1 Introduction

The hardening reaction taking place when aclay-rich sediment is fired is a fundamental char-acteristic of ceramic technology since the prehis-tory This process occurs by high-T low-P mineraltransformations that are mainly influenced by themineralogical composition of the raw clay-richmaterial its granulometry the firing temperatureas well as the kiln (oven) atmosphere conditions(Maggetti 1982)

Previous works have studied the mineralogicaland textural changes taking place following firingof raw clay (Tite amp Maniatis 1975 Freestone ampMiddleton 1987 Veniale 1990 Duminuco et al1996 Riccardi et al 1999) as well as the miner-alogical and structural modifications induced bythe presence of carbonates in the paste (Rye 1976Peters amp Iberg 1978 Maggetti 1982) Eventhough some reactions taking place when singlephases form at high temperature (T gt 750degC) arewell-established (Peters amp Iberg 1978) the influ-

Eur J Mineral2001 13 621-634

Carbonate and silicate phase reactions during ceramic firing

GIUSEPPE CULTRONE(1) CARLOS RODRIGUEZ-NAVARRO(1) EDUARDO SEBASTIAN(1)OLGA CAZALLA(1) AND MARIA JOSE DE LA TORRE(2)

(1) Departamento de Mineralogiacutea y Petrologiacutea - Universidad de GranadaFuente Nueva sn - 18002 Granada Spain

(2) Departamento de Geologiacutea Universidad de Jaeacuten Spain

Abstract Mineralogical textural and chemical analyses of clay-rich materials following firing evidence that initialmineralogical differences between two raw materials (one with carbonates and the other without) influence the tex-tural and mineralogical evolution of the ceramics as T increases from 700 to 1100degC Mineralogical and texturalchanges are interpreted considering local marked disequilibria in a system that resembles a small-scale high-T meta-morphic process (eg contact aureoles in pyrometamorphism) In such conditions rapid heating induces significantoverstepping in mineral reaction preventing stable phase formation and favoring metastable ones High-T transfor-mations in non-carbonate materials include microcline structure collapse andor partial transformation into sanidineand mullite plus sanidine formation at the expenses of muscovite andor illite at T sup3 800degC Mullite forms by mus-covite-out topotactic replacement following the orientation of mica crystals ie former (001)muscovite are ^ to(001)mullite This reaction is favored by minimization of free energy during phase transition Partial melting followedby fingered structure development at the carbonate-silicate reaction interface enhanced high-T Ca (and Mg) silicatesformation in carbonate-rich materials Gehlenite wollastonite diopside and anorthite form at carbonate-silicateinterfaces by combined mass transport (viscous flow) and reaction-diffusion processes These results may add to abetter understanding of the complex high-T transformations of silicate phases in both natural (eg pyrometamor-phism) and artificial (eg ceramic processing) systems This information is important to elucidate technologicalachievements and raw material sources of ancient civilizations and it can also be used to select appropriate clay com-position and firing temperatures for new bricks used in cultural heritage conservation interventions

Key-words Ceramics carbonates clay high-T reactions gehlenite mullite wollastonite reaction-diffusion fingersmuscovite-out reaction architectural conservation

0935-1221010013-0621 $ 350atilde 2001 E Schweizerbartrsquosche Verlagsbuchhandlung D-70176 Stuttgart

DOI 1011270935-122120010013-0621

E-mail carlosrngoliatugres

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

ence of and the interactions between various phas-es that coexist disappear or form are not wellestablished Little is known on the transformationsundergone by silicate and carbonate phases at thereaction interfaces or the phyllosilicate-out reac-tion In particular there is no clear understandingof the transport mechanisms of reactants (ie dif-fusion vs mass transportviscous flow) and theprocesses involved in mineral transformation athigh-T (ie solid-state reactions vs crystallizationfrom a melt) These issues are not solely relevantfor understanding ceramics as there is a close sim-ilarity between ceramic formation and the devel-opment of reaction textures resulting from markeddisequilibrium during pyrometamorphism (ie incontact aureoles or in xenoliths where heatingrates were very rapid Brearley 1986 Brearley ampRubie 1990 Preston et al 1999) They also mayhave important implications in understandingancient ceramic technologies or elucidating rawmaterial sources as well as in the designing ofnew ceramic materials in general or appropriate(ie compatible) conservation materials (iebricks) for architectural heritage conservationinterventions

It is the aim of this work to study the miner-alogical chemical and textural changes of bothresidual minerals and newly formed phases takingplace upon firing and in particular the mecha-nisms of mineral transformation of raw clay-richmaterials with and without carbonates

2 Materials and methods

Two Pliocene clay-rich materials were selectedone from Guadix (G) and the other form Viznar(V) two villages located in the vicinity of GranadaSpain Raw material was collected milled andsieved discarding the fraction with grain size gt 15mm Bricks paste was prepared adding 400 cc ofwater to 1000 g of raw material G and V brickswere fashioned using a wooden mould 245 acute 115acute 4 cm in size and fired in an air-ventilated electricoven (Hoerotec CR-35) at the following tempera-tures 700 800 900 1000 and 1100degC T wasraised at a heating rate of 3degC per minute first upto 100degC with one hour soaking time and later upto the peak T with a soaking time of three hoursFired brick samples were stored at constant T andrelative humidity (RH) conditions of 21degC and 55 respectively

Separation of the fractions lt 2 microm 2 to 20 andgt 20 microm was performed using a KUBOTA 2000centrifuge Grain size distribution in the lt 20 microm

fraction was determined using a laser-beam particlesize analyzer (GALAIh CIS-1) Grain size distribu-tion of gt 20 microm fraction was determined usingstandard ASTM sieves (50 microm up to 1 mm mesh f)

The mineralogy of the raw material as well asthe mineralogical changes taking place upon firingwere studied by powder X-ray diffraction (XRD)using a Philips PW-1710 diffractometer with auto-matic slit CuKa radiation (l = 15405 ) 3 to60deg2q explored area and 001deg2q s-1 goniometerspeed At least three samples (~ 1 g each) of eachbrick typefiring T were analyzed They were milledin agate mortar to lt 40 microm particle size XRD anal-ysis of the clay fraction (ie fraction with grain sizelt 2 microm) was performed using oriented aggregates(air-dried ethylene-glycol and dimethyl sulfoxidesolvated and 1 h heated at 550degC)

The bricks texture and microstructure as wellas the progress of mineral transformation andreactions upon firing were studied by means ofoptical microscopy (OM) and scanning electronmicroscopy (SEM Zeiss DMS 950) coupled withEDX microanalysis Two thin sections per sam-ple type and firing T were prepared A polishedthin section was prepared using one of themBoth SEM secondary electron (SE) and back-scattered electron (BSE) images were acquiredusing either small brick pieces (5 acute 5 acute 10 mm insize gold coated) or polished thin sections (car-bon coated) The same thin sections were used toanalyze small-scale compositional changes ofselected minerals following firing by means of anelectron microprobe (EMPA Cameca SX 50)The EMPA working conditions were 20 keVbeam energy 07 mA filament current and 2 micromspot-size diameter 14 analyses of muscovitecrystals (4700degC 6800degC and 41100degC) and25 of carbonates (6700degC 9800degC and101100degC) were performed Albite orthoclasepericlase wollastonite and oxides (Al2O3 Fe2O3and MnTiO2) were used as standards(Govindaraju 1989) Detection limit for majorelements after ZAF correction (Scott amp Love1983) was 001 wt

