Continental versus oceanic subduction zones

Subduction zones are tectonic expressions of convergent plate margins, where crustal rocks descend into and interact with the overlying mantle wedge. They are the geodynamic system that produces mafic arc volcanics above oceanic subduction zones but high- to ultrahigh-pressure metamorphic rocks in continental subduction zones. While the metamorphic rocks provide petrological records of orogenic processes when descending crustal rocks undergo dehydration and anataxis at forearc to subarc depths beneath the mantle wedge, the arc volcanics provide geochemical records of the mass transfer from the subducting slab to the mantle wedge in this period though the mantle wedge becomes partially melted at a later time. Whereas the mantle wedge overlying the subducting oceanic slab is of asthenospheric origin, that overlying the descending continental slab is of lithospheric origin, being ancient beneath cratons but juvenile beneath marginal arcs. In either case, the mantle wedge base is cooled down during the slab–wedge coupled subduction. Metamorphic dehydration is prominent during subduction of crustal rocks, giving rise to aqueous solutions that are enriched in fluid-mobile incompatible elements. Once the subducting slab is decoupled from the mantle wedge, the slab–mantle interface is heated by lateral incursion of the asthenospheric mantle to allow dehydration melting of rocks in the descending slab surface and the metasomatized mantle wedge base, respectively. Therefore, the tectonic regime of subduction zones changes in both time and space with respect to their structures, inputs, processes and products. Ophiolites record the tectonic conversion from seafloor spreading to oceanic subduction beneath continental margin, whereas ultrahigh-temperature metamorphic events mark the tectonic conversion from compression to extension in orogens.


INTRODUCTION
Subduction zones, where Earth's lithospheric plates sink into the mantle, are responsible for volcanic eruptions, earthquakes and mineral resources. Subduction delivers crustal materials to the mantle, playing a primary role in terrestrial element cycling. Although subduction zones are the dominant physical and chemical system of Earth's interior, their depth extends from the surface to deeper mantle. While trenchs, arc volcanics and blueschistto eclogite-facies metamorphic rocks are the surface expression of subduction zones, orogens composed of deformed, metamorphic and magmatic rocks are the products of subduction-zone processes. The descending of lithosphere in subduction zones provides an effective mechanism for the exchange of mass and energy between the crust and mantle. This is realized by delivery of raw materials to the subduction factory, where material derived from oceanic and continental lithospheres is interacted with the asthenospheric mantle [1][2][3]. Metamorphic dehydration and partial melting are particularly prominent during the subduction of crustal rocks to forearc and subarc depths [4,5]. These processes are recorded by high-pressure (HP, 0.6-2.8 GPa) blueschist-facies and HP to ultrahigh-pressure (UHP, >2. 6-2.8 GPa) eclogite-facies metamorphic rocks in collisional orogens [6][7][8].
Collisional orogens are produced by subduction of a continental block beneath another continental block or beneath a marginal arc terrane, resulting in mountain ranges such as the Alps and Himalayas. On the other hand, mafic arc volcanics are common in accretionary orogens, where an oceanic lithosphere is subducted beneath either another oceanic lithosphere or a continental lithosphere. However, only HP metamorphic rocks have been found in oceanic subduction zones [9], indicating that descending crustal rocks there have returned only from forearc depths of <60-80 km. In contrast, deeper UHP metamorphic units in continental subduction zones are exhumed from subarc depths of 80-160 km, though some of them are also sunken into the asthenospheric mantle. Therefore, the mafic arc volcanics above oceanic subduction zones only can be used for indirect investigation of the mass transfer from the subducting slab to the mantle wedge at the subarc depths [10], whereas the petrology and geochemistry of UHP metamorphic rocks in collisional orogens provide more direct clues of this process [3,11].
A first-order classification of continental subduction zones relies on the nature of overriding lithosphere. The first category involves the subduction of one less ancient continent beneath another more ancient continent (Fig. 1a), which is exemplified by the Dabie-Sulu orogenic belt in east-central China [12,13]. The second category is the subduction of one ancient continent beneath the juvenile arc terrane of continental margin (Fig. 1b), which is typical in the Himalayan orogen of southern Asia [14,15].
In both cases, subducting crustal slices were detached at different depths and then exhumed along subduction channels, giving rise to different types of metamorphic zones in collisional orogens [16,17]. It is these subduction-zone processes that result in a series of physical, chemical and geological changes, such as HP to UHP metamorphism, dehydration and hydration, partial melting and melt metasomatism, magmatism as well as the weakening, detachment and exhumation of crustal slices in the subduction channels [4,5,16].
This paper aims to highlight the most important aspects of continental versus oceanic subduction zones, such as their structures, inputs, processes and products. In particular, the geological and thermal structures of subduction zones are linked to the petrological and geochemical records of orogenic rocks. Such efforts result in an integrated interpretation of those aspects with respect to the tectonic evolution of subduction zones in both time and space. For convenience, subduction-zone fluids may broadly refer to aqueous solutions, hydrous silicate melts or supercritical fluids [18].

GEOLOGICAL ARCHITECTURE
Continental lithosphere is composed of continental crust and the uppermost mantle that are rigid relative to the underlying asthenospheric mantle. The continental crust is composed of a crystalline basement and a sedimentary cover. The crystalline basement is mainly composed of the upper crust of felsic granite and the lower crust of mafic granulite, with amphibolite-facies metamorphic rocks in between. Although the continental lithosphere shows a range of variations in structure and composition, its complexity does not exert a prominent control on the fundamental behavior of continental subduction zones. In general, mantle lithosphere controls the physics of subduction, supracrustal rocks control the chemistry of subduction and the intracrustal rocks play an intermediate role in both physics and chemistry.

Subducting slab
The thickness of continental crust only varies from 30 to 75 km, whereas the variation in the thickness of subcontinental lithospheric mantle (SCLM) can be as significant as from 50 to 300 km [19]. The difference in lithospheric thickness between the overlying plate and the descending slab is also greater in continental subduction zones than in oceanic subduction zones. While a significant proportion of continental margins is subducted to mantle depths, its buoyant property resists subduction and eventually results in REVIEW Zheng and Chen 497 breakoff at the transition between continental and oceanic lithosphere. Continental subduction zones always have shallow dips, but its precedingly subducted oceanic slab may become steeper due to gravitational sinking or mantle flow. As soon as the subducted slab descends at a steeper angle than before, it will cause the trench to retreat and the slab to roll back. The slab rollback leaves the space for the lateral filling of asthenosphere beneath the mantle wedge, resulting in heating of the both the slab surface and the wedge base [5]. Because Fe, Al and Ca are preferentially partitioned into mafic melts during partial melting of mantle peridotite, their underlying restite changes in lithochemistry toward sterile harzburgite. Thus, the lithosphere is usually considered less dense than the underlying asthenosphere. This becomes prominent for the continental lithosphere when the thick felsic crust of low density is produced by partial melting of the high-density mafic crust. It is this deficient density that makes the continental lithosphere buoyant overlying the asthenosphere. The density deficiency increases as the lithosphere thickens. Generally speaking, oceanic lithosphere has a mean thickness of ∼100 km and its mantle lithosphere is primarily composed of less fertile lherzolite in the upper part, residue after extraction of mafic melts for mid-ocean ridge basalts (MORB) and gabbros, and more fertile lherzolite in the middle and lower parts, simply cooled from the asthenospheric mantle. Because of the melt-residue relationship at the crust-mantle transition zone, the oceanic crust is not susceptible to detachment from the underlying mantle lithosphere during subduction. In contrast, continental lithosphere has a mean thickness of ∼150 km and its mantle lithosphere is generally composed of three layers: (i) sterile harzburgite in the upper part, residue after extraction of MORB-type mafic melts at spreading ridges; (ii) fertile lherzolite in the middle part, simply cooled from the asthenospheric mantle; and (iii) sterile harzburgite in the lower part, residue after extraction of mafic melts for arc-type mafic melts. Therefore, the continental crust is susceptible to subduction detachment from the underlying mantle lithosphere because of the weak link in petrogenesis between them at the crust-mantle transition zone.
The property of descending crust exerts a fundamental control on the subduction-zone processes. Crustal density and thickness largely determine whether subduction continues or fails. Although normal oceanic crust is invariably subductable, descending of continental crust into the trench leads to subduction slowdown and eventually to failure. This tectonism is called continental collision or terrane accretion, by which two continental blocks con-tact with each other. Contrasts in lithospheric bulk density (crust plus mantle) determine whether subduction continues or collision occurs [20]. A failure of subduction takes place when sufficient buoyant material is introduced to subduction zones, and the buoyancy of continental lithosphere prevails over the dragging of the leading oceanic crust. The continental crust is also geochemically distinct from the oceanic crust, transporting a significantly different material into subduction zones. The continental crust is normally enriched in melt-mobile incompatible trace elements and their radiogenic isotopes [21]. In contrast, the oceanic crust is only enriched in these elements and their radiogenic isotopes in surface sediment [22,23], whereas the majority of oceanic crust is composed of the mafic igneous rocks that are depleted in melt-mobile incompatible trace elements and their radiogenic isotopes [24,25]. As crustal rocks descend into subduction channels, they are progressively heated and squeezed for deformation, metamorphism and anatexis, modifying their mineralogy and hydrous property with time and space. The occurrence of HP blueschist to eclogitefacies metamorphic rocks in orogens indicates that subduction zones are cooler than the surrounding mantle. Thus, it is critical to know how thermodynamic equilibrium is approached for cold materials that are slowly heated and squeezed through phase transformations but kinetically retarded during prograde subduction.

