Abstract

New trace-element, radiogenic isotopic, and geochronologic data from the Troodos ophiolite, considered in concert with the large body of previously published data, give new insight into the tectonic history of this storied ophiolite, as well as demonstrating the variability of suprasubduction-zone ophiolites, and differences between them and commonly used modern analogs. Similar to earlier studies, we find that island-arc tholeiite of the lower pillow lava sequence erupted first, followed by boninite. We further divide boninitic rocks into boninite making up the upper pillow lava sequence, and depleted boninites that we consider late infill lavas. We obtained an Ar-Ar age from arc tholeiite of 90.6 ± 1.2 Ma, comparable to U-Pb ages from ophiolite plagiogranites. New biostratigraphic data indicate that most of the basal pelagic sedimentary rocks that conformably overlie the boninitic rocks are ca. 75 Ma. This suggests that voluminous eruption of boninitic rocks persisted until ca. 75 Ma. Limited eruption of boninitic lavas may have continued until 55.5 ± 0.9 Ma, based on the Ar-Ar age we obtained. The duration of arc magmatism at Troodos (at least 16 m.y., with some activity perhaps extending 35 m.y.) without the development of a mature arc edifice greatly exceeds that of other well-studied suprasubduction-zone ophiolites. We propose that Troodos was formed over a newly formed subduction zone, similar to many proposed models, but that the extended period of magmatism (boninitic) resulted from a prolonged period of ridge subduction.

INTRODUCTION

Research on the Troodos ophiolite of Cyprus has strongly influenced the evolution of the ophiolite concept and ideas on the significance of on-land sheets of oceanic lithosphere in orogenic belts (e.g., Moores and Vine, 1971; Dilek, 2003; Robinson et al., 2003). Accordingly, petrologic and geochemical studies of the Troodos ophiolite have also figured prominently in the wide acceptance of the role of suprasubduction-zone magmatism in the generation of many ophiolites (e.g., Shervais, 2001; Pearce, 2003; Nicolas and Boudier, 2003; Pearce and Robinson, 2010).

The interpretation of suprasubduction-zone generation of many ophiolites derives primarily from geochemical affinities of ophiolite lavas and dikes. Since Miyashiro (1975) initially proposed an island-arc setting for the Troodos ophiolite generation based on its major-element compositions, many subsequent studies, cited in this paper, have identified island-arc characteristics, based on minor- and trace-element compositions and isotopic ratios.

A potential problem in some of the early geochemical studies is the mobility of large ion lithophile elements (LILEs) during alteration, which may result in enrichment of these elements in nonarc volcanic rocks, and an erroneous interpretation of such rocks as arc generated, given that LILE enrichment is a fundamental characteristic of arc lavas. Analysis of fresh volcanic glass of pillow and dike margins, however, can largely surmount this difficulty, as evidenced by the extensive major-element data set of Pearce and Robinson (2010), and trace-element and isotopic analyses (e.g., Rautenschlein et al., 1985), which demonstrate an arc origin for the lavas.

Although previous studies have firmly established the suprasubduction-zone origin of the Troodos ophiolite, further exploration of the geochemistry and geochronology of the lavas offers additional insight into details of magmatic evolution of the Troodos ophiolite and other suprasubduction-zone ophiolites as well as providing a means by which to better compare the geochemistry of Troodos rocks to other ophiolites and modern seafloor rocks (particularly those in arc environments).

To exploit the improvements in technology and methodology and gain better insight into the tectonic settings of Troodos magmatism, we analyzed trace elements of the 12 freshest samples from throughout the ophiolite (Fig. 1), and also analyzed the Nd, Sr, and Pb isotope ratios of five of these samples. Whereas our samples span a greater geographical extent of the Troodos ophiolite than the previous studies by Rautenschlein et al. (1985), Cameron (1985), and others, our sample set is small compared to the 137 samples of glass analyzed by Pearce and Robinson (2010), although their study presented major-element data only.

In addition to geochemical data, we present radiolarian biostratigraphic data, some paleomagnetic data, and Ar-Ar age analyses of two fresh volcanic glasses that directly date tholeiite and boninite eruption in the Troodos ophiolite. The combination of new geochemical and geochronologic data, when integrated with published data, gives new insight into the sequence of the magmatism and its temporal extent.

GEOLOGIC SETTING

The Troodos ophiolite extends 100 km in E-W and 30 km in N-S dimensions (Geological Survey Department Cyprus, 1995). The Troodos massif forms a gentle dome structure elongated E-W (Fig. 1). Because of the superposition of this domal structure and erosion, the structurally lowest ultramafic rocks constitute the center of the Troodos massif and the highest elevations at Mount Olympus (1952 m altitude). The ultramafic rocks are flanked by gabbro associated with a small amount of plagiogranite, an extensive sheeted dike complex, pillow lava, and sediments, in ascending structural-stratigraphic order (Fig. 1).

