Late Miocene igneous rocks of Samos: the role of tectonism in petrogenesis in the southeastern Aegean

Abstract Late Miocene igneous rocks of Samos, in the southeastern Aegean Sea, comprise monzodiorite and minor granite of the Katavasis complex, trachyte and rhyolite of the Ambelos volcanic centre, and bimodal basalt–rhyolite at basin margins. Six new K–Ar ages, together with existing geochronology and biostratigraphy, show that the Katavasis complex and Ambelos centre date from 10–11 Ma and basalt–rhyolite from 8 Ma, correlating with cooling ages for the Katavasis complex and an unconformity in the basin fill. Monzodiorite, granite, trachyte and basalt all have similar radiogenic isotopes. Monzodiorite and basalt have similar trace element compositions and could result from 5–10% partial melting of enriched garnet lherzolite in the subcontinental lithosphere. Variations in trace elements suggest that trachyte and monzodiorite evolved by fractional crystallization from a parental magma similar to the younger basalt. The Katavasis and Ambelos rocks were synchronous with regional extension and listric faulting, which created opportunities for mid-crustal magma chambers and magma fractionation. Basalt extrusion was synchronous with the onset of north–south strike-slip faulting, which permitted more rapid transfer of magma to the surface. Late Miocene strike-slip faulting propagated from north to south in western Anatolia and the southeastern Aegean Sea, providing pathways for different types of mantle melts.

Miocene igneous rocks of the island of Samos in the southeastern Aegean Sea (Fig. 1) are part of a series of late Miocene-Quaternary extension-related rocks in a continental back-arc setting in the Aegean Sea (Pe-Piper & Piper 2006). In northwestern Anatolia and the central Aegean Sea, to the north of Samos, is an extensive area of shoshonites of early Miocene age (Pe-Piper & Piper 1992;Aldanmaz et al. 2000). I-type late Miocene granitoids in the Cyclades (Altherr et al. 1982) were interpreted by Altherr & Siebel (2002) as a consequence of extension and upwelling asthenospheric melts. The Pliocene-Quaternary south Aegean arc includes subduction-related andesite -dacite. In western Anatolia (and the island of Patmos), alkaline basalts of late Miocene age and younger are widespread (Aldanmaz et al. 2000).
Three groups of Late Miocene igneous rocks are present in Samos (Fig. 2). An intrusive complex of monzodiorite and granitoid dykes crops out at Katavasis in western Samos (Mezger et al. 1985). Potassic trachytes and minor rhyolite form a volcanic complex near Ambelos (Theodoropoulos 1979). Rhyolites and basalts crop out along the margins of two Miocene basins, in the eastern and western parts of the island (Robert & Cantagrel 1979). This wide range of rock types provides an opportunity to determine the petrogenetic relationships between these rock types. This interpretation is placed within the general context of the Late Miocene of the southeastern Aegean Sea, in the light of the tectonic evolution of the Aegean back-arc area.

Regional setting
The Miocene igneous rocks of Samos closely resemble those of the nearby islands of Patmos and Kos to the south, and the Bodrum peninsula of Turkey just east of Kos. Rather similar rocks are also present at Urla, east of the Karaburun peninsula, to the north of Samos.
During the Miocene, palaeomagnetic data and stretching lineations show that the west Aegean block (Fig. 1), bounded by the Scutari -Pec line and the Mid-Cycladic lineament, rotated clockwise as a coherent block with respect to the southeastern Aegean (Walcott & White 1998). The southeastern Aegean appears to have rotated counterclockwise together with western Anatolia, although local clockwise rotation (e.g. at Karaburun) points to complex local tectonics (Kissel et al. 1987;Kondopoulou 2000). Yilmaz et al. (2000), from detailed mapping in western Anatolia, showed that north-south-trending grabens developed in the Early Miocene. North-south extension with detachment faulting began in the Late Miocene, leading by early Pliocene to the formation of east -west-trending grabens.
South of Samos, on the island of Patmos (Fig. 1), trachyte and rhyolite, with minor basalt of the Old Volcanic Series (OVS) date from 7.4 Ma. Trachyte flows and minor phonolite of the Intermediate Volcanic Series (IVS) date from 6.1 to 7.2 Ma. The potassic Main Volcanic Series (MVS) of Patmos consists of lavas of trachyandesite (¼ shoshonite sensu stricto) and trachyte and includes hawaiite bombs (nepheline trachybasalt of Wyers & Barton 1986), with an age of about 5.6 Ma. Younger basaltic dykes and lavas have been dated at about 4 Ma (Robert & Cantagrel 1979) and are termed the Young Volcanic Series (YVS). Their Nd and Sr isotopic composition suggests that they are related to the alkaline basalts of western Anatolia (e.g. Altunkaynak & Dilek 2006).
On the island of Kos, the late Miocene Dikeos pluton of monzonite and quartz monzonite and dykes of lamprophyre (kersantite) to monzonitic porphyry are found in the central part of the island (Altherr et al. 1976(Altherr et al. , 1982Altherr & Siebel 2002). Andesite lava flows and ignimbrites have yielded ages of 10.0-10.6 Ma, similar to the age of the monzonite intrusion. Minor trachytic volcanic rocks and dykes are found in the NE part of the island and appear to be of Tortonian age (Besang et al. 1977;).
On the west coast of the nearby peninsula of Bodrum (Turkey), the Bodrum Volcanic Complex represents the remnants of a stratovolcano with basalt to trachyandesite flows and trachytic domes (Ulusoy et al. 2004). Several small monzodiorite bodies were emplaced as ring dykes. Radiometric ages from the volcanic complex range from 8.6 to 12 Ma, but mostly cluster around 10 Ma (Robert & Montigny 2001). Dykes of alkaline trachyte and ultrapotassic basalt, trending NW -SE, cut the central part of the volcanic complex and have yielded ages of 7.7-7.9 Ma (Robert & Cantagrel 1979). Primitive mafic rocks represented in the complex are ultrapotassic basalt dykes and sodic basalt flows.
In western Anatolia, rocks similar to those of Samos include 12.7 Ma K-rich basalt at Foça (Innocenti et al. 1982) and 11.3 Ma quartznormative basalt and hawaiite and 11.9 Ma trachyte and rhyolite SE of Urla (Borsi et al. 1972;. In western Anatolia alkaline rocks mostly of late Miocene to Pliocene age (Yılmaz 1990;Aldanmaz et al. 2000;Altunkaynak & Dilek 2006) include the late Miocene Egrigöl basalts, a 7 Ma volcano at Soke near Samos, and the Quaternary volcanic centre of Kula (Westaway et al. 2004; Fig. 1).

