The Reaction Talc * Forsterite = Enstatite + Hro: New Experimentalresults and Petrologicalimplications Ar,Rsour

AmericanMineralogist, Volume 83, pages 5l-57, 1998

The reaction talc * forsterite = enstatite + HrO: New experimentalresults and petrologicalimplications Ar,rsouR. PlwrBv

Departmentof Earth Sciences,University of Manchester,Oxford Road, ManchesterM13 gPL, U.K.

Ansrnacr The reaction talc * forsterite = enstatite + HrO has been investigatedbetween 6 and 20 kbar. Previous high-pressureexperimental studies suggestedvarious reaction positions, mostly with positive P-7 slopes. The new results show that the reaction has a negative slope in the pressurerange studied. It was bracketedbetween 640 and 680'C at l0 kbar, between 620 and 640 "C at 15 kbar, and between 580 and 600 'C at 20 kbar. The growth of antigorite at high pressuresprevented the reaction from being determinedabove 20 kbar. The reaction position is consistentwith the previous low-pressuredata of Chernosky et al. (1985), is in very good agreementwith the position calculated using the thermodynamic databaseof Berman (1988), and is in reasonableagreement with the calculation using the thermodynamic databaseof Holland and Powell (1990). The high reaction temperatures suggestedby previous high-pressureexperiments may reflect metastableor quench growth of talc in those experiments.Talc in equilibrium with forsterite is stable over a wide P-7 range in hydrothermally altered and metamorphosedulffamafic rocks and may be important for carrying HrO into subduction zones. However, talc dehydration occurs at too shallow a depth for the HrO to contribute directly to subduction zone volcanism.

INrnonucrroN and releasesits HrO through a series of dehydration re- Talc, Mg.SioO,o(OH)r,is a common mineral in hydro- actions. These reactions may play a major role in sub- thermally altered and metamorphosedultramafic rocks. In duction zone processesby providing the HrO necessary these rocks it is commonly associatedwith olivine and for volcanism or by triggering earthquakes.In addition, other hydrous minerals such as serpentineand amphibole. HrO releasedby these actions may become incorporated The maximum stability of end-membertalc, as given by in new high-pressurehydrous phases,e.g., densehydrous the reaction talc : enstatite + quartz + H,O, occurs at magnesiumsilicates or in nominally anhydrousminerals. -800 "C at 10-30 kbar (Chernoskyer al. 1985;Jenkins These phases may transport the HrO to much greater et al. 1991). However, in typical altered ultramafic rock depth than the stability of common metamorphicminerals in which talc coexistswith olivine, it breaksdown at low- such as talc. er temperaturesthrough reaction with the olivine. The Several experimental investigationsof reaction t have end-member reaction in the system MgO-SiOr-HrO been published. For example, as part of a comprehensive (MSH) is: study of phaseequilibria in the systemMSH at low pressures,Chernosky et al. (1985)determined its positionup talc + forsterite: 5enstatite+ H,O (1) to 6 kbar. Their results on this reaction and others in- Ta, MgisiaOr0(OH)2 Fo, Mg2SiOl En, MgSrOl volving talc formed the basis for the determinationof the This reaction is significant becauseit determinesnot only enthalpy of talc in the internally consistentthermodynam- the maximum stability of talc in equilibrium with for- ic databasesof Holland and Powell (1990) and Berman sterite but also the highest temperatureat which HrO can (1988).The experimentaldata of Chernoskyet al. (1985) be stored in typical altered ultramafic rock at moderateto and the position of reaction I calculated using the data- high pressures.The amphibole-bearingMSH assemblage basesare shown in Figure 1, togetherwith other previous anthophyllite + forsterite is stableto higher temperatures data on reaction1. than talc + forsterite but is not stable above -8 kbar The results of Chernoskyet al. (1985) on talc reactions (Chernosky et al. 1985). Serpentine is more abundant are generally self-consistentand so the calculatedreaction than talc at low pressuresand temperatures,but below I passesthrough their low-pressure brackets. However -15 kbar it breaks down at a temperature lower than this consistency does not extend to the previous high- reaction 1. Reaction I could therefore play an important pressureexperimental studies, which have produced re- role in affecting the conditions of HrO storage in the sults inconsistent with each other and with the calcula- Earth's mantle, parlicularly in subduction zones. Here, tions. The most-cited high-pressurereversal experiments hydrated oceanic lithosphereis subductedinto the mantle are those of Kitahara et al. (1966), who obtained phase- 0003-004x/98/0102-0051$05.00 sl 52 PAWLEY: TALC-FORSTERITE REACTIONS

