BONINITES THROUGH TIME AND SPACE : PETROGENESIS AND GEODYNAMIC SETTINGS

The article provides an overview of boninitic magmatism occurrences in space and time and shows that the boninite rock series were generated through the entire geological history of the Earth. In modern environments, the genesis of boninites is related to intra‐oceanic subduction initiation. Boninites are typical members of suprasubduc‐ tion zone ophiolite sequences in the Phanerozoic fold belts and also present in the early Precambrian greenstone belts. A comparative study on compositions of the early Precambrian and Phanerozoic boninites indicate their evolu‐ tion through time due to gradual transition from the early thick‐plate tectonics to the modern thin‐plate tectonics. A link between subduction initiation and mantle‐plume impingement at the oceanic lithosphere is discussed.


INTRODUCTION
Boninites attracted much attention at the end of the 20th century, primarily, due to deep-sea studies of forearc slopes in the modern plate convergence zones of the southwestern Pacific.At that time, however, detailed geochemical studies of many ophiolite sections revealed the presence of boninites unknown for midoceanic ridges, but abundant in juvenile island-arc basements.Following the common postulation that ophiolites are remnants of the ancient oceanic crust, ophiolites in fold belts were largely interpreted as relic sutures which mark sites of paleo-ocean closure.As the boninite findings were ever-growing in both the classical fragments of the 'ancient oceanic crust', e.g. the Troodos in Cyprus, and Semail in Oman, or 'classical sutures' such as, the Main Urals Fault in Russia, this inevitably raises the question on the ophiolite nature.
Another problem concerns the evolution of boninitic magmatism as boninites have been discovered in the early Precambrian greenstone belts.Until recently, it was believed that in the early stages of the Earth evolution, 'dry' komatiitic volcanism was predominant and then followed by 'wet' boninite volcanism [Hall, Hughes, 1993].However, the komatiite abundance in the Archean seems to be well overestimated.Komatiites are lacking in many belts and constitute less than 5 % of the volcanic rocks in most Archean greenstone belts [ de Wit, Ashwal, 1997].Detailed geochemical studies of mafic-ultramafic complexes in greenstone belts show that a majority of olivine spinifex-free 'komatiites' and 'komatiitic basalts' belong in fact to the boninite series rocks.According to [Furnes et al., 2014], boninitic magmatism has been occurred through the geological history of the Earth.However, aspects of its evolution have not been thoroughly discussed in the literature yet.
Thus, the aims of the paper are to provide an overview of the boninitic magmatism occurrences through space and time, track its evolution, and propose a geodynamic explanation of this highly informative phenomenon.

DEFINITION AND GEOCHEMICAL CHARACTERISTICS OF BONINITES AND THE BONINITE SERIES
The International Union of Geological Sciences (IUGS) classification of the high-Mg volcanic rocks defined a boninite as a volcanic rock with the following arbitral chemical composition recalculated to 100 wt %: SiO 2 >52 wt %; MgO>8 wt %, and TiO 2 <0.5 wt % [ Le Bas, 2000].The Ti content is strongly constrained in the classification as a critical requirement because of titanium is the most incompatible among the major elements, and its low content in primitive melts indicates that their mantle source was highly depleted.Boninite definitions in many publications are very similar to the above-mentioned one.According to [Crawford et al., 1989], SiO 2 content in boninite lavas exceeds 53 wt %, and Mg#>0.6,where Mg#=Mg/[Mg+Fe 2+ ].It was recommended in [Taylor et al., 1994], to classify volcanic and hypabyssal rocks as boninites if SiO 2 >53 wt %, TiO 2 <0.6 wt %, and 7<MgO<25 wt %.
The above definitions of boninites are based on chemistry, rather than on mineralogy of these rocks.Neither mineralogical nor petrographic characteristics could be included in the classical definition of boninites as the modal mineralogy of phenocrysts show highly variable compositions, and their textural patterns vary widely also.The traditional approach distinguishes between two major groups of boninites, high-Ca and low-Ca [Crawford et al., 1989], characterized by CaO/Al 2 O 3 >0.75 and <0.6, respectively.Boninites with intermediate ratios are grouped as low-Ca type 3 boninites.The upper pillow lavas of the Troodos (Cyprus) ophiolite are distinguished as the tectonic and petrographic type of high-Ca boninites.Low-Ca type 1 boninites are best represented by the Cenozoic lavas of New Caledonia.Low-Ca type 2 boninites include high-Mg andesite lavas of the Baja peninsula, California and the Shikoku Island, Japan.The latter of very special geochemical characteristics and tectonic settings have their own names, bajaites and sanukites, respectively [Rogers, Saunders, 1989;Tatsumi, Ishizaka, 1981].Low-Ca type 3 boninites are most fully represented in the Bonin Islands, Japan where the boninites have been described for the first time at the end of the 19th century.
Despite the fact that two clearly different geochemical groups (high-Ca and low-Ca boninites) are recognized, geodynamic settings for particular types of boninitic magmatism occurrences have not been specified yet [Meffre et al., 1996].The assemblage including both high-Ca and low-Ca boninites and intermediate-Ca boninites have been reported from the Mariana island arc [Arculus et al., 1992].In other occurrences, such as the Troodos, Oman and Josephine ophiolites, only high-Ca boninites are present [Cameron, 1985;Crawford et al., 1989;Ishikawa et al., 2002;Harper, 2003].Low-Ca and intermediate boninites are dominant in the Ordovician ophiolites of Newfoundland [Bédard, 1999] (Table 1).
