Formation of the Troodos Ophiolite at a triple junction: Evidence from trace elements in volcanic glass

    Formation of the Troodos Ophiolite at a triple junction: evidence from trace elements in volcanic glass M. Regelous, K.M. Haase, S. Freund, M. Keith, C. Weinzierl, C. Beier, P.A. Brandl, T. Endres, H. Schmidt PII: DOI: Reference:

S0009-2541(14)00379-9 doi: 10.1016/j.chemgeo.2014.08.006 CHEMGE 17315

To appear in:

Chemical Geology

Received date: Revised date: Accepted date:

14 January 2014 6 August 2014 7 August 2014

Please cite this article as: Regelous, M., Haase, K.M., Freund, S., Keith, M., Weinzierl, C., Beier, C., Brandl, P.A., Endres, T., Schmidt, H., Formation of the Troodos Ophiolite at a triple junction: evidence from trace elements in volcanic glass, Chemical Geology (2014), doi: 10.1016/j.chemgeo.2014.08.006

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT Formation of the Troodos Ophiolite at a triple junction: evidence from

RI PT

trace elements in volcanic glass

SC

M. Regelous*, K.M. Haase, S. Freund, M. Keith, C. Weinzierl, C. Beier, P.A. Brandl, T.

PT ED

MA

NU

Endres, H. Schmidt

GeoZentrum Nordbayern, Friedrich-Alexander-Universität Erlangen-Nürnberg, Schlossgarten

CE

5, 91054 Erlangen, Germany

email: tel.: fax:

AC

* Corresponding author:

[email protected] +49 9131 8526064 +49 9131 8529295

Word count (main text): 6574 Word count (abstract): 278

1

ACCEPTED MANUSCRIPT Abstract Fresh volcanic glasses from the extrusive section of the Troodos Ophiolite in Akaki Canyon are tholeiitic and basaltic to dacitic in composition. Compared to normal MORB they have

RI PT

extremely low fractionation corrected Na8, Fe8 and Ti8 and are enriched in fluid-mobile trace elements, including U, Ba, Rb, Sr and Pb, relative to non-fluid mobile elements of similar incompatibility. Trace element compositions of Akaki lavas define an array extending

SC

between 'back-arc lava'-like compositions, and the field defined by Troodos boninites from the upper part of the lava sequence. Troodos lavas were derived from a mantle source that underwent early melt depletion, and later enrichment by both fluids and small degree melts.

NU

These processes can explain the unusual negative correlation of Pb/Ce with Zr/Nb and Ba/Nb in Troodos extrusives. Although some Troodos lavas are similar in composition to lavas from

MA

back-arc spreading centres, the boninites from the upper parts of the lava pile do not appear to have exact compositional equivalents among lavas from fore-arcs, back-arcs or other tectonic settings where similar rocktypes have been recovered. We suggest that the geochemical

PT ED

evolution inferred for the mantle source of Troodos lavas, together with geological evidence is most consistent with an origin for the Troodos Ophiolite at a spreading centre close to a ridge-trench-trench, or ridge-trench-transform triple junction, where highly depleted, subduction-modified, fluid-enriched mantle wedge material was able to upwell and

CE

decompress to shallow depths in a 'fore-arc' location. In such a tectonic setting, arc volcanism is captured by the spreading centre, explaining the lack of evidence for subaerial arc

AC

magmatism in Troodos. Rapid lateral migration of the triple junction could account for the similar ages of other Tethyan supra-subduction zone ophiolites.

1. Introduction Oceanic crust created at mid-ocean ridge spreading centres covers approximately 60% of the Earth's surface, yet much of this crust lies at water depths of >2.5 km, and is covered by sediments. Since there have been very few deep-penetration (>500 m) boreholes into oceanic crust, much of our current understanding of the chemical and physical structure of the oceanic crust is based on the study of ophiolites, which are generally thought to represent fragments of obducted oceanic lithosphere. In many ophiolites, the full spectrum of rocktypes thought to characterise the oceanic crust, from submarine tholeiitic basalts, sheeted dykes, to gabbros overlying depleted mantle peridotites can be examined, temporal variations in the chemical composition of the extrusive crust can be determined, and models for the generation and 2

ACCEPTED MANUSCRIPT evolution of mid-ocean ridge basalts can be tested by analysis of associated melt residues (peridotites) and cumulates (gabbros). However, many ophiolites have chemical compositions unlike mid-ocean ridge basalts (MORB) erupted at presently-active mid-ocean ridges. In

RI PT

particular, high H2O contents and relative enrichments in fluid mobile elements such as Pb, U, K, Rb indicate that the lavas were apparently erupted at spreading ridges in the neighbourhood of subduction zones ('supra-subduction zone ophiolites'), although the precise tectonic setting in which they formed is debated (see recent reviews by Pearce, 2003;

SC

Robertson, 2004; Dilek and Furnes, 2011). It is therefore uncertain to what extent the geological structure of ophiolites, or the mantle melting and magma evolution processes

NU

inferred from ophiolites are representative of 'normal' oceanic crust. The occurrence of boninitic lavas in several ophiolites has been used to argue for a

MA

fore-arc origin, since this relatively rare rocktype characterises the Marianas fore-arc region (e.g. Stern and Bloomer, 1992; Ishizuka et al., 2011). However, boninites have recently been reported from active volcanoes in both back-arc (Resing et al., 2011) and arc settings (Cooper

PT ED

et al., 2010) and so are apparently not restricted to fore-arc regions. If ophiolites generally form in fore-arc settings, and if fore-arc crust is created during subduction initiation, then the mantle melting and melt evolution processes inferred from ophiolites may help in understanding the mechanisms of subduction initiation (Stern et al., 2012).

CE

Trace element studies of lavas from ophiolites can place important constraints on their petrogenesis and thus the tectonic environment of formation, since trace elements can be used to infer the degree of source depletion and fluid input, and the nature of the mantle melting

AC

process. However, most ophiolites have undergone extensive low- and moderate-temperature alteration (Alt and Teagle, 2000), which has significantly modified the chemical composition of the lavas. Many previous trace element studies of ophiolites were carried out on variably altered whole-rock samples. Since the more mobile elements that are characteristic of subduction zone environments are particularly affected by alteration, many previous studies of ophiolites have been restricted to using only the more immobile trace elements (e.g. Pearce 1975). Here we present major and trace element analyses of fresh, unaltered volcanic glasses from the extrusive section of the Troodos Ophiolite of Cyprus, determined using electron probe and laser-ablation ICP-MS. The use of fresh glasses together with microanalytical methods allows us to circumvent the problem of alteration. We have determined the chemical stratigraphy of Troodos lavas in the Akaki Canyon section in order to examine igneous processes responsible for changes in lava composition over 104-105 y 3

ACCEPTED MANUSCRIPT timescales. We use incompatible trace element compositions of the lavas to provide some

RI PT

constraints on the tectonic setting in which this fragment of oceanic crust was formed.

2. Geological setting and sampling

The Troodos Ophiolite covers an area of approximately 3000 km2 in the central part of the island of Cyprus in the eastern Mediterranean. It is one of the best preserved ophiolites and

SC

exposes a complete sequence of marine sediments, lavas, sheeted dikes, isotropic gabbros, layered gabbros and ultramafic rocks representing the oceanic crust and uppermost mantle

NU

(Gass, 1968). Zircon ages from plagiogranites and Ar-Ar ages of lavas indicate an age for the main part of the Troodos ophiolite of 90-92 Ma (Mukasa and Ludden, 1987; Osozawa et al.,

MA

2012). Structural and paleomagnetic studies have identified possible north-south striking former ridge axes in the northern part of the ophiolite (Varga and Moores, 1985; MacLeod et al., 1990), most notably the Solea Rift (Fig. 1a). The tectonic situation is more complex in the

PT ED

southeastern region, the Limassol Forest complex (Murton and Gass, 1986), which is separated from the northern part of the ophiolite by the Arakapas Fault Zone, interpreted as a fossil transform fault (Fig. 1a). Troodos lavas have previously been divided into the Upper and Lower Pillow Lavas. The latter consist of tholeiitic basalts, andesites, dacites and

CE

rhyolites, with the more evolved rocktypes apparently more common at lower stratigraphic levels. The Upper Pillow Lavas include picrite and boninite-like lavas; the most depleted boninites may be restricted to the southern margin of the ophiolite (Osozawa et al., 2012).

AC

The presence of a sheeted dyke complex indicates that the Troodos ophiolite formed in an extensional regime, whereas the major and trace element characteristics of the lavas, in particular the enrichment in volatiles and fluid-soluble elements as well as radiogenic isotope compositions shows that it formed in a 'supra-subduction zone' setting (e.g. Pearce and Robinson, 2010; Muenow et al., 1990; Sobolev et al., 1993; Rautenschlein et al., 1985). However, the exact geodynamic setting in which the Troodos ophiolite formed is still debated. An island arc origin of the ophiolite (Miyashiro, 1973) is unlikely on the basis of geological stucture (Gass, 1968) and major and trace element compositions of the lavas (e.g. Pearce and Robinson, 2010; Osozawa et al., 2012), but various geochemical, petrological, sedimentological and structural arguments have been used to argue for its formation in a juvenile ('proto') arc, back-arc, fore-arc, or plate edge setting (see recent review by Pearce and Robinson, 2010).