H a2O Chedoslyeral i1985) H Kilahanerdl (1966) Fo- o 15 H Yamamoroand Atimoto (1977) L i3'---O Ulrer and Tronmdod ( I 9951

700

Temperature('C)

FrcurB 1. Resultsof previousstudies on thereaction Ta f Temperature("C) Fo - En + HrO,and calculation of thereaction using the com- puterprograms THERMOCALC (Powell and Holland 1988) and TWQ (Berman1991). Program versions are THERMOCALC v2.5 (thermodynamicdata generated February 1997) and TWQ v1.0.Reaction temperatures were also calculated using TWQ v2.01(March 1997) and up to 30 kbarare within 2 'C of those calculatedusing vl 0 Thecalculation using v2.01 is notshown, however,as this versionof the databasedoes not includeanti- gorite,as requiredfor the other reactionsshown in Figure2. Fo En SiO2 : grew, : Closedsymbols Ta t Fo opensymbols En grew. Frcuns2. Equilibriain thesystem MSH around the invariant pointinvolving Th, Atg, Fo, En, and H.O, calculated using TWQ. Solidlines are stable equilibria, dashed lines metastable. Com- equilibrium brackets between 10 and 30 kbar. As shown patibility diagramsare shownfor stableassemblages in each in Figure 1, their bracket at 10 kbar occurs at a much field.The numberscorrespond to thereactions given in thetext. point higher temperaturethan the 6 kbar bracket of Chernosky Notethat stability levels relate to thisinvariant only. et al. (1985). Given that the data of Chernosky et al. (1985) suggestthat the maximum thermal stability of Ta The stabilities of other hydrous MSH phasesare also + Fo is at -5-6 kbar, with the reaction bending back to relevant to this study. For example, the experiments of lower temperaturesat higher pressures,the two sets of Ulmer and Trommsdorff (1995) showed that the serpen- resultscannot be reconciled.Chernosky et al. (1985)sug- tine mineral antigorite [Atg, Mgo.Si.oO*.(OH)ur]is stable gestedthat the apparentlyhigher-temperature brackets of at low temperaturesand high pressures.At low pressures Kitahara et al. (1966) are becauseof inaccurate calibra- and high temperatures,antigorite dehydratesthrough the tion of their piston-cylinder apparatus,whereas Berman reaction (1988) suggestedthat the error lies in the use of poorly antigorite : talc * forsterite + HrO (2) characterizedstarting materials synthesizedin very short experiments.This substantialdiscrepancy between high- which has been reversedbetween 2 and 15 kbar by Evans and low-pressure data indicates the need for a redeter- et al. (1976).Their bracketswere usedby Berman(1988) mination of reaction 1 at pressuresabove 6 kbar. in deriving the enthalpy data for antigorite in his data- Reaction I has been investigatedin other studiesof the base.The position of reaction 2 and all others around the MHS system. For example, in synthesis experiments at invariant point involving Ta, Atg, Fo, En, and H,O, as =29 kbar, Yamamoto and Akimoto (1977) obtained a re- calculated using the program TWQ, are shown in Figure action position -80 'C higher than that suggestedby Ki- 2. (Becauseantigorite is not presentin THERMOCALC, tahara et al. (1966), which is itself much higher than the a similar calculation cannot be performed using that calculatedcurve (Fig. 1). Yamamoto and Akimoto (1977) database.) accepted the possibility of disequilibrium in their short Anthophyllite is anothermineral potentially relevant to synthesis experiments and cited the need for reversal the presentstudy, becauseat low pressuresand high tem- expenments. peraturesTa + Fo react to anthophyllite rather than en- The assemblageTa + Fo was produced in the inves- statite. However the reaction is