A Calorimetric Study of Synthetic Amphiboles Along the Tremolite

Home , Coupled substitution, Hornblende, Tschermakite

American Mineralogist, Volume 79, pages I I I 0- I I 2 2, I 994

A calorimetric study of synthetic amphibolesalong the tremolite-tschermakitejoin and the heats of formation of magnesiohornblendeand tschermakite EucnNr A. Snnrr.rx Departmentof Geologicaland GeophysicalSciences and PrincetonMaterials Institute, Princeton University, Princeton, NewJersey 08544, U.S.A. Dlvro M. Jpxrrns Departmentof Geologicaland EnvironmentalSciences, State University of New York at Binghamton,Binghamton, NewYork 13901.U.S.A. Ar,nx,lNon c, NlvnorsxY Departmentof Geologicaland Geophysical Sciences and PrincetonMaterials Institute, Princeton University, Princeton, NewJersev 08544. U.S.A.

Arstnlcr A suite of synthetic calcic amphiboles in the tremolite-tschermakite (Tr-Ts) serieshas been studied by high-temperature drop-solution calorimetry in molten 2PbO.BrOr. The pseudobinaryjoin, which is shifted toward magnesiocummingtonite(Mc) by l0 mol0/0, extends from tremolite (Tr), Ca,rMgrrSirOrr(OH)r, to magnesiohornblende(MgHb), Ca,,MgrAlrsirorr(OH)r, and representsthe operation of one full Mg-Tschermak substitution, I6rMg,l4rsi= IoJdl,t+l[],into the tremolite fOrmula. Transmission electron microscope (TEM) examination of the amphiboles shows that all the samples contain a low concentration of (0 l0) chain multiplicity faults (CMFs), suggestingthat the increaseof Al in the amphibole does not significantly increaseor decreasethe tendency ofthese defects to form. The calorimetric measurementsindicate that the energy changeassociated with one full Mg-Tschermak substitution is quite small, on the order of -5 to -10 kJlmol. Enthalpies of formation from the elements at 298.15 K and I bar have been calculated using the new calorimetric data for the magnesiohornblende(MgHb) and tschermakite (Ts)end-members.ThevaluesofAlf forMgHbandTsare -12401.2 + 10.6and-12527.7 + 16.4 kJlmol, respectively.The entropy of MgHb, at 298.15 K and I bar, has also been calculated using four different site-mixing activity models involving the octahedral (M2) and tetrahedral (Tl) sites and the experimental phase-equilibrium data of Jenkins (1994). Two of the models, the two-site and four-site coupled models, are essentially indistin- guishableand gave the most consistentvalues for ^q"Hb,575.1 + 3.9 and 575.6 + 3.9 J/(mol. K), respectively.

Irvrnooucrrox Any attempt to relate amphibole composition to a spe- cific P and ?" or to an exchangereaction in P-Z space The general form of the Tschermak cation exchange requires accurate thermodynamic data for certain end- can be expressedas r61M,l4lsi= I6lAl,t41Al,where M nor- member amphibole compositions.For calcic amphiboles, mally refers to a divalent cation such as Fe, Mg, or Ca. these data have generally come from phase-equilibrium This exchangeis of fundamental importance in many sil- studies (e.g.,Welch and Pawley, l99l1, L6ger and Ferry, icate solid solutions, including the pyroxenes, amphi- l99l; Jenkins, 1994).Thermodynamic data for a large boles, micas, chlorites, and melilites. In amphiboles, this number of rock-forming minerals, including amphiboles, coupled substitution is of primary importance in two sol- are tabulated in various thermodynamic data bases(Ro- id-solution series,the orthorhombic ferromagnesianse- bie et al., 1979;Helgeson et al., 1978;Berman, 1988; ries and the monoclinic calcic series.Recent transmission Holland and Powell, 1990). analytical electron microscope (TEM-AEM) studies have In this paper we present results from a calorimetric study shown that this substitution dominates during exsolution of synthetic amphiboles in the tremolite-tschermakite (Tr- reactions between anthophyllite and gedrite in the ferro- Ts) series [CarMgrSirOrr(OH)r-CarMgrAloSiuOrr(OH)r]. magnesianseries (Smelik and Veblen, 1993)and between The primary goals of the study are to (l) investigate the actinolite and hornblende in the calcic series(Smelik et energychange associated with the Mg-Tschermak exchange al., l99l). in calcic amphibole and (2) determine independently the 0003-004x/94/ | | r2-t I I 0$02.00 1110 SMELIK ET AL.: CALORIMETRIC STUDY OF SYNTHETIC AMPHIBOLES 1l1t

Ca2Mg3Al2[Al2Si6]q2(OH)2 a. Ts

f MgHb .144 AMPH I44 o 28-2 6 19-5 AMPH 28-2 o 25-l AMPH 19-5 o 29-l tr 23-10,23-5 AMPH 25-1

AMPH 29-1

TREM 23-IO,TREM 23-5 Tr tM4lCa= tM4lMg MC Tr IVIC Ca2Mg5[Sfu]Qz(oH)2. Mg7[Sis]Q2(oH)2 Fig. l. (a) Plot of tremolite-tschermakite-magnesiocummingtonite(Tr-Ts-Mc) showing starting compositions for amphibole synthesis,made at intervals of 10 molo/oalong a pseudobinaryjoin shifted toward the Mc apex by l0 mol9o (circles).The solubility limit for Al in tremolite is near the MgHb end-member.(b) Representativeelectron microprobe analysesof the syntheticamphiboles. The dashedline at 0.5 Ts representsthe operation of one full Mg-Tschermaksubstitution, i.e., the Al content of magnesiohornblende. heatsof formation for the amphibole end-membersmag- (Jenkins,1987; Graham et al., 1989;Pawley et al., 1993). nesiohornblende (MgHb), CarMgoAlrSi'Orr(OH)r, and Therefore, the amphiboles were synthesizedat l0 molo/o tschermakite, and compare theseresults with those from increments along the join Ca, , Mg, , Si, Or, (OH)r - phase-equilibrium studies. By incorporating thermo- Ca, rMgorAlrsirorr(OH)r. Electron microprobe (EMP) chemical data from Berman (1988) and Holland and analysis of the amphiboles (discussedbelow) and previ- Powell (1990) in our analysis,the results presentedhere ous work on similar compositionsby Jenkins(199a) in- should be compatible with those internally consistent data dicate that the compositional offset is preservedalong the sets. join. Approximately I wto/oqvarlz was mixed into each experiment to compensate for the loss of silica to the ambient fluid during hydrothermal treatment (Jenkins, Enpnnnrrcxur. METHoDS 1987). Syntheseswere performed by loading a stoichio- Amphibole synthesis metric oxide-carbonatemixture into a Pt capsulesealed All amphiboles were synthesizedfrom mixtures of re- at one end, heating it with a butane torch to drive of CO, agent grade CaCOr, MgO, and AlrOr; SiO, was prepared from CaCOr, and then sealingthe other end after adding from desiccatedsilicic acid. Previous experimental work 25-30 wto/oHrO. by Jenkins(1988) and Cho and Ernst (1991) has shown Synthesesbelow l0 kbar were performed in internally that the solubility limit of Al in tremolite approachesthe heated vesselsusing Ar as the pressure medium. Pres- magnesium hornblende end-member composition, sureswere measuredwith both a PPI bourdon-tube and Ca,MgAlrSi?Orr(OH)r, which is equivalent to one full Harwood manganin resistance-coilgauges, the latter be- Mg-Tschermak substitution into the tremolite formula. ing factory calibrated againsta dead-weightpiston gauge. For the synthesisof amphiboles for calorimetry, we pur- Pressuremeasurements are believedto be accurateto + 50 posely chose a join parallel to the true join of tremolite bars. Temperatures were measured with two chromel- and magnesium-hornblende,offset by l0 molo/omagne- alumel thermocouplescalibrated against the freezingpoint siocummingtonite (Mc), to obtain the purest amphibole of NaCl and are believed accurate to + I K. Syntheses yields possible. The magnesium-cummingtonitecompo- above I 0 kbar were carried out in a piston-cylinder press nent is reflected by a constant substitution of 0.2 Mg for t/z in. rn diameter using NaCl as the pressure medium. Ca in the M4 site acrossthe entire join (Fig. la). We chose The syntheseswere performed in a hot piston-out fashion to offsetthejoin on the basisoffindings ofprevious work- by pressurizingthe assemblageto about 2 kbar below the ers, who reported a 5-l0o/o enrichment of Mg in the M4 desired pressureand then using the thermal expansionof site in their attempts to synthesizeend-member tremolite the assemblageto reach the desired pressure. Pressure ttt2 SMELIK ET AL.: CALORIMETRIC STUDY OF SYNTHETIC AMPHIBOLES

