Earth's Mantle Melting in the Presence of C-O-H-Bearing Fluid

2 Earth's Mantle Melting in thePresence of C-O-H-Bearing Fluid

KONSTANTIN D. LITASOV,',' ANTON SHATSKIY,1,2AND EIJI OHTANI1

lDepafiment of Earth and Planetary Matefials Science, Graduate School of Science,

zv s sobotev rn,,u"*Ti!!"i"yi'J,ifi[,;!fi,!r::;,;;";^r,Novosrbjrsk, Russjc

Summary

Recent experimental data on phase trans{orma-tions and melting in peridotite and eclogite sys-tems with a C-O-H fluid at pressures up to about30 GPa are reviewed with speclal attention to theeffect of redox conditions. The fundamental dif-ferences for partial melting in systems with HrO,CO, and reduced C-O-H fluid (CHo-HzO-Hz)are outlined. Melting in systems with HrO de-pends mainly on the total HrO content and iscontrolled by HrO solubility in nominally anhy-drous minerals. Partial melting occurs when thetotal HrO content of the system exceeds the HrOstorage capacity in the minerals oi the rock undergiven P-T-X-/O, conditions. Melting in systemswith COt is determined by carbonate stabilityand is strongly affected by the alkali (particu-1arly, KrO) content in the system. The onset ofmelting is relatively insensitive to the total CO,content. Studies of peridotite and eclogite sys-tems saturated with HrO and CO, show thatHrO -bearing phases, such as dense hydrous sil-icates, superhydrous phase B and phase D, cancontrol initial melting and iow degree partialmelts are silica-rich. Moreover, most solidi areflattening out at pressures above 6-8 GPa. Thesoiidi of peridotite and eclogite with coexistingreduced C-O-H fluid {presumably CHo * H2O, atthe oxygen fugacity near the Fe-FeO buffer) are 1o-

cated 300-500'C above the solidi of the svstems

with HrO and CO, at 15GPa. At the same rimethey are still 300-400oC lower than volatile-freesolidi. Thus, we provide a first calibration of thedependence o{ mantle melting at constant pres-sure on redox conditions. The stability boundaryof Fe-Ni alloy, which may coincide with thelithosphere-asthenosphere boundary under cra-tons (200-250 km), the 410 km discontinuity andthe transition zone itself, may be paramount to re-dox and decarbonation-dehydration melting andfreezing. Subducted carbonates rather than watermay control melting in the "big mantle wedge"model for stagnant slabs. We proposed a phe-nomenological model for the segregation of aslab-derived carbonated melt in the transitionzone, which can move as a diapir through thetransition zone and upper mantle and initiatemagmatism at the surface.

2.I Introduction

In a widely accepted pyrolitic model, the uppermantle consists of peridotite with -50-60%olivine, 20-30% pyroxenes/ and L5-25% garnet(e.g., Ringwood, 1979). Subducted oceanic platesare the major source of heterogeneities in thepresent-day mantle. They supply the mantlewith terrigenous sediments, altered basalts andvariably depleted peridotite. A significant partof this material, especially incompatible and

Physics and Chemistry of the Deep Earth,First Edition. Edited by Shun-ichiro KaratoO 2013 fohn Wiley & Sons, Ltd. Published 2013 by fohn Wiley & Sons, Ltd.

Earth's Mantle Mehing in the Presence of C-O-H-Bearing Flttid 39

volatile components, is involved in melting,island-arc volcanism, and the chemical alterationof the mantle wedge above subducting plates. Inturn, mantle wedge material becomes entrainedin mantle convection. Incompatible and volatilecomponents can recycle and escape from themantle in mid-ocean ridges and hotspots as wellas during continental magmatism. Indeed, thesematerials are also involved in the formation o{mantle heterogeneities in the form of depletedperidotites, pyroxenites and eclogites producedby deep-ievel melt fractionation and reactionswith the peridotites (Hofmann & White, 1982,Sobolev et al., 2007). A potential source ofmantle heterogeneiiy is the Earth's core, butthe scale of core-mantle exchange remainsuncertain. According to theoretical modeling,mantle heterogeneities can persist over geologicaltime, that is, for billions of years. They can bepreserved both in the rigid lithosphere and in theconvecting mantle. Evidences for lithosphericheterogeneities come {rom ancient pyroxenitexenoliths in a therzolite matrix. Heterogeneitiesin the convecting mantle can be detectedby subduction-related seismic anomalies anddifferent seismic reflectors (e.g., Stixrude, 2007).

Mantle heterogeneities are closely relatedto the distribution of volatile components andvolatile cycles in the Earth's interior. Deter-mining the role of volatiles, especially in theC-O-H-S system/ in deep mantle processes isone of the key issues in modern mantle petrology.One of the major tools for solving these issues isthe experimental studies at high pressures. Theabundance and behavior of hydrogen, carb{rn,and their compounds in the mantle have beenwidely discussed in a number of papers, whichare based on experimental studies o{ meltingphase relations in systems with volatiles atpressures up to 6-10GPa and some limiteddata for higher pressures {e.g., Poli & Schmidt,2002; Dasgupta et a1., 20O4; Kawamoto, 2004;Dasgupta & Hirschmann, 2006i Dasgupta &Hirschmann, 2010). Other important problemsrelated to the behavior of volatiles in the mantleare hydrogen and carbon solubiiity in silicates,which were studied at pressures up to 25*27 GPa

(Bolfan-Casanova, 2005; Shcheka et a1., 2005iLitasov & Ohtani, 2007; Hirschmann et a1.,

2009), the influence o{ hydrogen and carbonon the rheological and transport properties ofsilicates (Mei & Kohlstedt, 2OOOa,h; Karato &|ung, 2003; Hayden & Watson, 2008; Yoshinoet a1., 20101, and the {ormation of diamond influid-bearing systems {Palyanov et a1., 2005iPalyanov & Sokoi, 2009).

In recent years we have published severalstudies on volatile-bearing systems at pressuresof up to 20-30 GPa (Litasov & Ohtani, 2002;Litasov & Ohtani, 2005; Ghosh et a1., 2009iLitasov & Ohtani, 2009a,2010; Litasov, 2011).In this chapter, we analyse these and otherexperimental data on peridotite and eclogitesolidi and on melting in the presence of C-O-Hvolatiles. We consider the systems with HrO,CO2, H2O + CO2, as well as those with a reducedC-O-H fluid (presumably CHo + H2O). On thebasis of this analysis, the importance of eclogitesin mantle rnelting in the presence o{ volatiles isemphasized along with the key role of carbonatesin the deep-leve1 melting o{ subducted plates.

2.2 High-Pressure Experimental Techiquesfor Fluid-Bearing Systems

Simple systems with volatiles can be studiedin conventional high pressure experiments.Descriptions of the methods can be found in(Litasov & Ohtani, 2002; Litasov & Ohtani,2009a,bi Shatskiy et a1.,2011). However, buffer-ing techniques for controlling redox conditionsare required in many systems. Unbufferedexperiments, for example with HrO or COr, yieldlOr conditions, which are determined eitherby starting composition (i.e. mineral equilibriawith a fluid phase) or by the /O, imposed bythe cell assemblage (e.g., BN, LaCrO., graphite).Hydrogen may migrate in or out of the samplecapsule, which is particularly dangerous forsystems with CO, (Brooker et a1., 1998).In orderto protect sample from undesirable hydrogengain or loss different techniques and capsulematerials can be used. Among different metal

+, KONSTANTIN D. LITASOV, ANTON SHATSKIY AND EI]I OHTANI

Fig, 2..l Ce11 assembly for doublecapsule experiments at 12-16 GPa

capsules Au-Pd ailoy is considered to be leastpermeable for hydrogen {Nishihara et al., 2005).To reduce hydrogen diffusion into the samplecapsule a hematite {Fe2O3J sleeve can be used inpiston-cylinder or multianvil experiments {Liuet al.,2006).

Buffering techniques control f O, in the sampleduring experiment typically using metal-metaloxide pairs and can be considered as most suitablefor studies of volatile-bearing systems. In studiesof systems wlth reduced C-O-H fluid we usedmodified double capsule methods (Sokol et a1.,

2009, Sokol et a1., 2010, Litasov et a1., 2013b),where the thick-walled outer capsule was made ofbu{fer materiai (Molybdenum or Iron). In this casethe sample was placed into an Au-Pd or Pt capsuleand separated from the outer capsule talc, whichtransforms to enstatite and HrO-fluid or HrO-bearing silicate melt upon heating {Figure 2.1).

2.3 Temperature Profiles and OxidationState in the l\{antle

The temperature proffles in the mantle and redoxconditions largely control melting in the presenceof the voiatiles. The temperature profiles havelarge uncertainties even for average mantle due touncertainty in the estimated mantle compositionand parameters for equations of state for mantlephases. The thermai regimes of lithosphere andasthenosphere are dl{{erent. In the lithosphereheat is transported by conduction and temper-ature increases rapidly with depth, whereas the

::-- -^L, l 4rv-

W Lacro:

ro i)

f ]l l"tco e\* Oute; aal:uie -W bufferili-lil -errt- ,l '*"re lnr,er cacsuieK w;Ih sJrir:e

4mmw

temperature profiles in the asthenosphere corre-spond to nearly adiabatic conditions of convectingmantle. The depth of the transition from mainlyconductive to convective thermal regime variesfrom 10-20km beneath Mid-Ocean ridges to>250 km beneath the ancient cratonic areas {e.g.,Priestley & McKenzie, 2006). Figure 2.2 showsa series of geotherms constrained by mantlexenoiiths from continental basaits (South-EastAustraliai and kimberlites of the Udachnaya pipe(Sitrerla) and /ericho pipe (Canada).

The temperature in subducting oceanic platesis significantly lower than in the surroundlngmantle. The coldest paft m^y correspond tothe slab Moho {5-7km into the slab or siightlydeeper). The PT-profi1es of hottest modernslabs may be similar to cratonic geotherms

{Figure 2.2). Tlne PT-profiies o{ ancient subduc-tion might be significantly hotter than thoseo{ modern subduction. The PT-estimationsfor Precambrian metamorphic rocks related tosubduction environments indicate that theycorrespond to modern continentai rifts, whereasmost estimations for Phanerozoic rocks are con-sistent with modern hot subduction (Figure 2.2).Model PT-profi1es used {or comparison withvolatiie-bearing solidi of mantle rocks atesummarized in Figure 2.3. The average mantleadiabat corresponds to a potential temperatureof 1315'C + 100 - accordlng to thermobarom-etry and thermoelastic properties of mantlephases {McKenzie & Bickle, 1988, Katsuraet a1., 2004i Putirka, 2005). The PT-proffles of

Earth's Mantle Mehing in the Presence of C-O-H-Bearing Fluid 4L

t)cl3t[c)

g.oF

nepth {km)

0 50 100 150 200 250

1 200

1 000

ann

OUU

400

200

0

012345678Pressure {GPai

tig.2.2 PT-proffles in the shallow mantle. Mantlegeotherms based on xenoliths in continental basalts ofSoutheast Australia iO'Reilly & Griifin, 1985) and inkimberlites o{ the Udachnaya pipe, Siberia (Boyd et a1.,

1997) and the |ericho pipe, Canada {Kopylova et a/.,7999) are shown as well as a MORB adiabat withpotential temperature of 1315'C constrained usrng a

model o{ McKenzie et aL. 12005). The PT-profiles oisubduction zones under Northeast and Southwestlapan are shown after Peacock and Wang (1999).

Symbols show PT-data for subduction-related rocks inophiolite and high-pressure complexes of different age(in Ma) iMaruyama & Liou, 2005). The gray fie1d showsthe maximurn pressure and temperature o{ ultra-highpressure metamorphism in the Kokchetav complex(Korsakov & Hermann, 2006). The gray line is thesolidus o{ wet basalt (Poli & Schmidt, 2002). Thegraphite-diamond phase boundary is alter Kennedy andKennedy (1976).

modern subduction slabs are based on resentestimates by Syracuse et al. 12010). Examples ofcold subducting slabs are Tonga and NE |apan.Representative hot subduction zones includeSW |apan, and North and South American slabs

{van Keken et a1., 2002; Syracuse et al., 2010).

The temperature proffles of subducting slabs are

difficult to constrain at depth below 250-300 km,especially if the slabs are stagnant in the

transition zone. Empirical profl1e s oi cold and hotsubducting slabs, stagnant in the transition :oneand penetrating into lower mantle are sho\rn inFigure 2.3.

The oxidation state o{ the mantle is discussedin detail by Frost & McCammon (2008). Nlcstestimates of f 02 for basaltic rocks and shaliotmantle peridotites plot close to the FMQ {fayalite-magnetite-quartz) oxygen buffer. However, datafor garnet peridotite frorn kimberlite indicatea well-pronounced decrease in f O). with depth(Woodland 8l Koch, 2003). These data along withexperimental studies {Frost et al.,20O4; Rohrbachet aL.,2007) showed that the Fe3+ contents of sili-cates increased wj.th pressure even at ecluilibriumwith Fe-rneta1. This increase is due to the sta-bilization of Fe3+ in some high-pressure phases/particularly in Mg-perovskite. At -8 GPa, thecurve o{ the average f 02 fuom mantle peridotlteswould cross the stability line of Fe-Ni metal{Figure 2.4). ln the presence of a smal1 amountof this al1oy (-0.1 wt o/o rn the upper mantle at10-14GPa and -1.0 wt % in the iower mantle,Frost at aI., 2004i Rohrbach et a1., 20071, the sys-tem would be buffered near the IW (Iron-wustite)equilibrium {or 1-2 log units belora. this buffer)and the average fO2 corresponds to the bold curvein Figure 2.4. As yet, it is di{ficult to accurateiydetermine /O2 below the intersection with thestability line of the Fe-Nl alloy, because it de-pends not only on the disproportionation o{ Fe2+in silicates, but also on the original heterogeneityln oxygen distribution in the bulk Earth. Oxi-dized material of subducting slabs (and of themantle wedge entrained to convective mixing)may significantly affect the redox state and melt-ing regime of the upper and maybe lower mantle.The consequences of this processes is discussedin detail in the next secrions.

