Partial melting, fluid supercriticality

Partial melting,fluid supercriticality and element mobility in ultrahigh-pressure metamorphic rocks during continental collisionYong-Fei Zheng ⁎,Qiong-Xia Xia,Ren-Xu Chen,Xiao-Ying GaoCAS Key Laboratory of Crust –Mantle Materials and Environments,School of Earth and Space Sciences,University of Science and Technology of China,Hefei 230026,Chinaa b s t r a c ta r t i c l e i n f o Article history:Received 14April 2010Accepted 14April 2011Available online 21April 2011Keywords:Continental collision UHP metamorphismDecompression dehydration Partial melting Supercritical fluid Element mobilityPartial melting at continental lithosphere depths plays an important role in generating geochemical variations in igneous rocks.In particular,dehydration melting of ultrahigh-pressure (UHP)metamorphic rocks during continental collision provides a petrological link to intracrustal differentiation with respect to the compositional evolution of continental crust.While island arc magmatism represents one end-member of fluid-induced large-scale melting in the mantle wedge during subduction of the oceanic crust,the partial melting of UHP rocks can be viewed as the other end-member of fluid-induced small-scale anatexis during exhumation of the deeply subducted continental crust.This latter type of melting is also triggered by metamorphic dehydration in response to P –T changes during the continental collision.It results in local occurrences of hydrous melts (even supercritical fluids)as felsic veinlets between boundaries of and multiphase solid inclusions in UHP metamorphic minerals as well as local accumulation of veinlet-like felsic leucosomes in foliated UHP metamorphic rocks and metamorphically grown zircons in orogenic peridotites.Thus,very low-degree melts of UHP rocks provide a window into magmatic processes that operated in continental subduction zones.This article presents a review on available results from experimental petrology concerning the possibility of partial melting under conditions of continental subduction-zone metamorphism,and petrological evidence for the occurrence of dehydration-driven in-situ partial melting in natural UHP rocks during the continental collision.Although the deeply subducted continental crust is characterized by a relative lack of aqueous fluids,the partial melting in UHP rocks commonly takes place during decompression exhumation to result in local in-situ occurrences of felsic melts at small scales.This is caused by the local accumulation of aqueous fluids due to the breakdown of hydrous minerals and the exsolution of structural hydroxyl and molecular water from nominally anhydrous minerals in UHP rocks during the exhumation.The dehydration melting of UHP rocks would not only have bearing on the formation of supercritical fluids during subduction-zone metamorphism,but also contribute to element mobility and ultrapotassic magmatism in continental collision orogens.Therefore,the study of dehydration melting and its effects on element transport in UHP slabs,rocks and minerals is a key to chemical geodynamics of continental subduction zones.©2011Elsevier B.V.All rights reserved.Contents 1.Introduction ..............................................................3432.Experimental possibility of partial melting in UHP metamorphic rocks ...................................3443.Dehydration melting in Dabie –Sulu UHP metamorphic rocks .......................................3473.1.Geological setting ........................................................3473.2.Postulation and observation of dehydration melting ........................................3473.3.Multiphase solid inclusions as a record of dehydration melting ...................................3513.4.K-feldspar ±quartz inclusions as a record of phengite-dehydration melting ............................3514.Dehydration melting in other UHP metamorphic terranes.........................................3524.1.Kokchetav in Kazakhstan .....................................................3524.2.Erzgebirge in central Europe ...................................................3534.3.Western Gneiss Complex in Norway ...............................................3534.4.Western Alps in Italy ......................................................3544.5.Himalaya –Tibet in South Asia ..................... (354)Earth-Science Reviews 107(2011)342–374⁎Corresponding author.Tel.:+865513603554;fax:+865513603554.E-mail address:yfzheng@ (Y.