Mantle wedge
The mantle wedge is part of the mantle that lies above the subducting slab, where subducted inputs are chemically mixed with its base for metasomatism [19]. While the asthenospheric mantle is predominated above oceanic subduction zones, the SCLM is prominent above continental subduction zones [5]. In common models of the mantle composition, geochemically depleted MORB-type mantle is widespread in the upper mantle [26,27]; it is equivalent to the asthenospheric mantle prior to the effects of slab subduction [28]. As such, the asthenospheric mantle is fertile in lithochemistry and depleted in geochemistry (depletion in melt-mobile incompatible trace elements and their pertinent radiogenic isotopes). In contrast, the SCLM is heterogeneous in lithology and partially sterile due to extraction of mafic melts. On the other hand, the mantle wedge is ancient beneath cratonic block but it is juvenile beneath marginal arc terranes. Because of the age-dependent thickening, ancient lithosphere is more negatively buoyant than juvenile lithosphere with respect to the underlying asthenosphere. It follows that old lithosphere tends 498 Natl Sci Rev, 2016, Vol. 3, No. 4 REVIEW to subduct beneath young lithosphere if there is only the gravitational traction. Due to pulling of the preceding oceanic slab, nevertheless, the continental lithosphere may be subducted beneath the juvenile arc terrane.
Although the mantle wedge becomes partially sterile due to extraction of arc basaltic melts during the crust-mantle differentiation at convergent plate margins, it can be refertilized by interaction with subduction-zone fluids to generate maficultramafic metasomatites [5]. It is the fertile metasomatites that serve as the mantle sources of mafic arc-like magmas in syn-subduction, syn-exhumation or post-subduction stages [3,17,29,30]. The dynamics of this magmagenetic system is fundamentally different from that at divergent plate margins (e.g. mid-ocean ridges), where basaltic melts are produced by decompressional melting of the asthenospheric mantle. At convergent plate margins, the mantle wedge undergoes metasomatism by subducting slab-derived fluids at temperatures much lower than the wet solidus of peridotite [5]. Afterwards, the mantle wedge becomes partially melted when its temperature is elevated to meet or overstep the solidus of ultramafic metasomatites [31]. As such, it is not the fluxing effect of subduction-zone fluids, liberated from the subducting slab and rising to interact with the mantle wedge, that immediately triggers partial melting of the mantle wedge for arc magmatism. Nevertheless, it is this interaction between sinking slab, rising fluids and stagnant mantle wedge that defines the reprocessing system in the subduction factory.

THERMAL STRUCTURE
During subduction, the slab descends into the mantle and continuously exchange heat with the ambient mantle through advective and conductive processes [1][2][3]. The thermal structure of subduction zones may evolve with time [5]. This evolution is one of the dominant factors that control the physical and chemical properties of subducting slab and mantle wedge, and impact on many key geological processes such as crustal dehydration, crustal melting, fluid metasomatism, mantle melting, arc volcanism and seismic activity in subduction zones.
There are mainly two approaches of geodynamics to study the thermal structure of subduction zones: analytical models and numerical models. The analytical models (e.g. [32,33]) provide insights into the most dominant controlling physical parameters on the slab thermal structure, such as the slab age, convergent velocity, slab dip, shear stress and thermal conductivity. The numerical models (e.g. [34,35]) can further deal with more complicated environ-ments, such as viscosity change in the mantle wedge, coupling process between subducting slab and mantle wedge, and incorporation of petrology and mineralogy. Numerous geodynamic models have been developed (e.g. [32][33][34][35][36][37][38]). The results indicate that the thermal structure of subduction zones is influenced by a number of parameters [39][40][41][42], including: (i) subduction rate, (ii) slab age, (iii) slab dip, (iv) slab thickness, (v) slab width, (vi) the absolute motion of the overriding plate, (vii) the thickness of the overriding plate and (viii) the extent of the slabmantle coupling. It is noted that the mutual interaction between some of these parameters may exert a significant influence on the thermal structure of subduction zones.
Although most geodynamic models have been focusing on oceanic subduction zones [39][40][41][42], some of their results are also applicable to continental subduction zones (Fig. 2). This is because continental lithosphere is generally subducted along the cold geotherms of 5-10 • C/km, similarly to the majority of oceanic subduction zones in their early stage. As a consequence, kinetics would allow considerable overstepping of the equilibrium phase boundaries there due to sluggish reaction rates at lower temperatures. The difference is that continental lithosphere is thicker and colder than oceanic lithosphere [43]. With deepened understanding on continental subduction and collision, especially with the progress in the study of HP to UHP eclogite-facies metamorphic rocks [3,[6][7][8], it is becoming possible to directly construct and constrain the thermal structure of continental subduction zones. Nevertheless, the thermal evolution of continental subduction zones is more complicated than that of oceanic subduction zones because of the tectonic transition from oceanic to continental subduction.
According to the previous studies of geodynamic modeling, it appears that the convergent velocity exerts the first-order control on the thermal structure of subduction zones [39][40][41][42]. The slower the convergent rate, the higher the temperature at the slabmantle interface; the faster the convergent rate, the lower the temperature at the slab-mantle interface (Fig. 3). In addition, the thickness of the overriding plate also plays an essential role [44,45]. The thicker the overriding plate, the lower the temperature at the subducting slab surface; the thinner the overriding plate, the higher the temperature at the subducting slab surface (Fig. 3). Although slab age, slab dip, shear heating, phase change and fluid migration also affect the thermal structure of subduction zones, their effects are only concentrated in some specific areas and are generally much weaker than the effects of convergent velocity and plate thickness. Continental subduction zones are dominated by  The slab surface geotherms were calculated by different geodynamic models for oceanic subduction zones [33,35], but adjusted for continental subduction zones with differences mainly in convergent rate and the thickness of the overriding plate. Also plotted are lithostatic geotherms at 5 • C/km and 10 • C/km, respectively.
a cold regime, whereas oceanic subduction zones show a large range from cold to hot regimes [5]. As such, the thermal structure of continental subduction zones is affected by additional variables such as the size of subducting slab and the extent of slabmantle coupling. In other words, the kinematics of the subducting slab exerts the extra control on the thermal structure of subduction zones. It should be noticed that, when applying geodynamic modeling results to specific subduction zones, many factors in the realistic subduction environments may complicate the processes and cause deviations of the modeling results from field-based observations. For instance, there are considerable differences in both geotherm and composition at the lithosphere-asthenosphere boundary, but many of geodynamic models did not take this into account.
For this reason, many of the modeling results suggest significant increases in temperature for the slab surface at nearly isobaric conditions (e.g. [35,42]). Tectonically, however, the slab surface temperature would remain to be low if the subducting slab is coupled with the mantle wedge, whereas higher temperatures are caused by the slab-wedge decoupling [5]. As a consequence, the thermal structure of subduction zones often shifts from a cold regime in the early stage to warm and even hot regimes in the late stage. Therefore, one should be careful in applying the modeling results to extinct subduction zones that are intracontinental orogens at present. Nevertheless, with the development of new quantitative methods in geophysics and petrology, we have been obtaining more observational constraints on the thermal structure of subduction zones, providing better guidance for constructing the tectonic models of subduction zones.