2.4 An Outline of Experimental Studies atPressures below 6*7 GPa

Most studies of peridotite and eclclgite systetrswith a C-O-H fluld were carried out at pressuresbelow 6-7 GPa, where most basaltic meits and

= lJlJ Lr

ffilpturyrntirvt"'

I 800

1 600

1400

1 200

1 000

800

OUU

zvu

the deepest kimberlite and lamproite meltsare generated (Wyllie & Huang, 1976; Wyllte,1978, I987j Green & Wallace, 1988; Wallace& Creen, 1988; Green & Falloon, 1998; Wyllie& Ryabchikov, 2000). Major melting patternsand melt compositions were determined in thesystem peridotite-H2O-COz at pressures ofup to 4-5 GPa. The key obsewations include1i) carbonatite melts form at the peridotite solidus

KONSTANTIN D. LITASOV/ ANTON SHATSKIY AND EIII OHTANI

Depth {km)400 500 600

'1800

=3

1500 Nilc

1400 0

o

fas

oE,o

15 2B

Pressure {GPa)

Fig' 2.3 Mantle PT'profiles to 900 km depth. The gray ffeld shows the range of mantle adiabats with potentialtemperatures of 13 i 5- 141 5'C. The dashed line K-04 shows the 1400'C adiabat after Katsura et a\. 12004). Thedotted line SD-08 correspond to an average mantle thermal model after Stacey and Davis (2008). Riit (oceanic) andCraton geotherms are shown after data in Figure 2.2. Numbers show PT-proffles of hottest (1) medium (2) andcoldest {3) subduction slabs stagnant in the transition zone and penetrating into the lower mantle (2a and 3a) basedon estimates of van Keken et al.12002) and Syracuse et al.12010) for depths of 50 and 250km {shown by barscorresponding to slab sur{ace and slab Moho 1evels). K-parameters {or peak metamorphism in the Kokchetavmassif (see Figure 2.2). Phase boundaries and solidus of dry peridotite are shown after Litasov and Ohtani (2007).

1 200

1000

5UU

bUU

in the amphibole-phlogopite stability field, and(ii) decarbonation reactions of solid or liquidcarbonates with silicates result in the extractionof CO, (Figure 2.5). Green & Falloon (1998)argued that all types of mantle magma, frombasalts to kimberlites and carbonatites may beproduced by melting of peridotite-HrO-CO, atpressures of 2-7 GPa. Based on studies of the sys-tems peridotite-CO, and eclogite-Coz to 9 Gpa

,.,itn'-o\t*.",.l::Pr

,lJJ

$

@'L-*-)Q9

Wd {Rw)* Gttransition eone

Carrofiatite ! Magreslfe

Earth's Mantle Melting in the Presence of C-O-H-Bearing Fluid

Depth (km)0 25 50 75 100 12s

()

5i:0Id)F

oo

Pre$sure {€Pa}

Fig.2.4 Oxygen fugacity lf O"; relative to the IWbufferJ calcr,rlated for garnet peridotite along theT-proffle of a cratonic geotherm as a Iunction ofpressure. The curve {or Fe-Ni alloys was calculated forperidotite using ecluations in Frost and McCammon(20081. The data for garnet peridotites from theKaapvaal Craton (South Africa) is after Woodland andKoch (2003). The positions of the buffers FMQ (Fayalite+ 02 : Magnetite+Quartz) ar'd EMOC/D(En + Mst : Ol + Graphite/Diamond + O2), are alerKadik {2003), Stagno and Frost (2010).

(Dasgupta et aL.,2004i Dasgupta & Hirschmann,2006; Dasgupta & Hirschmann, 2010) modelsfor eclogite and peridotite melting beneathmid-ocean ridges were developed. Melting occurswhen smal1 portions of carbonatite melt risefrom deep levels and cause further melting uponascent with the participation of HrO.

Systems with a reduced C-O-H fluid werestudied experimentally at pressures of up to6-7 GPa. A peridotite system with a C-O-Hfluid bu{fered at WCWO (WC-W-WO2 : IW + 1)

was studied by Taylor and Green {1988) at0.9-3.6GPa. According to Taylor and Green(1988), the CH4/{CH4+H2O) ratio of the fluidmeasured by mass spectrometry decreases from0.8 to 0.3 at 1050-I250"C and 1.5-3.6GPa.The solidus temperature in a system with a

reduced fluid {CH4 > H2O) was 200-300"Chigher than in systems with HrO and CO, at

01234Pressure {GPa)

Fig. 2.5 Position of solidi in volatile-{ree andvolatile-bearing Hawaiian pyrolite a{ter Green andFalloon (1998). The decarbonation reactionFo + Di + CO2 : En + Dol is shown {or theCOr-bearing pyrolite. Reduced solidus is {or theHawaiian pyrolite equilibrated with a COo-HrO fluidatfO2: Iw+ 1. Sp1, spinel; Grt, garnet; Parg,pargasite; Fo, forsterite; Di, diopside; En, enstatite; Do1,

doiomite; Lz - Iherzolite. Reproduced with permissionof Cambridge University Press.

3.6 GPa (Figure 2.51. At the same time, fakobsson& Holloway (2008) studied peridotite with a

reduced C-O-H fluid buffered by IW at 5-12 GPaand observed melting at 1200*1250'C, whichis substantialiy below the solidus o{ the systemperidotite-C-O-H fluid at IW+1 (Figure 2.5).The fluid compositions measured in theseand other experiments (Taylor & Green, 1988;

|akobsson & Oskarsson, I990i Matveev et a1.,

1997; Sokol et al.,2004i Sokol et al.,20091 some-times agrees with the model fluid compositionscalculated from the equations of state for realgases (Saxena & Fei, 1987; Zhang & Duan, 2009)but sometimes not. For example, high hydrogenand 1ow CHo contents o{ quenching melt, whichdo not agree with calculations, are reported bySokol et aL.12009).

Hawaiian pyrolite + COH .,{Oot, /od-

44 KONSTANTTN D. lITASov, ANTON SHATSKIv AND EIII OHTANI

2.5 Diamond Formation in Fluid-BearingSystems

Some conclusions about fluid regime in the uppermantle were derived from laboratory studies ofdiamond cry stalhz ation in fl uid-b earing systems(Palyanov et a1.,2005i Palyanov & Sokol, 2009).The PT conditions of growth of lithosphericdiamonds {900-1600"C and 5-7 GPa) are muchlower than expected for the direct graphite-to-diamond transition (>1500"C and >13 GPa).This suggests the presence of a liquid/fluid cat-alyst for diamond formation in the lithosphericmantle. The solidus o{ volatile-free silicatemantle is located at much higher temperatures/and in addition silicate meit does not cataTyzediamond nucleation and growth at 5-8 GPa and1400-1600'C, because carbon solubility in sucha melt is negligibly low {Palyanov et a1., 2005).Only fluid or melt containing COr-HrOtalkalicarbonates r.rtay c ataly ze diamond crystallizationat these conditions lowest temperature condi-tions {Palyanov et a7., 1999, 2002) (Table 2.1).There{ore, natural diamonds may have crystal-lized {rom carbon-oversaturated fluid-bearingmedia at the PT-conditions of cratonic roots orthe shallow asthenosphere (Palyanov et al.,2O05).Carbon saturation could be realized by partialreduction o{ carbonatite meit by surroundinglithospheric mantle. Another possibility is oxi-dation of methane-rich fluid. However, it shouldbe emphasized that the effectiveness of diamonderystallization in reduced fluid is significantly

lower. Minimum PT conditions required {ordiamond nucleation exceed the PT range oflithospheric diamond {ormation (Table 2.1).

2.6 Melting Phase Relations in Peridotite andEclogite Systems at Pressures to 20-30 GPa

2.6.1 Systems withHrO

Phase relationships in peridotite-HrO systemsat pressures to 25-30GPa have been consideredin detail (Litasov & Ohtani, 2002; Kawamoto,2004). H2o-bearing eclogite systems have alsobeen studied (Schmidt & Poii, 1998; Litasovet aJ., 2004; Okamoto 8r. Maruyanla, 2004;Litasov & Ohtani, 2005). In these studies, thestability of dense hydrous magnesium silicatesin peridotite was determined (Figure 2.6). Theseresults and data on the HrO solubility or storagecapacity in mantle minerals {Bolfan-Casanovaet al., 2000; Bolfan-Casanova, 2005; Litasovet aL., 20O7 i Litasov & Ohtani, 2007; Mookherjee& Karato, 2010; Litasov et a1.,201lb) allow anempirical diagram to be constrained for the solidiof HrO-undersaturated peridotite, depending onthe HrO content of the system (Figure 2.6). Thesolidus position depends on the HrO storagecapacity of bulk peridotite. The storage capacityin this chapter is defined as the maximumHrO content of nominally anhydrous phasesat given P, T, X, and fO2 {Hirschmann et a1.,

2005). According to some preliminary data forreduced conditions and in equilibrium with

Table 2.1 Minimum pressures and temperatures of diamond nucleation and growth established in kineticexperiments.

System ReferenceP,

GPa

T,"C

Nucleation Growtht ^ur,hou rs

(/NarCOr-H, O-CO, -C

co2-H2o-cH20 /CO2-Cct4-Hto-caH2 0-ct s

H36 -02 -caS_C

Fe-Ni -S-C

5.7

5.7

5.7

5.1

6,3

6.3

6.3

120

r36t3s136

48

20

65

1 1s0

I 200

1 300

>1420>,]600

17 50

>1450

I 150

I 200

1200/1300r 300

>'l600I 650

1 450

(Pal'yanov et al., 1 9991

(Sokol et al., 2001)(Sokol ef a/,, 200 1 )

(Sokol ef a/,, 2001 )

(Sokol et a|.,2009)(Palyanov et q1.,2009)

(Palyanov et a|.,2006)

'produces reduced CHo-bearing fluid at experimental conditions.

Earth's Mantle Mehing jn the Presence of C-O-H*Bearing Fhtid

Depih {kn:)400 500 600

0s101524253035e'essure iGPa)

Fig.2.6 Solidiofperidotitewithdi{ferentHrOcontentsinthesystem.Thesolidiineindicatesthestabilityo{dense hydrous magnesium silicate phases and the storage capacity iin wt % H.O) at the solidus of the system.Dashed lines indicate the solidi in systerls with lower HrO contents (wt %). Gray lines indicate the major phasetransitions and mantle PT-profiles from Figure 2.3. 01, Olivine; Wd, wadsleyite; Rw, ringwoodite; Mpv,Mg-perovskrte; Fp, ferropericlase; Amp - amphibole; Serp serpentine. iConstrained using data from Ohtani et al.,2000; Litasov et a1.,2003,2011b; Litasov 8r. Ohtani, 20A7,2008).

45

1 800

C)

g

fs

lt

,r',nn$':f,U

lU,r(, X

{:a

-

CO2-bearing fluid/meit with 1ow HrO-activityolivine and wadsleyite may contain up to 10

times less HrO than in the system with HrC)only (Litasov et a1., 2009i Sokol et a1., 2U.0J.The parameterization and systematic study ofthese dependencies are particularly importantfor accurate modeling of HrO-bearing solidi andmelting in the deep ilantle (see also Keppler,Chapter 1, this volume). It should be noted thatunder supercritical conditions, there is no solidusin the systelrr. Under these conditions, a fluidalways coexists with solid silicates and the com-position of the fluid becomes rnore sllicate-richwith temperature, until it reaches a compositionsimilar to a silicate melt. They may, however, be

a "practical solidus," if the silicate content of thefluid changes from a "fluid-like" to a "melt-like"regime over a narrow temperature interval.

According to Figure 2.(t, the transition zoneaccumulates HrO even i{ it is only a minor ortrace component in the pyrolite mantle. At the410km and 660km discontinuities, the solidustemperature in HrO-bearing systems decreasesdramatically, and this may cause melting{Litasov & Ohtani, 2002). I{ peridotite contains0.1 wt % H)O (which is reasonable for thetransition zone), the solidus temperature at the410krn discontinuity drops by 200-250'C inthe olivine stability fie1d and at the 660 km dropsby approximately 600"C in the Mg-perovskite

'i,*--./{i l

[,$ffiis'*tudilnri'

iO\ST.\\TI\ D. IITASOV/ ANTON SHATSKIY AND EIII OHTAN]

:::..rir,:-, neld There is also major pressure:.::lience of solidus temperatures inside the:---.rne stabiiity fie1d at 10-14Gpa and with, -l 0.5 rvt % HrO in the system (Figure 2.6).The maximum HrO content in rnantle olivinesfrom kirnberlitic xenoliths (0.02_0.04 wt yo; e.g.-\latsyuk &. Langer, 2004) remains within theolivine storage capacity at 6Gpa and 1300'C,rvhich is 0.07-0.10 wto/o HrO. The maximumH. O content of wadsleyite and ringwoodite in thetransition zone is 0.4-0.5 wt % along the mantleadiabat {Litasov et a1., 2071b). Considering thepresence of other rninerals lup to 40% garnet andpyroxenes with up to 0.1 wt % H2O), the transi_tion zone may store up to 0.3-0.3 5 wt To H, O ataverage man tle relnperarure.