-F.Zheng).0012-8252/$–see front matter ©2011Elsevier B.V.All rights reserved.doi:10.1016/j.earscirev.2011.04.004Contents lists available at ScienceDirectEarth-Science Reviewsj o ur n a l h o m e p a ge :ww w.e l s ev i e r.c o m/l o c a t e /e a rs c i rev5.Partial melting in UHP metamorphic rocks during continental collision (355)6.Partial melting and supercriticalfluid (358)7.Partial melting and element mobility (360)7.1.General arguments (360)7.2.Element partitioning under subduction-zone conditions (361)7.3.Element transport during continental subduction-zone metamorphism (364)7.4.Rutile effect during subduction-zone metamorphism (365)7.5.Summary (367)8.Concluding remarks (367)Acknowledgments (368)References (368)1.IntroductionContinental collision is defined as a series of tectonic processes that encompass subduction and exhumation of the continental crust with regional metamorphism at low-pressure greenschist facies, high-pressure(HP)blueschist and eclogite facies,to ultrahigh-pressure(UHP)conditions in the stabilityfield of coesite and diamond.Lithosatic pressures are assumed to cause mineralogical reactions and phase changes in the crustal interior,resulting in formation of HP to UHP parageneses and index minerals(e.g., Schreyer,1995;Liou et al.,1998;Chopin,2003).The continental collision is associated with intercontinental orogenesis within the framework of plate tectonics theory.Thus,continental collision orogens are categorized into two types,depending on the nature of colliding crustal slabs(Zheng et al.,2009a,b).One is continent–continent collision orogens that are generated by collision between two ancient continents(e.g.,Dabie–Sulu of China,Western Gneiss Region of Norway),and the other is arc–continent collision orogens that are produced by collision between a juvenile arc and an ancient continent(e.g.,Alps of Western Europe,Himalaya of South Asia).The continental collision takes place within given temporal and spatial frameworks,which can be clearly distinguished from post-collisional processes.However,post-collisional magmatism is very common in the two types of continental collision orogens,often erasing various records of collisional orogenesis.A prominent progress in Earth sciences is discoveries of coesite and micro-diamond inclusions in metamorphic minerals of supracrustal rocks(e.g.,Chopin,1984;Smith,1984;Sobolev and Shatsky,1990;Xu et al.,1992).It provides mineralogical evidence for burial of crustal rocks to mantle depths of N120km.This has resulted in recognition of continental deep subduction and UHP metamorphism,leading to dramatic advance in our understanding of the plate tectonics theory.(e.g.,Schreyer,1995;Liou et al.,1998,2009;Chopin,2003;Jahn et al., 2003a;Rumble et al.,2003;Zheng et al.,2003a,2009a;Ernst and Liou, 2008).Some special textures of mineral exsolution were also found in UHP metamorphic rocks(e.g.,Dobrzhinetskaya et al.,1996;van Roermund and Drury,1998;Ye et al.,2000;Ogasawara et al.,2002; Song et al.,2005;Liu et al.,2007a,b),suggesting possible ultradeep metamorphism at depths in excess of200to300km.These deep processes would not only result in phase changes and mineralogical reactions within UHP slabs,but also bring about crust–mantle interactions during the continental collision.UHP metamorphic rocks would commonly evolve under nominally anhydrous conditions and thus were examined mainly in terms of phase relationships among crystalline minerals(Poli and Schmidt, 2002;Zheng,2009).Peak P–T conditions of UHP metamorphism during the continental collision are commonly estimated to be as high as 3.3–4.4GPa at700–950°C.At these conditions,dehydration melting in UHP slabs due to breakdown of hydrous minerals such as mica-and epidote-group minerals is possible(Hermann,2002a; Schmidt and Poli,2003;Patiño Douce,2005;Auzanneau et al.,2006). The generation of hydrous silicate melts by dehydration-driven in-situ partial melting has been increasingly recognized in UHP rocks (e.g.,Labrousse et al.,2002;Massonne,2003;Hwang et al.,2004; Whitney et al.,2004;Perchuk et al.,2005;Korsakov and Hermann, 2006;Lang and Gilotti,2007;Zhao et al.,2007;Xia et al.,2008; Ragozin et al.,2009;Liu et al.,2010).Partial melting of UHP metamorphic rocks can dramatically affect the