Very high-pressure orogenic garnet peridotitesJ. G. Liou*, R. Y. Zhang, and W. G. ErnstDepartment of Geological and Environmental Sciences, Stanford University, Stanford, CA 94305

Edited by Russell J. Hemley, Carnegie Institution of Washington, Washington, DC, and approved January 4, 2007 (received for review August 23, 2006)

Mantle-derived garnet peridotites are a minor component in many very high-pressure metamorphic terranes that formed during con-tinental subduction and collision. Some of these mantle rocks contain trace amounts of zircon and micrometer-sized inclusions. Theconstituent minerals exhibit pre- and postsubduction microstructures, including polymorphic transformation and mineral exsolution.Experimental, mineralogical, petrochemical, and geochronological characterizations using novel techniques with high spatial,temporal, and energy resolutions are resulting in unexpected discoveries of new phases, providing better constraints on deep mantleprocesses.

Data on the composition of thesubcontinental lithosphericmantle are essential for erect-ing realistic large-scale models

of the Earth’s geochemical and tectonicevolution (1). Our knowledge of mantlecomposition and petrochemical pro-cesses has been derived mainly fromstudies of xenoliths and xenocrysts inkimberlites, mantle-derived volcanicrocks, and experimental very high-pressure (VHP) phase equilibria, andfrom the interpretation of seismic tomo-graphic images. Recent studies of oro-genic peridotites provide additionalinsights regarding upper mantle pro-cesses at convergent lithospheric plateboundaries. It was found that many oro-genic peridotites were derived from adepleted, metasomatized mantle orcrustal cumulate, and later weresubjected to subduction-zone VHPmetamorphism (e.g., refs. 2–6). Someperidotites preserve a record of ultra-deep origin revealed by mineral ex-solution and the persistence of VHPpolymorphs (6–14), and several perido-tites contain dense hydrous magnesiansilicates (DHMS) that are stable only atmantle depths (15, 16). It was alsofound that some garnet peridotites, andtheir host continental crust, underwentcoeval subduction-zone VHP metamor-phism under pressure–temperature(P–T) conditions characterized by lowthermal gradients (�5°C/km), based onsensitive high-resolution ion microprobe(SHRIMP) U–Pb ages of zircon sepa-rates from both rock types (e.g., refs.17–20). Furthermore, VHP experimentshave revealed that numerous hydrousphases and nominally anhydrous miner-als containing substantial amounts ofH2O are stable under such conditions.Therefore, cold subduction zones arethe principal sites of H2O recycling backinto the mantle (for reviews, see refs. 21and 22). Such findings have advancedour knowledge of the thermal structureof subduction zones and of the recyclingof volatiles into the mantle.

These petrochemical findings lead tonew challenges posed by critical tectonicquestions: How were deep-seated (�200km) mantle rocks transported to shallowdepths? How were such peridotites in-corporated into subduction-zone oro-gens? How can we distinguish thepetrochemical/geochronological pro-cesses taking place in a mantle wedgesetting from those affecting deeply sub-ducted ultramafic rocks of the continen-tal lithosphere?

In the spirit of synergy of 21st centuryscience and technology, this article pre-sents an overview of VHP metamorphismof garnet peridotites and poses new chal-lenges for petrochemical and experimentalstudies of mantle-derived orogenic perido-tites. Specifically, we describe differencesin petrochemical features for mantle-wedge and subduction-zone processesthrough examination of micrometer-sizedminerals, exsolution textures, and poly-morphic transformations. A recent studyof garnet nodules in the Western GneissRegion of the Norwegian Caledonides (6)indicates that the interpretation of conti-nental subduction depths �200 km forsome VHP terranes may be incorrect, in-asmuch as the deep-mantle origin of theperidotites occurred before emplacementin the subduction zone. In the followingdiscussion, except for a few specific exam-ples, we focus mainly on our own pub-lished and unpublished research in theDabie–Sulu terrane of east-central China.

VHP MetamorphismPhysical Conditions of Metamorphism. Sincethe initial discoveries of coesite in su-pracrustal rocks (23, 24), VHP meta-morphism has become synonymous withthat portion of eclogite-facies conditionswithin the P–T stability field of coesite.Understanding VHP tectonics is viewedas a significant undertaking of consider-able importance, as underscored by theabundance of recent task groups, work-shops, conference sessions, and booksdevoted to the subject. VHP metamor-phism refers to the transformation ofcrustal rocks to assemblages containing

the index minerals coesite and/or dia-mond at a minimum P � 2.7 GPa atT � 600°C (Fig. 1); such metamorphismis now well recognized in the geologiccommunity (25, 26). The discovery oftracts of upper continental crust meta-morphosed under VHP conditions hasrevolutionized our understanding of col-lisional orogenic belts. The subductionof sialic materials to mantle depths playsa crucial role in crust–mantle interac-tions at convergent plate junctions. Oneof the most significant orogenic pro-cesses is the formation and subsequentexhumation of VHP rocks subducted todepths of 150 km or more. Several newVHP terranes (Fig. 2) have recentlybeen identified on the basis of partiallypreserved trace index minerals (e.g.,coesite with or without diamond) instrong containers such as zircon and/orgarnet.