Geology of Samos
The island of Samos (Fig. 2) exposes stacked metamorphic nappes of the Cycladic blueschist unit that were emplaced in the Eocene to early Oligocene (Ring et al. 1999). In the late Oligocene and Miocene, the area experienced crustal extension that emplaced the Kallithea nappe over the blueschists. The Katavasis complex is in fault contact with the Kallithea nappe, but it was regarded as the same palaeogeographical entity by Ring et al. (1999). At the same time, extensional sedimentary basins formed, with the oldest sediments being of Serravallian (late mid-Miocene) age. Volcanism was synchronous with basin formation. A contractional event within the basins (Boronkay & Doutsos 1994) is dated by a mid-Tortonian unconformity within the basins (9-8.6 Ma: Weidmann et al. 1984). Basin sedimentation continued until the early Pliocene. Basin-margin faults trend north-south. Apparently correlative faults in the marine areas both north and south of the Karaburun peninsula, imaged in seismic reflection profiles, show predominant strike-slip motion (Ocakoglu et al. 2004(Ocakoglu et al. , 2005. The Katavasis (or Kallithea) igneous complex has been interpreted as part of the younger group of plutonic rocks of the Cyclades . Mezger et al. (1985), in their detailed petrographic and geochemical study of the Katavasis dykes, focused on the problem of the origin of coexisting felsic and mafic magmas. They concluded that at least some of the different magma pulses were genetically unrelated, and that net-veined parts of the composite dykes were formed by multiple injections of felsic melt into mafic magma. They obtained a K -Ar age of 10.2 + 0.15 Ma on hornblende from monzodiorite. Emplacement of the Katavasis complex post-dated this intrusion age, but predated the mid-Tortonian contraction event (Ring et al. 1999).
The trachyte and minor rhyolite in the volcanic complex near Ambelos on the northern coast of the island (Theodoropoulos 1979) overlie basement rocks and have not been previously studied in detail.
Rhyolite and sodic basalt crop out along the margins of the Vathy and Karlovasi basins, in the eastern and western parts of the island, respectively (Robert & Cantagrel 1979). Three whole-rock K -Ar ages on basalts from the basin margins range from 7.8 + 0.5 to 8.3 + 0.4 Ma (Robert & Cantagrel 1979; Table 1). At the eastern margin of the Karlovasi basin, volcanic rocks are interbedded with Tortonian lacustrine sediments, ash-fall tuffs and tuffites. The volcanic rocks are overlain and underlain by calcareous marlstone in the main part of the basin (Meissner 1976;Stamatakis 1989a, b).
On the western margin of the Vathi basin, in eastern Samos, an extensive mafic sheet crops out (Robert & Cantagrel 1979). Where best exposed at Agios Pandeleimon (Fig. 2), the basalt is 12 m thick, with a weathered top, and is overlain by 40 m of felsic pyroclastic deposits including clasts of basalt. The jointed top of the felsic pyroclastic deposits contains overlying marl up to 50 cm below the regional top surface. At Pagondas and Pirgos (Fig. 2), there is a similar section of basalt with a weathered top overlain by a few tens of metres of felsic pyroclastic deposits. This volcanic unit appears to correlate with the Mytilene Formation in the centre of the basin.
Within the Vathi basin, the occurrence of tuffs is known precisely, as a result of stratigraphic studies of the setting of the famous mammalian faunas (Weidmann et al. 1984;Sen & Valet 1986;Fig. 3 Robert & Cantagrel (1979) the Pythagorion Formation in the western part of the basin is a 3-8 m thick subaqueous basalt flow overlain by felsic lahar tuffs with thin basalt flows. These volcanic rocks have yielded ages of 10.8-11.2 Ma. A thin tuffaceous turbidite in the overlying 400 m thick Hora Formation (lacustrine limestone) yielded an age of 9.0 + 0.3 Ma. The Hora Formation includes evaporites such as nitre, halite and sylvite (Stamatakis & Zagouroglou 1984). It is overlain unconformably by the 20 m thick clastic Mytilini Formation, which contains groups of tuffs with mean ages of 8.26 Ma, 7.35 Ma and 6.18 Ma. The base of the overlying Kokkarion Formation (50 m, principally limestone), which includes tuffaceous silts, was dated at 5.7 Ma by magnetostratigraphy (Sen & Valet 1986).
In the Karlovasi basin (Stamatakis 1989b), tuffs 250 m thick overlie 50 m of limestone in the Pythagorion Formation. Tuffs also occur near the base of the Hora Formation. These two formations were deposited in a saline alkaline lake. Authigenic zeolites, feldspars and silica polymorphs are abundant in the tuffs (Stamatakis 1989a, b;Pe-Piper & Tsoli-Katagas 1991).