TABLE1. Details of amphibole synthesis

Sample Composition r(K) P (kbar) r (h) Synthesis products'

TREM 23.10 Ca, 6Mg5rSisOr2(OH), 113q10) s.20(70) 239 amph, trace cpx, and trace opx TREM 23-5 Ca, sMgs,SisOr2(OH), 11 1 1(17) 6.10(10) 188 amph and trace cpx AMPH 29.1 Ca, sM95oAlo4SiTsOz(OH), 11 37(5) 6.35{5) 237 ampn AMPH 25-1 Cal sMg4sAlosSiT"Oz(OH), 11 48(1 0) 6.0q5) 109 amph AMPH 19-5 Cal sM946A112Si7oO"r(OH), 1134(5) 13.0(3) 120 amph, trace opx, and trace qtz AMPH 28-2 Ca, sMgliAll 6SiTrOz(OH), 11 15(s) 12.9(3) 147 ampn AMPH 14.4 Ca, sMg4rAl2oSiT oOz(OH), 1096(s) 12.0(5) 70 amph, cor (1.6 wtTo)*, and trace opx

/Vote:the error ranges for P-f conditions (shown in parentheses)take into account small P and ffluctuations during the course of the experiment. 'Amph : amphibole; cor : corundum; opx : orthopyroxene; cpx : clinopyroxene;qtz : quartz; trace : below detection limit on powder XRD (about 1.0 wto/"). ** Amount determinedby Rietveldmodal analysis.

transmission to the sample was calibrated by the reaction amphibole stoichiometry was achievedwhen the analysis albite : jadeite + qvartzat 873 K to a precisionof +250 was normalized to 23 O atoms. Representative EMP bars, and it was found that virtually no correction was analysesare plotted on the Tr-Ts-Mc triangle in Figure required to obtain the acceptedvalue of 16.25 kbar (Jo- lb for comparison with the end-member compositions hanneset al., l97l;, Holland, 1980).Temperatures were (Fig. la). Despite the scatter in the probe data, most of measured with chromel-alumel thermocouples newly the synthetic compositions cluster well along the offset made for each experiment and have an estimated uncer- Tr-Ts join. The most Al-rich one, AMPH l4-4, does not tainty of t 5 K. The synthesisconditions for the amphi- match the composition of the MgHb end-member. Op- boles are summarized in Table l. tical examination of this sample indicated the presence of unreacted corundum from the original oxide mix in Amphibole characterization the experiment products (Table l). The amount of co- Optical examination. All experimental products were rundum was quantifiable by Rietveld modal analysis, as examined with the petrographic microscope using an oil discussedbelow. of refractive index 1.60, which matches n, of the amphi- X-ray diffraction. All synthetic amphiboles were sub- bole, to identify the presenceofother phases.Only trace ject to Rietveld structure refinement. Step-scanswere amounts (

Trsle 2. Crystallographicdata for synthetictremolite-tscher- 9.85 makiteamphiboles 23-10 sampre a(A) b(A) c(A) B0 v(A') o{ , B 9.80 ,o rREM 23-5 9.807(2) 18.054(3) s.276(1) 104.56(1) 904.2(21 3-s 29_I TREM23-10 9.815(1) 18.0580) s.279(1) 104.58(1)905.49(9) O L4-4 AMPH 29-1 9.801(1) 18.042(21 s.279(1) 104.68(1)902.9(2) 25-l o AMPH 2s-1 9.784(1) 18.027(2) 5.280(1) 104.74(6) 900.6(2) ,99 ,"9 AMPH19-5 9.771(1) 18.013(2) s.284(1) 104.80(1)899.2(2) AMPH 28-2 9.767(1) 17.994(2) s.288{1) 1M.90(1) 898.0(2) AMPH 14-4 9.769(1) 17.995(2) 5.287(11104.91(1) 898.2(2) Note: the numbers in parenthesesindicate the enor in the last decimal 18.04 o place. o{ o 18.02 o 18.00 this study and we were surprisedto find that their refined oo unit-cell volumes differ by about 1.3 A'. We initially r7.98 thought that this might be due to different instrumental OO conditions for the data collections, but we rejected this hypothesison the basis of data for corundum refinements o (multiple scansat different conditions), which indicate no I systematic variation in cell parameters as a function of o o scanningcondition. However, there does appear to be as much as 0.lolo variation in the unit-cell volumes from 5.275 multiple scans,which could account for the 1.3-A' dif- ro4.9 oo ferencebetween TREM 23-5 and TREM 23-10. On the 104.8 o basis of that, the EMP results, and the TEM results, we o o do not feel that the higher cell volume for TREM 23-10 104.7 o indicates something fundamentally different about the 104.6 sample; it is simply on the high side of statistical variations for volume determinations. i04.5 All the unit-cell parametersvary linearly as a function of composition (Fig. 2). These trends suggesta complete o{ solid solution acrossthe join at the temperaturesof for- q) o mation (1093-1148 K). The linear decreasein cell vol- o ume with increasing Al content is consistent with other o oo studiesin the Tr-Ts system(Jenkins, 1988, 1994),and agreeswell with the trends observed for the Mg-Tscher- f,g mak substitution in the pyroxene system along the join 0.0 1.0 t.74 2.0 of diopside and Ca-Tschermak(Newton et al., 1977) and A{olcations Per 23 O in the mica system along the phlogopite-eastonitejoin Fig. 2. Unit-cell parametersfor synthetictremolite-tscher- (Circone least-squares et al., l99l). Extrapolation of a makiteamphiboles (with 10 molo/oMc) vs. total Al content.All regressionline through the volume data (R'z: 0.96) yields cellparameters show nearly linear trends with composition'sug- a cell volume for the hypothetical Ts end-member (with gestinga completesolid solution at the synthesisconditions. l0 molo/oMc) of 888.3 A', which agreeswell with the study of Jenkins(1994, 888 + 2 L). This cell volume should be regarded as a minimum volume for tschermakite, sincethere is no correction for the approximately Mareschet al. (1990),and Ahn et al. (1991) has shown l0 molo/oMg in the M4 site. varying degreesof defect concentration in synthetic am- Transmission electron microscopy.The synthetic amphiboles. The dominant defect type is the chain-multi- phiboles were also examined with a Philips CM2O-ST plicity fault (CMF), which occurs parallel to (010) of the transmission electron microscope(TEM) operatedat 200 amphibole structure.These planar faults can be either the keV. The samples were gently ground and dispersed in single-chain variety, which are essentiallyunit-cell slabs anhydrous alcohol. A drop ofeach suspensionwas then of pyroxene structure, or the multiple-chain variety (>2), applied to a 3.0-mm Ni grid with a holey carbon support which increasesthe proportion of the mica-like slab of film. the amphibole structure. Characteization of these faults The primary purpose of the TEM examination was to is important for two reasons.First, if the defect concen- determine the nature and concentration of defectsin the tration is high enough, it can significantly shift the am- amphiboles. Previous high-resolution TEM (HRTEM) phibole stoichiometry from the expectedvalue. Second, work by Mareschand Czank(1988), Graham et al. (1989), a highly defective structure may not be energeticallysim- ll14 SMELIK ET AL.: CALORIMETRIC STUDY OF SYNTHETIC AMPHIBOLES