The dependence of the HrO solubility onpressure and temperature in eclogite minerals isloglly studied. At pressures above the stabilityfield of phlogopite, lawsonite, and phengite,HrO can accumulate in accessory richterite,phlogopite, or nominally anhydrous phases.Some studies recorded the appearance of theK-bearing phase X at pressures above I5Gpawhen K-richterite disintegrated; this phase couldcontain up to 1.5 wt o/o HrO and it may occurboth in eclogitic and peridotitic assemblages ifthe system has an elevated K content (Konzett& Fei, 2000). Aluminous stishovite may alsobe a major reservoir for HrO in eclogite atpressures above 20 Gpa as it can contain upto 0.3 wt % HzO (Litasov et a1., 2007). pyropeand majorite garnet can contain up to 0.1 and0.13 wt ok HzO, respectively lKatayama et aJ.,2003; Mookherjee & Karato, 2010). Accordingto the available data, the HrO solubility inNa-clinopyroxene is lower than that in garnet.In the experiments conducted at 600_700"CiBromiley 8r. Keppler, 20041, it decreases {rorn,170ppm at 2 GPa to 100ppm at 10 Gpa.

The phase boundaries between olivine and wad_sleyite and ringwoodite and Mg-perovskite +periclase shi{t toward low and high pressures,respectively/ in the presence of HrO (Litasovet a1., 2005i Litasov et a1., 2006; Frost &. Dolejs,2007) due to the different H2O solubility in theminerals. These data can be used to estimate

rnantle water contents from the depths of the410km and 660km discontinuities, especially inthe regions close to subduction zones.

2.6.2 Systems with CO,

The peridotite-CO, and eclogite_CO2 systemswere studied in the simpliffed (CaO_MgO_A1203-SiO2NarO-COr) (Litasov & Ohtani,2009a; Keshav & Gudfinnsson, 2010; Litasov &Ohtani, 2010) and in complex, close to naturalcompositions {Ghosh et al., 2009i Kiseeva et aI.,2013) at pressures of up to 27_B2Gpa, mainlyalong the solidus temperatures. The solidi inthese systems are subparallel or have verygentle PT-slopes above 10 Gpa. It was found thatNarO and KrO play a key role in the meltingof carbonate-bearing peridotite ancl eclogite.The addition of 0.1 wt % KrO reduces thesolidus temperature by 500 C it 20Gpa in bothsystems. In addition to the solidus curves, thethermal stability o{ carbonate phases (magnesiteand aragonite) is important to constrain thebehavior of carbon in the mantle. Magnesite andaragonite stability in dif{erent systems, includingthose with H2O+CO, (Litasov et al.,20l1alis shown in Figure 2.7. There are rhree maiorregimes of magnesite stability under oxidizedconditions: {i) rnagnesite-bearing silicate systemswithout free silica phase, {ii) magnesite +SiO,assemblages (for example in eclogite), (iii) ftrO,saturated magnesite-bearing systems. In the ffrstregime magnesite stability is iimited by decarbon_ation and melting reactions involving silicates,

3uch- as MgCO. + MgSiO3 : MgzSiO+ + CO2.In the second regime, magnesite stability iscontrolled by melting reactions with ,ili.,phases MgCO, + SiO2 : MgSiOs + CO2. Theaddition o{ H2O to the system causes the thermalstability limit of magnesite to fa1l dramatically tothe level of the solidi of KrO-containing systems.The stability lines plot parallel to the pressureaxis. The reason for such a drastic reductionin magnesite stability is pooriy understood atpresent. In general, the thermal stability limit ofmagnesite in eclogite is lower than in peridorite(Litasov et aI., 20lLa).

Earth's Mantle Mehing in the presence of C-O-H-Bearing Fluid

Depth {km)300 400 500

47

2000

1800

1600

1400

1200

1000

800

600

In most peridotite and eclogite systems,magnesite is the only carbonate phase at pres-sures above 6-7 GPa. F{owever, in some eclogitesystems with an elevated CaO content, thestable phase is aragonite {Shirasaka & Takahashi,2003; Kiseeva et a7., 2013). The decarbonationreaction CaCO, + SiO2 - CaSiO, + CO2 occursat temperatures close to slightly above the mantleadiabat. In eclogite-CO, systems, aragonite isstable at temperatures of 1200'C and below {ata pressure below TGPa; Figure 2.7). Magnesite

decarbonation is replaced by melting at pressuresabove 6 GPa with the formation of a carbonatitemelt or a carbonate-bearing silicate melt. Thisboundary lies at - 10 GPa for aragonite.

2.6.3 AIkaIi carbonatite melting

Near-solidus melts of Na- and K-bearing car-bonated peridotite and eclogite systems, showstrong enrichments in alkalis. However, the pre-cise determination of the compositions of thesemelts is difffcult due to their very low modal

o!5!oo-Eot-

0 5 10_ 15 2A 25 30Pressure (GPa)

Fig. 2'7 Magnesite and aragonite stability in different carbonate-bearing systems. Dashed lines indicate magnesitedecarbonation and melting, according to Irving and Wyllie (1975); Katsura and Ito (1990). The magnesite-SiO, lineshows magnesite stabiiity limits for systems containing a free silica phase and roughly

"orr.rponi, to the reaction

MgCO, + SiO2 : MgSiO, f COr-bearing melt (Dasgupta & Hirschmann,2006l Litasov & Ohtani, 2009a,2010).The stability of magnesite in the peridotite and eclogite systems with H2o-Co, and the Dol : A4gs f Arg line areafter Litasov (201 1); Litasov et al. 12011a). A" level of decarbonation ,"r"iion, near 2-2.5 Gpa (see Figure 2.5 ). Graylines indicate average mantle and stagnant slab PT-profiles {Figure 2.3). The stability of aragonite for"eclogiticassemblages is shown after Kiseeva et al.12013) (coincides with peridotite Hro-Co, line) and Shirasaka andTakahashi (2003J (dotted line). The decarbonation reaction of aragonite roughiy coincides with themagnesite * SiO, 1ine.

Man!l93$9!:*

+8 KONSTANTIN D. LITASOv/ ANTON SHATSKIY AND EIJI oHTANI

abundance and because of the poor stability ofquench products during polishing and electronprobe microanalysis (Dasgupta & Hirschmann,2006; Ghosh et aI., 2009). Mass-balance calcula-tions of samples obtained below apparent solidiin these studies show clear deffcits o{ alkalis sug-gesting the presence of minor alkali-rich liquidor solid carbonate phases. Recently, we reportednew results on melting and subsolidus phase rela-tions ln alkali carbonatite systems lLitasov et a1.,

2013a). We used two starting compositions of Na-and K-rich {NarO :7 wt'h and KzO : 2 wt o/"

and vice versa) Mg-Ca carbonatite with minorSiO, and FeO and studied their phase relationsat pressures from 3 to 21 GPa. The experimen-tal results and phase compositions are shown inFigures 2.8 and 2.9. Major carbonate phases inboth the Na-carbonatite and K-carbonatite sys-

tems are aragonite and rnagnesite. Magnesite wasa liquidus phase together with silicates and wasfound among run products in a1l experimentsup to 1400-1600"C. Aragonite contains signif-icant amounts of NatO (up to 7 wt o/"), KtO

(up to 1 wtTo), and MgO {up to 8 wt %} in theNa-carbonatite system. Since it can store mostNarO in the system, only minor K-Na-bearingcarbonates were observed in Na-carbonatite insubsolidus runs. The maximum temperature sta-bility of Na-aragonite is near 1400"C at 16GPa.The solidus temperatlue is deffned by the sta-bility of double carbonate phases. The slope ofthe solidus is relatively steep at pressures below8-l0CPa (Figure 2.8). Above 10GPa the solidusbecomes fl.^t al a temperature near 1150"C.

Several K- and Na-bearing double carbonateswere observed in both the Na-carbonatite and K-carbonatite systems. The major phases observedin the experiments are (K,Na)rMS(COs)z (f IASI

and {K,Na)tCao(CO3)5 (K-Ca) (Figure 2.9). Thehigh-temperature stability limit o{ the K-Ca phaseis different in the Na- and K-carbonatite systems.The Na-bearing phase in the Na-carbonatitehas a slightly negative PT-slope of the upperphase boundary above 15CPa, whereas theK-bearing phase in the K-carbonatite is stable upto 1250'C at 21 GPa (Figure 2.9). Other subsolidus

Depth {km)300 400 500

5 10 '15 20 25Pressure (GPa)

Fig. 2.8 The solidi of Na- and K-bearing carbonatite from 3 to 21 GPa and the stability of alkali carbonate phases

{Litasov et a1.,2013a) (see Figure 2.10 ior compositions). GS-1 I shows the solidus of anhydrous carbonated pelitea{ter Grassi and Schmidt (20i 1). PT-proffles are a{ter Frgure 2.3. Abbreviations for double carbonates are as follows:NarMgr: NarMgr(COa)r, NarCa: NazCa{CO:)2; NarCar: NazCaz(COs):; (K, Na)rcao : (K, Na)rCaa(Co3)q; and

KrMg: KrMg(CO.)r.

sloa):isjc)g

E,0)

I

Earth's Mantle Mehing in the Presence of C-O-H-Beafing F\uid 49

(K,Na)rCa2{CQ),

(K,Na)2Ca(COs)2

N^ * K 1t1x

"s rriislcors,

{K,Na}2Mg2paj3

Fig. 2.9 Phase compositions in the carbonatitesystems at Ca-{Mg + Fe)-(Na + K). N and K - arestarting compositions of Na- and K-bearingcarbonatite, respectively. The dashed arrow shows themain trend of melt composition with increasingtemperature {Litasov et al., 20I3a). Reproduced wrthpermission of the Geological Society oI America.

double carbonate phases were observed only inthe Na-carbonatite system and include the ap-pearance of NarMgr{CO3 ), in a single experimentat 3 GPa and 750'C. NarCa2{CO"). (shortite com-position) was observed at 10 GPa and 1100'C, andNarCa{CO.), (nyerereite composition) at 16 GPaand 1100'C. Na-Ca-carbonates were reported tobe stable in a carbonated pelite system (Grassi &Schrnidt, 2011). The high-temperature stabilitylirnit of these phases as well as the solidus of thesystem were higher than for a1ka1i carbonatite(Figure 2.8).

2.6.4 Systems with a rcduced C-O-H fluidRecently, carbonated peridotite systems werestudied along a range of fO, conditions atpressures up to 23 CPa (Stagno & Frost, 2010;Rohrbach & Schmidt, 201 1 ) in order to model lO,dependence of carbonate stability and meltingtemperatures. It was shown that magnesite is sta-ble at f Ot: 2-B 1og units below the FMQ bufferand with pressure its stability slightly expandstowards more reduced conditions (Stagno &Frost, 2010) or remains constant (Rohrbach &

Schmidt, 2011). These papers show that reducedforms o{ carbon (diamond or carbide) are stableacross most of the upper mantle. Carbonatiremelt or carbonate can be stable only at depthsiess than 150km or in highly oxidized regions(Figure 2.5). These experiments were performedunder nominally hydrogen-free conditions.

We performed experiments in peridotite andeclogite systems with a reduced C-O-H fluidbuf{ered at MMO (Mo-MoOr) and IW at 3-16 Gpaand 1200*1600"C (Litasov, 2AlI; Litasov et al.,2013b). Roughly, MMO is located one log unirabove the IW buffer atT-f Oz diagram. The exper-iments were performed using a modified doublecapsule rnethod. The solidus temperature in thesystem peridotite-reduced C-O-H fluid for bothbuffers is substantially higher than the solidi insystems with HrO and CO, {Figure 2.10), but still400-500"C below the melting curve of " dry" peri-dotite at 16 GPa. The eclogite solidi are 50-100"Cbelow the peridotite solidus. Within the studiedpressure interval {3-16 GPa), the solidus does notflatten out, unlike in systems with HrO and COr.In both systems/ the ffrst melt generated near thesolidus has 40*50 wt % SiO, in dry residue.

The fluid composition was not determined inthe experiments. The pressure dependence of thefluid composition at 1200"C was calculated {romthe equations o{ state in lZhang and Duan, 2009)

{Figure 2.11). The major component at the fO,deffned by the MMO buffer is HrO, which con-tent increases with temperature and pressure. Themajor component at the f 02 deffned by the IWbuffer is CH., but HrO becornes predominantwith rlsing temperature and pressure. Additionalcomponents o{ the fluids are ethane or hydrogen.

2.7 Melting Behavior in Different MantieSystems with Volatiles

Experimental studies of peridotite and eclogitesystems with C-O-H voiatiles show a wide vari-ation in the position of the solidi depending onf O, as well as on the bulk composition of the sys-tem. In general, the solidi in the eclogite systemsare below those in the peridotite ones or coin-cide with them, as is the case with COr-bearing

(K,Na)rCao(CO

Na-carban6tlte:.1. 3.0 and 6.5 GPac 10.5 CPat 15.5 GPa+ 21.0 cpa

K-6arbonaliie:A 3.0 and 6.5 GPaO 10.5 CPa: 15.5 GPao 21.0 GPa

ilmFwr"/nffidqqirm{t*.n'i::'Jf *il*nrn#rii r n

50 KONSTANTIN D. TITASOV/ ANTON SHATSKIY AND EIII OHTANI

n^^rt- /t.a,\ l-lonth /lrm\L/sPrll\nll:,1 usvur l^rilJ200 300 4c0 5c0 0 100 20c 300 4.00

o.^^^3 IOLIJ

;:J- lAnn

0)aE , n,,n

f-

Fig. 2.10 Solidi in the systems peridotite C O H fluid (A) and eclogite C-O H fluid (B) at lO, buffered by MMOMo-MoO2) and IW (Fe FeO). TG, Solidus in the system peridotite + C-O-H fluid at iO2 : IW + 1 {Taylor &Green, 1988) (Figure 2.5). Ior comparison a dash-and-dot line in B rndicates the peridotite sohdus bu{fered at IWirorn A. Gray lines indicate the main phase transitions, volatile-free solidi, and mantle geotherms from Figure 2.3.Ol, Olivine; Gt, garnet; Op, orthopyroxene; Cp, clinopyroxene; Wd, wadsleyite; Rw, ringwoodite; Fl, fluid.Reproduced with permission of Nature.