Bulk chemical analyses of ceramic materialsbefore and after firing at each target T were per-formed by means of X-ray fluorescence (XRFPhilips PW-1480) 1 g per raw material or firedbrick sample was finely ground and well mixedin agate mortar before being pressed into Al hold-er for disk preparation ZAF correction was per-formed systematically (Scott amp Love 1983)International standards (Govindaraju 1989) wereused thoroughly The estimated detection limitfor major elements was 001 wt

622

3 Results

31 Granulometry

V samples show a higher concentration of par-ticles lt 2 microm when compared with G samples(Fig 1) However the differences are not signifi-cant Both samples show a maximum in particlesize around 100 microm Carbonates are the mostabundant phase in the larger size fraction in Vsamples while schist pieces (with Ms plus Qtz andFs mineral symbols after Kretz (1983)) made upthe larger size fraction of G samples

Carbonate and silicate phase reactions during ceramic firing 623

Fig 2 Optical microscopy micrographs of a) calcite (Cal) grain fired at 800degC showing cracks which penetrates thebrick matrix (plane light) Reaction rims are observed at the carbonate-phyllosilicate (matrix) and carbonate-quartz(Qtz) interfaces and b) muscovite (Ms) crystal before (cross polars) and c) after heating up to 800degC (plane light)Mullite (Mul) observed in the latter

Fig 1 Grain-size distribution curves for Guadix andViznar raw material

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

32 Optical microscopy analysis

G and V samples fired at 700degC do not showsignificant mineralogical transformations Thetemper grain-size (ie phenocrysts) reaches up to15 mm No major textural changes are visible inthe carbonates at 700degC Phyllosilicates in thepaste are oriented due to compression duringmolding At 800degC carbonate decomposition in Vsamples is almost complete and some cracksappear surrounding the former carbonate grains(Fig 2a) The matrix in both G and V samplesturns dark with low birefringence as T increasesMuscovite crystals in both G and V samples startto transform into mullite at 800degC and some bub-bles develop within these crystals at higher T (Fig2b and 2c) Muscovite replacement by mullite iscompleted at 1000degC Mullite preserves the mor-phology of replaced muscovite In fact mullite(001) planes are ^ to (001) planes of former mus-covite grains as deduced by the optical axes ori-entation study At 1000-1100degC the matrix is fullyvitrified and brown colored (in both G and V sam-ples) Phyllosilicates in the matrix do not exist atthese T only some pseudomorphs after formercrystals can be observed An empty space withscattered grayish fragments at the pore edges canbe found where calcite was originally presentCracks observed in V samples at T lt 900degC disap-pear at 1100degC due to extended vitrification Poresbecome more abundant and they are no longerellipsoidal but spherical However an orientationof the remaining temper grains in planes parallel tothe brick largest faces (ie where compressionwas applied during molding) remains

33 XRD analysis

Raw G samples show significant amounts ofquartz and phyllosilicates with little feldspar Thelt 2 microm fraction includes illite and smectite withsmall amounts of kaolinite and paragonite (Table

1) There is no chlorite which could have beenmasked by the d(001) peaks of smectite (~ 14 Aring)and kaolinite (71 Aring) as evidenced by these peaksdisappearing after 90acute heating at 550degC Raw Vsamples have significant amounts of calcite anddolomite The clay fraction includes illite andsmectite with trace amounts of paragonite andkaolinite (Table 1) Therefore the main differenceamong the two raw materials lays in the miner-alogical composition of the gt 2 microm fraction

Regarding the phyllosilicates evolution uponfiring only the diffraction peak at 10 Aring corre-sponding to a dehydroxylated illite-like phaseremains at 700degC This peak intensity is reducedupon firing at higher T till it disappears at ~900degC (Fig 3a and 3b) Microcline main diffrac-tion peak reduces its intensity in G samples Thismineral disappears at T gt 1000degC when traceamounts of sanidine are formed Mullite is detect-ed in both G and V samples at T ~ 900degC increas-ing its main diffraction peak (d(210) = 339 Aring)intensity at higher T Mullite diffraction peaks aremore intense in phyllosilicate-rich G samples thanin carbonate-rich V samples Other phases under-go a significant increase in their diffraction peaksintensities this is the case of hematites in G sam-ples fired at T gt 1000degC New phases appear in Vsamples upon firing a) gehlenite (a melilite-groupphase) appears at 800degC increasing its mainBragg peak intensity at 900degC and reducing it athigher T b) wollastonite and diopside appear at1000degC and they show a significant increase intheir main diffraction peak at higher TDecomposition of calcite and dolomite begins at Tgt 700degC and both phases disappear at T gt 800degCQuartz remains as the most abundant phase at anyT in both V and G samples In V samples the pla-gioclase increases its Ca content (ie a more An-rich phase) as well as its main Bragg peakintensity as T increases This increase is more sig-nificant at T gt 1000degC The existence of an amor-phous phase (ie vitreous phase) is evidenced by

624

Table 1 Results of XRD analysis of the raw materials and the fraction with lt 2 m m grain size Mineral symbols afterKretz (1983) (in Table 3)

a rise of the background noise in the XRD patternat T gt 900degC in the G samples and at T gt 800degCin the V samples

34 SEM-EDX analysis

Significant textural changes most evident withrespect to pore morphology and volume weredetected upon firing An overall porosity reductionis observed at T gt 1000degC when the vitreous phasefills the pores

Samples fired at 700 and 800degC still preservethe laminar habit of phyllosilicates although mus-covite crystals clearly exfoliate along basal planes

most probably due to dehydroxylation (Fig 4a)The interlocking among particles is limited (Fig4d) V samples show some cracks At these T noclear evidence of sintering or partial melting isdetected However there is indirect evidence (ieincrease in background noise of XRD patterns)that some vitrification is reached at T lt 900degC incarbonate-rich V samples Vitrification is clearlyobserved in all samples at T gt 900degC ie phyl-losilicates appear deformed (their edges turnsmooth) andor aggregated (Fig 4b and 4e) andthe pores turn ellipsoidal with smooth surfaces AtT gt 1000degC these effects are less extended in thecarbonate-rich V samples (Fig 4c) than in the sil-

Carbonate and silicate phase reactions during ceramic firing 625

Fig 3 Guadix (a) and Viznar (b) samples powder X-ray difraction patterns Legend (mineral symbols after Kretz1983) Sm = smectite Ill = illite Pg = paragonite Qtz = quartz Cal = calcite Kln = kaolinite Phy = phyllosilicatesFs = feldspar Dol = dolomite Gh = gehlenite Hem = hematite Wo = wollastonite Di = diopside An = anorthiteMul = mullite Sa = sanidine CuKa X-ray radiation l = 15064 Aring

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

icate-rich G samples (Fig 4f) A higher degree ofparticle interlocking results in a porosity reduc-tion At 1100degC vitrification is significant in allsamples regardless of initial compositional differ-ences Pores coalesce as spherical cells due to par-tial or extended melting of clay particles in thematrix