Fig. 3. High-resolutionTEM (HRTEM) imageof Al-free sampleTREM 23-10taken down [1001. The (020)fringes are clearlyvisible, running from top to bottomin the image.Planar defectsare indicatedby arrows.These defects are single-chain faultsand representnarrow slabs ofdiopsideJike material. Tri- ple-chainfaults were also observed. Fig. 4. HRTEM imageof sampleAMPH 25-l with approx- imately0.8 total Al cationspfu. This image,also taken down a, showsfour individual triple-chainfaults embedded in an oth- ilar to one poor in defects, even if the composition is erwiseperfect amphibole structure. This microstructure,con- nearly identical. sistingof severalrandomly scattered CMFs in a givencrystal, Representative TEM images of the synthetic amphi- wasthe most commonly observed one in thesynthetic tremolite- boles are presentedin Figures3-5. Most of the images tschermakiteamphiboles. weretaken down a [u}wlzone, usually[100] or Il0l], so that the CMFs were parallel to the electron beam and bearing samples, but they did not approach the defect easily visible. All the synthetic amphiboles were found to concentration in synthetic end-member tremolite sam- contain CMFs. Figures 3, 4, and 5 are HRTEM images plesexamined by Ahn et al. (1991).Pawley et al. (1993) of Al-free TREM 23-10, AMPH 25-l (0.8 total Al), and also reported a relatively low CMF concentration in the AMPH l4-4 (1.74total Al), respectively.The most com- Mg-enriched synthetic tremolite (TrrrMcr) used in their monly observeddefect types for the amphiboles were sin- calorimetric study. These authors further indicated that gle-chain errors (pyroxene material) and triple-chains the increaseof richterite component virtually eliminated faults. The most common observation was severalplanar the CMFs from the amphibole structure. The presentre- defectsscattered randomly throughout a given crystal. No sults for tremolite-tschermakite amphiboles indicate that examplesof ordered polysomesof the biopyribole series the Mg-Tschermak substitution doesnot appearto inhib- (e.g., clinojimthompsonite or chesterite)were observed. it the formation of planar defects, as the richterite sub- For the vast majority of crystals, the estimated volume stitution does. percent of defective material was less than 5o/o.The TEM Mareschet al. (1990),in a HRTEM study of synthetic observations did indicate, however, that the tremolite tremolite, found the defect characterto vary with synthe- compositions(TREM 23-5, TREM 23-10)had a slightly sis conditions. For synthesisconditions comparable with higher frequency and abundanceof defects than the Al- ours, they describearmored diopside relics, up to l0 nm SMELIK ET AL.: CALORIMETRIC STUDY OF SYNTHETIC AMPHIBOLES 1115 wide, on which the amphibole appearsto nucleate.They also noted that CMFs are minor in both the diopside and tremolite. We have found no such diopside relics in our TEM examinations.Maresch et al. (1990) also studied synthetic anthophyllite-gedrite samples and found that there was a significantdecrease of CMFs with only a slight increaseof gedritecomponent into the anthophyllite. They were unable to determine whether the causeof this was due to the edenite substitution, the Tschermak substitution, or a combination of both. Even though our samples belong to the calcic, rather than the ferromagnesianse- ries, the presentTEM observationssuggest that the edenite substitution may be largely responsiblefor the elimi- nation of CMFs in synthetic orthorhombic amphiboles, rather than the Tschermak substitution. As mentioned above, one major concernis whether the AMPH l4-4 with amphibole stoichiometry is significantly shifted as a re- Fig. 5. Bright-fieldTEM imageof sample 1.74Al pfu. The centralpart of rhe crystal(indicated by sult of the presenceof defects.Simple calculationsfollow- about arrows)contains a considerablenumber of CMFs.This example ing the method of Veblen and Buseck (1979) show that representsthe mostextreme defect concentration observed, which per- even the most defective crystals in our study, with still representsless than 5-100/o of thetotal volume of thecrystal. haps 100/odefect concentration, do not causea significant Similar defectconcentrations were observed, though rarely, in compositional shift in the amphibole. Furthermore, since all the syntheticamphiboles, regardless ofthe Al content. in most crystals there was a combination of single-chain and triple-chain defects,any net compositional shift would be even smaller, since these defect types affect the stoichiometry in opposite directions. ature of the sample from room temperatureto the calori- Although the EMP results presented above (Fig. lb) meter temperature. This experiment differs from a reg- show good agreementwith the chosencompositional join, ular-solution experiment, in which the sample is the microprobe provides no information about compo- equilibrated at calorimeter temperature before dissolu- sitional heterogeneityon the submicron scale.Therefore, tion. The calibration constant, relating microvolts to we have examined severalof the samplesusing analytical joules, was determined by the Pt-drop method (Navrots- electron microscopy (AEM) to check the compositions of ky, 1977).Overall, the methodologyis essentiallyiden- nondefective and defective amphibole areas. For these tical to that used by Pawley et al. (1993) for tremolite- analyses,we useda probe sizeof about 10 nm. The anal- richterite amphiboles. yseswere collected with an EDAX Si(Li) detector with a Recentwork by Navrotsky et al. (1994)has addressed fixed take-offangle of 25'. We were unable to detect any the problem of the evolution of volatiles during calori- significant compositional differences between defective metric experiments on hydrous phasesand carbonates. and nondefectiveareas using AEM. This is due in part to They found that performing calorimetric experimentsun- the poor precision of AEM, compared with EMP tech- der a flowing gas atmosphereresulted in a negligible en- nique, and also to the fact that the small compositional thalpy of interaction between the evolved HrO or CO, differencesthat must exist in the defective areasare with- and the solvent,suggesting that very little, ifany, ofthe in a region <10 nm, which is the optimal resolutionof volatile remains dissolved in the solvent. Under flowing the AEM system. gas conditions, they were able to obtain reproducible results, with excellent base-line behavior, for several hy- Drop-solution calorimetry drous silicatesand carbonates.The presentdrop-solution The calorimetric technique used in this study is based experiments were performed in the same manner, under on the method describedby Navrotsky (1977), employ- an atmosphereof flowing Ar at 30-50 mllmin. ing a Calvet-type twin microcalorimeter. Two Pt cruci- For TREM 23-5,tlrresample was prepared for calorim- bles,each containing 30 g of 2PbO.BrO. solvent,are held etry in two ways. The first seriesof drops employed the at a constant temperature of 97 5 K in the calorimeter. use of small silica-glasscapsules, in which 6-12 mg of Relative changesin heat flow within the calorimeter,from amphibole powder was placed.This technique,described one sample chamber to the other, are measuredby two by Burnley et al. (1992), was developedto overcome dis- thermopiles, connected in opposition. The samples are solution problems that commonly occur with Mg-rich introduced into the solvent by dropping them from room compounds. Of course, using the capsulesintroduces the temperature(-298 K) through a long, silica-glasstube, need for an additional calibration constant, and ulti- hence the term drop solution. The heat effect is a com- mately increasesthe likelihood of additional errors in the bination of the heat of solution of the amphibole in the measurements. The results of these first experiments lead borate and the heat effect from raising the temper- showed significant scatter (a little over 2o/oof the total I116 SMELIK ET AL.: CALORIMETRIC STUDY OF SYNTHETIC AMPHIBOLES

TABLE3. Enthalpiesof drop solution for synthetic tremolite-tschermakiteamphiboles with 10 mol% Mc