not directly relevant to tigation of serpentinestability by Ulmer and Tiommsdorff this study, becausebelow 700'C anthophylliteis only (1995) and again appearedto be stable to higher temper- stable below 8 kbar, breaking down at higher pressures atures than calculated. Those experiments, although not to Ta + En (Chernoskyet al. 1985). designed specifically to investigate reaction l, reinforce This paper reports the reinvestigation of reaction I be- the need for a redeterminationof this reaction by reversal tween 6 and 20 kbar. The results are in very good agree- expenments. ment with the position of the reaction calculated using PAWLEY: TALC-FORSTERITE REACTIONS 53

TWQ and are close to the position calculated using TeeLe 1. Exoerimentalresults THERMOCALC. The reasonsfor the discrepancies with EXOer- previous experimentsare examined. ment Starting Duration number material P (kbar) f ("C) (hours) ExpBmlrBNrAL PRocEDURES The reaction talc + forsterite = enstatite + H2Ousing natural talc Starting materials for the experiments on reaction I 23 TFE 6 0 670 280 No reaction were 1:1:5mixtures of naturalor synthetictalc, synthetic 21 TFE 6 0 679 312 No reaction 19 Expt 4 8.0 660 353 Ta + Fo'f. forsterite, and synthetic enstatite. The natural talc has 8 TFE 10 O 620 792 Ta + Fo'l been used in previous studies(Pawley et al. 1995; Pawley TFE 10.0 640 258 Ta + Fo 1" and Wood 1995) and is of ideal composition 7 TFE 1o.o 640 845 Ta + Fo 1" except for ^ TFE 10 0 660 237 No reactiont a I atom%oreplacement of Mg by Fe. The synthetic sam- TFE 10O 680 163 En 1 ples were preparedfrom oxide mixes, talc at 2 kbar, 650 I TFE 150 600 74O Ta+Fot + 1+ "C for 120 h 10 TFE 15.0 620 580 Ta Fo and enstatiteat 5 kbar, 900'C for 48 h. The tl TF 15 0 620 674 No reaction forsterite was kindly provided by S.J. Kohn (University to TFE 15 0 640 238 En 1 of Bristol). Characterizationby X-ray diffraction (XRD) 18 TFE 2Oo 580 431 Ta + Fo 1 1+ ntgl 17 TFE 205 600 574 En 1 (+ Atg) and optical examination showed well-crystallized sam- 'I TFE 2OO 660 93 En 1 ples, with negligible impurities in the talc and forsterite 20 TFE 250 550 185 Ta + En + Atg and a trace of quartz in the enstatite.The reaction of this TFE 25 0 530 312 Ta + En + Atg quartz during the experimentshas an insignificant effect The reaction talc + forsterite = enstatite + HrO using synthetic talc on reaction l. 13 Expt 12 10 0 660 484 No reaction 14 STFE 10 0 680 390 Enrs Starting materials were sealedin 2 mm diameter plat- tc Expt 14 10.0 680 520 Enl inum capsuleswith 6-9 wt% distilled H,O. This ensured 12 STFE 15.0 620 690 ta + Fo I that there was -5 wt%oHrO in excessof that required to The reaction talc + antigorite = enstatite + H2O react the entire sample to Ta + Fo. 24 Expt 20 20.0 620 163 Ta + Atg t 2O.O 640 289 En 1 All experimentswere carried out in a half-inch piston- 25 Exptzz cylinder apparatuswith NaCl as the pressuremedium. A Nolesi Starting materials: TFE : natural talc + synthetic torsterite + synthetic enstatite TF : natural talc + synthetic forsterite + seeds of non-end-loadedpiston-cylinder apparatus was used for enstatite. experimentsup to 15 kbar. These experimentswere run Results:All resultsare determinedfrom changesin XRD peak height "hot-piston-out", whereby they were pressurizedto -1 ratiosfrom the startingmaterial: 1 : >30% increase,J : >30% de- crease Behavior of Ta and Fo is consistent excepl as footnoted The kbar above the desired experimentpressure, heated to ex- presenceol a smallamount of Atg can be ignoredin experiments17 and periment temperature,and then the pressurebled off. For 18 (Atg in brackets)but not in experiments2O and22 (no brackets)See for furtherdiscussion experimentsabove 15 kbar piston-cylinder text 'l an end-loaded t Fo: En but no chanoein Ta: En. apparatuswas used, and experimentswere run "hot-pis- t ryoc.lrapg info _En-butTa:En J ton-in", in which the pressurewas taken to experiment tFo:hn tDUt ta:En.r. pressurewhile heating. 