Field geology and sampling
The section at Katavasis ( Fig. 2; Mezger et al. 1985) consists of schists cut by irregular monzodiorite dykes up to several metres thick striking SSW (1608E), orthogonal to sparse stretching lineations in the underlying nappes (Ring et al. 1999, Fig. 8). Granitoid rocks occur in composite dykes,  from Meissner 1976;Weidmann et al. 1984;Mezger et al. 1985;Sen & Valet 1986;Stamatakis 1989a, b;Ring et al. 1999). some contain oriented mafic inclusions, and in places the rocks form net veins within the monzodiorite. Diabase sills also occur. Sampling was directed towards avoiding rocks where interaction of mafic and felsic magmas was obvious, and thus focused on granite and monzodiorite, at the expense of monzonite or granodiorite.
The rocks at Ambelos ( Fig. 2; Theodoropoulos 1979) consist principally of trachytes. Minor rhyolites occur at the apparently intrusive contact of the trachytes and schists. The rocks are interpreted as a subvolcanic intrusive complex.
On the eastern margin of the Karlovasi basin, in western Samos, the best outcrops are NW of Koumeika (Fig. 2), where several tens of metres of rhyolitic tuffs and flow-banded rhyolite are exposed in road cuts. In the same area, small basalt cones and marls with tufa (indicating hot spring activity) are exposed. The rhyolitic rocks are intensely altered, mostly along veins, which are associated with either zeolitization or silicification. Volcaniclastic conglomerate immediately overlying the rhyolites contain clasts of both basalt and fresh (dark grey) flow-banded rhyolite. Basalt from the western margin of the Karlovasi basin was sampled at Moni Agios Georgios (Fig. 2). Samples from the western margin of the Vathi basin were collected from Agios Pandeleimon, Pagondas and Pirgos (Fig. 2).

Laboratory analyses
All mineral analyses were made with a JEOL-733 electron microprobe with four wavelength spectrometers and a Tracor Northern 145 eV energydispersive detector. Operating conditions were 15 kV at 5 nA beam current. Geological standards were used. Data were reduced using a Tracor Northern ZAF matrix correction program.
A total of 39 representative igneous rock samples were analysed for 10 major and minor element oxides and 14 trace elements on a Philips PW 1400 sequential X-ray fluorescence spectrometer using a Rh-anode X-ray tube. Rare earth elements (REE) were determined on selected samples by instrumental neutron activation analysis. Analytical precision is as given by Pe-Piper & Piper (2002, appendix 1). Lead and Sm-Nd isotopic ratios were determined by Geospec Consultants Limited, Edmonton, using methods summarized by Pe-Piper & Piper (2001). Geochronology was carried out on whole-rock samples, or mineral separates, as noted in Table 1, by the K -Ar method at Kruger Laboratories Inc.
A full set of whole-rock geochemical analyses, radiogenic isotope analyses, mineral analyses and details of the petrography of analysed samples is available online at http://www.geolsoc.org.uk/ SUP 18299. A hard copy can be obtained from the Society Library.

Geochronology
Six new K -Ar ages have been obtained (Table 1; Fig. 3). One age on hornblende in monzodiorite from Katavasis was 10.7 + 0.9 Ma, confirming the early Tortonian age obtained by Mezger et al. (1985) for the Katavasis complex. Two ages from biotite in monzodiorite and quartz monzodiorite at Katavasis were a little younger, 8.2 and 8.1 + 0.4 Ma, and probably represent cooling ages associated with the juxtaposition of the Katavasis complex and the Kallithea nappe. Samples of rhyolite and trachyte at Ambelos yielded consistent ages of 10.2 and 9.9 + 0.3 Ma, respectively, which, within the range of analytical error, is synchronous with or slightly younger than the Katavasis intrusive complex.
One new age has been obtained from the basin margin rhyolites and basalts: a rhyolite from Koumeika yielded a whole-rock age of 8.7 + 0.4 Ma. The adjacent and overlying basalts have yielded ages between 8.3 and 7.8 Ma (Robert & Cantagrel 1979). These mid-Tortonian ages are synchronous with, or a little younger than, the unconformity between the Hora and Mytilene formations (Fig. 3).