In Situ VHP Metamorphism. The volumetri-cally predominant rocks of VHP ter-ranes are felsic gneisses and schists,many of which lack obvious evidenceof mantle-depth metamorphism. Recentobservations (27) indicate that not allgarnet peridotites and eclogites arefault-bounded, as was previouslythought; some such VHP rocks preserveevidence that their contacts withgneissic rocks have retained structuralcoherence throughout subduction, meta-morphism, and exhumation. Mineralogi-cal indicators of VHP metamorphismhave been found in a variety of wallrock lithologies, including gneisses,

Author contributions: J.G.L. and R.Y.Z. designed research;J.G.L. and R.Y.Z. performed research; W.G.E. analyzed data;and J.G.L., R.Y.Z., and W.G.E. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Abbreviations: DHMS, dense hydrous magnesian silicates;P–T, pressure–temperature; REE, rare earth element;SHRIMP, sensitive high-resolution ion microprobe; TEM,transmission electron microscopy; VHP, very high pressure.

*To whom correspondence should be addressed. E-mail:liou@pangea.stanford.edu.

© 2007 by The National Academy of Sciences of the USA

9116–9121 � PNAS � May 29, 2007 � vol. 104 � no. 22 www.pnas.org�cgi�doi�10.1073�pnas.0607300104

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 9,

202

0

quartzites, and marbles (27–30). De-tailed studies of mineral compositionsin Dabie felsic gneisses and schists showthat they were metamorphosed togetherwith intercalated coesite-bearing eclogiteand garnet peridotite bodies under simi-lar P–T conditions.

Evidence of mantle-depth meta-morphism is typically preserved as raremineral inclusions and relict phase as-semblages within host rocks that laterreequilibrated under crustal conditions.Among the various types of evidence,zircons from VHP rocks provide themost useful information with regard

to the P–T time path of a subductioncomplex, inasmuch as this mineral isextremely stable and resistant overa wide range of conditions. Duringgrowth stages, individual zircon zonaldomains may include and preserve in-clusions of minerals in equilibrium withthe matrix phase assemblage. For in-stance, zircons that crystallized at man-tle depths in equilibrium with garnetdisplay characteristic heavy rare-earthelement (REE) depletions and lack anEu anomaly, whereas those grown atcrustal depths in equilibrium with pla-gioclase have pronounced heavy REE

enrichments and a marked negative Euanomaly (31, 32). Consequently, identifi-cation of mineral inclusions and charac-terization of REE patterns of zonedzircons have been used in conjunctionwith ion microprobe U–Pb dating toelucidate the P–T time paths for someVHP terranes (e.g., refs. 33 and 34).

New isotopic ages support the hypoth-esis that Dabie–Sulu eclogites, garnetperidotites, and the surrounding wallrocks were subjected to coeval VHPmetamorphism at 220–240 Ma. Meta-morphic overgrowths on zircons fromeclogites and country rock gneisses andschists yield virtually identical U–PbTriassic ages (e.g., refs. 19, 35, and 36),demonstrating that all units were meta-morphosed at the same time. Zirconseparates from Dabie–Sulu VHP rocksretain low-P mineral-bearing inheritedcores, VHP mineral-bearing (e.g., coes-ite) mantles, and rims that containlow-P minerals such as quartz and pla-gioclase (37, 38). Ion microprobe U–Pbanalyses of these zoned zircons haveidentified three discrete age groups,shown schematically in Fig. 1: (i) thelatest Proterozoic protolith ages (�680Ma) in the inherited cores, (ii) a culmi-nating VHP metamorphic event in thecoesite-bearing mantles at 220–240 Ma,and (iii) a late amphibolite-facies retro-gressive overprint in rims at 210 � 10Ma. The presence of anomalously low�18O in VHP minerals, not only in coes-ite-bearing eclogites but also in the wallrocks, suggests that both Dabie–Sulu

Fig. 2. Distribution and peak metamorphic ages of recognized VHP terranes worldwide (modified after ref. 27).

0

> 680 Ma

Coesite

Quartz

Diamond

Graphite

GR

HGR

AMGS

Ep-EC Amp-EC

Dry EC

40

80

120

160

200 400 600 800 1000

Pre

ssur

e [G

Pa]

Dep

th k

mTemperature [ C]

1.0

2.0

3.0

4.0

5.0

6.0

BS

o

EA

Lw-E

C5 C/k

m

o

210 ± 4 Ma

> 680 Ma

231±4 Ma

211± 4 Ma

Protolith age: > 680 Ma Qtz-bearing inherited coreUHP metamorphic age: 231±4 Ma Coe-bearing mantleRetrograde metamorphism: 211±4 Ma Qtz-bearing rim

Exhumation rate: > 5 km/Ma

Forb

idde

n zo

ne

Peridotite

Eclogite

(B)

231± 4 Ma

S

S

G

G

GG

(A)

(A)

(A)

Dab

ie-S

ulu

Fig. 1. P–T conditions of VHP mafic–ultramafic rocks. (Left) (A), P–T fields of VHP metamorphism,‘‘forbidden-zone’’ (17), and stability of coesite and diamond; (B), P–T time paths for Dabie–Sulu eclogiteand garnet peridotites. (Right) Zoned zircon domains with SHRIMP U–Pb ages for Sulu paragneiss.

