American Mineralogist, Volume 77, pages 784-794, 1992

Experimental investigation of melts from a carbonated phlogopite lherzolite:Implications for metasomatism in the continental lithospheric mantle

Yvns TrrrsAULT.* Ar-aN D. EoclnDepartment of Geology, University of Western Ontario, London, Ontario N6A 5B7, Canada

Frr,rcrry E. Lr,oyoDepartment of Geoscience, University of Reading, Reading RG6 2AB, England

Ansrnlcr

The composition of near-solidus melts formed from a model mantle of carbonatedphlogopite therzolite and phlogopite lherzolite was investigated at 3.0 GPa. At I100 "C,the carbonated phlogopite lherzolite yielded 4 wto/o of alkaline dolomitic melt coexistingwith garnet-rich phlogopite lherzolite, whereas, at 1225 "C, the phlogopite lherzolite pro-duced 7 wo/o of hydrous potassic and calcic silicate melt in equilibrium with titaniferousphlogopite-bearing lherzolite.

The reactivity of these melts with peridotite was determined at 2.0 GPa and 1000 'C.Alkaline dolomitic melts can metasomatize harzburgite to olivine-rich phlogopite wehrlite,and infiltration in wehrlite may produce calcite-bearing phlogopite dunite. The interactionof harzburgite with the hydrous silicate melt results in enrichment in clinopyroxene andphlogopite.

A model of cyclic metasomatism active beneath a continental rift is proposed. A horizonof carbonated phlogopite lherzolite, originally formed at the base of the lithosphere by therelease of dense alkaline fluids from a hot mantle plume, migrates upward by means ofcycles of melting, migrating, solidifying, and reacting as rifting progresses. At a depth of100 km, fractional melting of the carbonated phlogopite lherzolite horizon yields succes-sively an alkaline dolomitic melt and a hydrous potassic and calcic silicate melt. Theinfiltration of these distinct melts into depleted lithospheric mantle ar 65-km depth resultsin a decoupled metasomatic event. First, metasomatism by the dolomitic melt creates atrend from harzburgite to olivine-rich wehrlite. The subsequent infiltration of the silicatemelt enriches the metasomatized rocks in clinopyroxene and phlogopite. The variety ofrocks that result bears similarities with a suite of mantle xenoliths from the West Eifelvolcanic field, Germany.

INtnooucrroN

Carbonatitic and hydrous mafic silicate liquids maysegregate from their mantle sources and migrate by infil-tration even at melt fractions of l0lo or less (e.g., Mc-Kenzie, 1985; Hunter and McKenzie, 1989). Such smallvolumes of melts will probably experience thermal crisis(e.g., Spera, 1984) or be consumed in metasomatic reac-tions during migration into the overlying colder litho-spheric mantle (McKenzie, 1989). These considerationsemphasize the importance of (l) characterizing the com-position of volatile-bearing low-degree partial meltsformed from hydrated and carbonated mantle sources,and (2) investigating the reactivity of these liquids withdepleted peridotitic material.

This paper reports the results of an investigation ofnear-solidus melt fractions formed from phlogopite-bear-ing mantle compositions. Two types of experiments were

* Present address: Department of Geology, Arizona State Uni-versity, Tempe, Arizona 85287, U.S.A.

0003-004x/92l0708-0784$02.00

performed: (l) partial melting of a model mantle of car-bonated phlogopite lherzolite and phlogopite lherzolite;and (2) melt-peridotite interaction experiments on harz-burgitic and wehrlitic materials. The results complementprevious investigations that have dealt with the nature(e.9., Olafsson and Eggler, 1983; Wallace and Green, 1988)and reactivity (e.9., Meen, 1987; Green and Wallace, 1988;Meen et al., 1989) of melts formed from amphibole-bear-ing carbonated peridotitic sources.

ExprnrtrnNTAr- AND ANALvTTCAL TECHNTeuEs

Starting materials

The model mantle compositions (Table l) were pre-pared using mineral separates, Spec Pure CaCO., andFisher Certified NarCOr. Fresh olivine, orthopyroxene,clinopyroxene, and spinel grains were separated from alherzolite xenolith (E48Y; Lloyd et al., l99l) from theWest Eifel volcanic field, Germany. Phlogopite is from asingle xenocryst also sampled from the West Eifel vol-canic field. The dried materials were mixed and ground

784

THIBAULT ET AL.: MELTING OF PHLOGOPITE LHERZOLITE

TABLE 1. Compositions of synthesized starting materials

785

CPL CPL" PIZ Carmet

Chemical composition (wtoA)sio,Tio,Al2o3Cr2ogFeO,,MnOMgoNioCaONaroKrOHrOCo.

TotalXun .

45.070.294.050.696.440 . 1 1

33.920 . 1 16.090.600.580.221.83

100.000.90

44.930.294.060.696.510 .11

34.230 .115.800.600.59o.221.86

100.000.90

47.O40.304.23o.726.720.12

35.400 . 1 14.320.200.610.230.00

r00.000.90

46.940.304.240.726.79o.12

35.760.113.980.190.620.230.00

100.000.90

48.310.011.520.157.430.13

42.O00.15o.280.020.000.000.00

100.000.91

42.330.032.360.657.62o.12

40.96o.225.500.210.000.000.00

100.000.91

1 . 1 60.280.2'l0.004.250 .15

15 .160.00

28.572.943.991 .61

41.68r00.00

0.86

45.811.67

12.',t40.006.510.00

11.750.00

11.292.276.202.360.00

100.000.76

Notej CPL" and PLz'are corrected bulk compositions calculated by subtraction of 10% of the total amount of clinopyroxene. HrO and CO" arecalculated assuming perfect stoichiometry in the carbonated and hydrous minerals used in the synthesis. XMstu represents Mg/(Mg + Fe,").

together under acetone to a grain size of approximatelyl0 pm.

The two model mantle sources have mineral contentsrepresentative of carbonated phlogopite lherzolite (CPL;Table l) and phlogopite lherzolite (PLZ; Table l). Theproportion of olivine incorporated in the mixtures waskept low (x40o/o) in order to enhance the proportion ofthe minor phases, thereby facilitating their identificationand analysis (cf. Green and Ringwood, 1970). The rela-tive abundance ofphlogopite and carbonates in the source,which is essentially related to the amount of COr, HrO,and alkalis, has a significant effect on near-solidus melt-ing relationships (e.g., Holloway and Eggler, 1976). How-ever, no unequivocal samples of natural metasomatizedcarbonate-bearing peridotite, upon which a starting com-position may be chosen, have been reported. This isprobably because of the very rapid decomposition of car-bonates in mantle xenoliths during ascent in their hostmagma (e.g., Canil, 1990). Therefore, the COr, HrO, andtotal alkali contents ofthe CPL source were kept close tothose of the starting material used by Wallace and Green(1988), which allowed a consistent comparison with re-cent experimental results on carbonated amphibole-bear-ing peridotite. However, the KrO-NarO value is signifi-cantly higher in CPL, reflecting a characteristic of manyxenoliths of phlogopite-bearing garnet lherzolite (e.g., Er-lank et al., 1987, their Table IVb).