The Troodos ophiolite has been subdivided into the northern main massif and its axial sequence, including Mt. Olympus, and the southern Limassol Forest complex and its Arakapas sequence, bounded by the E-W–trending Arakapas transform fault zone (Fig. 1; Geological Survey Department Cyprus, 1995; Gass et al., 1994). Ultramafic rocks, gabbro, and sheeted dikes also crop out in the Limassol Forest complex, but tectonism has severely disrupted the original tectonomagmatic relationships. The Troulli inlier is isolated from the ophiolite, but it is clearly an eastern extension of the main massif. Another isolated massif with the same lithology as the main massif crops out on the Akamas Peninsula (Fig. 1).

The Geological Survey Department Cyprus (1995) divided the pillow lavas into (1) a basal group, (2) lower pillow lavas, (3) upper pillow lavas, mostly in the axial sequence, and (4) lavas and volcanic breccias in the Arakapas sequence whose stratigraphic relationship with the other three groups is not clear. The basal group lavas display greenschist-facies hydrothermal metamorphism (Geological Survey Department Cyprus, 1995) and are easily distinguished from the other lavas. However, the first author could not lithologically distinguish the upper and lower pillow lavas (cf. Murton, 1989), and sampled lavas are classified (Table 1) by their subunits as defined by the Geological Survey Department Cyprus (1995) mapping, although the stratigraphic position of several samples in the axial sequence can be described (see Table 1). Volcanic breccias are distributed along the Arakapas valley and its western and eastern extensions, and they characterize the transform fault zone. Associated lavas, especially those above the breccias, are expected to occupy the stratigraphically highest level. Gass et al. (1994) called these infill lavas, and we follow that designation here. Weathered infill lavas tend to be gray colored. Gass et al. (1994) classified pillow lavas cropping out along the southern margin of the Limassol Forest complex as upper pillow lavas instead of infill lavas.

The lack of easily identified marker horizons with the volcanic rocks, lack of distinctive field characteristics of volcanic units, and the enormous extent of the exposures and structural complexity of the ophiolite necessitate the application of a chemostratigraphic classification of the lavas. For Figure 1, we used a somewhat different division of lavas than previously proposed that highlights key geochemical differences as well as stratigraphic setting, as will be explained in the sections on geochemistry. The problem with such an approach is that the classification depends heavily on geochemical sampling density, so it is subject to some uncertainty, and differences in interpretation between different researchers are inevitable, as well as some ambiguity regarding the affinity of some outcrops. For the purposes of our study, we believe that uncertainties in location of chemostratigraphic contacts, or the definition of them, do not significantly impact the main interpretation and conclusions of this study. We divided the lavas into the lower pillow lava sequence (suite A; island-arc tholeiite), upper pillow lava sequence (suite B; boninite), and the infill lava (suite C; depleted boninite).

The dikes of the sheeted complex strike N-S, parallel to the paleo–spreading axis, as concluded by many authors (e.g., Gass et al., 1995). However, three fossil axial valleys on the northern flank of the ophiolite may record discrete eastward ridge jumps, based on their field relationships (Varga and Moores, 1985; Allerton and Vine, 1991; Moores et al., 1990). In Figure 1, three antiformal axes in lavas on the northern flank correspond to these grabens; the antiformal dip configuration results from tilting of lavas in the hanging wall of graben-bounding listric normal faults. The relationship between the lava stratigraphy and the proposed ridge jumps is not clear, and we will revisit and discuss these relationships later in the context of the relation of suites A and B.

GEOCHEMISTRY

Analytical Methods

Major-element compositions for 12 samples (Table 1) were measured on fused glass beads, using X-ray fluorescence (XRF) spectrometry at Hirosaki University and Tohoku University. The analytical procedure used has been described in detail by Fujimaki and Aoki (1987) and Yajima et al. (2001). An additional eight samples (Table 1) were analyzed for major elements by electron microprobe at Tohoku University; these samples were not analyzed for trace elements.

Trace elements (Table 1) were analyzed by inductively coupled plasma–mass spectrometry (ICP-MS) (Yokogawa Analytical Systems HP4500) at the University of the Ryukyus, with a 115In internal standard. A 50 mg powdered sample was dissolved in a mixture of HF-HClO4, and the evaporated sample was re-dissolved in HNO3, as described in Shinjo (1999). The uncertainty in the analyses is generally better than 5% for most trace elements. Calibration curves were constructed using rock standard JB-1 and a blank solution. In order to evaluate our analytical result, we also analyzed JA-1 and JB-2, standard rock samples provided by the Geological Survey of Japan. JB-2 is a recently erupted basalt from Oshima Island of the Izu-Bonin arc, and its geochemical data favorably compare with Troodos tholeiitic samples. Deviation of our data from the recommended values (Imai, 1990) is mostly less than 5% (Table 1), but error in the recommended values themselves may account for those element contents that show greater than 10% difference.

Nd, Sr, and Pb isotope ratios (Table 2; total: five samples) were measured in a multicollector VG Sector 54 thermal ionization mass spectrometer (TIMS) at the Geological Survey of Japan, AIST. Sample treatment, normalization values for Sr and Nd isotope ratios, their within-run precision, their standard sample ratios, mass fractionation correction for Pb isotope ratios, and their internal precision are described in Hoang and Uto (2006). Measured isotope ratios were corrected to initial ratios based on the 91.6 Ma U-Pb zircon age obtained from the ophiolite (Mukasa and Ludden, 1987; we use 90 Ma for all samples) and corresponding element concentrations obtained by ICP-MS (Table 2).