1n {:Pracrr :ro l()Fr',

(a)

compositlons (Ghosh et al., 2009, Ktseeva et al.,20i31. The factors determining the position of thesolidi at pressures above 6 GPa in systems withH,O and CO, reveal fundamental differences. Inthe HrO-bearing systems, the solidus depends onthe H.O solubllity in nominally anhydrous sil-icates. The solidi in the system peridotite-HrOrr ithin the transition zone can 1ie 300-400 Cabove those of HrO-bearing eclogite because of:1-ie high HrO solubility in wadsleyite and ring-rtoodite (Figure 2.3). The position of the solidus inCO. -containing systems depends on the presence.ri alkalis and HrO. A small amount oi KrO can re-

luce the solldus temperature of carbonate-bearing:clogrte or peridotite by 400-500"C at 20GPa.

10 15n----.^..-^ /^n^\' '""-"'- \-- */

(b)

The CO, content of the system itseli should nothave significant influence on the solidus. The sta-bility of carLronates (magneslte, aragonite) is notsignificantly dependent on the a1ka1i content, butdecreases dramatically if H2O is added to the sys-tem (Figure 2.7). In peridotite and eclogite systemswith a hypothetical C-O-H fluid, the solidi areat higher temperature than for systems with CO,and HrO. Nevertheless, they are still substan-tial1y (300-400'C) lower than the "dry" solidi at16GPa (Figure 2.10). Under reduced conditions,prevalent in the most mantle, the first melt {orPT-profiles between the average adiabat and sub-duction would be a metallic liquid in the Fe-S-Csystem (Morard et a1., 2007i Nakajima et a1.,

Ni-H20Fe-CHa

Earth's Mantle Melting in the Ptesence o;f C-O-H-Beafing Fluid 5 1

0.8

5 o.o;o9 04

X4.2

05101520Pressure (GPa)

Fig. 2.1 I Composition of the C-O-H fluid at 1200"Cat lO, bufiered at IW (Fe), Mo MoOr, MMO (Mo), andNi-NiO, NNO {Ni)r calculated {rom the equations ofstate for real gases (Zhang & Duan, 2009). Only HrOcontent is shown for NNO. Reproduced withpermission o{ Elsevier.

2009). The melting occurs when the line IW : 0 iscrossed, because the system is no longer bufferedat IW. This boundary cannot be correlated witha depth of 200-250km as yet, it may certainlylie above or below. On crossing the line IW: 0,the solidus temperature shi{ts gradually towardincreasingly oxidized systems as f 02 changes; italso depends on the HrO solubility in silicates.This process may cause redox melting and isconsidered in detail in the subsequent section.

The eclogite solidus will always be below theperidotite solidus, which implies the preferen-tial melting o{ eclogite. The difference in thesolidi and carbonate-stability temperatures is upto 100*200'C at pressures above 6GPa. In thepresence of mantle heterogeneities and for melt-ing caused by p1ume, eclogites will be the ffrstto melt, leading to an enrichment of eclogitecomponent in the meit. This conclusion is con-sistent with modern models for mantle mag-matism, where eclogites and hybrid pyroxenites(produced by the reaction of basaltic melts withmantle peridotites) play an important role in dif-ferent mantle sources (Hofmann & White, 1982;Sobolev et al.,2007). fudging by the trace-element

composition of olivine phenocrysts, pyroxenitesand eclogites account for 10-30% of MORB,up to 60% and more of OIB and continentalbasalts, and 20-30% of komatiites. Such meltingmay also take place under "dry" conditions, be-cause the solidus temperature of "dry" eclogite isalso below that o{ peridotite. However, volatilesplay a substantial role in almost all rnodels forbasaltic magma generation {for exampie, MORBor OIB) even if their contents are 1ow. Accord-ing to recent reviews (Hirschmann et a1., 2009iDasgupta & Hirschinann, 20i0), the most de-pleted MORB sources contain 30*120ppm CO,and 50-150ppm HrO, and OIB sources contain120-1800 ppm CO, and 350-1000 pprn HrO.

Analysis of the melt compositions obtained bypartial melting of peridotite and eclogite at pres-sllres below 7 GPa can be directly compared withnatural igneous rocks with the exception of reac-tive kimberlite and other alkaline volcanics thatmay have been modified upon ascent. Melting ofvolatile-{ree mantle is possible at 20-50 km depthunder the mid-ocean ridges or by significant in-crease of temperature, for example, in very hotmantle plumes. The 1ow-degree partial melts ofperidotite correspond to tholeiite basalts at thedepth levels to 100 km.

Partial melts obtained in experiments at pres-sures above 6-7 GPa cannot be directly comparedwith natural magmas. Therefore, we can onlyconstrain hypothetic models for deep rneltingand magrnatism. There are several important is-sues for melt compositions formed by meltingof volatile-bearing mantle lithologies. Meltingof HrO-bearing peridotite and eclogite produceandesitic, basaltic, and komatiite-like melts at20-40% melting. In the systems with COr, 1ow-degree partial melts are carbonatitic in a widepressure range at least to 30 GPa. Similar car-bonatite melts can be formed from carbonatedperidotite or eclogite at HrO-undersaturated con-ditions. We should emphasize that primary man-tle carbonatites are extremely rare. Most of themwere formed either by liquid immlscibility or byfractionatlon of alkali magma (Mitchell, 2005).Under HrO-saturated conditions low-degree par-tial melts of carbonated peridotite and eclogite

ffitlqi{HirdlFifi"'rr'! ${r 'r!'r!'^ilrt r "'$r$ir"!r!'ir{r- '"" 'r'-'-.*- "-'-:-":lFril

52 KoNSTANTTN D. LITASOI/, ANTON sHATSKIv AND EIJI or{TANr

are similar in composition to those in the systemswith F{rO due to the enhanced thermal stability o{carbonates relative to hydrous silicates. Basalticmelts are also {ormed in systems with reducedfluid (fakobsson & Hol1oway,2008; Litasov et aL.,

2013b).The brief review of melt compositions above

indicates that in most environments low degreepartial melting of volatile-bearing mantle litholo-gies leads to {ormation of silicate melt, withcompositions close to basaitic magma. Forma-tion of carbonatite- or kimberlite-like melts ispossible only under relatively HrO-poor oxidizedconditions {in the stability field of carbonates orH2O-CO2 fluid with molar H2O/(H2O + CO2) n-tio less than 0.3). This suggestion is consistentwith low concentrations o{ HrO (<20-50 ppm) inminerals from diamondiferous xenoliths and in-clusions in diamonds {Matsyuk & Langer, 2004;Matveev & Stachei, 2007).

2.8 Redox Melting, Redox Freezing, andDiamond Formation

Although some diamonds contain native iron{Sobolev et al., 1981), they cannot be inequilibrium with Fe at mantle PT-conditions.Fe-carbides {FerC and FerC3} are stable at anyPT-parameters below the solidus o{ Fe-C system(e.g. Scott et aL., 20011. The assemblage of Fe anddiamond is possible only i{ Fe is in the moltenstate and carbon solubility in this metal melt islimited. Accordingly, the presence of diamond inthe mantle including parts of the lower mantlelevels suggests the absence of native iron there.This supports the idea that at least some regionsin the mantle are more oxidized than expected{rorn standard models described above. The slowrate of exchange processes in the mantle may beconsistent with super-deep diamond formationin oxidized ancient subduction domains.

Our data for melting in volatile-bearing sys-tems constrain the conditions {or redox melting inthe Earth's mantle. Classic redox melting impliesoxidation of CH*-HrO fluid, which causes forma-tlon of HrO and CO2 and drastically decreasesthe melting temperature of rocks (Figure 2.12).

Depth (km)0 100 20a 300 400 500

600

0510152APressure (GPa)

Fi1.2.12 Mantle solidus at the contact of oxidizedasthenosphere and reduced cratonic lithosphere. Thearrow show metasomatic or magmatic modiffcation ofcratonic roots (gray fie1d) and crystallization o{asthenosphere melt by interaction with reduced anddepleted cratonic peridotite. This wouid be the majormechanism o{ diamond crystallization in cratonicroots. The PT-profi1es and solidi are from Figures 2.3,2.6, and2.1.0.

Recently, Foiey (2011) suggested that two typesof redox melting operate in the mantle. He calledthe classic mechanism hydrous redox melting(HRM) and emphasized the importance o{ carbon-ate redox meiting (CRM). The HRM correspondsto the transition from methane to hydrous flu-ids (at f Oz:IW : 0 - IW + 1.5) and character-izes ancient subduction, lithosphere erosion andmetasomatism. The CRM corresponds to transi-tion from hydrous to COr-fluids lat f Or: IW+4.5 - IW*5.5 or FMQ-1.5-FMQ-0.5) and istypical {or Phanerozoic subduction and magma-tism. An important observation is the extensionof the region of the so-called water maximumwith increasing pressure. At high pressure (above

$ rooo

.fr r+oo

!)

f; rzootr

1000

oxLdized sclidus

-nnttta

Eatth's Mantle Meltlng in the Presence o;f C-O-H-Beafing Fluid

7-10GPa) the difierence between /O, values iorHRM and CRM increases up to several ordersof magnitude {several log units). Redox meltingcan operate only at fO, parameters that drasti-ca1ly change fluid composition. With increaslngpressure, the expanding range oI fO, values forthe water maximum may reach nearly to theIW bu{fer, thus at pressures above approximately10CPa, redox melting becomes less and less im-portant in the mantle. FIowever, modeling fluidcomposition is difficult at present since the effectof melting and the influence of dissolved silicatesare not well known.

2.9 The Big Mantle Wedge Model andCarbonates

2.9.1 The model

Recent studies suggest the existence of a so-called "big mantle wedge," when a water-bearingsubducted plate stagnates and dehydrates in thetransition zone (Figure 2.13). When a stagnantslab is heated, meiting takes place and a u,"ater-bearing melt rises upwards, causing local melting

lntracontinental Back lsland Mid-volcanism arc arc ocean

basin *l ridge

.*- 410 km

600 km

Fig. 2.13 Schematic model {or a "big mantle wedge"caused by devoiatilization of a stagnant slab in thetransition zone. Carbonate and rninor HrO can betransported to transition zone depth, wheredecarbonation melting can occur. Segregation ofcarbonate-bearing melt can cause ascent o{ buoyantcarbonate blob-diapirs from transition zone towardsthc surface.

ln the upper mantle and magmatism on rhesurface {Ivanov, 2007; Maruyama et al., 1009,Ohtani 8r Zhao, 2A09; Zhao & Ohtani, lO09This modei was appiied to continental volcanrsrnin the China Craton (above the stagnant paclncpiate in tire transition zoneJ (Ohtani &. Zhao,2009) and to the Siberian Traps (Ivanov, 2OOr-'.The Farallon s1ab, which is subducting underthe Nortir America deep into the lower man-tle, may provide an example for this model. Here,a "wet" plume may separate from the bottom o{the transition zone approximately at the 650 kmdiscontinuity (van der Lee et aL.,2AA81,

The original model rnay require some modifi-cations. First, the HrO content preserved whenthe plate is subducted below the level of island-arc volcanism should not exceed 0.1 wt % in theupper 10-20km of the plate and may be higheronly in the coldest plates (Kerrick 8r. Cc'nnolly,200i; Poli &. Schmidt, 2002). These contents areclose to those in the source regions of OIB andenriched MORB and cannot cause serious con-sequences such as large-scale mantle melting.Second, as discussed in the previous sections,the transition zone can be considered as a reser-voir {"sponge") {or HrO, and HrO-bearing silicaternelts may not be able to cross the 410krn dis-continuity (Bercovici & Karato, 2003). Third, ahypothetical HrO-bearing fluid/melt separatedfrom a subducted plate has to react with reducedmantle rocks and transforms into CHo or H2,rvhich, presumably, have very low silicate so1-

ubility and low wetting angles. Therefore, theability of this fluid or melt to migrate will belimited. Some enrichment o{ transition zone b__v

HrO can be also realized via transport by thrnoverlying low-viscous boundary iayer above slab-rnantle interface, where 0.1-0.4 wt Yo HrO canbe stored during subduction {e.g Iwamori, JOO',Tonegawa et a1., 2008). Ftrowever, even ii tiris i.the case, this water would not initiate significantmelting in the deep upper mantle rvhere olir inecan incorporate up to 0.5-0.8 wt % H. O.

The situation may change if we talie thc. roleof carbonates in the melting o{ subductin-e slabsand their influence on the behavior oi H.O rnroaccount. Carbonates are preserved in srdtments,

iI

ffiff1qn'"11,'r'''

54 KONSTANTIN D. LITASOV/ ANTON SHATSKIY AND EIII OHTANI

the oceanic crust, and the upper peridotite layer,when melting in the mantle wedge occurs alongalmost any subduction PT-profiles except thehottest ones (Figures 2.3 and2.7). The existence ofcarbonatite melt in the deep mantle is confirmedby studies o{ super-deep diamonds and their lnclu-sions. Some carbonate inclusions were found inclose association with transition zone and lowernrantle minerals (Stachel et a1., 2000; Brenkeret a1., 2007; Kaminsky et a1., 20091 Moreover,indirect comparison of Ca-perovskite inclusionswith experimentally synthesized phases also re-

veals their likely equilibrium with carbonatitemelt in the lower mantle (Walter et a1., 20O8).