OROGENIC PROCESSES
By building on the different aspects of subduction zones, an orogenesis refers to a series of processes that result in lithospheric tectonics composed of crustal thickening, deformation, metamorphism, magmatism, surface uplift and erosion. Generally, orogens can be categorized into two types [46][47][48]. One is the accretionary orogen due to the subduction of oceanic slab and the other is the collisional orogen due to the subduction of continental slab. The former is typified by circum-Pacific orogens (thus also referred to as the Pacific-type or oceanic-type orogens), whereas the latter is typified by Alpine-Himalayan orogens (thus also referred to as the Alpine-type or continental-type). The accretionary orogens are abundant of syn-subduction arc volcanics, ophiolites and HP blueschist-to eclogite-facies metamorphic rocks but short of UHP REVIEW eclogite-facies metamorphic rocks. In contrast, the collisional orogens often contain UHP eclogitefacies metamorphic rocks and syn-exhumation alkaline igneous rocks, but lack syn-subduction arc volcanics. In either case, orogenesis only refers to bulk processes from subduction to exhumation, distinctive from postorogenic processes that are independent of synorogenic tectonism. Nevertheless, the two types of orogens may be converted into intracontinental orogens away from active plate margins [17]. On the other hand, the accretionary orogens often develop into composite accretionary-type collisional orogens, leading to large-scale crustal thrusting, thickening and uplift. Although intracontinental orogens are generated at an early time by extinct orogenesis, they can be reactivated at a later time by the tectonic conversion from compression to extension due either to transmission of far-field stresses from plate margins or to founding of thickened orogenic roots in the postorogenic stage. The thermal regime of orogens may change from cold in the subduction stage through warm in the exhumation stage to hot in the postorogenic stage.
Orogenic processes involve a series of deformation, metamorphism and magmatism that take place in a period from slab subduction to crustal exhumation. They result in tectonic imbrication and crustal detachment at different depths [49,50], with the occurrence of ophiolites, HP blueschists and eclog-  LG, low-grade metamorphism; GS, greenschist facies; AM, amphibolite facies; GR, granulite facies; UHT, ultrahigh-temperature metamorphism, corresponding to the granulite facies at T > 900 • C. These facies are subdivided into Barrovian series in the warm and Buchan series in the hot regime. Highto ultrahigh-pressure metamorphic facies occur along cold to ultracold geotherms (5-10 • C/km): BS, blueschist facies; EC, eclogite facies at pressure below the coesite stability field; UHP, ultrahigh-pressure metamorphism, corresponding to the eclogite facies at pressure above the quartz stability field. Wet solidus is for the oceanic basalt-water system. Mineral abbreviations: sill, sillimanite; ky, kyanite; jd, jadeite; ab, albite; qz, quartz; coe, coesite; dia, diamond; gr, graphite.
ites, and UHP eclogites at convergent plate margins. While subduction is a long-lasting process in oceanic subduction zones, it is a short-lived process in continental subduction zones. In either case, it leads to prograde metamorphism with synchronous increases in pressure (P) and temperature (T). On the other hand, exhumation may proceed in various styles such as decompressional cooling, isothermal decompression and decompressional heating. Despite the difference in exhumation paths, retrograde metamorphism in continental subduction zones is temporally associated with decompression. Postorogenic processes take place at a later time and are independent of the orogenesis. They are often referred to as post-collisional or post-subduction processes, whereas late-orogenic or late-collisional processes belong to syn-exhumation tectonism. Although the post-collisional tectonism is of anorogenic nature, it has the geological inheritance in both space and composition from preexisting collisional orogens. Orogens may suffer significant reworking at the time when there is a tectonic conversion from compression to extension due to re-ordering of lithotectonic units through processes such as lithospheric rebound and orogenic collapse. The synorogenic episode evolves into a postorogenic phase during which the nature of metamorphism and magmatism changes due to the combination of slab rollback and breakoff, mantle lithosphere delamination and orogenic extension [17].
Since the discovery of UHP metamorphic index minerals such as coesite and diamond in supracrustal rocks [51][52][53][54], geologists have realized that the continental crust was subducted to subarc depths of 80-160 km and suffered the UHP eclogite-facies metamorphism, and subsequently returned to the surface. Thus, UHP metamorphic rocks provide direct records of crustal P-T changes at the slab surface in subduction zones (Fig. 3). Metamorphic facies are defined by sets of mineral assemblages in crustal rocks of different compositions that recur through space and time (Fig. 4a). On the other hand, geodynamic modeling provides possible pictures for the thermal structure of subduction zones responsible for the metamorphic facies (Fig. 4b).
While lithostatic pressure is sufficient to account for the HP to UHP metamorphic facies along cold to ultracold geotherms, lithostatic temperature is not sufficient to account for high-temperature (HT, 800-900 • C) to ultrahigh-temperature (UHT, 900-1100 • C) metamorphic facies along warm to hot geotherms.
So far, 30 UHP metamorphic terranes have been identified on Earth, with their metamorphic age varying from 7-8 Ma to 620-630 Ma (Table 1). For these UHP eclogite-facies metamorphic rocks,  their peak pressures are usually reached before peak temperatures [4]. Their retrograde metamorphism is often indicated by decompression with an increase to and then a decrease from the maximum temperature. In general, peak metamorphic temperatures of subducted oceanic crust are relatively low (usually < 600 • C), whereas peak metamorphic temperatures of subducted continental crust may be as high as 700-800 • C (even up to 1000 • C, but mostly 650-750 • C). The P-T paths for subduction-related HP to UHP metamorphic rocks can be categorized into three groups [55]: cold loops (to 600 • C), tepid loops (to 800 • C) and hot loops (to 1000 • C). The cold paths are the hallmark of HP blueschist and eclogite terranes, which mainly occur in cold oceanic subduction zones. The majority of continental UHP terranes follow the tepid paths to maximum temperatures of ∼800 • C during exhumation, whereas the hot paths are relatively uncommon in the cycle of collisional orogenesis.
A significant increase in temperature leads to partial melting of UHP metamorphic rocks in collisional orogens [4]. This can significantly affect the physicochemical properties of crustal rocks and their constituent minerals [56,57]. As a consequence, partial melting has a remarkable influence on the exhumation mechanism of deeply subducted crustal slices [58]. In addition, the partial melting of deeply subducted continental crust is an important mechanism for intracrustal differentiation, and thus can considerably influence the geochemical behaviors of various elements and isotopes in subduction zones [4,59]. The resulted melts may be an important material source of intracrustal igneous rocks [60,61]. Because UHP metamorphic rocks experienced deep subduction to subarc depths of >80 km, they may entrain ultramafic rocks such as garnet peridotite and harzburgite offscrapped from the overlying mantle wedge base during exhumation (e.g. REVIEW target to investigate the crust-mantle interaction in continental subduction channel [64,65]. High-T (HT) to ultrahigh-T (UHT) overprinting of ultrahigh-P (UHP) metamorphic rocks have been increasingly found in collisional orogens [66]. These include the Caledonides in Greenland, the Erzgebirge in Germany, the Kokchetav in Kazakhstan, the North Dabie in China and the Rhodope in Greece. Nevertheless, HT-to-UHT metamorphic rocks also occur in intracontinental orogens where UHP metamorphic rocks have not been identified [67,68]. In either case, they are represented by low-to moderate-pressure metamorphic rocks that formed in warm to hot geotherms (Fig. 4). Petrologically, UHT metamorphism is indicated either by robust geothermobarometry or by the presence of a diagnostic mineral assemblage in an appropriate bulk composition, such as assemblages with sapphirine + quartz, orthopyroxene + sillimanite ± quartz, osumulite and spinel + quartz [69][70][71]. Such assemblages are commonly preserved in extremely Mg-Al-rich rocks of restitic origin. Occasionally, widespread assemblages like garnet + orthopyroxene, ternary feldspars, (F-Ti) pargasite or metamorphic inverted pigeonite are taken to be indicators of UHT metamorphism. Because the formation of UHT metamorphic rocks has been bearing not only on orogenic processes, but also on postorogenic reworking, these rocks have been intensively investigated by various approaches. The results indicate that UHT metamorphic events may take place in the geological history of Archean to Cenozoic [67] and their timescale may vary from <10 Myr to >30 Myr [68,72]. In either case, HT-to-UHT metamorphism plays a fundamental role in the development and stabilization of continents in orogens. However, it remains to be resolved how the tectonic switch takes place in subduction zones where the UHP metamorphic regime of cold to ultracold geotherms is switched to the UHT metamorphic regime of warm to hot geotherms.

FLUID ACTIVITY
There are considerable amounts of water in hydrous minerals and nominally anhydrous minerals (Table 2). The stability of these minerals at subduction zones shapes fluid activity at different depths [5]. As a consequence, plate convergence during oceanic subduction and continental collision results in not only dehydration and hydration, but also partial melting and melt metasomatism in subduction channel [4,16]. The addition of fluids released from subducting slab to the mantle wedge leads to its metasomatism [18,73]. On the other hand, the liber-ation of fluids from the subducting slab would modify the compositions of the slab itself [74,75]. As a result, the both mantle wedge and subducting slab would be geochemically influenced by subductionzone fluids. Therefore, understanding the geochemical compositions and physicochemical property of subduction-zone fluids represents one of the most fundamental issues in the study of subduction-zone processes.

Field-based studies
Structural shearing is evident in rock fragments detached from subducting continental slab and offscrapped from the mantle wedge, indicating the generation of aqueous solutions and even hydrous melts in the subduction channel. These fluids provide a medium for crust-mantle interaction [3]. Thus, fluid generation and crust-mantle interaction are closely related to each other during subduction and exhumation of continental crust [64,76,77]. Partial melting of crustal rocks may occur at subarc depths of 100-160 km, producing hydrous melts and even supercritical fluids [4,78,79]. Therefore, the geochemical composition of fluids in continental subduction zones provides an analogue to that in oceanic subduction zones, and certainly sheds light on basic issues such as the crust-mantle interaction in subduction zones, mantle geochemical heterogeneity, arc-like magmatism and continental growth [4,18,80].
How fluids escape from the subducting crust is critical for understanding how the overlying mantle wedge melts for arc volcanism [5]. The stability of hydrous minerals at forearc to subarc depths is the key to fluid activity. Whereas some hydrous minerals are only stable until forearc depths in warm to hot subduction zones, the other hydrous minerals are more refractory and can retain their water to subarc depths in cold to ultracold subduction zones. Arc volcanism is usually linked to dehydration reactions at subarc depths, assuming that water released from the subducting crust forms 'hydrous curtains' where this water rises into the overlying mantle wedge [81]. The geographical curve of arc volcanoes above oceanic subduction zones is often explained as the surface projection of slab dehydration reactions during subduction. However, the experimental study [82] indicates continuous dehydration of the subducting crust from forearc depths of 60-80 km (for hot subduction zones) to postarc depths of >300 km (for cold subduction zones). In this regard, the dehydration and anataxis of subducting slab, albeit substantial to enrichment of the mantle wedge with water and crustal components, would have no direct correspondence in both space and REVIEW Zheng and Chen 503 time to the onset of the mantle wedge melting. Instead, partial melting of the hydrated mantle wedge for arc volcanism is triggered by the asthenospheric heating at a later time [5].
It is usually assumed that the deeply subducting continental crust does not release enough water to hydrate the overlying mantle wedge; the absence of syn-subduction arc magmatism was taken as the geological evidence for the lack of fluid activity in continental subduction zones [12,[83][84][85]. Compared with oceanic subduction zones, where altered mafic igneous rocks contain abundant water prior to subduction, descending crust in continental subduction zones was often considered to be drier than the oceanic crust. Recall that the oceanic crust is predominated by mafic lithology whereas the continental crust is predominated by felsic lithology. After their subduction to forearc depths for HP blueschist to eclogite-facies metamorphism, hydrous minerals in the former are dominated by serpentine and amphibole whereas those in the latter are dominated by biotite and white mica. However, there is no significant difference in the amount of hydrous minerals between the oceanic and continental crust-metamorphosed rocks. Therefore, a similar amount of water may be released from subducting crustal rocks at subarc depths. As a consequence, the mantle wedge overlying the descending oceanic and continental slabs would be equally metasomatized by subduction-zone fluids. Nevertheless, the occurrence of syn-subduction arc volcanics above oceanic subduction zones indicates that the mantle wedge was heated still in the subduction stage for partial melting of the metasomatites, whereas the absence of syn-subduction arc magmatism above continental subduction zones suggests a significant delay of the heating for partial melting of the metasomatites [5].