Boninites are spatially and genetically related to primitive island-arc low-Ti lavas, thus generating a need for recognition of a separate magmatic series known as 'the boninite series' [Pearce, Robinson, 2010;and others].The less differentiated rocks of the series are known from the literature as low-Ti tholeiites, LOTI, [Brown, Jenner, 1989] or low-Ti ophiolite basalts [Sun, Nesbitt, 1978], and they are typical of numerous ophiolites.In many cases, the boninite series volcanic rocks are represented mainly by primitive lavas (MgO>12 wt %) that can be termed as picrites or mistakenly classified as komatiitic basalts or basaltic komatiites [Cameron et al., 1979].
Crystallization of low-Ti tholeiite melts can be schematically given as follows: olivine → clinopyroxene → plagioclase, and the scheme for MORB melts is: olivine → plagioclase → clinopyroxene [Cameron et al., 1980;Natland, 1981].This difference is due to drastic distinc-tions between water-saturated melts of the boninite series and dry melts of the komatiitic and tholeiitic series.The komatiitic series trend is generated by dry mantle melting and directed towards a field of MOR tholeiites, while the boninite series trend reflects the evolution of the primary composition towards the quartz apex, which is typical of wet melting at gradually decreasing temperature and pressure.Distinctions in the compositional trends of the high-Mg volcanic series are well depicted in the projection of the normative basalt tetrahedron Ol-Pl-Qt plotted by using the data on the boninite series of the North Karelian greenstone belt and the komatiitic series of the Kostomuksha greenstone belt in the Baltic Shield (Fig. 1).
The trace elements patterns of the boninites demonstrate strong depletion of the mantle source and concurrent evidences of their suprasubduction genesis, such as, for instance, negative anomalies of Nb(Ta) (Fig. 2).Different types of the boninites have spectra of the same type with distinct negative anomalies of Nb(Ta) and Ti, positive anomalies of Sr and Zr(Hf) and apparent enrichment in large-ionic lithophile (LIL) elements (Rb, Ba, Cs, U, and Th) relative to N-MORB.Such a kind of the trace elements patterns clearly suggests that (i) petrogenesis of boninite series should be excluded any and even a minimum crustal input, and (ii) melting requires a mantle source more depleted in comparison to the mantle lherzolite that generates MORB melts.In other words, melts of the boninite series seem to be derived from a depleted harzburgite mantle source [Hickey, Frey, 1982;Crawford et al., 1989].
On the other hand, there is geochemical evidence that there is a direct link between the boninite series rocks and island-arc tholeiites (IAT), taking into account the commonly accepted idea that juvenile continental crust portions are generated in the island-arc zones (Fig. 3).Thus, the boninite series rocks are of crucial importance for deciphering crust-forming processes through the geological history of the Earth.

BONINITES IN THE GEOLOGICAL TIME AND SPACE
The global occurrences of boninite series rocks are shown on Figure 4 and at Table 2. Boninites have been discovered in all the continents, except South America, which is attributable to the fact that South America is still relatively poorly covered by geological studies.The boninite series rocks are described in various sequences through the geologically documented history of the Earth, from the most ancient Eoarchean Isua belt in the southwestern Greenland to the recent eruptions of submarine volcanoes in the northeastern Lau basin.The only lacuna in their history is the Mesoproterozoic, the period of relatively passive tectonic events, possibly T a b l e 1. Selected analyses representative of boninite series of different ages which are mentioned in the text and were used for the geochemical plots Т а б л и ц а 1. T a b l e 1 (end) Т а б л и ц а 1 (окончание) N o t e. HCB -High-Ca boninite, ICB -intermediate-Ca boninite, LCB -Low-Ca boninite; LOTI -Low-Ti basalt/picrite.Subdivision of boninites is after [Crawford et al., 1989].П р и м е ч а н и е.HCB -высококальциевые бониниты, ICB -умереннокальциевые бониниты, LCB -низкокальциевые бониниты, LOTI -низкотитанистые базальты/пикриты.Разделение бонинитов по [Crawford et al., 1989].related to the existence of the stable supercontinent Nuna [Cawood, Hawkesworth, 2014].
Contemporary settings for generation of the boninite series rocks are known in the areas of intraoceanic island-arc systems of the southwestern Pacific, specifically the Izu-Bonini-Mariana (IBM) and Tonga-Kermadec island arcs.Boninites compose mostly forearc slopes, forming the lowest stratigraphic levels in the island-arc architecture.These regions are natural laboratories for comprehensive studies of the boninitic magmatism coupled with geodynamic modeling of its genesis.It is noteworthy that since the Paleocene onwards there is a strong link between boninites and ophiolites, as confirmed by the data on the Cape Vogel Peninsula, Papua New Guinea (see Table 2).Actually, this genetic link has long been known [Sun, Nesbitt, 1978;Cameron et al., 1979] and led to introduction of the term 'suprasubduction zone (SSZ) ophiolite' in the 1980s to acknowledge that some ophiolites are more closely related to island arcs than to ocean ridges [Pearce, 1982].This term was readily accepted in the literature, and the majority of the known ophiolite complexes, including the largest ones with the com-pletely preserved sequences, e.g.Troodos, Oman, were classified into the SSZ type.