BSE images of mineral and textural evolutionof samples with high carbonate content (ie Vsamples which undergo the most significant min-eralogical changes) give an interesting picture oflocal phase transformations taking place at grainboundaries No textural or mineralogical changescan be detected in brick fired at 700degC (Fig 5a)Carbonates comprise the larger-size fraction inthese samples which allows easy detection ofhigh-T reaction rims In fact at 800degC 2-5 micromthick reaction rims are observed developing at thecalcite-quartzphyllosilicate interface (Fig 5b)Rims composition corresponds to a calcium sili-cate phase presumably wollastonite (EDXresults) A dark rim surrounded by a light outershell is observed where dolomite was originallypresent (Fig 5c) However no compositional vari-ation among internal (greyish) and external(white) rims is detected We were not able tounambiguously detect the presence of gehlenite

within these reaction rims (ie some EDX analy-ses of 2-5 microm thick rims at the carbonate-silicateinterface showed Al together with Ca and Si butpossible contamination from underlying phyllosil-icates was not discarded) However the presenceof gehlenite was detected by XRD at this T(800degC) Muscovite phenocrysts (Fig 5d) show amorphology and composition typical for unalteredwhite mica At 1100degC muscovite crystals showsecondary bubbles development between layers(Fig 5e) The mica composition is homogeneous(no BSE image gray-level differences along thecrystals) and it is similar to that of an unchangedmuscovite but with slightly lower K concentra-tion At 1100degC carbonates (ie dolomite in Fig5f) shows an outer thick white Ca- and Si-rich rim(ie wollastonite) which includes small (micronsized) bright-white Ca- Si- and Al-rich particles(ie gehlenite or plagioclase) There is a strikingfeature of this reaction rim it shows a fingeredgeometry at the outer edge The darker dolomitecore shows a significant Mg enrichment

35 EMPA analysis

Due to the fact that only silicate phases in sam-ples with carbonates undergo mayor composition-

626

Fig 4 SEM secondary electron photomicrographs of Viznar samples fired at 800degC (a) 1000degC (b) and 1100degC(c) and Guadix samples fired at 700degC (d) 900degC (e) and 1000degC (f)

al changes upon firing only V samples were ana-lyzed using this technique In particular composi-tional changes of phyllosilicates and carbonateswere studied using polished thin sections of firedV bricks (one section per selected target T ie700 800 1100degC)

Phyllosilicates (Table 2) show a muscoviticcomposition at 700degC where some K is replaced byNa even though the overall composition does notseem to correspond to paragonite Neverthelessparagonite was detected in the raw clay (Fig 3b)

Fe Mg and Ti replace some Al in the octahedralsheets Replacement of OH- by F- is almost absentAt 800degC a slight increase in Fe content is detectedOnly analyses M9 and M10 show higher concen-trations of Na and Al At 1100degC the amount of Kis reduced while Ca and Si content is increasedThe total oxide content is close to 100 a valueconsistent with complete dehydroxylation of thephyllosilicates Analyses M11-M14 in Table 2 arein agreement with high-T phengite compositionaldata from Worden et al (1987)

Carbonate and silicate phase reactions during ceramic firing 627

Fig 5 BSE images and EDX analyses of carbonate-rich Viznar samples a) fired at 700degC b) fired at 800degC whenWo (EDX analysis) formed at the Cal- silicates interface c) detail of former carbonate grain showing a white outerreaction rim (Wo or Gh) a crack and an inner core with an outer darker shell (see text for details) d) Ms (see EDXanalysis) at 700degC e) former Ms after transformation into a Fs + Mul + melt mixture (see EDX analysis) f) DolPhy(+ Qtz) interface showing a reaction rim with fingered geometry formed at 1100degC g) detail of the outer reactionrim in previous figure EDX analysis shows the rim composition (ie Wo)

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

Calcite (Table 3) shows no chemical changesat 700degC The maximum MgO content is 104 wt Trace amounts of Fe Si Al and Mn are alsodetected in calcite (as well as in dolomite) At800degC both carbonates are fully transformed eitherinto burnt lime (CaO) or into a mixture of burtlime and periclase (MgO) (eg analysis C14)Gehlenite-like compositions are detected in someanalyses (eg C10) At 1100degC the presence ofnewly formed phases in particular wollastonite isclearly confirmed (eg C17) Ca content in thelime plus periclase mixture is reduced significant-ly due to the incorporation of this element intonewly formed high-T silicate phases As T increas-es a clear compositional trend toward Si- and Al-rich calcium phases (ie toward gehlenite andanorthite) is observed (Fig 6a) a fact consistentwith XRD and SEM-EDX results A clear trendtoward wollastonite-like compositions as well as a

trend toward MgO rich composition is alsoobserved as T increases (Fig 6b) Some EMPAanalyses suggest a larnite -like compositionHowever XRD results do not show the presenceof larnite

36 XRF analysis

According to the classification proposed forclay materials in ceramic industry by Fabbri ampFiori (1985) G raw material composition falls intothe quartz-feldspar sands category while V sam-ples fall out of any defined category due to theirlow silica and high calcium contents (Table 4) Asexpected firing does not affect major elementconcentration in the brick The only significantchange is detected in the loss on ignition values Itis evidenced that G samples lost almost all waterbetween 700 and 900degC while V samples show a

628

Table 2 EMPA analysis results of phyllosilicates composition (in wt )

Table 3 Composition of selected carbonate grains ( EMPA results in wt )

more gradual loss of water and CO2 as T increases(Table 4)

4 Discussion

Firing up to 700degC induce no significant min-eralogicaltextural transformations in both thecarbonate-rich V and the silicate-rich G sampleswith the exception of the clay minerals Clay min-erals structure is reported to collapse due to dehy-droxylation to resemble an illite-like structure(XRD data) when a T between 450 and 550degC isreached (Evans amp White 1958) A dehydroxylatedphyllosilicate phase structurally different to thehydrated one (Guggenheim et al 1987) is report-ed to exist up to 950degC when complete breakdownof the dehydroxylated illite occurs (Peters amp Iberg1978) Our data indicate that dehydroxylation isnot completed at T ~ 900degC which indicates thatthe kinetics of this process is slower than previ-ously estimated (Evans amp White 1958) (OH)-

released at T gt 700degC may play a significant rolein texture and mineral evolution at high-T as willbe discussed later On the other hand muscoviteshows a slight K deficit at 700degC probably due tothe high diffusion capacity of this element alongmuscovite (001) basal planes at T gt 500degC(Sanchez-Navas amp Galindo-Zaldivar 1993)Sanchez-Navas (1999) showed high-T K-diffu-sion-out of mica is induced by water adsorbedalong mica basal planes resulting from phyllosili-cate dehydroxylation

XRD data shows that dolomite most intenseBragg peak in V bricks significantly reduced itsintensity due to thermal decomposition at lower Tthan calcite main Bragg peak (Fig 3b) Dolomite

Carbonate and silicate phase reactions during ceramic firing 629

Table 4 Bulk composition of Guadix (G) and Viznar (V) samples (XRF results in wt )

LOI loss on ignition

Fig 6 a) ACS (Al2O3-CaO-SiO2) and b) CMS (CaO-MgO-SiO2) compositional diagrams of carbonates andreaction products (EMPA results)

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

starts to decompose according to (1) at ~ 700degC (1atm pressure) while calcite decomposes at higherT (830 ndash 870degC) according to (3) (Boynton 1980)Shoval (1988) proposed a two steps thermaldecomposition for dolomite involving formationof periclase and calcite (2) followed by calcitedecomposition at higher T (3)

dolomite lime periclaseCaMg(CO3)2 reg CaO + MgO + 2CO2 (1)

calcite CaMg(CO3)2 reg CaCO3 + MgO + CO2 (2)

CaCO3 reg CaO + CO2 (3)

However our XRD results showing no increase incalcite 104 Bragg peak intensity as dolomitedecomposes at 700degC (Fig 3b) are not consistentwith Shovalacutes two-steps process

Carbonate decomposition results in shrinkageof the brick (Barahona et al 1985) Upon storageat 22degC and 55 RH burnt lime (CaO) readilytransforms into portlandite according to the fol-lowing reaction (Boynton 1980)

portlanditeCaO + H2O reg Ca(OH)2 (4)