Geodynamic modeling
Besides the geological, geochemical and geophysical investigations of subduction channel processes, geodynamic modeling becomes a common tool in this aspect (e.g. [39,86]). With few exceptions, previous geodynamic models of slab subduction and continental collision were often based on purely thermo-mechanical principles but neglected the effects of fluid activity [40,87,88]. These models provide a series of quantitative constraints on subduction channel processes, which contribute greatly to understanding of their operative mechanisms. However, with neglecting dehydration and hydration processes in these models, weakening of the overriding lithosphere is rather limited. Thereby, the rocks from the overriding mantle wedge are generally not incorporated into the subduction channel model. In addition, these models did not take into account hydration, partial melting and magmatism in the mantle wedge, leading to significant discrepancies between the geodynamic models and geological observations (e.g. [16]  In order to improve the previous models, Li et al. [89] investigate the detailed structure and dynamics of both oceanic and continental subduction channels. They conducted a suite of high-resolution petrological-thermo-mechanical simulations by taking into account fluid and melt activities. The results suggest that subduction channels are composed of a tectonic mélange formed by crustal rocks detached from the subducting slab and hydrated mantle rocks offscrapped from the overriding mantle wedge base. These rocks are mingled and mixed at mantle depths, and then either extrude sub-vertically upward as diapirs through the mantle wedge to the crust of the overriding plate or exhume along the subduction channel to the surface near the suture zone (Fig. 5). The models also suggest significant differences between oceanic and continental subduction channel processes. A basic aspect is that continental lithosphere cannot subduct for long time as oceanic lithosphere does, due to the much lower density of continental crust than oceanic crust. Therefore, the exhumation of HP to UHP metamorphic rocks are generally associated with continental collision for mountain building. In contrast, the oceanic subduction can be much more continuous if there is no island arc terrane, seamount or oceanic plateau to plugging the subduction.
Another important aspect from the geodynamic model of Li et al. [89] is the fluid activity in continental subduction channel. If a high water content is taken for subducting oceanic crust, the oceanic subduction channel is characterized by strong dehydration from the descending slab, hydration of the overriding mantle wedge and partial melting of the hydrated mantle wedge for arc magmatism. In contrast, the hydration and melting activities in the continental subduction channel may be limited compared with the oceanic subduction channel if a much lower water content is taken for the subducting continental crust. However, if the oceanic crust underwent metamorphic dehydration at forearc depths of <60-80 km, its water content becomes similar to the continental crust prior to UHP eclogite-facies metamorphism [5]. As a consequence, the behavior of fluid activity in both oceanic and continental subduction channels becomes similar at subarc depths of 80-160 km, where the thermal structure of subduction zones plays a more important role in affecting the metamorphic dehydration and partial melting.

MASS TRANSFER AT THE SLAB-MANTLE INTERFACE
It is well known that oceanic crust is rich in water due to seawater-hydrothermal alteration of MORB during their eruption along spreading ridges at low and high temperatures. In contrast, continental crust contains much less water in the shallowest layer prior to subduction. Nevertheless, the amount of water in both crust becomes similar to each other with increasing subduction depth to >30 km, where all pore water has been expelled by metamorphic dehydration from subducting crustal rocks at conditions of HP blueschist to amphibolite facies. Abundances of H 2 O, CO 2 and fluid-mobile incompatible trace elements (especially K and U) are high in continental crust, leading to formation of hydrous minerals in crustal rocks. Thus, continental crust may contain a similar proportion of water to oceanic crust after being metamorphosed to UHP eclogite facies at the subarc depths. The breakdown of hydrous minerals at the subarc depths is a major cause for the mass transfer in both oceanic and continental subduction zones. While small amounts of hydrous minerals in the subducting crust are substantial to water cycling through subduction zones (Fig. 6), much larger volumes of nominally anhydrous minerals carry considerable amounts of water into the mantle [75].
Subducting crustal rocks transport most of the fluid-mobile incompatible elements into the mantle wedge (Fig. 7). This can be considered in the   [74]). Natural high-pressure and ultrahighpressure metamorphic rocks can be used to infer forearc and subarc processes, respectively.
following five aspects. First, the progressive heating and squeezing of these rocks cause progressive phase transformations that increase density and decrease water content [3,6]. Second, the solubility of incompatible elements in subduction-zone fluids increases with temperature and pressure [18,74]. For example, large ion lithophile elements (LILE) are watersoluble at forearc depths and thus transportable by aqueous solutions, whereas light rare earth elements (LREE) are water-insoluble and only can be trans-ported by hydrous melts at subarc depths. Third, hydrophile elements fractionate significantly as subduction and dehydration proceeds at the subarc depths. Fourth, partial melting of these rocks become profound at 750-800 • C due to breakdown of hydrous minerals [4] and a similar process has been reported to take place at UHP conditions corresponding to the subarc depths [90,91]. Although many arc lavas are associated with cold subduction zones, their trace-element systematics show signatures from subducted metasediment [92,93]. This indicates the geochemical transfer from the subducted metasediment to the mantle wedge by hydrous melts at the subarc depths. Fifth, the kinetic retardation may only be a problem in the early stage of subduction when the subducting slab is coupled with the mantle wedge, but the slab-wedge decoupling leads to heating from the laterally filled asthenospheric mantle [5]. This advances dehydration reactions in both the subducted slab surface and the mantle wedge base, transferring abundant water and dissolved materials from the subducting crust to the mantle.
Partial melting of UHP metamorphic rocks has been increasingly recognized in collisional orogens [4,18,76,77]. Petrological evidence for the partial melting is prominent if felsic veins or veinlets of anatectic origin are observed at outcrops or in thin sections (e.g. [77,94,95]). Microstructures on the thin section also serve as petrographic evidence for partial melting, which includes the occurrence of felsic multiphase solid inclusions in mafic minerals such as garnet and clinopyroxene, patchy microstructures highlighted by newly growing garnets, felsic melt film and melting reactions between phengite and quartz (e.g. [76,77,96]). In addition to aqueous solutions, hydrous melts produced under HP to UHP conditions have also transferred crustal components to the overlying mantle wedge. Experimental studies indicate that partial melts of mafic to felsic lithologies can be either adakitic or non-adakitic granitic melts, depending on melting pressure or depth (e.g. [97,98]). A trivial amount of CO 2 can significantly reduce the melting temperature of peridotite and lead to pronounced enrichment of meltmobile incompatible trace elements in carbonate melts. These silica-saturated or silica-undersaturated melts can react with the mantle wedge peridotite in subduction channels to generate mafic to ultramafic metasomatites [3,99].
Lawsonite is a low-T/HP hydrous mineral that is produced in cold subduction zones [100,101]. The upper pressure limit for the stability of lawsonitebearing mineral assemblages may reach 8-10 GPa at 750-900 • C. Because of its high water contents (11.2 wt%) and its enrichment in LREE and REVIEW Sr [102][103][104], it is of specific importance in the mass transfer from the subducting slab to the mantle wedge. Lawsonite-rich rocks are termed as lawsonitite. They contain > 70% lawsonite and have compositions close to the system CaO-Al 2 O 3 -SiO 2 -H 2 O (CASH). They occur together with Chl-Tlc-Amp-rich (±carbonate) rocks in the HP terrane of Alpine Corsica [105]. Thermodynamic calculations [106] suggest that lawsonitites are stable at much higher pressures than the experimental P-limit for lawsonite-bearing HP-UHP metamorphic rocks of mafic composition. The lawsonitites contain a much higher water content (>7.5 wt%) than all major subducted rock types, including metasedimentary, mafic and ultramafic igneous rocks at UHP conditions (e.g. 4 GPa and 550-800 • C). They also have higher trace-element (including Cr and Ti) contents than common oceanic-type blueschists and eclogites. The preference for LREE and Sr over heavy rare earth elements (HREE) in lawsonite has been confirmed by an experimental study of trace-element partition coefficients [107]. Therefore, lawsonite breakdown at subarc depths can deliver LREE-and Sr-enriched fluids to metasomatize the overlying SCLM wedge, generating ultramafic metasomatites with enrichment of LREE and Sr. Partial melting of such metasomatites can give rise to not only synorogenic, but also postorogenic mafic igneous rocks with arc-style geochemical signatures.