Petrography of the major rock types
The lithologies identified in the Katavasis intrusive complex are monzodiorite, quartz monzodiorite, tonalite and leucogranite. The main ferromagnesian minerals in the monzodiorite, quartz monzodiorite and tonalite are hornblende and biotite, with rare clinopyroxene; titanite is also common; and the principal opaque mineral is magnetite, although hematite and pyrite are also present. Some rocks are very heterogeneous, with 'pools' of quartz, plagioclase and titanite. All crystals in these pools contain abundant apatite inclusions. The only ferromagnesian mineral present in the leucogranites is biotite, with accessory minerals titanite, allanite and rarely actinolite and actinolitic hornblende.
The phenocrysts commonly present in the Ambelos trachyte are plagioclase, K-feldspar, clinopyroxene, biotite and quartz, generally forming glomeroporphyries. The groundmass ranges from 72% to 82% of the whole rock and consists generally of the same minerals. The trachytes in places contain many xenoliths, including granite, chloritized gabbro or diorite, very fine-grained igneous lithologies, garnetiferous pelites and quartzfeldspathic schists.
The basin margin basalt contains principally plagioclase and clinopyroxene. The associated rhyolite shows a variety of textures: granophyric, spherulitic, vitrophyric, and vesicular, and the glassy groundmass may make up to 90% of the rock. Common phenocrysts are K-feldspar, quartz and plagioclase.
Clinopyroxene is the only ferromagnesian mineral present in the basalt, showing complex zoning and ranging in composition from diopside to augite (Fig. 4). Three early clinopyroxene populations are present in basalt, forming cores to phenocrysts: green augite (Ts component 13-30%), colourless diopside (Di component 80 -86%) and brown augite. All of these cores may be resorbed and are mantled with intermediate augite compositions, with variable Cr 2 O 3 (from below detection limit to 1%). These augites are similar to some normally zoned phenocrysts and to the microphenocryts in these rocks.
Clinopyroxene is also the main ferromagnesian mineral in the trachyte. Phenocrysts are of light green diopside (Mg/(Mg þ Fe) 0.62 -0.75), in some cases with augite cores with a high Ts component. Some samples contain dark green Fe-rich augite or hedenbergite crystals with opaque rims, partly replaced by dusty opaques and clays. Reverse zoning is common.

Amphibole
The granite contains ,2% amphibole, either actinolite or actinolitic hornblende. The monzodiorite contains ,60% amphibole, including edenitic hornblende, ferroan pargasitic hornblende and magnesiohornblende, some of which shows sub-solidus alteration to actinolite. Al iv is as high as 1.8 in some samples. Estimates of the temperature of crystallization, using the Blundy & Holland (1990) geothermometer, are about 900 8C, which yields a pressure estimate of 1 kbar for the most aluminous cores, using the Anderson & Smith (1995) geobarometer. Anderson & Smith noted that the geobarometer is less reliable for high temperatures, low oxygen fugacity and plagioclase outside the range of An 25 -35 , all conditions that may apply to the Katavasis monzodiorite. Amphibole is absent from the basalts, trachytes and rhyolites.

Biotite
Biotite is present in monzodiorite, granite and trachytes, belonging to the phlogopite -annite series (Fig. 5). Some biotite in granite has Al-rich rims. Biotite in granite is much less magnesian than biotite in monzodiorite for the same FeO T content. The rare biotite from trachyte has lower FeO T / FeO T þ MgO) compared with those from monzodiorite, indicating that the trachyte crystallized under high fO 2 , and has high TiO 2 content. All biotites fall in the calc-alkali field in the ternary discrimination diagram of Abdel-Rahmen (1994).

Sulphides and Fe -Ti oxides
The monzodiorite contains equant crystals of magnetite, some with chalcopyrite inclusions that contain globular inclusions of pyrite. Rutile occurs along fractures or as corona rims on magnetite. Some magnetite is developed in fractures. Rutile crystals with hematite and chalcopyrite inclusions, the latter containing both hematite and pyrite inclusions, occur in some samples (e.g. SV47).
The granite contains magnetite and rutile either as coronas upon magnetite or as scattered hypidiomorphic crystals. Some magnetite occurs in fractures and some is closely associated with brown biotite. Late leucogranitic veins (SV37) contain very few opaque crystals of ilmenite. The mineral assemblage titanite þ magnetite þ quartz found in the granites indicates high oxygen fugacity (Wones 1989).
Titano-magnetite is the most common Fe -Ti oxide mineral in the trachyte and may contain globular inclusions of chalcopyrite. Idiomorphic crystals of pyrite are partially replaced by rutile. Small inclusions of pyrite are present in clinopyroxene phenocrysts. Ambelos rhyolite contains only magnetite, which may be intergrown with rutile.
The basalts generally contain hypidiomorphic and equant crystals of titano-magnetite. Rutile is the dominant opaque mineral in altered samples (e.g. SV71b). The basalt from Agios Georgios with unusual feldspar mineralogy contains composite crystals of magnetite -ilmenite throughout the groundmass and magnetite crystals as overgrowths on chromian spinel.
The most common opaque mineral in the altered rhyolite from Koumeika is pyrite, as hypidiomorphic crystals, large idiomorphic crystals, small equant anhedral grains dispersed throughout the groundmass, or intergrown with rutile. Chalcopyrite as hypidiomorphic crystals dispersed throughout the groundmass is less common. Hematite in some samples forms xenomorphic grains intergrown with rutile, mostly confined to spherulitic grain boundaries.

Titanite and allanite
In the monzodiorite, pleochroic titanite is abundant (up to 5%), as hypidiomorphic to xenomorphic crystals with magnetite and plagioclase inclusions. Granite contains minor titanite, with higher Al content than in monzodiorite. Some titanite from monzodiorite has much higher P REE than those from granite. In titanite from monzodiorite, P REE generally decreases from core to rim, but increases in the titanite from granite. The size, shape and P REE content of titanite in monzodiorite suggest early crystallization. The lower P REE in titanite from granite may be a consequence of allanite crystallization. Titanite is rare in trachytes, but chemically similar to titanite from monzodiorite.  Mezger et al. (1985) reported allanite in both the mafic and felsic rocks of the Katavasis complex. The common pleochroic allanite in the granite shows a wide range of compositions and no systematic zoning. Apatite is common in monzodiorite and leucogranite.

