American Mineralogist, Volume 77, pages I166-I 17I, 1992

Kinetics of the marcasite-pyritetransformation: An infrared spectroscopicstudy

Ar-rsrarn R. LnNNrn, Davro J. YaucneN Department of Geology, University of Manchester,Oxford Road, ManchesterMl3 9PL, England

Ansrn-lcr The kinetics of the solid-state transformation of marcasiteto pyrite have been investi- gated using infrared spectroscopy.Kinetic analysis using the Johnson-Mehl or Avrami- Erofe'ev equation is based on measurementsof infrared absorption intensity made on marcasite samples heated in the temperature range of 698-735 K. From the Arrhenius equation, an activation energyof 253 + 8 kJlmol and frequencyfactor of 4.6 x l0r5/min is calculated for the conversion from marcasite to pyrite. The transformation is modeled by a second-orderJohnson-Mehl equation suggestinga nucleation and growth model for the formation of pyrite in the marcasite matrix.

INrnooucrroN The differencein A@ values at 298.15 K for marcasite and pyrite is small; for pyrite, AGP: -160.060 kJ/mol, Pyrite and marcasite are naturally occurring phasesin whereasfor marcasite,AGP : - 156.159kJ/mol. At 700 the Fe-S system,dimorphs of composition FeSr.The py- K, AGPfor pyrite : -137.074 kJlmol, for marcasite,it rite crystal structure is cubic (spacegroup Pd3), with the is -133.617 kJlmol (Chaseet al., 1985),based on the octahedrally coordinated Fe atoms at the corners and data of Gronvold and Westrum (1976\. The formation face centersof the cube unit cell. Disulfide pairs lie at the and persistenceof marcasite is thus due not to its ther- center ofthe cube and at the midpoints ofthe cube edges modynamic stability but to kinetic factors. and are oriented such that their axes are parallel to four Whereaspyrite has been synthesizedboth by high-tem- nonintersectingbody diagonalsofthe cubic lattice. Each perature reaction betweenelemental Fe and S and by pre- S atom is tetrahedrally coordinated to three Fe atoms and cipitation from aqueous solution, the synthesisof mar- one S atom. casite has been achieved only by reaction of Fe2*and S Marcasite has an orthorhombic unit cell (spacegroup species in acidic aqueous solutions (Murowchick and Pnnm) and, like pyrite, has Fe atoms in octahedral co- Barnes, 1986; Schoonen and Barnes, l99la, l99lb, ordination with S, and S atoms tetrahedrally coordinated l99lc). to three Fe atoms and one S atom. The differencebetween Tossell et al. (198l) proposedthat the formation of the marcasiteand pyrite structuresis found in the linking marcasite from acidic solutions is a function of reaction of the Fe-centeredoctahedra. In the marcasite structure, mechanism. A valencecounting schemesuggests that the theseoctahedra share two edgesin planesnormal to (001); formation of marcasite in preferenceto the more stable in pyrite, the octahedra are linked at corners (Vaughan pyrite (with a system of 14 electron dianions) may be a and Craig, 1978). The pyrite structure can be obtained result of H* interaction with the Sj dianion electrons. from that ofmarcasite by rotation ofhalfofthe S, groups The effectof the H* ion is to withdraw two electronsfrom through 90'. the metal 3do-dianion III*g system, giving a distorted Marcasite readily inverts to pyrite when heated under geometry to the three Fe atoms coordinated to each S vacuumto temperaturesabove 673 K (Fleet,1970; Gron- atom of the disulfide group. Loss of H* would then ex- vold and Westrum, 1976; Kjekshus and Rakke, 1975). pand the M-A-M angle to the value of 97" observed in Murowchick and Barnes (1986), in contrast, reported in- marcasrte. version of very fine-grainedsynthetic marcasiteto pyrite This model was invoked by Murowchick and Barnes at ambient temperatures,and Rising (1973) found that (1986) to interpret experiments that showed a relation- marcasitereadily transformed to pyrite above 433 K un- ship between the pK, of the polysulfide solution species der hydrothermal reaction conditions. To determine and the pH of marcasite formation. Murowchick and whether marcasite has a field of stability below 673 K, Barnes argued that, in acidic solutions, the addition of an investigation of the thermodynamic stability of the HrS, to the growing FeS, surfaceresults in the marcasite two iron disulphides utilizing adiabatic-shield calorime- structure. The electron valency of HrS, is effectively the try was made by Gronvold and Westrum (1976). This sameas that of a l2-electron valent dianion pair. Binding study demonstratedthat marcasiteis metastablewith re- of HrS, to the crystal surfacehas an electron withdrawing spectto pyrite from 5 to 700 K. Above 700 K, marcasite effect that stabilizes the marcasite structure. undergoesirreversible exothermic transformation to py- In comparison, when HS; is the predominant poly- rite. sulfide species,it is proposedthat the proton has no struc- 0003-oo4x/9 2/ 1| | 2- | 166$02.00 I 166 LENNIE AND VAUGHAN: MARCASITE-PYRITE TRANSFORMATION rl67 turally determining effect, and so the more stable pyrite Infrared spectroscopy was selected for quantitative structure is formed. analysis of the fraction of marcasite transformed to py- Schoonenand Barnes(l99lc) concludedfrom experi- rite. Portions of the milled sampleswere diluted by add- mental work that the rate of direct nucleation of pyrite ing oven-dried spectralgrade KBr in