TECTONIC TRANSITION FROM OCEANIC TO CONTINENTAL SUBDUCTION
In general, the convergence of two continental plates starts from oceanic subduction and accretion, followed by continental subduction. This results in the tectonic transition from oceanic subduction to continental subduction, which has been recognized for a long time by the occurrence of ophiolites in the Alpine and Himalayan orogens [108,109]. The second type of lithotectonic evidence is provided by the identification of oceanic-type eclogites identified in the Hong'an orogen [110][111][112] and the North Qaidam orogen [113][114][115]. The third type of lithotectonic evidence is the occurrence of oceanic island basalts (OIB)-like mafic igneous rocks in collisional orogens [116].
The Cenozoic Alpine and Himalayan orogens provide two typical examples for such a composite orogenic process [108,109], where the subduction of Tethyan oceanic crust is following by the subduction of continental crust. However, the composite orogenesis commonly results in collisional tectonics that superimpose on the accretion tectonics [71], leading to the complexity in identifying the tectonic processes of two-stage orogenesis. The Qinling-Tongbai-Hong'an-Dabie-Sulu orogenic belt in central China represents another type of composite orogens that assembled the North and South China Blocks in the Early Mesozoic. This belt comprises a Paleozoic accretionary orogenic system in the north and a Mesozoic collision orogenic system in the south [111,117,118]. The older accretionary orogenic system shows little reworking by the younger collisional event. Thus, it serves as an ideal place to unravel the tectonic evolution from oceanic subduction/accretion to continental collision.
The Tongbai-Hong'an orogens are located in the central part of the composite Qinling-Tongbai-Hong'an-Dabie-Sulu orogenic belt. As outlined by Liu et al. [119], the tectonic evolution of the two orogens involves four stages. Based on the tectonic affinity and metamorphic P-T paths of the Kuanping, Erlangping and Qinling Groups in the Early Paleozoic accretionary orogenic system, an arc-continent collision is inferred to be induced by the subduction of oceanic slab during the period 440-420 Ma, creating the Andean-type active continental margin on the newly accreted North Qinling terrane to the North China Block. For the Late Paleozoic (340-310 Ma) accretionary process, a model of paired metamorphic belts can be used to explain the occurrence of Carboniferous eclogites within the Triassic continental HP metamorphic terrane. By taking into account the isotope-age spectra of different tectonic units in the Early Mesozoic orogenic system, it is possible that there are the diachronous subduction and exhumation of continental slices from 255 to 210 Ma [120]. The present architecture and tectonic style of the composite Qinling-Tongbai-Hong'an-Dabie-Sulu orogenic belt may result from tectonic extension in the Late Mesozoic (140-120 Ma). Taken together, the Qinling orogen is the composite accretionary-type collisional orogenesis in the west whereas the Dabie-Sulu orogens are the collisional orogenesis in the east, with the Tongbai-Hong'an orogens as the transitional orogen in the middle.
While eclogite is the hallmark of cold subduction zones, its mafic protolith can be of either continental or oceanic origin. Whereas continentaltype eclogites are indicated by their arc-style traceelement and relatively enriched Nd isotope compositions, oceanic-type eclogites are usually indicated by their MORB-style trace-element and relatively depleted Nd isotope compositions [113,121]. If a suite of eclogites from the same orogen show a large variation in trace-element composition from arcstyle to MORB-style in the spidergram (sometimes with OIB-like eclogites) with relatively depleted Nd isotope composition, such composite compositional features indicate that these eclogites were metamorphosed from backarc basin basalts. This has been documented not only for Late Paleozoic eclogites in Chinese Southwest Tianshan [121] and the Huwan shear zone in the Hong'an orogen [112], but also for Early Paleozoic eclogites in North Qilian [122]. A zirconological study of UHP eclogites from North Qaidam also suggests that their protoliths were transformed from backarc basin basalts in the Early Paleozoic [115]. It appears that these subduction zones were converted from extinct backarc rift zones. It is intriguing how such tectonism takes place along continental margins, which merits to study by geodynamic modeling.
It is commonly assumed that continental deep subduction to mantle depths is dragged by the negative buoyancy of descending oceanic slab [6,123]. Although UHP eclogites with oceanic crust-like protoliths also occur in collisional orogens such as the Western Alps in Italy [109] and Southwest Tianshan in China [124], no UHP terranes have been observed in oceanic subduction zones [9]. This is not because these UHP terranes fail to form at subarc depths, but because they are too dense to be exhumed from such depths. In fact, there is a tectonic continuum from the subduction of oceanic lithosphere to the collision of continental lithosphere with continental lithosphere or marginal arc terranes. Because modern oceanic crust has the old-est age of <200 Ma, the time interval between the growth of precedingly subducted oceanic crust and the oceanic-type eclogite-facies metamorphic event is smaller than 200 Myr in collisional orogens. If the time interval is much larger than 200 Myr, the oceanic-type eclogite would be derived from reworking of the previously subducted oceanic crust in ancient accretionary orogens.

CRUST-MANTLE INTERACTION
Plate subduction is the most important mechanism for exchanging mass and energy between the mantle and the crust. Crustal materials can be recycled into the mantle to cause the crust-mantle interaction, resulting in mantle heterogeneities in geochemistry and lithology. Although continental subduction zones are characterized by the absence of syn-subduction arc magmatism, it does not mean that there was no metasomatism of the mantle wedge by subducting crust-derived fluids (Fig. 8). Because of the low geotherms there, mantle metasomatites generated by the fluid-peridotite reaction did not undergo partial melting in the subduction stage [5]. Instead, they become partially melted at a later time, either during the exhumation of deeply subducted continental crust to produce alkaline igneous rocks or in the post-subduction stage to produce postcollisional calc-alkaline igneous rocks (e.g. [29,61]). These rocks provide an analogue to the product of the crust-mantle interaction in oceanic subduction zones.

Experimental petrology
The metasomatic reaction between hydrous siliceous melt and mantle peridotite at subarc depths was first proposed by Nicholls and Ringwood [125]. This process was then explored experimentally by Sekine and Wyllie [126,127] using either mixtures of granite and peridotite with water or reaction couples of hydrous granitic melt adjacent peridotite. The results indicate ubiquitous occurrence of orthopyroxene and jadeitic pyroxene in the reaction zones. Afterwards, many experiments have been made to study melt-peridotite reactions at different pressures (e.g. [98,99,[128][129][130]). Two types of lithology were placed in contact with peridotite. One is hydrous felsic rocks including synthetic and natural granites, and the other is mafic rocks including MORB, amphibolite, eclogite, amphibole pillow lava and andesite. Run products in the first case are mainly phlogopite-and garnet-bearing pyroxenites, whereas those in the REVIEW second case are garnet pyroxenite, garnet lherzolite, olivine-pyroxenite and olivine-amphibolite. Experimental melts are generally high-Mg adakite, Fe-and Ti-rich calc-alkaline basalts and primitive alkaline basalts. The composition of run products changes with mineral and chemical compositions of starting materials, melting degrees and melt/peridotite ratios. Zoning of reaction products is common for the second type of reactions. At temperatures below the solidus of peridotite, partial melting of eclogite produces felsic melts and garnet pyroxenite residues with a reaction transition zone of (garnet) pyroxenite. At temperatures above the solidus of peridotite, the peridotite forms a zone of dunite-harzburgite-lherzolite toward the reaction melt.
There are also two types of melt-peridotite reactions in terms of lithochemistry (e.g. [128][129][130][131][132]). One is the reaction between silica-saturated melt and peridotite, which consumes olivine to generate orthopyroxene and clinopyroxene and result in fertile lherzolite and garnet pyroxenite. There is a dense reaction zone around olivine and orthopyroxene grains that isolates melts from further reaction with them. It is possible that melt would be localized into narrow channels to ascend by buoyancy. The other is the reaction between silica-undersaturated melt and peridotite, which consumes orthopyroxene to generate fine-grained olivine. The reaction products are mainly wehrlites. This type of reaction generates a loose structure in peridotite by dissolving away orthopyroxene that facilitates melt migration and melt-peridotite interaction. As the reaction continues, the continuous dissolution of orthopyroxene results in a gradual increase of SiO 2 content in the melt if the melt cannot be timely supplemented. The silica-undersaturated melt would eventually transform into a silica-saturated melt, which continues to react with the peridotite until the melt is completely consumed or rises to shallow levels.
Although available experiments are mostly dedicated to the crust-mantle interaction in oceanic subduction zones, results with the low geotherms at subarc depths provide an analogue to the meltperidotite reaction in continental subduction zones. Nevertheless, the significant difference in rock composition and property between continental and oceanic subduction channels can lead to a series of variations in deep physical and chemical processes as well as crust-mantle interacted products [16]. Thus, it is crucial to perform experiments on all possible reactions between peridotite and crustal material that includes rocks and their derived melts at appropriate P-T conditions corresponding to subarc and postarc depths in subduction channels.