rheology of deeply subducted crust and thus play a crucial role in accelerating the exhumation of UHP slabs(Hermann et al.,2001;Labrousse et al., 2002;Chopin,2003).Thermal–mechanical models support a link among continental subduction,partial melting,and crustalflow in the overriding continental lithosphere and suggest that partial melting may be a significant process in exhumation of UHP rocks and collisional orogenic evolution in general(Whitney et al.,2009).Partial melting of the lower crustal rocks during HP granulite-facies metamorphism is well established and its genetic links to migmatite and granite have been explored in terms of metamorphic petrology and geochemistry(Clemens,1990;Brown,1994;Villaseca et al.,2001; Brown,2007).Partial melting of thermally mature zones of thickened crust can result in the generation of a layer of low-viscosity rocks,which can significantly affect the rheological behavior of crustal rocks in orogenic belts(e.g.,Arzi,1978;Brown and Solar,1998;Whitney et al., 2003).This would favor mechanical decoupling between the subducting plate and the overlying thickened orogenic crust(Vanderhaeghe and Teyssier,2001;Wallis et al.,2005),density-driven intracrustal differ-entiation at mantle depths(Patiño Douce,2005),and tectonic collapse of collisional orogens(Rey et al.,2001;Vanderhaeghe and Teyssier, 2001;Skjerlie and Patiño Douce,2002).The generation,transport and final fate of crustal melts are controlled by tectonic forces(Vanderhaeghe and Teyssier,2001;Brown,2007),and thus may have different dynamic and temporal relationships with tectonic events during collisional orogenesis(Keay et al.,2001;Whitney et al.,2003).However,it is hardly straightforward to demonstrate explicitly that the incipient melting indeed took place in these UHP rocks because most of UHP metamorphic rocks have experienced extensive retrograde reaction and reequilibration during exhumation.This is particularly so for those UHP rocks that experienced amphibilite-facies overprinting.If aqueousfluids were locally present in felsic UHP lithologies to approach water saturation at appropriate P–T condi-tions,partial melting would take place even under reduced geother-mal gradients(e.g.,Compagnoni and Rolfo,1999;Chopin,2003). Nevertheless,isothermal or heating decompression is generally associated with the initial exhumation of UHP slabs(Carswell and Zhang,1999;Zong et al.,2007;Liu et al.,2010;Xia et al.,2010;Gao et al.,2011;Janak et al.,2011;Zheng et al.,2011).This involves high geothermal gradients and thus is called as the“hot”exhumation in comparison to the continental“cold”subduction.Partial melting can take place during the“hot”exhumation of deeply subducted continental crust from peak UHP depths if sufficient water becomes available from decomposition of hydrous minerals(Hermann,2002a; Patiño Douce,2005;Auzanneau et al.,2006).As soon as the melting occurs in UHP rocks,it transforms water bound in hydrous minerals and nominally anhydrous minerals to aqueousfluids and hydrous melts,which may be also an efficient catalyst for the tectonic exhumation of UHP slices from mantle depths.343Y.-F.Zheng et al./Earth-Science Reviews107(2011)342–374At the pressures and temperatures of the coesite or diamond stability field,silicate –water systems that carry hydrous melts and aqueous fluids are beyond a second critical endpoint to become completely miscible (Stalder et al.,2000;Wyllie and Ryabchikov,2000;Kessel et al.,2005a ).Thus,the primary fluids that were trapped during deep subduction-zone metamorphism can form uniform,highly concentrated,high-density but low-viscosity supercritical fluids that are capable of dissolving many fluid-immobile elements.As a consequence,the partial melting of UHP rocks is a prerequisite to initiate the supercriticality of metamorphic fluid/melt with respect to element transfer in continental subduction zones.In addition,partial melting is always incongruent,and accessory minerals are not always peritectic products.This leads to differential partition of trace elements between fluid,melt and residue containing the accessory minerals.Thus,understanding the conditions and timing of incipient melting in UHP metamorphic rocks in as many individual terranes as possible can provide insights into processes and mechanisms that operated in deep subduction zones during the continental collision.In view