These ffndings suggest that subducted plates re-

main oxidized to the level o{ the lower mantleand their buffering capacity is not exhausted bythe surrounding reduced mantle. This may be

related to the limited buffering capacity of themantle (i.e. low contents of metallic Fe) or theslow rate of solid-state redox reactions (as in-ferred from the low dif{usion rates of oxygen insiiicates) (Dohmen et a1., 2002i Dobson et al.,2008). Comparison of solidi for Na- and K- bearingcarbonatite with subduction PT-proff1es indicatesthat if the oxygen {ugacity of the subducting slabremains su{ficiently high to stabilize carbonate,melting of alkali carbonate may occur at a rangeo{ depths depending on the PT-profile of subduc-tion. In some cases, they may survive to lowermantle depths, but in most cases a likely region{or melting of subducted alkali-bearing carbon-ated eclogite is the transition zone (Figures 2.7,

2.8) as the solidi o{ carbonate-bearing systemsbecome flatter above about 10 GPa (Figure 2.7).

2.9.2 The mechanism of melt segregation andmovement

Since the principle o{ segregation and the move-ment o{ hydrous and carbonatite melts are gen-

erally similar, we will consider the details of theproposed model using a carbonatite melt withsome simple approximations. The average CO,content of the upper 500m of the oceanic crustis estimated to be 3 wt % (-7 wt "k carbonatei

Staudigel et a1., L989). According to recent esti-mations by fohnston et al. l20l1l decarbonationefficiency at the island arc may be 20*80"/o,which means that in the coldest subduction zonesmost carbonates can be transported down beyond150km. At the conditions of the transition zonethese concentration of CO2 will lead to the forma-tion of carbonatite melt with a carbonate contentof about 83 wt % (Shatskiy et a1., 2013) and avolume fraction (rp) o{ up to 0.1, which is twotimes lower than the equilibrium (maximum)value (Laporte & Watson, 1995i Hammouda &Laporte, 2000), but 5-10 times higher than theiimit of carbonatite melt interconnectivity insilicate mantle (Hunter & McKenzie, 1989, Mi-narik & Watson, 1995). The density of carbonatitemelt lp*"y) at the transition zone conditions is3.0g/ cm3 based on the partial molar volumeof CO2 and the thermal expansion of carbonatemeit (Liu & Lange, 2003; Ghosh et a1., 2007iSakamaki et a1., 2011). This density is signifi-cantly lower than for mantle rocks at transitionzone conditions (3.6-3.8 g/cm3). Consequently,buoyancy-driven porous flow should resuit incarbonatite melt segregation at the slab-mantleinterface (figure 2.1 4al.

The compaction-driven flow of a low-viscositymelt through a creeping matrix has two limitingregimes: {i) a hydraulically limited regime, wherethe rate of fluid flow is controlled by the ma-trix permeability, and (ii) a rheologically limitedregime, where the rate o{ melt flow is controlledby the matrix rheology (Connolly et a1.,2009).Inthe first regime the melt ascent velocity (v) canbe expressed as:

'l'orous -k.Ap.gQ ' 4tttott

(2.1)

where Ap is the density contrast between meltand residual silicate matrix (-600 kg/m"), g isthe gravity (9.8 m/s2), 4 is the viscosity oi melt(0.015-0.005Pa s) {Treiman & Schedl, 1983;

Genge et a1., 1995, Dobson et a1., 1996), and k isthe permeability, which depends on melt fractionlrp), grarn size ld), and the geometrical factor (C):

. ^i1 )Z

C12.2)

Earth's Mantle Mehing in the Presence of C-O-H-Bearing Flutd 55

lnf iltraction driven bysurface tension

Flux of CO2 < 1011 g/year.penetraction depth -1 m

over first year

Buoyantcy-divenporous flow

> 500 m/year

Pressure-solution creepof infiltrated layer

Buoyant melt diapir,Rate of 2-km body

-0.5 m/year

(b) (c)

Fig.2.14 Schematic illustration of the formation o{ a

carbonatite diapir as a result of partial melting in theCOr-rich uppermost layer o{ the subducting slab in thetransition zone. Three mechanisms of the melttransport are involved: ( 1 ) buoyancy-driven porous flowwithin partially molten slab; (2) surface tension-driveninffltration o{ tt drytt overlaying mantle;(3) buoyancy-driven diapir ascent accompanied bypressure-solution creep o{ the inffltrated layer. See text{or details.

Empirical relationships with values of n ry 3 andC ry 10 to 270 were inferred {rom experimentaldata (Wark & Watson, 1998; ConnoIly et aL.,2009iZhu et aI.,2011). Considerlng recent experimen-tal results on the porous flow of basaitic melt(Conno11y et a1.,2O09i Zhu et a1.,2Oll) and thelower viscosity of carbonatite melt {by 3-4 ordersof magnitude), both porous flow regimes should

provide suificientiy fast rates (i.e. >500 mlyearl{or the efficient partial melt extraction from theuppermost 500-m layer of the subducting slab(Figure 2.14a,2.I5a). The corresponding CO, fluxwould be:

f po,our: L .\Nq .vpo,ourpM"ttcYJ;/loou, lz.BJ

where I ry 5 km is the width of partially moltenarea along the solidus temperature lCurrre et a1.,

2004) {Figure 2.14a1, I4l: 38 485km is the ap-proximate total length o{ modern subductionzone segments (van Keken et a1.,2011), CYJ::33wt% is the CO2 content in the melt {shatskiyet a1.,2013). These values yield t'oorou, >1016 g o{CO2f year {Figure 2. I 5b).

Sur{ace-energy considerations (Watson, 1982:,

Stevenson, 1986; Riiey et aL, 1990) suggest thata carbonatite magma body in chemical equilib-rium with its surroundings (overiying volatiie,poor "dry" mantle) wiil tend to dissipate bywetting the dry grain edges of the host mate-rial, because the dihedral angle measured at thecontact of silicate minerals and carbonatite meltis much lower than 60 (Hunter and McKenzie,1989; Minarik and Watson, 7995; Yoshino et a-1.,

2007). Since the capillary force is nondirectional,this process would counteract melt segregation,which is supported by the {orces driving direc-tional fl uid migration. High-pressure experimentssupport a fast lnffltration rate of carbonatite melton a millimetre scale (Hammouda & Laporte,2000). However, this rate would diminish rapidlyif infiltration distance extends to geologicallyrelevant kilometre scale due to the increase ofdiffusion distance and to the blurring o{ interfa-cial energy difference at the interface with the dryrock (Figure 2.L5a). This explains why {ull dissi-pation of carbonatite magma chambers into the" dry" mantle does not occur in reality and whymelt segregation, for example of carbonatite andkimberlite magmas happens.

Infiltration involves chemical solution of sili-cate at grain edges and simultaneous precipitationof silicate crystals within the melt reservoir. Tobalance the flux of the equilibrium melt intothe nonporous solid aggregate, an equal di{fusive

@

w

56 KONSTANTIN D. LITASOV/ ANTON SHATSKIY AND EI'I OHTANI

Porous fiow within slab

lx

0123456log(t fyear])

(a)

counterflux of solid through the melt must exist(see Flgure 2.3 in {Hammouda & Laporte, 2000)).The rate of melt infiltration can be expressed{rom characteristic diffusion distance (x) {Crank,1975) as:

v,n, : dxldt: ",qDn. Q 4)

Silicate diffusivity {D) in carbonatite melt wasestimated to be 2 x 10-e m2/s at the PT con-ditions of the mantle transition zone (Shatskiyet a1., 2013). Overall CO, flux associated withthe melt infiltration into surrounding mantle canbe expressed as:

fi"|, : utnt .w .x . pMett. ,pr. cfll]1rco"/,, lz.s)

rvhere x - 0.06 mfyear is the average subduct-ing rate {van Keken et a1.,20II), pl : 0.2 is theequilibrium (maximum) carbonatite melt volumeiraction in the silicate mantle near the interfaceu-ith the melt chamber (Hammouda & Laporte,)OOOI, CcMn -33 wt % is the CO, content in themelt (Shatskty et a1.,2013). According to our es-timations using Eq. 5, the maximum initial CO,t-lux into pristine surrounding mantle is about2 x 1011 gfyear, which is three orders of magni-tude lower than the annual CO, supply from the

Porous flow wrthin slab

Ascent of diapirSubduction of sediments

lnkp,tv or,_,. :..qfp, *. _<enr€

_

18

C 1Ao '"0

.o1 14

OO ro

o-9 ro

c'ao

F

(!M.Oaa -/-

Fig. 2.15 Dynamics of moving carbonatite melt (a) and of CO, fluxes (b) in a subducting slab. Four processes areconsidered. "Subduction" implies CO, deiivery by subducting slab {in the uppermost 500m layer) and rate ofresulting carbonatite melt production in the transition zone. "Porous flow" denotes the rate of melt segregation inthe partially molten layer across the solidus o{ carbonated rock. "Inffltration" gives the rate and flux ofimpregnation ol " dry" overlying mantle by carbonatite melt segregated at the slab-mantle interface. "Diap:u,,denotes the rate and maximum flux of carbonatite melt upon ascent of a diapir with l km radius.

0123456Log (t [year])

(b)

subducting slab (staudigel et aL., 1989; Hayes &Waldbauer, 2006; Dasgupta & Hirschmann, 2010)

{Figure 2.15b). The actual tki' iteven smailer, be-cause the melt fraction decreases rapidly with in-creasing p enetration distance. Ftxthernor e, ffnl

2

diminishes rapidly with time as the surroundinggets saturated by carbonatite melt (Flgure 2.15b).

Buoyant ascent of the melt diapir is possible ifthe viscosity of the country rocks is 1ow enoughlFyie, 19731. For a sphere moving under gravity ina viscous medium, the velocity (v) of the spherels given by:

vDiapir :2lrzlp, - Qrll(94o,y--on tn), 12.6)

where g is the gravitational acceleration/ r is theradius of the sphere, ps - gr is the density con-trast between the sphere and the medium and4Dry_mantle is the viscosity of the medium. Assum-ing a mantle viscosity of 3 x 1021 Pa.s lForte et a1.,

1991) the ascent rate of a sphere with z: l kmwill be negligibly slow, i.e. on the order of 10-snfyear. In addition, the carbonatite melt cham-ber must be surrounded by a low-viscosity silicatelayer enriched in carbonatite melt (Figure 2.15c).Low viscosity of this layer would be due to thespecific deformation regime. Dislocation creep

Earth's Mantle Mehing in the Presence of C-O-H-Bearing Fluid 57

limited by grain-boundary diffusion in solid-stateis the predominant deformation mechanism ofthe "dry" silicate mantle. However, deformationof rock wetted by solvent occurs by a pressure-solution creep, which is li.mited by silicate dif-{usion in the liquid {Weyl, 1959). The principleof this mechanism is as follows: stress concentra-tion at the grain contacts causes local dissolution,diffusion of the dissolved material out of the in-terface, and deposition at the less stressed faces ofthe grains. Since the dif{usion in the liquid is sev-erai orders of magnitude faster than that in solid,the pressure-solution creep will be much fasterthan diffusion creep. The viscosity of wetted rockis given by:

dsRT4Wet-mantle: ADCoMw'

where A is the constant of about 10 {or equiax-ial polycrystals (Vickers & Greenffe, 1967iKruzhanov & Stockhert, 19981, C0 is the silicatesolubility in carbonatite melt (silicate molefraction is about 0.24 at temperature of adrabat),M is the molar volume of silicate (3.93 m3/mo1for wadsleyite {Katsura et a1., 2009)). Followingthe fluid fi1m model, the effective grain boundarywidth, w, is of the order of 1-l0nm and thediffusion coefficient of solute in ff1m is thesame as an order of magnitude lower than thatin the bulk melt (Dysthe et a1., 2A03). Suchan analysis yields a viscosity of 8 x 1016 Pa s

providing the ascent rate of A5mfyear for r:1 km {Figure 2.15a). Diapirs with such ascentrate could consume ail carbonatite, which couldbe potentially extracted from the subductingslab (Figure 2.15b). Nominally, rising diapirscould support an upward CO2 flux, two orders ofmagnitude higher than the subduction CO2 flux.Note that the maximum rate of diapir ascentthrough the " dty" mantle cannot exceed theinfiltration rate. According to our preliminaryestimations this rate approaches to 0.5 m/year

{Figure 2.I5a). However, diapirs following behindthe first one would not have such a rate limitationif carbonatite saturated conduits are established.

Rohrbach & Schmidt (201 1) suggested that car-bonatite melt is unstable, when at depths greater

than 250 lcm, and is reduced to immobile diamondor carbide since the mantle redox conditions cor-respond to the stability of (Fe,Ni) metal. In ourmodel, most of carbonatite melt at the slab-mantle interface would segregate into magmadiapir rather than disperse into surrounding man-tle (Figure 2.15b). Moreover, since the carbonatereservoir is replenished by continuous subduc-tion, the limited amount of iron in the mantleabove the slab may be totally oxidized with time.