One model mantle protolith for the experiments onmelt-peridotite interaction was harzburgitic (HAR; Tablel), a composition abundant and widespread in the mantlexenolith population (e.g., Harte and Hawkesworth, 1989).A spinel wehrlite protolith (WHR; Table l) was also pre-pared.

Hydrous silicate melt compositions were synthesizedin a two-stage process. First, mixtures of the requiredamounts of dried SiOr, TiOr, AlrOr, MgO, CaCOr,NarCOr, and KrCO. were melted in air at 1350'C for 30min to remove CO, and quenched to a glass. Then,mixtures of the appropriate amounts of synthesized glass,

natural fayalite (source of FeO), and brucite (source ofHrO) were sealed in Pt capsules and melted in a piston-cylinder apparatus at 1.2 GPa and 1350 "C for 5 min.The products were hydrous glasses whose compositionswere checked by microprobe analyses. Carbonate meltcompositions were synthesized as simple mixtures of driedCaCOr, NarCO, and KrCO. and natural dolomite andbrucite.

Experimental procedures

The pressure-temperature conditions were achieved ina piston-cylinder apparatus and furnace assemblies witha diameter of 1.27 cm (Boyd and England, 1960), usingthe so-called piston-out technique after achieving the de-sired temperature of the experiment. The pressure-trans-mitting medium consisted of a talc-Pyrex sleeve. Agraphite furnace was used, and the temperature was mon-itored using a Pt-PteoRhr. thermocouple. No frictionalcorrection was applied to pressure nor a pressure correc-tion to the emf of the thermocouple. In principle, pres-sure and temperature could be controlled to within +0.05GPa and a 5 'C of the stated values. The starting materialwas sealed into AgroPdro capsules. Normally, experimen-tal times ranged from 28 to 8 h.

An estimation of the,fo, *as made in experiments whereolivine, orthopyroxene, and spinel coexist in the product,using the semiempirical O geobarometer proposed byBallhaus et al. (l 99 l). The calculation ofthe Fe3+ contentof spinel assumed perfect RrOo stoichiometry (e.g., Mat-tioli et al., 1989; Ballhaus et al., l99l). The log .fo,ob-tained is within 0.5 log units of the fayalite + magnetite+ quartz (FMQ O buffer.

The liquid compositions at low degrees of partial melt-ing of the two model mantle sources, CPL and PLZ,weredetermined using a sandwich technique, in which 10- I 8wto/o of an estimated melt composition is added betweentwo layers of the peridotite source. This technique en-sures that large areas of melt, whose composition is notaffected by quench overgrowths on primary residual crys-

786 THIBAULT ET AL.: MELTING OF PHLOGOPITE LHERZOLITE

TlsLe 2. Experimental results

Partial melting experiments

Starting material (h)T

("c) Stable ohases

CPLCPLCPLCPLCPLCPLCPLCPLCPLCPLcPL (SDW1)cPL (SDW2)cPL (SDW3A)cPL (SDW3B)

PLZPLZPLZPLZPLZPLZPLZPLZ (SDW1)PLZ (SDW2A)PLZ (SDW2B)

900925975

1 000950->1025950-> 1 050950->1075950-> 1 1 00950-> 1 1 25950-> 1 1 50

1 1001 1 0 01 1001 1 0 0

11251 1501 1751200122512351250122512251225

28282028

21_>520->521->4.520->520->420->4

9.59.59.51

282820201 9201 0881

Melt-peridotite interaction experiments

ol, opx, cpx, phl, magol, opx, cpx, phl, magol, opx, cpx, phl, magol, opx, cpx, phl, magol, opx, cpx, phl, mag, dolol, opx, cpx, phl, dolol, opx, cpx, phl, gar, liqol, opx, cpx, phl, gar, liqol, opx, cpx, phl, gar, liqol, opx, cpx, phl, gar, liqol, opx, cpx, gar, liqol, opx, cpx, phl, gar, liqol, opx, cpx, phl, gar, liqol, opx, cpx, phl, gar, liq

ol, opx, cpx, phl, garol, opx, cpx, phl, garol, opx, cpx, phl, gar, liqol, opx, cpx, phl, gar, liqol, opx, cpx, phl, gar, liqol, opx, cpx, phl, gar, liqol, opx, cpx, phl, gar, liqol, opx, cpx, phl, gar, liqol, opx, cpx, phl, gar, liqol, opx, cpx, phl, gar, liq

Starting material (h)T

fc) Weight ratio- Stable phases

HARWHRCarmet-HARCarmet-WHRSilmet-HAR

10001 0001 0001000

1200->1000

28282020l->,22

0:1 000:1 00

90:1 090:1075:25

ol, opxol, opx, spol, opx, cpx, phl, vol, cpx, sp, cc, phl, vol, opx, cpx, phl

/Vote.' The notations m-> n and x->? for temperature and times, respectively, mean that the experiment was conducted x h at m'C, then broughtto n 'C gradually in z h, and then maintained y h at n "C. Abbreviations: ol : olivine; opx : orthopyroxene; cpx : clinopyroxene; phl : phlogopite;gar: gamet; mag: magnesite; dol : dolomite; liq: liquid; sp: spinel; cc: calcite; v: vapor phase. Stable phases do not include relict phasesobserved in trace amounts. (SDWX) refers to the Xth iteration for the sandwich experiment,* Expresses the melvperidotite ratio of the mixture in weight percent.

tals, are available for defocused electron beam micro-probe analyses (Fujii and Scarfe, 1985). An iterativemethod was adopted in which the composition of themelt, detemined after an initial sandwich experiment, isused as a guide for the synthesis of the melt layer of thesubsequent sandwich experiment (cf. Wallace and Green,1988). The composition of the quench liquid is consid-ered to be near that of the melt ideally in equilibriumwith the residual mantle source when: (l) the mineralassemblage in the sandwich experiment is the same asthat of the nonsandwich experiment at the same pressure-temperature conditions, and (2) the composition of thecorresponding residual minerals in the two types of ex-periments are very close. The iteration experiments weredone with durations of 9.5 and 8 h. When the two criteriadescribed above were fulfilled, an additional sandwichexperiment of I h was performed to minimize Fe losseffect on the determined melt composition (cf. Mengeland Green, 1989).