4

ACCEPTED MANUSCRIPT Although several previous geochemical studies of Troodos lavas have reported trace element data, most of these analyses were carried out on whole-rock samples, and the interpretation of data for such samples is complicated by the fact that the lavas have

RI PT

undergone pervasive low- and moderate-temperature alteration. Indeed, the initial debate over whether the Troodos lavas were typical MORB or subduction zone lavas was largely centred over the effects of alteration. This problem can be avoided by analysing fresh volcanic glasses, which are abundant in some sections through the Troodos lava pile.

SC

We carried out major and trace element analyses of fresh volcanic glasses from a 3.5 km long section of the Akaki Canyon on the NE flank of the Troodos Ophiolite (Fig. 1b). The

NU

lavas from this locality have previously been divided into the Upper and Lower Pillow lavas (Bear, 1975; Gass and Smewing, 1973) and into Low-Ti and High-Ti lavas (Bednarz and

MA

Schmincke, 1994) on the basis of mineralogy, colour, major element content and the abundance of dikes. Previous detailed lithological and structural mapping (Bednarz and Schmincke, 1994; Schmincke et al., 1983) of the Akaki Canyon section defined twelve

PT ED

different extrusive lava units (labelled A to L; Fig. 1b). In the north, the youngest, highly altered lavas (unit A) are overlain by northward-dipping carbonate-rich sediments. In the south, the lavas pass into the Basal Group, which consists of 50-90% dykes with thin screens of altered lavas. Fresh glass is rare in this unit, which grades southwards into the sheeted dike

CE

section. In the Akaki Canyon section, the boundary between the Upper and Lower Pillow Lavas occurs approximately between Units B and C, whereas the boundary between older High-Ti and younger Low-Ti lavas lies between Units J and K. Our samples were recovered

AC

from units B to L (Fig. 1b); no fresh glass was found in Unit A. We sampled only fresh glasses where these were available, and so the stratigraphic coverage of our sample set is not complete or uniformly distributed within the section, and is biased towards pillow lavas and hyaloclastites in comparison to sheet flows, within which fresh glass is less often present.

3. Analytical methods We found that glassy lava margins are often partially altered along fine cracks and contain abundant small quench crystals, requiring the use of microanalytical geochemical techniques. Fresh volcanic glass fragments were hand-picked, embedded in epoxy and polished for electron microprobe analyses. Analysis of major elements (SiO2, TiO2, Al2O3, FeOT, MnO, MgO, CaO, Na2O, K2O, P2O5, SO3, Cl and F) was carried out using a JEOL JXA 8200 Superprobe electron microprobe at the GeoZentrum Nordbayern, University of Erlangen. The instrument was operated using an accelerating voltage of 15 kV, a beam 5

ACCEPTED MANUSCRIPT current of 15 nA and a defocused beam 10 m in diameter. Counting times were set to 20 s and 10 s for peaks and backgrounds for all elements except for Cl and F (40 and 20 s respectively). Data in Table 1 represent averages of 10 spot analyses of a single glass chip.

RI PT

Repeated measurements of the international standards VG-2 and VG-A99 were used to evaluate accuracy and precision, which is better than 3.5 % for all elements except MnO, P2O5, SO3, Cl and F. More details on the major element analytical methods can be found in Brandl et al. (2013).

SC

Trace element concentrations were measured using laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) at the GeoZentrum Nordbayern, on the same glass

NU

chips used for major element analysis after removal of the carbon coating. Samples were ablated using a UP193FX excimer laser from New Wave Research operated with a beam

MA

diameter of 50 µm, coupled to Agilent 7500i quadrupole mass spectrometer. As internal standard we used the SiO2 values of the samples as determined by electron microprobe. Data in Table 1 represent averages of 3-5 spot analyses of the same glass chip. Glass standard

PT ED

NIST614 was used for external calibration, and accuracy and reproducibility was monitored from repeated measurements of BCR-2g.

4. Results

CE

Most lavas from the Akaki Canyon are aphyric but olivine, plagioclase, clinopyroxene and FeTi-oxides phenocrysts and microphenocrysts occur in some flow units. The Akaki

AC

Canyon glasses analysed in this study have MgO contents of between 1.1 and 8.3 weight % (Fig. 2). At 8% MgO, the glasses have higher Al2O3, slightly higher SiO2, and significantly lower TiO2, FeO and Na2O, compared to MORB tholeiites from 'normal' open ocean spreading ridges such as the East Pacific Rise (Fig. 2). These major element characteristics are however typical of many of the lavas erupted at spreading centres in the vicinity of subduction zones, including lavas from back-arc spreading centres, as discussed in detail in section 5. The range in K2O at a given MgO content, which is also observed in the Mehegan (1988) dataset (Fig. 2) may result from alteration, although Cl concentrations, which are highly sensitive to alteration are negatively correlated with MgO (Fig. 2). Low major element totals in some samples (as low as 93%) may result from the high water contents of Troodos glasses, which can be as high as 6% (Sobolev et al., 1993; Portnyagin et al., 1997); this is supported by the negative correlation between MgO and the difference between the major element total and 100%.

6

ACCEPTED MANUSCRIPT The major element compositions of the glasses from the Akaki Canyon section overlap with those from previous analyses of glass from other localities on both the northern and southern margins of the Troodos ophiolite (Cameron, 1985; Robinson et al., 1983;

RI PT

Schminke et al., 1983; Rautenschlein et al., 1985; Pearce and Robinson, 2010; see Fig. 3) over a distance of about 90 km. Highly evolved lavas with MgO < 1.0%, which occur sporadically in Troodos (Pearce and Robinson, 2010) were not found in the Akaki section. The average Mg# of the Akaki Canyon glasses analysed in this study (about 44) is very

SC

similar to that of the sheeted dike glass margins (about 45), and significantly lower than the average Mg# value of N-MORB of about 56. Due to the lack of fresh glass, our sample set

NU

does not include any of the more Mg-rich lavas which occur in the uppermost stratigraphic levels of the Akaki section. Nevertheless, the average composition of the lavas from Akaki,

MA

and from the sections elsewhere in Troodos studied by Mehegan (1988), is apparently more Si-rich and depleted in Mg compared to average 'open ocean' MORB, and highly fractionated lavas (andesites, dacites) such as found in Troodos, are uncommon along most of the present-

PT ED

day spreading ridge system.

Lavas from lower in the stratigraphy of the Akaki section (lower latitude) generally have lower MgO and higher SiO2 than those from the upper part of the section, but more mafic lavas occur sporadically throughout the section (Fig. 3). Some of these are dykes or

CE

sills, but some may represent flows from adjacent overlapping volcanic centres. The restricted range in major and trace element composition of the glasses within parts of the profile (e.g. between latitudes 35.022 and 35.030, see Fig. 3) suggest that these samples may belong to a

AC

single eruptive event.

Concentrations of highly incompatible trace elements such as Th vary by a factor of approximately 4 (Fig. 4), and are negatively correlated with MgO. This is also the case for the more mobile elements Rb, Ba, U, and this together with the similarity of the trace element patterns in Fig. 4 suggests that the concentrations of these elements have not been significantly affected by alteration. Our trace element data overlap with those of Rautenschlein et al. (1985), who did not however analyse samples from all volcanic units in the Akaki section (Fig. 4). La/Sm ratios vary between 0.8 and 1.1 (Fig. 5), which overlaps the average N-MORB value of 0.95 (Sun and McDonough, 1989). However, the relative concentrations of fluid-mobile elements, Cs, Rb, Ba, U and Pb in Akaki lavas are higher than in typical MORB (Fig. 4). The Troodos glasses have Pb/Ce and U/Nb ratios in the range 0.114 to 0.398, and 0.079 to 0.141 respectively (Fig. 5a); for comparison, most oceanic basalts have Pb/Ce in the range 0.03±0.01 and U/Nb 0.022±0.006 (Hofmann et al., 1986). 7

ACCEPTED MANUSCRIPT Ba/Th ratios of Troodos glasses range between 89 and 148 (Fig. 5b) which is higher than the average MORB value of 50, lower than most island arc lavas (>300), but comparable to the range observed in some back-arc lavas, as discussed in more detail in section 5. Compared to

RI PT

the average MORB composition of Hofmann (1988), Troodos lavas have lower REE concentrations (despite their lower MgO), and a pronounced negative Nb, Ta anomaly (Fig. 4) and are enriched in the fluid mobile elements Rb, Ba, Cs, U and Pb. The least fractionated lavas also have a positive Sr anomaly (Fig. 4), and Sr/Nd ratios correlate positively with

SC

MgO, whereas other incompatible trace element ratios (Ba/Th, Zr/Nb, La/Sm) do not vary systematically with MgO.