SAMPIE TREM 23-5 TREM 23-10 AMPH 29.1 AMPH 25-1 AMPH 19-5 AMPH 28-2 AMPH 14.4 Al- 0.0 0.0 0.4 0.8 1.2 1.6 1.74 AH AS'd (kJ/moD 931.7 946.2 936.4 937.9 928.5 938.4 93it.9 937.2- 948.6 945.6 939.5 935.1 949.2 941.1 973.5' 952.3 948.2 967.4 936.6 950.2 946.7 975.5 952.6 953.4 967.6 949.7 953.8 949.0 977.O 956.8 953.7 968.5 949.8 957.1 950.2 986.4' 959.2 955.9 978.8 959.2 961.3 951.4 966.9 967.2 970.4 967.1 952.8 977.9 969.5 954.3 Mean 963.5 954.6 951.5 959.9 950.9 955.8 947.4 Std Err" 18.8 5.3 7.2 13.9 12.4 7.2 4.8 . These values for TREM 23-5 were obtained using SiO, glass capsules; the remainingTREM 23-5 values and all other values were obtained using oellets. " The standard error equals 2ol\aN.

heat effect) and showed some evidence of slow dissolu- vrotsky, 1992), but the nonlinear trend of the mica data tion; therefore the capsuletechnique was abandoned. extrapolate to a slightly higher Af/o.., *, for end-member A second batch of experiments for TREM 23-5, and eastonite. all the remaining calorimetric experiments, were per- pressed pellets formed using of the amphiboles, which Drscussrox have been shown to dissolvewell in molten 2PbO.BrO3 (Pawleyet al., 1993).The pelletswere about 3 mm in Mfu,oo-, for MgHb and Ts diameter and weighed from l0 to 20 mg. Each sample One of the primary goalsof this study is to obtain, from was carefully weighed before it was dropped into the cal- calorimetric data, values for the enthalpies of formation orimeter. The calorimetric peak normally returnedto base from the elementsfor the amphibole end-membersmag- line in 30-50 minutes, indicating complete, and rapid, nesiohornblendeand tschermakite at 298.15 K. In order dissolution. to extract this information we first need values for Afl*.o *, for thesecompositions. The mean valuesfor AFld.oo*, from Rrsur,rs Table 3 are plotted against the Al content of the amphi- The enthalpiesof drop-solution for synthetic tremolite- bole in Figure 6. Unlike Graham and Navrotsky (1986) tschermakite amphiboles (in kJlmol) are given in Table and Pawley et al. (1993),we had only compositionsfor 3. Six to eight enthalpy measurementswere made for the Tr-rich half of the full Tr-Ts join, from Tr to MgHb, each composition. For sample TREM 23-5, the mean becausemore Al-rich compositions could not be synthe- value for the enthalpy of drop solution, Ma.op *r, includes sized. The lack of calorimetric data for Xr" > 0.5 and the experiments performed with capsulesand with pellets. almost constant Ma,op"orfor 0 < Xr" < 0.5 preclude a The uncertainty for this sample,given as 2 sd of the mean, definite statement about enthalpies of mixing along the is rather high, about 2o/oof the total heat effect,and may Tr-Ts join. The presentdata do not justify anything more contain errors due to problems with the silica-glasscap- than a least-squaresregression through the data (Fig. 6). sules.We unfortunately exhaustedthe material before we Becauseofthe large and variable uncertainties,the lin- could complete a full seriesof pellet drops for this sam- ear regressionwas perfbrrned by weighting each A110."o.", ple. by the inverse ofits uncertainty squared(e.g., Bevington For the other tremolite sample and the remaining am- and Robinson, 1992, p. 106).The equation for the line phibole compositions, the uncertainties are in the 0.5- is Af/o,oo,o,: 955.0 (+4.4) - 3.0 (+3.6;41,",.The straight l.5olorange, expressed as 5-13 kJ/mol, which agreewell line fit is rather poor, with a low R value (0.563). Because with the magrritude of uncertainties encountered in the the energy change associated with the Mg-Tschermak Pt-calibration experiments. The magnitude of the heat substitution is quite small, we feel the least-squaresre- effect is in the 950 kJlmol range for these amphiboles, gression line is the best approach for extracting A/1d.op*r and it becomes immediately obvious by examining the values for the end-member MgHb and for a tentative mean values of A-Flo.oo*, as a function of Al content that extrapolation to Ts. The question still remains, however, there appears to be only a small energeticdifference of whether the measured enthalpy values for the composi- about -10 kJlmol or less betweenthe Al-free and Al- tions with l0 molo/oMc can be used to representtrue Tr- rich compositions (Table 3). This small reduction in the Ts amphiboles. enthalpy ofsolution is consistent with the effect that the To check the possible effect of the l0 molo/oMc com- Tschermak substitution has on the enthalpy of solution ponent, a series of drop-solution experiments was per- of akermanite(Charlu et al., l98l) and diopside(Newton formed on a sample of very pure natural tremolite, kindly et al., 1977). It also agreeswith the initial effect of the provided by Colin Graham. This tremolite was used by Tschermak substitution in phlogopite (Circone and Na- Graham et al. (1984) in a study of H isotopic exchange SMELIKET AL.:CALORIMETRIC STUDY OF SYNTHETIC AMPHIBOLES lltT betweenamphibole and HrO. It was also usedin a phase- = - 980 AHa,op"or955.0 3.0(Ald R=0.563 equilibrium study of the tremolite breakdown reaction tremolite:2diopside + 1.5enstatite + Pquartz+ HrO szo by Welch and Pawley(1991). These authors reported an ! o averagecomposition for this tremolite of Ca,nrFeoor- €1soo MgrouAloo,SirerO22(OH)r,on the basisof 3l EMP anal- E yses.This tremolite wasalso used by Pawleyet al. (1993) F sso in a calorimetric study of synthetic amphiboles along the \{ join. tremolite-richterite The mean value of Aflo.op,orfor R noo this sample was 959.6 + ll.2 kJlmol (8 drops), which compareswell with the results of Pawley (1993, et al. 930 958.0 kJ/mol). These enthalpy values are in general 0 0.4 0.8 r.2 1.6 2 agreementwith our measuredvalues for the Mc-substi- tuted tremolite shown in Table 3. On the basis of this Al.o, result, we conclude that the small amount of Mg in the Fig. 6. Plot of AFld,op*rvs. A1,..for synthetictremolite- M4 site has a minimal effecton the enthalpiesof solution, tschermakiteamphiboles. A weightedleast-squares regression and therefore we may calculate values of A11o,oo for "", line is fittedthrough the drop-solutiondata. The energychange MgHb and Ts using the equation above. The results are associatedwith the Mg-Tschermakexchange, reflected by the 949.0 t 8.5 and 943.0 + l5.l kJlmol, respectively. heatofdrop-solution data, is very smalland negative, suggested by the slightnegative slope of the line. The MgHb end-member A.EIPfor MgHb and Ts is indicatedby the solidsquare. Two pointsfor synthetictrem- : The enthalpies of formation for MgHb and Ts may be olite (A1,., 0) are shown,TREM 23-5 andTREM 23-10,the latterhawing error. calculated using a combination of the present calorimet- the smaller ric data and data from the literature. The thermodynamic cycle is based on the following equations for MgHb and for A-F/o,oo.o,show close agreement,independent of the Ts, respectively: heat-capacity expression chosen. The calculated values for quartz are complicated by the enthalpy of the a-0 2CaMgSirOu+ MgO + AlrO3 + 3SiO, + Mg(OH), transition at 844 K. A summary of published values for : CarMgAlAlSi?Orr(OH), (l) AII*', rangingfrom 0.335 to l.2l kJ/mol, hasbeen given and by Ghiorso et al. (1979). Recently,Hemingway (1987) provided new heat-capacitydata for quartz and reevalu- 2CaMgSirOu+ 2Al2O3+ 2SiO, + Mg(OH), ated the literature in an attempt to resolve the published : CarMg.AlrAlrSi6Orr(OH)r. (2) disparitiesin heat-capacitydata in the 425-1025 K range. The data required for the calculations include enthal- He concluded that all recent tabulations of thermody- pies of formation from the elements and enthalpies of namic data for quartz are low by about l-2o/o. In light of drop solution in lead borate at -97 3 K for diopside, peri- these results, we have included in Table 4 calculations clase,corundum, quartz, and brucite. For this paper we using the heat-capacity expression for quartz given by have chosen heat of formation values from two of the Hemingway (1987) and have incorporated his value for published internally consistentthermodynamic data sets, AH..p(0.625 kJ/mol) in all the Hr-, - -Flrn,', calculations. thoseof Holland and Powell (1990)and Berman (1988). Becausethe variations in AlIfL; *' values are small, the The heat of solution data for the reactant minerals errors resulting from differencesin values from Table 4 available from the literature are summarized in Table 4. are small, compared with the elTor in the heats of drop For quartz and diopside, both heats ofdrop solution and solution measuredfor the synthetic amphiboles.We have heats of solution in 2PbO.BrO, have been reported, chosen the following values as the preferred values of whereasfor periclaseand corundum only heats of solu- A/Id.oo for the calculation of amphibole heats of for- "or tion are available, and for brucite only drop-solution ex- mation: diopside and quartz, the valuesmeasured by Chai periments have been done. To calculate a heat of drop and Navrotsky(1993), 40.02 + 0.22and231.4 ! l.l kJ/ solution from solution data, it is only necessaryto add mol, respectively;periclase and corundum, the calculated the value of heat content, Hr*, - Hrnr,r, for the mineral values for Ma.op.orusing the solution data of Navrotsky of interest. The value for Hr*, - Hr'",, (Z*, : calorimeter et al. (1994, 1986)combined with the heat-capacityex- temperature)varies dependingupon which heat-capacity pressionsof Holland and Powell(1990), 36.51 and 107.63 expressionis chosen.To assessthe extent of variability, kJ/mol, respectively;brucite, the measuredvalue of Na- we have calculated heats ofdrop solution for all five re- vrotsky et al. (1994), 144.1 + 3.6 kJlmol, which was actant minerals in Equations I and 2, using different pub- obtained using the methodology employed for the syn- lished values for heats of solution and several heat-ca- thetic amphiboles in this study. pacity expressions(Table 4), and have compared these Table 5 shows the calculated values of A.F{ for MgHb with available measuredvalues for AFld.oo*r. and Ts, using the preferred thermodynamic data for the For all the phasesexcept quartz, the calculated values reactant minerals from Table 4 and the AH! data for the 1l18 SMELIK ET AL.: CALORIMETRIC STUDY OF SYNTHETIC AMPHIBOLES