Attainment of equilibrium

In the products of the melting experiments, most crys-talline phases are compositionally uniform. Only largeclinopyroxene grains (> l0 pm) have slight zonation withrelict composition in the cores. The rims of these largergrains have a composition similar to very small homo-geneous clinopyroxene crystals (-2 p.m) in close contactwith the melt, suggesting that the clinopyroxene rims havelikely reacted fully with the liquid (cf. Fujii and Scarfe,1985). Moreover, the unreacted central portion of theclinopyroxene crystals is quite small (< 100/o of the grain).Assuming that 100/o of the total amount of clinopyroxene(estimated maximum amount of relict cores) was not partof the system, new bulk compositions of the CPL andPLZ model sources could be calculated (Table l). Becausethese corrected bulk compositions are very close to theoriginal ones, we considered that adequate equilibriumconditions have been attained. When needed in melting-reaction calculation, the original and corrected bulk com-

THIBAULT ET AL.: MELTING OF PHLOGOPITE LHERZOLITE 787

TABLE 3. Comparison of representative compositions of phases obtained in nonsandwich and sandwich melting experiments

cPL (1 100 .C) PLZ (122s.C)SourcePhaseType

cpx phlgarphl

N-SDW SDW3B N.SDW SDW3B N-SDW SDW3B N-SDW SDW2B N.SDW SDW2B N-SDW SDW2B

sio,Tio,Al203Cr2O3FeO.,MnOMgoCaONarolGo

Total

52.91 52.950.09 0.093.74 3.500.46 0.512.97 2.820.08 0.16

16.72 16.3822.26 22.330.85 0.810.00 0.00

100.08 99.55

37.45 38.585.52 5.45

17.71 17.370.34 0.614.54 4.820.04 0.10

20.02 19.740.54 0.350.10 0 .139.43 9.82

95.69 96.97

Compositions in wt0641.38 41.44 52.261.72 1 .41 0.14

21.41 20.96 4.481.32 1.39 0.637.20 7.55 2.920.33 0.34 0.11

19.27 19.19 ' , t7.217.60 7.56 20.900.00 0.00 0.730.00 0.00 0.00

100.23 99.84 99.38

Cations per O atoms-(24) (241 (6)

37.81 38.767.04 6.56

16.12 16.931.54 0.394.04 4.220.06 0.0s

19.03 18.870.21 0.(M0.03 0.10

10.43 10.4496.31 96.36

(221

42.30 42.120.69 0.s6

20.44 21.513.12 2.495.16 5.930.25 0.26

20.81 19.977.13 6.960.00 0.000.00 0.00

99.90 99.80

52.630.154.200.982.940.06

17.2020.890.830.03

99.91

(6) (241(221(22)(221(6)(6) (24)

Si(alAl

r6lAl

TiCrFeMnMgCaNAK

TotalX"n.*

1 .9210 1 .93220.0790 0.06780.0811 0.08280.0025 0.00230.0132 0.01460.0902 0.08590.0025 0.00490.9047 0.89060.8660 0.87320.0598 0.05710.0000 0.00004.0200 4.01140.91 0.91

5.3058 5.40272.6942 2.59730.2639 0.27050.5882 0.57400.0381 0.06750.5379 0.56450.0048 0.01194.2271 4.11980.0820 0.05250.0275 0.03531.7045 1.7545

15.4740 15.45050.89 0.88

5.9137 5.95480.0863 0.04523.5208 3.50570.1849 0.15240.1491 0.15790.8605 0.90730.0399 0.04144.1042 4.10971.1638 1 .16400.0000 0.00000.0000 0.0000

16.0232 16.03840.83 0.82

1.9040 1.90890.0960 0.09110.0964 0.08850.0038 0.00410.0181 0.02810.0890 0.08920.0034 0.00180.9345 0.92970.8159 0.81190.0516 0.05840.0000 0.00144.0127 4.01300.91 0.91

5.3577 5.45632.6423 2.54370.0512 0.26600.7503 0.69450.1728 0.0,|{}40.4785 0.49680.007s 0.00604.0190 3.95880.0323 0.00600.0082 0.02731.8861 1.8750

15.4059 15.37380.89 0.89

6.0191 6.00030.0000 0.00003.4289 3.61250.0738 0.06000.3510 0.28040.6141 0.70650.0301 0.03144.4131 4.23981.0871 1.06240.0000 0.00000.0000 0.0000't6.0172 15.993it0.88 0.86

A/ote.'Abbreviations as in Table 2; N-SDW represents nonsandwich experiment; SDWX refers to the Xth iteration for the sandwich experiment (seeTable 2).

* Number of O atoms in parentheses at head of columns.

positions were both used in order to semiquantify theeffects of this potential problem. In the products of theinteraction experiments no compositional zonation wereobserved.

Analytical methods

Longitudinal sections of the experimental charges weremounted and polished for microprobe analysis. To avoidpossible dissolution of alkali-bearing carbonates, HrO wasnot used in the preparation of polished sections for theproducts of the experiments on the CPL composition (cf.Wallace and Green, 1988). Wavelength dispersive anal-yses of all phases were conducted on a JEOL JXA-8600electron microprobe. All elements were analyzed at 15-kV accelerating voltage with a l0-nA beam current. Ma-trix corrections were made using the Tracor NorthernZAF program.

Rnsulrs FoR CPL MANTLE souRcr lr 3.0 GPa

Phase relationships

Temperatures for the melting experiments on the CPLmantle composition at 3.0 GPa ranged from 900 to I 150'C. Olivine, orthopyroxene, clinopyroxene, and phlogo-pite were stable at those temperatures (Table 2). Garnetwas abundant in experiments from 1075 to I 150 'C but

could not be observed at subsolidus conditions (Table 2),probably because of lack of time for nucleation at suchlow temperatures. Well-crystallized carbonates were sta-ble up to 1050 "C. At higher temperatures only smallinterstitial acicular grains of carbonate were observed,probably quenched from a COr-rich melt. The solidus forthe CPL composition at 3.0 GPa occurred, therefore, near1060 "C. This temperature significantly exceeds the oneobtained by Wallace and Green (1988) for the solidus oftheir carbonated peridotite (< I 000 'C) at pressures great-er than amphibole breakdown (=carbonated phlogopitelherzolite). This is probably because of the fact that CPLmelts under fluid-absent conditions, in contrast to thesource used by Wallace and Green (1988), which coexistswith free HrO released by amphibole breakdown at highpressure.

Chemical composition of minerals

Olivine and orthopyroxene have relatively constantcompositions throughout the investigated temperatureinterval. Clinopyroxene is an aluminian chromian diop-side (Table 3). At suprasolidus conditions, the clinopy-roxene grains show lower Na content (NarO I 0.8 wto/o)than the subsolidus ones (NarO = 1.5 wto/o).

Phlogopite has high TiO, contents (Table 3). As melt-ing proceeds, the ratio of TilK in phlogopite increases

788 THIBAULT ET AL.: MELTING OF PHLOGOPITE LHERZOLITE

0.l()

0.35

Ti/K(atomic) o.ro

0.25

0.20

0.40

0.35

Ti/K o.3o(atomic)

0.2s

0.20

PLz phloqooite b)

bulkPLZ

II

I

CPL solidus

+CPL phlogonite

bulk CPL- - - a - - a

850 900 950 1000 1050 1100 ll50 1200

Temperature (oC)

PLZ solidusIY

2.0n

Ti02(wtVo\

1.00

0.000.00 r.00

CrrO, (wt%o)

Fig.2. Diagram of TiO, (wto/o) against CrrO3 (wto/o) in garnetobserved in the melting experiments on the CPL and the PLZmodel mantle sources.