Liou et al. PNAS � May 29, 2007 � vol. 104 � no. 22 � 9117

SP

EC

IAL

FE

AT

UR

E:

PE

RS

PE

CT

IVE

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 9,

202

0

mafic–ultramafic and felsic rocks re-mained in mutual contact throughoutsubduction and that the entire terraneunderwent Triassic VHP closed-systemmetamorphism (e.g., refs. 39 and 40).

Mantle-Derived and Crust-Hosted GarnetPeridotites. Garnet-bearing ultramaficrocks are widespread as a minor butsignificant component of the Dabie–Sulu VHP terrane (e.g., ref. 41). Al-though most surface exposures areheavily serpentinized, fresh samplesfrom quarries and drill-hole cores of theChinese Continental Scientific DrillingProject consist of garnet lherzolite, harz-burgite with or without minor wehrlite,and dunite. Dabie–Sulu garnet perido-tites are classified as mantle-derived(type A) or crust-hosted (type B) on thebasis of structural, geochemical, and iso-topic characteristics (Table 1 and Fig.3). Type B igneous intrusions occur asminor ultramafic cumulates associatedwith dominant metagabbroic layers ofvarious compositions, whereas type Aperidotites represent depleted, metaso-matized mantle fragments, some ofwhich contain minor eclogite and garnetclinopyroxenite pods. Members of bothtypes were subjected to Triassic subduc-tion-zone VHP metamorphism; somewere metamorphosed at mantle depthsunder P–T conditions (�5°C/km) involv-ing pressures up to �6.7 GPa and T �950°C (15, 41–43) (Fig. 1).

Morphologies of zoned zircon grains,and SHRIMP U–Pb isotopic analysesobtained from their cores and rims, pro-vide important constraints regardingwhether or not the mantle-derived gar-net peridotites experienced subduction-zone metamorphism. Most zircons fromChinese garnet peridotites are of therounded isometric form without inher-ited cores, implying a metamorphic ori-gin. SHRIMP U–Pb dating of zirconsfrom both the peridotites and the en-closed eclogite lenses in the Sulu regionhas yielded VHP metamorphic ages of220–240 Ma (18–20, 44); the data areconsistent with VHP ages of 230 � 10Ma for the country rocks. However, re-connaissance study of Hf isotopic andU–Pb upper-intercept ages of zirconsyield Early Proterozoic, even Archean,model ages for certain other Sulu garnetperidotites (44), suggesting that somebodies had long resident times in themantle wedge before involvement inthe Triassic subduction.

Microstructures of VHP MineralsExsolution of Mineral Lamellae. Studies ofmicrominerals and exsolution structureshave revealed numerous preserved,deep-seated features formed at muchhigher P than values estimated usingconventional (e.g., Grt–Opx) geobarom-eters (27, 45). The best example may bethe electrifying report of micrometer-sized FeTiO3 rods and plates of chro-

mite in olivine from the Alpe Aramigarnet lherzolite in the Central Alps (7).Dobrzhinetskaya et al. (7) hypothesizedthat the exsolved FeTiO3 lamellae wereoriginally a high-P perovskite polymorphof ilmenite and that the precursor phaseformed at 10–15 GPa (300–450 km).From the abundance, morphology, crys-tallography, and topotaxy of these ox-ides, they argued that the inferred veryhigh solubility of highly charged cations(Ti and Cr) reflects previously unrecog-nized mantle conditions in recoveredrock samples. Despite considerable con-troversy, subsequent experiments underthese conditions have confirmed that �1vol % of TiO2 can be accommodated inolivine (46, 47). The additional observa-tion of exsolved Ca-poor pyroxene dis-playing antiphase domains in diopside(8, 48, 49) also supports the idea of anextremely deep origin (�300 km) forthe Alpe Arami garnet peridotite.

Precursor majoritic garnet was postu-lated based on the identification of py-roxene lamellae exsolved from garnet ina Norwegian orogenic garnet peridotite(9, 10). Multistage processes for forma-tion of the majoritic garnet nodulessubsequently were described for the hy-pothesized �180 km depth of origin ofthe Otroy peridotite (10). Experimentalsimulation of the exhumation path ofmantle material shows that high-T(1,400°C) decompression of lherzolitefrom 14 to 12 GPa results in exsolutionof interstitial blebs of diopside andMg2SiO4 (wadsleyite) lamellae from aparental majoritic garnet (47, 50). Later,research involving critical analyses ofREE concentrations in minerals andother geochemical characteristics, aswell as recalculation of the volume ofexsolved pyroxene inclusions in garnet,led researchers to reinterpret the originof Otroy garnet peridotite (6). Evi-dently, Archean (�3.5 Ga) deep-seatedmantle peridotites containing majoriticgarnets underwent exsolution duringupwelling from a depth of 350 km ormore and were subjected to extensivepartial melting; the residue formed agarnet-bearing cratonic root. Theselithospheric mantle fragments of Pro-terozoic age (�1.8–1.4 Ga) were thenincorporated into subducting sialic crust

-100km

-200km

Continental lithosphere

SN

s

s

s

G

Spinel peridotites

Garnet peridotites

Footwall mantle

Ancient relict mantle

Mafic rocks

Ultramafic rocks

fluid + volatile

Mantle wedge

Crustal

G

G

Footwall m

antle

Continental crust

Oceanic crust

Mantlewedge

Sino-Korean craton Yangtze craton

Fig. 3. A schematic model for Triassic subduction of the Yangtze beneath the Sino-Korean cratons,showing the tectonic setting for mantle-derived (type A) and crustal-hosted (type B) garnet peridotites(for details, see ref. 2).