According to [Sklyarov et al., 2016], there are four main types of boninites in the ophiolite sequences (Fig. 5).Type 1 bonninites spatially coexists with ophiolites, although compose (or belong to) other tectonic units.Type 2 boninites presents as later constituents of ophiolite sequences, such as crosscutting dikes or lavas on top of section.Type 3 includes island-arc tholeiites and basaltic andesites coupled with boninites, which are replaced by younger MORB or BABB affinities.Type 4 boninites occupies the whole mafic portion of ophiolite sequences, together with island-arc tholeiites and basaltic andesites, and all the components of such sequences (gabbro, dykes, and lavas) have clearly boninitic affinities.
These types of the boninite-bearing ophiolite sequences are recorded through the entire Phanerozoic and Neoproterozoic.The Neoproterozoic boninite Fig. 1.A comparison between normative compositional trends of the boninite and komatiite volcanic series in ternary projection olivine-plagioclase-quartz from diopside [Walker et al., 1979].The both volcanic series are from the Archean Karelian Craton.Modified after [Shchipansky, 2008].

Fig. 2. Trace elements patterns of different boninite groups from Izu-Bonin arc by comparison with N-MORB.
Note the negative anomalies of Nb (Ta) and, conversely, positive anomalies of Zr (Hf) and Sr, both typical of supra-subduction zone volcanics.Strong depletion in REE and HFSE of the boninite compositions with respect to N-MORB is clearly visible.Averaged boninite compositions are from [Pearce et al., 1992].N-MORB and primitive mantle values are from [Hofmann, 1988].
series are widely distributed in the Altai-Sayan area of the southern frame of the Siberian Craton [Simonov et al., 1994;Dobretsov et al., 2005].More ancient boninite series are discovered in the Archean and Paleoproterozoic metamorphosed fold belts that are often jointly termed as 'greenstone belts', regardless of the metamorphic grade.
As a rule, the rock assemblages composing the greenstone belts were strongly tectonically transformed and dismembered that gives a zero chance for preservation of a complete ophiolite sequence.Nonetheless, the Early Precambrian greenstone belts maintain vestiges indicating that rock assemblages of oceanic provenance were involved in the tectogenesis of the belts [Shchipansky, 2008;Rosen et al., 2008;Furnes et al., 2015].These are mainly the isotopic and geochemical signatures of the lack of crustal contamination or affiliation of the mafic-ultramafic assemblages of the greenstone belts to the ensimatic island-arc settings.In the latter case, the presence of the boninite series rocks is a critical support to the suprasubduction ophiolite interpretation of the greenstone belts sequences.
Besides the isotope-geochemical data, some boninite series from the Archean greenstone belts do preserve field evidences on the ocean lithosphere extension.We reported a fragment of the suprasubduction ophiolites with sheeted dikes in gabbroids and metalavas of the boninite series (~2.8 billion years) in the Iringora locality of the North-Karelian greenstone belt [Shchipansky et al., 2001[Shchipansky et al., , 2004]].Later on, relics of a sheeted-dike complex were discovered in the Garbenchifer formation from the Eoarchean Isua supracrustal belt, southwestern Greenland [Furnes et al., 2007[Furnes et al., , 2009]].This suggests that the boninite series rocks and the SSZ spreading have been genetically related since the early stages of the Earth's geological history.
The early Precambrian boninite series differ from the Phanerozoic ones by the presence of both boninites and komatiites in some greenstone belts, such as the Bogoin (Paleoproterozoic), Abitibi (Neoarchean), and Koolyanobbing (Mesoarchean) belts, which is related to mantle plume fingerprinting into intra-oceanic convergence zones (see Table 2).It should be noted, however, that data on the Phanerozoic boninite series often Fig. 3. Spidergram for trace-element compositions of the continental crust and juvenile arc melts (boninite, arc tholeiite), modified after [Stern, 2002].
Elements on the horizontal axis are listed in order of their incompatibility in the mantle relative to melt; elements on the left are strongly partitioned into the melt whereas those on the right are strongly partitioned into a peridotite source.Note the characteristic enrichments of subduction zone outputs relative to MORB with respect to fluid-mobile LIL elements and the relative depletion of these in HFSE, and HREE.Note also the overall similarity of continental crust to juvenile arc crust.

PETROGENESIS OF THE BONINITE SERIES ROCKS, AND THEIR EVOLUTION IN TIME
The boninite series rocks are unique because of their genesis requires the set of conditions which is possible only in specific geodynamic settings within spatially limited locations.Indeed, the origin of a boninite source requires prior depletion of an upper mantle reservoir by basalt melt extraction in one or several episodes, which means that mantle harzburgite was such a source [Hickey, Frey, 1982;Sun, Nesbitt, 1978;Duncan, Green, 1980, 1987].Boninitic melts are primitive and, at the same time, highly silicic with very low absolute concentrations of incompatible trace elements (Nb, Ta, and Ti) and rare earth elements; this suggests melting of residue peridotite at relatively shallow depths.
In fact, contents of Cr and Ni in the boninite series rocks are high.The strong and fast depletion in Cr suggests that chromite was among the main liquidus phases at the early stages of the primary melt fractionation.Depletion in Ni and simultaneously increasing contents of SiO 2 indicate that olivine played an important role in magmas fractionation to Mg#~0.70.As noted above, the subsequent crystallization phase was clinopyroxene replaced then by plagioclase.
The boninite series lavas are considerably enriched in LILE and LREE in comparison to HFSE.Such geochemical characteristics suggest active infiltration of hydrous fluid into the mantle source of the boninite melts [Pearce et al., 1984[Pearce et al., , 1992;;Pearce, 1982Pearce, , 2003;;Crawford et al., 1989;Bloomer et al., 1995;Stern et al., 1991].The presence of fluid water in the mantle wedge is also a prerequisite to lower the solidus melting temperature of a refractory source.