Orogenic peridotites
Field-based studies of petrology and geochemistry indicate that subduction-zone fluids would react with the mantle wedge peridotite in subduction channel, resulting in two types of metasomatites [3,5]. The first is serpentinized to chloritized peridotites due to reaction with aqueous solutions at subsolidus conditions, and the second is pyroxenite to hornblendite due to reaction with hydrous felsic melts at and above solidus conditions. In the first case, Fe and Mg are leached out from the peridotite mainly at forearc depths and concentrated in the altered products. Once these altered peridotites undergo partial melting at greater depths for mafic magmatism, high-Mg olivine is produced in residual harzburgites. In the second case, olivine-poor metasomatites are generated at subarc to postarc depths due to incorporation of felsic melts, and their partial melting gives rise to variable compositions of mafic melts. In either case, the fluid-peridotite reaction is heterogeneous and the metasomatites are more susceptible to partial melting than the peridotite upon heating.
Peridotite fragments are common in collisional orogens, providing a window to see the geodynamic processes of continental subduction/exhumation and crust-mantle interaction. This is typified by peridotites from the Dabie-Sulu orogenic belt [65]. The Dabie-Sulu orogenic peridotites are commonly classified into 'crustal' type and 'mantle' type [63]. The former was crystallized from mafic magmas that underwent magmatic differentiation in the lower continental crust before subduction, whereas the latter was originally located in the SCLM wedge that is of cratonic origin in the case of the North China Block. These two types of orogenic peridotites exhibit a series of differences in mineral texture, geochemical composition and metamorphic evolution. Most orogenic peridotites belong to the 'mantle' type; they were offscrapped by the subducting continental crust into subduction channel and exhumed together with felsic UHP metamorphic rocks to the crustal level [3]. During these processes, they suffered multistage metamorphism and metasomatism, making them the best sample for studying the physicochemical effects of crust-mantle interaction at subarc depths [18,64].
The Dabie-Sulu orogenic peridotites typically contain various metasomatic minerals, implying that the mantle wedge overlying the deeply subducted continental crust would have undergone multistage metasomatism during the preceding oceanic subduction to the later continental subduction. Metasomatic agents are dominated by one species of subduction-zone fluids such as aqueous solutions, hydrous melts and even supercritical fluids that are derived from metamorphic dehydration and partial melting of the subducted crust [64,[133][134][135]. Rutile, zircon, Ti-clinohumite and Ti-chondrodite are common metasomatic minerals in these peridotites, providing mineralogical evidence for mobilization of high field strength elements (HFSE) from the subducted crust into the mantle wedge. These metasomatic agents are generally in equilibrium with metasomatic minerals such as orthopyroxene, phlogopite and magnesite, indicating that the incoming fluids are rich in Si, Al, K, F, CO 2 , LILE, LREE and HFSE. Zircon in orogenic peridotites is an important mineral that enables us to constrain not only the source of metasomatic agents, but also the metasomatic age. Metasomatic zircon domains show U-Pb ages of 212-231 Ma [65,136], slightly younger than the UHP metamorphic age of 225-240 Ma [3]. These zircon domains commonly have high Th/U ratios (>0.1), demonstrating that they were crystallized from hydrous melts produced in the early stage of exhumation. Few zircon grains show old U-Pb ages of Neoproterozoic to Archean, representing relict detrital zircons that were inherited from the crustal rocks of UHP metamorphism [3,18,136].

Mafic igneous rocks
It has been a paradigm that the subduction of oceanic crust and subsequent crust-mantle interaction are the principal geological processes for the mantle chemical heterogeneity and the traditional plate tectonics theory assumed that only the oceanic subduction zone is the most active region of crustmantle interaction. However, in addition to the subduction of oceanic crust, continental crust can be also subducted into the mantle to experience UHP metamorphism at subarc depths and then exhumed back to shallow crustal levels. Thus, it is intriguing to know whether recycling of subducted continental crust and crust-mantle interaction also occur in continental subduction zones. Because the subduction of continental crust is dragged by the subduction of oceanic crust, it is substantial to examine whether the oceanic crust was involved prior to the continental subduction. Mafic igneous rocks in subduction zones are important carriers for studying crust-mantle interaction. While syn-subduction arc volcanics are common in oceanic subduction zones, they are generally absent above continental subduction zones. However, post-collisional igneous rocks are widespread in continental subduction zones, providing excellent targets to study the crustmantle interaction during collisional orogenesis.
A series of comprehensive studies have been devoted to post-collisional mafic igneous rocks from the Hong'an-Dabie-Sulu orogens and the southeastern edge of the North China Block in eastcentral China [29,116,137]. The results indicate two types of the crust-mantle interaction in the continental subduction zone, which are represented by two types of mafic igneous rocks with distinct geochemical compositions [138]. The first type of rocks exhibit arc-style trace-element distribution patterns (i.e. enrichment of LILE, LREE and Pb, but depletion of HFSE) and enriched radiogenic Sr-Nd isotope compositions. In contrast, the second type of rocks show OIB-style trace-element distribution patterns (i.e. enrichment of LILE and LREE, but no depletion of HFSE) and depleted radiogenic Sr-Nd isotope compositions. Both of them have variable zircon O isotope compositions, which are different from those of normal mantle zircon, and contain relict protolith zircons. These geochemical features indicate that the two types of mafic igneous rocks were originated from the different natures of mantle sources. The mantle source for the second type of rocks would be generated by reaction of the overlying juvenile SCLM wedge peridotite with felsic melts originated from the precedingly subducted oceanic crust [116], whereas the mantle source for the first type of rocks would be generated by reaction of the overlying ancient SCLM wedge peridotite of the North China Block with felsic melts from subsequently subducted continental crust of the South China Block [29]. In either case, the enrichment of melt-mobile incompatible trace elements in the mafic igneous rocks requires twostage processes of partial melting for geochemical differentiation. The first stage is partial melting of the oceanic/continental crust in the syn-subduction stage for the generation of felsic melts and the second stage is partial melting of the mantle metasomatites in the post-subduction stage for mafic magmatism. As a consequence, the two types of crust-mantle interaction in the oceanic-continental subduction zone are realized by the melt-peridotite reaction in the same subduction channel. Therefore, the post-collisional mafic igneous rocks provide petrological and geochemical records of the slabmantle interactions in collisional orogens.

OPHIOLITE CONUNDRUM
As the fragments of oceanic lithosphere that were tectonically emplaced at continental margins, the occurrence of ophiolites in collisional orogens is regarded as the lithotectonic evidence for the tectonic transition from oceanic to continental subduction. However, igneous secessions in ophiolites are commonly viewed as the vertical structure of oceanic lithosphere at spreading ridges. From REVIEW bottom to top, a typical and complete ophiolite secession is composed of depleted mantle peridotite with tectonite at the base, layered ultramafic-mafic cumulates, massive (isotropic) gabbro, sheeted dikes and extrusive volcanic rocks represented by pillow basalts [139,140]. The secession is typically overlain by pelagic sediments and chert, indicating their deposition in deep-sea environments.
In the early study of ophiolites, they were categorized into Tethyan and Cordilleran types [141,142]. Tethyan-type ophiolites are typical in Oman, Troodos, Pindos, Vourinos and Muslim Bagh, where they are placed onto passive continental margins and are typically overlain by sediments characteristic of passive-margin settings (limestone, dolomite). In contrast, Cordilleran-type ophiolites are common in Coast Range of California, Trinity and Cape Vogel, where they are associated with active continental margins and are typically underlain by accretionary complexes and overlain by clastic sediments deposited in a forearc basin setting (turbidites, mudstones, conglomerates). The two types of ophiolites are similar in that they commonly exhibit complete or incomplete ophiolite secessions in seafloor spreading environments such as midocean and backarc rifts (e.g. [143,144]). On the other hand, they are geochemically characterized by arc-style trace-element compositions, showing enrichment in fluid-mobile incompatible trace elements such as LILE and LREE but depletion in fluid-immobile incompatible trace elements such as HFSE and HREE (e.g. [145][146][147]). The ophiolite conundrum describes the conflict in many ophiolite complexes between (i) the petrological evidence for intraplate rifting-style magmatism at divergent plate margins and (ii) the geochemical evidence for the presence of volcanic arc-style signatures at convergent plate margins (e.g. [144,[146][147][148]).
On the other hand, a corollary of plate tectonics is that spreading ridge systems ultimately interact with trenches during subduction [149,150]. Although the majority of spreading ridges moves obliquely toward continental margins, it does not preclude the possibility that some fossil ridges move in a parallel way toward continental margins. In this case, such ridges may be converted into the trenches without development of slab windows. As a consequence, the fossil ridges would have the weakest interaction with the continental margins. As the fragments of oceanic lithosphere on continental crust, the occurrence of ophiolites testifies to the operation of plate tectonics in two ways [151]: their generation requires seafloor spreading and their emplacement requires plate subduction. However, this seemingly irreconcilable conflict between the divergent and convergent environments for ophiolites can be re- solved by a ridge-trench conversion model, in which fossil-spreading ridges are converted to new subduction zones along continental margins. Therefore, one specific aspect of plate tectonics can be deciphered by examining ophiolites, which are an unequivocal index of subduction tectonics.