Nomenclature
We follow Altherr & Siebel (2002) in naming the mafic plutonic rocks as monzodiorite. The analysed granitoid rocks classify either as granite or as quartz syenite in the chemical classification system of Streckeisen & Le Maitre (1979). In the IUGS chemical classification system (Fig. 6) all the analysed felsic volcanic rocks classify as rhyolites, the intermediate rocks as trachytes and the mafic rocks as basalts (Pagondas, Agios Pandeleimon), trachybasalts (Koumeika, Pirgos) and basaltic trachyandesites (Agios Georgios and one sample from Koumeika). We use the general term basalt for all mafic rocks.

Katavasis monzodiorite and granite
The analysed monzodiorites are nepheline normative, except for the most felsic rocks (57% SiO 2 ), which are hypersthene normative (see Table 2). Trace element variation, normalized to primitive mantle (Fig. 7) is very uniform, with a progressive decrease in normative abundance from the large ion lithophile elements (LILE) to the high field strength elements (HFSE), prominent troughs at Nb and Ta, and minor troughs at P, Zr, Hf and Ti for some samples. REE show progressive decrease in normative light (LREE) and middle REE (MREE), but lesser fractionation of the heavy REE (HREE; Fig. 8).
All analysed granites are slightly peraluminous (mol Al 2 O 3 . mol (CaO þ Na 2 O þ K 2 O)), with a small amount of normative corundum. Using the tectonic environment discriminant diagrams of Pearce et al. (1984) they classify either as post-collision or volcanic arc granites. Compared with the monzodiorite, the granites show enrichment in LILE and overall rather lower HFSE, including strong depletion in Nb, P, Ti and Y. The granites have lower contents of most REE than the monzodiorites (Fig. 8), but similar La content (as discussed in detail by Mezger et al. (1985)).

Ambelos trachyte and rhyolite
The Ambelos trachytes have a rather narrow range of geochemistry, with 62-65% SiO 2 (Table 3). They are hypersthene normative. Trace element patterns differ from those for monzodiorite in that most incompatible elements are more abundant and troughs for Sr, P and Ti are more pronounced (Fig. 7). REE patterns are similar to those of the monzodiorite, except that the LREE are a little more abundant (Fig. 8). Unlike the granites, the associated rhyolites show strong Eu depletion and enrichment in HREE compared with trachyte. Compared with the trachytes, the rhyolites show strong depletion in Ba, Sr, P, Eu and Ti, but some enrichment in LILE, Nb and Ta (Fig. 7).

Basin margin basalt and rhyolite
Most basalt is nepheline-normative and has similar major element composition to the monzodiorite,  (Fig. 7). The basalt has higher Ni and Cr and lower Y than the monzodiorite, reflecting the presence of clinopyroxene and possible pseudomorphs after olivine in the former and amphibole in the latter. REE abundances are also very similar to those for monzodiorite (Fig. 8).
Rhyolite associated with the basalt has similar abundances of most HFSE, but is enriched in LILE, Nb and Ta, and strongly depleted in Ba, Sr, P, Eu and Ti (Fig. 7). The LREE abundance in the rhyolite is slightly lower than in the basalt, but HREE abundance is similar (Fig. 8).
Alteration in basalts was evaluated from two samples from the same outcrop, SV71A (fresh) and SV71A (altered) (

Lead isotopes
The Samos rocks exhibit a narrow range of Pb isotopic ratios, varying between 18.86 and 19.10 in 206 Pb/ 204 Pb. Most data points form a welldefined line in a 207 Pb/ 204 Pb v. 206 Pb/ 204 Pb diagram ( Fig. 9) with two outlying values for one basalt and Ambelos rhyolite, which have high 206 Pb/ 204 Pb. The outlying basalt (SV71A) shows no unusual characteristics in its trace element composition, except that Cs is particularly high. The m value of the source rock reservoir appears to be around 10.0, indicating involvement of upper crustal material, either as subducted sediment or by assimilation and fractional crystallization in the upper crust. Data distribution in 208 Pb/ 204 Pb v. 206 Pb/ 204 Pb space (not illustrated) implies an average present-day Th/U ratio of 2.3, lower than the average crustal value of 3.78 (Stacey & Kramers 1975). Pb isotopic values show no systematic variation with SiO 2 content.

Nd -Sm isotopes
The monzodiorite, granite, Ambelos trachyte and basin margin basalt all show very similar, 1 Nd of 21 to 22 (Fig. 10) and model ages of 0.7-0.9 Ga. These values are similar to those shown by rocks of similar age in Bodrum in western Turkey (Robert et al. 1992), some basalts from Patmos (Wyers & Barton 1987), monzodiorite from Kos (Altherr & Siebel 2002), and some mildly alkaline volcanic rocks from western Anatolia (Altunkaynak & Dilek 2006). The rhyolites from both Ambelos and the basin margin have more negative 1 Nd (24.4, 26.1) and higher model ages (0.9, 1.9 Ga), together with high 207 Pb/ 204 Pb.