Table 1. Dual origin of Dabie–Sulu garnet peridotites in Triassic VHP terrane

Characteristic Type A (mantle-derived) Type B (crust-hosted)

Rock type Garnet peridotite � minor eclogite lenses Metagabbroic eclogites � peridotite cumulate layersOrigin Mantle wedge or subducted subcontinent lithosphere Continental crust of the downgoing plateEmplacement Faulted into subduction zone Mafic–ultramafic intrusion into crust prior to subductionAge of protolith Archean to Proterozoic Late Proterozoic to Paleozoic (�700–450 Ma)Age of zircons �220–240 Ma �700–220 MaExtent of crustal metasomatism Minor Substantial

9118 � www.pnas.org�cgi�doi�10.1073�pnas.0607300104 Liou et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 9,

202

0

during the Caledonian continental colli-sion at �400 Ma. The conclusion ofpresubduction exsolution was based ondetailed Re-Os and Nd-Sm age data andon REE concentrations and partitioningbetween exsolved pyroxenes and thehost garnet.

Numerous examples of garnet andclinopyroxene exsolution from Dabie–Sulu peridotites and eclogites have beenhypothesized to have been derived froma majoritic garnet precursor formed atconsiderable mantle depths. Coarse-grained clinopyroxenes from Rizhaogarnet clinopyroxenites contain up to25 vol % exsolved garnet and 4 vol %ilmenite (Fig. 4A). Petrologic and exper-imental studies suggest that the precur-sor of such intergrowths was majoriticgarnet in which Ca2�Ti4� 3 2Al3�,Mg2�Si4� 3 2Al3�, and Na�Ti4 3Ca2�Al3� in octahedral sites (14, 42,43). Exsolved needles of pyroxene,rutile, and apatite along garnet (111)planes from the Yangkou eclogite layerwithin peridotite (11, 45, 51) are basedon optical and SEM observations.

Similar exsolution lamellae, suggestingthat a majoritic garnet precursor formedat depths �200 km, have also beenreported in the Erzgebirge massif ofGermany, another VHP terrane (13).Despite numerous reports of majoriticgarnets, however, formation of VHPsolid-solution phases (e.g., majorite)may have taken place either in the deepupper mantle, and then later been se-questered in the mantle wedge such asin Norway, or in a subduction zone suchas in the Kokchetav of northern Kazak-stan (see below). This problem remainsto be satisfactorily investigated.

Supersilicic titanite was suggested as aprecursor phase of coesite lamellae in ti-tanite from a Kokchetav impure marblethat formed at P � 6 GPa (52). Otherexsolution lamellae of quartz or K-feldspar with or without phengite indiopside from diamond-bearing marbleand gneiss, and quartz exsolution ineclogitic omphacite (53), together with

biologic C-isotope signatures for micro-diamonds (54), allow the conclusion thatsubducted crustal felsic and carbonateprotoliths reached mantle depths �180km. Transmission electron microscopy(TEM) identification of nanometricinclusions of aragonite (CaCO3) andmagnesite (MgCO3) in microdiamonds,together with the experimental stabilityof these carbonates, further suggeststhat the diamond-bearing rocks of theKokchetav massif were subducted to adepth of �190–280 km (55). Appar-ently, exsolution occurred during de-compression/exhumation of VHP rocks.In addition, nanometer-thick (�2 nm)lamellae of �-PbO2-type TiO2 occur be-tween multiple twinned rutile crystals inboth diamond-bearing felsic rocks fromthe Erzgebirge (56) and coesite-bearingDabie eclogites (57). The occurrence ofthese lamellae implies subduction ofcontinental materials to a depth �200km. These observations, togetherwith inferred supersilicic titanite inKokchetav marble, allow the conclu-sion that some continental supercrustalrocks have been subducted to depthsof at least 300 km before being re-turned to the surface. These depths ofmetamorphism demonstrate that coun-try rocks, although perhaps not asdeeply buried as some garnet perido-tites, have ascended from astonishingsubduction depths.

Exsolution of Hydrous Phases. K-bearingpargasite [KCa2(Mg,Fe)4AlSi6Al2O22(OH,F)2] lamellae in clinopyroxene in-clusions within garnet megacrysts, andphlogopite lamellae in lherzolitic diop-side from Sulu, show topotactic inter-growths and are confined to the cores ofthe host clinopyroxene (14, 58). TheseK- and OH-bearing exsolved phases sug-gest that the primary clinopyroxene mayhave incorporated a considerableamount of K2O and H2O under VHPconditions, as documented in clinopy-roxene inclusions in Kokchetav zircons(59). This conclusion is consistent with

VHP experiments demonstrating thesolubility of K in clinopyroxene (60, 61).