the ratio of 300 parts and marcasite in slightly acidic solutions is insignificant by weight of KBr to one part by weight of sample. The below 573 K. Instead,most pyrite and marcasitein hy- diluted sampleswere then thoroughly mixed in the agate drothermal ores form by sulfidation of iron monosulfide mill vessel.Disks were pressedfrom 200 mg of the sam- precursors.They interpreted the Murowchick and Barnes ple plus KBr mixture in a Specac13 mm infrared die at (1986)relationship of polysulfidepK to formation pH in an appliedpressure of 9.5 tons.The small ratio of sample terms of surface sulfide specieson the growing pyrite or to KBr was necessaryto avoid high absorption valuesand marcasitecrystal, rather than in terms of solution sulfide thus possible departuresfrom the Beer-Lambert relation specres. in the spectralrange ofinterest. In a crystallographic study (Fleet, 1970) of the struc- Srl.tlolnos tural aspectsof the marcasite-pyrite transformation, il was found from X-ray measurementsthat the inversion To obtain a calibration curve for measurementof the to pyrite was partially complete after 12 h at 698 K and concentration of pyrite and marcasitein the heated sam- essentiallycomplete at 748 K after lessthan 4 h of heat- ples, a seriesof pyrite-marcasite standardsranging from ing. Attempts by Kjekshus and Rakke (1975) to detect 100 wto/omarcasite to 100 wto/opyrite were prepared by the inversion using differential thermal analysis (DTA) mixing weighedportions ofground natural samples.These were, however, unsuccessful. mixtures were ground in the agateball mill, diluted with The present study was undertaken to measurethe ac- KBr in a l/300 wt ratio, and mixed with KBr in the agate tivation energy of this high-temperature transformation mill. Disks of standard mixtures were pressedfrom 200 and to determine the mechanism bv which this transfor- mg of sampleas describedabove. Absorption of moisture mation takes place. was minimized by storageof samples,KBr' and prepared disks in a vacuum desiccator. ExpnnlvrnNTAl APPRoACH Infrared spectrawere recordedon a Perkin-Elmer 17l0 Marcasite samplesfrom the Harwood Mineral Collec- FTIR spectrometerover the spectralrange 400G-220 cm-' '. tion of the University of Manchesterand pyrite obtained with 50 scansper sample at a resolution of I cm Minerals were characterizedby X-ray from Manchester Rnsur-rs powder diffraction and microprobe analyses. Analysis of marcasite samplesheated in evacuatedsil- The infrared spectra of pyrite and marcasite between ica tubes by X-ray powder diffractometry confirmed that 600 and 300 cm-' are shown in Figure l. The absorption I transformation to pyrite only occurs over the experimen- peak of pyrite at 350 cm and the marcasite absorption I tal temperaturerange. Diffraction peakscorresponding to peak at 359 cm were selectedfor measurementof ab- pyrrhotite were not detected. sorption. Thesepeaks are well separated,unlike the more To establish the fraction of marcasite transformed to strongly absorbing peaks of marcasite and pyrite in the pyrite, a sample of marcasite was coarselyground in an region of 425 cm-r, which overlap significantly. agate mortar and pestle under amyl acetate.Portions of Standard mixtures of marcasite and pyrite, ranging from the ground marcasite were loaded into silica tubes and pure marcasite to pure pyrite, were analyzedby the KBr held in placewith silica wool. Thesesilica tubeswere then disk method. The absorption intensities were obtained sealedunder vacuum, using an Or-CHo flame. from the recorded spectraby subtraction ofthe base-line Heating of silica tubes containing the marcasite sam- absorbancefrom the peak absorbance.The base line at ples was carried out in a horizontal tube furnace pre- the peak wavenumber was established by means of a heated to the required temperature.After the silica tubes curve-fitting computer Program. were placed in the furnace, the temperature was moni- These absorbanceshave been plotted against the weight tored by a measuring thermocouple close to the end of percent of the mineral contained in the KBr disk in Fig- the silica tube containing the marcasitesample. The mea- ure 2. The plots of absorption intensity values against suring thermocouple was calibrated against the melting FeS, concentration have been fitted using least-squares points ofSn and Pb. linear regressionanalysis and show a straight-line rela- Each tube was removed from the furnace after a timed tionship in accordancewith the Beer-Lambert law (At- heating period and immediately quenchedin HrO to halt kins. 1986). the inversion reaction. This procedurewas carried out at It has been assumedthat the disk thicknessis the same four temperatures(735, 723,710, and 698 K) and over in all samples,allowing combination of extinction coef- a range of heating times at each temperature. After the ficient and pathJength terms in the Beer-Lambert law to timed heating and quenchingprocess, the marcasitesam- give the slope ofthe calibration graph- Disk thicknesses ples were removed from the evacuated silica tubes and were checkedusing a micrometer. ground to a fine powder in a small agateball mill (Speca- The fraction of marcasite converted to pyrite in the mill6000). heated sampleswas obtained by measuringabsorbances, I 168 LENNIE AND VAUGHAN: MARCASITE-PYRITE TRANSFORMATION