NU

The Akaki lavas with the lowest Pb/Ce ratios tend to have the lowest Ba/Th and highest Zr/Nb, Ba/Nb, U/Nb and La/Nb ratios, and overlap in composition with many lavas

MA

from back-arc spreading centres, for example the Lau Basin (Fig. 6, 7). Akaki lavas with the highest Pb/Ce extend towards the field defined by Troodos boninites (Fig. 6, 7). Although Troodos boninites have Pb/Ce ratios similar to many arc lavas (e.g. those from Tonga), they

PT ED

have significantly lower Ba/Nb, U/Nb, Ba/Th, Zr/Nb, Ba/Zr (Fig. 6, 7). Incompatible element ratios such as La/Sm and Ba/Nb show no systematic stratigraphic change within the section of the Akaki Canyon sampled in this study (Fig. 3). The 2 samples from Unit B (Upper Pillow Lavas) have similar compositions to those from the

AC

5. Discussion

CE

underlying units (Lower Pillow Lavas).

5.1. Implications for the chemical stratigraphy of Troodos lavas Previous geochemical studies of the Troodos lava pile exposed in the Akaki Canyon and elsewhere on Cyprus have divided the volcanic section into two or three different lava series, in addition to the boninitic lavas that are found at the highest stratigraphic levels in some sections, and suggested that these represent separate, unrelated lava sequences. Most recently, Pearce and Robinson (2010) combined data for glasses from 11 sections through the lava pile on both the northern and southern margins of the Troodos complex, and argued that the Troodos lavas could be divided into a boninitic Upper Series and a tholeiitic Lower Series, the latter comprising 3 Units. In this classification scheme, our samples lie within the fields defined by Units II and III, the two younger units of the Lower Series (Fig. 8). Much of the heterogeneity in major and trace element composition identified by previous studies in

8

ACCEPTED MANUSCRIPT this area, most of which were carried out on whole-rock samples, may be due to the effects of alteration and crystal accumulation. Compared to the boninitic Upper Lavas, the most magnesian tholeiitic lavas in our

RI PT

dataset have lower SiO2, slightly higher TiO2, Al2O3, CaO and Na2O at an MgO value of 8 wt% (Fig. 2). Trace element compositions of the Akaki lavas lie between the fields defined by Lau back-arc lavas and the boninitic Upper Lavas, which have high Pb/Ce and low Zr/Nb and La/Nb (Fig. 6, 7). The possible genetic relationships between the Upper and Lower Lavas are

SC

explored in more detail below. As already noted, there are no systematic variations in incompatible trace element ratios with stratigraphic depth in the Akaki section (Fig. 3),

NU

although the tholeiitic lavas with trace element compositions most similar to the boninitic Upper Lavas (high Pb/Ce, low Zr/Nb) mainly occur higher in the section (Fig. 3).

MA

The major element compositions of the Akaki lavas overlap with those of lavas from other sections through the lava pile on both the northern and southern margins of the Troodos Ophiolite (Fig. 2). On the basis of the Pearce and Robinson (2010) data set, there are no

PT ED

obvious systematic geographical variations in the major element composition of Troodos lavas, although lavas from the southern margin of the ophiolite are apparently more MgOrich. This could reflect limited stratigraphic sampling of individual sections, since in the Akaki Canyon there is an overall progression to more MgO rich lavas with increasing

CE

stratigraphic height above the sheeted dyke section, and due to the lack of fresh glass in the youngest part of the Akaki section, we did not analyse any samples of the olivine-rich lavas immediately underlying the sediments, which outcrop to the east of Agrokipia (Fig. 1b).

AC

Osozawa et al. (2012) propose that the youngest flows on the southern margin of the Troodos extrusive section are composed of high MgO, boninitic 'infill' lavas that are associated with the Arakapas Fault Zone, and are significantly younger than the youngest flows on the northern margin. Further detailed sampling of additional sections through the Troodos lava pile is required to confirm this, and to determine how representative the lava section exposed in the Akaki Canyon is of the extrusive part of the Troodos crust. 5.2. The role of fractional crystallisation and contamination in Akaki Canyon lavas Pearce and Robinson (2010) suggested that the lavas of the Lower Series could be related by fractional crystallisation. For our new set of glass samples from the Akaki Canyon, major element concentrations are generally well correlated with MgO (Fig. 2), consistent with a single tholeiitic liquid line of descent. The low MgO contents of Akaki lavas, the inflections in the MgO-TiO2 and MgO-FeO diagrams, and the relatively uniform incompatible trace 9

ACCEPTED MANUSCRIPT element ratios of the lavas suggest that much of the major element variation within the lavas is due to fractional crystallisation. The decreasing FeO, TiO2 and increasing SiO2 at MgO <3.5% reflect the typical tholeiitic late onset of crystallisation of titanomagnetite (Pearce and

RI PT

Robinson, 2010). Some of the differences in major element composition between the Troodos lavas and 'open-ocean' MORB result from the higher water contents of the former, which influences the fractionation path followed by primitive melts. At higher water contents, plagioclase crystallisation is suppressed so that clinopyroxene precedes plagioclase on the

SC

liquidus in the most primitive magmas. In the Troodos lavas, the MgO-Al2O3 diagram (Fig. 2) indicates the onset of plagioclase crystallisation occurs at around 6.0% MgO, a lower MgO

NU

value than for the 'wet' and the 'damp' Lau Basin back-arc lavas analysed by Bezos et al. (2009), and the 'moist' lavas of the Oman Ophiolite (MacLeod et al., 2013), indicating that the

MA

Troodos lavas were positively sodden in comparison. This is supported by measurements of H2O in Troodos lavas (Muenow et al., 1990; Sobolev et al., 1993, Portnyagin et al., 1997) which lie in the range 2-6%. Decreasing CaO and CaO/Al2O3 with decreasing MgO over the

PT ED

whole range of MgO contents indicate that clinopyroxene was an important fractionating phase.

Cl/K ratios of the least evolved Akaki glasses average 0.25, similar to MORB glasses from fast-spreading ridges. Cl/Nb ratios increase by a factor of 4-5 as MgO contents decrease

CE

from 8.0 to 1.0% (Fig. 2). This variation cannot be explained by crystal fractionation of the observed mineral phases, nor by alteration of the glasses. Although Cl concentrations are sensitive to interaction with seawater, alteration is unlikely to result in a systematic variation

AC

of Cl and Cl/Nb with MgO, whereas both Cl and Cl/Nb are negatively correlated with MgO in Troodos glasses (Fig. 2). The variation in Cl/Nb suggests that the more evolved magmas interacted with Cl-enriched, hydrothermally-altered crustal rocks in shallow crustal magma chambers, similar to many MORB lavas from fast-spreading ridges which extend to higher Cl and Cl/Nb (Jambon et al., 1995; Michael and Cornell, 1998). On the other hand, incompatible element ratios such as Ba/Nb, U/Nb, Pb/Ce are not correlated with MgO, nor with Cl/Nb, but correlate with each other (e.g. Fig. 6), indicating that the variations within these trace element ratios are not due to crystal fractionation or crustal assimilation processes and instead reflect source variations, as discussed further below. Some incompatible element ratios, for example La/Sm and Ba/Th, have a restricted range (Fig. 5b), however there are significant variations in other trace element ratios, for example Pb/Ce and Zr/Nb (Fig. 6), that require source heterogeneity.

10

ACCEPTED MANUSCRIPT The least evolved Akaki lavas analysed in this study (with 7.0-9.0% MgO) have FeO, TiO2 and Na2O values of 6.5-8.0, 0.56-0.81, and 1.76-2.01 respectively, much lower than found in 'open ocean' MORB (Fig. 9). Lavas from the Upper Series lavas of Troodos (Pearce

RI PT

and Robinson, 2010) extend to even lower Na2O and TiO2 (Fig. 9). Relatively low Fe8, Ti8 and Na8 (oxide concentrations normalised to 8% MgO) also characterise lavas from many back-arc spreading centres, but values as low as the Akaki lavas are rare in back-arc basins (Fig. 9) and are more commonly associated with island-arc lavas. However, some lavas from

SC

the Fonualei Spreading Centre, located close behind the arc in the Tonga back-arc have similarly low Fe8, Ti8 and Na8 (Escrig et al., 2012). The differences in major element

NU

composition at similar MgO content between Troodos lavas and 'open-ocean' MORB likely reflect differences in the source and/or the melting processes. In general, low Fe8, Ti8 and Na8

MA

values are favoured by high T, low P melting of relatively depleted mantle in the presence of

PT ED

H2O (e.g. Langmuir et al., 2006; see Fig. 9), as discussed in more detail below.

5.3. Relationship between the tholeiitic and boninitic lavas As discussed above, much of the range in incompatible trace element ratios in the Akaki glasses results from melting processes and source composition, rather than low-

CE

pressure fractionation and contamination processes. In many incompatible trace element ratio diagrams, the Akaki glasses define arrays extending between compositions similar to those of lavas from back-arc spreading centres which contain a small subduction component (e.g.