Similar exsolution lamellae of rutileplus sodic amphibole plus apatite in gar-net from the Qaidam garnet peridotiteof western China (Fig. 4 B and C) havebeen suggested (62) to represent decom-pression products from depths �200 kmof supersilicic majorite crystals typifiedby high concentrations of Na2O (0.3wt %) and hydroxyl (up to 1,000 ppm).Thus, in addition to DHMS and nomi-nally anhydrous silicates, majoritic gar-net and supersilicic clinopyroxene couldbe important reservoirs of H2O at man-tle depths.

Polymorphic Transformations. Intergrowthsof ortho- and clinoenstatite lamellae arecommon within Chinese orogenic garnetperidotites. Clinoenstatite lamellae inorthoenstatite may have formed eitherby inversion from orthoenstatite or bya displacive transformation from VHPclinoenstatite during decompression (8,12). Experiments indicate that orthoen-statite transforms to VHP clinoenstatiteat P � 8 GPa, 900°C, corresponding to amantle depth of �300 km (e.g., refs. 63and 64) (Fig. 5). The growth of high-Pclinoenstatite in mantle-derived perido-tite, and the inferred majoritic garnetprecursor, may have formed at greatdepth in the mantle wedge long beforeinsertion into the downgoing continentallithospheric plate, then recrystallizedduring subduction-zone metamorphism.

Synergy of VHP Metamorphic Studieswith Mineral PhysicsThe above review concludes that somesections of continental crust havereached subduction depths approachingor exceeding 200 km, involving passive-margin lithologies, including carbonate,felsic, pelitic, and minor mafic–ultra-mafic protoliths. Some peridotites nowhosted in continental orogens may havebeen formed even deeper. However, therecognition of mineral exsolution inmantle-derived garnet peridotites result-

Fig. 4. Microstructures of VHP minerals. (A) Exsolution lamellae of garnet plus ilmenite in relict clinopyroxene from Sulu peridotite (for details, see ref. 14).(B and C) Exsolution lamellae of rutile plus amphibole in peridotite garnet from northern Qaidam VHP terrane (for details, see ref. 62).

Liou et al. PNAS � May 29, 2007 � vol. 104 � no. 22 � 9119

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 9,

202

0

ing from decompression in a mantle-wedge setting, or due to exhumation ina subduction zone, remains to be deter-mined, except in cases for which ageconstraints, such as those in the WesternGneiss Region, are conclusive. With therecent breakthroughs in VHP technol-ogy and new-generation synchrotron,neutron, and laser facilities for charac-terization of nano-sized materials, newin-depth research on VHP minerals androcks is now within reach; such studiesare just beginning (see refs. 65, 66, and83). Three general fields of mineralphysics research are highlighted belowto illustrate some of the opportunities.These fields all have relevance to afuller understanding of crust–mantletectonics.

Identification of Nano-Sized Minerals. Ahost of peculiar minerals and mineralcompositions have been described inthe Western Gneiss Region (38 miner-als in all) (67) and in the Dora Mairamassif (�7 minerals) (68) VHP rocksby conventional methods. Novel tech-niques developed for experimentalstudies with high spatial, temporal, andenergy resolution should be applied formicroanalysis of solid and f luid inclu-sions in naturally occurring VHP min-erals, including unexplored trace butwidespread opaque phases (e.g., Fe-Nisulfides) in garnet peridotites and co-rundum-bearing garnetite (69). Ubiqui-tous micrometer-sized mineral andf luid inclusions in tough, rigid VHPmineral hosts, including zircon, dia-mond, garnet, and pyroxenes (e.g., refs.65 and 70–72), require analytical elec-tron microscopy to characterize theirstructures and compositions. For exam-

ple, kokchetavite, a new hexagonalpolymorph of K-feldspar, was discoveredas a metastable phase together with�-cristobalite plus phengite plus sili-ceous glass with or without phlogopite/titanite/calcite/zircon as multiphasecloudy inclusions in clinopyroxene andgarnet from a diamond-grade Kokchetavgarnet-pyroxene rock (73). This prob-lematic phase (2- to 7-�m-sized plates)can be misidentified as K-feldspar byconventional techniques and misinter-preted as an exsolved phase.

Focused ion beam techniques andTEM studies of microdiamonds fromthe Erzgebirge (74) have revealednumerous nanometric crystalline inclu-sions, including phases of known stoi-chiometries such as SiO2 and Al2SiO5and minerals with different combina-tions of Si, K, P, Ti, Fe, and O2. Thesephases need to be investigated by em-ploying synchrotron radiation. Suchapproaches are only now beginning tobe applied to VHP rocks. For exam-ple, metamorphic diamonds from theErzgebirge have been examined usingsynchrotron infrared absorption, Ra-man scattering, and f luorescence spec-troscopy. The characteristic features ofCOC and COH bonds, molecularH2O, OH� and CO3

2� radicals, and Nimpurities all support the concept ofdiamond crystallization from a COH-rich supercritical f luid (66, 75).