A, 0.20

q '1 0.08

0.04

0.0 0.1 0.2 0.3 U) weight percentof FeS2 in KBr |r Fig. 2. Calibration graphs for marcasite and pyrite in mar- casite + pyrite standard mixtures in KBr disks. Filled circles: absorption ofmarcasite at v : 359 cm I against weight percent 600 500 400 300 marcasite in KBr disks. The marcasite calibration plot has an interceptof0.005, a slopeof0.275, and a correlationcoefrcient wavenumber (cm-t) (r) of 0.93. Open circles:absorption of pyrite at v : 349 cm I againstweight percentpyrite Fig. l. Infraredtransmittance spectra of marcasite pyrite in KBr disks. The pyrite calibration and plot usingiron disulfideincorporated into pressed has an interceptof 0.020, slope of 0.590, and correlation KBr disks:spec- (r) trum a : marcasite,spectrum 6 : pyrite. coefficient of 0.96.

deriving the weight percent value from the calibration | (l _ graph, and then calculating the fraction ofpyrite (a) from dq/dt: Knt,n o) (2) the following relation: describes the isothermal kinetics of a wide variety of sol- id-state reactions (Hancock and Sharp, 1972). a: P/(P + M) 0) ln addi- tion, many theoretically derived kinetic equations can be where P and M are, respectively, the weight percent of reduced to this equation. pyrite and weight percent of marcasite measured in the Integration and rearrangement ofEquation 2 gives KBr disk. Calculationof a in this way reducesthe effect ot: I - (k.l),"] (3) of variations in absorbancesdue to particle size varia- exp[- tions and uneven distribution of iron sulfide in the KBr which is known as the Johnson-Mehl or Avrami-Erofe'ev matrix. The standarderror of estimatein a was calculated equation. For rate curves conforming to Equation 3, if from differencesbetween prepared and measured frac- time is plotted on a logarithmic scale, then the shape of pyrite tions of in the standardmixtures. the curve is determined only by m, and k fixes the posi- From the calibration graphs, the concentrations of tion on the time axis (Burke, 1965; Redfern, 1987). Fig- marcasite pyrite and in the heated and then ouenched ure 3 shows the marcasite transformation data plotted as sampleshave been calculated. These analyses are plofted a against ln /. The curves are of similar shape, indicating in Figure 3 (as a against ln t). similar kinetic behavior. Reaction curves having the same shape for a range of Krxrrrc ANALysrs experimental temperatures are termed isokinetic and have It has been found empirically that a rate equation of the same value of m if the Johnson-Mehl equation is the form obeyed. To evaluate m and k, the Johnson-Mehl equation LENNIE AND VAUGHAN: MARCASITE-PYRITE TRANSFORMATION l 169