A common source for monzodiorite, basalt and trachyte
A plot of 1 Nd v. SiO 2 for Late Miocene igneous rocks of the southeastern Aegean (Fig. 7a) shows no systematic change in 1 Nd v. SiO 2 for the igneous rocks of Samos, with the exception of rhyolites. This indicates that crustal assimilation with fractional crystallization or mixing with crustal melts is unlikely to account for most of the isotope variation. Altherr & Siebel (2002) argued that the difference between a kersantite (lamprophyric) dyke (1 Nd ¼ þ 0.55) and various monzonite and more evolved lamprophyric dykes in Kos implied some mixing process between kersantitic melts and crustal materials, and the Bodrum rocks show a similar trend of decreasing 1 Nd with increasing SiO 2 (Robert et al. 1992). However, in the plutonic rocks of Samos, as in the intermediate rocks of Patmos and Bodrum, there is no systematic change in 1 Nd with SiO 2 . Such isotopic variation thus supports interpretations based on geochemistry (Robert et al. 1992) that the late Miocene basalts and trachytes of the southeastern Aegean were derived principally from the mantle and that there was no significant crustal assimilation involved in the fractionation of the trachytes. Conversely, the more negative 1 Nd and higher radiogenic Pb in the rhyolites suggests a crustal contribution to the felsic magma.
In addition to the isotopic evidence, the uniformity of incompatible trace element distribution for basalt and monzodiorite (Fig. 7) and their similarity in REE distribution (Fig. 8) argue for a common source. The REE distribution (Figs 8 and 11) suggests partial melting of enriched mantle in the presence of garnet. The trachyte also has similar incompatible trace element composition to the basalt and monzodiorite, except that elements fractionated by feldspars (Ba, Sr, Eu), clinopyroxene (Cr, Ni, Eu), amphibole (Y), apatite (P), and titanite (Ti) are relatively depleted and other elements are relatively enriched (Figs 7 and 8). It is concluded that they too are derived from the same mantle source.

The evidence for magma mixing
Petrographic and mineral chemical data, in particular, the complex zoning patterns shown by the clinopyroxenes and plagioclase phenocrysts, provide compelling evidence that mixing of small magma batches has occurred in the mafic volcanic rocks. At least three types of mafic magma were mixed in the basalts: one of these was relatively primitive and carried phenocrysts of diopsidic clinopyroxene and anorthitic plagioclase, whereas the other two were more evolved and carried phenocrysts of either augitic or salitic clinopyroxene and more albitic plagioclase. After mixing, pyroxene and plagioclase of intermediate composition precipitated from the hybrid magma as microphenocrysts or mantles around the xenocrysts. The green augite cores with high Al 2 O 3 content represent crystallization at elevated pressures, whereas the diopside overgrowths with high Cr 2 O 3 represent crystallisation from a less evolved magma batch at lower pressures (see Pe-Piper 1984). Reverse zoning in clinopyroxene from trachyte may be related to the appearance of magnetite at the liquidus.
In the Katavasis complex, Mezger et al. (1985) showed that chemical variations could be best explained by multiple intrusions of small magma batches and that Sr isotope ratios suggested that some magma batches were genetically unrelated.
New Nd isotope data from this study and from Altherr & Siebel (2002), however, show a remarkable overall uniformity in isotopic composition (Fig. 10). The minor variations in isotope composition could be a result of different source points in a compositionally inhomogeneous enriched mantle (Pe-Piper & Piper 2001) and/or minor interaction with crustal melts (Altherr & Siebel 2002).

The role of fractionation
Using geochemical characteristic such as MgO/ (MgO þ FeO T ) ratio (Fig. 12d) and Ni and Cr  Tables 2 -4. abundance (Fig. 13b), the basalts appear more primitive than the monzodiorite. Element ratios including La/Sm, Nd/Sm and Nb/Zr show no systematic differences between basalt and monzodiorite, although Sm/Yb tends to be higher in basalt (Fig. 11). Ba/Rb ratio shows more scatter in unaltered basalt than in monzodiorite, whereas V/Zr and Y/Zr show more scatter in the monzodiorite (Fig. 12).
Geochemical relationships between monzodiorite, basalt and trachyte have been examined using binary plots of trace elements that are influenced by fractionation of particular minerals. Comparisons are made between the set of little altered basalt (outlined by the circle in Fig. 13) and the monzodiorite and trachyte samples to examine the nature of fractionation assuming that the basalt most closely represents parental magma. Low Cr and Ni abundances in both monzodiorite and trachyte imply predominant clinopyroxene (or perhaps olivine) fractionation (Fig. 13b). Values of Zr and Nb in monzodiorite are scattered around values for basalt (Fig. 13c), consistent with the variable amounts of titanite, zircon and various ferromagnesian minerals in the monzodiorite. Trachyte compositions could be derived from parental basalt magma by fractionation of amphibole (Fig. 12c). Fractionation of clinopyroxene or feldspar is suggested by some element plots (Fig. 13c -e), and some plagioclase fractionation is confirmed by the small Eu anomaly in the trachyte (Fig. 8) and the fractionation of Sr against Ba (Fig. 13f). Ti content is high in biotite from the trachyte (Fig. 5), but whole-rock TiO 2 is lower in the trachyte (Fig. 13a), probably as a result of magnetite or titanite fractionation. The trachyte is depleted in P relative to basalt and monzodiorite (Fig. 7), probably as a result of apatite fractionation. Variations in U and Th suggest that zircon fractionation may be involved in petrogenesis of monzodiorite.
The fractionation relationships discussed above suggest that the monzodiorites and trachytes (of similar age) were derived from parental magma similar to that of the basin margin basalts. The trachyte resulted from fractionation of hornblende, plagioclase and magnetite, together with lesser apatite and titanite fractionation. These minerals are the principal mineral phases in the monzodiorite, suggesting that much of the monzodiorite represents residual fractionating material retained in a magma chamber.
This genetic link between monzodiorite and trachyte is further supported by the opaque mineralogy of the two rock types. In the monzodiorite, ilmenite is altered to rutile (with sulfide inclusions) and there are two generations of magnetite: equant crystals and in fractures. The presence of pyrite suggests low fO 2 during the early stages of crystallization of the dioritic magma, but during later stages the presence of magnetite and titanite suggests high fO 2 . Pyrite and chalcopyrite inclusions in titanomagnetite in the trachytes suggest evolution of fO 2 similar to that in the monzodiorite.