Melt composition at 3.0 GPa and 1100 t

Sandwich experiments using the iterative method de-scribed earlier were performed at I100 'C and 3.0 GPa.The synthesized melt composition for the initial sand-wich experiment was a sodic dolomitic melt (Table 4,analysis A). Three iterative steps were needed to obtainresidual mineral compositions very close to those in thenonsandwich experiment at I100 "C (Table 3). The com-position of the melt (Table 4, analysis B) was estimatedby defocused electron beam microprobe analyses oflargemelt pools consisting of aggregates of acicular quenchedcarbonate grains. The melt has a strong dolomitic affinity,very low contents of SiOr, AlrO., and TiO, and a signif-icant amount of alkalis. Because K was supplied by thepartial breakdown of phlogopite, and no other hydrousminerals were observed, the synthesized melt composi-tion in each iteration step contained HrO in an amountdefining a HrO/KrO weight ratio near that of hydrousphlogopite (HrO/KrO = 0.38). Finally, the X*r.*, valueof the melt is high. Although this can be partially ex-plained by Fe loss to the AgroPdro capsule, Green andWallace (1988) have also reported that Mg-Fe fraction-ation between solid silicates and alkaline dolomitic car-bonatite melt is significantly smaller than for silicate melts.

The composition of all the residual minerals and of theliquid were used together with the bulk CPL composition(either original or corrected) in a least-squares mass-bal-ance calculation to estimate the weight proportion of thephases stable at I 100 "C. The calculation suggests thatthe CPL model source yields 4 wto/o of alkaline dolomiticmelt, and, as observed in the experimental product, theproportion of garnet is quite high (10 wto/o). The reactionscontrolling the phlogopite-present melting interval of theCPL model source at 3.0 GPa can be qualitatively definedAS

dolomite.o'o + phlogopite"" * clinopyroxeness- dolomite-",, * (K,Na,OH)-o, * olivine"" +

pyropess (rich in Ti) + residual phlogopite*(rich in Ti) + residual clinopyroxene* (l)

0.151100 tt25 1150 tr75 1200 1225 1250 1275

Temperature (oC)

Fig. l. Variation of the Ti,{K atomic ratio of phlogopite againsttemperature in the melting experiments on (a) the CPL and (b)the PLZ model mantle sources. Because the amount of K isrelatively constant in phlogopite, Ti-K variation represents es-sentially the change in the abundance of Ti. The dashed linesrepresent the TilK ratios calculated for the model sources bulkcompositions. As long as phlogopite represents the only K-bear-ing phase, these bulk TilK ratios should define the maximumTi content that phlogopite can contain.

markedly to values in excess of 0.29, calculated for thebulk CPL mantle composition (Fig. la). A hieh ratio ofTi/K for suprasolidus phlogopite suggests that the coex-isting melt contains a significant amount of K with a lowTilK ratio.

As the solidus is crossed, garnet becomes abundant. Itspyrope component ranges from 64.8 to 67.3 molo/0. Asshown in Figure 2, garnet from the melting experimentson the CPL composition is rich in TiO, (x) wto/o) andrelatively poor in CrrO, (<2 wto/o).

The stable carbonate phase is magnesite (MgCO, * 86molo/o) from 900 to 1025 .C. It is then replaced by do-lomite (MgCO. = 48 molo/o) up to 1050'C (cf. Brey etal., 1983; Olafsson and Eggler, 1983; Falloon and Green,1989).

Gauet

a a ^ ^

^ ^ a

l a

II t l

I t

THIBAULT ET AL.:

TABLE 4. Average melt compositions

MELTING OF PHLOGOPITE LHERZOLITE 789

NormcrPW (D)

46sio,Tio,Al2o3Cr"O.FeO,,MnOMgoCaONaroKrOHrOCO,

TotalXtn.*

2.570.720.380.00

5.61 4.540.13 0.1 6

15.40 15.1224.95 21.608.19 4.93

7.012.66

45.72 40.31r00.00 100.00

0.83 0.86

44.83 47.962.12 1 .68

15.48 10.970.29

7.36 5.660.10 0 .15

11.46 13.216.06 11 .591 .13 1 .048.30 5.403.16 2.05

100.00 100.000.74 0.81

1100 tt25 1150 rl75 1200

OrAnLcNeDiorMttl

7.259.s3

19.814.86

38.8314.60

1 .593.25

Wo

3E

36

/Vote.- A and C are synthesized melt compositions used in the initialsandwich experiments on the CPL (A) and PLZ (C) sources. B and D arefinalcompositions of melts obtained by sandwich experiments on the CPL(B) and PLZ (D) sources. D was normalized to 1 00 wt% Geal total anhy-drous: 96.01 wt%). lts norm was obtained with a calculated ratio of Fe.O"/FeO (Kress and Carmichael, 1988) for fo. of the OFM buffer (Myers andEugster, 1983) at 1225 "C. H,O was calculated for a H,O/K,O weight ratioof 0.38. CO, was calculated by the stoichiometry of the carbonates in Aand the deficiency in the total of oxides in B. Na"O in D is calculated usingthe phase proportions obtained by least-squares mass balan@.

where ss means solid solution. In addition to the com-plete melting of dolomite at the solidus, Reaction I de-scribes the partial breakdown of phlogopite and clino-pyroxene (adeite component) to yield K, Na, and OH tothe melt (cf. Wallace and Green, 1988). SiOr, AlrOr, MgO,FeO, and TiOr, also released from the breakdown ofphlogopite and clinopyroxene, are combined to form ol-ivine and pyrope. This explains the high abundance ofgarnet rich in TiO, (Fie. 2).

Implications for the subsequent experiments

If carbonated phlogopite lherzolite composition com-parable to CPL exists in the upper mantle, alkaline dolo-mitic melt may be formed and easily migrate upward byinfiltration caused by its low viscosity (Hunter andMcKenzie, 1989), leaving behind a residual COr-free gar-net-bearing phlogopite lherzolite. The experiments on thePLZ composition were performed to investigate the com-position of the near-solidus melt formed from such a re-sidual mantle source.

Rnsur-rs FoR PLZ MANTLE souRcE lr 3.0 GP,q'

Phase relationships

Olivine, orthopyroxene, clinopyroxene, garnet, andphlogopite were stable throughout the investigated tem-perature interval ranging from ll25 to 1250'C (Table2). On the basis ofthe appearance ofinterstitial glass anda significant decrease in the abundance ofphlogopite, thesolidus was estimated at ll7 5 "C (Table 2), a temperaturein accordance with the solidus temperature obtained byWendlandt and Eggler (1980, their Fig. l) on a compa-rable phlogopite lherzolite.

0.032

Na/Q 0.028

0.024

0.02

0.12

(Al+Cr)/Qo.rt

0.10

1100 rl25 1150 1175

0.091100 rrzs 1150 ll75 1200 1225 1250 r27s

Temperature (oC)

Fig. 3. Variation of (a) Wo : l00x[Cal(Ca + Mg + Fe +Mn)], (b) Na/Q : Na/(Ca + Mg + Fe + Mn), and (c) (Al +Cr)/Q : (Al + Cr)/(Ca + Mg + Fe + Mn) in clinopvroxeneagainst temperature for the melting experiments on the PLZmodel mantle source. All ratios are atomlc.