Most extensive redox interaction is expected atthe beginning of carbonate melting, when thefirst diapir rises up through the pristine FeO-

bearing mantle from about 550 to 250km depth.A11 metallic iron on the route of this diapirwould inevitably react with carbonatite melt:2Fe + CO, -+ 2FeO * C. Carbon solubility incarbonate melt is about 0.3 wta/o at the man-tle conditions (Palyanov et a1., 20051. Therefore,carbonatite melt soon becomes supersaturatedwith carbon, which causes diamond precipita-tion {Figure 2.16). Reduction o{ CO2 should alsobe accompanied by silicate precipitation from themelt. This "redoxfreezing" gradually reduces sizeof the carbonatite diapir and its ascent rate byabout 30% and 50"/", respectively, for an initial

500 450 400 350 300 250

Depth, kn

Fig. 2.16 Carbon loss of the carbonatite melt duringdiapiric ascent due to carbonate reduction by metalliciron dispersed in the mantle. We assumeC(Ieo ) : 0.1 wt % in the depth range of 660-250 kmllrost et al.,20081.

12.7)

\o

$.d

€l)

u

I -l-jBl

1

'.1

Er.-l'|

21II

al

,|

u -r-c3u

Initial C as CO,'' ix melt

*c4*gi;;t ,rrj

u,ql.*e::'lit'.t*;

58 KoNSTANTIN D. LITASOV/ ANTON sHATSKIv AND Ettr oHTANI

radius of the diapir of 1 km. Note that diapirs withz < 0.5 km will be completely reduced during thefirst ascent. The interaction of initial diapirs withthe reduced mantle will create a network of ox-idized conduits saturated by carbonatite melt orcarbonate, through which subsequent diapirs canarise rapidly. As far as the rate of diapir ascentis one order of magnitude Iaster than the convec-tion in the surrounding mantle {Stern, 2OO2l, thelatter could not signiffcantly modify the oxidizednetwork if consecutive diapirs emerge every 104

years. In such a way a continuous CO2 (carbon,atite melt) transport from the transition zone tothe shallow mantle may be established. This sys,tem should contribute to the volatile budget inthe source regions of basalts, kimberlites, and car-bonatites, which return CO, back to the Earth'ssurface (Figure 2.15). We also propose that a simi-lar mechanism governs the delivery of primordialcarbon and hydrogen from the core-mantle bound-ary or at least from 660km depth.

In the above discussion we assumed the meltcomposition to be anhydrous carbonatite accord-ing to {Kerrick & Connoiiy, 2001}. We also esti-mated the possible contamination of carbonatitemelt by HrO during ascent through the tran-sition zone from a stagnant slab surface (-560km depth). The maximum HrO solubility inwadsleyite and ringwoodite in equilibrium withhydrous melt/fluid is about 0.35 wt 7" at transi-tion zone P-T conditions {Litasov et a1.,2}Ilbl.As we discussed above, the ascent of carbonatitediapirs involves extensive silicate recrystalliza-tion through intergranular carbonatite melt atthe diapir front. Re-equilibration of wadsleyitewith hydrous-carbonatite melt reduces its HzOcontent by about 70% {Shatskiy et a1., 20Ob).

If carbonatite melt was initially dry, this valuewould approach to 100%. For a spherical meltdiapir with z : I km the maximum HrO contentin a carbonatite melt can vary in the range of25-32 wt o/o:

M#;'3: n .t .(560 - 410) . t03 pwa,a*

. nc[!{- lroo :4 - 6 x to15 g,

MDiopir:413 'n '13 ' Pmen: 1.3 x 10tu g,

cy:[: m\iS/lmo,"uj, + My:[).100:25 + 32 wt.o/"

Yet, no or minor hydration can be expected forsubsequent diapirs travelling through a networkof dried conduits. We expect that hydration of car-bonatite melt will increase the rate of segregationand diapir ascent rate up to one order of magni-tude due to the increased diffusivity and solubilityof silicate components in the HrO-bearing melt(Shatskiy et aL., 2013).

2.lO Concluding Remarks

Fundamental dif{erences between deep {> 200 km)mantle melting for systems containing H zO, COz,and a reduced C-O-H fluid are outlined. Melt-ing in the HrO-bearing systems is controlledby hydrogen solubility in nominally anhydroussilicates and occurs when srlicates are supersat-urated with HrO at deffnite P, T, X, arrd fOr.Melting in CO2-bearing systems is determined byalkali carbonate stability and controlled mainlyby NarO, KrO, and H2O. Studies of the peri-dotite and eclogite systems containing volatilesshow that most solidi flatten out at pressuresabove 6-8 GPa, which may cause melting whenthe solidus intersects the PT-profiles of subduc-tion and average mantle. Mantle melting in thepresence of voiatiles strongly depends not onlyon PT-conditions, but also on the redox state.An increase in f O, causes redox melting in de-ffned parts of the mantle. The stability boundaryof Fe-Ni metal and the 410 and 560km dis-continuities are most important {or redox anddecarbonation-dehydration melting. We also ar-gue that subducted carbonates should play amajorrole in the "big mantle wedge" model for stagnantor deeply-sinking slabs and we propose a newmechanism f or generating slab-derived carbonate-bearing diapirs in the transition zone.

Acknowledgements

The authors wish to thank Hans Keppler andHikaru Iwamori for numerous useful comments,

Earth's Mantle Mehing in the Presence of C-O-H-Beadng Fluid 59

which substantially improved the manuscript.The study was conducted as part o{ the G-COEprogram "Advanced Science and TechnologyCenter {or the Dynamic Earth" at TohokuUniversity, supported by the Russian Foundationfor Basic Research (grant no. 09-05-00917 and12-05-33008) and Integer project of SiberianBranch of RAS no.97 Ior 2012-2014.

References

Bercovici, D. & Karato, S., 2003. Whole-mantle convec-tion and the transition-zone water fflter, Nature,425,39-44.

Bolfan-Casanova, N., 2005. Water in the Earth's mantle,Mineral Mag, 69, 229 -257.

Bolfan-Casanova, N., Keppler, H. & Rubie, D.C., 2000.Water partitioning between nominally anhydrousminerals in the MgO-SiO2-H2O system up to24GPa:implications for the distribution of water in theEarth's mantle, Earth Planet Sci Lett, t82,209-22I.

Boyd, F.R., Pokhilenko, N.P., Pearson, D.G., Mertzman,S.A., Sobolev, N.V. & Finger, L.W., 1997. Corn-position of the Siberian cratonic mantle: evidencefrom Udachnaya peridotite xenoliths, Contrib Min-er al P etro1, 128, 228-246.

Brenker, F.E., Vol1mer, C., Yincze, L., Vekemans,8., Szymanski, A., Janssens, K., Szaloki, I., Nas-dala, L., |oswig, W. & Kaminsky, F., 20O7. Car-bonates from the lower part of transition zoneor even the lower mantle/ Earth Planet Sci Lett,260,1-9.

Bromiley, G.D. & Keppler, H., 2004. An experimentalinvestigation of hydroxyl solubility in jadeite andNa-rich clinopyroxenes, Contrib Mineral Peuol, 147,1 89-200.

Brooker, R., Ho11oway, |.R. & Hervig, R., 1998. Reduc-tion in piston-cylinder experiments: The detectionof carbon inffitration into platinum capsules, AmerMineral, 83,985-994.

Conno1ly, |.A.D., Schmidt, M.W., Solferino, G. & Bag-dassarov, N., 2009. Permeability of asthenosphericmantle and melt extraction rates at mid-ocean ridges,Nature, 462, 209-212.

Crank, f., 7975. The mathematics of diffusion, Claren-don Press, Ox{ord.

Currie, C.A., Wang, K., Hyndman, R.D. & He, |.H.,2004. The thermal e{fects of steady-state slab-drivenmantle flow above a subducting plate: the Cascadia

subduction zone and backarc, Earth Planet Sci Lett,223,35-48.

Dasgupta, R. & Hirschmann, M.M., 2006. Melting in theEarth's deep upper mantle caused by carbon dioxide,Nature,440, 659-662.

Dasgupta/ R. & Hirschmann, M.M., 2010. The deepcarbon cycle and melting in Earth's inteflor, EarthPlanet Sci Lett,298, l-13.

Dasgupta, R., Hirschmann, M.M. &Withers, A.C.,2004.Deep global cycling of carbon constrained by thesolidus of anhydrous, carbonated eclogite under up-per mantle conditions, Earth Planet Sci Lett, 227,73-85.

Dobson, D.P., Dohmen, R. & Wiedenbeck, M., 2008.Self-dif{usion o{ oxygen and silicon in MgSiOu per-ovskite, Earth Planet Sci Lett, 270, 125-129 .

Dobson, D.P., |ones, A.P., Rabe, R., Sekine, T., Kurita,K., Taniguchi, T., Kondo, T., Kato, T., Shimomura, O.8t lJrakawa, S., 1996. In-situ measurement of viscos-ity and density of carbonate melts at high pressure,Earth PLanet Sci Lett, I43,207-215.

Dohmen, R., Chakraborty, S. & Becker, H.W., 2002.Si and O difiusion in olivine and implications forcharactetizing plastic flow in the mantle, GeophysRes Lett, 29 2030, doi:10.I029 I 2O02cL0i 5480.

Dysthe, D.K., Renard, F., Ieder, 1., Jamtveit,B., Meakin, P. & fossang,7.,2003. High-resolution measurements of pressure solution creep,Phys Rev, 68, 011603, doi: 011610.011103/Phys-RevE.0l 1668.01 1603.

Foley, S.F., 201 1. A Reappraisal o{ Redox Melting in theEarth's Mantle as a Function of Tectonic Setting andTime, I Petrol,52, 1363-1391.

Forte, A.M., Peltier, W.R. & Dziewonski, A.M., 1991.Inferences o{ mantle viscosity {rom tectonic platevelocities, Geophys Res Lett, L8, I747-I750.

Frost, D.f. & Dolejs, D., 2007. Experimental determi-nation of the ef{ect of HrO on the 410-km seismicdiscontinuity, Earth Planet Sci Lett, 256, 182-195.

Frost, D.J., Liebske, C., Langenhorst, F., Mccammon,C.A., Tronnes, R.G. & Rubie,D.C.,2004. Experimen-tal evidence for the existence of iron-rich metal in theEarth's lower mantle, Nature, 428,409*412.

Frost, D.f., Mann, U., Asahara, Y. & Rubie, D.C., 2008.The redox state of the mantle during and just af-ter core formation, Phil Trans Roy Soc A, 366,43ls-4337.

Frost, D.|. & McCammon, C.A., 2008. The redox stateof Earth's mantle, Ann Rev Earth Planet Sci, 36,389-420.

60 KONSTANTIN D, LITASoV/ ANToN SHATSKIY AND EIII oHTANI

fvfe, W.S., 1973. Granulite facies partial melting andarchean qast, Phil Trans Roy Soc Lond A, 273,147-46r.

Genge, M.)., Price, G.D. & |ones, A.P., 1995. Molecular-dynamics simulations of CaCO, melts to man-tle pressures and temperatures - implications forcarbonatite magmas, Earth Planet Sci Lett, l3l,225-238.

Ghosh, S., Ohtani, E., Litasov, K., Suzuki, A. & Sal<a-

maki, T., 2007. Stabiiity o{ carbonated magmas at thebase o{ the Earth's upper mant1e., Geophys Res Lett,34, L223r2.

Ghosh, S., Ohtani, E., Litasov, K.D. & Terasaki, H.,2009. Solidus of carbonated peridotite from l0 to20 GPa and origin of magnesiocarbonatite melt in theEarth's deep mantle/ Chem Geo1,262, 17-28.

Grassi, D. & Schmidt, M.w., 2011. The melting o{carbonated pelites {rom 70 to 700km depth, I Petrol,52,76s-789.

Green, D.H. & Fa11oon, T.i., 1998. Pyrolite: A Ringwoodconcept and its current expression. In L lackson (ed.),

The Earth's Mantle. Cambridge University Press,

Cambridge, pp. 31 1-378.

Green, D.H. & Wallace, M.8., 1988. Mantle metasoma-tism by ephemeral carbonatite melts, Nature, 336,459-462.

Hammouda, T. & Laporte, D., 2000. Ultrafast man-tle impregnation by carbonatite melts, Geology, 28,283-28s.

Hayden, L.A. & Watson, E.B., 2008. Grain boundarymobility of carbon in Earth's mantle: A possible car-bon flux from the core, Proc Nat Acad Sci, 105,8s37-8541.

Hayes, f.M. & Waldbauer, J.R., 2006. The carbon cycleand associated redox processes through tirne, PhilTrans Roy Soc 8,36L,931-950.

Hrrschmann, M.M., Aubaud, C. & Withers, A.C., 2005.Storage capacity o{ HrO ln nominally anhydrous min-erals in the upper mantle, Earth Planet S ci Lett, 236,167-181.

Hirschmann, M.M., Tenner, T., Aubaud, C. 8d With-ers, A.C., 2009. Dehydration melting o{ nominallyanhydrous mantle: The primacy of partitioning, P-hys

Earth Planet Inter,176, 54-68.

Hofmann, A.W. 8d White, W.M., 1982. Mantle plumesfrom ancient oceanic ct.ustt Eatth Planet Sci Lett,57,421-436.

Hunter, R.H. & McKenzie, D., 1989. The equilibriumgeometry of carbonate melts in rocks of mantle com-position, Earth Planet Sci Lett,92,347-356.

Irving, A.f. & Wy11ie, P.1., 1975. Subsolidus and melt-ing relationships for calcite, magnesite and thejoin CaCOr-MgCO, to 36Kb, Geochim CosmochimActa, 39, 35-53.

Ivanov, A.y., 2007. Evaluation oi different models forthe origin o{ the Siberian traps. In G.R. Foulger,D.M. furdy (eds), Plate, plumes and planetary pro-cesses, Geological Society of America Special Paper,pp.569-691 .

Iwamori, H., 2007. Transportation o{ HrO beneath the

fapan arcs and its implications {or global water circu-Iation, Chem Geol,239, i82 198.