This process generates crystallization pressure inconfined spaces (ie the brick pores originallyoccupied by CaO) resulting in crack development(as observed in Fig 2a) The dark-color rimaround former carbonate grains (Fig 5c) is proba-bly due to limited portlandite carbonation in thepresence of CO2 also resulting in volume increase(Moorehead 1985) and crack development(Butterworth amp Honeyborne 1952) As a commonpractice to avoid these undesired side-effectsbricks that contain carbonates were traditionallyimmersed in water to dissolve and eliminate exist-ing CaO or Ca(OH)2

This effect is less evident in the case of MgOPericlase in the presence of H2O (room T) under-goes a slow transformation (it takes months oreven years Webb 1952) into brucite (Mg(OH)2)and may eventually transform into hydromagne-site (Mg5(CO3)4(OH)24H2O) according to thefollowing reactions (Garavelli et al 1990)

MgO + H2O reg Mg(OH)2 (5)

hydromagnesite5Mg(OH)2 + 4CO2 reg Mg5(CO3)4(OH)24H2O (6)

EMPA results indicate the presence of Ca and Mgoxides however our XRD results do not clearlyshow the presence of Ca and Mg oxides or hydrox-ides mainly due to their low concentrations (lt 10wt ) and because their main diffraction peaks aremasked by other major phase peaks (eg quartzfeldspars hematites) The hydromagnesite mainBragg peak at 581 Aring is not present either

At T sup3 800degC and in the absence of carbonateschanges are constrained to textural modificationsand reactions where a low-T phase is decomposedand fully transformed into a high-T phase withoutmajor compositional changes (Riccardi et al1999) In particular muscovite undergoes a solid-state phase change into a mixture of mullite and K-feldspar (or a melt) at T ranging from 800 to1000degC The composition of high-T muscovite isclose to K-feldspar (EMPA results) although OManalysis points to the presence of mullite Brearley(1986) and Worden et al (1987) showed fine inter-growth of alternating nanometer up to a fewmicrometers thick bands of K-feldspar (or a meltwith composition close to that of K-feldspar) andmullite in pyrometamorphic muscovite which canexplain our EMPA results

The most significant textural and mineralogi-cal changes are observed in samples with carbon-ates when fired at T gt 800degC Also melting T andextent seems to be connected with the presence orabsence of carbonates In general melting starts atlower T (~ 800degC) when carbonates are present(Tite amp Maniatis 1975) Ca and Mg from carbon-ates may act as melting agents (Segnit ampAnderson 1972) but they are reported to some-how limit the extent of vitrification at T gt 1000degC(Everhart 1957 Nuacutentildeez et al 1992) In fact weobserved that vitrification is more extended athigher T when no carbonates exist This may beexplained considering that the phyllosilicate in thematrix of non-carbonate G samples not contribut-ing to the formation of high-T Ca (or Mg) silicatesand releasing significant amounts of H2O due todehydroxylation they can contribute to extensivemelting at high-T as demonstrated by Brearley ampRubie (1990) On the other hand the higher SiO2content in these samples provides a higher amountof potential silicate-rich melt

BSE images of reaction rims between carbon-ates and silicates show that formation of Ca-sili-cates starts at 800degC This is confirmed by XRD(Fig 3b) and EMPA analyses (Table 3) Gehlenitestarts to form at this T by grain-boundary reactionbetween CaO Al2O3 and SiO2 the first derivingfrom former carbonates and the later two fromalready dehydroxylated phyllosilicates (ie a

630

somehow amorphous mixture of 3SiO2Al2O3according to Peters amp Iberg 1978) as follows

illite calcite gehleniteKAl2(Si3Al)O10(OH)2 + 6CaCO3 reg 3Ca2Al2SiO7+ 6CO2 + 2H2O + K2O + 3SiO2 (7)

The formation of gehlenite is striking becauseone could expect the formation of an Al-richpyroxene such as fassaite (Dondi et al 1998)Nevertheless it should be indicated that gehleniteformation is a good example of the so-calledGoldsmithacutes ldquoease of crystallizationrdquo principlewhich states that a silicate phase with all Al in 4-coordination is easier to form (at least in the lab)than a phase that includes Al in both 4- and 6-coor-dination when (OH)- are not present (Goldsmith1953) This is due to the higher energy barrier nec-essary to overcome when Al has to enter in a 6-coordination if compared with a 4-coordination If(OH)- are present this energy barrier is reduceddue to the smaller electrostatic repulsion among(OH)- groups than among O2- in octahedral coor-dination Gehlenite has all Al in tetrahedral coor-dination (Warren 1930) therefore crystallizingpreferentially if compared with other anhydrousCa-Al silicates This simple principle also appliesto the high-T formation of other anhydrous Al-sil-icates as we will discuss later

Wollastonite formed at the carbonate-quartzinterface through the following reaction

wollastoniteCaO + SiO2 reg CaSiO3 (8)

Wollastonite and gehlenite appear at 800degC aT ~100degC lower than previously reported (Petersamp Iberg 1978 Dondi et al 1998) This is proba-bly due to a) the low resolution of the analyticaltechniques used in the past (ie XRD) which didnot allow the detection of the small initial reactionrim between carbonates and silicates as it is evi-denced by BSE images and EMPA analysesandor b) most previous works on high-T phaseformation in ceramic processing use single com-ponents as reactants without considering addition-al phases or previous reaction by-products (Petersamp Iberg 1978) We observed that in a closer-to-reality multicomponent system where interactionsbetween reactants and products occur differentformation T are obtained

At the dolomite-quartz interface diopsidestarts to form at 900degC by the following reaction

dolomite diopside2SiO2 + CaMg(CO3)2 reg CaMgSi2O6 + 2CO (9)

At this T anorthite starts to form at the expensesof calcite and illite (plus quartz)

illite calciteKAl2(Si3Al)O10(OH)2 + 2CaCO3 + 4SiO2 regK-feldspar anorthite

2KAlSi3O8 + 2Ca2Al2SiO8 + 2CO2 + H2O (10)

At higher temperatures (1000 to 1100degC)phyllosilicates have already disappeared in allsamples (V and G) being transformed into sani-dine and mullite plus a melt according to the fol-lowing reaction

illite (muscovite) sanidine 3KAl2(Si3Al)O10(OH)2 + 2SiO2 reg 3KAlSi3O8 +

mulliteAl6Si2O13 + 3H2O + melt (11)

Formation of a melt during the previous reaction(Brearley amp Rubie 1990) is consistent with theobserved ldquocellular structurerdquo of muscovite (Tite ampManiatis 1975)

Brearley (1986) proposed a somehow differentreaction for mullite formation at the expenses ofphengitic muscovite resulting in K-feldspar +mullite + biotite No biotite is detected in ourexperiments most probably due to the low Fe con-tent in V and G micas Brearley amp Rubie (1990)experimentally demonstrated that high-T mus-covite breakdown in the presence of quartz isinfluenced by the crystal size The smallest crys-tals have a strong tendency to react with H2Oresulting in enhanced melt formation while largemuscovite crystals form limited amounts of meltThis may account for the significant vitrificationof the matrix (specially in G samples) while largemuscovite grains (up to some mm in size) still donot show significant melting

Mullite which according to OM analysis firstappears at 800degC is the second most abundantphase (after quartz) at 1100degC in the richer inphyllosilicate G samples It is interesting to men-tion that K-feldspar progressively disappears as Tincreases This is not consistent with publisheddata (Maggetti 1982) It seems that low-T K-feldspar (microcline) is not stable at high-T there-fore it transforms into a high-T K-feldspar (iesanidine) or reacts with lime to form anorthiteBoth sanidine and anorthite are detected in thehigh-T bricks However a higher concentration ofsanidine should be expected to form speciallyconsidering reaction (11) A possible explanationfor the low K-feldspar content at high-T is that K-feldspar might not have enough time to develop itscrystalline structure (sanidine) therefore being