of the natural occurrences of partial melting in UHP mineral assemblages,it is probably useful to ask what phases were entrapped in UHP metamorphic minerals?What is the element and isotope composition of the mineral phases in these melting products?Do these lithological records mirror the original composition of the trapped melt?Was the trapped melt representative of that melt in the UHP metamorphic rock?Is the dehydration melting of UHP metamorphic rocks actually common during continental subduction-zone metamor-phism?Are only the conditions required for preservation unusual?How is the partial melting associated with element transport and mass transfer in continental subduction zones?Are there further mesoscale (hand specimen)to macroscale (outcrops)records of partial melting in UHP metamorphic terranes during the continental collision?Can we directly date the products of partial melting relative to their host rocks?How is the partial melting associated with crust –mantle interactions during the continental collision?De finite answers to these questions have been on the way of modern petrology for UHP metamorphic rocks and their associated igneous rocks in continental subduction zones.This review article firstly outlines experimental results for partial melting under HP and UHP conditions.Then natural occurrences of partial melting in UHP metamorphic rocks during the continental collision are summarized from a number of typical UHP terranes on Earth.Extensive discussions are devoted to link of partial melting with fluid super-criticality and element mobility under the P –T conditions of continental subduction-zone metamorphism.2.Experimental possibility of partial melting in UHP metamorphic rocksA lot of experimental studies have been devoted to partial melting of crustal rocks under subduction-zone metamorphic conditions.Inprinciple,there are two series of experiments that were made either in the presence of excess water (e.g.,Hill and Boettcher,1970;Lambert and Wyllie,1972,1974;Huang and Wyllie,1975,1981;Johannes,1984;Carroll and Wyllie,1990;Schmidt,1993;Nichols et al.,1994;Thompson and Ellis,1994;Ryabchikov et al.,1996;Schmidt and Thompson,1996;Holtz et al.,2001;Kessel et al.,2005a,b;Thomsen and Schmidt,2008a,b )or in the presence of hydrous minerals without direct involvement of free water under subsolidus conditions (e.g.,Vielzeuf and Holloway,1988;Rapp et al.,1991;Irifune et al.,1994;Rapp,1995;Rapp and Watson,1995;Stevens et al.,1997;Hermann and Green,2001;Hermann,2002a;Skjerlie and Patiño Douce,2002;Schmidt et al.,2004a;Patiño Douce,2005;Tropper et al.,2005;Auzanneau et al.,2006;Hermann et al.,2006;Hermann and Spandler,2008;Wu et al.,2009).The former is commonly referred to the hydration (water-saturated,wet)melting,whereas the latter is usually called the dehydration (water-unsaturated,nominally dry)melting.The results indicate that the wet melting of both ma fic and felsic rocks can begin at temperatures of 650°C at 1.5GPa to 750°C at 3.0GPa (Fig.1).Sincea considerable quantity of water (0.5to 1.5wt.%)is always present in the hydrous and nominally anhydrous minerals of any UHP lithology (Zheng,2009),their dehydration is capable of producing a local sink with suf ficient amounts of aqueous fluids for very low degrees of partial melting in continental crustal rocks.Many studies of experimental petrology have also demonstrated that there are three distinct types of high-T fluids (i.e.aqueous fluid,hydrous melt and supercritical fluid)under subduction-zone meta-morphic conditions (e.g.,Shen and Keppler,1997;Bureau and Keppler,1999;Stalder et al.,2000;Manning,2004;Hermann et al.,2006;Hack et al.,2007).The wet solidus is thus de fined as reaction curve where silicate rocks are partially melted at given P –T conditions,with transformation of an aqueous fluid to a hydrous melt.For silicate –water systems,with increasing P –T values the silicate becomes partially melted when crossing the wet solidus and completely melted when crossing the liquidus (Fig.2).Below the wet solidus,the hydrous melt does not occur in the systems and thus only the aqueous fluid is active.The hydrous melt starts to occur when the P –T values increase crossing the wet (water-saturated)solidus or the dry (water-unsaturated)solidus.Along the wet solidus,the water content of hydrous granitic melts increases with pressure (Huang and