Glass Samples

Lavas comprise mostly pillow lavas with some massive sheet flows, and they are intruded by dikes and sills (Fig. 2 A1). Mutual crosscutting and overlapping relationships show the cogenetic nature of these extrusive and intrusive rocks. In spite of the widespread weathering and local hydrothermal alteration in the lavas, largely unaltered glasses of these lavas and intrusives can be found (Fig. 2 A2, B1, B2). The glasses are black in outcrop, contrasting with altered gray-, brown-, red-, and green-colored rocks inside the glass margins of extrusives and intrusives. Microscopically, such glasses are not devitrified (colorless in plane-polarized light and completely extinct in cross-polarized light), and contain only clinopyroxene microcrystals or microcrystallites (Fig. 2 A2), although we found olivine pseudomorphs (iddingsite) and pores filled by smectite in some thin sections. Gass et al. (1994) reported similar and characteristic microcrystallites in their plate 2.12 (the same photo is in fig. 11.1-a inRogers et al., 1989). Sample CY152 appeared exceptionally fresh (Fig. 2 B1, B2), so we analyzed both the glass rim (CY152-f1) and glassy basalt interior (CY152-f2; Fig. 2 B2), and we also used this sample for Ar-Ar dating. Oxygen isotope data from the Akaki River section samples show no evidence of alteration in nearly all of the samples (Rautenschlein et al., 1985). Although sample CY144 has a low (90.99) total wt% (Table 1) that apparently reflects alteration, the glass appeared fresh petrographically, so we used this sample.

Rock Series Based on Major-Element Composition

Major-element chemistry of the samples allowed us to classify rock series (Fig. 3). Our samples are subalkaline basalt to andesite in composition (Fig. 3A). Although two tholeiitic samples plot near the low-K/medium-K series boundary, all other samples fall into the low-K series field. The Troodos glass can be divided into at least two groups: one comprising tholeiitic samples, and the other plotting in the field of the calc-alkaline rock series (Fig. 3B; Miyashiro, 1975). These samples are boninitic, as detailed in the following.

Our calc-alkaline samples have higher MgO and Mg# (Mg/[Mg + Fe]; 0.57–0.73) compared with tholeiitic samples and modern Izu-Bonin arc volcanic front lavas (Fig. 3C). The calc-alkaline samples also have high Cr and Ni abundances (Table 1). Sample CY119 has an exceptionally high MgO content (Fig. 3C) relative to its silica content. Such high-Mg rocks were also analyzed by Robinson and Malpas (1990, see their fig. 6).

The high-Mg calc-alkaline rocks are characterized by relatively low TiO2. In MgO-TiO2 space (Fig. 3D), data are separated into suite A (“lower suite” in Robinson et al., 1983; their samples were more silicic than corresponding rocks of our sample set), suite B (“upper suite” in Robinson et al., 1983), and suite C in Robinson and Malpas (1990) and Robinson et al. (2003). These researchers interpreted suite C rocks as boninitic.

Pearce and Robinson (2010) showed that the calc-alkaline and tholeiitic rock series classification shown in Figure 3B (Miyashiro diagram) does not characterize the crystallization trends exhibited by the Troodos glasses. They concluded that a better rock series classification would class Troodos lavas as tholeiitic, depleted tholeiitic (suites A and B), and boninitic (suite C).

Combined with other geochemical characteristics reviewed herein, we will employ a classification that differs somewhat from that of Pearce and Robinson (2010). This classification consists of a tholeiitic suite A of the lower pillow lava sequence, a boninite suite B or upper pillow lava, instead of depleted tholeiite as per Pearce and Robinson (2010), and a suite C of depleted boninite that occurs as stratigraphically high units erupted into paleotopographic lows along the Arakapas fault zone, which we refer to as infill lavas after Gass et al. (1995). The classification is based primarily on the FeO/MgO versus SiO2 and TiO2 versus MgO concentrations from our data, in which suites B and C are similar and highly distinct from suite A. In addition, rare earth element (REE) patterns clearly separate the depleted suites B and C from suite A, consistent with previously published REE data on Troodos, particularly Cameron (1985), and, to a lesser extent Rautenschlein et al. (1985) (Fig. 4).

Crawford et al. (1989) proposed a division of boninite suites into two classes, high-Ca boninites and low-Ca boninites. The Troodos boninites represent a reference suite of high-Ca boninites, which have notably lower SiO2, Na2O, and K2O, and higher CaO and FeO* contents than low-Ca boninites. High-Ca boninites have also been found at the northern termination of the Tonga Trench and at the southern termination of the Vanuatu Trench (Crawford et al., 1989). Some of the boninite from the Bonin Islands is high-Ca boninite (e. g., Maehara and Maeda, 2004).