Carbonate and silicate phase reactions during ceramic firing 631

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

concentrated as an amorphous (sanidine-like)phase responsible in part for the rise in the back-ground noise in the XRD results Mullite forma-tion seems to be kinetically faster resulting in awell-developed crystalline phase which givessharp diffraction peaks Goldsmith (1953) indi-cates that mullite could be another example of hisldquosimplexity principlerdquo for the ldquoeaserdquo of crystal-lization if compared with other aluminum sili-cates In fact the mullite structure has a largeramount of Al in 4-coordination and less Al in 6-coordination than sillimanite (Rehak et al 1998)for instance which may account for the abundanceof the former and the non-existence of the latter inthe studied bricks

In carbonate-rich V samples fired at T gt1000degC a significant reduction of gehlenite maindiffraction peak is observed According to Petersamp Iberg (1978) this is due to the reaction of thisphase with quartz forming wollastonite and anor-thite This is consistent with the increase in themain Bragg peak intensity of these later mineralsat 1100degC

At the dolomite-silicate (ie Qtz and Phy)interface it is common to observe wollastoniteeither newly formed or resulting from the destruc-tion of gehlenite The texture of the interface israther complex Protrusion or fingers from thereaction zone penetrate into areas were completetransformation into wollastonite has occurred Theinterface geometry resembles the so-called ldquovis-cous fingersrdquo (Garcia-Ruiz 1992) occurringwhen two fluids or semi-plastic solids of differentcomposition and viscosity are placed in contactInitially the contact zone is a flat surface but lateron it reaches a fingered fractal geometry(Nittmamm et al 1985) In the carbonate-silicateinterface a similar process may have taken placewhere mass transport of the reactants may havebeen enhanced by partial melting The formationof a silicate-rich melt due to clay-minerals melting(Brearley amp Rubie 1990) in the external part ofthe ring in contact with a second inner melt formedby preexisting gehlenite or the eutectic CaO-SiO2would result in viscous finger development Thefractal geometry of the developing viscous fingersmay have enhanced diffusion of mobile ions suchas Ca toward silicate reactive sites Both EDXdata showing Mg enrichment in the dolomite core(Fig 5f) and the compositional trend towardsMgO-rich terms in the CMS diagram (Fig 6b)confirm that Ca is preferentially leached out of theformer dolomite grain Finger formation willresult in a significant increase in interface surfacearea resulting in higher reaction rates Another

alternative explanation for this fingered geometrydevelopment involves partial melting at the car-bonate-silicate interface and mass transport (ieviscous flow) through the clay-quartz porousmatrix resulting in the so-called ldquoscalloped struc-turerdquo proposed by Ortoleva et al (1987) for reac-tion-diffusion processes While the viscous fingermodel requires two melts the second model onlyrequires one melt It is not clear if at any momentconditions for two melt occurrence are reachedHence the second model seems more plausible

In any case the process which is originally dif-fusion driven is transformed into a mass (ie fin-gering) plus diffusion transport which greatlyenhances Ca-silicate formation This is consistentwith the thick reaction rims (up to 250 microm thicksee Fig 5f) that surround former carbonate grainsThese thick Ca-silicate shells would hardly devel-op just by diffusion in the short time-scale of theexperiment (3 hours at the peak T) In fact at800degC (conditions where melting would be verylimited) Ca (and Mg)-silicate rims formed aroundcalcite or dolomite are 2-5 microm thick At 1100degCwhen melting is extensive the reaction rate is sohigh that rims are two orders of magnitude thick-er Therefore mass transport (ie viscous flow)seems to have played a key role in high-T Ca-(Mg-) silicate formation Mass transport is also consis-tent with observation of some areas with differentcomposition (unreacted silicates) trapped betweenfingers resulting in the island geometry observedin Fig 5g It could be expected that without fin-gering the process of Ca- or (Ca-Mg)-silicate for-mation would be self-limiting ie once a rim ofthe newly formed silicate appears diffusion wouldbe jeopardized This shell would preclude or limitthe transport of Ca (andor Mg) from the carbon-ate or it would limit the transport of the lessmobile Si (and Al) from the surrounding silicatetowards the reaction front Therefore it is neces-sary that one of the above described transport-reaction mechanisms takes place This may alsoexplain why phase formation takes place at lowerT than reported previously The presence of multi-ple interfaces among different crystals mayenhance finger development resulting in kineticchanges favoring phase formation at lower T

5 Conclusions

It is observed that even though the overallcomposition of a ceramic material does not under-go major changes upon firing significant miner-alogical and textural modifications do occur

632

These changes are connected with two reactiontypes a) solid-state replacement of a phase by anew one with little or no compositional variation(ie the muscovite and illite case) and minimizingenergy changes by preserving the original textureand orientation of the replaced phase and b) reac-tion-diffusion at the interface between two miner-al grains with contrasting composition (iecarbonate-silicate interfaces) The latter reaction isenhanced if massviscous transport resulting frompartial melting and finger structure developmentoccurs Non-carbonated clay-rich materials (G)show phyllosilicate destruction at T ranging from700 to 900degC followed by vitrification which issignificant at T gt 1000degC Carbonate-rich materi-als (V) show major phase transformation uponphyllosillicates destruction including new Ca (orMg) silicates and aluminum-silicates formation Inthis later case the vitrification starts at lower Tand it is limited at higher T Therefore it is possi-ble to establish the firing T of ceramic materialswith composition broadly similar to that in ourwork based on the mineralogy Thus a) presenceof phyllosilicates indicates a low 700 to 900degC fir-ing T for the dehydroxylated illite (or muscovite)phase b) dolomite disappears at sup3 700degC whilecalcite is still present at 800degC c) gehleniteappears at 800degC reducing its concentration above1000degC d) diopside and wollastonite appear at800degC (in very low concentrations) as reactionrims between carbonates and silicates Formationof these minerals includes ldquofingeringrdquo develop-ment e) mullite appears at T gt 800degC while littleamounts of sanidine form as microcline disap-pears and f) anorthite develops at T gt1000degC incarbonate-rich bricks

When considering equilibrium thermodynamicdata and calculated reaction T for most of theabove mentioned phases and reactions (Holland ampPowell 1985 and references herein) it isobserved that rapid heating induces significantoverstepping Therefore phases formed are inmost cases metastable a fact consistent with anal-yses of pyrometamorphic assemblages (Brearleyamp Rubie 1990 and references herein)

The formation of non-equilibrium fractal mor-phologies at the carbonate-silicate reaction inter-faces may explain why high-T silicates incarbonate-rich ceramics develop very fast Thisgrain-to-grain reaction model may help explainhigh-T silicate formation in nature (eg contactmetamorphism pyrometamorphism in xenoliths)

The gained knowledge on mineral transforma-tion and the dynamics of interface mass and diffu-sion transport facilitated through ldquofingeringrdquo or

ldquoscallopingrdquo opens new ways to understand thecomplexities of ceramic processing in particularand high-T mineral transformation in general bothin natural (ie geologic settings) and artificialcontexts (eg archaeometric analysis of ancientceramics or the design of new bricks ndashreplicas ofancient onesndash for architectural heritage conserva-tion) However care must be taken when extrapo-lating these results to other raw materialcompositions andor ceramic processing methods

Acknowledgements This work has been finan-cially supported by the CGICYT projectMAT2000-1457 and the Research Group RNM0179 of the Junta de Andaluciacutea (Spain) We thankthe Centro de Instrumentacioacuten Cientifica (CIC) ofthe Granada University for allowing us to use theirSEM-EDX EMPA and XRF facilities We thankK Elert JM Garciacutea-Ruiz M Rodriguez-Gallego and A Sanchez-Navas for fruitful discus-sions and comments We also thank the in-depthreview by Dr F Goumltz-Neunhoeffer and an anony-mous referee