5FO:EnJbutta:tsnl In both apparatus, pressure was maintained within +0.05 kbar dunng the experiment by means of a small motorized variable-volume oil reservoir attached to the Only the ratio of Fo to En peak heights (hereafterFo:En) pressuresystem. Pressurewas calibrated at 600 "C using could be used, becausetalc peak heights sometimesgave the reaction albite : jadeite + qu,urtz(Holland 1980) and inconsistent results, as explained in the next section. no pressurecorrection was found to be necessaryfor ei- When reaction was observed, the change in Fo:En was ther apparatus.The pressureuncertainty is +0.5 kbar. always >30Vo. Temperaturesin the experimentswere measuredusing Rnsur.rs Pt/PtlORh thermocoupleswith no correction for pressure, oC. = and controlled within + I Capsuleswere flattenedand The reaction talc * forsterite enstatite * HrO placed perpendicularto the thermocouple,separated by a Results of experimentson the reaction Ta + Fo : En thin ruby disc to prevent puncturing during the experi- + H,O are listed in Table 1 and plotted in Figure 3. The ment. Temperaturesin the capsuleswere not calibrated experiments using natural talc appear first followed by independently,but an experiment on the melting of albite those using synthetic talc. Also listed are the results of + HrO was utilized to check that temperatureswere not two experimentson the reaction underestimated.At 1l kbar,680'C, 10'below the melting talc * antigorite : enstatite + HrO. (3) point determinedby Goldsmith and Jenkins (1985), albite + H,O did not melt. In several experiments,the apparentreaction direction Experiment products were examined optically and by suggestedby changesin Ta:En is different from that im- XRD. In many cases,reaction direction could be deter- plied by changesin Fo:En (seeTable l). In thoseincon- mined from the optical examination but was inferred sistent experimentsusing natural talc, the apparentTa:En more unambiguously by comparing ratios of XRD peak either did not change or went down, whereas Fo:En in- heights betweenstarting material and experimentproduct. creased or did not change. Thus, the reaction position 54 PAWLEY: TALC-FORSTERITE REACTIONS

I Ta + Fo orew In the original experiment,the changein Ta:En suggested delerminedfrom changes rt\oteacton'--.,-^- ^,_ ___-^i__----- I ; -- l in Fo:EnXRD peak heights that talc had increasedwhereas the follow-up experiment u tn grew + Ta + Fo qrew ; actually produced a decreasein Ta:En (experiment 15 : .: I Oerermrneorom cnanges .- No reaclon I in Ta:En XRD peak helghts enstatite is stable at the conditions > En grew rerun of 14). Clearly 6 of those experiments,as indicated by the Fo:En changes. Thus, whether the assemblageTa + Fo is stable or not,

d-' it appearsthat talc initially recrystallizes during all ex- ! 5 periments, giving XRD peaks with different intensities 920 l from the starting material. The effect on natural talc is $rs clearly different from the effect on synthetic talc, and L 10 hence their peak heights behave differently. Experiment 10, in which the starting material contained 5 -50Vo enstatite,showed an apparent decreasein Ta:En, 0 suggestingthat enstatite is stable at 15 kbar and 620 "C. 450 500 550 600 650 700 750 To test this result, experiment 11 was carried out at the crystals in the start- Temperature ('C) same conditions using enstatiteseed ing material. If enstatite were stable at these P-T condi- FtcunB 3. Results of experimentson the reaction Ta + Fo tions, then an increasein its amount would be more no- : En + H.O (squares).Reaction direction was inferred from ticeable from seeds.No reaction was observed, a result changesin Fo:En peak XRD height ratios relative to the starting that is not conclusive, but may support the interpretation material. Arrows below squaresshow reaction direction inferred from the increase in Fo:En observed in experiment l0 from Ta:En peak heights in experiments using natural talc; ar- + rows above squaresshow Ta:En changesin experimentsusing that Ta Fo is the stable assemblageat theseconditions. synthetic talc. Experimentsin which reaction direction could not Experiments involving antigorite be inferred becauseof signilicant antigorite growth are shown as crosseswithin squares,and experiments on the reaction Ta + At high pressuresand low temperatures,experimental Atg - En * HrO are shown as circles (filled : Ta + Atg grew, products