Chondrite-Normalized REE Patterns

The island-arc tholeiite (suite A), boninite (suite B), and depleted boninite (suite C) show contrasting REE patterns (Fig. 4A). The tholeiites are characterized by depletion of light (L) REEs relative to heavy (H) REEs and REE absolute abundances more than ten times chondrite values. The REE patterns of the arc tholeiite samples closely resemble normal mid-ocean-ridge basalt (N-MORB), so the REE pattern cannot discriminate island-arc tholeiite from N-MORB. The Mariana forearc basalt and Oshima basalt REE patterns (Fig. 4D) are also comparable to our tholeiitic REE patterns (Fig. 4A).

Most of our boninite samples of suites B and C show a “spoon-shaped” REE pattern, with lower overall REE abundance less than ten times chondrite values (Fig. 4A). Our boninite samples appear to comprise two groups: suite B boninite and suite C depleted boninite, with REE patterns similar to those reported in Cameron (1985) (Fig. 4B). The depleted boninites (suite C) comprise five samples, all of which are from the Arakapas transform fault zone, and have lower Ce-Nd contents than suite B boninite. However, HREE concentrations of both types of boninite overlap. The “spoon-shaped” REE pattern is well developed in the depleted boninites. Suite B samples from the lower pillow lava sequence along the Akaki River section (Fig. 1) reported by Rautenschlein et al. (1985) also show a slightly “spoon-shaped” REE pattern (Fig. 4C). The “U-shaped” REE pattern (Fig. 4D) of the Chichijima boninites differs markedly from those of Troodos boninites, as absolute HREE concentrations are lower and MREE/HREE values are higher in Chichijima boninites.

MORB-Normalized Trace-Element Patterns

The most distinctive feature in incompatible elements is enrichment of large ion lithophile elements (LILE; such as Rb, Ba, K, Sr), Th, U, and Pb relative to high field strength elements (HFSEs; Nb, Ta, Zr, Hf, Ti, Y), which are shown in MORB-normalized multi-element patterns (Fig. 5A). In addition, high Pb/Ce and Ba/La (LILE/LREE) are evident in the patterns. These characteristics are typical features of subduction zone–related magmas, and because these are glass analyses, the enrichment is not an artifact of alteration.

Thorium (Th) is a key element, because it is enriched in arc lavas, but it is immobile during metamorphism up to the melting temperature (Pearce, 2003) and during seawater alteration, contrasting with U, which is mobile in fluids (Hawkesworth et al., 1997). Enriched Th relative to Nb and Ta (Fig. 5A) is indicative of a suprasubduction-zone setting.

Three island-arc tholeiite patterns (suite A) are similar and very closely coincident in HFSEs but show a bit more scatter for LILEs (Fig. 5A). However, every incompatible element is enriched relative to the suite B and C boninites (Fig. 5A). Most HFSE and REE abundances of suite A are very close to N-MORB (i.e., normalized value of 1; Fig. 5A). Ratios of Nb (or Ta)/La are lower in Troodos tholeiite than N-MORB values, which is also a diagnostic characteristic of arc rocks. Collectively, the Troodos arc tholeiite patterns closely resemble typical intra-oceanic-arc tholeiite of the Izu-Bonin arc (Oshima, JB2, and Hahajima basalt) and that of the Mariana forearc (Fig. 5B).

Troodos boninite patterns are fairly closely grouped for LILEs, but HFSEs patterns are depleted (Fig. 5A), corresponding to suite B “boninite” or suite C “depleted boninite” REE patterns (Fig. 4A). Both types show enrichment of LILEs, Th, and Pb over HFSEs and LREEs. Nb/La values of suite C depleted boninites are comparable to N-MORB and similar to Chichijima boninite (Fig. 5B), whereas Nb/La values of suite B boninite are intermediate between suite C depleted boninite and tholeiite. The type low-Ca boninites of Chichijima have characteristic Zr and Hf enrichment relative to middle (M) REEs (Fig. 5B), unlike the Troodos boninites.

Sr, Nd, and Pb Isotope Ratios

Island-arc tholeiite sample CY27 (suite A) has ɛNd(t) of 6.9 and ɛSr(t) of −9.6 (Fig. 6A; Table 2). The depleted boninites (suite C) have lower ɛɛNd(t) (3.1–6.6) but much higher 87Sr/86Sr(t) 0.704801–0.706199, whereas the boninites (suite B) have age-corrected Nd and Sr isotope ratios comparable to the tholeiites. Our data from suite A tholeiite and suite B boninites are similar to those reported in Rautenschlein et al. (1985), along the Akaki River section. Rogers et al. (1989) published Nd and Sr isotope analyses of six boninites including three suite C samples from the Limassol Forest complex. Rogers et al. (1989) suite C data are characterized by lower 143Nd/144Nd(t) between 0.512577 and 0.512816, and higher 87Sr/86Sr(t) between 0.707155 and 0.707519. Thus, Nd and Sr isotopic ratios clearly distinguish suite C from suites A and B.