TABLE4, Calorimetricdata for silicates near 973 K from the literature

Hrut - Hn"' aF,5.3- aF,s*

Diopside 149.18 149.97 150.14 87.65+ 1.05 236.83 237.62 237.79 85.90+ 1.05 235.08 235.87 236.04 85.86+ 1.55 235.04 236.00 23583 85.4'l + 1.47 234,59 235.38 235.55

Periclase 31.71 31.68 31.71 4.94 + 0.33 36.65 36.62 36.6s 4.81 + 0.58 36.52 36.49 36.52 4.87 + 0.62 36.58 36.55 36.58 4.80+ 1.00 36.51 36.48 36.51 Corundum 74.84 74.91 74.87 32.34 + 0.33 107.1I 107.25 107.21 32.55 + 0.46 107.39 107.46 107.42 32.76 + 0.33 107.60 107.67 107.63 43.48 42.33 42.36 45.13 -5.15 + 0.29 38.33 37.18 37.21 -3.18 + 0.42 40.30 39.15 39.18 -4.23 + 0.21 39.25 38.10 38.13 -3.51 + 0.18 39.97 38.82 38.85

Brucite 69.62 70 31 70.33

/Voteiboldfaced values represent high and low values of AFfiSd for each mineral.ltalicized values are the preferredvalues used in our calculations. Unitsof measureare kJ/mol. -Heat-capacityexpressionsused:R:Robieetal.(1979);B:Berman(1988);H-P:HollandandPowell(1990);H:Hemingway(1987,forquartz only). '* Calorimetricexoeriments done at 986 K. t Drop solution under static air conditions near 973 K. + Drop solution under static air conditions neat 977 K. $ Drop solutionunder flowing Ar gas near975 K.

reactantminerals from Holland and Powell (1990) and no previous studies, either phase-equilibrium or calori- Berman(1988). metric, aimed at determining thermodynamic properties For magnesiohornblende,the calculated enthalpy of for the magnesiohornblendecomposition. Holland and formation at 298.15 K is - 12401.2kJlmol. For tscher- Powell (1990),however, have derived a value of - 12420.3 makite, the calculatedenthalpy of formation is - 12527.7 + 12.7 kJlmol for AfI! of MgHb, for inclusion in their kJlmol. If the enthalpies for these amphiboles are calcu- internally consistentthermodynamic data base(their Ta- lated using the minimum and maximum thermodynamic ble 7, reliability level 2). Their value agreesclosely with values shown in Table 4, a range of about 25 kJ/mol the value we calculate using the older quartz heat-capac- (-12390.9 to -12419.5 and - 12519.3to -12543.5 kJ/ ity data but is somewhat more negative than our pre- mol, for MgHb and Ts, respectively) is obtained. This ferred Af19.This diference probably reflectsHolland and range can be largely attributed to variations in the quartz Powell's(1990) use ofthe older heat-capacityexpressions and diopside calorimetric data. The differencesthat result for quartz in their least-squaresfitting process. using the Berman (1988) AH! data are insignificant, shift- The enthalpy of formation for the end-member tscher- ing the values by a maximum of about I kJ/mol. It is makite composition has been derived in several phase- difficult to determine accurately the error range associ- equilibrium studies, and the published values are com- ated with these calculations becausesome studies give pared with our preferred value in Table 6. Our results uncertainties as I or 2 sd, whereasothers give uncertain- agreewell with the resultsofthe recentphase-equilibrium ties as the standard deviation of the mean, and some study ofJenkins (1994)and the study ofl,6ger and Ferry studies give no errors at all. By taking the squareroot of (1991), which combinesexperimental data with natural the sum of the squaresof all the reported errors for all mineral parageneses.The At{ reported by Chermak and the data used in the calculations (using the Holland and Rimstidt (1989), obtained by summing the estimated Powell AH? data), we obtain uncertainties of + 10.6 and polyhedral contributions for each oxide and hydroxide + 16.4 kJlmol for MgHb and Ts, respectively. component, is significantly more negative than our preferred value, which corroborates that these authors ad- mitted difficulties at predicting thermodynamic proper- Comparisonwith values obtained from ties for calcium aluminum amphiboles and pyroxenes phase equilibrium using the polyhedral approach. The value of Lieberman Our preferred values for the enthalpies of formation and Kamber (1991),derived by least-squaresfitting to a for MgHb and Ts are compared with values from the combination of natural paragenesesand experimental literature in Table 6. As far as we know. there have been data, is considerably more positive than our value. SMELIK ET AL.: CALORIMETRIC STUDY OF SYNTHETIC AMPHIBOLES I119

Taate 4.-Continued TABLE5. Calculated enthalpies of formation for MgHb and Ts at 1 bar and 298.15 K usingcalorimetric data