References

Barahona E Huertas F Pozzuoli A Linares J(1985) Firing properties of ceramic clays fromGranada province Spain Miner Petrogr Acta 29A577-590

Boynton RS (1980) Chemistry and technology oflime and limestone 2nd Ed Wiley New York

Brearley AJ (1986) An electron optical study of mus-covite breakdown in pelitic xenoliths duringpyrometamorphism Miner Mag 357 385-397

Brearley AJ amp Rubie DC (1990) Effects of H2O onthe disequilibrium breakdown of muscovite +quartz J Petrol 31 925-956

Butterworth B amp Honeyborne DB (1952) Bricksand clays of the hasting beds Trans Brit CeramSoc 51 211-259

Dondi M Ercolani G Fabbri B Marsigli M (1998)An approach to the chemistry of pyroxenes formedduring the firing of Ca-rich silicate ceramics ClayMinerals 33 443-452

Duminuco P Riccardi MP Messiga B Setti M(1996) Modificazioni tessiturali e mineralogichecome indicatori della dinamica del processo di cot-tura di manufatti ceramici Ceramurgia 26 5 281-288

Evans JL amp White J (1958) Further studies of thethermal decomposition of clays Trans Brit CeramSoc 57 6 298

Everhart JO (1957) Use of auxiliary fluxes to improvestructural clay bodies Bull Am Ceram Soc 36268-271

Carbonate and silicate phase reactions during ceramic firing 633

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

ence of and the interactions between various phas-es that coexist disappear or form are not wellestablished Little is known on the transformationsundergone by silicate and carbonate phases at thereaction interfaces or the phyllosilicate-out reac-tion In particular there is no clear understandingof the transport mechanisms of reactants (ie dif-fusion vs mass transportviscous flow) and theprocesses involved in mineral transformation athigh-T (ie solid-state reactions vs crystallizationfrom a melt) These issues are not solely relevantfor understanding ceramics as there is a close sim-ilarity between ceramic formation and the devel-opment of reaction textures resulting from markeddisequilibrium during pyrometamorphism (ie incontact aureoles or in xenoliths where heatingrates were very rapid Brearley 1986 Brearley ampRubie 1990 Preston et al 1999) They also mayhave important implications in understandingancient ceramic technologies or elucidating rawmaterial sources as well as in the designing ofnew ceramic materials in general or appropriate(ie compatible) conservation materials (iebricks) for architectural heritage conservationinterventions

It is the aim of this work to study the miner-alogical chemical and textural changes of bothresidual minerals and newly formed phases takingplace upon firing and in particular the mecha-nisms of mineral transformation of raw clay-richmaterials with and without carbonates

2 Materials and methods

Two Pliocene clay-rich materials were selectedone from Guadix (G) and the other form Viznar(V) two villages located in the vicinity of GranadaSpain Raw material was collected milled andsieved discarding the fraction with grain size gt 15mm Bricks paste was prepared adding 400 cc ofwater to 1000 g of raw material G and V brickswere fashioned using a wooden mould 245 acute 115acute 4 cm in size and fired in an air-ventilated electricoven (Hoerotec CR-35) at the following tempera-tures 700 800 900 1000 and 1100degC T wasraised at a heating rate of 3degC per minute first upto 100degC with one hour soaking time and later upto the peak T with a soaking time of three hoursFired brick samples were stored at constant T andrelative humidity (RH) conditions of 21degC and 55 respectively

Separation of the fractions lt 2 microm 2 to 20 andgt 20 microm was performed using a KUBOTA 2000centrifuge Grain size distribution in the lt 20 microm

fraction was determined using a laser-beam particlesize analyzer (GALAIh CIS-1) Grain size distribu-tion of gt 20 microm fraction was determined usingstandard ASTM sieves (50 microm up to 1 mm mesh f)

The mineralogy of the raw material as well asthe mineralogical changes taking place upon firingwere studied by powder X-ray diffraction (XRD)using a Philips PW-1710 diffractometer with auto-matic slit CuKa radiation (l = 15405 ) 3 to60deg2q explored area and 001deg2q s-1 goniometerspeed At least three samples (~ 1 g each) of eachbrick typefiring T were analyzed They were milledin agate mortar to lt 40 microm particle size XRD anal-ysis of the clay fraction (ie fraction with grain sizelt 2 microm) was performed using oriented aggregates(air-dried ethylene-glycol and dimethyl sulfoxidesolvated and 1 h heated at 550degC)

The bricks texture and microstructure as wellas the progress of mineral transformation andreactions upon firing were studied by means ofoptical microscopy (OM) and scanning electronmicroscopy (SEM Zeiss DMS 950) coupled withEDX microanalysis Two thin sections per sam-ple type and firing T were prepared A polishedthin section was prepared using one of themBoth SEM secondary electron (SE) and back-scattered electron (BSE) images were acquiredusing either small brick pieces (5 acute 5 acute 10 mm insize gold coated) or polished thin sections (car-bon coated) The same thin sections were used toanalyze small-scale compositional changes ofselected minerals following firing by means of anelectron microprobe (EMPA Cameca SX 50)The EMPA working conditions were 20 keVbeam energy 07 mA filament current and 2 micromspot-size diameter 14 analyses of muscovitecrystals (4700degC 6800degC and 41100degC) and25 of carbonates (6700degC 9800degC and101100degC) were performed Albite orthoclasepericlase wollastonite and oxides (Al2O3 Fe2O3and MnTiO2) were used as standards(Govindaraju 1989) Detection limit for majorelements after ZAF correction (Scott amp Love1983) was 001 wt

Bulk chemical analyses of ceramic materialsbefore and after firing at each target T were per-formed by means of X-ray fluorescence (XRFPhilips PW-1480) 1 g per raw material or firedbrick sample was finely ground and well mixedin agate mortar before being pressed into Al hold-er for disk preparation ZAF correction was per-formed systematically (Scott amp Love 1983)International standards (Govindaraju 1989) wereused thoroughly The estimated detection limitfor major elements was 001 wt

622

3 Results

31 Granulometry

V samples show a higher concentration of par-ticles lt 2 microm when compared with G samples(Fig 1) However the differences are not signifi-cant Both samples show a maximum in particlesize around 100 microm Carbonates are the mostabundant phase in the larger size fraction in Vsamples while schist pieces (with Ms plus Qtz andFs mineral symbols after Kretz (1983)) made upthe larger size fraction of G samples

Carbonate and silicate phase reactions during ceramic firing 623

Fig 2 Optical microscopy micrographs of a) calcite (Cal) grain fired at 800degC showing cracks which penetrates thebrick matrix (plane light) Reaction rims are observed at the carbonate-phyllosilicate (matrix) and carbonate-quartz(Qtz) interfaces and b) muscovite (Ms) crystal before (cross polars) and c) after heating up to 800degC (plane light)Mullite (Mul) observed in the latter