Izu-Bonin volcanic front lavas, Hahajima basalts, and Mariana forearc basalts have 87Sr/86Sr similar to Troodos tholeiites, but they have higher 143Nd/144Nd (Fig. 6A). Chichijima boninite samples have lower Nd and higher Sr isotope ratios compared to relevant arc lavas; such higher Sr isotopic ratios are observed in the Troodos depleted boninites (suite C).

In Pb isotope space (Figs. 6C and 6D), our Troodos tholeiites and boninites plot above the 0 Ma Northern Hemisphere Reference Line (NHRL). Among our samples, tholeiite sample CY27 (suite A) has the lowest age-corrected 206Pb/204Pb and 208Pb/204Pb, but its 207Pb/204Pb ratio is intermediate between boninites (suite B) and depleted boninites (suite C). Depleted boninites (suite C) have the highest age-corrected 207Pb/204Pb. Pb isotope data of Rautenschlein et al. (1985) fall in the range outlined by our samples, excluding one sample (Fig. 6D).

Pb isotope ratios for the Cypriot massive sulfides are similar to those of fresh Troodos glasses reported by Rautenschlein et al. (1985), and this suggests that the Pb incorporated in the sulfides is derived from the Troodos basement (Booij et al., 2000), and the above mentioned Pb ratios are primary. Our Pb isotope data are quite similar to those of the Izu-Bonin boninites and Izu-Bonin volcanic front lavas (Figs. 6C and 6D). On the Pb-Nd isotopic diagram (Fig. 6B), all our samples and those of Rautenschlein et al. (1985) plot close to the field of Izu-Bonin boninites.

Xu and Castillo (2004) reported Sr-Nd-Pb isotope compositions of gabbro, its mineral separates (plagioclase and clinopyroxene), and basaltic glass from the Troodos ophiolite. Although their Sr and Nd isotope ratios are similar to our tholeiites and boninites (Fig. 6A), age-corrected 206Pb/204Pb values of their samples are lower than our samples, plotting in the field of Indian Ocean MORBs (Figs. 6C and 6D). Xu and Castillo (2004) thus suggested the Tethyan asthenosphere had the Indian Ocean MORB-type isotopic signature.

GEOCHRONOLOGY AND STRATIGRAPHY

Ar-Ar Ages

Ar-Ar analyses were done at National Taiwan University. For full details on analytical methodology, including irradiation, see Lee et al. (2009).

We obtained Ar-Ar ages from tholeiitic sample CY27, suite A, and depleted boninitic sample CY152, suite C (Fig. 1). CY27 was from a fresh glass margin of a basaltic-andesitic sill, and CY152 was from the exceptionally fresh glassy basaltic-andesite interior of a pillow. Outcrop and thin section photos of sample CY152 are shown in Figure 2B. Our Ar data for each heating step are summarized in Table 3, and the data and statistics associated with the isochrons are presented in Table 4.

Plateau and inverse isochron ages were obtained from both samples. Tholeiitic sample CY27 (suite A) has plateau age of 89.8 ± 0.5 Ma and isochron age of 90.6 ± 1.2 Ma (Fig. 7, left), and boninitic sample CY152 (suite C) has plateau age of 56.7 ± 0.7 Ma and isochron age of 55.5 ± 0.9 Ma (Fig. 7, right).

Samples CY27 and CY152 both exhibit good plateaus in their Ar release spectra (Fig. 7, top). The isochrons are tightly constrained for both sample CY27 and CY152 (Fig. 7, bottom), with only the 400 and 600 °C steps, plotting significantly for sample CY152 off of them (Fig. 7, bottom right). The 40Ar/36Ar intercepts for both samples are close to air, suggesting minimal excess argon. However, the slightly higher 40Ar/36Ar value for sample CY152, coupled with the fact that the isochron age (and plateau age) is younger than the total gas age, suggests some excess argon in this sample. Accordingly, we prefer the isochron age for sample CY152. For sample CY27, the isochron and plateau ages are indistinguishable within their uncertainties, but we will refer primarily to the isochron age for this sample in the following discussion. The 90.6 ± 1.2 Ma Ar-Ar age of the tholeiite, suite A, agrees well with the 90.3 ± 0.7 Ma and 92.4 ± 0.7 Ma U-Pb zircon ages of plagiogranites of the Troodos ophiolite (Mukasa and Ludden, 1987). We will further address the significance of the age data in the discussion section, including providing corroborating stratigraphic data to support the 55.5 Ma age included in suite C.

Sedimentary Cover and Radiolarian Ages

Robertson and Hudson (1974) described the sedimentary cover of the Troodos ophiolite, and Robertson (1975) reported umbers from the base of such cover. Umberiferous chert with radiolarians (main part of the Perapedhi Formation) were observed at site 44CR on the northeastern flank of Troodos, at site 18MR at the Troulli inlier, at sites 111CR and 125CR to the east of the Arakapas transform fault zone, and at sites 132CR, 137CR, and139CR around the southern margin of Limassol Forest complex (Fig. 1). Although there is no exposure of overlying sediments at site 44CR, the other umberiferous chert localities are overlain by white mudstone (included in the Perapedhi Formation), which also contains radiolaria. Umberiferous cherts at 111CR and 125CR overlie volcanic breccia, and the others directly overlie pillow lavas. All contacts appear conformable and unfaulted.