Chemical composition of minerals

Olivine and orthopyroxene have relatively constantcompositions. With increasing temperature, clinopyrox-ene shows a systematic decrease in the proportion of wol-

rastonite component (Fig. 3a). Also observed in clino-pyroxene, from the solidus temperature to 1250 "C, area decrease in Na (Fig. 3b) and an increase in (A1 ., + Cr)(Fig. 3c), relative to the quadrilateral components (Ca *

M g + F e + M n ) .Phlogopite is rich in TiO, with the ratio of TilK in-

creasing steadily with temperature (Fig. lb). At suprasoli-dus conditions, TilK ratios in excess of 0.29 calculatedfor the bulk PLZ composition (Fig. lb) suggest that themelt is enriched in K with Ti partitioned preferentially

in the coexisting phlogoPite.The pyrope component of garnet ranges from 68.9 to

72.5 molo/o. Compared to garnet observed in the melting

solidus

I

b)

790

of the CPL composition, PLZ garnet is poorer in TiO,and generally richer in CrrO, (Fig. 2).

Melt composition at 3.0 GPa tnd, 1225 "CSandwich experiments were performedat 1225 "C. The

composition of the synthesized glass used in the initialsandwich experiment (Table 4, analysis C) was based onelectron microprobe analyses of small melt pools fromthe nonsandwich experiment at 1225 "C. The amount ofHrO added to the synthesized glass was defined using aHrO/IIO weight ratio of 0.38, typical of hydrous phlog-opite.

Two iterative steps were needed to obtain a good ap-proximation of the equilibrium partial melt of the pLZsource at 1225'C. The melt layer showed a homogeneoustexture consisting ofglass and small acicular quench grainsof clinopyroxene. Analyses of the quench liquid (glass +quench clinopyroxene) were obtained with the micro-probe using a defocused electron beam with a diameterof 10-30 pm. The average composition is presented inTable 4 (analysis D). The melt is strongly potassic (K,OE 5 wto/o) and silica undersaturated. The major norma-tive components are diopside, leucite, olivine, and an-orthite, reflecting the major role of clinopyroxene andphlogopite in the melting reaction. The X.r.",o, value of0.81 is significantly lower than for the melt obtained fromthe CPL source. The composition found by Mengel andGreen (1989) for a melt coexisting with phlogopite-bear-ing peridotitic assemblage at 2.8 GPa and I195 "C usingcomparable sandwich techniques bears similarities to themelt formed fromPLZ at 1225 "C but is clearly less po-tassic (KrO r 1.6 wto/o). This difference probably resultsfrom the fact that, in contrast to the melting conditionsof this study, the HrO/alkali ratio in Mengel and Green'ssource is such that there is an excess of HrO relative tothe amount needed to form phlogopite when amphibolebreaks down at high pressure.

The weight proportions of the phases stable at Il25and 1225 "C were calculated by least-squares mass bal-ance. Based on the results, a generalized melting reactiondescribing the phlogopite-present melting interval of thePLZ model source at 3.0 GPa would be

3.8(4.3) phlogopite + 3.7(3.5) clinopyroxene+ 0.8(0.8) garnet

- l. l(1.2) olivine + 0.0(0.4) orthopyroxene+ 7.1(7.0) l iquid (2)

where the values in parentheses are calculated using thecorrected bulk composition of PLZ (Table l,PLZC). Thedegree of partial melting at 1225 t is, therefore, esti-mated at 7 wto/o.In contrast to the relations observed forthe CPL composition (Reaction l), garnet is a reactant ofmelting Reaction 2 due to the high compatibility of Al,Or,SiOr, and TiO, in the silicate melt. This explains the low-er TiO, and higher CrrO, contents of PLZ garnet (Fig. 2).

THIBAULT ET AL.: MELTING OF PHLOGOPITE LHERZOLITE

Rrsur-rs oF THE MELT-pERrDorrrE INTERACTIoNEXPERIMENTS

The melt-peridotite interaction experiments were per-formed in order to investigate the reactivity of the meltcompositions described in the earlier sections with de-pleted peridotitic material. For this purpose, dolomitic(Carmet) and silicate (Silmet) melt compositions, and twoperidotitic protoliths were synthesized (Table 1). The in-teraction experiments were performed at 2.0 GPa and1000'C. Details of the experimental conditions and re-sults are presented in Table 2. Reactions were obtainedby least-squares mass-balance calculation of the propor-tion ofthe phases stable before and after each interaction.

Carmet-HAR interaction experiment

In addition to olivine and orthopyroxene present in theharzburgitic protolith, the product of the Carmet-HARinteraction experiment contains clinopyroxene andphlogopite (Table 2). Moreover, the absence of carbon-ates suggests the existence of a COr-rich low-density fluid.The metasomatic reaction between Carmet and HAR is(in weight proportion)

24.8 (orthopyroxene)ro" * 9.9 Carmet- 14.2 olivine + 13.6 clinopyroxene

+ 2.7 phlogopite + 4.2 CO.,. (3)

This important transformation of the harzburgite by Car-met is related to the breakdown in peridotitic systems ofdolomite at pressures less than 2.1 GPa (e.g., Wyllie andRutter, 1986; Wallace and Green, 1988; Falloon andGreen, 1989) according to the well-documented reaction:

dolomile + enstatite- diopside + forsterite + CO, (4)

(e.g., Wyllie and Huang, 1976;'Eggler, 1978; Brey et al.,1983). At 2.0 GP4 the decarbonation reaction (Reaction4) occurs around 950'C in carbonated and hydrated pe-ridotite (e.g., Falloon and Green, 1989). Therefore, inter-action of Carmet and harzburgite at temperatures lessthan 950 "C should probably result in the formation ofdolomite and phlogopite without consumption of ortho-pyroxene (dolomite-bearing phlogopite harzburgite),rather than the products ofReaction 3.

Carmet-WHR interaction experiment

The resulting mineral assemblage of the Carmet-WHRinteraction experiment (Table 2) contains a calcite solidsolution (CaCO, = 90 molo/o) and phlogopite, in additionto olivine, clinopyroxene, and spinel present in the wehr-litic protolith. The metasomatic reaction between Carmetand WHR is (in weight proportion)

3.1 (clinopyroxene)*r* + 0.3 (spinel)*r** 9.8 Carmet

- 2.5 olivine + 2.6 phlogopite + 7.0 calcite+ 1.1 co,. (5)

THIBAULT ET AL.: MELTING OF PHLOGOPITE LHERZOLITE 791

Silmet-HAR interaction experiment

The experimental conditions for the Silmet-HAR in-

teraction (Table 2) were essentially the same as those forthe Carmet-HAR experiment (2.0 GPa, 1000 "C), exceptthat the starting mixture was first held at 1225 "C for I h

to induce Silmet melting and then cooled slowly to 1000'C (cf. Meen, 1987). The final product of the experimentusing the Silmet composition and the harzburgite proto-

lith consists of olivine, orthopyroxene, clinopyroxene, andphlogopite (Table 2). The reaction for this interaction is(in weight proportion)

3.3 (olivine)"AR + 0.1 (orthopyroxene)ro^+ 10.0 Silmet

- 7.2 clitopyroxene + 6.2 phlogopite. (6)

In Reaction 6 most of the components required for theformation of the newly introduced clinopyroxene andphlogopite come from the crystallization of the silicatemelt. Only the small excess of KrO, AlrO3, SiOr, and HrOcontained in Silmet, reflecting the incongruent melting ofphlogopite to form olivine and liquid in the phlogopitelherzolite melting experiment, reacts with olivine of theharzburgite to form metasomatic phlogopite.