Where was the source?
Bulk chemical composition and specific trace element ratios in the mafic rocks of Samos are consistent with partial melting of enriched hydrous mantle peridotite within the stability field of phlogopite, amphibole and garnet (e.g. Turner et al. 1996). The ratios of REE (Fig. 11) suggest 5-10% partial melting. The Sr and Nd isotope data (Fig. 10) do not support significant involvement of asthenospheric melts, which are important in the more alkalic magmas in the YVS of Patmos and in western Anatolia (Güleç 1991;Seyitoglu et al. 1997;Aldanmaz et al. 2000). Rather, the source in enriched subcontinental lithosphere is similar to that of the lamprophyres of Kos (Altherr & Seibel 2002) and Bodrum (Robert et al. 1992). The isotopic data for the Katavasis granites indicate a common source with the monzodiorite, although the details of their evolution is uncertain, as noted by Mezger et al. (1985).

Evolution of the rhyolites
The Nd and Pb isotope composition of the rhyolites, compared with that of the associated trachyte and basalt, suggests that both have evolved through crustal assimilation. Their relatively high 1 Nd (24 to 26.5) indicates that any old crustal component is likely to be minor; for example, the Hercynian paragneisses of the central Aegean have 1 Nd of near 211 (Tarney et al. 1998) and granites of Naxos have 1 Nd of 27 to 210 (Pe-Piper 2000).
The basin margin rhyolite shows simple REE behaviour with a strong Eu anomaly and contains 3-8 ppm Cr, suggesting mixing with 1 -2% basaltic magma. This may indicate that rhyolite eruption was triggered by injection of hot mafic magma into a fractionating magma chamber, as argued, for example, by Druitt et al. (1999) at Santorini. Strong depletion in Ba, Sr, Eu and Y points to significant plagioclase and clinopyroxene and/or amphibole fractionation, implying that fractionation took place within the crust, but at what depth is uncertain.
Complex REE patterns in the Ambelos rhyolite ( Fig. 8) and some binary element plots ( Fig. 13c and d) suggest the importance of accessory mineral fractionation. This rhyolite is much less voluminous than the basin margin rhyolite, and Ni and Cr are below detection.

Tectonic implications
The change from trachyte to basalt As in Samos, many back-arc volcanic rocks of the Aegean area correlate in time with basin subsidence , suggesting a relationship between faulting and extrusion. Ring et al. (1999) showed that Neogene sediments in the Pirgos area of Samos occupy a half-graben bounded by a WNW -ESE striking master fault, corresponding to their D 3 phase, which they interpreted as correlative with the Kallithea detachment. This is then cut by D 4 north-south-or NE-SW-striking faults, which appear to localize the basin margin basalts  (2002)). and correspond in time to the major unconformity between the Hora and Mytilene formations. These faults show transpressional kinematics (Boronkay & Doutsos 1994) and are correlated with a 'sinistral wrench corridor' extending east of Karaburun proposed by Ring et al. (1999;Fig. 14). The D 4 strike-slip faulting was mapped on either side of the Karaburun peninsula by Ocakoglu et al. (2004Ocakoglu et al. ( , 2005. The faulting appears to continue northeastward into the late Miocene Zeytindag, Ö renli-Egiller and Altınova basins mapped by Yılmaz et al. (2000), which cross-cut earlier structures and contain late Miocene lacustrine sediments. These sediments interbed with the Egrigöl basalt, approximately dated by Rb -Sr at 9-6 Ma (Borsi et al. 1972;Ercan et al. 1985) and erupted from NE-SW-trending fissures (Yılmaz et al. 2000). Also on this trend are the 10 -12 Ma bimodal basalt and rhyolite at Urla and 4-5.6 Ma basalts of the Young Volcanic Series at Patmos, perhaps indicating that this sinistral strike-slip system propagated southward through time. This 'wrench corridor' may also have been responsible for the localization of north-south lamprophyre dykes in Kos of uncertain age that cut the 10 Ma Dikeos monzonite and the faulted margin of the monzonite against country rock. Older, middle Miocene strike-slip faulting is suggested by the northsouth-trending dykes of the Mytilene Formation in southeastern Lesbos (Pe-Piper & Piper 2007) and the parallel orientation of rhyolite domes and fissure eruptions in Chios Fig. 14).
The change from extensional listric faulting (D 3 ) to strike-slip faulting (D 4 ) in Samos is analogous to the pattern seen in the Pliocene-Quaternary of the south Aegean arc, where evolved andesites and dacites are associated with listric faulting in the western arc, with a much higher proportion of basalt and rhyolite associated with strike-slip faulting in the eastern part of the arc . In Samos, evolved Ambelos trachytes were extruded during listric extensional faulting and bimodal basalts and rhyolites during strike-slip faulting. Large-scale strike-slip faulting provides crust-penetrating pathways for basaltic magma. In contrast, mid-crustal extensional detachment faulting associated with upper crustal listric faulting creates barriers to the upward migration of magma and promotes fractional crystallization in the resulting mid-crustal magma chambers, with the production of monzodiorite and trachyte.