In addition to the white mudstone, red, green, and white mudstones (also included in the Perapedhi Formation) conformably overlie pillow lavas at the other MR sites (Fig. 1). Some white mudstone is tuffaceous, and coarse-grained silicic tuff (the Kannaviou Formation; Robertson, 1977; Geological Survey Department Cyprus, 1995) is distributed along the southwestern margin of the Troodos main massif. These mudstones, also intercalated with thick silicic tuff, contain well-preserved radiolarians (e.g., Osozawa and Okamura, 1993).

The umberiferous chert at the Kalavasos mine yielded a Turonian (ca. 89–94 Ma) radiolarian age (near the sample point of 137CR; Blome and Irwin, 1985; Pseudodictyomitra pseudomacrocephala is the key index fossil; Sanfilippo and Riedel, 1985), but according to Urquhart and Robertson (2000), their sample was not from the ophiolite but from the Moni mélange. The radiolarian assemblage collected from umberiferous chert at Perapedhi east of 47MR (Bragina and Bragin, 1996) also contains old species, Theocorys antique, which was extinct at the Santonian and Campanian boundary (ca. 83.5 Ma) (Sanfilippo and Riedel, 1985), and Amphipyndax pseudoconulus. However, all the radiolarian samples we analyzed, including the umberiferous chert underlying white mudstone, contained Amphipyndax tylotus, which is an index fossil of the late Campanian to Maastrichtian (ca. 75–66 Ma) (Sanfilippo and Riedel, 1985; Osozawa and Okamura, 1993). Other radiolarians are A. pseudoconulus, which is an index fossil of the early Campanian but coexists with A. tylotus in samples from 111CR, 125CR, and 44CR, and Dictyomitra koslovae, Myllocercion acineton, and others. Considering the coexistence of A. tylotus and A. pseudoconulus, all cover sediments we analyzed can be considered late Campanian (ca. 75 Ma; Yamasaki, 1987).

DISCUSSION

Spatial Distribution of Chemostratigraphic Types and Order of Eruption

Our assessment of the chemostratigraphy, described in the geochemistry sections, combines our data with previously published data sets to arrive at the conclusion that our suite A tholeiitic rocks and suite B boninitic rocks crop out primarily along the northern flanks of the Troodos massif, whereas suite C depleted boninites appear restricted to an infill geometry along the Arakapas fault zone. We note that the complex chemostratigraphic relationships at this large ophiolite can lead to differing interpretations, but there is agreement that tholeiite eruption preceded boninite eruption at Troodos, and this relationship and associated details will form a key part of our analysis of the tectonomagmatic evolution of this ophiolite.

Chronology of Troodos Ophiolite Formation

Here, we discuss our new geochronologic data in relation to our geochemical data and in the context of the generation, evolution, and emplacement of the ophiolite. Whereas the 90.6 ± 1.2 Ma Ar-Ar date from tholeiite glass (suite A) is comparable to U-Pb zircon ages of plagiogranites by Mukasa and Ludden (1987), the 55.5 ± 0.7 Ma age for boninitic sample CY152 (included in suite C) requires additional explanation, as does the difference between the ca. 91 Ma ages given previously and the ca. 75 Ma age of the basal cover strata overlying much of the ophiolite.

The tholeiitic lower pillow lava sequence (suite A) has a genetic link with a brittlely deformed early suite of plutonic rocks and sheeted dikes (therefore spreading related), and the boninitic upper pillow lava sequence (suite B) has a genetic link with an undeformed later suite of cumulate (e.g., Dilek et al., 1990). Because plagiogranite belongs to the early suite of plutonic complex, sheeted dikes, and lower pillow lavas (Dilek and Thy, 2009), the agreement between the tholeiite (suite A) age and the plagiogranite ages indicates a reasonable volcanic date.

In contrast, a hypothetical ca. 91 Ma age for boninite (suite B) and depleted boninite (suite C) would require a major unconformity or tectonic contact between lavas of suites B and C and their cover sediments (the Perapedhi Formation), because of late Campanian (ca. 75 Ma) radiolarian ages obtained from the base of this unit. This contradicts the conformable nature of the contact of the cover over both suites B and C. A reasonable interpretation of these relationships is that fairly boninitic volcanism persisted until ca. 75 Ma, although additional Ar-Ar dating of suite B and additional suite C boninite samples is desired for confirmation. The occurrence of siliceous tuff intercalated with the umbers suggests some volcanism persisting to ca. 75 Ma. This, however, does not require that boninitic eruption continued until then, for significant gaps in age between siliceous volcanism and earlier main-stage ophiolite formation are documented in other ophiolites, such as the Coast Range ophiolite of California, where siliceous volcanism may have occurred >20 m.y. after main-stage ophiolite development (e.g., Hopson et al., 2008). Whereas these relationships suggest the possibility of Troodos volcanism lasing until ca. 75 Ma, they do not explain the 55 Ma boninite age. We address this issue next.