While the ideas concerning the source of the boninite series melts are commonly consistent, there is a disagreement in estimated P-T conditions for genera-Fig. 5. Models illustrating four tectonic types of boninite occurrences in ophiolitic sections, after [Sklyarov et al., 2016].
It has been experimentally determined [Kushiro, 1990] that adding 4.4-6.6 wt % H 2 O into the mantle substance reduces the melting start pressure by almost 2 kbar and lowers the temperature by ~150 °C.Experiments with refractory harzburgite under anhydrous and H 2 O-undersaturated conditions [Falloon, Danyushevsky, 2000] show that the high-Ca boninite petrogenesis requires temperatures as high as ~1480 °C and a pressure of about 1.5-2 GPa in the presence of 2-3 wt % H 2 O.
Notwithstanding significant uncertainties in the boninite series melt models (which compositions are dependent on many factors, including the mantle depletion degree and the fluid melting mode), there is a general consensus that their genesis requires anomalously high temperatures and the presence of a water-containing fluid in considerable quantities.In this case, temperature anomalies mean temperature values considerably exceeding those in the ambient upper mantle, which are sufficient for MORB generation.In the recently introduced theory on decompression melting of the upper mantle, a potential mantle temperature (Tp), i.e. a temperature at the lowest point of the mantle column, is set at ~1350 °С for generation of the parental melt for N-MORB [Langmuir et al., 1992;O'Hara, Herzberg, 2002].This melt corresponds to a picrite composition (13 wt % MgO), and MORB13 composition model is perfectly matched to the fractionation trend of tholeiitic melts forming N-MORB [Niu, O'Hara, 2009].Figure 6 illustrates the idea that the generation conditions of the boninite series rocks and MORB are different.Comparison of the MORB and boninite compositions with similar Mg -number gives grounds to challenge the geodynamic model suggesting that boninites could form in zones of stretching above the mid-ocean ridges in supra-subduction environments (see also [Metcalf, Shervais, 2008]).What petrogenetic conditions may be responsible for the boninite series rock genesis?Did any evolutionary changes occur in such conditions through the Earth's history?Answers to the above questions can be derived from the theory of upper mantle decompression melting and its key concept of accumulated fractional melting.If the latter is the case, the mantle source begins to melt in small amounts during decompression; typically, melt is extracted after 1-2 % melting by buoyancy-driven porous flow; and the residue continues to melt in small increments during decompression.This process takes place repeatedly as decompression progresses from an initial high to a final low melting pressure.The small melt fractions mix in transitional crustal magma chambers to make an aggregate primary magma melt.Each melt droplet contributing to the aggregate is in equilibrium with a specific source which composition varies from the initial relatively fertile to the depleted final composition.An aggregate fractional melt is not in equilibrium with its residue; only the final drop of the melt extracted is in equilibrium with the residue [Herzberg, Rudnick, 2012].In solutions of the mass balance equation, mean values of residue, pressure and melt amount are used for simplicity.This facilitates solving Fig. 6.A comparison between trace element compositions of the parental melt for modern MORB expressed as MORB13 [Niu, O'Hara, 2009] and the recent boninite with ~13 wt.% MgO from the Tonga fore-arc [Falloon et al., 2008].
It is clear that a petrogenesis of boninite series seems to be unrelated with an extensional environment characteristic of a MORBforming magmatic column.Primitive mantle values are from [Hofmann, 1988].
general problems of decompression melting with regard to the mantle columns formed in differing, i.e. plume or mid-ocean ridge, settings.Using the data on the most primitive compositions, that are mainly controlled by olivine fractionation, it becomes possible to calculate parental melt compositions and potential mantle temperatures corresponding to the melting initiation depth of the given mantle column [Herzberg et al., 2007;Herzberg, Asimow, 2008].The diagram is simplified by showing a dry solidus applicable to both MORB and the boninite series, because a hydrous melting will take place below the dry solidus and the boninite series have a more depleted mantle solidus than MORBs.Tp and Tliq for boninite series rocks have been calculated from the appropriated compositions using software PRIMELT2.XLS [Herzberg, Asimow, 2008].Data source: Troodos [Pearce, Robinson, 2010]; Tonga [Falloon et al., 2008]; Abitibi [Kerrich et al., 1998]; North Karelian greenstone belt [Shchipansky et al., 2004].Data for the Kostomuksha komatiites were taken from [Puchtel et al., 1998].
presence of hydrous fluid.So, this phenomenon may be thought of as a result of the secular cooling of the Earth.The gradient of 50-70 °C/Ga suggests evolutionary changes in geodynamic processes in the Earth's history, rather than any significant transformations in tectonic mechanisms responsible for the continental crust growth.A nearly similar difference in the potential temperatures for the plume magmatism in the Earth's history is also noteworthy [Herzberg et al., 2007].
The evolution of boninitic volcanism in the Earth's history is clearly revealed in Figure 8 with regard to the definition of potential mantle temperature, envisaging a relationship between a mantle melting degree and a depth whereat the magma column is initiated due to decompression melting.The petrogenetic grid is used to show the calculated primary melt compositions of the Late Archean (North-Karelian and Abitibi belts), Paleoproterozoic (Flin Flon belt), Neoproterozoic (Jiangnan belt) and Meso-Cenozoic (Troodos, Tonga) boninite series, and a few fractionation trends are given for illustration of differentiation paths.It is clear that the early Precambrian boninite series were generated at higher degrees of mantle harzburgite melting (30-40 %), and the mantle melting columns occurred at considerably larger depths (3.5-4.0GPa) than during the Phanerozoic eon (2.5-3.0GPa).The knowledge of the P-T differences of the boninite series petrogenesis is important for understanding the evolution of geodynamic processes in the Earth's history that will be discussed below.