Jakobsson, S. & Hoiloway, J.R., 2008. Mantle melting inequilibrium with an lron-Wustite-Graphite bui{eredCOH-fl uid, C ontrib Miner al P ett o1, 155, 247 -25 6.

jakobsson, S. & Oskarsson, N., 1990. Experimentaldetermination o{ fluid compositions in the systemC-O-H at high P and T and low fo2, Geochim Cos-mochim Acta, 54, 355-362.

fohnston, F.K.B., Turchyn, A.V. & Edmonds, M., 20i1.Decarbonation efffciency in subduction zones: Impli-cations for warm Cretaceous climates, Earth PlanetSci Lett,303, 143-152.

Kadik, A.A., 2003. Mantle-derived reduced fluids: Rela-tionship to the chemical dif{erentiation o{ planetarymatter/ Geochetn Int, 41,844 855.

Kaminsky, F., Wirth, R., Matsyuk, S., Schreiber, A. &Thomas, R., 2009. Nyerereite and nahcolite inclu-sions in diamond: evidence for lower-mantle carbon-atitic magmas, Mineral Mag, 73, 797 -816.

Karato, S.L & |ung, H.,2003. E{fects of pressure on high-temperature dislocation creep in olivine, Phil Mag,83, 401-414.

Katayama, I., Hirose, K., Yurimoto, H. & Nakashima,S., 2003. Water solubility in majoritic garnet in sub-ducting oceanic ctt)st, Geophys Res Lett,30, 2i55,doi:2 1 1 O. 1029/2003910 1 8 127.

Katsura, T. & Ito, 8., 1990. Melting and subsolidusrelations in the MgSiOr-MgCO, system at highpressures: implications to evolution of the Earth'satmosphere, Earth Planet Sci Lett,99, 110-II7.

Katsura/ T., Shatskiy, A., Manthilake, M., Zhai, 5.,Yartazakt, D., Matsuzaki, T., Yoshino, T., Yoneda,A., Ito, E., Sugita, M., Tomioka, N., Nozawa, A.& Funakoshi, K., 2009. P-V-T relations of wadsieyitedetermined by in situ X-ray dllhaction in a large-volume high-pressure apparatus, Geophys Res Lett,36, L11307, doi: 1 13 10. I 1029 I 1200991038107.

Katsura, T., Yamada, H., Nishikawa, O., Song, M.S.,Kubo, A., Shinmei, T., Yokoshi, S., Aizawa, Y.,

Earth's Mantle Mehing in the Presence of C*O-H-Bearing Fluid 61

Yoshino, T., Walter, M.J., Ito, E. & Fr-rnakoshi,

K., 2004. Olivine-wadsleyite transition in the sys-

tem (Mg,Fe)rSiao, I Geophys Res, 109, 802209, doi02210.01029 I 02003ib00243 8.

Kawamoto, T.,2004. Hydrous phase stability and partialmelt chemistry in HrO-saturated KLB-I peridotiteup to the uppermost lower mantle conditions, PftysEarth Planet Inter, 143-144, 387 -395.

Kennedy, C.S. & Kennedy, G.C., 1976. The equillbriumboundary between graphite and diamond, I GeophysRes, 81, 2467-2470.

Kerrick, D.M. & Connolly, J.A.D., 2001. Metamorphrcdevolatilization of subducted oceanic metabasalts:implications {or seismicity, arc magmatism andvolatile recycling, Earth Planet Sci Lett, L89 , 19-29.

Keshav, S. & Gudfinnsson, G.H., 2010. Experimen-tally dictated stability of carbonated oceanic crustto moderately great depths in the Earth: Resultsfrom the solidus determination in the system CaO-MgO-AlrOr-SiO2-CO2, I Geophys Res, 115, 805205,doi:05210.0 1029 10200918006457 .

Kiseeva,8.S., Litasov, K.D., Yaxley, G.M., Ohtani, E. &Kamenetsky, V.S., 2013. Melting phase relations ofcarbonated eclogite at 9-ZI GPa and alkali-rich meltsin tlre deep mantle, I Petol, submitted.

Konzett, J. & Fei, Y.W., 2000. Transport and storage o{potassium in the Earth's upper rrantle and transitionzone: an experimental study to 23GPa in simpliffedand natural bulk compositions, I Petro1,41, 583-603.

Kopylova, M.G., Russell, I.K. & Cookenboo, H., 1999.Petrology of peridotite and pyroxenite xenoliths fromthe jerlcho kimberlite: Implications for the thermalstate of the mantle beneath the Slave craton, North-ern Canada, I Petrol, 40,79-704.

Korsakov, A.V. & Hermann, 1., 2006. Silicate and car-bonate melt inclusions associated with diamonds indeeply subducted carbonate rocks, Earth Planet SciLett, 241, 104-1 18.

Kruzhanov, V. & Stockhert, B., 1998. On the kinetics ofelementary processes o{ pressure solution, Pure ApplGeophys,152,667-683.

Laporte, D. & Watson, E.8., 1995. Experimental andtheoreticai constraints on melt distribution in crustalsources - the effect of crystalline anisotropy on meltinteconnectivity, Chem Geol, 124, I 6 1-l 84.

Litasov, K.D., 2011. Physicochemical conditions {ormelting in the Earth's mantle containing a C-O Hfluid (from experimental datal, Russ Geol Ceophys,5t, 475-492.

Litasov, K.D., Shatskiy, A. & Ohtani, E.,2073a. Meltingphase relations of peridotite and eclogite coexistingwith reduced C-O-H lluid at 3-16 GPa, Earth PLanet

Sci Lett, submitted.

Litasov, K.D., Kagi, H., Shatskiy, A., Ohtani, E., Laksh-tanov, D.L., Bass, f.D. & Ito, E., 2007. High hydrogensolubiiity in Al-rich stishovite and water transport inthe lower mantle, Earth Planet Sci Lett,262,620-634.

Litasov, K.D. & Ohtani, E., 2002. Phase relations andmelt compositions in CMAS-pyrolite-HrO system upto 25 GPa, Phys Earth PLanet Inter, I34, LO5-127 .

Litasov, K.D., Ohtani, 8., Langenhorst, I., Yurimoto,H., Kubo, T. & Kondo, T., 2003. Water solubilityin Mg-perovskites, and water storage capacity in thelower mantle, Earth Planet Sci Lett,2ll, 189-203.

Litasov, K.D., Ohtani, E., Suzuki, A., Kawazoe, T. &Funakoshi, K., 2004. Absence of density crossoverbetween basalt and peridotite in the cold slabs passingthrough 660km discontinuity, Geophys Res Lett,3l,L24607, 24610.21029 12200491021306.

Litasov, K.D. & Ohtani, E., 2005. Phase relations inhydrous MORB at l8 28GPa: implications for het-erogeneity of the lower mantle, Phys Earth PLanet

Inter, I50, 239-263.

Litasov, K.D. & Ohtani, E., 2007. E{fect of water onthe phase relatlons in Earth's mantle and deep watercycle. In E. Ohtani led.), Advances in High-PressureMinerdlogy, Geological Society of America SpecialPapers, pp. 115 156.

Litasov, K.D. & Ohtani, E., 2008. Systematic study o{hydrogen incorporation into Fe-bearing wadsleyrteand water storage capacity of the transition zone,Pr oceedings of Sth International W orkshop on W aterDynamics, Americanlnstitute of Physics Conf erenceProceeding,s, 987, 1 13-1 18.

Litasov, K.D. & Ohtani, E.,2009a. Solidus and phaserelations of carbonated peridotite in the system CaO-AlrOu-MgO-SiOr-NarO-CO2 to the lower mantledepths, Phys Earth Planet Inter, 177, 46-58.

Litasov, K.D. & Ohtani,E.,2O09b. Phase relations in theperidotite-carbonate-chloride system at 7.0-15.5 GPaand the role oi chlorides in the origin of kimberliteand diamond, Chem Geol, 262, 29-41.

Litasov, K.D. & Ohtani, E., 2010. The soiidus o{ carbon-ated eclogite in the system CaO-AlrOu-MgO-SiOr-NarO-CO, to 32GPa and carbonatite liquid in thedeep mantle, Earth Planet Sci Lett,295, 115-126.

Litasov, K.D., Ohtani, E. & Sano, A.,2006.Influence ofwater on major phase transitions in the Earth's man-tie. In S.D. Jacobsen & S. van der LeeledsJ, EarthDeep

62 KONSTANTIN D. LITASOV/ ANTON SHATSKIY AND TIII oHTANI

Water Cyde, American Geophysical Union Geophys-ical Monograph, Washington DC.

Litasov, K.D., Ohtani, E., Sano, A., Suzuki, A.& Funakoshi, K., 2005. Wet subduction velsuscold subduction, Geophys Res Lett, 32, LI3312,13310.11029 I 1200591022921 .

Litasov, K.D., Shatskiy, A., Ohtani, E. & Katsura, T.,2011a. Systematic study of hydrogen incorporationinto Fe-{ree wadsleyite, Phys Chem Minera}, 38,7s-84.

Litasov, K.D., Shatskiy, A., Ohtani, E. & Yaxley, G.M.,2013b. The solidus of alkaline carbonatite in the deepmantle, Geology, 4L, 79*82.

Litasov, K.D., Shatskiy, A.F., Palyanov, Y.N., Sokol,A.G., Katsura, T. & Ohtani,l,., 2009. Hydrogen in-corporation into forsterite in MgrSiOo-KrMg(C03 )2-HrO and MgrSiO4-H2O-C at7.5-14.OCPa, Russ GeolGeophys, 50, 1 129-1 138.

Litasov, K.D., Shatskiy, A.F. & Pokhilenko, N.P.,2011b. Phase relations and melting in the systems ofperidotite-HrO-CO, and eclogite-HrO-CO, at pres-sures up to 27 GPa, Dok Earth Sci, 437, 498-502.

Liu, Q. & Lange, R.A., 2003. New density measurementson carbonate liquids and the partial molar volume ofthe CaCO. component, Contrib Mineral Petrol, 146,3 70-38 1.

Liu, X., O'Neili, H.S.C. 8rBerry, A.1.,2006. The e{fects ofsmall amounts of HrO, CO, and NarO on the partialmelting of spinel lherzolite in the system'CaO-MgO-A12O3-SiO2+HrOfCOr*NarO at 1.1GPa, I Petrol,47, 409-434.

Maruyama, S., Hasegawa, A., Santosh, M., Kogiso, T.,Omori, S., Nakamura, H., Kawai, K. &Zhao, D.,2009.The dynamics of big mantle wedge, magma factory,and metamorphic-metasomatic {actory in subductionzones, Gond Res,16, 474-430.

Maruyama, S. & Liou, J.G., 2005. From snowball toPhancorozic Earth, Int Geol Rev, 47 , 775-79L

Matsyuk, S.S. & Langer, K., 2004. Hydroxyl in olivinesfrom mantle xenoliths in klmberlites of the Siberianplatform, Contrib Mineral Petrol, !47, 413-437.

Matveev, S., Ballhaus, C., Fricke, K., Truckenbrodt, f. &Ziegenbein, D., 7997. Volatiles ln the Earth's mantle.1. Synthesis of CHO-fluids at 7273 K and 2.4GPa,Geochim C osmochim Acta, 61, 308 1-308B.

Matveev, S. 8r Stachel, -1., 2007. FTIR spectroscopy ofOH in olivine: A new tool in kimberlite exploration,Geochim Cosmochim Acta, 7 l, 5528-5543.

McKenzie, D. & Bickle, M.J., 1988. The volume andcomposition of melt generated by extension o{ thelithosphere, I Petrol, 29, 625-679.

McKenzie, D., fackson, J. & Priestley, K., 2005. Ther-mal structure of oceanic and continental lithosphere,Earth Planet Sci Lett,233, 337-349.

Mei, S. & Kohlstedt, D.L., 2000a. Influence of water onplastic deformation of olivine aggregates 7. Diffusioncreep regime, I Geophys Res, 105, 21457-21469.

Mei, S. & Kohlstedt, D.L., 2000b. Influence o{ water onplastic de{ormation of olivine aggreg^tes 2. Disloca-tion creep regime, I Geophys Res, 105, 2147I-21481.

Minarik, W.G. & Watson, 8.8., Iggs.Interconnectivityof carbonate melt at low melt fraction, Earth PlanetSci Lett, 133, 423-437.

Mltchell, R.H., 2005. Carbonatites and carbonatites andcarbonatites, C anad Miner al, 43, 2049-2068.

Mookherjee, M. 8d Karato, S., 2010. Solubility of wa-ter in pyrope-rich garnet at high pressures andtemperature, Geophys Res Lett, 37, L03310, doi:03 3 1o.0 1029 I 02009 g10 4 1289 .

Morard, G., Sanloup, C., Fiquet, G., Mezouar, M., Rey,N., Poloni, R. & Beck, P., ZO07 . Structure of eutecticIe-FeS melts to pressures up to 17GPa: Implica-tions {or planetary cores, Earth Planet Sci Lett,263,128-139.

Nakajima, Y., Takahashi, E., Suzuki, T. & Funakoshi,K., 2009. "Carbon in the core" revisited, Phys EarthPlanet Inter, 17 4, 202-211.

Nishihara, Y., Shinmei, T. & Karato, S., 2006. Grain-growth kinetics in wadsleyite: Effects of chemicalenvironment, Phys Earth Planet Inter , 154, 30-43 .

O'Reilly, S.Y. & Grifffn, W.L., 1985. A xenolith-derivedgeotherm {or Southeastern Australia and its geophys-ical implications, Tectonophy sics, 1ll, 47 -63.

Ohtani, E., Mizob ata, H. & Yurimoto, H., 2000. Stabilityof dense hydrous magnesium silicate phases in thesystems MgrSiO4-H2O and MgSiOr-HrO at pressuresup to 27 GPa, Phys Chem MineraL,27, 533-544.