Implications of the interaction experiments

Figure 4a represents a summary of the mineralogicaleffects on a harzburgitic protolith, caused by the hypo-thetical infiltration of Carmet at 2.0 GPa and 1000 'C,

depicted on an olivine-orthopyroxene-clinopyroxene ter-nary diagram. The harzburgite in Figure 4a contains 70volo/o of olivine and 30 volo/o of orthopyroxene, a realisticaverage of the proportions found in many mantle xeno-liths of harzburgitic compositions (e.g., Boyd, 1989). Ap-proximately 12 g of Carmet would be needed to trans-form 100 g of harzburgite to an olivine-rich wehrlite (Fig.4a) with about 3 volo/o of phlogopite. Because of the highratio of Mg/Fe of the dolomitic melt, the bulk X'ro.,., ofthe wehrlite should be close to that of the protolith (cf.Green and Wallace, 1988). If Carmet infiltration persists,the metasomatized mantle would then evolve toward acalcite- and phlogopite-bearing dunite (Fig. 4a).

The mineralogical effects of Silmet infiltration in theharzburgite is transformation to an olivine-poor phlogo-pite lherzolite (Fig. 4a). Because Silmet reacts primarilywith olivine in the Silmet-HAR interaction experiment,the same type of reaction (Reaction 6) should probably

occur if Silmet infiltrated an olivine-rich wehrlite. Theeffect would then be to drive the composition toward anolivine-poor phlogopite wehrlite (Fig. 4a).

Pernor,ocrcAl, IMPLICATToNS

In this section the experimental results are integratedinto a petrological framework of cyclic metasomatic pro-

cesses potentially active in the lithospheric mantle of a

continental rift.

Orthopyroxene Clinopyroxene

Gecs, West EifelGermany

Orthopyroxene Clinopyroxene

Fig. 4. (a) Effects of the progressive metasomatism (arrows)

of a harzburgitic protolith at 2.0 GPa and I 000 "C caused by the

infiltration of Carmet or Silmet. The trend from hanburgite (fiiled

square) to olivine-rich wehrlite (empty square) caused by Carmet

infiltration is gaduated at intervals, indicating the amount of

Carmet (in grams) needed for the transformation of 100 g of

harzburgite protolith. More details in text. (b) Modal composi-

tions of representative samples of the major types of peridotitic

xenoliths collected from the mafic ashes in a quarry southeast of

Gees, West Eifel, Germany, plotted in an olivine-orthopyroxene-clinopyroxene ternary diagram (cf. Fenwick, 1991). Each sampleis labeled by a number representing the weighted average of theXro,"., values calculated for all the silicate minerals. Filled circles

are orthopyroxene-bearing xenoliths, and empty circles are or-

thopyroxene-free xenoliths. All data used for the construction of

this diagram are taken from Edgar et al. (1989) and Lloyd et al.( 1991). The metasomatic trends shown in a are superposed in b.

Multicycle metasomatic processes below a

continental rift

A thermal framework. Figure 5 shows four geotherms

representing the thermal evolution of the upper mantle

b)

Olivine

Olivine

0 8 6

792 THIBAULT ET AL.: MELTING OF PHLOGOPITE LHERZOLITE

r20 g

F

160 .:a

O i -

E t8 c

L 6

G 2Es

a <

2 a

3 a4 Es a6 ! ,

2 *

3 8

4 is a

E

6 9

600 800 1000 L200 1400

Temperature ( o C)Fig. 5. Group ofgeotherms representing different hypothet-

ical stages in the thermal evolution of the lithospheric mantlebelow a continental rift. Geotherm a is considered characteristicofthe lithospheric mantle before initiation ofrifting and is takenfrom McKenzie (1989, his Fig. 4d). Geotherms b, c, and d depictthe thermal disturbances during progressive rifting related to theadiabatic ascent of a hot mantle plume with a potential temper-ature (7ro) of 1380 "C. Geotherms b and c are not derived fromany data or calculations but are considered reasonable inter-mediates between geotherms a and d. The temperatures definedby geotherm d up to 3.0 GPa are comparable to the ones pro-posed by Seck and Wedepohl (1983) and Fuchs and Wedepohl(1983) for the Rhenish Massif, Germany. The potential temper-ature is the temperature that the mantle would have if it couldrise to the Earth's surface under perfect adiabatic conditions (e.g.,McKenzie and Bickle, 1988).

below a hypothetical active continental rift. The progres-sive thinning ofthe lithosphere and the associated ther-mal disturbances are depicted as the result of the adia-batic upwelling of a mantle plume with a potentialtemperature of 1380 .C. Simplified schematic sectionsdepicting the proposed processes occurring in a litho-spheric mantle of harzburgitic composition are shown inFigure 6.

Deep rnetasomatic cycles (depth >100 km). In Figure6a, the fertile mantle plume releases fluids as it reachesthe l ithosphere-asthenosphere boundary (LAB) (e.g.,Wyllie, 1988, his Figs. 45 and 46). Diamond microinclu-sions whose compositions resemble that of potassic mag-mas but with volatile contents (essentially HrO and COr)up to 400/o may represent such high-pressure (>5.0 GPa)fluids (Navon et al., 1988). Because these volatile-richinclusions are rich in KrO, CaO, SiOr, and FeO (Navonet al., 1988, their Table l), they should be effective intransforming harzburgite into carbonated phlogopitelherzolite.

As the LAB progressively rises (Fig. 6b), the carbon-ated phlogopite lherzolite partially melts because of heatconduction from the underlying plume and decompres-sion caused by uplift (e.g., McKenzie, 1989). The low-

Fig. 6. Hypothetical schematic sections representing thethermal conditions in the lithospheric mantle below a continen-tal rift. The sketches at the center of each section illustrate theproposed metasomatic processes in a portion of the lithosphericmantle originally harzburyitic in composition. The intersectionsof the isotherms with the central sketches define the pressure-temperature conditions corresponding to the geotherms of Fig-ure 5: (a) geotherm a, (b) geotherm b, (c) geotherm c, and (d)geotherm d. The curvatures of the isotherms are purely quali-tative but emphasize the thermal disturbances accompanyinglithospheric thinning relative to the original thermal conditionsshown in a. LAB refers to the lithosphere-asthenosphere bound-ary. The pattemed areas represent the metasomatic horizons.Large empty arrows : adiabatic upwelling of hot mantle plume(2" 1380 "C); small filled arrows: infiltration of small fractionsof melts or dense fluids; small empty arrows : release of COr-rich low-density fluid; Carmet: alkaline dolomitic melt; Silmet: hydrous potassic and calcic silicate melt. More details in text.

temperature melting components are then remobilized insmall-volume volatile-rich melts, which, as they migrateupward, resolidify and react with the overlying colderharzburgite. Through such cycles of melting, migrating,solidifying, and reacting, the metasomatic front eventu-ally reaches a depth of 100 km (=3.0 GPa). The temper-ature at that depth being approximately 1050'C (Fig. 6b),the interaction of the COr- and HrO-bearing melts withharzburgite results in the formation of a dolomite-bearingphlogopite lherzolite.