contained antigorite. This is consistent with open : En grew). Also shown are calculatedstable curves and these conditions being at higher pressures than the the metastableextension of reaction I from Figure 2, two cal- reaction culated P-T paths of a subducting slab (Peacock 1990) and the range in depths (in kbar of lithostatic pressure)of subducting talc * forsterite: antigorite + enstatite (4) slabsbelow subductionzone volcanoes(Gill 1981) Seetext for forms of antigorite-containingexperimen- further discussion. Three different tal products were produced. At 20 kbar, 580 'C (experiment 18) there is a significant decreasein the amount of suggestedby a comparison of Ta with En peaks would En relative to both Ta and Fo from the starting material be at a lower temperaturethan indicated by a comparison and some antigorite is also present. This experiment is of Fo with En peaks. The temperaturedifference is -20 probably at a temperaturebelow the metastableextension "C at 10 and 15 kbar. of reaction 1, becausesuch a strong increasein Ta + Fo In one experiment using synthetic talc, peak height ra- is unlikely to occur far out of its stability field. However, tios were also inconsistent.In this case, the relative be- it is possiblethat no reaction occurs on reaction 1; instead havior of Ta:En and Fo:En was opposite to that of the Atg + Ta formed from En + HrO in reaction 3, which natural talc experimentsin that Ta:En increasedwhereas also producesan increasein Fo:En. Fo:En decreased.Thus, if only Ta:Enchanges were used These proportions of phasescontrast with the appear- to infer reaction position, at 10 kbar the position of the ance of experiment 17 (carried out at 20 'C higher), reaction involving natural talc would apparentlybe more which,, relative to Ta * Fo, contains a little more anti- than 20'C below the position involving synthetic talc. gorite than experiment 18 and considerably more ensta- Such a difference is considered unlikely becauseit has tite. It is not possible to determine the absolute propor- not been observedin previous studies on the higher-tem- tions of the phases,so it is not clear whether the increase peraturereaction talc : enstatite + coesite + HrO. Work in En:(Ta + Fo) in experiment 17 relative to experiment in this laboratory has shown no difference in the position 18 is becauseof crossing reaction I or an increasein the of that reaction using natural and synthetic talc at 30 kbar extent of reaction 4. It is assumedhere that the conffast (Chinnery, unpublished data). Howeveq other studies in (Ta + Fo):En between experiments l7 and l8 is good have yielded a lower-temperaturereaction for the dehy- reason for placing the metastableextension of reaction 1 dration of synthetic talc than of natural talc (e.g., Bose between these two P-7 points. and Ganguly 1995).This is one reasonto suspectthat Ta: The changes in ratios of Ta, Fo, and En XRD peak En changesfrom the starting material do not give a ffue heights in experiments 17 and 18 are roughly consistent indicationof reactiondirection. with those in the lower-pressureexperiments. This sug- The unreliability of Ta:En peak height ratios was con- geststhat the appearanceof antigorite can be ignored, as firmed by rerunning an experiment using synthetic talc. discussedabove. However, the same is not true for ex- PAWLEY: TALC-FORSTERITE REACTIONS 55 periments 20 and 22 at 25 kbar in which Fo disappeared, periment demonstratedthat measuredtemperature could but there was no changein Ta:En XRD peak height ratios. not be more than 10'C too low. Thus the new experi- These are the only experiments using natural talc in mental results support the assertionsof Chernosky et al. which a decreasein Fo:En is not accompaniedby an ob- (1985) and Berman (1988)that the positionof reactionI vious decreasein Ta:En. Clearly, forsterite is reacting out determined by Kitahara et al. (1966) is incorrect. The in an antigorite-producingreaction. The compatibility di- results of Kitahara et al. (1966) may be in error partly agrams (Fig. 2) show that in this field the stable assem- becauseof inaccuratepiston-cylinder calibration, as sug- blage is either En + Atg * Ta or Atg + Ta + H,O, gestedby Chernoskyet al. (1985),and panly becauseof depending on the HrO content of the bulk composition. the poorly crystalline nature of their synthetic starting The nominal HrO content places the bulk composition materials,as suggestedby Berman (1988). However the very close to the Atg-Ta join, so at equilibrium, the sam- increasing discrepancyfrom the calculated reaction with ple should contain no more than a minor amount of en- increasing pressure suggeststhat these are not the only statite. Given longer experiments,it is expected that en- sourcesof error. There arejust three reversalexperiments statite would almost disappearfrom experiments20 and of Kitahara et al. (1966) in which Ta + Fo were appar- 22. ently produced stably from En + HrO. However, the The products of experiments20 and 22, in which for- amount of water in these experiments is not quoted. I sterite had reactedout, were used as starting material for suggest that it may have been sufficiently large that a the two experimentson reaction 3. Reaction direction was significant amount of MgO and SiO, dissolvedin the fluid unambiguously determined in both experiments from during the experiment, with the resulting solid composi- changesin heights of Ta + Atg peaksrelative to En peaks tion being displacedfrom enstatitetoward forsterite (SiO, in XRD patterns.Consequently, these experiments brack- is more soluble than MgO). This produced the trace of et reaction 3 between 620 and 640 "C at 20 kbar. This is forsterite observedin the experimentsand the talc formed consistentwith the position of this reaction calculatedus- on quench. ing TWQ @igs. 2 and 3). The discrepancy with the observations of Ulmer and Trommsdorff(1995) of a positive P-Z slope of reaction DrscussroN I up to -20 kbar requires some consideration.The ex- positive Comparison with calculated reactionsand with previous periments that force the reaction to have a slope + + En experimental results produced talc from a starting material of Atg Fo (+ brucite to provide H,O) at 18 kbar, and 690 and 710 : This reinvestigation of the reaction Ta + Fo En + "C (Fig. 1). According to the present study and the cal- HrO was carried out becauseprevious studies had pro- culations using TWQ, at theseP-Z conditions, which are duced a wide range of reaction positions at high pres- at temperaturesabove the invariant point (Figs. 2 and 3), sures,none of which agreed with the position calculated the antigorite in the starting material should break down using existing thermodynamic data. The new results do to Fo * En. The amount of antigorite in the experiments agree with the calculatedreaction. As shown in Figure 3, of Ulmer and Trommsdorff (1995) was observed to de- the position of reaction 1 calculatedusing TWQ is in very creasebut the amount of enstatite did not increase.The good agreementwith the brackets obtained in this study: antigorite used was a natural sample, which contained it passesthrough the l0 kbar bracket and is at the high- minor impurities. Perhaps an Al-rich talc formed when temperatureend of the brackets at 15 and 20 kbar. The antigorite broke down (chlorite formed at higher temper- reaction position calculated using THERMOCALC is 7 atures). Alternatively, talc may have formed on quench 'C higher than the TWQ reaction at l0 kbar and 22'C (or metastably) as antigorite broke down, because its higher at 20 kbar, and so diverges slightly from the ex- structure is more similar to talc than to enstatite or for- perimental brackets with increasing pressure. However, sterite. Becausethe experimentsof Ulmer and Tiomms- the divergence is not great, considering the uncertainties dorff did not aim to determine talc stability, the position in the thermodynamic data used in the calculations,par- of reaction I does not affect their conclusions about the ticularly in the equation of stateof HrO. The brackets on high-pressurestability of antigorite, but it does affect the reaction 3 are also consistentwith its calculatedposition stability of antigorite close to the invariant point. Clari- using TWQ, and all experimentsthat produced antigorite fication of this discrepancy between studies clearly