Continent
The ridge-trench conversion takes place when a spreading ridge is moved in a parallel way to an active continental margin during the plate convergence. The ridge-trench conversion model for ophiolite occurrence at the continental margin can be outlined in the following four stages (Fig. 9). (i) Seafloor spreading generates a mid-ocean ridge in coupling with subduction of an oceanic slab beneath an active continental margin (Fig. 9a). Although these two tectonic processes are complementary to each other in geodynamics, the former gives rise to a secession of mafic igneous rocks with MORB-style geochemical signatures [144] whereas the latter may produce Andean-type magmatism if the subductionzone geotherm is high enough for partial melting of hydrated and metasomatized peridotites in the mantle wedge [5]. (ii) A new spreading ridge is generated in the oceanic plate, pushing the old ridge toward the continental margin in coupling with subduction of the old oceanic slab beneath it. The fossil ridge is eventually moved in a parallel way to the continental margin (Fig. 9b). (iii) The subduction is inhibited in the old subduction system due to the arrival of the fossil ridge at the continental margin, REVIEW Zheng and Chen 511 converting the weak zone along the ridge-slab interface into the trench (Fig. 9c), where the young oceanic slab is subducted beneath the fossil ridge and its attached oceanic slab close to the continental margin. (iv) The later descending slab dehydrates and partially melts at forearc to subarc depths in the new subduction system, liberating fluids and their dissolved components into the overlying fossil ridge and its attached mantle wedge close to the continental margin. These mafic-ultramafic rocks were offscrapped by the subducting slab into the subduction channel and eventually exhumed to the continental margin (Fig. 9d). The combination of these steps results in the ophiolites showing the petrological sequence of spreading ridges but the geochemical signature of subduction zones. While the emplacement mechanism of ophiolites is commonly ascribed to obduction of the mafic-ultramafic secession onto the continental margin, it generally overlooks the tectonic conversion from fossil-spreading ridges to new subduction zones. This led Dilek and Furnes [152] to categorized the mechanisms of ophiolite formation into subduction-unrelated (in mid-ocean or backarc rifts) and subduction-related (with volcanic arc signature) ophiolites. However, it is the ridge-trench conversion that causes the one side of the oceanic plate to be subducted beneath the other side of the oceanic plate that is readily located beneath the continental margin. As a consequence, the old subduction system is replaced by the new subduction system, with the fossil ridge and its attached oceanic slab close to the continental margin. Because the ridge-trench conversion would have only occurred along the continental margin, this is the reason why ophiolites are common along Tethys zones. Furthermore, once the ridge-parallel subduction zone is created, the descending slab dehydrates at forearc to subarc depths to hydrate the overlying fossil ridge and its attached oceanic slab, transferring fluid-mobile incompatible elements (crustal components) to the igneous secession. This explains why many ophiolites show arc-style geochemical signatures.
Ophiolites are abundant in collisional orogens, indicating their transition from accretionary orogens. Whereas the peak ages of their generation generally predate the orogenic events, their peak ages of emplacement are usually consistent with collisional events. Although mid-ocean ridge-generated oceanic lithosphere has been extensively subducted, the occurrence of ophiolites is dictated by preservation of their fragments in collisional orogens. While the generation of mafic-ultramafic igneous secessions marks the crust-mantle differentiation at divergent plate margins, the placement of such seces-sions marks the tectonic conversion at convergent plate margins. The former event can be dated by the whole-rock Sm-Nd isochron method and the latter event can be dated by the U-Pb isotope analysis of zircon from diorite or plagiogranite of the synmagmatic plutonic sequence. An intermediate time can be constrained by the ages of radiolarian fossils in overlying pelagic chert deposited in the trench during the oceanic subduction. Age gaps between the deposition of pelagic sediments and ophiolite emplacement in continental margins and/or island arc terranes are commonly less than 25 Myr [153]. Such a short duration is mostly consistent with the life span of many backarc basins (<20 Myr). This indicates a longer duration than 25 Myr for the tectonic conversion from mid-ocean ridges to subduction zones in a parallel way. The emplacement of some ophiolites would have occurred soon after their creation in backarc basins, suggesting that these ophiolites represent the fragments of young oceanic lithosphere that were detached while still hot. However, the placement of the other ophiolites would have occurred long after their creation in mid-ocean ridges, suggesting that such ophiolites represent the fragments of old oceanic lithosphere that were detached while readily cold. The difference in the age of subducting oceanic lithosphere is correlated with the thermal structure of subduction zones, which has a bearing on the dehydration and melting of descending slab at different depths [5,86].

MAGMATISM IN OROGENS
Magmatism in orogens has the intimate connection with growth and reworking of continental crust in the history of Earth. However, the relationship between the origin of magmatism and the growth of continental crust is still enigmatic. Various compositions of magmatic rocks may be produced by tectonic processes from oceanic subduction through continental collision to post-collisional reworking in extinct subduction zones. However, there are challenges to distinguish between syn-subduction arc and post-subduction arc-like magmatic rocks, between syn-exhumation and post-collisional alkaline magmatic rocks, and between orogenic and anorogenic magmatic rocks. A resolution to these issues has great bearing on the development of plate tectonics from oceanic subduction zones to continental collision orogens [17].
Orogenic magmatism refers to the magmatism that takes place from a long-lasting period of subduction through a short-lived period of collisional orogensis. These processes are recorded by geochemistry of the product orogenic magmas, which represent the end result of multistage and REVIEW multicomponent processes. These processes can be categorized into two major stages, which exert important influences on the style and composition of the resulted magmas [17]: (i) subduction, leading to the first episode of dehydration and anataxis of the descending crust. This produces the metasomatic agents for reaction with the overlying mantle wedge; (ii) collision, resulting in tectonic imbrication and detachment of the subducting lithosphere at different depths, with the second episode of dehydration and anataxis of the exhuming UHP rocks. Each of these processes may involve different magma sources and melting regimes. The conversion from an arc system to a collisional system is indicated not only in the age of magmatic events, but also in the nature of magma sources. While calcalkaline arc volcanics are the typical product of synsubduction magmatism, alkaline igneous complexes are commonly produced by syn-exhumation magmatism. Although arc-style geochemical signatures are shared in both cases, the magma sources of older arc volcanics contain much more components of the oceanic crust than the continental crust. In contrast, the magma sources of younger igneous rocks would be dominated not only by the SCLM wedge peridotite in major elements, but also by the deeply subducted continental crust in melt-mobile incompatible trace elements and their pertinent radiogenic isotopes [3].
Magmatic rocks in collisional orogens may have not only highly variable lithologies from ultramafic through mafic and felsic to alkali, but also highly variable radiogenic isotope compositions from enriched to depleted [138,[154][155][156][157], and their traceelement geochemistry is predominated by arc-style distribution patterns in the spidergram. While felsic magmatics are derived from reworking of deeply subducted continental crust [154,158,159], mafic magmatics originate from ultramafic metasomatites that were generated by reaction of the SCLM wedge peridotite with felsic melts derived from partial melting of the deeply subducted continental crust [29,30,116,137,138]. Although postorogenic magmatics mostly show the arc-style trace-element signature, their generation is generally independent of synorogenic tectonism. This can be distinguished by the following observations: (i) whole-rock Sm-Nd or zircon Lu-Hf isotope compositions, which are relatively depleted for syn-subduction arc magmatics showing positive to less negative ε Nd (t) or ε Hf (t) values characteristic of the juvenile mafic crust but considerably enriched for postorogenic arc-like magmatics showing more negative ε Nd (t) or ε Hf (t) values typical of the ancient continental crust; (ii) synmagmatic zircon abundances, which are large in the postorogenic magmatics but very small in the arc magmatics. This is because arc melts are characterized by depletion in HFSE including Zr and thus rarely allow crystallizing zircon in the synmagmatic stage. As such, it is impossible for the arc-like magmatics to be the product of arc magmatism if abundant zircon grains are separatable from them; (iii) lithological occurrences, with different volumes for different lithologies. Oceanic arc volcanics are dominated by basalts, with small amounts of andesites and rhyolites; continental arc volcanics are dominated by andesites and basalts, with small volumes of rhyolites. In contrast, the postorogenic magmatics are generally dominated by felsic rocks such as granites and rhyolites. Furthermore, adakitic rocks with relatively enriched Nd-Hf isotope compositions may be produced by partial melting of the previously subducted oceanic crust in ancient accretionary orogens.
On the other hand, caution must be taken when linking the relatively depleted Sr-Nd-Hf isotope compositions of granites to a contribution from the mantle. This is because granitic melts cannot be directly produced by partial melting of mantle peridotite according to experimental petrology [160,161]. Thus, any of depleted mantle isotope signatures in granites must be transferred to the juvenile mafic crust at first and then partial melting of the mafic crust gives rise to felsic rocks with relatively depleted radiogenic isotope compositions [17]. As such, elevated ε Nd (t) values for mafic rocks indicate increased contributions from the depleted MORB mantle, whereas elevated ε Nd (t) values for felsic rocks only mean increased contributions from the juvenile crust. This is also the basic mechanism for the relatively depleted radiogenic Nd isotope composition of Paleozoic granites in the Central Asian Orogenic Belt. In fact, only mafic igneous rocks are the characteristic product of mantle melting, but their geochemistry may vary significantly depending on the nature of melted mantle domains. It is MORB-like from the depleted domains of asthenospheric origin but arc-like from the enriched domains of metasomatic origin. Although partial melting of the subducting oceanic crust (eclogite) can produce adakitic melts with relatively depleted radiogenic isotope compositions, their melt-mobile incompatible trace-element abundances cannot be elevated to the level of common arc volcanics by the single-stage process of partial melting.