Relationship to regional tectonics
The changes in fault patterns and volcanism in the late Miocene of the southeastern Aegean are a result of the progressive convergence of Gondwanan continental crust with the Anatolian -Aegean microplates (Fig. 15), as reviewed for example by Aksu et al. (2005). By the Miocene, there was already collision of Arabia with eastern Anatolia. However, oceanic subduction was active south of Cyprus at least in the early Miocene (Robertson 2000). Widespread north-south extension in the Aegean Sea region began in the early Miocene (Gautier & Brun 1994;Dinter 1998), but less extension appears to have taken place in western Anatolia (Yılmaz et al. 2000). The widespread north-south graben faulting in western Anatolia mapped by Yılmaz et al. (2000) may therefore have had a component of wrench faulting to accommodate the Aegean extension, as suggested for the late Miocene by Ring et al. (1999). In western Anatolia, north -south extension with mid-crustal detachment developed in the late Oligocene (Ç emen et al. 2006). Roll-back continued along the subduction zone south of the Aegean Sea (Royden 1993) across oceanic crust, whereas collision had already taken place south of Cyprus (Robertson 2000).
Palaeomagnetic data from Anatolia show no evidence for significant rotation from Eocene to Miocene, but with rotation beginning probably in the late Miocene (Platzman et al. 1998) or Pliocene (Kissel et al. 2003). Arabia started moving northward faster than Africa at about 12 Ma with the development of the Dead Sea fault zone (Garfunkel 1981), and many researchers have recognized this as the time of final collision and the onset of orogen-parallel movement of Anatolia (Ş engör   al. 1985). The initiation of the North Anatolian Fault is generally interpreted to date from about this time (Barka 1992), but slip along the East Anatolian Fault did not begin until the latest Miocene (Hempton 1985).
The rotation of the west Aegean block relative to the SE Aegean -Anatolian block throughout the Miocene (Walcott & White 1998) would have resulted in sinistral strike-slip motion along the block boundary, most clearly seen along the mid-Cycladic lineament in the central Cyclades . Walcott & White (1998) suggested that elsewhere the shear across the boundary between the two blocks was more distributed. This distributed shear is manifested in the mid-Miocene faulting in Lesbos and Chios, and the better documented late Miocene ( -?early Pliocene) sinistral shear extending from the Ö renli-Egiller    Thompson & Fowler (1986); (e), (f ) from Tindle & Pearce (1981). amp, amphibole (hornblende); ap, apatite; bi, biotite; cpx, clinopyroxene; fel, feldspar; Kfs, K-feldspar; mgt, magnetite; ol, olivine; opx, orthopyroxene; plg, plagioclase; tit, titanite; zrn, zircon. basin in the north through Samos to the Dikeos monzonite of Kos in the south. The initial development of the Samos basins dates from the late Serravallian to early Tortonian (12-11 Ma), and the Katavasis complex and oldest tuffs date from this time. The mid-Tortonian, at about 10 Ma, was a time of change in tectonic style throughout the Aegean (Le Pichon & Angelier 1979), including NE-SW compression in Crete that affected sediments as young as latest Serravallianearly Tortonian (12-11 Ma) (Meulenkamp & Hilgen 1986; phase 1 of ten Veen & Meijer 1998). Palaeomagnetic data suggest that the rotation rate of the west Aegean block may have increased at this time. The Ambelos volcanic rocks and the mid-Tortonian volcanic rocks of Kos date from this time. The deformation at the late Tortonian unconformity in the Vathy basin and the extrusion of the basin margin basalts of Samos is a little younger than this mid-Tortonian event, but correlative volcanic rocks are present in Kos. Volcanism in Patmos is a little younger and ended when subduction-related volcanism began in the south Aegean arc.

Implications
Late Miocene igneous rocks of Samos are part of a suite of rocks in the southeastern Aegean Sea derived from partial melting of enriched subcontinental lithospheric mantle during regional extension. Monzodiorite, granite, trachyte and basalt magmas all had a common source and experienced no significant crustal assimilation. Older monzodiorite, granite and trachyte were emplaced during listric faulting, which created conditions for midcrustal magma chambers to form, within which trachyte fractionated. Basalts and associated rhyolites were extruded during later strike-slip faulting, which provided efficient pathways for magma to rise through the crust. This sinistral north-south faulting marked the diffuse eastern margin of the west Anatolian block and appears to have  Ring et al. (1999). West Aegean block from Walcott & White (1998). Crete from ten Veen & Meijer (1998). Aegean Sea from Dinter (1998). Western Anatolia from Yılmaz et al. (2000). North Anatolian Fault from Armijo et al. (1999). propagated southward from the Karaburun peninsula through Samos to Patmos and Kos in the late Miocene and (?)early Pliocene.