re- lie in the calculated stability field of that mineral. quires a careful comparison of reaction positions deter- All the previous high-pressurebrackets on reaction 1 mined using simple compositions in synthetic systems are at higher temperaturesthan obtained in this study, the with those determined using complex natural difference increasing from a few degrees Celsius at 10 compositions. kbar to -100 'C at 20 kbar suggestingthat perhapstem- peraturesin this study have been underestimated.How- Inrpr,rca,rroNs FoR NATURALTALC-BEARTNG ever, this can be ruled out becausethe sample rurange- ASSEMBLAGES ment in the piston-cylinder apparatus is such that Talc typically forms by the hydrothermal alteration of thermocouple and sample are equidistant from the fur- ultramafic rocks or metamorphismof serpentinite.Much nace hot-spot. Moreover, the albite + HrO melting ex- of the oceanic crust close to spreadingcenters experienc- 56 PAWLEY: TALC-FORSTERITE REACTIONS es hydrothermal alteration. Although most of the crust is et al. 1991),decreasing to -710 "C at 50 kbar (Pawley basaltic, ulffamafic rocks are also present at shallow lev- and Wood 1995). This result suggeststhat, in suitably els in the crust, where they can become serpentinized silica-rich compositions,talc could transport HrO to con- (e.g., Cannat 1993).The extent of hydrothermal alteration siderable depths in subduction zones. Silica effichment of the mantle below the oceanic crust is unknown but of peridotite is not common at low metamorphic grades. may also be significant. Most of this oceanic lithosphere However, SiO, solubility in HrO increaseswith increasing eventually is subducted.The HrO incorporatedin hydrous pressure and temperature.As a result, silica metasoma- minerals is releasedand may trigger melting and subduc- tism of mantle peridotite may occur at high pressuresand tion zone magmatism.The possibility of talc contributing temperaturesin subductionzones (Manning 1994, 1995). H,O to subduction zone magmatism can be evaluatedby Thus, metasomatism could converl Ta-Fo assemblages comparing its breakdownreactions with P-Tpaths of sub- into Ta-En assemblages,with a correspondingexpansion ducting slabs and with P-Z conditions of subductionzone of the talc srability field. magma genesis.Two calculated P-T paths for the top of a subducting slab (Peacock 1990) are shown in Figure 3, AcrNowr-BncMENTS along with the approximate range in lithostatic pressure Financial support for this study was provided by NERC grant GR3/ of the slab below subductionzone volcanoes (Gill 1981). 10308 Helpful comments by JosephChernosky and Bernard Evans are The calculated P-7 paths are for two different ages of gratefully acknowledged subducting lithosphere: 10 Ma (path A) and 50 Ma (path B). RnronnNcns crrED Figure 3 shows that both P-Z paths begin in the anti- Berman,R G (1988) Internally-consistentthermodynamic data for min- gorite stability field, consistent with the common abun- eralsin the systemNa,O-K.O-CaO-MgO-FeO-Fe.O.-Al,O.-SiO,-TiO,- dance of serpentine in altered peridotite. However the H,O-CO, Journalof Petrology,29,445-522 compatibility diagrams(Fig.2) show that the H,O content -(1991) Thermobarometry using multiequilibrium calculations: a required to stabilize antigorite is greaterthan that required new techniquewith petrologic applications CanadianMineralogist, 29, to stabilize talc. 833-855 Consequently,if conditions in the slab Bose, K. and Ganguly, J (1995) Experimental and theoreticalstudies of are HrO-undersaturated,talc might be stabilized instead the stabilities of talc, antigorite and phase A at high pressureswith of antigorite. applcations to subductionprocesses Earth and PlanetaryScience Let- Under the relatively warm P-T path calculated for the ters, 136. 