Shallow decoupled metasomatic cycle (depth <100 km).In Figure 6c, the temperature of the horizon of dolomite-bearing phlogopite lherzolite at a depth of 100 km is raisedto about ll50 "C. An alkaline dolomitic melt can formand infiltrate the overlying depleted lithosphere, leavingresidual COr-free garnet-bearing phlogopite therzolite,which returns to subsolidus conditions (e.g., PLZ, Table2). As the alkaline dolomitic melt reaches a depth of about65 km (tf.g GPa, <950 "C), it solidifies and reacts withharzburgite, resulting in the formation of a dolomite-bearing phlogopite harzburgite. The metasomatic front isnow decoupled and forms two distinct horizons: a dolo-mite-bearing phlogopite harzburgite at 65 km and a COr-free garnet-bearing phlogopite lherzolite at 100 km.

With progressive rifting, the temperature at a depth of

65 km reaches 1000 "C (Fig. 6d). The dolomite-bearingphlogopite harzburgite is then transformed to olivine-richphlogopite lherzolite and wehrlite, with the release of COr-rich low-density fluid (see Reaction 3). At 100 km, thetemperature of the garnet-bearing phlogopite lherzolitehorizon exceeds the solidus, forming a hydrous potassicand calcic silicate melt. When this melt infiltrates themetasome at the depth of 65 km, an enrichment in cli-nopyroxene and phlogopite takes place, according to Re-action 6. Because the silicate melt has a lower Mg/Feratio, a decrease in the bulk Xrr..,o, is also expected. Thefinal product of this shallow decoupled metasomatic cycleis a wide variety of rock types comparable to those de-picted in Figure 4a.

The Eifel mantle xenoliths-an example of a shallowdecoupled metasomatic event?

Recently, Edgar et al. (1989) and Lloyd et al. (1991)studied a group of 225 ultramafic xenoliths sampled frommafic ashes near Gees village, in the West Eifel sodi-potassic volcanic region, Germany. The majority of thexenoliths (=87o/o) are peridotitic, ranging from spinelharzburgite and spinel lherzolite to olivine-rich and ol-ivine-poor spinel wehrlite, with various amounts ofphlogopite (Lloyd et al., l99l). In wehrlite, as the amountof olivine decreases, the abundance of titaniferous phlog-opite increases. In Figure 4b, peridotitic samples studiedby Edgar et al. (1989) and Lloyd et al. (1991) are plottedon an olivine-orthopyroxene-clinopyroxene ternary dia-gram.

It is suggested that the proposed decoupled metaso-matic process is compatible with the compositional vari-ation observed in the Gees xenoliths. The major pointssupporting this suggestion are

L The compositional variations of the xenoliths areconsistent with the trends that the decoupled metaso-matic process may create (Fig. ab) at thermal conditionsofgeotherm d (Fig. 5) and pressures near or less than 2.0GPa. These conditions are probably characteristic of themantle represented by the Gees xenoliths because (a)geotherm d is based on data from the Rhenish Massif,and (b) the peridotite xenoliths contain chromium spineland are garnet free, suggesting pressures less than 2.5 GPa(e.g., Carroll Webb and Wood, 1986).

2. The common source of the metasomatic agents is acarbonated phlogopite lherzolite horizon near the depthof 100 km. Such a horizon at this depth was proposed byMertes and Schmincke (1985) as the source for some ofthe potassic lavas erupted in the West Eifel area.

3. Except for one sample, all the olivine-rich rocks (>65volo/o olivine) following the trend from harzburgite to ol-ivine-rich wehrlite have constant average silicate X.r.*.values of 0.90 and 0.91 (Fig. 4b), compatible with themetasomatic effects of a primitive carbonatite melt. Allthe olivine-poor xenoliths have significantly lower Xrr.",o,(=0.86, Fig. 4b), in accordance with a silicate melt infil-tration event. The olivine-rich rock with an Xrr.",o, valueof0.86 is, in contrast to the others, very rich in phlogo-

793

pite (= I I volo/o; Lloyd et al-, l99l), and therefore mayhave been infiltrated by a K-rich derivative of the alkali-rich silicate melt.

AcrNowlnncMENTS

Financial support was provided by a NSERC grant to A.D.E. and aFCAR scholarship to Y.T. Funding from NATO as a Collaborative Re-search Grant is acknowledged. Danilo Vukadinovic and Sandro Conticelliare thanked for helpful discussions. Constructive reviews by DaYid H.Green, David Hay, and Richard F. Wendlandt are appreciated.

RnrnnrNcns crrED

Ballhaus, C., Berry, R.F., and Green, D.H. (1991) High pressure experi-menul calibration of the olivine-orthopyroxene-spinel oxygen geoba-

rometer: Implications for the oxidation state ofthe upper mantle. Con-

tributions to Mineralogy and Petrology, 1O7,27-40.Boyd, F.R. (1989) Compositional distinction between oceanic and cra-

tonic lithosphere. Earth and Planetary Science ktters, 96,15-26.Boyd, F.R., and En€iand, J.L. (1960) Apparatus for phase equilibrium

measurements at pressures up to 50 kbar and temperatures to 1750 {'

Journaf of Geophysical Research, 65, 7 4l-7 48.Brey, G, Brice, W.R., Ellis, D.J., Green, D.H., Harris, K.L., and Ry-

abchikov, I.D. (1983) Pyroxene-carbonate reactions in the upper man-

tle. Earth and Planetary Science l€tters, 62,63-74.Canil, D. (1990) Experimental study bearing on the absence ofcarbonate

in mantle-derived xenoliths. Geology, 18, 101 l-1013.Carroll Webb, S.A., and Wood, B.J. (1986) Spinel-pyroxene-garnet rela-

tionships and their dependence on CrlAl ratio. Contributions to Min-

eralogy and Petrology, 92, 47 1-480.Edgar, A.D., Lloyd, F.E., Forslth, D.M., and Barnett, R'L. (1989) Origin

of glass in upper mantle xenoliths from the quatemary volcanics of

Gees, West Eifel, Gerrnany. Contributions to Mineralogy and Petrol-

ogy, 103,277-286.Eggler, D.H. (1978) The effect of CO, upon partial melting of peridotite

in the system NarO-CaO-AlrO!-MgO-SiOrCO, to 35 kb, with an anal-ysis of melting in a peridotite-HrO-CO, system. American Journal of

Science.278. 305-343.Erlank, A.J., Waters, F.G., Hawkesworth, C.J', Haggerty, S.E., Allsopp'

H.L., Rickard, R.S., and Menzies, M.A. (1987) Evidence of mantle

metasomatism in peridotite nodules from the Kimberley pipes, SouthAfrica. In M.A. Menzies and C.J. Hawkeswonh, Eds., Mantle meta-

somatism, p.221-311. Academic Press, London.Falloon, T.J., and Green, D.H. (1989) The solidus of carbonated fertile

peridotite. Earth and Planetary Science ktters, 94,364-370.Fenwick, C.S. (199 l) Petrographic and geochemical heterogeneity in man-

tle xenoliths from the Quaternary lavas of the West Eifel: knplicationsfor mantle metasomatism beneath the Rhenish Massif, 187 p. Ph.D.

thesis, University of Reading, Reading, U.K.Fuchs, K., and Wedepohl, K.H. (1983) Relation of geophysical and p€t-

rological models of upper mantle structure of the Rhenish Massif. InK. Fuchs, K. von Gehlen, H. Miilzer, H. Murawski, and A. Semmel,Eds., Plateau uplift-The Rhenish shield-A case history, p.352-363.