Wyllie,1981).With a further P –T increase to cross the liquidus,the silicate becomes completely melted and thus only the hydrous melt is active.Although both the aqueous fluid and the hydrous melt coexist above the wet solidus,there are immiscible gaps between them.The second critical endpoint is de fined by the intersection of the wet solidus with the line of complete miscibility between the aqueous fluid and the hydrous melt for silicate –water systems (Fig.2).Thus,the silicate-bearing fluid at given P –T conditions exists at elevated pressures that meet the wet solidus at the second critical endpoint (Fig.2),above that the hydrous melt is completely mixed with the aqueous fluid to form the supercritical fluid (e.g.,Shen and Keppler,1997;Bureau and Keppler,1999;Stalder et al.,2000;Manning,2004;Hermann et al.,2006;Hack et al.,2007).The position of the second critical endpoint substantially depends on bulk rock composition,pilation of melting reactions in ma fic rocks and average pelite (revised after Schmidt and Poli,2003).Black lines denote the reactions with similar P –T locations in both bulk compositions,blue lines denotes the ma fic rocks,and red lines denote the pelite.The red and blue areas denote that P –T fields of biotite-dominated and amphibole-dominated melting,respectively.At subduction-zone melting pressures (i.e.,1GPa),the first melt in the pelite appears at the phengite-out reaction and in the ma fic rocks at the zoisite/epidote-out reaction.The bold lines denote the wet solidus,which involves amphibole or biotite in addition to phengite +clionpyroxene +coesite/quartz below 2.5GPa and plagioclase instead of clinopyroxene below 1.5–1.7GPa.Up to 5–6GPa classical melting occurs where a solidus separates a crystal +fluid field from a melt-present field (the fluid having distinctively higher water contents than the melt).A continuous increase of solute in the fluid occurs at high pressures (5.5GPa)and at least the K-bearing phases (phengite or at higher pressures K-hollandite)dissolve in a fluid of continuously evolving composition.Mineral abbreviations:Grt =garnet,Phg =phengite,Amp =amphibole,Zo =zoisite,Bt =biotite,Omp =omphacite,Ab =albite,Qtz =quartz,Coe =coesite,Cpx =clinopyroxene.344Y.-F.Zheng et al./Earth-Science Reviews 107(2011)342–374because both the solidus and the locus of complete miscibility are principally dependent on bulk composition.The solubility of species in the granitic system is likely to be higher than in the pure quartz system.However,in analogy to granitic melts,it is expected that SiO 2accounts for at least 50%of the total amount of dissolved oxides.Close to the second critical endpoint,small changes in pressure or temperature can result in large changes in density (thus solubility),allowing many properties of a supercritical fluid to be “fine-tuned ”between the aqueous fluid and the hydrous melt.The experimental results indicate that ma fic rocks have higher solidus temperatures than felsic rocks in the same contents of water (Fig.2),with only tens of degree differences in the wet solidus between ma fic and felsic rocks at pressures below 1.5GPa.At the pressures above 1.5GPa,however,the wet solidus occurs at about 100to 300°C lower temperatures than the dry solidus,depending on difference in lithochemistry.On the other hand,the P –T region between the wet and dry solidi where partial melting may take place either in the absence of hydrous melts or in the presence of aqueous fluids is of particular importance in continental subduction zones.Indeed,the continental crust provides more favorable conditions than the mantle for interactions with water.In most common rock types (metagranitoids,metapelites,metagreywackes,metabasalts)several hydrous minerals may coexist with an aqueous fluid and/or a hydrous melt even at relatively high temperatures and pressures (Vielzeuf and Schmidt,2001;Green and Adam,2003;Kessel et al.,2005a,b;Hermann et al.,2006;Hermann and Spandler,2008;Klimm et al.,2008).This proliferation of hydrous phases renders partial melting phase relationships particularly complex.Considerable attention has been paid to the effect of temperature on crustal melting,and recent studies also indicate the in fluence of pressure.In the range of 0.5–3.5GPa and 600–900°C,the two series of experiments consistently highlight the importance of biotite or amphibole breakdown associated