H Reference Preferredvalue' Amphibole Data base (kJ/mol) Navrotskyand Coons(1976)". MgHb H.P -'t240't.2 Newton et al. (1977) -12402-3 Kiselevaet al (1979) B Ts H-P -12527.7 Navrotskyet al. (1980).. -'t2528.7 237.40! 1.10 Chai and Navrotsky(1993) B (1975) Charluet al. Nofei H-P : Hollandand Powell(1990); B: Berman(1988). Navrotskyand Coons(1976)-' "Calculated using preferred values of AHEBd and AHm'd from Daviesand Navrotsky(1981)-- Table 4. Navrotskyet al. (1994) Charluet al. (1975) Newtonet al (1980) Navrotskyet al. (1986) 39.98 Charluet al. (1975) (298.15 41.95 Navrotskyand Coons(1976)-. reaction at To and Po K, I bar), AS. is the total 40.90 Kiselevaet al. (1979) entropy change of the reaction, and AV, and K" are the 41.62 Akaogiand Navrotsky(1984) volume changeand equilibrium constant for the reaction, 40.02+ 0.22 Chaiand Navrotsky(1993) 111.50+ 1.59 Kiselevaand Ogorodova (1984)t respectively. 120.04+ 2.22 Clemenset al. (1987)t BecauseReactions 3 and 4 involve only solids and con- + 117.76 1.54 Circoneand Navrotsky(1992)+ serve HrO, and since no Co data are available for MgHb, 144.10+ 3.60 Navrotskyet al. (1994)$ we will assumeas a first approximation that Aq : 0 for both reactions and that AZ, (solids) is constant over the P-T range(l-8 kbar and 873-1073 K) of Jenkins'sre- Site-rnixing models and entropy calculations versal experiments.These simplifications allow Equation 5 to be rewritten as Jenkins (1994) examined experimentally two divariant reactions involving Al substitution in tremolitic amphi- AG, : 0 : AGg - A^SKr - T) + LV,(P - Po) bole to derive activity-composition relationships for am- + RZln K". (6) phiboles in the Tr-Ts pseudobinary system, as well as Noting that AGp : AH? - Z.ASP and substituting into standard-statethermodynamic properties for end-mem- Expression6, we arrive at ber tschermakite. To complement Jenkins's study, we have used our AIf data, determined independently from AG,:0 : LH? - rA,S9+ LV,(P - &) + Rrh K". (7) calorimetry, combined with the experimentalreversal data For Reactions 3 and 4, the entropy term in Equation 6, of Jenkins (1994), to derive the standard entropy for AS!, can be expandedas shown in Equation 8: MgHb and test various site-mixing models for aluminum tremolites. ASf : AS!,n.* - Sfl,r", (8) We have reformulated the divariant Reactions 2 and 3 where AS!*.." is Sg, + ,Sl" - SBt - ,S&,,for Reaction 3 studied by Jenkins (1994) in terms of the MgHb com- and Sg. + ,S8"- S$. for Reaction 4. Substituting this into ponent, rather than the Ts component, to obtain the two Equation 7 and rearranginggives the following expression reactions for the entropy of MgHb at 298.15K and I bar: Ca,MgAlrSi,Orr(OH), + CaMgSirO6+ SiO, Sflnru amphibole (MgHb) diopside quartz - - : + (3) _ At19 IAS8*"." + A4(P %) + Rin ln K, CarMgrSirOrr(OH), CaAlrSirO, _ \at/o\ mphibol€ (Tr) plagidare I'

CarMgoAlrSirOrr(OH), * MgrSiOn Using Equation 9, we can test various site-mixing models amphibole (MgHb) fonterite for MgHb to see which, if any, give constant values of Sflorro.The equilibrium constant, K,, for Reaction 3 can : CarMgrSirOrr(OH),+ MgAlrO4. @) be expandedas amphibole (Tr) spinel (afli'n)(aol;") K"-- (10) Following the methodologyof Jenkins(199a), at equilib- (a.i&tiiXariXag%,) rium we may write the following expressionfor the Gibbs free energy ofreaction at P and T: and for Reaction4 as AG,: : - (l l) o ^G? l" ^t. o, * J'"^vdP plagioclase, + R?"ln K. (5) Electron microprobe analyses of diopside, quartz, forsterite, and spinel by Jenkins (1994) indicated whereAG! is the changein the Gibbs freeenergy of the that these phasesare essentiallypure. By accepting unit l 120 SMELIKET AL.: CALORIMETRIC STUDY OF SYNTHETIC AMPHIBOLES

TABLE6. Comparison with AHP and I values from the literature for MgHb and Ts

AHto s0 kJ/mol J/(mol.K) References(methods)' MgHb ',t0.6 -12401.2 + 575.3 + 3.9 Thisstudy'- (C) -12420.29 + 12.70 551.0 Hollandand Powell 0990) (LS) Ts -12527.7 + '16.4 Thisstudy'- (C) -12534.4 + 13.0 542.5 + 12.7 Jenkins(1 994)'- (PE) - 12578.949 514.687 Maderand Berman (1992) (PE) -12449.O94 542.72 Liebermanand Kamber(1991)t (LS, PE) -12606.0 + 36.0 Chermakand Rimstidt(1989)+ [rH) -12534.67 538.6 L6gerand Ferry(1991) (PE) - Methods: C : calorimetry; LS : least squares; PE : phase equilibrium;TH : theoretical. '- Using Holland and Powell (1990) data base; entropy is average of two-site and four-site activity-compositionmodels. f Using Berman (1988)data base. + Calculatedbased on summation of polyhedralcontributions wilh all OH as Mg(OH),.

activity for the pure solids at P and Z of interest as the The strategy here is to solve Equation 9 using the ex- standard state, Equations l0 and I I can be expressedin perimental reversaldata for Reactions 3 and 4 of Jenkins terms of amphibole activities only: (1994) to seewhich of the four site-mixing activity-composition models gives the most constant value for S$*0. ,. aTo" I\" : ------(r2) We have usedthe value for the AHPfor MgHb from Table s MsHb 6 and enthalpy of formation values for the other phases Four mixing models will be consideredhere: (l) coupled from Holland and Powell (1990), except for tremolite, substitution of octahedral and tetrahedral Al onto only which came from Jenkins et al. (1991), to calculate Alf. one M2 site (one-sitecoupled model), (2) random mixing We used the entropies for diopside, tremolite, anorthite, of Al on both M2 sites but not on the tetrahedral sites qlrarIz, forsterite, and spinel from Holland and Powell (two-site coupled model), (3) random mixing of Al and (1990) to calculate AS3**. Using the cell volume for Si on the four Tl sites but not the M2 sites (four-site AMPH 28-2 (Table 2) and the planar extrapolation meth- coupled model), and (4) complete random mixing of Al od described by Jenkins (1994), we obtain a volume of on the two M2 sites and Al and Si on the four Tl sites 270.8+ 0.1 cm3/mol(899.16 + 0.33A,; for MgHb. Us- (random mixing model). These models represent ideal- ing the molar volumes from the Holland and Powell ized site-mixing configurations that serve as a starting (1990) data set for all other phases,we calculate for LV, point for the following calculations and are not meant to a value of + 13.8 cm3 (+ 1.38 J/bar) for Reaction 3 and - imply long-range ordering arrangementsthat would be 1.98 cm3(-0. 198 J/bar) for Reaction4. inconsistent with amphibole crystallography or crystal The results for the four site-mixing models are sum- chemistry. Ideal activity-composition models for these marized in Table 7. The uncertainties reflect the propa- mixing schemeshave been derived according to the pro- gation of all errors entering into the calculation of Sfln*6 cedureof Price (1985).The resultantexpressions for rK" by Equation 9. The mean values representthe weighted for the four mixing models are given below, with X de- averageSflrro t I sd of the mean. Although all activity- fined as X{,'z(mole fraction of Al on the M2 site): composition modelsexamined provide a satisfactoryrange ofconstant entropy values,the two-site coupled and four- (l) one-sitecoupled model: site coupled mixing models produce the least amount of (t - 2x) scatter (smallest standard deviations). These two mixing n^:,, 2x models also yield nearly identical mean values for Sfloo*. The averagevalue of Sflrrrofrom thesetwo mixing models (2) two-sitecoupled model: is listed in Table 6. (t - Y1z n": Coucr,usroxs 4x.t _ x\ These results represent the first data on heat of for- (3) four-sitecoupled model: mation from calorimetric measurementsfor calcic amphiboles approximately along the join betweentremolite (t - %), n": ,t"tr-t