Characterization of Mineral Exsolution andPhase Transformations. Exsolution inter-growths are common in minerals ofdecompressed VHP rocks; however,the exsolution mechanisms are poorlyunderstood. Each lamellae-bearing hostmineral preserves information con-cerning the composition and physicalconditions of formation of the homo-geneous precursor phase, as well as aportion of the inferred P–T path dur-ing decompression. Compositional andstructural characterization of lamellae–host mineral pairs will provide impor-tant new constraints on the physicalconditions of crystallization/recrystalli-zation. Such studies should guide sub-sequent experimentation to delineatethe P–T conditions and mechanismsfor the formation of the primary VHPminerals.

Experimental Phase Relations and Composi-tional Variations. Experimental investiga-tions of the KMASH, CMASH, andKNCMASH systems have revealed thepossibility of occurrences of hydrousphases in VHP pelitic and peridotiticrocks (76–78). Experimental investiga-tion of hydroxyl solubility in clinopyrox-ene, orthopyroxene, and olivine (79, 80)indicates that the OH content of these

minerals is highly dependent on pressureand temperature. Many hydroxyl-bearingphases, such as OH-topaz and phase A,are only stable under mantle P–T condi-tions involving geotherms considerablyless than 5°C/km. Because such condi-tions are essentially transient and inevi-tably are followed by a period ofthermal relaxation, these phases havelittle chance of surviving T increase andP decrease on return to the surfacethrough erosional and tectonic pro-cesses. However, one possibility for thepreservation of these phases is asminute inclusions in high mechanical-strength, impervious containers likegarnet, zircon, or diamond (69).

Interpretation of the occurrence of ma-joritic garnets in orogenic peridotites re-quires better experimental data, and somesyntheses using natural peridotite sampleshave been accomplished (47, 51). For ex-ample, as shown in Fig. 6, the depths forthe formation of majoritic garnet arebased on reconnaissance investigationof the pseudobinary system MgSiO3–Mg3Al2Si3O12 (81). Although experimentsinvolved testing the effects of Fe and Ca,the effect of Ti has not been explored.Solubilities of Ti, K, OH, and other traceelements in olivine, garnet, and pyroxenein mafic–ultramafic systems need to beexamined, inasmuch as natural analogsexhibit numerous microstructures. Thesephases also contain minute inclusions asyet unidentified that may turn out to beVHP phases or DHMS previously synthe-sized only in diamond-cell or multianvilexperiments (e.g., see ref. 82). Discoveryof natural representatives of these syn-thetic phases would be a major step for-ward in understanding mantle processes,

18 A

Pre

ssur

e (G

Pa)

LCLEN(P2 /c)

HCLEN (C2 /c)

MgSiO

OREN(Pbc)a

3

1

9

8

7

6

5

4

3

2

1

XFe=0.1

0 400 800 1200 1600

Temperature ( C)o

Fig. 5. Phase transformations in enstatite. (Upper)TEM image of enstatite with clinoenstatite lamellaefrom Sulu garnet peridotite. (Lower) P–T path forsuch transformation. (For details, see ref. 12.)

Fig. 6. Schematic P–X (Cpx–Grt) diagram for iso-thermal decompression path of majoritic garnet toform garnet plus ilmenite lamellae in clinopyrox-ene host (Fig. 4A) (for details, see ref. 14).

9120 � www.pnas.org�cgi�doi�10.1073�pnas.0607300104 Liou et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 9,

202

0

including the role of hydrous phases asstorage sites for H2O.

Summary StatementThe search for mantle minerals in oro-genic VHP garnet peridotites should beconducted with the same sophisticatedtechniques (microRaman spectroscopy,

synchrotron x-ray diffraction, and high-resolution TEM) used by experimentaliststo identify synthetic analogs. Such studiesare needed to bridge the gap betweenmantle petrology and mineral physics. Areview of recent progress involving suchan experimental approach is detailed byDobrzhinetskaya (83).

We thank Dave Mao, Larissa Dobrzhi-netskaya, Harry Green, and Nick Sobolevfor support and reviews, including reviewof a draft version of this manuscript.This work was supported by Stanford Uni-versity and by National Science Founda-tion Continental Dynamic ProgramGrants EAR 00-03355 and 05-06901 (toJ.G.L.).

1. Griffin WL, O’Reilly SY, Ryan CG (1999) inMantle Petrology: Field Observations and Experi-mentation, eds Fei Y, Berka CM, Mysen BO(Geochemical Society, Houston), pp 13–45.

2. Zhang RY, Liou JG (1998) Episodes 21:229–234.3. Zhang RY, Liou JG, Yang JS, Yui TF (2000) J

Metamorph Geol 18:149–166.4. Zhang RY, Liou JG, Yang JS, Liu L, Jahn BM

(2004) Int Geol Rev 46:981–1004.5. Jahn BM, Rumble D, Liou JG (2003) in Ultrahigh-

Pressure Metamorphism, eds Carswell DA, Com-pagnoni R (Eotvos Univ Press, Budapest), pp365–414.

