volcanic rocks are focused within larger eruption centers such as the Delihalil volcano in the Toprakkale volcanic field and a number of smaller volcanic cones within the Karasu volcanic field along clearly defined fault segments. Although poorly constrained, there is some indication that volcanism within the Karasu volcanic field may have migrated toward the Arabian plate over the last 200 k.y. (Fig. 12).

Important questions in understanding the tectonic evolution of the A³ triple junction are: (1) When did the modern fault systems bounding the triple junction become active, and (2) can the Quaternary lavas associated with the fault system provide chronological constraints on fault dynamics? Given that Quaternary volcanism is clearly localized along fault segments and releasing and restraining bends of the East Anatolian fault zone (Tatar et al., 2004; Kavak et al., 2009), it seems likely that volcanism postdated initiation of major strike-slip activity related to Anatolian tectonic escape, and, based on our new results, associated magmatism did not occur until 2.63 ± 0.17 Ma. While this provides a lower limit on the tectonic escape of the Anatolia microplate along the western terminus of the East Anatolian fault zone, others have suggested initiation of the East Anatolian fault zone by ca. 4 Ma and possibly as early as 7 Ma near the Gulf of İskenderun (e.g., Westaway, 2003, 2004; Seyrek et al., 2008). A reasonable upper limit to the inception of the East Anatolian fault zone is probably a few million years greater than the ⁴⁰Ar/³⁹Ar basalt ages, judging from estimates of the time required for initial melt generation and transfer through the crust (Cesare et al., 2009; Ferrari, 2004; Karakas et al., 2017). These estimates agree with independent geologic evidence supporting the transition from regional transpression to transtension and from “proto-escape” to active escape of the Anatolia plate along the East Anatolian fault zone around 5 Ma (e.g., Yılmaz et al., 2006).

Prior to the late Miocene, a subduction/collision plate boundary existed between Nubia (Africa-Arabia) and Eurasia (Anatolia), and this is expressed geologically in the field by imbricated Kızıldağ and Baer Bassit ophiolite sequences (Bitlis suture). This plate boundary was disrupted in the early–mid-Miocene by an apparent tear in the downgoing slab (Nubia), and eventual breakoff of the Bitlis slab segment (Portner et al., 2018; Reid et al., 2019). Today, this boundary is represented by subduction of discrete segments of the Africa plate beneath Anatolia, as trench retreat is active along the Hellenic and Cyprus trenches, respectively. Based on seismic tomography in the eastern Mediterranean (e.g., Biryol et al., 2011; Portner et al., 2018), the following three distinct subduction/collision domains are observed: (1) Along the Hellenic arc, an intact slab extends continuously into the lower mantle below the Aegean; (2) in central Anatolia, subduction occurs near the Cyprus trench, and the slab appears to be mostly intact, although severely distorted with several apparent minor tears; and (3) northward of the Bitlis suture, which represents the oldest stage of the ongoing subduction/breakoff process, no slab is readily identifiable, and only relatively small, fast anomalies interpreted as detached or broken fragments of the Bitlis slab are visible in the upper mantle.

While tearing and breakoff of parts of the subducting Nubia slab were occurring at the former Bitlis subduction/collision zone, Arabia was beginning to separate and rotate away from Africa, and the northward propagation of the Red Sea rift ultimately led to the propagation of the Dead Sea fault zone into our study area by the late Miocene, just prior to the Pliocene formation of the East Anatolian fault zone (e.g., Reilinger et al., 2006; Bosworth and Stockli, 2016). Extrusion of the Anatolia plate since 5 Ma, currently moving at 21 mm yr^–1 (relative to a stable Eurasia) and partly driven by the 15 mm yr^–1 advance of Arabia and its collision with Eurasia along the Bitlis and Zagros sutures (e.g., Reilinger et al., 2006), results in a dynamic tectonic coupling along the more slowly (5 mm yr^–1) advancing African plate. The inception and continuing development of the A³ triple junction plate boundaries are intimately tied to the evolving Dead Sea–East Anatolian fault zone system and the ways in which these faults respond to deformation related to tectonic escape of the Anatolia plate. The oblique nature of the relative plate motions in the fault-fault-trench triple junction system is inherently unstable (Dewey et al., 1986) and will continue to cause the A³ triple junction to migrate until a more stable configuration is reached.

Although transfer of left-lateral deformation from the Dead Sea fault zone to the East Anatolian fault zone is fairly well accepted, and the East Anatolian fault zone represents a major component of the A³ triple junction, a less-well-known feature is the offshore linkage of the northern East Anatolian fault zone to the Kyrenia-Misis fault and the way in which the main strand of the East Anatolian fault zone links to the Cyprus trench (e.g., Duman and Emre, 2013). The exposed broad zone of distributed deformation, transtension, and basin development between the two main strands of the East Anatolian fault zone clearly continues offshore, represented by the Adana-Cilicia and İskenderun basins. Indeed, continuing rollback of the subducting Cyprean slab and associated trench retreat may explain some of the plate kinematics, including the location, possible migration, and occurrence of mafic magmatism associated with the A³ triple junction.