Fig 1 Grain-size distribution curves for Guadix andViznar raw material

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

32 Optical microscopy analysis

G and V samples fired at 700degC do not showsignificant mineralogical transformations Thetemper grain-size (ie phenocrysts) reaches up to15 mm No major textural changes are visible inthe carbonates at 700degC Phyllosilicates in thepaste are oriented due to compression duringmolding At 800degC carbonate decomposition in Vsamples is almost complete and some cracksappear surrounding the former carbonate grains(Fig 2a) The matrix in both G and V samplesturns dark with low birefringence as T increasesMuscovite crystals in both G and V samples startto transform into mullite at 800degC and some bub-bles develop within these crystals at higher T (Fig2b and 2c) Muscovite replacement by mullite iscompleted at 1000degC Mullite preserves the mor-phology of replaced muscovite In fact mullite(001) planes are ^ to (001) planes of former mus-covite grains as deduced by the optical axes ori-entation study At 1000-1100degC the matrix is fullyvitrified and brown colored (in both G and V sam-ples) Phyllosilicates in the matrix do not exist atthese T only some pseudomorphs after formercrystals can be observed An empty space withscattered grayish fragments at the pore edges canbe found where calcite was originally presentCracks observed in V samples at T lt 900degC disap-pear at 1100degC due to extended vitrification Poresbecome more abundant and they are no longerellipsoidal but spherical However an orientationof the remaining temper grains in planes parallel tothe brick largest faces (ie where compressionwas applied during molding) remains

33 XRD analysis

Raw G samples show significant amounts ofquartz and phyllosilicates with little feldspar Thelt 2 microm fraction includes illite and smectite withsmall amounts of kaolinite and paragonite (Table

1) There is no chlorite which could have beenmasked by the d(001) peaks of smectite (~ 14 Aring)and kaolinite (71 Aring) as evidenced by these peaksdisappearing after 90acute heating at 550degC Raw Vsamples have significant amounts of calcite anddolomite The clay fraction includes illite andsmectite with trace amounts of paragonite andkaolinite (Table 1) Therefore the main differenceamong the two raw materials lays in the miner-alogical composition of the gt 2 microm fraction

Regarding the phyllosilicates evolution uponfiring only the diffraction peak at 10 Aring corre-sponding to a dehydroxylated illite-like phaseremains at 700degC This peak intensity is reducedupon firing at higher T till it disappears at ~900degC (Fig 3a and 3b) Microcline main diffrac-tion peak reduces its intensity in G samples Thismineral disappears at T gt 1000degC when traceamounts of sanidine are formed Mullite is detect-ed in both G and V samples at T ~ 900degC increas-ing its main diffraction peak (d(210) = 339 Aring)intensity at higher T Mullite diffraction peaks aremore intense in phyllosilicate-rich G samples thanin carbonate-rich V samples Other phases under-go a significant increase in their diffraction peaksintensities this is the case of hematites in G sam-ples fired at T gt 1000degC New phases appear in Vsamples upon firing a) gehlenite (a melilite-groupphase) appears at 800degC increasing its mainBragg peak intensity at 900degC and reducing it athigher T b) wollastonite and diopside appear at1000degC and they show a significant increase intheir main diffraction peak at higher TDecomposition of calcite and dolomite begins at Tgt 700degC and both phases disappear at T gt 800degCQuartz remains as the most abundant phase at anyT in both V and G samples In V samples the pla-gioclase increases its Ca content (ie a more An-rich phase) as well as its main Bragg peakintensity as T increases This increase is more sig-nificant at T gt 1000degC The existence of an amor-phous phase (ie vitreous phase) is evidenced by

624

Table 1 Results of XRD analysis of the raw materials and the fraction with lt 2 m m grain size Mineral symbols afterKretz (1983) (in Table 3)

a rise of the background noise in the XRD patternat T gt 900degC in the G samples and at T gt 800degCin the V samples

34 SEM-EDX analysis

Significant textural changes most evident withrespect to pore morphology and volume weredetected upon firing An overall porosity reductionis observed at T gt 1000degC when the vitreous phasefills the pores

Samples fired at 700 and 800degC still preservethe laminar habit of phyllosilicates although mus-covite crystals clearly exfoliate along basal planes

most probably due to dehydroxylation (Fig 4a)The interlocking among particles is limited (Fig4d) V samples show some cracks At these T noclear evidence of sintering or partial melting isdetected However there is indirect evidence (ieincrease in background noise of XRD patterns)that some vitrification is reached at T lt 900degC incarbonate-rich V samples Vitrification is clearlyobserved in all samples at T gt 900degC ie phyl-losilicates appear deformed (their edges turnsmooth) andor aggregated (Fig 4b and 4e) andthe pores turn ellipsoidal with smooth surfaces AtT gt 1000degC these effects are less extended in thecarbonate-rich V samples (Fig 4c) than in the sil-

Carbonate and silicate phase reactions during ceramic firing 625

Fig 3 Guadix (a) and Viznar (b) samples powder X-ray difraction patterns Legend (mineral symbols after Kretz1983) Sm = smectite Ill = illite Pg = paragonite Qtz = quartz Cal = calcite Kln = kaolinite Phy = phyllosilicatesFs = feldspar Dol = dolomite Gh = gehlenite Hem = hematite Wo = wollastonite Di = diopside An = anorthiteMul = mullite Sa = sanidine CuKa X-ray radiation l = 15064 Aring

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

icate-rich G samples (Fig 4f) A higher degree ofparticle interlocking results in a porosity reduc-tion At 1100degC vitrification is significant in allsamples regardless of initial compositional differ-ences Pores coalesce as spherical cells due to par-tial or extended melting of clay particles in thematrix

BSE images of mineral and textural evolutionof samples with high carbonate content (ie Vsamples which undergo the most significant min-eralogical changes) give an interesting picture oflocal phase transformations taking place at grainboundaries No textural or mineralogical changescan be detected in brick fired at 700degC (Fig 5a)Carbonates comprise the larger-size fraction inthese samples which allows easy detection ofhigh-T reaction rims In fact at 800degC 2-5 micromthick reaction rims are observed developing at thecalcite-quartzphyllosilicate interface (Fig 5b)Rims composition corresponds to a calcium sili-cate phase presumably wollastonite (EDXresults) A dark rim surrounded by a light outershell is observed where dolomite was originallypresent (Fig 5c) However no compositional vari-ation among internal (greyish) and external(white) rims is detected We were not able tounambiguously detect the presence of gehlenite

within these reaction rims (ie some EDX analy-ses of 2-5 microm thick rims at the carbonate-silicateinterface showed Al together with Ca and Si butpossible contamination from underlying phyllosil-icates was not discarded) However the presenceof gehlenite was detected by XRD at this T(800degC) Muscovite phenocrysts (Fig 5d) show amorphology and composition typical for unalteredwhite mica At 1100degC muscovite crystals showsecondary bubbles development between layers(Fig 5e) The mica composition is homogeneous(no BSE image gray-level differences along thecrystals) and it is similar to that of an unchangedmuscovite but with slightly lower K concentra-tion At 1100degC carbonates (ie dolomite in Fig5f) shows an outer thick white Ca- and Si-rich rim(ie wollastonite) which includes small (micronsized) bright-white Ca- Si- and Al-rich particles(ie gehlenite or plagioclase) There is a strikingfeature of this reaction rim it shows a fingeredgeometry at the outer edge The darker dolomitecore shows a significant Mg enrichment

35 EMPA analysis

Due to the fact that only silicate phases in sam-ples with carbonates undergo mayor composition-

626

Fig 4 SEM secondary electron photomicrographs of Viznar samples fired at 800degC (a) 1000degC (b) and 1100degC(c) and Guadix samples fired at 700degC (d) 900degC (e) and 1000degC (f)

al changes upon firing only V samples were ana-lyzed using this technique In particular composi-tional changes of phyllosilicates and carbonateswere studied using polished thin sections of firedV bricks (one section per selected target T ie700 800 1100degC)

Phyllosilicates (Table 2) show a muscoviticcomposition at 700degC where some K is replaced byNa even though the overall composition does notseem to correspond to paragonite Neverthelessparagonite was detected in the raw clay (Fig 3b)