AC

lavas from the Eastern Lau Spreading Centre (ELSC) in the Lau Basin back-arc, Escrig et al., 2009; Bezos et al., 2009), towards the broad field defined by the boninitic Troodos Upper Lavas analysed by König et al. (2008), see Fig. 6, 7. This observation is interesting because it suggests that the combination of mantle source and melting processes that produced the boninitic Upper Lavas were also to a lesser extent important during the formation of the tholeiitic Lower Lavas, and that the two suites are therefore related, in contrast to some previous suggestions (Pearce, 1975; Dilek and Flower, 2003). The tholeiitic Akaki lavas lying furthest from the field defined by Troodos boninites (those samples with the lowest Pb/Ce and highest Zr/Nb) have similar HREE and HFSE systematics to N-MORB, with ratios of incompatible, fluid-immobile elements, e.g. Zr/Nb (Fig. 7), Sm/Yb and Zr/Yb (Fig. 10) which overlap with the depleted end of the range of MORB compositions. However compared to MORB all Akaki lavas are enriched in fluidmobile elements Pb, U, Ba and Rb, relative to fluid-immobile elements of similar 11

ACCEPTED MANUSCRIPT incompatibility The high Pb/Ce (0.114-0.398) and U/Nb (0.079-0.141) ratios of Akaki lavas are not as extreme as those of island-arc lavas, e.g. those from Tonga, but are similar to some lavas from back-arc spreading centres from the Lau Basin, such as those from the ELSC,

RI PT

which contain a small 'subduction component' (Fig. 5a). At the other end of the Akaki array, the most 'boninite-like' glasses with the highest Pb/Ce also have slightly higher Ba/Th, Zr/Ba, but lower Ba/Nb and U/Nb (Fig. 6, 7). They have significantly higher Nb/Ce and lower Zr/Nb ratios, which overlap with enriched MORB

SC

values and in this respect differ from most lavas from back-arc spreading centres (Fig. 6). Despite their high Pb/Ce ratios, which overlap with those of Tonga lavas and imply a greater

NU

subduction input, they lie far from the field defined by most arc lavas, having very much lower La/Nb and Zr/Nb ratios, which overlap with those for highly enriched MORB (Fig. 7b).

MA

In addition, the significant enrichment in Pb is not accompanied by high Ba/Th, Ba/Nb, U/Th or U/Nb ratios, although Ba and U are like Pb fluid-mobile (Fig. 5, 6). In the Pb/Ce - Nb/Ce diagram (Fig. 6a), the Akaki lavas define a curved array with

PT ED

positive slope, extending between MORB/BAB compositions towards the field of Troodos boninites. The curvature of the array indicates that it does not result from simple mixing, and implies that variations in source composition and/or the melting process created the range in Akaki lava compositions, ranging from back-arc-like tholeiitic lavas containing a small

CE

subduction component, to more 'boninitic-like' tholeiitic lavas having higher Pb/Ce, Zr/Nb. In several diagrams, an extension of the Troodos data array towards 'less boninitic' compositions does not intersect the field of MORB. For example, in Fig. 6a, 7a and 7b, a projection of the

AC

Troodos array to higher La/Nb, lower Pb/Ce, away from the boninite field does not pass through the MORB field. This requires a multistage evolution of Troodos lavas from 'normal' upper mantle compositions, as discussed below.

5.4. Mantle source characteristics of Akaki lavas The unusual trace element compositions of the most boninite-like Akaki lavas requires a complex, multi-stage source history, including previous melt extraction event(s), subsequent or concomitant enrichment in fluid-soluble elements, and later addition of small-degree melts. These processes are indicated by (1) relatively low HREE concentrations, low Na8, Ti8 and low Sm/Yb and Zr/Yb ratios relative to mid-ocean ridge and ocean island basalts, which require prior source depletion; (2) an enrichment in fluid-soluble elements relative to fluidinsoluble elements of similar incompatibility (e.g. high Pb/Ce, Sr/Nd) which indicates fluid enrichment; (3) low Zr/Nb and moderate La/Sm ratios indicate source enrichment in the 12

ACCEPTED MANUSCRIPT highly incompatible, fluid-insoluble elements, including Nb, Th and La. This latter enrichment process apparently occurred by small-degree melt re-enrichment of the mantle source, and its effect is most evident in the boninites which were derived from the sources

RI PT

most refractory in major elements, i.e. those having undergone the most melt extraction. Below, we discuss these processes and their relative timing in more detail.

In Fig. 10a, Akaki lavas lie at the depleted end of the MORB array, with low Sm/Yb and Zr/Yb. The boninitic glasses of the Upper Lava Series extend to even more depleted

SC

compositions, but lie on an extension of the MORB array which suggests that the low Sm/Yb, Zr/Yb ratios result from melt extraction. Assuming that the Akaki lavas result from relatively

NU

large (10-30%) degrees of mantle melting, then their source must have been previously depleted by 1-2% melt extraction compared to the mantle that melts beneath 'normal'

MA

spreading ridges (Fig. 10a).

In addition to this melt depletion event there is clear evidence for an enrichment in fluid-soluble elements, including Ba, Rb, U and Pb. All Akaki lavas, including those lying

PT ED

furthest from the boninitic field have higher Pb/Ce, U/Nb than MORB (Fig. 5a) and overlap with the field for back-arc basin basalts. The Akaki lavas with the lowest Zr/Yb tend to have the highest Ba/Yb and Pb/Yb (Fig. 10). Either the enrichment in fluid mobile elements Ba and Pb occurred after the melt depletion event that was responsible for low Zr/Yb, and thus

CE

affected the most depleted samples to the greatest extent, or sources which received most fluid were melted and depleted to the greatest extent. In contrast, if variable source depletion had followed a uniform enrichment in fluid, an array with positive slope in Fig. 10b would be

AC

expected. Assuming that Nb and Ce are not fluid-mobile, then the observed Pb/Ce and U/Nb ratios of Akaki glasses require that 65-90% of the Pb, and 72-84% of the U in these samples is fluid-derived (Fig. 5a). Melting of a melt-depleted but fluid-enriched mantle source can explain the trace element compositions of the Akaki lavas with the lowest Pb/Ce ratios, which have trace element compositions very similar to those of many back-arc basin basalts, for example those from the Lau Basin. However, an additional enrichment process is required to explain the compositions of Troodos boninites and the Akaki lavas with the highest Pb/Ce ratios. Despite the very low Sm/Yb and Zr/Yb ratios of these lavas (Fig. 10a), which imply a highly depleted mantle source, they also have low La/Nb and Zr/Nb values, which overlap with those of EMORB (Fig. 7). Low La/Nb ratios could result from smaller degrees of melting, or from the presence of a greater proportion of a component enriched in the highly incompatible elements such as Nb. The fact that some Troodos boninites have lower La/Nb ratios than many 13

ACCEPTED MANUSCRIPT EMORB (Fig. 7), yet low absolute Zr, Yb concentrations and low Sm/Yb ratios suggests that the latter explanation is more likely. We therefore suggest that Akaki magmas were produced by melting of a highly

RI PT

depleted, subduction-modified mantle source, which was variably re-enriched by small degree melts. Previous studies of boninites from other locations have also demonstrated that the highly depleted mantle sources of these lavas record small-degree melt refertilisation (e.g. Hickey and Frey, 1982; Falloon and Crawford, 1991; Bedard, 1999). This melt enrichment

SC

process would affect the trace element budget of the most depleted sources (boninite source) to the greatest extent. This process is especially evident in Nb and Ta, because for these

NU

elements there is the greatest contrast in concentration between the depleted source and the reenriching melt. In contrast, the highly incompatible but fluid soluble elements Ba, U, Rb were

MA

already present in the source at elevated concentrations due to fluid addition. Thus the effect of subsequent small-degree melt enrichment is to decrease the Ba/Nb and U/Nb ratios, whilst having little effect on element ratios such as U/Th, Ba/Th. This small-degree melt

PT ED

refertilisation process can also account for the REE systematics of Troodos boninites, which have high La/Ce ratios but low Sm/Yb ratios compared to average MORB (Fig. 11). Refertilisation by small-degree melts leads to an enrichment in LREE relative to the heavy and middle REE, thus creating the 'spoon-shaped' REE patterns observed for the most

CE

depleted Troodos lavas (König et al., 2008; Osozawa et al., 2012) from an originally highly LREE depleted source. Our model is shown more quantitatively in Fig. 11, in which the depleted upper mantle composition of Workman and Hart (2008) is depleted further by 2%

AC

fractional melting to reproduce the low Sm/Yb and Zr/Yb ratios of the boninitic samples. Fluid addition then results in an enrichment in Rb, Ba, Sr and Pb, and subsequent addition of a small degree melt of the Workman and Hart (2008) upper mantle composition to the meltdepleted, fluid-enriched mantle source results in greater enrichment of Nb, Ta relative to Ba, Rb, and thus a decrease in the Ba/Nb ratio (Fig. 6), as well as LREE enrichment. Like most melting models, the end result is highly model-dependent, but it is the simplest model that reproduces reasonably well the compositions of the Troodos lavas and the unusual orientation of the data arrays in Fig. 6, 7. The boninitic samples with the lowest Zr/Nb also have the highest Pb/Ce (Fig. 6), implying that the small-degree melt inferred to be responsible for Nb enrichment also carried significant amounts of Pb. König et al. (2008) have shown that Troodos boninites apparently contain a greater proportion of Pb derived from subducted sediment. Combined trace element

14

ACCEPTED MANUSCRIPT and isotope data are needed to determine whether the sedimentary Pb is introduced as fluid or melt.