6. Spengler D, Van Roermund HLM, Drury MR,Ottolini L, Mason RD, Davies GR (2006) Nature440:913–917.

7. Dobrzhinetskaya LF, Green HW, Wang S (1996)Science 271:1841–1845.

8. Bozilov KN, Green HW, Dobrzhinetskaya LF(1999) Science 284:128–132.

9. Van Roermund HLM, Drury MR, Barnhoorn A,De Ronde A (2000) J Metamorph Geol 18:135–147.

10. Van Roermund HLM, Drury MR, Barnhoorn A,De Ronde A (2001) J Metamorph Geol 42:117–130.

11. Ye K, Cong BL, Ye DN (2000) Nature 407:734–736.

12. Zhang RY, Shau YH, Liou JG, Lo CH (2002) AmMineral 87:867–874.

13. Massonne H, Bautsch H (2002) Int Geol Rev44:779–796.

14. Zhang RY, Liou JG (2003) Am Mineral 88:1591–1600.

15. Zhang RY, Liou JG, Cong BL (1995) J Petrol36:1011–1037.

16. Zhang RY, Li T, Rumble D, Yui T-F, Li L, YangJS, Pan Y, Liou JG (2007) J Metamorph Geol25:149–164.

17. Liou JG, Hacker BR, Zhang RY (2000) Science287:1215–1216.

18. Zhang RY, Yang JS, Wooden JL, Liou JG, Li TF(2005) Earth Planet Sci Lett 237:729–734.

19. Zhao R, Liou JG, Zhang RY (2005) Int Geol Rev47:805–814.

20. Zhao R, Liou JG, Zhang RY, Li T (2006) inUltrahigh-Pressure Metamorphism: Deep Continen-tal Subduction, eds Hacker BR, McClelland WC,Liou JG, Geol Soc Am Spec Pap 403, pp 115–125.

21. Ohtani E (2005) Elements 1:25–30.22. Maruyama S, Liou JG (2005) Int Geol Rev 47:775–

791.23. Chopin C (1984) Contrib Mineral Petrol 86:107–

118.24. Smith DC (1984) Nature 310:641–644.25. Coleman RG, Wang XM (1995) in Ultrahigh-

Pressure Metamorphism (Cambridge Univ Press,New York), pp 1–32.

26. Liou JG, Zhang RY (2002) in Encyclopedia ofPhysical Sciences and Technology (AcademiaPress, Ghent, Belgium), 3rd Ed, Vol 17, pp 227–244.

27. Liou JG, Zhang RY, Ernst WG, Rumble D,Maruyama S (1998) Rev Mineral 37:33–96.

28. Sobolev NV, Shatsky VS, Vavilov MA, GoryainovSV (1994) Dokl Akad Nauk 334:488–492.

29. Zhang RY, Liou JG, Shu JF (2002) Am Mineral87:445–453.

30. Carswell DA, Wilson RN, Zhai M (2000) Lithos52:121–155.

31. Hermann J, Rubatto D, Korsakov A (2001) Con-trib Mineral Petrol 141:66–82.

32. Rubatto D, Williams IS, Buick IS (2001) ContribMineral Petrol 140:458–468.

33. Katayama I, Maruyama S, Parkinson CD, TeradaK, Sano Y (2001) Earth Planet Sci Lett 188:185–198.

34. Mattinson CG, Wooden JL, Liou JG, Bird DK,Wu CL (2006) Am J Sci 306:683–711.

35. Hacker BR, Ratschbacher L, Webb L, McWil-liams MO, Ireland T, Calvert A, Dong S, WenkHR, Chateigner D (2000) J Geophys Res105:13339–13364.

36. Liu FL, Xu Z, Liou JG, Song B (2004) J Meta-morph Geol 22:315–326.

37. Liu F, Xu Z, Katayama I, Yang J, Maruyama S,Liou JG (2001) Lithos 59:199–215.

38. Liu F, Xu Z, Liou JG, Katayama I, Masago H,Maruyama S, Yang J (2002) Eur J Mineral 14:499–512.

39. Rumble D, Liou JG, Jahn BM (2003) in Treatise onGeochemistry, eds Rudnick RL, Holland HD,Turekian KK (Elsevier, Oxford), Vol 3, pp 293–319.

