American Mineralogist, Volume 80, pages 1252-1260, 1995 Experimental determination of the reactionsmagnesite * quartz : enstatite + CO, and magnesite: periclase+ CO2oand enthalpiesof formation of enstatite and magnesite ANonr,q,M. Kozror.,* Rornnt C. NnwroN Department of the GeophysicalSciences, University of Chicago,Chicago, Illinois 60637, U.S.A. Arsrnlcr The equilibriumP-Z curveof the reaction MgCO. + SiOr: MgSiOr + CO, (l) magnesite qmrtz enstatite was determined by reversed reactions in well-calibrated, solid-pressuremedium (NaCl) piston-cylinder apparatus.External hematite + magnetite + HrO /o, buffers were used in all experiments. A reversedpoint at 715 "C ofthe reaction MgCOr: MeO + CO2 Q) magnesite periclaa was obtained in an internally heated Ar pressurevessel, using a coarse-glainedsynthetic periclase to inhibit quenching back reaction, along with synthetic magnesite and silver oxalate.At7l5 "CourpressureforReactionI is 10.4+ 0.2kbar,andforReaction2it is 590 + 30 bars. From these two reactions we derive LH'l.2eg: 33.41 + 0.83 kJ for enstatite from the oxides. Using an extrapolatedheat-capacity equation for magnesite,we obtain LH!,,2'B: -116.68 kJ (from the oxides)and -llll.68 kJ from the elementsfor magnesite,somewhat lessnegative than the values of either the Holland and Powell (1990) or the Berman (1988) data sets. A calculated 2 kbar temperature for the magnesite + quartz reactionis 492 oC,about 25 'C lower than the data set predictions but in agreement with a prediction based on natural occurrencesof magnesite, quartz, and enstatite in metacarbonatesfrom the Alps by Trommsdorffand Connolly (1990). INrnonucrroN land and Powell, 1990); uncertainties in its thermody- properties propagate Enstatite,MgSiOr, is one of the most important of Earth namic throughout the data sets,with materials. It is the dominant member of orthopyroxene far-reaching effectson all the tabulated data. solid solutions, which is the secondmost abundant min- Despite the chemical simplicity of enstatite, its ther- properties eral speciesin the upper mantle and a definitive phaseof modynamic have eluded exact definition, with the major peridotite assemblages.The presenceof ortho- adverseconsequences for calculations of mineral equilib- pyroxene in quartzofeldspathic and mafic crustal rocks ria. The relative insolubility of MgO in aqueousacids and virtually defines the granulite facies of metamorphism. salt melts has hindered measurementof the enthalpy of (Charlu Enstatite is a component of the COr-metasomatic ultra- formation et al., 1975; Torgeson and Sahama, mafic rocks that have figured importantly in defrnition of 1948)and, hence,of the free energyof formation. Newton (1987) the P-7"history of alpine metamorphism (e.9.,Evans and summarized the large spread of measuredvalues properties Trommsdorfl 1974). Orthopyroxene is a characteristic of these and concluded that measured ther- pre- phasein felsic to mafic volcanic rocks and the large lay- mochemical data for enstatite were not sufficiently practical ered intrusions. Much of its importance to petrogenesis cise for utility. lies in its major role as a participant in mineral reactions Many workers (e.9.,Helgeson et al., 1978) have real- that are the basisfor geothermometry,geobarometry, and ized that if suitable high-quality experimental reactions properties oxygen fugacity analysis.As the simplest and most easily are available, thermodynamic of minerals can characterized magnesiansilicate, enstatite anchors the self- be derived from these data more accurately than from physicochemical consistentthermodynamic data sets(Berman, 1988; Hol- other measurements.For enstatite, the experimental databaseis marginal for this purpose. The only stable formation reaction of enstatite that can be * Presentaddress: Department of Geology,University of Day- investigated in accessiblepressure-temperature space is ton, Dayton, Ohio 45469, U.S.A. the decarbonation reaction of magnesitein the presence 0003-004x/95/ | | r2-r252s02.00 r252 KOZIOL AND NEWTON:EXPERIMENTAL STUDY OF MAGNESITE t253 of quartz: MgO + SiO,: MgSiOr. (3) periclas qurtz MgCO, + SiOr: MgSiO. + COr. (l) enstatite magnesite quflz eDstatitc BecauseReaction 3 does not involve magnesite, the unknown high-temperature thermal properties of mag- This reaction has not previously been reversed because nesite are not an issue in deriving the free energy offor- of very slow reactions in silicate systemsin which CO, is mation of enstatite.The heat-capacityfunctions of ensta- the only volatile substance.An equilibrium temperature tite, periclase, and quartz are very well-measured of 521 "C at2kbar was estimatedby Johannes(1969) by quantrtles. projection of other experimental reactionsin mixed HrO- CO, systemsto pure COr. Using this point, the