RI PT

5.5. Trace element constraints on the tectonic setting of Troodos The combination of pillow lavas containing a 'subduction' signature, underlain by a well-developed sheeted dyke complex has led to a general agreement that the Troodos Ophiolite formed in 'supra-subduction zone' spreading environment. However, it is debated

SC

whether this was an arc, proto-arc, fore-arc, or a back-arc setting (see recent review by Pearce and Robinson, 2010). Tectonic models for the origin of the Troodos Ophiolite must be able to

NU

explain the petrology, chemistry and chemical stratigraphy of the extrusives, namely the progression from tholeiitic andesites and basaltic andesites in the lower section, locally

MA

overlain by more MgO-rich rocktypes including picritic basalts and depleted boninites, all of which are enriched in fluid-mobile Ba, Rb, U, Pb relative to elements of similar incompatibility compared to 'normal, open-ocean' MORB. Previously, few trace element

PT ED

analyses of fresh Troodos glasses existed (König et al., 2008; Osozawa et al., 2012; Rautenschlein et al., 1985; Robinson et al., 1983; Thy and Xenophontas, 1991), which could be used to examine these relationships in more detail. In addition, since the earliest geochemical studies of the Troodos lavas were carried out, there has become available far

CE

more high-quality trace element data for lavas from fore-arcs as well as back-arc spreading centres. A detailed comparison of the trace element systematics of fresh, unaltered glasses from Troodos with those from other tectonic settings is therefore warranted, in order to place

AC

new constraints on the tectonic environment in which the Troodos Ophiolite was formed. In trace element diagrams, Akaki lavas define arrays extending from 'back-arc lava like' compositions with moderate depletion in REE and HFSE and enrichment in fluid-mobile elements such as Rb, Ba, U, Pb, towards the broad field defined by Troodos boninites, with lower (depleted) Zr/Yb and relative enrichment in Nb and Pb (e.g. Fig. 6-8, 11). Boninites are relatively rare volcanic rocks with high SiO2 (>52%) at high MgO (>8%) and low TiO2 (<0.5%, Le Bas, 2000), which are thought to result from melting of refractory (clinopyroxenepoor lherzolite or harzburgite) and H2O-rich mantle, possibly at elevated temperature, leaving a harzburgitic residue. Based on the melting model of Langmuir et al. (2006), the low Fe8, Ti8 and Na8 of Akaki glasses (Fig. 9) could be explained by melting of mantle that is depleted, water-rich, and possibly hotter (1350-1400°C), compared to the mantle beneath normal spreading ridges and most back-arc basins (Pearce and Robinson, 2010).

15

ACCEPTED MANUSCRIPT The tectonic environments in which boninite magmas may form are unclear. Recentlyerupted boninites or 'boninite-like' lavas have been reported from the northern termination of the Tonga Arc and rear-arc (Falloon et al., 2007; 2008; Crawford et al., 1989; Falloon and

RI PT

Crawford, 1991; Danyushevsky et al., 1995; Resing et al., 2011; Sobolev and Danyushevsky, 1994), the Valu Fa Ridge (Kamenetsky et al., 1997), Central Tonga Arc (Cooper et al., 2010), and from the southern New Hebrides Arc (Monzier et al., 1993). Izu-Bonin-Marianas (IBM) boninites outcrop in the fore-arc region and are of Eocene age, and their exact tectonic

SC

environment of formation is debated (Deschamps and Lallemand, 2003; Ishizuka et al., 2011). We have compiled existing trace element data for boninites and 'boninite-like' lavas

NU

from these locations for comparison with Troodos lavas and associated boninites. The IBM boninites and 'fore-arc basalts' are enriched in fluid-mobile elements Pb, U, Ba and overlap

MA

with Troodos tholeiites (and many back-arc lavas) in composition (Fig. 12). However, IBM boninites do not have the extreme low, depleted Zr/Yb ratios seen in Troodos boninites. Boninites from the northern termination of the Tonga Trench (Falloon et al., 2007; 2008)

PT ED

extend to high Pb/Ce, but are also not as depleted in terms of Zr/Yb as the Troodos boninites. Some Papua New Guinea (PNG) boninites show evidence for source re-enrichment (König et al., 2008), as also inferred for Troodos boninites. However, the PNG boninites have not only relatively high Nb/Yb, but also high Zr/Yb, unlike Troodos lavas (Fig. 12). None of these lava

CE

series therefore has the combination of extremely low Zr/Yb (melt depletion), high Pb/Ce (fluid enrichment), and relatively high Nb (re-enrichment by small degree melts) that

(Fig. 12).

AC

characterises the Troodos boninites analysed by König et al. (2008) and Osozawa et al. (2013)

Our trace element data indicate that the Troodos Ophiolite was formed in a tectonic setting in which a spreading axis was underlain by highly-depleted, subduction modified mantle. A currently popular model is that many boninite-bearing ophiolites may have formed during subduction initiation, and represent obducted fore-arc fragments (e.g. Stern et al., 2012). Pearce and Robinson (2010) proposed a STEP fault (trench-transform 'plate edge') setting for the Troodos Ophiolite, whereas Osozawa et al. (2012) suggest that ridge subduction may account for the chemical characteristics of Troodos crust; however both these models appeal to subduction initiation to explain the location of a supra-subduction zone spreading ridge in such settings. Deschamps and Lallemand (2003) pointed out that many boninites, possibly including those now preserved in the IBM forearc, were apparently erupted close to ridge-trench-trench or ridge-trench-transform triple junctions. We suggest that the Troodos Ophiolite formed in 16

ACCEPTED MANUSCRIPT such a triple junction setting, where subduction modified mantle, having been previously depleted by melt extraction due to fluid addition beneath the arc, underwent further melting during upwelling to shallow levels beneath the ridge axis, producing extremely depleted

RI PT

boninitic magmas. (Fig. 13). This could arise from propagation of a back-arc spreading ridge into the arc or fore-arc region, similar to the environment of formation of boninites inferred by Falloon and Crawford (1991); Monzier et al. (1993), Meffre et al. (1996), and specifically for Troodos boninitic lavas by Flower and Levine (1987). In such settings, the distinction

SC

between 'back-arc' and 'fore-arc' may be meaningless, since in modern settings where a spreading ridge approaches closely to the trench, arc magmatism is often captured by the

NU

ridge (e.g. Fonualei Ridge, Hunter Ridge). This could explain why no evidence for subaerial volcanic activity is observed in the sediments overlying the Troodos lavas. Oceanic crust

MA

situated on the overriding plate close to the trench and underlain by depleted, buoyant mantle would have a significant chance of eventual obduction and preservation. Boninites or boninitic-like lavas have been reported from similar active settings on Fonualei Ridge (Escrig

PT ED

et al., 2012), and Valu Fa Ridge (Kamenetsky et al., 1997), where a spreading ridge is propagating into arc crust. We note that such triple junctions can potentially migrate rapidly along the trench, leaving behind remnants of supra-subduction zone 'fore-arc' oceanic crust of a narrow age range and potentially explaining the similar ages of other Tethyan ophiolites. In

CE

our model, the Troodos Ophiolite was not associated with 'subduction initiation', but was formed in a broadly 'fore-arc' setting. Improved understanding of the 3D geochemical structure of the Troodos lava pile would give further insights into the previous tectonic setting

AC

of this fragment of oceanic crust, in particular its location and orientation relative to a former subduction zone.

17

ACCEPTED MANUSCRIPT 6. Summary Compared to normal MORB, tholeiitic lavas from the Akaki Canyon section of the Troodos Ophiolite have extremely low Na2O, FeO and TiO2 at a given MgO and are enriched in fluid-

RI PT

mobile trace elements, including U, Ba, Rb, Sr and Pb, relative to fluid-immobile elements of similar incompatibility. Although some Troodos lavas are similar in composition to lavas from back-arc spreading centres, the boninites from the upper parts of the lava pile do not

SC

appear to have exact compositional equivalents among lavas from fore-arcs, back-arcs or other tectonic settings where similar rocktypes have been recovered. Troodos lavas were derived from a mantle source that underwent previous melt depletion, and enrichment by both

NU

fluids and small degree melts. Although the general geological structure of the Troodos crust may therefore resemble that produced at normal fast-spreading mid-ocean ridges, the

MA

chemical composition of primitive Troodos magmas and their subsequent low-pressure evolution was very different. We suggest that the geochemical evolution inferred for the mantle source of Troodos lavas, together with geological evidence is most consistent with an

PT ED

origin for the Troodos Ophiolite at a spreading centre close to a ridge-trench-trench or ridgetrench-transform triple junction, where highly depleted, subduction-modified, fluid-enriched mantle wedge material was able to upwell and decompress to shallow depths. In this interpretation, the Troodos Ophiolite and possibly other Tethyan ophiolites of similar age

CE

were not created as a result of subduction initiation, but were formed in a 'fore-arc' location

AC

immediately behind an active trench.