GEODYNAMIC SETTINGS OF BONINITE OCCURRENCE
Almost all the boninitic magmatism occurrences, regardless of its age, are genetically associated with intra-oceanic subduction zones (see Table 2).The modern boninitic volcanism occurrences revealed by the recent studies are localized only in the intra-oceanic plate convergence zones, and there is no proven example that would evidence any other geodynamic settings for their generation.Considerations of petrogenesis of Fig. 8. Change in parental melts of boninite series compositions in time on the petrogenic grid MgO vs. FeO [Herzberg, Asimow, 2008].
the boninite series rock melts also imply their emplacement above the zones wherein the oceanic lithosphere slabs are dehydrating and sinking into the mantle.It is established that in the modern intraoceanic island arcs (Izu-Bonin-Mariana, and Tonga-Kermadec), the boninite series volcanic rocks build up the peridotite-gabbro sequences of the forearc slopes and indicate the initiation and growth of juvenile island-arcs [Pearce et al., 1992;Stern, Bloomer, 1992;Stern, 2004].
It should be noted that a genetic link between the boninite series and ophiolites was recognized well before the deep-sea drilling and dredging studies of the forearc slopes of the Tonga and Izu-Bonin-Mariana arcs [Miyashiro, 1973;Cameron et al., 1979].Numerous discoveries of the boninite series rocks [Pearce, 2003[Pearce, , 2008;;Stern, 2004;Metcalf, Shervais, 2008;Sklyarov et al., 2016] have given grounds to conclude that the majority of the world's ophiolites mark paleo-zones of spreading in suprasubduction environments at the ancient oceanic plate margins, rather than past spreading at ancient mid-ocean ridges.Thus, boninites are petrologically unique rocks providing a perfect indication of paleogeodynamic settings in the oceanic plate convergence zones through the Earth's history.
The new paradigm of geodynamic settings for the boninite series generation was based on the hypothesis that SSZ ophiolites are related to the initiation of intraoceanic island arcs [Stern et al., 1991;Stern, Bloomer, 1992].This raised the questions of how and where subduction zones were initiated [Stern, 2002[Stern, , 2004]].In terms of mechanics, a prerequisite for subduction initiation is gravitational instability in the oceanic lithosphere, which can lead to its tearing followed by decompression melting of the upper mantle and sinking of one plate under another.This phenomenon, as well as the genetic link between boninites and ophiolites, is encapsulated as the subduction initiation rule (SIR) [Whattam, Stern, 2011].
In the model proposed in [Stern, 2004], this condition is satisfied when plates of differing temperature and density are juxtaposed across a transform fault or fracture zone, i.e. theoretically, in case of the interaction between plates of different ages, i.e. an old cold plate and a young hot plate.Along the fault, the gravitationally instability of the ancient crust makes it sink with a down-dip component of motion, while a lateral component is lacking.The model suggest that upon initiation of sinking of the slab/plate, the overlying slab/plate is subject to stretching, the partially depleted mantle is rising to melt due to its decompression, and a large input of water from the sinking slab sets up conditions for extremely high extents of fusion.Later on, the slab motion vector acquires a lateral component, which stops the decompression, and, consequently, terminates melting of the depleted mantle.Thus, a true subduction regime is established for generation of the normal island-arc tholeiitic and calc-alkaline volcanic series.
Another subduction initiation concept is proposed in [Niu et al., 2003].It establishes a link between locations of subduction zones and density inhomogeneities at the borders of the normal oceanic lithosphere and the oceanic lithosphere thickened by mantle plume impingement, i.e. oceanic plateaus or hotspot tracersaseismic ridges or seamounts.This concept is supported by the data on the Tonga boninites that are confined to the intersection of the aseismic ridge of the Louisville hotspot and the paleotrough of the Tonga-Kermadec arc [Turner, Hawkesworth, 1997].There is also evidence that the initial stage of the Izu-Bonin-Mariana arc development was associated with mantle plume fingerprinting at the Manus back-arc basin [Macpherson, Hall, 2001].
The boninite series volcanic rocks often have enriched mantle-plume isotopic and geochemical signatures [Sobolev, Danyushevsky, 1994;Taylor, Nesbitt, 1998].The involvement of the mantle-plume thickened lithosphere in the petrogenesis of boninites would thus resolve the long debate on anomalously high temperatures for the initiation of partial melting of the depleted refractory mantle harzburgite as high heat inputs can be ensured by the ascending hot mantle plumes [Falloon, Danyushevsky, 2000].
The presence of mantle-plume products is also revealed in the SSZ ophiolite sequences, which isotopic and geochemical compositions are well studied, such as the Pindos, Josephine, Koch, Magnitorsk (Southern Urals, Russia) ophiolites, etc. (see Table 2).In the early Precambrian sequences, the boninite series volcanic rocks are also associated with mantle-plume derivatives, komatiites, as well as OIB-type metavolcanic rocks [Shchipansky, 2008].