formation Exprnrvrnxur, METHoDS propertiesof enstatitemay be derived in conjunction with the experimentally determined reaction Piston-cylinder experiments Experiments on reaction magnesite quartz, MgCOr: MgO + CO,. (2) the of and Reaction l, were made in a 0.75 in. diameter piston- magnesite prcriclare cylinder apparatus with NaCl pressure medium and a The two determinations of the univariant P-?" curve of graphite heater element. Calibrated matched pairs of Reaction 2 are somewhat discrepant.The broad reversed W-25o/oRe vs. W-30/oRe thermocouples measured the brackets of Harker and Tuttle (1955) place the reaction temperaturesand served as automatic temperature con- 'C curve approximately 15 higher than the two tight re- trol sensors.The manufacturer's statedprecision is better versedbrackets of Johannesand Metz (1968). Additional than + I "C, and the temperature-controllervariation was uncertainty arises from the poorly known heat capacity less than +2 "C. However, the need for a double-capsule of magnesiteabove 400 {: the data sets use extrapola- high-f", buffer to insure purity of the CO, gas phase re- tions of lower temperature data, and none of these ex- sulted in a separation of as much as I mm between the trapolations consider the possible onset ofcarbonate ro- thermocouple tip and the sample charge. The samples tational disorder at higher temperatures.All these were positioned in a level interval of the temperature uncertainties compound the difficulty in extracting the gradient, on the basis ofprevious gradient measurements, enthalpy and free energy of formation of enstatite from but the temperature uncertainty required independent phaseequilibria. fixed-point calibrations as describedbelow. Experiments Both the Berman(1988) and Holland and Powell(1990) were conducted under piston-out conditions, and maxi- data setsyield a 2 kbar equilibrium point near 520.C for mum nominal pressure variation during an experiment Reaction l, in apparentagreement with Johannes's(1969) was +70 bars. There is almost no frictional correction projection. However, Trommsdorff and Connolly (l 990) for NaCl under these conditions; the pressuresare be- have challengedthe data set predictions. Their revision lieved accurateto 4200 bars at l0 kbar. is based on analysis ofnatural paragenesesin carbonate Samples consisted of homogenized powder mixes of rocks from the Alps. According to their assessment,Re- synthetic enstatite, synthetic magnesite,and natural quartz action I should lie at significantly lower temperatures than sealed in I mm diameter platinum tube segmentswith the reaction of dolomite + quafiz to diopside and COr. weighed amounts of silver oxalate (AgrCrOo)as the CO, The latter reaction was experimentally determined by source. Magnesite was synthesizedfrom Baker hydrous Slaughteret al. (1975) to be closeto 520.C at 2 kbar. magnesium carbonate reagentby recrystallizing under CO, Trommsdorff and Connolly's revision locates a 2 kbar pressureat 400 'C and about I kbar in a Morey vessel equilibrium point of Reaction I near 490 .C. Their con- for 2 d. The result was well-crystallized magnesitewith sequent revision of PTX topology in the system CaO- averagegrain size of about l5 pm. The unit-cell constants MgO-SiOr-HrO-CO, is in better agreementwith the ob- were determined by powder X-ray difractometry, scan- served metamorphic zonation in metacarbonatesin the ning at Vq"/min with an internal quartz standard. Five central Alps than the topologies predicted by the widely major peaks in the range 30-55' 20 (CuKa, radiation) used data sets.They suggestthat the discrepancywith the were usedto calculatethe cell constantsby a least-squares Berman (1988) data set can be rectified by assuminga method:ao: 4.636+ 0.001and bo: 15.016+ 0.001 free energy of formation of magnesite2 kJ more positive A. Enstatite was synthesizedfrom this magnesitemixed than that used by Berman (1988). This assumptionis intimately with finely ground natural quartz in sealed equivalent to relocating Reaction 2 to considerablyhigh- platinum containersat 1000'C and 15 kbar for 6 h. The er pressuresand lower temperaturesthan those predicted product was nicely crystallized enstatite averaging30 pm by Berman. The discrepancyin Reaction I could, how- in length with the following unit-cell constants,based on ever, be equally well accountedfor by an adjustment of seven X-ray diffraction peaks in the range 26-40'20: ao the free energy of formation of enstatite. :18.226 + 0.003,r0:8.811 + 0.002,andco:5.178 The present work seeksto resolve these uncertainties + 0.001 A. Another sample of synthetic enstatite used in using improved experimental determinations of Reac- some experiments was prepared by Eckert et al. (1992) tions I and 2. A linear combination of these