with garnet-forming reactions in metapelites,metagreywackes and metabasalts for the formation of hydrous silicate melts with increasing pressure (Wyllie and Wolf,1993;Vielzeuf and Montel,1994;Stevens et al.,1997;Vielzeuf and Schmidt,2001;Klimm et al.,2008).Because the incongruent breakdown of biotite produces water-undersaturated granitic melt,partial melting of magnesian metapelites and metagreywackes begins between 780and 830°C at 0.5and 1.0GPa,respectively (Stevens et al.,1997).At higher pressures (N 2GPa),a few experimental melting studies were per-formed crossing the stability of mica or amphibole with changing pressure (Patiño Douce and McCarthy,1998;Hermann,2002a;Schmidt et al.,2004a;Patiño Douce,2005;Auzanneau et al.,2006).The results suggest that the higher the pressure,the lower the solidi.The experimental study of Hermann and Spandler (2008)indicates that the wet solidus of metapelite has a minimum temperature of 700°C at 2.0–2.5GPa and is located between 700and 750°C at 3.5GPa.Partial melting becomes increasingly likely at depths greater than about 100km (P ≈2.7GPa),where the negative slope of the zoisite+K-feldspar dehydration-melting solidus could lead to decompression melting (Patiño Douce,2005).Wu et al.(2009)observed high degrees of partial melting textures in their experiments at 3.5GPa/1000°C,6GPa/1100°C,9GPa/1200°C and 14GPa/1600°C,in which quenched melts were almost completely recrystallized into euhedral or subeuhedral crystals with grain size of several microns;coesite/stishovite and garnet were coexisted with melts as the liquidus phases.These observations suggest that solidus temperatures of the subducted continental crust are controlled by local accumulation of water in UHP metamorphic rocks due to continuous dehydration reactions of hydrous minerals such as phengite,epidote and lawsonite at mantle depths.In addition,the exsolution of molecular water and structural hydroxyl from nominally anhydrous minerals also contribute to the water accumulation (Zheng,2009).Taken together,whether partial melting takes place or not depends on the storage capacity of water in UHP rock systems,which is dictated by the storage capacity of water in individual minerals.Partial melting is expected to occur when the bulk water content of rock systems exceeds the storage capacity of water in the rocks themselves under given P-T conditions.Dehydration by breakdown of hydrous minerals is a key to partial melting of UHP metamorphic rocks.Partial melting due to phengite breakdown is very common (Vielzeuf and Holloway,1988;Hermann,2002a;Schmidt et al.,2004a;Auzanneau et al.,2006).Allanite breakdown is also associated with partial melting of host rocks (Hermann,2002b ).With P –T changes in continental subduction zones,transformation occurs between such hydrous minerals as amphibole and epidote (zoisite)as well as biotite and muscovite (phengite);the beginning of melting in the coexistence of two low-and high-P minerals is about 150to 200°C lower than that in the case of only one mineral (Vielzeuf and Schmidt,2001).In most cases,however,the continental crust is undersaturated with water.This is illustrated by the fact that K-feldspar (metastable with respect to phengite for most UHP conditions)occurs in many UHP metagranites.It is only in highly altered and weathered continental crust that there is considerable fluid release during prograde subduction.Only in very rare cases,there is signi ficant dehydration in continental subduction zones at temperatures N 700°C that could lead to hydration melting in subducting slabs.This is particular so for UHP parametamorphic rocks at elevated geotherms (e.g.,the mid-T/UHP eclogite in the Dabie orogen;Gao et al.,2011).Phengite is thermodynamically stable in a large pressure range,which,plus the kinetic reasons,would not breakdown,but recrystallize,under common UHP conditions.Experimental studies of metapelites suggest that phengite is the major hydrous mineral potentially involved in melting at P N 3.0GPa (Massonne,1999).An experimental study of partial meling in the K 2O –CaO –MgO –Al 2O 3–SiO 2–H 2O (KCMASH)system indicates that dehydration melting of phengite canoccurFig. 