Acknowledgments

We thank H. Brätz and R. Klemd for help with the LA ICPMS trace element analyses. We are grateful to the two reviewers and editor L. Reisberg for their constructive comments.

References Alt, J.C., Teagle, D.A.H., 2000. Hydrothermal alteration and fluid fluxes in ophiolites and oceanic crust. Geol. Soc. Am. Spec. Pap. 349, 273-282. Batiza, R., Niu, Y., 1992. Petrology and magma chamber processes at the East Pacific Rise 9°30'N. J. Geophys. Res. 97, 6779-6797. Bedard, J.H., 1999. Petrogenesis of boninites from the Betts Cove Ophiolite, Newfoundland, Canada: identification of subducted source components. J. Petrol. 40, 1853-1889.

18

ACCEPTED MANUSCRIPT Bednarz, U., Schmincke, H.-U., 1994. Petrological and chemical evolution of the northeastern Troodos extrusive series, Cyprus. J. Petrol. 35, 489-523. Bear, L.M., 1975. The geology and mineral resources of the Akaki-Lythrodondha area. Geol.

RI PT

Survey Department Cyprus Memoir 3, pp 122. Bezos, A., Escrig, S., Langmuir, C.H., Michael, P.J., Asimow, P.D., 2009. Origins of chemical diversity of back-arc basalts: a segment-scale study of the Eastern Lau Spreading Centre. J. Geophys. Res. 114, B06212.

SC

Brandl, P.A., Regelous, M., Beier, C., Haase, K., 2013. High mantle temperatures following rifting caused by continental insulation. Nature Geosci. 6, 391-394.

NU

Cameron, W.E., 1985. Petrology and origin of primitive lavas from the Troodos Ophiolite, Cyprus. Contrib. Mineral. Petrol. 89, 239-255.

MA

Cooper, L.M., Plank, T., Arculus, R.J., Hauri, E.H., Hall, P.S., Parman, S.W., 2010. High-Ca boninites from the active Tonga Arc. J. Geophys. Res. 115, B10206. Crawford, A.J., Falloon, T.J. Green, D.H., 1989. Classification, petrogenesis and tectonic

Hymann pp 1-49.

PT ED

setting of boninites. In: Crawford, A.J. (ed.) Boninites and related rocks. London, Unwin

Danyushevsky, L.V., Sobolev, A., Falloon, T., 1995. North Tongan high-Ca boninite petrogenesis: the role of Samoan plume and subduction zone-transform fault transition. J.

CE

Geodynam. 20, 219-241.

Deschamps A., Lallemand, S., 2003. Geodynamic setting of Izu-Bonin-Mariana boninites. Geol. Soc. Lond. Spec. Publ. 219, 163-185.

AC

Dilek, Y., Flower, M.F.J., 2003. Arc-trench roll-back and forearc accretion: 2. A model template for ophiolites in Albania, Cyrus and Oman. Geol. Soc. Lond. Spec. Publ. 218, 43-68.

Dilek, Y., Furnes, H., 2011. Ophiolite genesis and global tectonics: geochemical and tectonic fingerprinting of ancient oceanic lithosphere. Geol. Soc. Am. Bull. 123, 387-411. Escrig, S., Bezos, A., Goldstein, S.L., Langmuir, C.H., Michael, P.J., 2009. Mantle source variations beneath the Eastern Lau Spreading Centre and the nature of subduction components in the Lau Basin - Tonga Arc system. Geochem. Geophys. Geosys. 10, Q04014. Escrig, S., Bezos, A., Langmuir, C.H., Michael, P., Arculus, R.J., 2012. Characterising the effect of mantle source, subduction input and melting in the Fonualei Spreading Center, Lau Basin: constraints on the origin of the boninitic signature of the back-arc lavas. Geochem. Geophys. Geosys. 13, doi:10.1029/2012GC004130. 19

ACCEPTED MANUSCRIPT Falloon, T.J., Crawford, A.J., 1991. The petrogenesis of high-calcium boninite lavas dredged from the northern Tonga ridge. Earth Planet. Sci. Lett. 102, 375-394. Falloon, T.J., Danyushevsky, L.V., Crawford, T.J., Maas, R., Woodhead, J.D., Eggins, S.M.,

RI PT

Bloomer, S.H., Wright, D.J., Zlobin, S.K., Stacey, A.R., 2007. Multiple mantle plume components involved in the petrogenesis of subduction-related lavas from the northern termination of the Tonga Arc and northern Lau Basin: evidence from the geochemistry of arc and backarc submarine volcanics. Geochem. Geophys. Geosys. 8, Q09003.

SC

Falloon, T.J., Danyushevsky, L.V., Crawford, A.J., Meffre, S., Woodhead, J.D., Bloomer, S.H., 2008. Boninites and adakites from the northern termination of the Tonga Trench:

NU

implications for adakite petrogenesis. J. Petrol. 49, 697-715. Flower, M.F.J., Levine, H.M., 1987. Petrogenesis of a tholeiitic-boninitic sequence from

Mineral. Petrol. 97, 509-524.

MA

Ayios Mamas, Troodos Ophiolite: evidence for splitting of a volcanic arc? Contrib.

Gass, I.G., 1968. Is the Troodos Massif of Cyprus a fragment of Mesozoic ocean floor?

PT ED

Nature 220, 39-42.

Gass, I.G., Smewing, J.D., 1973. Intrusion, extrusion and metamorphism at constructive margins: evidence from the Troodos Massif, Cyprus. Nature 242, 26-29. Hickey, R.L., Frey, F.A., 1982. Geochemical characteristics of boninite series volcanics:

CE

implications for their source. Geochim. Cosmochim. Acta 46, 2099-2115. Hofmann, A.W., Jochum, K.P., Seufert, M., White, W.M., 1986. Nb and Pb in oceanic basalts: new constraints on mantle evolution. Earth Planet. Sci. Lett. 79, 33-45.

AC

Hofmann, A.W., 1988. Chemical differentiation of the Earth: the relationship between mantle, continental crust, and oceanic crust. Earth Planet. Sci. Lett. 90, 297-314. Ishizuka, O., Tani, K., Reagan, M.K., Kanayama, K., Umino, S., Harigane, Y., Sakamoto, I., Miyajima, Y., Yuasa, M., Dunkley, D.J., 2011. The timescales of subduction initiation and subsequent evolution of an oceanic island arc. Earth Planet. Sci. Lett. 306, 229-240. Jambon, A., Deruelle, B., Dreibus, G., Pineau, F., 1995. Chlorine and bromine abundance in MORB: the contrasting behaviour of the Mid-Atlantic Ridge and East Pacific Rise and implications for chlorine geodynamic cycle. Chem. Geol. 126, 101-117. Jenner, F.E., O'Neill, H.St.C., 2012. Analysis of 60 elements in 616 ocean floor basaltic glasses. Geochem. Geophys. Geosys. 13, Q02005. Kamenetsky, V.S., Crawford, A.J., Eggins, S., Mühe, R., 1997. Phenocryst and melt inclusion chemistry of near-axis seamounts, Valu Fa Ridge, Lau Basin: insight into mantle wedge melting and the addition of subduction components. Earth Planet. Sci. Lett. 151, 205-223. 20

ACCEPTED MANUSCRIPT König, S., Münker, C., Schuth, S., Garbe-Schönberg, D., 2008. Mobility of tungsten in subduction zones. Earth Planet. Sci. Lett. 274, 82-92. König, S., Münker, C., Schuth, S., Luguet, A., Hofmann, J.E., Kuduon, J., 2010. Boninites as

RI PT

windows into trace element mobility in subduction zones. Geochim. Cosmochim. Acta 74, 684-704.

Langmuir, C.H., Bezos, A., Escrig, S., Parman, S.W., 2006. Chemical systematics and hydrous melting of the mantle in back-arc basins. AGU Geophys. Monogr. 166, 87-146.

SC

MacLeod, C.J., Allerton, S., Gass, I.G., Xenophontas, C., 1990. Structure of a fossil ridgetransform intersection in the Troodos Ophiolite. Nature 348, 717-720.

NU

MacLeod, C.J., Lissenberg, C.J., Bibby, L.E., 2013. 'Moist MORB' axial magmatism in the Oman Ophiolite: the evidence against a mid-ocean ridge origin. Geology 41, 459-462.

MA

Meffre, S., Aitchison, J.C., Crawford, A.J., 1996. Geochemical evolution and tectonic significance of boninites and tholeiites from the Koh Ophiolite, New Caledonia. Tectonics 15, 67-83.