Ohtani, E. & Zhao, D., 2009. The role of water in thedeep upper mantle and transition zone: dehydrationof stagnant slabs and its effects on the big mantlewedge, Russ Geol Geophys, 50, 1073-1078.

Okamoto, K. & Maruyama, S., 2004. The eclogite-gametite transformation in the MORB+HrO system,Phys Earth Planet Inter, 143-144,283-296.

Palyanov, Y.N., Borzdov, Y.M., Khokhryakov, A.F.,Kupriyanov, I.N. & Sobolev, N.V., 2005. Sulffdemelts-graphite interaction at HPHT conditions: Im-plications for diamond genesis, Earth Planet Sci Lett,250,269-280.

Earth's Mantle Melting in the presence of C-O-H-Bearing F\uid 63

Palyanov, Y.N., Kupriyanov, I.N., Borzdov, y.M., Sokol,A.C. & Khokhryakov, A.F., 2009. Diamond crys-tallization from a sulfur-carbon system at HPHTconditions, C ry st G r owth D e s ign, 9, 29 22-29 26.

Palyanov, Y.N. & Sokol, A.G., 2009. The effect ofcomposition of mantle fluids/melts on diamond {or-mation processes, Lithos, lL2, 690-700.

Palyanov, Y.N., Sokol, A.G., Borzdov, y.M. &Khokhryakov, A.F., 2002. Fluid-bearing alkaline car-bonate melts as the medium for the formation ofdiamonds in the Earth's mantle: an experimentalsttdy, Lithos, 60, 745-759.

Palyanov, Y.N., Sokol, A.G., Borzdov, y.M.,Khokhryakov, A.F. & Sobolev, N.V., 1999. Diamondformation from mantle carbonate fl:uids, Nature,400,417-418.

Palyanov, Y.N., Sokol, A.G. & Sobolev, N.V., 2005.Experimental modeling of mantle diamond-formingprocesses/ Russ Geol Geophys,46, IZTI-IZB4.

Peacock, S.M. & Wang, K., 1999. Seismic consequencesof warm versus cool subduction metamorphism:Examples from southwest and northeast Japan, Scj-ence, 286, 937-939.

Po1i, S. & Schmidt, M.W.,2002. petrology of subductedslabs, Anz Rev Earth Planet Sci, 30, 207 -ZBS .

Priestley, K. & McKenzie,D.,2006. The thermal struc-ture of the lithosphere from shear wave velocities,EarLh ?lanet Sci t ett,244, 285-301.

Putirka, K.D., 2005. Mantle potential temperatures atHawaii, Iceland, and the mid-ocean ridge system,as inferred {rom olivine phenocrysts: Evidence forthermally driven mantle plumes, Geochem GeophysGeosys, 6, Q05108, doi fi .fO29 12005gc000915.

Riley, G.N., Kohlstedt, D.L. & Richter, F.M., 1990.Melt migration in a silicate liquid-olivine system - anexperimental test o{ compaction theory, Geophys ResLett, 17,2101-2104.

Ringwood, A.E., 1979. Origin of the Earth and Moon.Berlin: Springer-Verlag.

Rohrbach, A., Ballhaus, C., Golla-schindler, U., Ulmer,P., Kamenetsky, V.S. & Kuzmin, D.V., 2007. Metalsaturation in the upper mantle, Nature,449,456-459.

Rohrbach, A. & Schmidt, M.W., 2011. Redox freezingand melting in the Earth's deep mantle resulting {romcarbon-iron redox coupling, N atur e, 47 2, 2Og -212.

Sakamaki, T., Ohtani, E., Urakawa, S., Terasaki, H. &Katayama, Y., 201 i . Density of carbonated peridotitemagma at high pressure using an X-ray absorptionmethod, Amer Mineral, 96, 553-557.

Saxena, S.K. & Fei, y., 1987. High-pressure and high_temperature fluid fugacities, Geochim CosmochjmActa, 51,783-791.

Schmidt, M.W. & Poli, S., 1998. Experimentally basedwater budgets {or dehydrating slabs and consequences{or arc magma generation, Earth planet Sci Lett, 163,361-379.

Scott, H.P., Williams, Q., & Knittle, E., 2001. Stabilityand equation of state o{ Fe.C to 78 Gpa: Implicatronsfor carbon in the Earth,s core, Geophy Res Lett, Zg,1875-1878.

Shatskiy, A., Katsura/ T., Litasov, K.D., Shcherbakova,A.V., Borzdov, Y.M., Yamazakr, D., yoneda, A.,Ohtani, E. & Ito, E., 2011. High pressure genera-tion using scaled-up Kawai-cell, phys Earth planetInter, in press.

Shatskiy, A., Litasov, K.D., Borzdov, y.M., Katsura,T., Yamazal<i, D. & Ohtani, 8., 2012. Dissolution-precipitation as a possible mechanism of C-O-Hbearing fluid/melt segregation in the upwelling man-tle Phys Earth Planet Inter, submitted.

Shatskiy, A., Litasov, K.D., Matsuzaki, T., Shinoda, K.,Yamazaki, D., Yoneda, A., Ito, E. &Katsura, T.,2O0g.Single crystal $owth of wadsleyite, Amer Mineral,94,1130-1136.

Shcheka, S.S., Wiedenbeck, M., Irost, D.f. & Keppler,H., 2006. Carbon solubility in mantle minerals, Eart.hPlanet Sci Lett, 245, 730-7 42.

Shirasaka, M. 8r. Takahashi, E., 2003. A genesis of car-bonatitic melt within subducting oceanic crust: highpressure experiments in the system MORB_CaCO.,Bth International Kimbeilite Confercnce Long Ab_str act I Yictotia, Canada, 1-5.

Sobolev, A.V., Hofmann, A.W., Kuzmin, D.V., yaxley,G.M., Arndt, N.T., Chung, S.L., Danyushevsky, L.V.,Elliott, T., Frey, F.A., Garcia, M.O., Gurenko, A.A.,Kamenetsky, V.S., Kerr, A.C., Krivolutskaya, N.A.,Matvienkov, V.V., Nikogosian, I.K., Rocholi, A., Sig-urdsson, I.A., Sushchevskaya, N.M. & Teklay, M.,2007. 'fhe amount of recycled crust in sources ofmantle-derived melts, Science, Z16, 412-417.

Sobolev, N.V., Yeffmova, E.S. & pospelova, L.N., 19g1.Native iron in Yakutian diamonds and its paragene-sis, Russ GeoJ Geophys,22,25-29.

Sokol, A.G., Palyanov, Y.N., Kupriyanov, I.N., Litast_rv,K.D. & Polovinka, M.P., 2010. Ef{ect of oxygen fu-gacity on the HrO storage capacity o{ forsterite inthe carbon-saturated system s, Geochim C osmochimActa,74, 4793-4806.

6-i KoNSTANTIN D. LITASoV, ANTON sHATSKIv AND rrlr oHTANI

Sokol, A.G., Palyanov, Y.N., Palyanova, G.A.,Khokhryakov, A.F. & Borzdov, Y.M., 2001. Diamondand graphite crystallization from C-O-H fluids underhigh pressure and high temperature conditions, DraRelat Mater, 10, 2131-2136.

Sokol, A.G., Palyanov, Y.N., Palyanova, G.A. &Tomilenko, A.A., 2004. Diamond crystallizatron influid and carbonate-fluid systems under mantle P-Tconditions: I. Fluid Comp Geochem [nt,42,830-838.

Sokol, A.G., Palyanova, G.A., Palyanov, Y.N.,Tomilenko, A.A. & Melenevsky, V.N., 2009. Iluidregime and diamond formation in the reduced man-tle: Experimental constraints, Geochim CosmochimActa, 73, 5820*5834.

Stacey, F.D. & Davis, P.M., 2008. Physics of the Earth,4th edition. Cambridge University Press, Cambridge,UK.

Stachel, T., Brey, G.P. & Harris, f.W., 2000. Kankandiamonds (Guinea) I: from the lithosphere down tothe transition zone, Contrib Mineral PetroJ, 140,1-15.

Stagno, V. & Frost, D.1., 2010. Carbon speciation in theasthenosphere: Experimental measurements of theredox conditions at which carbonate-bearing meltscoexist with graphite or diamond in peridotite as-

semblages, Earth Planet Sci Lett,300,72-84.Staudigel, H., Hart, S.R., Schmincke, H.U. & Smith,

B.M., 1989. Cretaceous ocean crust at'DSDP site-417and site-4l8 - carbon uptake from weathering versusloss by magmatic outgassing, Geochim CosmochimActa,53,3091 3094.

Stern, R.|., 2002. Subduction zones, Rev Geophys,40,1012, doi: 1010.1029 1200119000108.

Stevenson, D.J., 1986. On the role of sur{ace-tention inthe migration of melts and fluids, Geophys Res Lett,r3,1149-1152.

Stixrude, L.,2007. Seismic properties of rocks and min-erals, and structure of the Earth. In G. Schubert (ed.),

Treatise on Geophysics, v.2, Elsevier, pp.7-32.Syracuse, E.M., van Keken, P.E. & Abers, G.A., 2010.

The global range o{ subduction zone thermal models,Phys Earth Planet Inter,183, 73-90.

Taylor, W.R. 8{ Green, D.H., 1988. Measurement ofreduced peridotite C-O-H solidus and implications {orredox melting of the mantle, Nature,332, 349-352.

Treiman, A.H. & Schedl, A., 1983. Properties o{ car-bonatite magma and processes in carbonatite magmachambers, I Geo1, 91, 437-447.

Tonegawa, T., Hirahara, K., Shibutani, T., Iwamori, H.,Kanamori, H. & Shiomi, K., 2008. Water flow to

the mantle transition zone inferred from a receiverfunction image of the Paciffc slab, Earth Planet SciLett,274, 346-354.

van der Lee, S., Regenauer-Lieb, K. 8r. Yuen, D.A., 2008.The role of water in connecting past and futureepisodes of subduction, Earth Planet Sci Lett,273,1s-27.

van Keken, P.E., Hacker, B.R., Syracuse, E.M. &Abers, G.A., 2011. Subduction {actory: 4. Depth-dependent flux of HrO {rom subducting slabs world-wide, / Geophys Res-Solid Earth, 116, 1012, doi:I0I0.1029 I 2001rg000 108.

van Keken, P.E., Kiefer, B. & Peacock, S.M., 2002. High-resolution models o{ subduction zones: Implicationsfor mineral dehydration reactions and the transportof water into the deep mantle, Geochem CeophysGeosys, 3, 1056, doi: I 0 I 0. 1 029/2001 gc000256.

Vickers, W. & Greenffe, P., 1967" Dif{usion-creep inmagnesium alloys, I Nucl Mater,24,249-260.

Wallace, M.E. & Green, D.H., 1988. An experimentaldetermination o{ primary carbonatite magma compo-sition, N atur e, 335, 3 43 -Z 46.

Walter, M.|., Bulanova, G.P., Armstrong, L.S., Keshav,S., Blundy, f.D., Gudffnnsson, G., Lord, O.T., Lennie,A.R., Clark, S.M., Smith, C.B. & Gobbo, L., 2008. Pri-mary carbonatite melt from deeply subducted oceanicctust, N atute, 454, 622-630.

Wark, D.A. & Watson, E.B., 1998. Grain-scale perme-abilities of texturally ecluilibrated, monomineralicrocks, Earth Planet Sci Lett,164,591,605.

Watson, E.B., 1982. Melt inffltration and magma evolu-tion, G eolo gy, 10, 23 6 -240.

Weyl, P.K., 1959. Pressure solution and the force o{crystallization-a phenomenological theory, I Geo-phys Res, 64, 2007-2025.

Woodland, A.B. & Koch, M., 2003. Variation in oxygenfugacity with depth in the upper mantle beneath theKaapvaal craton, South Afr Eatth Planet Sci Lett, 214,295-310.

Wyllie, P.1., 1978. Mantle fluid compositions bu{iered inperidotlte-COr-HrO by carbonates, amphibole, andphlogopite, I GeoJ,86, 687-713.

Wyl1ie, P.1., 1987. Discussion of recent papers on car-bonated peridotite, bearing on mantle metasomatismand magmatisrr, Earth Planet Sci Lett,82, 391-397.

Wyllie, P.J. & Huang, W.L., 1976. Carbonation andmelting reactions in system CaO-MgO-SiOr-CO, atmantle pressures with geophysical and petrologicalapplications, Contrib Mineral Petrol, 54, 79-107 .

Earth's Mantle Melting in the Presence of C-O-H-Bearing Fluid 65

Wyllie, P.J. & Ryabchikov, I.D., 2000. Volatile com-ponents/ magmas/ and critical fluids in upwellingmantle, I Peftol, 41,7195-1206.

Yoshino, T., Laumonier, M., Mclsaac, E. & Katsura,T., 2010. Electrical conductivity of basaltic and car-bonatite melt-bearing peridotites at high pressures:Implications for melt distribution and melt fractionin the upper mantle, Earth Planet Sci Lett, 295,s93-602.

Zhang, C. & Duan, 2.H., 2009 . A model for C-O-H fluidin the Earth's mantle, Geochim Cosmochim Acta,73,2089-2102.

Zhao, D.P. & Ohtani, E.,2009. Deep slab subductionand dehydration and their geodpramic consequences:Evidence from seismology and mineral phys ics, GondRes,16, 401-413.

Zhu, W.L., Gaetani, G.A., Fusseis, F., Montesi, L.G.J.& De Carlo, F., 2011. Mlcrotomography of partiallymolten rocks: Three-dimensional melt distributionin mantle peridotite, Science,332, 8B-91.