Whereas the Perapedhi Formation conformably overlies much of the suite B and C volcanic rocks, as noted already, the younger Lefkara Formation overlies volcanic rocks in the vicinity of the 55 Ma boninite sample. The Lefkara Formation (Robertson and Hudson, 1974) is a thickly bedded chalk unit that nearly encircles the Troodos ophiolite in map distribution (Fig. 1). If the base of the Lefkara Formation is Maastrichtian, as interpreted by Gass et al. (1994), this would indicate that our 55 Ma boninite age is erroneous. Recent biostratigraphic work on the Lefkara Formation shows that the Maastrichtian strata are either lacking or patchy in distribution. The age for most of the Lefkara Formation is middle Paleogene to early Miocene (Kähler and Stow, 1998). Strata as old as late Paleocene are found only in the Ayios Nocolaos area (Fig. 1), along the southwestern margin of the Troodos main massif (far from sample 152), whereas the bases of other sections are early Eocene (Kähler and Stow, 1998). Thus, the revised Lefkara Formation biostratigraphy indicates that its base is mostly younger than the 55 Ma boninite. This fact, coupled with the data indicating a robust age (see geochronology sections), suggests that the age is valid.

Although the Ar-data and stratigraphic data suggest the 55 Ma boninite age is valid, the widespread distribution of Campanian ages of basal cover strata (Fig. 1) suggests that post-Campanian mafic volcanism is rare, even for the infill unit suite C. The collective geochemical, stratigraphic, and geochronologic data suggest that Troodos ophiolite formation began with arc tholeiite magmatism at ca. 91 Ma. The onset of boninitic magmatism occurred sometime after this, and voluminous boninite eruption may have continued until ca. 75 Ma, after which minor boninitic magmatism continued until at least 55 Ma.

Metamorphic Sole of the Troodos Ophiolite? Does it Constrain Subduction Initiation There?

Metamorphic soles, i.e., thin sheets of high-grade, mostly mafic, metamorphic rocks, crop out beneath many ophiolites and are interpreted as the product of subduction initiation in young (hot) oceanic lithosphere (Williams and Smyth, 1973; Spray, 1984; Jamieson, 1986; Hacker, 1990). As such, analysis of the high-grade metamorphic rocks associated with the Troodos ophiolite has some relevance in evaluating possible connections between the formation and/or emplacement of the Troodos ophiolite and subduction initiation. The Troodos ophiolite lacks a coherent metamorphic sole, but scattered outcrops of high-grade metamorphic rocks have been interpreted as a dismembered sole (Malpas et al., 1992; Chan et al., 2007). Controversy exists over whether these rocks represent a dismembered metamorphic sole or whether they formed in a different tectonic environment, such as oceanic transform fault (Spray and Roddick, 1981; Malpas et al., 1992; Chan et al., 2007). The occurrence of metasedimentary rocks, as well as the more common metabasites, may suggest a metamorphic sole origin, rather than a deep fracture zone or ocean core complex, because the latter settings would not result in sufficient burial of metasediments to produce amphibolites-grade metamorphism in them.

The 40Ar/39Ar ages of hornblendes from these amphibolites, interpreted as Troodos metamorphic sole remnants, span 14 m.y. (ca. 90 and 80 Ma for northeastern coast of the Akamas Peninsula, and 76 Ma for two samples near Pafos; Fig. 1), which is an unusually long age range for metamorphic soles (Chan et al., 2007). This spread of ages suggests multiple alternatives, including the following. (1) The high-grade rocks did not form in a metamorphic sole environment and were juxtaposed with the ophiolite by subsequent tectonic and possibly sedimentary processes (if olistostrome, blocks in mélange for example). (2) The amphibolite samples include rocks formed in a metamorphic sole environment as well as rocks formed in other environments, such as deep along oceanic fracture zones. (3) The amphibolite formed in a metamorphic sole environment but in an environment associated with slow subduction initiation. The latter has been proposed to explain the ∼15 m.y. range of high-temperature–high-pressure metamorphic ages associated with high-grade metamorphic rocks of the Franciscan Complex, which may represent a dismembered metamorphic sole formed during subduction initiation (Anczkiewicz et al., 2004). That conclusion is consistent with the results of thermal modeling (Cooper et al., 2011). (4) Amphibolite formed in a subduction zone beneath the ophiolite over an extended period of time owing to the close approach of another spreading ridge on the downgoing plate (Aoya et al., 2002). The position of ridge crest segments on the downgoing plate coupled with relative plate motions (nearly parallel to fracture zones) may have led to a series of ridge crest–subduction zone interactions that may have spanned a significant amount of time (Wakabayashi, 2004). The last scenario may best explain the generation of boninites over an extended period of time, given that such rocks require high temperatures of formation.