2.Temperature –pressure relationships for wet solidi and second critical endpoints (denoted by solid circles)for various rock types in subduction zones (revised after Hack et al.,2007).Melting curves have negative slopes at pressures below 0.5–1.0GPa,indicating that increasing T in the silicate –water systems shifts the systems into the stability field of crystalline rocks whereas increasing P shifts the systems into the stability field of hydrous melts.In contrast,melting curves have positive slopes at pressures above 1.0GPa,indicating that increasing T in the silicate –water systems shifts the systems into the stability field of hydrous melts whereas increasing P shifts the systems into the stability field of crystalline rocks.Dashed lines denote the critical isopleths for different compositions of rock that are extrapolated from experimental data for the granite –water system (Bureau and Keppler,1999)and the basalt –water system (Kessel et al.,2005b;Mibe et al.,2005).By de finition,a critical isopleth starts at pressures higher than the second critical endpoint.Numbers 1to 7refer to wet solidi for:1=granite (Huang and Wyllie,1981;Holtz et al.,2001);2=pelite (Nichols et al.,1994);3=basalt (Mibe et al.,2005);4=basalt (Kessel et al.,2005b );5=pelite (Schmidt et al.,2004a,b );6=peridotite (Green,1973);7=peridotite (Grove et al.,2006).345Y.-F.Zheng et al./Earth-Science Reviews 107(2011)342–374under conditions of about800°C and2.0GPa up to at least1000°C and 5.5GPa(Hermann and Green,2001).According to the experiments of Schmidt et al.(2004a)for a bulk composition typical of mid-ocean ridge basalt(MORB),phengitic muscovite in the assemblage garnet+ omphacite+phengite+coesite+water even starts to melt at about 750°C at 3.5GPa.In phengite-dominated melting,first melts are potassic high-silica granites;at higher pressures(N5GPa),phengite disappears through dissolution instead of melting,producing supercrit-ical K-rich(but Na-poor)fluid(Schmidt et al.,2004a).Experiments of Irifune et al.(1994)indicate that near-solidus melts at6–10GPa have very high K/Na ratios and high SiO2contents.Thus,ultrapotassic melts can be produced by partial melting at the bottom of cratonic lithosphere, with SiO2contents depending on precursor lithochemistry(felsic,mafic or ultramafic).Domanik and Holloway(2000)investigated the stability and phase relations of phengitic muscovite in UHP metapelite at6.5–11GPa and 750–1050°C.The starting material is a natural coesite-bearing UHP metamorphic rock from the Dabie orogen that contains a total of2.1wt.% H2O and6.3wt.%CO2bound in hydrous and carbonate minerals.Phengite reacts out of the phase assemblage with increasing temperature between 900°C and975°C at6.5–8GPa,in association with a significant increase in the K2O and MgO contents of quench material and with the appearance of kyanite in the phase assemblage at T≥975°C.The morphology and texture of the quench phases in these runs are essentially undistinguish-able;unambiguous evidence of melting,such as glass or glassy rims on crystals,was not observed.In general,all their runs performed at P≤8GPa regardless of temperature are characterized by the presence of a dense, solute-filled,fluid phase containing only about60mol%volatile compo-nents.Domanik and Holloway(1996)also observed the lack of a clear distinction between thefluids present at temperatures above and below the phengite“melting”reaction.In those experiments,residual quench phases at P≤8GPa and T≤900°C(i.e.,below the devolatilization temperature of lawsonite,MgMgAl-pumpellyite,and topaz-OH)tended to be composed almost entirely of SiO2and Al2O3,whereas at highertemperatures,they generally resembled the quench phases observed by Domanik and Holloway(2000).Only at the highest temperature (8GPa,1150°C)examined by Domanik and Holloway(1996)was K2O-rich glass observed coexisting with K-bearing hydrous quench phases.Similarly,Ono(1998)reported thatfluids present in runs at 6–15GPa and700–1400°C in a synthetic sedimentary bulk composition contain large amounts of dissolved elements.Over most of the P–T ranges in which Ono(1998)reported,phengite reacts out of the phase assemblage with increasing temperature(i.e.,approximately6GPa and 1300°C to8GPa and1000°C),with occurrence of only quenchedfluid in the high-T run products.Small patches of SiO2-rich glass coexists with quenchedfluid in his one run at6GPa and1400°C.Massonne(1992) reported the dissolution of phlogopite in a KMASH system at P N1.5GPa, accompanied by the production of a dense aqueousfluid phase.In view of the experimental results of Bureau and Keppler(1999)that there is complete miscibility between hydrous silicate melts and aqueousfluids at mantle pressures for a wide range of silicate melt compositions,it is possible that phengite in these systems may be decomposed above its second critical endpoint at these P–T conditions,and therefore,there would be no distinction between hydrous melt and aqueousfluid produced by the phengite decomposition at UHP conditions.During exhumation,on the other hand,phengite would dehydrate with concurrent melting at different P–T conditions(Fig.3and Table1). As experimentally demonstrated by Auzanneau et al.