2.0 0.99 1.0 0.90

0.70 0.0 x 0.50

-1.0 0.30

50 100 s00 1000 20003000 t -2.0 Fig. 3. Fractionof pyrite (a) formedby transformationof 0.10 marcasiteagainst logarithm of time in minutes.Measurements at 735 K are representedby fllled circles,at 723 K by open squares,at 7 10K by opencircles, and at 698K by filledsquares. -3.0 The curvesare obtained from theJohnson-Mehl equation using valuesofn and k calculatedfrom experimentaldata. 6 lnt is converted to Fig. 4. Plot of ln [-ln(l - a)] againstlogarithm of time (in ln[-ln(l - a)]: m-lnk* m.llt (4) minutes).Measurements at 735 K arerepresented by filled cir- cles,at 723K by opensquares, at 7 I 0 K by opencircles, and at by taking logarithms of Equation 3. 698 K by filled squares. A graph of ln[-ln(l - a)] vs. ln I will be linear if the reaction conforms to the Johnson-Mehl equation. The id-state transformation results of Fleet (1970) referred to value of z is obtained from the slope ofthe graph, and in the introduction.Kjekshus and Rakke(1975) observed k is derived from the y intercept value from the relation that at 573 K, heat treatment under vacuum for up to 14 months does not induce changesin marcasite samples, k: exp(intercept/m). (5) whereasat 623 Kafter the sameperiod, mixtures of mar- The data from the experiment are plotted in Figure 4 casite and pyrite are obtained. At 613 K, mixtures of with ln / as abscissaand lnfin(l - a)] as ordinate. marcasite and pyrite are produced after heating for one From least-squareslinear regressionanalysis of these month. plots, values for m and k have been obtained, and they Estimates of the time taken for a given fraction (a) of are given in Table l. These plots show no pronounced pyrite to be produced from marcasite can be made using curvature, which suggestsconformation to the Johnson- the activation energyand frequencyfactor obtainedabove. Mehl equation throughout the transformation process. A rate constant is derived from the Arrhenius equation At temperaturesof 735, 723, and 7 l0 K, the values for for the temperature of interest, and the time taken for a rn are similar, indicating that the transformation over this given pyrite fraction (a) to be obtained is calculatedfrom temperaturerange can be modeled by one kinetic process. the following form of the Johnson-Mehl equation: Although the value of rn derived from the measurements 1: lexp{m-t 'ln[-ln(l - a)l])/k. (7) at 698 K is greaterthan those derived for the higher tem- peratures, it is proposed that this is a result of experi- Calculations using the values of -8" and I derived above mental variation rather than a fundamental change in and with m: 2 sugest that transformation of half of the reaction mechanism. marcasite to pyrite takes place in approximately 6 d at with the logarithmic form of the Arrhenius equation, 673 K, in 205 d at 623 K, and in 40 yr at 573 K. Theseestimated values, which should be usedas a guide ln k: ln A - E^/RT (6) only, are consistent with the reported experimental val- values of ln k obtained from the empirical integratedrate equation are plotted against 1/Z in Figure 5. These give Treu 1. Datafrom plot of In[-ln(l - a)] againstIn t an activation energy, obtained from the gradient (-E^/ + : r R), of 253 8 kJ/mol, with a frequency factor A 4.6 m (correlation k (min') x lOrs/min. r(K) (slope) coefficient) (fromintercept) 735 2.2 + O,3 0.96 0.00475+ 0.00030 DrscussroN 723 2.4 + O.3 0.97 0.00238+ 0.00013 The experimental data presented here are consistent 710 2.1+ 0.3 0.99 0.00101+ 0.00005 698 2.9 + 0.3 0.94 0.00054+ 0.00003 with the previously reported time, temperature, and sol- I 170 LENNIE AND VAUGHAN: MARCASITE-PYRITE TRANSFORMATION