Apparently, the occurrence of mantle-plume derivatives in the intra-oceanic subduction initiation zones does not seem to be random.It is recognized that a rising mantle plume can decrease the strength of the lithosphere, which may lead to the breakup of the continents [Courtillot et al., 1999].In addition, an emplacement of mantle plume head at the lithosphere can significantly change its density characteristics.The ingress of melts generated by the enriched deep source into the upper layers of the mantle and the oceanic lithosphere would lead to refertilization of the previously depleted mantle.While cooling, the upper mantle transformed by mantle plume impingement becomes denser, and a new lithospheric segment may gradually become negatively buoyant.The reason is that OIB volcanic rocks are considerably enriched in Fe and Ti.Besides, Fe-Ti basalts/gabbros are known to being eclogitizated faster than their magnesium equivalents.Figure 9 shows the experimentally determined garnet stability fields for basalt compositions differing in magnesium numbers.Clearly, ferriferous compositions show the earliest occurrence of garnet in quantities considerably exceeding those in case of olivine tholeiites and, especially, high-Mg basalts.This metamorphic transformation at the crustal base of the thickened oceanic island builds seems to be a critical factor disturbing the gravitational stability of the lithosphere.Garnet is denser by almost 15 % than pyroxenes, amphiboles and olivine, which means that 'density eclogitization' can take place in the reworked lower lithosphere.In this case, even if only the lower crust is foundered, tearing of the lithosphere seems to be inevitable .Actually, potential gravitational instability of the lithosphere seems to be directly dependent on a volume of the mantle-plume productivity and the thermal status of reworked lithosphere segment.It is not common that such factors are favorably combined; anyway, lithosphere breakup and initiation of new subduction zones is highly probable.It is noteworthy that the polarity of subduction is predetermined by the thermal status of the lithosphere -hot lithosphere is positively buoyant, while cold lithosphere is negatively buoyant.
The model in Figure 10 shows the global-scale chemical and density inhomogeneities of the lithosphere as a result of a superplume event.The term 'superplume' was coined by Roger Larson in the early 1990's, when the author has advocated that mantle-plume volcanism was the strongest in the mid-Cretaceous time (124-83 Ma) [Larson, 1991].This event resulted in the emergence of giant oceanic plateaus, such as the Ontong Java, Manihiki, Kerguelen, and Columbia-Caribbean.It is believed that four currently active hotspots in the Pacific Ocean, which are traceable in the oceanic seafloor, are associated with this superplume.Few other tracks seem to have been cut off in the subduction zones.Other hotspot tracks in the Pacific are younger (Paleogene-Neogene) and may be not associated with deep plumes [Clouard, Bonneville, 2001].Accordingly the numerical simulation of a termochemical superplume at the core-mantle boundary, protuberance emissions of the deep matter to the surface are represented by hot jets or large igneous provinces which impacts are observed even at the upper mantle [Farnetani, Hofmann, 2011].This phenomenon may be responsible for the occurrence of the first-order chemical and density inhomogeneities and the initiation of the most extended plate convergence zones.This scenario seems to be realistic for the initiation of the Izu-Bonin-Mariana arc in the early Eocene.
Another subduction initiation scenario assumes a closer link between subduction and the geodynamics of individual hotspots.As already mentioned, the youngest boninitic magmatism occurrences in the northern Tonga are spatially associated with the Louisville hotspot track.Its forearc structure includes the Vitiaz paleo-trench (age of ~4 Ma) and the modern trench located further in the ocean at a distance of about 500 km.A unique double-subduction picture is revealed in this region by detailed seismotomographic surveys.A fragment of the younger and more gently dipping slab of the oceanic lithosphere is clearly visible above the current subduction zone [Chen, Brudzinski, 2001].The shallow-depth slab detachment was discussed in detail in [Shchipansky, 2008] as the factor disturbing a lithospheric barrier and creating extremely favorable conditions for generation of the ophiolitic boninite series (Fig. 11).
It is quite possible that the shallow-depth slab break-off resulted from the compositional density inhomogeneity of the sinking lithosphere, such as local charging by OIB magmatism products.The key geodynamic implications of this phenomenon are, first, strong short-term thermal disturbance over the narrow slab window, and, second, fast uplift of its over-Fig.9. Experimentally determined Grt-in curves for compositionally different basalts: Fe-rich basalt (Mg#=41), olivine tholeiite (Mg#=55), and high-Mg basalt (Mg#=69).Note that relative abundances of garnet shown at the inset vary significantly also.Modified after data of [Molina, Poli, 2000].
riding plate [van de Zedde, Wortel, 2001;Buiter et al., 2002].This mechanism provides a reasonable explanation for the short-term (3-5 Ma) volcanism which was by far more voluminous that the steady-state subduction volcanism [Stern, 2002[Stern, , 2004]].Due to uplifting of the suprasubduction plate, an ophiolite 'platform' replaced the hanging plate and provided a basement for the nascent island-arc build.It is essential that the uplift of the ophiolite 'platform' could have attained the morphological stability as the previous episode of Fig. 10.Sketch illustrating an emergence of large refertilizited, or reworked, mantle (RM) region due to a superplume event.
This gives rise to a creation of lateral compositional buoyancy contrast over normal oceanic lithosphere (DM) and, thus, to a subduction initiation with time.Superplume cartoon is after the numerical simulation of a thermo-chemical plume [Farnetani, Hofmann, 2011].