109-121 (1993) subduction of young oceanic crust, reaction 2, Atg : Ta Cannat, M Emplacementof mantle rocks in the seafloor at mid- -600 oceanridges. Journal of GeophysicalResearch, 98,4163-4172. + Fo + H.O, is crossedat "C, and so the amount Chernosky,JV, Day, HW., and Caruso,L.J. (1985) Equilibria in the of talc in the rock increasesas antigorite reacts out. The system MgO-SiO,-H,0: experimentaldetermination of the stability of talc then reacts with forsterite in reaction I at -670 "C, Mg-anthophyllite American Mineralogist, 7O, 223-236. which is close to the highest possible thermal stability of Evans, B W., Johannes,W., Oterdoom, H., and Trommsdorfl V (1976) Stability of chrysotile and antigorite in the serpentinitemultisystem. talc in equilibrium with forsterite. Under more typical SchweizerischeMineralogische und PetrographischeMitteilungen, 56, cooler paths such as B, Ta + Fo does not survive above '79-93 -15-20 kbar, where reaction 4, Ta + Fo : Atg * En, is Gill, J (1981) Orogenicandesites and plate tectonics,390 p Springer crossed.This reaction marks the highest pressureat which Verlag, New York (1985) the assemblageTa + Fo is stable. Clearly it occurs at a Coldsmith, J and Jenkins, D The hydrothermal melting of low and high albite. American Mineralogist, 70,924 933. much shallower depth than the depth of magma genera- Holland, TJ.B (1980) The reaction albite : jadeire + quartz determined tion in subduction zones, so the dehydration of talc in experimentallyin the range600-1200'C AmericanMineralogist, 65, equilibrium with forsterite cannot be directly responsible 129-134 for triggering subduction zone magmatism under any P- Holland,TJ.B. and Powell,R (1990)An enlargedand updatedinternally I path. However, consistentthermodynamic dataset with uncartaintiesand correlations: under warm P-Z paths, HrO produced the systemK,O-Na,O-CaO-MgO-MnO-FeO-Fe,O.-Al,O.-TiO,-SiO.-C- in reaction I may rise into a part of the subducting slab H,-O, Journalof MetamorphicGeology, 8, 89-124 or overlying mantle wedge with a different bulk compo- Jenkins,DM, Holland,TJ.B., and Clare,A.K. (1991)Experimental de- sition, allowing other assemblagesinvolving hydrous terminationof the pressure-temperaturestability fietd and thermochem- phasesto form. Under cool P-Z paths, the antigorite pro- ical propertiesof synthetic tremolite American Mineralogist, 76, 458 469. duced in reaction 4 is able to transport HrO to a greater Kitahara, S , Takenouchi,S , and Kennedy,G C ( 1966) Phaserelations in depth until it dehydratesthrough reaction 5, Atg : p6 1 the system MgO-SiOr-HrO at high temperaturesand pressuresAmer- En + H,O. According to Ulmer and Trommsdorff (1995), ican Journalof Science,264,223-233 this scenario may provide the H,O for subduction zone Manning,C.E (1994)The solubilityof quartzin H,O in the lower crust volcanism. and uppermantle. Geochimica et CosmochimicaActa, 58, 4831-4839. -(1995) Phase-equilibriumcontrols on SiO, metasomatismby aque- WhereasTa + Fo has a relatively low P-I stability, in ous fluid in subductionzones: reaction af constantpressure and tem- bulk compositions more silica rich than enstatite,talc is perature InternationalGeology Review, 37, 1074-1093 still present above the Ta + Fo reactions and does not Pawley,AR and Wood, B J. (1995) The high-pressurestability of talc break down until the reaction talc : enstatite + quartzl and l0 A phase: Potential storagesites for HrO in subductionzones. American Mineralogrst,80, 998 1003 coesite+ -800'C HrO is encountered.This occursat in Pawley, A R , Redfern, S A T, and Wood, B J (1995) Thermal expansiv- the quartz stabiliry field (Chernosky er al. 1985; Jenkins ities and compressibilitiesof hydrous phasesin the systemMgO-SiO,- PAWLEY: TALC-FORSTERITE REACTIONS 57

H,O: talc, phase A and 10A phase Contributronsto Mineralogy and Ulmer, P and Trommsdorff, V. (1995) Serpentinestability to mandedepths Petrology, 122, 301-307 and subduction-relatedmagmatism. Science, 268, 858-861. Peacock,S.M. (1990) Fluid processesin subductionzones. Science,248, Yamamoto,K. and Akimoto, S. (1977) The systemMgO-SiO'-HrO at high for hydroxyl-chondrodite, 329-337 pressuresand temperatures-stability field hydroxyl-chnohumite and l0 A-phase. American Jdurnal of Science, Powell, R and Holland, T.JB. (1988) An internally consistentthermo- 271. 288-312. dynamic datasetwith uncertaintiesand corelations: 3. Applications to geobarometry,worked examples and a compuier program. Journal of MaNuscnrrr RECETvEDApnn- 14, 1997 Metamorphic Geology, 6, 173-204. Manuscnn'r AccEFTEDSeprsMsen 10, 1997