J4

-7.00

-8.00

1.340 r.380 r.420 1000 T (K) Fig. 5. Arrheniusplot for calculationofthe activationener- Fig. 6. Modelof marcasite(a) pyrite(b) gy. Valuesof k derivedfrom the Johnson-Mehlequatron are and structuralcom- plottedas ln k against1000/f (K). ponentsshowing inversion ofthe centraldisulfide pair required to transformmarcasite to pyrite.S atomsare represented by the opencircles, Fe atomsby the filledcircles.

ues of Kjekshusand Rakke (1975). That would suggest that a similar transformation mechanism occurs at lower sure long-term preservation of the marcasite structure in temperatures,although at considerably slower rates. natural systems. Murowchick and Barnes(1986) suggested that the tem- If marcasiteis slightly S deficient comparedwith pyrite, perature-dependentdistribution of marcasite in the Sal- as was proposedby Buerger(1934), then readyavailabil- ton Sea geothermal field indicates that temperatures of ity of aqueousS speciesin hydrothermal systemscould below 433 K are needed to preserve natural marcasite speedthe transformation to pyrite by a diffusion process over a multimillion-year time span. This theory is based (H. L. Barnes,personal communication, 1992). on studies by Rising (1973) and McKibbin and Elders However, from values of la tabulated by Hancock and (1e85). Sharp (1972) for various solid-statereactions, the mar- A half-life for the solid-state transformation of mar- casite to pyrite transformation in this experiment corre- casite to pyrite estimated from the above data at 473 K spondsto a nucleationand growth mechanismrather than is 3 x 100yr, whereasat 433 K, calculationsgive a half- to a diffusion process.Although Christian (1975) cau- life of 1.3 x lOeyr. Thesehalflives show a consistency tioned against proposing a mechanism simply from the with the proposal of Murowchick and Barnes (1936). value of m obtained in the kinetic analysisofa transfor- Again, these figures should be used with caution, taking mation, examination ofthe crystal structuresof marcasite note of the limitations of extrapolating higher tempera- and pyrite suggeststhat a nucleation and growth model ture kinetic data to lower temperatures,and the possible for the isochemicaltransformation is more plausible than dangers of inferring that the reaction mechanism is the transformation by diffusion. Furthermore, the suggestion same throughout the temperature range. It is, however, by Fleet (1970) that the structural reorganizationin- kinetic factors and not thermodynamic stabilitv that en- volved in the transformationis minimal implies negligi- LENNIE AND VAUGHAN: MARCASITE-PYRITE TRANSFORMATION tl7 | ble diffusion of Fe and S within the transforming crystal Donald, R.A, and Syverud,A.N. (1985)JANAF thermochemicalta- and ChemicalReference Data, I 4, I I 98- I I 99' matnx. bles.Journal ofPhysical Christian, J.W (l 975) The theory of transformationsin metalsand alloys' A nucleation model is in agreementwith the observa- Part l: Equilibrium and generalkinetic theory (2nd edition). Pergamon tion that rotation of half of the S, groups in marcasiteby Press,Oxford, England. 