Note that in this case different portions of the upper mantle, i.e. lherzolitic asthenospheric and harzburgitic lithospheric, should be partially molten.Crustal part of the detached slab provides a source for aqueous fluids enriched by subduction-related chemical components.Loci of low pressure and high pressure melting of metabasic oceanic rocks are shown also.
intense mantle melting leaved the mantle strongly depleted.This resulted in the emergence of a lithospheric keel capable to resist the convective instability of the surrounding mantle.
Figure 12 shows the above-described model based on the modern geodynamics of the northern termina-tion of the Tonga arc.The top panel shows the subduction initiation and generation of boninites in the forearc region.The bottom panel shows the modern subduction zone and paleoslab fragments under the emerging Lau basin.In the samples dredged from the Lau basin, the presence of the subduction fluid compo- The model is largely based on the available data from the modern Tonga arc [Sobolev, Danyushevsky, 1994;Turner, Hawkesworth, 1997;Chen, Brudzinski, 2001;Niu et al., 2003;Falloon et al., 2007Falloon et al., , 2008;;Resing et al., 2011]: a -lithosphere charged with Fe-rich melts gets negative buoyancy as it cooled, and starts to sink.Lithosphere tearing leads to raise asthenospheric mantle and triggers intensive melting that includes boninite series melts (see fig. 10); b -as a result of the shallow level slab break-off, a new subduction should develop and, as a consequence, a new trench occurs at a distance of a few hundred kilometers oceanward from the older trench.Detached slab fragments lead to a lithosphere extension, thus forming a tight back-arc trough characterized by occurrences of local multi-spreading centers where younger boninites can also be formed.See the text for more details.
nent is evidenced by higher contents of H 2 O in glasses and enrichment in large-ion lithophile (LIL) elements [Falloon et al., 1992;Danushevsky et al., 1993;Pearce et al., 1994].The fact that compositions of these basalts significantly differ can be explained by contribution to their petrogenesis of various mantle sources, including mantle plume input.Besides, active eruptions of boninites take place at present time [Resing et al., 2011].Thus, boninites occur in a variety of combinations in different tectonic settings in time and space, i.e. in the fore-and back-arc environments.This indicates that, firstly, it is an established phenomenon, and, secondly, the subduction initiation process may be not as simple as described by the 'subduction initiation rule' (SIR) model [Whattam, Stern, 2011].
It seems likely that the subduction regime should be stabilized after a certain period of accommodation associated with the slab break-off event/ or events and resultant contrasting tectonic regimes on the surface.During this period, movements caused by the frontal subduction compression are rapidly replaced by strikeslip motion, which leads to the occurrence of a complex rift system with both hot spot volcanism and volcanic seamounts [Falloon et al., 2007[Falloon et al., , 2008]].
Considering the ophiolite sequences worldwide, it has been noted that, although their magmatic chemostratigraphic progression is almost similar [Pearce, Robinson, 2010;Whattam, Stern, 2011], the relationship between boninite series and other members of ophiolite sections are variable and may significantly differ from the ideal ophiolite sequence, known as a 'penrose ophiolite' [Sklyarov et al., 2016].In the literature, there are numerous cases showing the lateral variability of the ophiolite sequences within the local regions and therefore suggests non-stationary settings of ophiolite-forming processes.These increasingly compel the conclusion that ophiolites are commonly associated with the subduction initiation environ-ments, rather than with spreading in the mid-ocean ridges.

SUBDUCTION INITIATION IN THE EARLY PRECAMBRIAN
As described above, generation of the early Precambrian boninite series was developed by mantle harzburgite melting of high degrees (30-40 %), and the mantle melt columns started at greater depth (3.5-4.0GPa) as comparing with those from Phanerozoic eon (2.5-3 GPa).What are the geodynamic implications of these facts?
It is well known that the early Precambrian cratons are underlain by the depleted subcontinental lithospheric mantle (SCLM) or a lithospheric keel (root) extending into the diamond stability field.This phenomenon has never reoccurred in the later history of the Earth, and its origin has been widely discussed in the literature [e.g.Herzberg, Rudnick, 2012;Shchipansky, 2012, and references therein].In the Phanerozoic, strong depletion of the upper mantle is associated with generation of the boninite series in the subduction initiation zones.The above overview of the global boninite occurrences gives abundant evidence of boninitic magmatism in the early Precambrian, and the discoveries of ancient boninites are progressively growing.Accordingly to the recently published assessment, the volumes of the Archean boninite volcanics and komatiites appear to be almost roughly similar (Fig. 13) [Furnes et al., 2014].Discovery of the boninite series rocks with fragments of sheeted dikes and IATtype metabasites in the oldest preserved complex Isua strongly suggests that subduction dates back from the early Earth's geological history.
A prerequisite for subduction initiation is the oceanic lithosphere breakup through its entire thickness.It follows thus from Figure 6 that, first, the early Precam-Fig.13.Percentage lithology of komatiite and boninitic volcanism represented at given age intervals, relative to total of mafic and ultramafic magmatism in the Earth history, according to [Furnes et al., 2014].
brian lithosphere as a unit should have rheological properties providing for brittle or brittle-plastic deformations.In other words, such lithosphere can be considered as a rigid body capable of resisting the convective instability that is an attribute of plate tectonics [Sleep, 1992].The Archaean oceanic lithosphere thickness is estimated at 85-120 km, whereas the modern lithosphere is ~60 km thick [Herzberg et al., 2010].
Second, unlike the Phanerozoic boninite series, parental melts of the early Precambrian series were generated at depths of ~120-130 km, i.e. in the diamond stability field.Given the fact that the primitive melts of the ancient boninite series clearly have the subduction influence signatures, it is reasonable to believe that deep sinking of the slabs into the early Precambrian mantle was highly possible.