PT ED

Mehegan, J.M., 1988. Temporal, spatial and chemical evolution of the Troodos Ophiolite, Cyprus: supra-subduction zone magmatism in the Tethyan Sea. PhD thesis, Dalhousie University, Halifax.

Michael, P.J., Cornell, W.C., 1998. Influence of spreading rate and magma supply on

CE

crystallisation and assimilation beneath mid-ocean ridges: evidence from chlorine and major element chemistry of mid-ocean ridge basalts. J. Geophys. Res. 103, 18325-18356. Miyashiro, A., 1973. The Troodos ophiolitic complex was probably formed in an island arc.

AC

Earth Planet. Sci. Lett. 19, 218-224. Monzier, M., Danyushevsky, L.V., Crawford, A.J., Bellon, H., Cotten, J., 1993. High-Mg andesites from the southern termination of the New Hebrides island arc (SW Pacific). J. Volc. Geotherm. Res. 57, 193-217. Moores, E.M., Robinson, P.T., Malpas, J., Xenophontas, C., 1984. Model for the origin of the Troodos Massif, Cyprus, and other mideast ophiolites. Geology 12, 500-503. Muenow, D.W., Garcia, M.O., Aggrey, K.E., Bednarz, U., Schmincke, H.-U., 1990. Volatiles in submarine glasses as a discriminant of tectonic origin: application to the Troodos Ophiolite. Nature 343, 159-161. Mukasa, S.B., Ludden, J.N., 1987. Uranium-lead isotopic ages of plagiogranites from the Troodos ophiolite, Cyprus, and their tectonic significance. Geology 15, 825-828. Murton, B.J., Gass, I.G., 1986. Western Limassol Forest Complex, Cyprus: part of an Upper Cretaceous leaky transform fault. Geology 14, 255-258. 21

ACCEPTED MANUSCRIPT Osozawa, S., Shinjo, R., Lo, C.-H., Jahn, B.-M., Hoang, N., Sasaki, M., Ishikawa, K., Kano, H., Hoshi, H., Xenophontos, C., Wakabayashi, J., 2012. Geochemistry and geochronology of the Troodos Ophiolite: an SSZ ophiolite generated by subduction

RI PT

initiation and an extended episode of ridge subduction? Lithosphere 4, 497-510. Pearce, J.A., 1975. Basalt geochemistry used to investigate past tectonic environments on Cyprus. Tectonophys. 25, 41-67.

Pearce, J.A., 2003. Supra-subduction zone ophiolites: the search for modern analogues. Geol.

SC

Soc. Am. Spec. Paper 373, 269-293.

Pearce, J.A., Robinson, P.T., 2010. The Troodos ophiolitic complex probably formed in a

NU

subduction initiation, slab edge setting. Gondwana Res. 18, 60-81. Plank, T., Langmuir, C.H., 1998. The chemical composition of subducting sediment and its

MA

consequences for the crust and mantle. Chem. Geol. 145, 325-394. Portnyagin, M.V., Danyushevsky, L.V., Kamanetsky, V.S., 1997. Coexistence of two distinct mantle sources during formation of ophiolites: a case study of primitive pillow-lavas

PT ED

from the lowest part of the volcanic section of the Troodos Ophiolite, Cyprus. Contrib. Mineral. Petrol. 128, 287-301.

Rautenschlein, M., Jenner, G.A., Hertogen, J., Hofmann, A.W., Kerrich, R., Schmincke, H.U., White, W.M., 1985. Isotopic and trace element composition of volcanic glasses from

CE

the Akaki Canyon, Cyprus: implications for the origin of the Troodos Ophiolite. Earth Planet. Sci. Lett. 75, 369-383. Reagan, M.K., Ishizuka, O., Stern, R.J., Kelley, K.A., Ohara, Y., Blichert-Toft, J., Bloomer,

AC

S.H., Cash, J., Freyer, P., Hanan, B.B., Hickey-Vargas, R., Ishii, T., Kimura, J.-I., Peate, D.W., Rowe, M.C., Woods, M., 2010. Fore-arc basalts and subduction initiation in the Izu-Bonin-Mariana system. Geochem. Geophys. Geosys. 11, Q03X12. Regelous, M., Niu, Y., Wendt, J.I., Batiza, R., Greig, A., Collerson, K.D., 1999. Variations in the geochemistry of magmatism on the East Pacific Rise at 10°30'N since 800 ka. Earth Planet. Sci. Lett. 168, 45-63. Resing, J.A., Rubin, K.H., Embley, R.W., Lupton, J.E., Baker, E.T., Dziak, R.P., Baumberger, T., Lilley, M.D., Huber, J.A., Shank, T.M., Butterfield, D.A., Clague, D.A., Keller, N.S., Merle, S.G., Buck, N.J., Michael, P.J., Soule, A., Caress, D.W., Walker, S.L., Davis, R., Cowen, J.P., Reysenbach, A.L., Thomas, H., 2011. Active submarine eruption of boninite in the northeastern Lau Basin. Nature Geosci. 4, 799-806.

22

ACCEPTED MANUSCRIPT Robertson, A., 2004. Development of concepts concerning the genesis and emplacement of Tethyan ophiolites in the Eastern Mediterranean and Oman regions. Earth Sci. Rev. 66, 331-387.

RI PT

Robinson, P.T., Melson, W.G., O'Hearn, T., Schmincke, H.-U., 1983. Volcanic glass compositions of the Troodos Ophiolite, Cyprus. Geology 11, 400-404.

Schmincke, H.-U., Rautenschlein, M., Robinson, P.T., Mehegan, J.M., 1983. Troodos extrusive series of Cyprus: a comparison with oceanic crust. Geology 11, 405-409.

SC

Sobolev, A.V., Portnyagin, M.V., Dmitriev, L.V., Tsameryan, O.P., Danyushevsky, L.V., Konokova, N.N., Shimizu, N., Robinson, P.T., 1993. Petrology of ultramafic lavas and

NU

associated rocks of the Troodos Massif, Cyprus, Petrology 1, 331-361. Sobolev, A.V., Danyushevsky, L.V., 1994. Petrology and geochemistry of boninites from the

MA

north termination of the Tonga Trench: constraints on the generation conditions of primary high-Ca boninite magmas. J. Petrol. 35, 1183-1211. Stern, R.J., Reagan, M., Ishizuka, O., Ohara, Y., Whattam, S., 2012. To understand

PT ED

subduction initiation, study forearc crust: to understand forearc crust, study ophiolites. Lithosphere 4, 469-483.

Sun, S.-s., McDonough, W.F., 1989. Chemical and isotopic systematics of oceanic basalts: implications for mantle composition and processes. Geol. Soc. Lond. Spec. Publ. 42, 313-

CE

345.

Thy, P., Xenophontos, C., 1991. Crystallisation orders and phase chemistry of glassy lavas from the pillow sequences, Troodos Ophiolite, Cyprus. J. Petrol. 32, 403-428.

AC

Varga, R.J., Moores, E.M., 1985. Spreading structure of the Troodos ophiolite. Geology 13, 846-850.

Workman, R.K., Hart, S.R., 2005. Major and trace element composition of the depleted MORB mantle (DMM). Earth Planet. Sci. Lett. 231, 53-72.

Figure captions Figure 1. (a) Simplified geological map of the Troodos Ophiolite in Cyprus modified from Osozawa et al. (2013) showing the location of Akaki Canyon and the area in Fig. 1b, (b) topographic map of Akaki Canyon area showing the locations of the samples analysed in this study, together with volcanic units A-L as defined by Schmicke et al. (1983) and the sediment lava boundary from Bear (1975). Units A and B correspond approximately to the Upper Pillow Lavas as mapped by Bear (1975). Grid squares in (b) are 1 km across.

23

ACCEPTED MANUSCRIPT Figure 2. Major element variations in volcanic glasses from the Akaki Canyon (black symbols, data from this study) and from elsewhere in Troodos (blue symbols, data from Mehegan, 1988). Shown for comparison are data for basalts, basaltic andesites and

RI PT

andesites from the East Pacific Rise spreading centre (Batiza and Niu, 1992; Regelous et al., 1999).

Figure 3. Variations in major and trace element composition with latitude (stratigraphic

SC

position within the Akaki Canyon section, which youngs from south to north). MgO tends to increase, and SiO2 decreases upsection, whereas incompatible trace element ratios show

NU

no systematic variation with age. The sampled section does not extend to the youngest flows (the 'Upper Series' of Pearce and Robinson, 2010) which are more MgO-rich and

MA

locally picritic, because of the lack of fresh glass.

Figure 4. Trace element compositions of Akaki glasses (data from this study and

PT ED

Rautenschlein et al., 1985) normalised to average MORB (Sun and McDonough, 1989). Compared to basalts from 'normal' active spreading centres, Akaki lavas are enriched in Rb, Ba, Th, U, Pb but have lower Nb, Ta concentrations.