40. Zheng YF, Fu B, Gong B, Li L (2003) Earth SciRev 62:105–161.

41. Zhang RY, Liou JG, Cong B (1994) J MetamorphGeol 12:169–186.

42. Zhang RY, Liou JG, Yang JS, Ye K (2003) JMetamorph Geol 21:539–550.

43. Zhang RY, Liou JG, Zheng JP, Yui ZF, GriffinWL, O’Reilly SY (2005) Am Mineral 90:801–813.

44. Zheng JP, Griffin WL, O’Reilly SY, Yang JS, LiTF, Zhang M, Zhang RY, Liou JG Earth PlanetSci Lett, in press.

45. Zhang RY, Liou JG (1999) Int Geol Rev 41:981–993.

46. Dobrzhinetskaya LF, Bozhilov KN, Green HW(2000) Chem Geol 163:325–338.

47. Bozilov KN, Green HW, Dobrzhinetskaya LF(2003) Am Mineral 88:596–603.

48. Green HW, Dobrzhinetskaya LF, Bozhilov KN(1997) Science 278:704–707.

49. Dobrzhinetskaya LF, Green HW, Renfro AP,Bozhilov KN, Spengler D, Van Roermund HLM(2004) Terra Nova 16:325–330.

50. Dobrzhinetskaya LF, Green HW, Renfro AP,Bozhilov KN (2005) in Advances in High-PressureTechnology for Geophysical Applications, eds ChenJ, Wang Y, Duffy TS, Shen G, DobrzhinetskayaLF (Elsevier, Amsterdam), pp 265–287.

51. Zhang RY, Zhai SM, Fei YW, Liou JG (2003)Earth Planet Sci Lett 216:591–601.

52. Ogasawara Y, Fukasawa K, Maruyama S (2002)Am Mineral 87:454–461.

53. Katayama I, Zayachkovsky AA, Maruyama S(2002) in The Diamond-Bearing Kokchetav Massif,Kazakhstan, eds Parkinson CD, Katayama I, LiouJG, Maruyama S (Universal Academy Press, To-kyo), pp 147–156.

54. Sobolev NV, Shatsky VS (1990) Nature 343:742–746.

55. Dobrzhinetskaya LF, Wirth R, Green HW (2006)Earth Planet Sci Lett 243:85–93.

56. Hwang SL, Shen P, Chu HT, Yui TF (2000)Science 288:321–324.

57. Wu X, Meng D, Han Y (2005) Am Mineral90:1458–1461.

58. Chen J, Xu Z (2005) Eur J Mineral 17:895–903.59. Katayama I, Parkinson CD, Okamoto K, Naka-

jima Y, Maruyama S (2000) Am Mineral 85:1368–1374.

60. Luth RW (1997) Am Mineral 82:1198–1209.61. Harlow GE (1997) Am Mineral 82:259–269.62. Song S, Zhang L, Chen J, Liou JG, Niu Y (2005)

Am Mineral 90:814–820.63. Pacalo REG, Gasparik T (1990) J Geophys Res

95:15853–15858.64. Ulmer P, Stalder R (2001) Am Mineral 86:1267–

1274.65. Dobrzhinetskaya LF, Wirth R, Green HW (2005)

Earth Planet Sci Lett 243:85–93.66. Dobrzhinetskaya LF, Liu Z, Cartigny P, Zhang J,

Tchkhetia D, Hemley RJ, Green HW (2006) EarthPlanet Sci Lett 248:325–334.

67. Smith D (1988) in Eclogites and Eclogite-FaciesRocks, ed Smith DC (Elsevier, Amsterdam), pp1–206.

68. Chopin C, Sobolev NV (1995) in Ultrahigh-Pressure Metamorphism, eds Coleman RG, WangXM (Cambridge Univ Press, Cambridge, UK), pp96–131.

69. Zhang RY, Liou JG, Zheng JP (2004) ContribMineral Petrol 146:21–31.

70. Dobrzhinetskaya LF, Wirth R, Green HW (2005)Terra Nova 17:472–477.

71. Hwang SL, Shen P, Chu HT, Yui TF, Liou JG,Sobolev NV, Shatsky VS (2005) Earth Planet SciLett 231:295–306.

72. Hwang SL, Chu HT, Yui TF, Shen PY, Schertl HP,Liou JG, Sobolev NV (2006) Earth Planet Sci Lett243:94–106.

73. Hwang SL, Shen P, Chu HT, Yui TF, Liou JG,Sobolev NV, Zhang RY, Shatsky VS, Zayachk-ovsky AA, Nadezhdina ED (2004) Contrib MineralPetrol 148:380–389.

74. Dobrzhinetskaya LF, Green HW, Weschler M,Darus M, Wang YC, Massonne H, Stochert B(2003) Earth Planet Sci Lett 210:399–410.

75. Dobrzhinetskaya LF, Wirth R, Green HW (2005)Terra Nova 17:472–477.

76. Domanik KJ, Holloway JR (1996) Geochim Cos-mochim Acta 60:4133–4150.

77. Ulmer P, Trommsdorff V (1999) Geochem SocSpec Pub 6, pp 259–281.

78. Dobrzhinetskaya LF, Green HW (2007) J Meta-morph Geol 25:83–96.

79. Bromiley GD, Kepler H (2004) Contrib MineralPetrol 147:189–200.

80. Grant KJ, Kohn SC, Brooker RA (2006) ContribMineral Petrol 151:651–664.

81. Akaogi M, Akimoto S (1977) Phys Earth PlanetInteriors 15:90–106.

82. Bose K, Navrotsky A (1998) J Geophys Res103:9713–9719.

83. Dobrzhinetskaya LF, Wirth R, Green HW, II(2007) Proc Natl Acad Sci USA 104:9128–9132.

Liou et al. PNAS � May 29, 2007 � vol. 104 � no. 22 � 9121

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 9,

202

0