(2006),there is a rapid increase of melt fraction with decreasing pressure at2.5–2.0GPa which corresponds to the transition from a phengite+clinopyroxene to a biotite+plagioclase assemblage.At high pressure(P N2.5GPa),the characteristic phases are phengite and clinopyroxene whereas biotite and plagioclase become stable at low pressure(P b2.0GPa).Therefore, the breakdown of phengite in metagreywackes during decompression allows a melting pulse at the eclogite/amphibolite facies transition and crystallization of biotite and plagioclase.This contrasts to the traditional view that the assemblage of biotite and plagioclase is the product of amphibolite-facies retrogression due to influx of aqueousfluids(e.g., Heinrich,1982).Furthermore,hydrous ultrapotassic melts can be produced in this way,which were subsequently transformed to felsic veinlets composed of granitic minerals such as quartz,feldspar and biotite(Hermann et al.,2006).While melting is possible during the decompression exhumation at eclogite-facies conditions if the rock K2O content is above2wt.%,the water liberated by the breakdown of hydrous minerals is used for amphibolitization under lower P–T conditions if the rock K2O content is very low(Poli,1993;Schmidt,1993).In either case, the phengite dehydration melting of continental crustal rocks under UHP conditions is capable of generating K-rich melts.Because of the difference in the wet solidus between mafic and felsic rocks at the pressures higher than1.5GPa(Fig.2),a contrasting behavior of partial melting would occur between the different compositions of lithology in the local presence of water.Felsic and mafic rocks behave differently because phengite,a major hydrous phase in metapelite and metagreywacke,is stable to pressures as high as about8GPa(Ono,1998)while amphibole and zoisite,the major hydrous minerals in mafic eclogites,become unstable at about2.5and 3.0GPa,respectively,without being replaced by the other hydrous minerals at near solidus conditions(Vielzeuf and Schmidt,2001).The consequence is that mafic rocks may lose their H2O component during a prograde path to become nominally anhydrous eclogites(+free fluid)whereas metapelites and metagreywackes may still maintain H2O.At pressures below2.5GPa,both metapelites and metabasalts contain the pair of hydrous minerals with the similar water contents. In both cases,thefirst melts are produced through the volumetrically less-important phase,i.e.,phengite in metapelites and epidote/zoisite in metabasalts.These melts are dacitic and probably amount to less than10vol.%.The melt productivity depends on the amountof Fig.3.P–T paths and dehydration-driven in-situ partial melting of felsic to mafic UHP metamorphic rocks in the Dabie–Sulu orogenic belt.UHP and HP metamorphic domains are separated by the P–T boundary for the quartz/coesite equilibrium.The P–T paths of Dabie–Sulu UHP rocks in low-T/UHP,mid-T/UHP and high-T/UHP zones are constructed,respectively,following the thermobarometric estimates of Zheng et al. (2011),Gao et al.(2011)and Xu et al.(2006).The UHP metamorphic rocks are characterized by convex-type P–T paths of exhumation,so that dehydration melting due to phengite breakdown during exhumation of the deeply subducted continental crust from mantle depths would take place at points A,B and C following the experimental data in Table1.Wet solidus for the system granite+H2O is after Huang and Wyllie(1981)and Holtz et al.(2001),and that for the system basalt+H2O is after MIbe et al.(2005).The position of the second critical endpoint forfluid supercriticality in the system granite+H2O is after Hermann et al.(2006)and that in the system basalt+H2O is after Mibe et al.(2005).346Y.-F.Zheng et al./Earth-Science Reviews107(2011)342–374。