CE

Figure 5. Variations in (a) Pb/Ce with U/Nb, and (b) La/Sm with Ba/Th (note log scales) for Akaki glasses (this study) and Troodos boninites (Osozawa et al., 2013; König et al., 2008), together with data for the Eastern Lau Basin Spreading Centre (ELSC; Bezos et al.,

AC

2009; Escrig et al., 2009) and Fonualei Rift (Escrig et al., 2012), MORB (Jenner and O'Neill, 2012) and Tonga island arc lavas from Tofua Island (Regelous et al., in prep.). Akaki lavas have higher Ba/Th, Pb/Ce and U/Nb than MORB, and La/Sm ratios which overlap with the depleted end of the MORB array. Dashed lines in (a) show the percentage of Pb and U that must be added by fluid to produce high Pb/Ce and U/Nb ratios assuming initial upper mantle ratios of 0.03 and 0.022 respectively (see text). Star symbol shows average global subducting sediment composition (Plank and Langmuir, 1998). Data for Akaki Canyon glasses from Rautenschlein et al. (1985) shown in (b).

Figure 6. Variations of Nb/Ce, Ba/Nb, Zr/Ba and Zr/Nb with Pb/Ce (note log scales). Akaki lavas with the highest Pb/Ce ratios also have the highest Nb/Ce, lowest Zr/Nb, and show only slightly lower Zr/Ba, unlike arc lavas and many lavas from back-arc spreading centres. Curved array in (a) indicates that the range in Pb/Ce is not the result of simple 24

ACCEPTED MANUSCRIPT mixing processes. Extrapolation of the Troodos array in (b) to 'less boninitic' compositions does not intersect the MORB array, indicating a multi-stage history of the source of these

RI PT

lavas. Sources for literature data as for Fig. 5.

Figure 7. Variation of Zr/Nb and Zr/Ba with La/Nb (note log scales). Despite their depleted major element compositions and low Yb concentrations, Troodos boninites and some Akaki lavas have trace element ratios which overlap with the 'enriched' end of the MORB

SC

array. The field defined by Akaki tholeiitic lavas extends to higher La/Nb, Zr/Nb and low Zr/Ba, similar to some back-arc lavas, but unlike either 'open-ocean' MORB or island arc

NU

lavas. Sources for literature data as for Fig. 5.

MA

Figure 8. MgO and TiO2 for Troodos glasses from Akaki (black symbols, data from this study), with data for volcanic glasses from elsewhere in Troodos (Mehegan, 1988; Pearce and Robinson, 2010; blue symbols) for comparison. Akaki lavas sampled in this study fall

PT ED

into the 2 youngest units of the older Lower Series as defined by Pearce and Robinson (2010).

Figure 9. Fractionation-corrected Ti8, Na8, Fe8 for lavas from active back-arc spreading

CE

centres (from Langmuir et al., 2006), together with measured Na2O, TiO2, FeO for Akaki glasses (red symbols, data from this study and Mehegan, 1988) and other Troodos glasses (blue symbols, data from Mehegan, 1988) with between 7-9% MgO (i.e. no fractionation

AC

correction applied). Field for 'normal' Atlantic MORB shown for comparison. All Troodos lavas have lower Ti8, Na8, Fe8 than Atlantic MORB, and also lower than most back-arc lavas. The more MgO-rich Troodos Upper Series lavas have most extreme values. Green symbols are taken from the melting model of Langmuir et al. (2006) and show qualitatively the effects of temperature, source fertility (Na, Ti), and water on pooled melt compositions. The low Fe8, Ti8 and Na8 of Akaki glasses could be explained by melting of mantle that is relatively hot (1350-1400°C), depleted, and water-rich, compared to the mantle beneath normal spreading ridges and most back-arc basins, as also inferred by Pearce and Robinson (2010). See text for further discussion.

Figure 10. Variations of Sm/Yb, Pb/Yb and Nb/Yb with Zr/Yb (note log scales). Assuming that Zr and Yb are fluid immobile, low Zr/Yb ratios of Akaki lavas and Troodos boninites reflect a depleted mantle source, and the displacement of the Troodos samples above the 25

ACCEPTED MANUSCRIPT 'mantle array' indicates that Pb and some Nb, but no Sm is added via subduction input. Lines show compositions of 1-30% melts of upper mantle (Workman and Hart, 2005) that has undergone prior depletion as a result of 1% and 3% equilibrium (batch) melting.

RI PT

Sources for literature data as for Figure 5.

Figure 11. Trace element compositions of Akaki lava ET4-01-1 (this study) and boninite CY152f1 (Osozawa et al., 2013) normalised to average MORB (Sun and McDonough,

SC

1988). Also shown is composition of 20% melt of DMM (Workman and Hart, 2005) having undergone 2% prior depletion by fractional melting (curve a.), which reproduces

NU

the low Zr/Yb, Sm/Yb ratios of the boninite. Troodos lavas have undergone enrichment of fluid-mobile elements Ba, Rb, Pb, Sr and the boninites record a later small-degree melt

MA

refertilisation event which increased Nb, Ta concentrations relative to Rb, Ba, and La relative to Ce. This results from the greater contrast in concentration in Nb and Ta compared to Ba, Zr between the depleted, fluid-enriched source, and the small degree

PT ED

refertilising melt (here represented by 0.5% melt of DMM, curve b.), which therefore has a greater effect on the budget of the former elements. This process is illustrated by the dotted curves, which show trace element compositions produced by mixing of variable amounts of small degree melt (b.) to the previously melt-depleted but fluid-enriched source. See text

CE

for more details.

Figure 12. Zr/Yb, Nb/Yb, Ba/Nb and Pb/Ce in Troodos lavas from Akaki and Troodos

AC

boninites (König et al., 2008; Osozawa et al., 2013), together with fore-arc basalts and boninites from IBM (Reagan et al., 2010), and boninites from PNG and from northern Tonga for comparison. Although some boninites from other localities have high Pb/Ce which overlap with Troodos lavas, none have the extreme depletion in Zr/Yb coupled with low Ba/Nb and high Nb/Yb that characterises Troodos boninites. Some boninites from PNG have elevated Nb/Yb due to re-enrichment (König et al., 2008), but also high Zr/Nb.

Figure 13. Schematic diagram showing inferred tectonic setting of formation of the Troodos Ophiolite at a ridge-trench-trench (a), or ridge-trench-transform (b) triple junction. In both scenarios, a back-arc spreading centre propagates into arc crust, resulting in upwelling and melting of melt-depleted, fluid-enriched mantle wedge, 'capturing' adjacent arc volcanism and creating 'supra-subduction zone' oceanic crust in a fore-arc location (grey shaded region). Such triple-junction settings have been proposed as the primary site of boninite 26

ACCEPTED MANUSCRIPT genesis (Deschamps and Lallemand, 2003). In (b), supra-subduction zone ophiolite crust adjacent to the transform would not be covered by younger arc lavas. Migration of the triple junction along the plate boundary (dashed arrow) could potentially produce

AC

CE

PT ED

MA

NU

SC

RI PT

ophiolitic 'fore-arc' crust of similar age along the length of the trench or transform.

27

ACCEPTED MANUSCRIPT

AC

CE

PT ED

MA

NU

SC

RI PT

Figure 1s

28

ACCEPTED MANUSCRIPT

AC

CE

PT ED

MA

NU

SC

RI PT

Figure 1b

29

ACCEPTED MANUSCRIPT

AC

CE

PT ED

MA

NU

SC

RI PT

Figure 2

30

ACCEPTED MANUSCRIPT

AC

CE

PT ED

MA

NU

SC

RI PT

Figure 3

31

ACCEPTED MANUSCRIPT

AC

CE

PT ED

MA

NU

SC

RI PT

Figure 4

32

ACCEPTED MANUSCRIPT

AC

CE

PT ED

MA

NU

SC

RI PT

Figure 5

33

ACCEPTED MANUSCRIPT

AC

CE

PT ED

MA

NU

SC

RI PT

Figure 6

34

ACCEPTED MANUSCRIPT

AC

CE

PT ED

MA

NU

SC

RI PT

Figure 7

35

ACCEPTED MANUSCRIPT

AC

CE

PT ED

MA

NU

SC

RI PT

Figure 8

36

ACCEPTED MANUSCRIPT

AC

CE

PT ED

MA

NU

SC

RI PT

Figure 9

37

ACCEPTED MANUSCRIPT

AC

CE

PT ED

MA

NU

SC

RI PT

Figure 10

38

ACCEPTED MANUSCRIPT

AC

CE

PT ED

MA

NU

SC

RI PT

Figure 11

39

ACCEPTED MANUSCRIPT

AC

CE

PT ED

MA

NU

SC

RI PT

Figure 12

40

ACCEPTED MANUSCRIPT

AC

CE

PT ED

MA

NU

SC

RI PT

Figure 13

41

ACCEPTED MANUSCRIPT

Trace element data for fresh volcanic glasses from the Troodos Ophiolite.

PT

Troodos lavas derived from melt-depleted mantle source which was enriched by fluids and small-degree melts Troodos Ophiolite formed at a ridge-trench-trench or ridge-trench-transform triple junction

AC

CE

PT

ED

MA NU SC

RI

Model could explain origin of fore-arc crust and similar ages of many Tethyan ophiolites

82