We suggest that initiation of the triple junction at a position near the village of Kahramanmaraş began around 5 Ma, and it subsequently migrated to the southwest to its current position near the Amik Basin (Fig. 13). This interpretation is consistent with an overall southwesterly shift in the locus of deformation along the East Anatolian fault zone toward areas containing young and shallow basins. This interpretation also implies that plate segmentation within the lithosphere may not correspond to exposed crustal sequences representing former plate positions. For example, the trend of pronounced seismicity west of Kahramanmaraş (e.g., Duman and Emre, 2013) could represent either a zone of previous weakness or one of incipient faulting. Coincident with the A³ triple junction migration, there was linked migration of the Africa-Anatolia plate margin and a jump of the Cyprean trench to its current southerly position (Fig. 13). Geologic evidence for such a model within the Adana Basin includes the abrupt termination of late Miocene compressional (Bitlis suture) sedimentation features and the establishment of Pliocene and younger high-angle translational faults and extensional/transtensional basins (e.g., Burton-Ferguson et al., 2005). Such a model is also consistent with a period of major Pliocene surface uplift within Cyprus (e.g., McCallum and Robertson, 1995; Kinnaird et al., 2011; Schildgen et al., 2012). This younger than 5 Ma southward trench migration could have been related to slab detachment through ongoing horizontal tearing and segmentation within the Cyprean slab (e.g., Portner et al., 2018). Cyprean slab detachment is thought to be a major factor driving the post–8 Ma surface uplift of the central Taurus Mountains (e.g., Schildgen et al., 2014; Meijers et al., 2018; 2020). The same or similar tectonic forces responsible for the partial uplift of Cyprus will ultimately lead to future uplift of the Eratosthenes Seamount.

Plate-boundary reorganization from 5 Ma to present would have modified the regional deformation and the state of stress and promoted the tapping of sublithospheric melt along lithospheric-scale shear zones associated with extension/transtension (Fig. 13). Cyprean slab rollback and the southwesterly migration of the A³ triple junction are not only consistent with establishment of the southern strand of the East Anatolian fault zone as a primary plate boundary, but they are also consistent with the spatial distribution of mafic volcanism. The spatially restricted occurrence of Quaternary basalts between the current and former positions of the A³ triple junction suggests that the migration of the triple junction, in this case associated with trench migration (or jump), triggered deeply sourced volcanism. Tectonic conditions favor continued migration of the A³ triple junction along the Dead Sea transform fault as rollback of the Cyprean slab proceeds and the Anatolia-Africa-Arabia plate boundaries continue to evolve.

From 5 Ma to the present day, the intersecting Africa-Anatolia-Arabia plate boundaries have adjusted to the tectonic escape of the Anatolia plate and rollback of the Cyprean slab. This process led to the initiation and subsequent migration of the A³ triple junction and development of major transform faults, broad zones of transtension, and localized volcanism. Quaternary basalts and basanites within the Toprakkale and Karasu volcanic fields are a consequence of these plate adjustments within this dynamic setting. Radiogenic isotope data (Sr, Nd, Hf, Pb) indicate that the basanites were derived from the most-depleted mantle source, but all samples indicate derivation from a garnet-bearing OIB-like source. The seismically imaged depth (∼60 km) of the lithosphere-asthenosphere boundary directly beneath the northwestern margin of the Arabia plate is consistent with the inferred depth of melting of all basalts. Basanites from the Toprakkale volcanic field, however, indicate a deeper (∼90 km) melting source, interpreted to have originated from either partial melting of the adjacent Cyprean slab lithospheric mantle, or melting of detached fragments of Arabian lithospheric mantle that foundered within the asthenosphere. A mafic lava with a ⁴⁰Ar/³⁹Ar plateau age of 2.63 ± 0.17 Ma from the Karasu volcanic field represents the earliest stage of volcanism associated with the developing A³ triple junction. This time period is also consistent with more regional data from the northwestern margin of the Arabia plate supporting an increase in magmatic activity from 2 Ma. Episodic eruption of lavas in the last 2.6 m.y. along discrete segments of the East Anatolian fault zone clearly demonstrates that volcanism is being localized along and/or between transform faults related to the A³ triple junction. The Dead Sea fault zone and East Anatolian fault zone plate boundaries are high-angle shear zones that likely penetrate into the lithospheric mantle, providing the “leaky” conduits along which melts infiltrate to reach the surface, analogous to the restricted volcanism along transform faults occurring in oceanic crust. The spatially restricted distribution of basalts proximal to the A³ triple junction indicates the potential for a causal relationship, because they occur within a zone of transtension near the triple junction and are not more widespread along the length of the East Anatolian fault zone. Tectonic conditions favor continued migration of the A³ triple junction as this motion is directly tied to the ongoing tectonic escape of Anatolia and rollback of the Cyprean slab.

This study was partially funded with collaborative support of the National Science Foundation (NSF) as part of the Continental Dynamics program (NSF EAR-1109762 and 1109826). We thank Nuretdin Kaymakçı for fruitful scientific discussions and introducing some of us to the field area. We also thank Gilles Brocard for maps and scientific discussions. Critical reviews and suggestions for improving the manuscript by Samuele Agostini, Richard Arculus, and David Moecher are gratefully acknowledged. Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the U.S. government.