Third, extensive melting of the early Precambrian upper mantle for the period of subduction initiation implies that it was subject to strong depletion through the diamond stability field.During the accommodation stage, the new sinking slab/slabs are cut off the astenospheric heat input thus leading to quickly cooling of overriding forearc lithosphere.This model provides an explanation of the origin and stability of the cold diamond-bearing SCLM that has not been subject to any convective perturbations, at least, since 3.0 billion years [Boyd et al., 1985].
The differences between the modern and early Precambrian subduction initiation settings are schematically shown in Figure 14.The main distinction is that thin-plate tectonics covers the period from the Neopro-terozoic to the present time, and thick-plate tectonics refers to the early Precambrian.Today, the oceanic crust is ~6-7 km thick on average, and this requires ~6-7 % melting.In general, about 1 km of oceanic crust is produced for every 1 % of partial melting (i.e. 1 km/1 % melting) [Herzberg, Rudnick, 2012].According to available petrological estimations, the Archean MORB-type oceanic crust thickness ranges from ~20-25 km [Abbott et al., 1994] to 40-60 km [Herzberg et al., 2010].In other words, the Archaean oceanic crust should have been, roughly, 3 to 6 times thicker than the modern crust.Such estimations scatter is due to the problem of validity of mantle potential temperature (Tp) computations for the ambient Archaean upper mantle as preserved off-arcs mafic volcanics of that age are very seldom [Pearce, 2008;and others].
Anyway, thickening of the Archean oceanic crust is a prerequisite for generation of tonalite-trondhjemitegranodiorite (TTG) gneisses composing the bulk Archean continental crust.Both tonalite granitoids and their high-pressure analogues (adakites) are observed in modern suprasubduction settings, although in much smaller amounts than the basalt-andesite-rhyolite series.

CONCLUSION
The boninite series and its most fractionated endmember, boninites, are not just a simple unit of the suprasubduction sequence, but also a vivid indicator of initiation of intra-oceanic convergence zones.Being directly related to ophiolites, boninites and low-Ti basalts/picrites give direct evidence that the oceanic lithosphere stretching was related to subduction initiation.This means that the age of SSZ ophiolites or, more precisely, SIR ophiolites [Whattam, Stern, 2011] may show the age of paleo-oceans only in the first approximation.The oceanic lithosphere itself may be much older.Indeed, the modern boninites of the northern Tonga arc, the Eocene IBM boninites and ophiolites of Papua New Guinea cannot evidently correspond to timing of the Pacific or Indo-Austarlian plate life.Moreover, the Phanerozoic folded belts show multi-suturation signatures rather than mono-suture architecture.
The records of MOR-type ophiolites are very seldom as comparing with the large occurrences of the SIR ophiolites including the well-known sequences of Troodos, Semail, or Newfoundland [Shervais, 2001;and others].The completely developed MOR-type includes ophiolites of 12-9 Ma age in the Macquarie Island located at the boundary between the Pacific and Australian plates, and the Neoproterozoic ophiolites of Gabal Gerf, the Arab-Nubian shield [Pearce, 2008;Whattam, Stern, 2011, and references therein].A rarity of the MOR type in the ophiolite continuum stems well from the fact that physically it is almost impossible to emplace true MORB crust at a convergent plate boundary; rather sediments and fragments of seamounts may be scraped off of the sinking plate [Stern, 2004].
It is now recognized that SIR ophiolites are volumetrically major in the Earth's geological history (see Table 2 and [Furnes et al., 2014]).As shown above, a prerequisite for subduction initiation is a lateral compositional buoyancy contrast within the lithosphere which would led to its collapse and consequently to form zones of intra-oceanic plate convergence.An impingement of mantle plume into the pre-existing lithosphere appears to be the most obvious among the known mechanisms controlling the global geodynamics of the Earth.However, mantle plumes themselves do not seem to be responsible for continental crust growing, even in the case of Iceland plateau placed directly above the hot mantle plume head [Martin et al., 2008].However recently subducted oceanic plateaus crust revealed the capability to produce a juvenile continental crust similar to Archean TTG via its partial hydrous melting [Hastie et al., 2010].It is vital also, that highpressure adakites are also found in association with the Tonga boninites related to the hot spot track [Falloon et al., 2008].Furthermore, the mere fact that boninites tend to be compositionally close to the bulk composition of continental crust suggesting that the crustal growth processes began with the subduction initiation (see Fig. 3).
Thus, lacking direct influence on crustal growing at the intra-oceanic island-arc zones, mantle plume events could have acted as 'remote triggers' of theses process and predetermined subsequent loci of subduction initiation zones.Such a kind of the self-organized geodynamic system of the Earth may have originated as early as Eoarchean, and its further evolution was mainly operated by the secular cooling of the planet.
extensional zone in a forearc setting possibly related to the MOR centers subduction

Figure 7
Figure7illustrates relationships between potential and liquidus temperatures of the most primitive compositions of the modern (Tonga forearc and Troodos and upper pillow lavas) and Archean (Abitibi and North Karelian greenstone belts) boninite series.The potential mantle temperatures for their generation differ by 150-200 °C.This difference refers to the assemblages generated by mantle sources of an almost similar type (mantle harzburgite) due to its melting in the

Fig. 12 .
Fig. 12.An illustration of the model of boninite formation when a reworked lithosphere caped by OIB volcanics initiates a non-stationary subduction.