TABLE7. Calculatedvalues of S$n"ovalues using different activity-compositionmodels PTX datapoint' One-sitecoupled model Two-site coupled model Four-sitecoupled model Random mixing model Reaction 3 '| 571.44+ 14.91 576.68+ 14.90 577.U + 14.90 565.92+ 15.19 2 570.55+ 13.65 575.68+ 13.62 576.79+ 13.60 566.11+ 14.16 3 570.97+ 12.34 575.91+ 12.30 576.95+ 12.29 568.48+ 12.73 4 572.56+ 13.85 577.21+ 13.82 578.13+ 13.81 572.13 + 14.07 5 572.02 + 12-56 576.52+ 12.53 577.37 + 12.52 572.39 + 12.82 o 573.38+ 14.12 577.44+ 14.11 578.11+ 14.11 575.59+ 14.17 7 572.02+ 13.13 575.97+ 13.00 576.61+ 12.96 574.02+ 13.58 Reaction 4 I 575.37+ 12.18 574.48 + 12.05 573.89+ 12.04 580.35+ 12.13 9 570.62+ 12.61 572.47 + 12.60 572.46 + 12.59 575.81+ 12.63 10 571.53+ 12.33 S72.Ag+ 12.90 572.68 + 12.29 576.85+ 12.35 11 570.25 + 12.78 572.32 + 12.75 572.37 + 12.74 575.29 + 12.82 Mean values 571.89+ 3.94 575.09+ 3.92 575.57+ 3.92 573.38 + 3.99 Nofe.'units of measurefor models are J/(mol.K). . PTX data points are reversal experimentsof Jenkins (1994).

linear regressionthrough the calorimetric data suggestsa imental petrology laboratory in the Department of Geological Sciencesat ofNew York at Binghamton, supported by NSF grant decreasein A.F155;*,with increasing Al content. This the State University slight EAR-9104266 to D.M.J. The TEM and EMP characterizationwas done trend is also observed in other silicate systemsin which at the electron microbeam facility of the Princeton Materials Institute, the same substitution is operating, such as initially along and the calorimetry was performed in the calorimetry laboratory in the the phlogopite-eastonitejoin (Circone and Navrotsky, Department of Geological and Geophysical Sciences, both at Princeton grant DE-FG02- 1992)and thejoin of diopsideand Ca-Tschermakpyrox- University. This research was supported by DOE 85ER13437to A.N. ene (Newton et al., 1977). Becauseof the limited range available for this study (only the Al-poor of compositions Rmnnrxcns crrED half of the join), the data for the Al-rich end-member tschermakite were determined by linear extrapolation. Ahn, J.H., Cho, M., Jenkins, D.M., and Buseck, P.R. (1991) Structural American Mineralogist, 76, The thermodynamic data calculated from the mea- defectsin synthetic tremolitic amphiboles. l8l l-1823. surementsof the heat of drop solution and other calori- Akaogi, M., and Navrotsky, A. (l 984) The quartz-coesite-stishovitetrans- metric data from the literature agreewell with previous formations: New calorimetric measurementsand calculation of phase values derived from phase-equilibrium studies.Our final diagrams.Physics ofthe Earth and PlanetaryInteriors, 36,124-134. preferred values for A/{ for MgHb and Ts are -12401.2 Berman, R.G. (1988) Internally-consistentthermodynamic data for rnin- in Na'O-KrO-CaO-MgO-FeO-Fe'O.-AlrOr-SiOr-TiO'- + -12527.7 + 16.4kJ/mol, respectively.Four erals the system 10.6 and H,O-CO'. Journal of Petrology, 29, 445-522. ideal activity-composition relationshipswere testedusing Bevington, P.R., and Robinson, D.K. (1992) Data reduction and error the calorimetric data. Although all four models gave sat- analysisfor the physical sciences(2nd edition), 328 p. McGraw-Hill, isfactory results, the two-site and four-site coupled mod- New York. (1988) phase using the Rietveld method. els gavethe leastscattered values for Shs'b,575.1 + 3.9 Bish, D.L. Quantitative analysis Journal ofApplied Crystallography,21, 86-91. + Thesevalues signifi- and 575.6 3.9, respectively. are Bish, D.L., and Post,J.E. (1993) Quantitative mineralogicalanalysis using cantly higher than the one published value in Holland the Rietveld full-pattern fitting method. American Mineralogist, 78, and Powell's(1990) data base. 932-940. Although these results are insufficient for the precise Burnley, P., Navrotsky, A., and Bose, K. (1992) Heat of formation of natural talc by drop solution calorimetry: A test ofa new phase calcic syntheticand calculation of equilibria involving aluminous technique (abs.).Eos, 73, 522. amphiboles, they representanother step in the persistent Chai, L., and Navrotsky, A. (1993) Thermochemistry ofcarbonate-py- effort to understand the activity-composition relation- roxene equilibria. Contributions to Mineralogy and Petrology, 114, 139- ships in complex amphibole solid solutions. It is hoped 147. (1975) Enthalpies of for- that the results from this paper can be incorporated into Charlu, T.V., Newton, R.C., and Kleppa, O.J. mation at 970 K of compounds in the system MgO-AlrOr-SiO' from the ever-evolving internally consistent thermodynamic high temperature solution calorimetry. Geochimica et Cosmochimica data setsand provide some critical data neededfor ther- Acta.39. 1487-1497. modynamic modeling of mineral reactions involving -(1981) Thermochemistry of synthetic CarAlrSiO, (gehlenite)- tschermakitic calcic amphiboles. Ca,MgSi,O, (ikermanite) melilites. Geochimica et Cosmochimica Acta, 45. 1609-1617. (1989) the thermodynamic AcxNowr,nncMENTs Chermak, J.A., and Rimstidt, J.D. Estimating properties (AG?and Alf) ofsilicate minerals at 298 K from the sum E.A.S. would like to thank Ed Vicenzi, for help with the microprobe of polyhedral contributions. American Mineralogist, 74, lO23- lO3l. analyses,and Vincent lamberti and PeterHeaney, for helpful discussions. Cho, M., and Ernst, W.G. (1991) An experimentaldetermination of calcic Constructive reviews by D. Perkins and H.T. Haselton,Jr. improved the amphibole solid solution along the join rremolite-tschermakite.Amer- manuscript. The arnphibole synthesiswork was carried out at the exper- ican Mineralogist,76, 985-1001. tr22 SMELIK ET AL.: CALORIMETRIC STUDY OF SYNTHETIC AMPHIBOLES