The ages derived from the amphibolites spatially associated with the Troodos ophiolite are identical to the age range we have interpreted for most of the volcanic rocks of the ophiolite (ca. 91–75 Ma). This similarity between igneous formation ages of the ophiolite and metamorphic ages of the amphibolite mirrors the near-synchronous ophiolite formation and metamorphic sole metamorphism ages that appear to characterize many suprasubduction-zone ophiolites (Wakabayashi and Dilek, 2000, 2003; Wakabayashi et al., 2010), although Troodos is unusual, or perhaps unique, for the long duration of both amphibolites-grade metamorphism and the arc magmatism, possibly as a consequence of slow initial subduction rate and/or the staggered subduction of multiple ridge segments.

Similarities and Differences among Troodos, Other Suprasubduction-Zone Ophiolites, and Proposed Modern Analogs

Many models of suprasubduction-zone ophiolite generation propose generation of the ophiolite shortly after subduction initiation, with the ophiolite forming during slab rollback and eruption of refractory boninites following initial arc tholeiite eruption (e.g., Stern and Bloomer, 1992; Shervais, 2001; Dilek and Thy, 2009). Whereas the eruptive sequence of tholeiite before boninite characterizes the Troodos ophiolite as well, at least one well-preserved suprasubduction-zone ophiolite, the Betts Cove ophiolite of Newfoundland, has an eruptive sequence of boninite before tholeiite (Bédard et al., 1998), as does the Izu-Bonin arc itself (e.g., Ishizuka et al., 2006).

Based on our new geochronologic and biostratigraphic data, the Troodos ophiolite may represent a much longer record of suprasubduction-zone magmatism than the <10 m.y. (usually <5 m.y.) recorded in most suprasubduction-zone ophiolites (e.g., Shervais, 2001; Tremblay et al., 2011). Voluminous magmatism appears to have persisted at Troodos from ca. 91 Ma to 75 Ma, with some boninitic magmatism taking place as late as 55 Ma. Although the relationship, and even existence of a metamorphic sole for the Troodos ophiolite is somewhat controversial, metamorphic ages and metamorphic relationships are permissive of high-temperature metamorphism that spanned approximately the same period of time as the magmatism. This is also unusual for metamorphic soles, which commonly have a limited age range (e.g., Hacker, 1991; Wakabayashi and Dilek, 2000).

These relationships illustrate a significant degree of variability in basic characteristics of suprasubduction-zone ophiolites and suggest that the formational history of such ophiolites as well as their subsequent emplacement may vary more than commonly surmised. Bédard et al. (1998) raised this point for ophiolites in a single orogenic belt by pointing out differences in the eruptive order of boninite before tholeiite in the Betts Cove ophiolite versus tholeiite before boninite in the Thetford Mines and Bay of Islands ophiolites of the Northern Appalachians. Pearce (2003) also presented a much broader range of modern seafloor possible analogs of suprasubduction-zone ophiolites than is commonly considered by ophiolite geologists.

The unusually long duration of arc magmatism without development of an arc edifice in the Troodos ophiolite may require a somewhat different tectonic model than that proposed for other suprasubduction-zone ophiolites. Next, we present a brief, speculative model that may explain the unusual features of the Troodos ophiolite (Fig. 8).

Adopting other models of suprasubduction-zone ophiolite generation, we propose as others have, that the Troodos ophiolite formed following subduction initiation with subduction rollback and creation of first arc tholeiite and then boninite during subduction rollback, extension above the subducting plate, and an increasing degree of depletion of the mantle wedge in the zone of partial melting above the sinking slab (e.g., Stern and Bloomer, 1992; Shervais, 2001; Dilek and Flower, 2003; Tremblay et al., 2011). Whether subduction initiation was spontaneous or induced (Stern, 2004) cannot be determined with existing data, and it is also unclear whether subduction initiated in young or old oceanic lithosphere.

The extended history of boninite eruption exhibiting progressively greater depletion, as well as the somewhat fragmentary evidence of long-lived, high-grade metamorphism, suggests the arrival of an additional heat source, probably a spreading ridge, on the downgoing plate. Such a scenario resembles that proposed by Shervais (2001), except that we propose that suprasubduction-zone oceanic crust production did not end with ridge subduction. Instead, ridge subduction promoted a further partial melting of already depleted mantle, producing boninite, and then depleted boninite. The ridge subduction event was a prolonged one, resulting in an extended period of boninitic magmatism. Such a prolonged ridge subduction event can result from (1) the subducting ridge axis being subparallel to the plate convergence vector or (2) the subduction of a series transform-offset ridge segments wherein the convergence vector leads to subduction of the ridge segments along approximately the same reach of the trench (Fig. 8C; e.g., Wakabayashi, 2004).

Scientific support was provided by the Cyprus Geological Survey, and accommodation was provided by the Cyprus American Archaeological Research Institute, Nicosia, and Axiothea Hotel, Paphos, Cyprus, all in 1997 (for the first author). We thank R. Russo, J. Shervais, P. Robinson, and J. Bédard for helpful comments. We thank H. Fujimaki, Tohoku University, and K. Uto, AIST, for their assistance at analytical facilities. Y. Kato, University of the Ryukyus, is also appreciated for his guidance. Measurement of paleomagnetism was done by T. Koitabashi. Financial support was provided by the Overseas Research Fund offered by the Ministry of Education, Culture, Sports, Science and Technology, Japan, in 1997.