90' would result in the pyrite structure (Tossell et al., Fleet. M.E. (1970) Structural aspectsofthe marcasite-pyritetransforma- l98l). This would involve breakingone of the threeS-Fe tion. Canadian Mineralogist, 10, 225-231. Westrum, E.F., Jr' (1976) Heat capacitiesof iron di- link in order to allow Gronvold. F., and bonds on eachend ofthe disulfide sulfides.Thermodynamics of marcasitefrom 5 to 700 K, pyrite from rotation to occur. The Fe-S bonds most likely to break 300 to 780 K, and the transformation of marcasiteto pyrite. Joumal are those forming the four-membered rings arising from of ChemicalThermodynamics, 8, 1039-1048. edge-sharingoctahedra in the marcasite structure. These Hancock, J.D., and Sharp, J.H. (1972) Method of comparing solid-state decompositionofkaolinite, bruc- featuresunder most strain during heat- kinetic data and its application to the are the structural ofthe American Ceramic Society, 55, 74-77' group ite and BaCO.. Journal ing. Breaking these bonds would allow the S-S to Kjekshus,A., and Rakke, T. (1975)Compounds with marcasitetype crys- invert, and new Fe bonds to attach to the S opposite to tal structure. XI. High temperature studies of chalcogenides'Acta the former Fe-bondingposition on the S (Fig. 6). It is Chemica ScandinavicaA, 29, 443-452. (1985) mineralization unlikely that the S-S bonds in the structure are broken. McKibbin, M.A., and Elders, W.A. Fe-Zn-Cu-Pb geothermalsystem, Imperial Valley, California. Eco- (1975) values of lzl from the in the Salton Sea Christian summarized nomic Geology,80, 539-559. Johnson-Mehlequation that describevarious models of Murowchick, J.B., and Barnes,H L. (1986) Marcasite precipitation from polymorphic transformations. A value of m : 2 corre- hydrothermal solutions.Geochimica et CosmochimicaActa, 50' 2615- sponds to a grain-edgenucleation model. The pyrite nu- 2629. (1987) The kinetics of dehydroxylation of kaolinite. Clay cleii formed by such a processcould then grow along and Redfem, S.A.T. Minerals,22,447-456. outward from the marcasitegrain edgeforming the com- Rising, B.A. (1973) Phaserelations among pyrite, marcasite,and pyrrho- monly observedintergrowths (Fleet, 1970) of marcasite rite below 300 eC, 192 p. Ph.D. thesis,Pennsylvania State Univenity, { l0l } parallelto {001} of pyrite. University Park, Pennsylvania. Schoonen,M.A.A., and Barnes, H.L. (l99la) Reactions forming pyrite of FeS,below 100 lC. Geo- ACKNOWLEDGMENTS and marcasitefrom solution: L Nucleation chimicaet CosmochimicaActa, 55, 1495-1504. Thanks are due to Dennis Rawlinson of the Chemistry Department, - (l 99 lb) Reactionsforming pyrite and marcasitefrom solution: II. ManchesterUniversity, for valuable assistancewith the infrared analysis. Via FeS precursorsbelow 100 "C. Geochimica et CosmochimicaActa, Discussionswith Simon Redfern on collection and analysisofkinetic data 55,1505-1514. provided an important input to this work. We thank H.L. Barnesfor his -(l99lc) Mechanisms of pyrite and marcasiteformation from so- constructive review, which helped to clarify interpretation of the data. lution: III. Hydrothemal processes.Geochimica et CosmochimicaActa, This project was supportedby NERC grant GR3/6823 A. 55,3491-3504. Tossell,J A, Vaughan,D J., and Burdett,J.K. (1981)Pynte, marcasrte, RnrnnnNcBscrrnn and arsenopyritetype minerals: Crystal chemical and structural prin- ciples. Physicsand Chemistry of Minerals, 7, 117-184- Atkins, P.W. (1986) Physical chemistry (3rd edition) Oxford University Vaughan,D.J., and Craig, J.R. (1978)Mineral chemistry of metal sulfrdes' Press,Oxford, England. Cambridge University Press,Cambridge, England Buerger,M J. (1934) The pyrite-marcasiterelation. American Mineralo- gist,19,37-61. Burke, J (1965) The kinetics ofphase transformationsin metals' Perga- mon Press,Oxford. MeNuscnrrr RECEIvEDJeNuenv 13, 1992 Chase.M W.. Jr., Davies, C.A., Downey, J.R., Jr., Frurip, D.J , Mc- Me.Nuscnrrr AccEPTEDJtxa 25, 1992