Circone, S., and Navrotsky, A. (1992) Substitution oft6lAl in phlogopite: Lieberman, J.E.,and Kamber, B. (1991)Consistent thermodynamic prop- High-temperature solution calorimetry, heat capacities,and thermo- erties for tschermakite from experimental and empirical constraints dynamic properties of the phlogopite-eastonitejoin. American Miner- (abs.).Eos, 72, 564. aloglst, 77, I I 9 l-l 205. Miider, U.K., and Berman, R.G. (1992) Arnphibole thermobarometry:A Circone, S., Navrotsky, A., Kirkpatrick, R.J., and Graham, C.M. (1991) thermodynamic approach.Current research:Part E. GeologicalSurvey Substitution of t6alAl in phlogopite: Mica characterization, unit-cell of Canada,Paper 92-lE, 393-400. variation, ,?Al and -Si MAS-NMR spectroscopy,and Al-Si distribu- Maresch,W.V, and Czank, M. (1988) Crystal chemistry, growth kinetics tion in the tetrahedral sheet.American Mineralogist, 76, 1485-1501. and phaserelationships of structurally disordered(Mn'+, Mg) amphi- Clernens,J.D., Circone, S., Navrotsky, A., and McMillan, P.F. (1987) boles. Fonschritte der Mineralogie, 66, 69-121. Phlogopite: High temperature solution calorimetry, thermodynamic Maresch, W.V., Czank, M., and Seifert, F. (1990) Realbau (defect struc- properties,Al-Si and stackingdisorder, and phaseequilibria. Geochim- ture) in synthetic tremolite and anthophyllite-gedrite solid solutions. ica et Cosmochimica Acta, 51,2569-2578. Terra Abstracts. 2. 24. Davies, P.K., and Navrotsky, A. (1981) Thermodynamics of solid solu- Navrotsky, A,. (1977) Progressand new directions in high temperature tion formation in NiO-MgO and NiO-ZnO. Joumal of Solid State calorimetry Physicsand Chemistry of Minerals, 2, 89-104. Chemistry,38,264-276. Navrotsky, A., and Coons, W.E. (1976) Thermochemistry of some py- Ghiorso, M.S., Carmichael,I.S.E., and Moret, L.K. (1979) Inverted high- roxenesand related compounds. Geochimica et Cosmochimica Acta, temperaturequartz: Unit cell parametersand propertiesofthe d-p in- 40, l28l-1288. version. Contributions to Mineralogy and Petrology, 68,307-323. Navrotsky, A., Hon, R., Weill, D.F., and Henry, D.J. (1980) Tbermo- Graham, C.M., and Navrotsky, A. (1986) Thermochemistry of the trem- chernistryof glassesand liquids in the systemsCaMgSi,06-CaAlrsirOs- olite-edenite using fluorine analogues,and applications to amphibole- NaAlSi.O., SiO,-CaAlrSi,O.-NaAlSiOr, and SiO,-AlrO,-CaO-Na,O. plagioclase-quartzequilibria. Contributions to Mineralogy and Petrol- Geochimica et Cosmochimica Act^. 44. | 409- | 423. ogy,93, l8-32. Nawotsky, A., Wechsler,B.A,. Geisinger,K., and Seifert,F. (1986) Ther- Graham, C.M., Harmon, R.S., and Sheppard,S.M.F. (1984) Experimen- mochemistry of MgAl,Oo-AlqOodefect spinels.Journal of the Ameri- tal hydrogen isotope studies: Hydrogen isotop€ exchange between am- can Ceramic Society, 69, 418-422. phibole and water. American Mineralogist, 69, 128-138 Nawotsky, A., Rapp, R.P., Smelik, E.A., Burnley, P., Circone, S., Chai, Graham, C.M., Maresch, W.V., Welch, M.D., and Pawley, A.R. (1989) L., Bose,K., and Westrich, H.R. (1994) The behavior of HrO and COt Experimental studies on amphiboles: A review with thermodynamic in high-temperature lead borate solution calorimetry of volatile-bearing perspectives.European Journal of Mineralogy, l, 53 5-5 55. phases.American Mineralogist, 79, 1099-1 109. Helgeson,H.C , Delany, J.M., Nesbitt, H.W., and Bird, D K. (1978) Sum- Nemon, R.C, Charlu, T.V., and Kleppa, O.J. (1977) Thermochemistry mary and critique of the thermodynamic properties of rock-forming of high pressure garnets and clinopyroxenes in the system CaO-MgO- minerals. American Journal of Science.278A. l-229. Al,O3-SiOr. Geochimica et Cosmochimica Acl^, 41, 369-377. Hemingway, B.S. (1987) Quartz: }{eat capacitiesfrom 340 to 1000 K and -(1980) Thermochemistry of the high structural state plagioclases. revised values for the thermodynamic properties. American Mineral- Geochimica et Cosmochimica Acta, 44, 933-941. ogist,72,273-279- Pawley, A.R., Graham, C.M., and Navrotsky, A. (1993) Tremolite-rich- Holland, T.J.B. (l 980) The reaction albite : jadeite + quartz derermined terite arnphiboles: Synthesis,compositional and structural character- experimentally in the range 600-1200 qC.American Mineralogist, 65, ization, and thermochemistry.American Mineralogist, 78, 23-35. t29-r34. Pouchou,J.L., and Pichoir, F. (198 S)'PAP' g(pZ) procedurefor improved Holland, T.J.B., and Powell, R. (1990) An enlargedand updated inter- quantitative microanalysis.Microbeam Analysis, 1985, 104-106. nally consistent thermodynamic dataset with uncertainties and corre- Price, J.G. (1985) Ideal site mixing in solid solutions,with an application lations: The system KrO-Na,O-CaO-MgO-MnO-FeO-Fe,O,-AlrOr- to two-feldspargeothemometry American Mineralogist, 70, 696-7 0 l. TiOr-SiO,-H,-Or. Journal of Metamorphic Geology, 8, 89-124. Raudsepp,M., Turncock, A.C., and Hamhome, F.C. (1991) Amphibole Jenkins, D.M. (1987) Synthesisand characterizationoftrernolite in the synthesis at low pressure:What grows and what doesn't. European systemHrO-CaO-MgO-SiOr. American Mineralogist, 72, 7O7-7 | 5. Journal of Mineralogy, 3, 983-1004. -(1988) Experimental study of the join tremolite-tschermakite:A Robie, R.A., Hemingway, B.S.,and Fischer,J.R. (1979) Thermodynamic reinvestigation.Contributions to Mineralogy and Petrology, 99, 392- properties ofminerals and related substancesaf 298.15 K and I bar 400. (105Pascals) pressure and at higher temperatures.U.S. GeologicalSur- -(1994) Experimental reversal of the aluminum content in tremo- vey, Bulletin 1452,456p. litic amphiboles in the system HrO-CaO-MgO-AlrOr-SiO,. American Smelik, E.A., and Veblen, D.R. (1993) A transmission and analltrcal Journaf of Sciencn.294. 593-620. electron microscope study of exsolution microstructuresand mecha- Jenkins,D.M, Holland,T.J.B., and Clare,A.K. (1991)Experimental de- nisms in the orthoamphiboles anthophyllite and gedrite. American termination of the pressure-temperature stability field and therrno- Mineralogist, 78, 5l l-532. chemical propertiesof synthetic tremolite. American Mineralogist, 76, Smelik, E.A., Nyman, M.W., and Veblen,D.R. (1991)Pervasive exso- 4s8-469 lution within the calcic amphibole series:TEM evidencefor a misci- Johannes,W., Bell, P.M., Mao, H.K., Boettcher, A.L., Chipman, D.W., bility gap betweenactinolite and hornblendein natural samples.Amer- Hays, J.F., Newton, R.C., and Seifert, F. (1971) An interlaborarory ican Mineralogist, 76, | 184-1204. comparison ofpiston-cylinder pressureand calibration using the albite Veblen, D.R., and Buseck,P.R. (1979) Chain-width order and disorder breakdown reaction Contributions to Mineralogy and Petrology, 32, in biopyriboles. American Mineralogist, 64, 687-7 O0. 24-38. Welch, MD., and Pawley, A.R. (1991) Tremolite: New enthalpy and Kiseleva, I.A., and Ogorodova, LP. (1984) High-temperature solution entropy data from a phase equilibrium study of the reaction tremolite calorimetry for determining the enthalpiesof formation for hydroxyl- : 2 diopside + 1.5 orthoenstatite + B-quartz + HrO. American Min- containing minerals such as talc and tremolite. Geochemistry Inter- eralogist,7 6, 193l-1939. national. 21. 36-46 Young, R.A. (1993) Introduction to the Rietveld method. In R.A. Young, Kiseleva,I.A., Ogorodova,L.P., Topor, N.D., and Chigareva,O.G. (1979) Ed., The Rietveld method, p.39-42. Oxford University Press,Oxford, A thermochemical study of the CaO-MgO-SiO, system.Geochemistry U.K. International. 16. 122-134. L6ger,A., and Ferry, J.M. (1991) Highly atuminous hornblendefrom low- pressure metacarbonates and a preliminary thermodynamic model for the Al content of calcic amphibole. American Mineralogist, 76, lOO2- M,rNuscnrP.r RECEIVEDAPrIr 4, 1994 l0l 7. Mem;scnrrr AccEp,rEDJvrv 25, 1994