THERMAL EVOLUTION OF OROGENS
Post-subduction magmatics are abundant in both accretionary and collisional orogens, but their origin has been enigmatic with respect to the heat REVIEW Zheng and Chen 513 source for partial melting. So far, there is no evidence for the generation of P-T conditions under which crustal rocks undergo extensive partial melting for orogen-scale magmatism in post-subduction stages. Although felsic magmatics of large volumes are present in fossil orogens, their composition is not complementary to mafic granulites that are usually considered as the residue of crustal melting. On the other hand, many granulites preserve mineral assemblages that formed at UHT metamorphic conditions, corresponding to partial melting of the lower continental crust. Many studies have been devoted to quantification of the peak metamorphic P-T conditions with improved thermodynamic constraints on the mineral P-T stability [162,163]. When these P-T data are compared with geotherms suggested by geodynamic models that simulate the thermal behavior of continental lithosphere in various tectonic settings, it is found that these geodynamic models for orogenic processes cannot yield the P-T conditions recorded by UHT metamorphic rocks. Continental geotherms at the Moho depth only can have the highest temperature of 650 • C [19]. As a consequence, the heat source for orogen-scale magmatism also remains enigmatic. A number of additional heat sources are proposed to account for UHT metamorphic conditions in fossil orogens. These include: (i) high radioactive heat production in thickened crust; (ii) high mantle heat input to backarc basins; (iii) mechanical heating in ductile shear zones; (iv) high heat influx from the asthenospheric mantle. As argued by Clark et al. [163], the former three sources can elevate the Moho temperatures to 700-850 • C, but they cannot achieve temperatures as high as 950-1100 • C. Thus, the UHT conditions require a heat supply from somewhere below the mantle lithosphere, appealing for a thermal regime that does not occur in the lithostatic temperature at the Moho depth. Together with the orogen-scale magmatism, the high heat influx is necessary for partial melting of mafic lithologies such as eclogite, granulite and amphibolite in the lower continental crust. Therefore, the asthenopsheric heating model is favored for the additional heat supply, which can be achieved by lithospheric foundering beneath thickened orogens (Fig. 10). In this regard, the bottom-up mantle action would probably play a dominant role in regulating the lithospheric evolution by serving as an important thermal energy-transfer system. This is also the basic cause of the switch in the thermal structure of orogens from a warm regime in the early stage to a hot regime in the late stage.
In general, there are three feasible mechanisms for the asthenospheric heating: (i) delamination of subduction-thickened lithospheric roots along the  Figure 10. The asthenospheric heating model for coupled magmatism and ultrahigh-temperature metamorphism in thickened orogens. Modified from geodynamic models of Molnar et al. [170] and Pysklywec et al. [171]. (a) Lithospheric thickening through subduction and collision; (b) the lower continental crust and underling lithospheric mantle is heated by underplating of the asthenospheric mantle due to foundering of the orogenic root.
Moho [164]; (ii) breakoff of deeply subducting slab at the transition between oceanic and continental lithospheres [165]; (iii) convective removal of the mechanically unstable thick root of the mantle lithosphere [166]. The delamination of the entire mantle part of the lithosphere from the crust is a possible mechanism to induce extensive partial melting of the lower crustal lithologies; so is the slab breakoff. Either of these two mechanisms involves rapid heating of the orogenic root at subarc depths. The slab breakoff even results in overprinting of UHP metamorphic rocks by UHT metamorphism still in active subduction zones. However, either the delamination or the breakoff is a vertical foundering process, leaving the space for asthenospheric upwelling and thus its partial melting due to decompression [167]. This would give rise to geochemically depleted mafic igneous rocks that show MORB-style trace-element and radiogenic isotope compositions. However, such igneous rocks have never been observed in association with UHT metamorphic rocks in fossil orogens. On the other hand, the development of a subduction zone beneath a continental block shuts off the mantle region in the continental interior from subduction, leading to lateral flow of the asthenospheric mantle beneath two sides of the orogenic root. Therefore, it is more likely that the lithospheric roots in fossil subduction zones are REVIEW convectively removed by lateral erosion of the asthenospheric mantle [168][169][170]. Because the convective removal is a sluggish process, it takes time to transfer the asthenospheric heat to the lower continental crust for the UHT metamorphism and consequent magmatism in the post-collisional stage.
In order for the asthenospheric heating model to work in fossil orogens (Fig. 10), the lower continental crust has to be dehydrated and partially melted in a layered way from bottom to top. Thus, the orogenic root is first heated from its base by lateral convection of the asthenospheric mantle at temperatures above the dry solidus (900-1100 • C), resulting in a deep crust hot zone in which dehydration and anatexis of the thickened lithospheric bottom take place under fluid-absent conditions (dehydration melting). This produces fluids (dominated by aqueous solutions with minor amounts of hydrous melts) that rise toward the overlying lower continental crust, leaving a less mafic, lower-density and granulitic residue layer in the lowest continental crust. The fluids metasomatize the overlying crustal rocks at different temperatures, leading to (i) amphibolitefacies overprinting if the temperature is below the wet solidus; and (ii) partial melting under fluidpresent conditions (hydration melting) if the temperature is at and above the wet solidus. While lowdegree partial melting produces amphibolite-facies migmatites, high-degree partial melting gives rise to large volumes of felsic melts that are able to buoyantly rise into the upper continental crust. This leaves a more mafic but higher-density, granulitic residue layer in the majority of the lower continental crust. The higher-density granulite residue is gravitationally unstable and thus susceptible to foundering into the underlying asthenospheric mantle, whereas the lower-density granulite residue is susceptible to tectonic exhumation during post-subduction orogensis. This is the reason why the exposed granulite is not complementary in geochemical composition to the emplaced granites.
It is intriguing which lithotectonic units were juxtaposed against the asthenospheric mantle subsequent to the foundering of orogenic roots. If they are a slab section composed of continental crust and mantle, bimodal magmatism is expected to occur due to partial melting of the both crustal and mantle rocks. If it is the largely thinned mantle lithosphere that contains mafic-ultramafic metasomatites, mafic magmatism would be abundant due to heating of the asthenospheric mantle. If it is only the half thinned mantle lithosphere of refractory property, there could be no mafic magmatism due to partial melting of either the asthenospheric or the lithospheric mantle. In either case, the lithospheric foundering leads to HT-to-UHT metamor-phism of the lower continental crust and subsequent felsic magmatism. An anticlockwise P-T path is produced for UHT metamorphic rocks, with nearly isobaric heating in the early stage but isothermal decompression in the late stage. Therefore, the UHT metamorphism is temporally and spatially associated with orogen-scale magmatism, being the hallmark of lithospheric foundering at convergent plate boundaries. As such, it marks the tectonic conversion not only in individual orogens from compression to extension, but also in supercontinental evolution from amalgamation to breakup.

CONCLUDING REMARKS
Subduction zones are central for understanding what drives plate motions and how continental crust is formed. While they are part of the dynamic Earth systems with unparalleled scale and complexity in both space and time, their tectonic evolution can be generalized by the formation of HP to UHP metamorphic rocks in the early stage and mafic arc volcanics in the late stage. These two types of rocks are the complementary end products of subduction-zone processes. As such, the early stage of subduction-zone processes can be revealed by studying the metamorphic dehydration of crustal rocks at high-to ultrahigh-pressure conditions, whereas their late stage can be deciphered by studying the dehydration melting of crustal and mantle rocks at subarc to postarc depths. The transport of water and crustal components from the subducting crust to the mantle wedge is substantial to the mass transfer at the slab-mantle interface in subduction channels. Continental subduction zones provide a natural laboratory to study the mass transfer, which is key to the understanding of geochemical fluxes between Earth's surface and interior. The study of UHP metamorphic rocks and post-subduction igneous rocks from collisional orogens also provides an important supplement to the arc volcanics in accretionary orogens. In particular, lithospheric foundering may develop in thickened orogens, where the mechanically unstable thick root of the orogenic lithosphere is eroded into the asthenospheric mantle. Thus, the post-orogenic magmatism and UHT metamorphism develop in response to the asthenospheric heating, with the timescale depending on the mechanism of asthenospheric underplating. Such coupled processes of orogen-scale magmatism and UHT metamorphism are generally later than and thus independent of the orogenesis, and their coupled occurrence marks the disappearance of negative buoyancy in orogenic roots. Therefore, refinements and discoveries of REVIEW Zheng and Chen 515 fundamental processes in intracontinental orogens can be expected to result from an integrated interpretation of field-based observations and geodynamic models on continental subduction zones and their preceded oceanic subduction zones. The thermal structure of subduction zones may convert from cold in the early stage to hot regimes in the late stage. The time interval in the conversion of thermal regime may vary from less than one million years to more than tens of millions of years. This is primarily dictated by the property of the mantle wedge and the rate of plate convergence. In modern oceanic subduction zones, the cooled, metasomatized mantle wedge can be externally heated by thermal conduction of the asthenospheric mantle either from the backarc position or from below after rollback of the subducting slab, or both. These two ways of asthenospheric heating would generally take place still in the subduction stage, resulting in the syn-subduction arc volcanism. In continental subduction zones and fossil to extinct oceanic subduction zones, however, the asthenospheric heating is later than the subduction period because the backarc mantle is also cold and the slab-mantle coupling is kept for a long time. Thus, the heat transfer from the asthenospheric mantle to the cold, metasomatized SCLM wedge does not take place in the subduction stage; instead, it happens either during exhumation of deeply subducted crustal slices due to the slab-mantle decoupling or in the post-collisional stage due to foundering of orogenic roots. Therefore, the arc magmatism is not immediately triggered by fluxing of the slab-derived fluids into the mantle wedge, but the metasomatized mantle wedge would be externally heated for partial melting still in the subdcution stage despite variable timescales of delay in thousands to millions of years from dehydration of the subducting slab. In contrast, the post-subduction magmatism and its preceding UHT metamorphism take place considerably later than the syn-subduction tectonism, suggesting that the time of heat supply for partial melting of the measomatized SCLM wedge is considerably later than the subduction period. As such, the thermal structure of some orogens converts from a warm regime in the synorogenic stage to a hot regime in the postorogenic stage.
Because of the complexity of oceanic-continental subduction zones, there is still an inexhaustible set of central questions that await to be answered in the future. An unprioritized 'top 10' list of these questions can be outlined as follows: (i) How is oceanic subduction transformed into continental subduction in collisional orogens? (ii) How much water is carried into subduction zones by nominally anhydrous minerals in the descending slab? (iii) What is the thermal structure of oceanic-continental subduction zones? (iv) What is the geological nature and physicochemical property of subduction-zone fluids at different depths? (v) How does the composition of subducting crustal rocks change due to reaction with evolving fluids as they descend together into the mantle? (vi) How do slab-derived fluids move into and react with the overlying mantle wedge? (vii) How does the precedingly subducted oceanic slab interact with the mantle wedge? (viii) What kind of mantle metasomatites is generated by the fluid-peridotite reaction at the slab-mantle interface in oceanic-continental subduction channels? (ix) How are geochemical signatures transferred from the subducting slab to the mantle wedge? (x) What is the stability of metasomatic minerals at subarc to postarc depths? Robust answers to these questions can greatly advance our understanding of how the subduction factory operates and how the operational mechanism changes with time.