Springer Verlag, Berlin.Fujii, T., and Scarfe, C.M. (1985) Composition of liquids coexisting with

spinel lherzolite at l0 kbar and the genesis of MORBs. Contributionsto Mineralogy and Petrology, 90, 18-28.

Green, D.H., and Ringwood, A.E. (1970) Mineralogy of peridotitic com-positions under upper mantle conditions. Physics of the Eanh andPlanetary Interiors, 3, 359-37 l.

Green, D.H., and Wallace, M.E. (1988) Mantle metasomatism by ephem-eral carbonatite melts. Nature, 336,459-462.

Harte, B., and Hawkesworth, C.J. (1989) Mantle domains and mantlexenoliths. In J. Ross, A.L. Jaques, J. Ferguson, D.H. Green, S.Y.O'Reilly, R.V. Danchin, and A.S.A. Janse, Eds., Kimberlites and re-lated rocks: Their mantle/crust setting, diamonds and diamond explo-ration. Geological Society ofAustralia, Special Publication 14, vol- 2,p. 649 -686. Blackwell, Victoria, British Columbia, Canada.

THIBAULT ET AL.: MELTING OF PHLOGOPITE LHERZOLITE

794

Holloway, J.R., and Eggler, D.H. (1976) Fluid-absent melting of perido-tite containing phlogopite and dolomite. Carnegie Institution of Wash-ington Year Book, 75, 636-639.

Hunter, R.H., and McKenzie, D. (1989) The equilibrium geometry ofcarbonate melts in rocks of mantle composition. Earth and PlanetaryScience Irtters. 92. 347 -356.

Kress, V C., and Carmichael, I.S.E. (1988) Stoichiometry ofthe iron ox-idation reaction in silicate melts. American Mineralogist, 73, 1267-t274.

Lloyd, F.E., Edgar, A.D., Forslth, D.M., and Barnett, R.L. (1991) Theparagenesis of upper mantle xenoliths from the Quaternary volcanicssouth-east of Gees, West Eifel, Germany. Mineralogical Magazine,55,95-l 12.

Mattioli, G.S., Baker, M.8., Rutter, M.J., and Stolper, E.M. (1989) Uppermantle oxygen fugacity and its relationship to metasomatism. Journalof Geology, 97, 521-536.

McKenzie, D. (1985) The extraction of magma from the crust and mantle.Earth and Planetary Science l€tters, 74,8l-91.

- (1989) Some remarks on the movement of small melt fractions inthe mantle. Earth and Planetary Science Irtters, 95,53-72.

McKenzie, D., and Bickle, M.J. (1988) The volume and composition ofmelt generated by extension of the lithosphere. Joumal of Petrology,29,625-679.

Meen, J.K. (1987) Mantle metasomatism and carbonatites; an experi-mental study of a complex relationship. Geological Society of America,Special Paper 215, 9l-100.

Meen, J.K., Ayers, J.C., and Fregeau, E.J. (1989) A model of mantlemetasomatism by carbonated alkaline melts: Trace-element and iso-topic compositions of mantle source regions of carbonatite and othercontinental igneous rocks. In K. Bell, Ed., Carbonatites-Genesis andevolution, p. 464-499. Unwin Hyman, l-ondon.

Mengel, K., and Green, D.H. (1989) Stability of amphibole and phlogo-pite in metasomatized peridotite under water-saturated and water-un-dersaturated conditions. In J. Ross, A.L. Jaques, J. Ferguson, D.H.Green, S.Y O'Reilly, R.V. Danchin, and A.S.A. Janse, Eds., Kimber-lites and related rocks: Their composition, occurrence, origin and em-placement. Geological Society ofAustralia, Special Publication 14, vol.1, p. 571-581. Blackwell, Victoria, British Columbia, Canada.

Mertes, H., and Schmincke, H.-U. (1985) Mafic potassic lavas of the

THIBAULT ET AL.: MELTING OF PHLOGOPITE LHERZOLITE

Quatemary West Eifel volcanic field. Contributions to Mineralogy andPetrology, 89, 330-345.

Myers, J., and Eugster, H.P. (1983) The system Fe-Si-O: Oxygen buffercalibrations to I 500 K. Contributions to Mineralogy and Petrology, 82,75-90.

Navon, O., Hutcheon, I.D., Rossman, G.R., and Wasserburg, G.J. (1988)Mantle-derived fluids in diamond micro-inclusions. Nature. 335. 784-789.

Olafsson, M., and Eggler, D.H. (1983) Phase relations of amphibole, am-phibole-carbonate, and phlogopite-carbonate peridotite: Petrologicconstraints on the asthenosphere. Earth and Planetary Science IJtters,64 ,305 -315 .

Seck, H.A., and Wedepohl, K.H. (1983) Mantle xenoliths in the RhenishMassif and the northern Hessian Depression In K. Fuchs, K vonGehlen, H. Miilzer, H. Murawski, and A. Semmel, Eds., Plateau up-lift-The Rhenish shield-A case history, p. 343-351. Springer-Verlag,Berlin.

Spera, F.J. (1984) Carbon dioxide in petrogenesis III: Role ofvolatiles inthe ascent ofalkaline magma with special reference to xenolith-bearingmafic lavas. Contributions to Mineralogy and Petrology, 88,217-232.

Wallace, M.8., and Green, D.H. (1988) An experimental determinationof primary carbonatite magma composition. Nature, 335,343-346.

Wendlandt, R.F., and Eggler, D.H. (1980) The origins of potassic mag-mas: 2. Stability of phlogopite in natural spinel lherzolite and in thesystem KAISiO,-MgO-SiO,-H'O-CO, at high pressures and high tem-peratures. American Journal of Science,280, 421-458.

Wyllie, P.J. (1988) Magma genesis, plate tectonics, and chemical differ-entiation of the earth. Reviews of Geophysics, 26, 370-404.

Wyllie, P.J., and Huang, W.-L. (1976) Carbonation and melting reactionsin the system CaO-MgO-SiO,-CO, at mantle pressures with geophysi-cal and petrological applications. Contributions to Mineralogy and Pe-trology, 54,79-107 .

Wyllie, P.J., and Rutter, M. (1986) Experimental data on the solidus forperidotite-CO,, with applications to alkaline mag,matism and mantlemetasomatism. Eos. 67. 390.

MeNuscnrrr RrcErvED SrprrMsen 6, 1991MeNuscrurr ACCEPTED MencH 13, 1992