Fe Mg and Ti replace some Al in the octahedralsheets Replacement of OH- by F- is almost absentAt 800degC a slight increase in Fe content is detectedOnly analyses M9 and M10 show higher concen-trations of Na and Al At 1100degC the amount of Kis reduced while Ca and Si content is increasedThe total oxide content is close to 100 a valueconsistent with complete dehydroxylation of thephyllosilicates Analyses M11-M14 in Table 2 arein agreement with high-T phengite compositionaldata from Worden et al (1987)

Carbonate and silicate phase reactions during ceramic firing 627

Fig 5 BSE images and EDX analyses of carbonate-rich Viznar samples a) fired at 700degC b) fired at 800degC whenWo (EDX analysis) formed at the Cal- silicates interface c) detail of former carbonate grain showing a white outerreaction rim (Wo or Gh) a crack and an inner core with an outer darker shell (see text for details) d) Ms (see EDXanalysis) at 700degC e) former Ms after transformation into a Fs + Mul + melt mixture (see EDX analysis) f) DolPhy(+ Qtz) interface showing a reaction rim with fingered geometry formed at 1100degC g) detail of the outer reactionrim in previous figure EDX analysis shows the rim composition (ie Wo)

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

Calcite (Table 3) shows no chemical changesat 700degC The maximum MgO content is 104 wt Trace amounts of Fe Si Al and Mn are alsodetected in calcite (as well as in dolomite) At800degC both carbonates are fully transformed eitherinto burnt lime (CaO) or into a mixture of burtlime and periclase (MgO) (eg analysis C14)Gehlenite-like compositions are detected in someanalyses (eg C10) At 1100degC the presence ofnewly formed phases in particular wollastonite isclearly confirmed (eg C17) Ca content in thelime plus periclase mixture is reduced significant-ly due to the incorporation of this element intonewly formed high-T silicate phases As T increas-es a clear compositional trend toward Si- and Al-rich calcium phases (ie toward gehlenite andanorthite) is observed (Fig 6a) a fact consistentwith XRD and SEM-EDX results A clear trendtoward wollastonite-like compositions as well as a

trend toward MgO rich composition is alsoobserved as T increases (Fig 6b) Some EMPAanalyses suggest a larnite -like compositionHowever XRD results do not show the presenceof larnite

36 XRF analysis

According to the classification proposed forclay materials in ceramic industry by Fabbri ampFiori (1985) G raw material composition falls intothe quartz-feldspar sands category while V sam-ples fall out of any defined category due to theirlow silica and high calcium contents (Table 4) Asexpected firing does not affect major elementconcentration in the brick The only significantchange is detected in the loss on ignition values Itis evidenced that G samples lost almost all waterbetween 700 and 900degC while V samples show a

628

Table 2 EMPA analysis results of phyllosilicates composition (in wt )

Table 3 Composition of selected carbonate grains ( EMPA results in wt )

more gradual loss of water and CO2 as T increases(Table 4)

4 Discussion

Firing up to 700degC induce no significant min-eralogicaltextural transformations in both thecarbonate-rich V and the silicate-rich G sampleswith the exception of the clay minerals Clay min-erals structure is reported to collapse due to dehy-droxylation to resemble an illite-like structure(XRD data) when a T between 450 and 550degC isreached (Evans amp White 1958) A dehydroxylatedphyllosilicate phase structurally different to thehydrated one (Guggenheim et al 1987) is report-ed to exist up to 950degC when complete breakdownof the dehydroxylated illite occurs (Peters amp Iberg1978) Our data indicate that dehydroxylation isnot completed at T ~ 900degC which indicates thatthe kinetics of this process is slower than previ-ously estimated (Evans amp White 1958) (OH)-

released at T gt 700degC may play a significant rolein texture and mineral evolution at high-T as willbe discussed later On the other hand muscoviteshows a slight K deficit at 700degC probably due tothe high diffusion capacity of this element alongmuscovite (001) basal planes at T gt 500degC(Sanchez-Navas amp Galindo-Zaldivar 1993)Sanchez-Navas (1999) showed high-T K-diffu-sion-out of mica is induced by water adsorbedalong mica basal planes resulting from phyllosili-cate dehydroxylation

XRD data shows that dolomite most intenseBragg peak in V bricks significantly reduced itsintensity due to thermal decomposition at lower Tthan calcite main Bragg peak (Fig 3b) Dolomite

Carbonate and silicate phase reactions during ceramic firing 629

Table 4 Bulk composition of Guadix (G) and Viznar (V) samples (XRF results in wt )

LOI loss on ignition

Fig 6 a) ACS (Al2O3-CaO-SiO2) and b) CMS (CaO-MgO-SiO2) compositional diagrams of carbonates andreaction products (EMPA results)

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

starts to decompose according to (1) at ~ 700degC (1atm pressure) while calcite decomposes at higherT (830 ndash 870degC) according to (3) (Boynton 1980)Shoval (1988) proposed a two steps thermaldecomposition for dolomite involving formationof periclase and calcite (2) followed by calcitedecomposition at higher T (3)

dolomite lime periclaseCaMg(CO3)2 reg CaO + MgO + 2CO2 (1)

calcite CaMg(CO3)2 reg CaCO3 + MgO + CO2 (2)

CaCO3 reg CaO + CO2 (3)

However our XRD results showing no increase incalcite 104 Bragg peak intensity as dolomitedecomposes at 700degC (Fig 3b) are not consistentwith Shovalacutes two-steps process

Carbonate decomposition results in shrinkageof the brick (Barahona et al 1985) Upon storageat 22degC and 55 RH burnt lime (CaO) readilytransforms into portlandite according to the fol-lowing reaction (Boynton 1980)

portlanditeCaO + H2O reg Ca(OH)2 (4)

This process generates crystallization pressure inconfined spaces (ie the brick pores originallyoccupied by CaO) resulting in crack development(as observed in Fig 2a) The dark-color rimaround former carbonate grains (Fig 5c) is proba-bly due to limited portlandite carbonation in thepresence of CO2 also resulting in volume increase(Moorehead 1985) and crack development(Butterworth amp Honeyborne 1952) As a commonpractice to avoid these undesired side-effectsbricks that contain carbonates were traditionallyimmersed in water to dissolve and eliminate exist-ing CaO or Ca(OH)2

This effect is less evident in the case of MgOPericlase in the presence of H2O (room T) under-goes a slow transformation (it takes months oreven years Webb 1952) into brucite (Mg(OH)2)and may eventually transform into hydromagne-site (Mg5(CO3)4(OH)24H2O) according to thefollowing reactions (Garavelli et al 1990)

MgO + H2O reg Mg(OH)2 (5)

hydromagnesite5Mg(OH)2 + 4CO2 reg Mg5(CO3)4(OH)24H2O (6)

EMPA results indicate the presence of Ca and Mgoxides however our XRD results do not clearlyshow the presence of Ca and Mg oxides or hydrox-ides mainly due to their low concentrations (lt 10wt ) and because their main diffraction peaks aremasked by other major phase peaks (eg quartzfeldspars hematites) The hydromagnesite mainBragg peak at 581 Aring is not present either

At T sup3 800degC and in the absence of carbonateschanges are constrained to textural modificationsand reactions where a low-T phase is decomposedand fully transformed into a high-T phase withoutmajor compositional changes (Riccardi et al1999) In particular muscovite undergoes a solid-state phase change into a mixture of mullite and K-feldspar (or a melt) at T ranging from 800 to1000degC The composition of high-T muscovite isclose to K-feldspar (EMPA results) although OManalysis points to the presence of mullite Brearley(1986) and Worden et al (1987) showed fine inter-growth of alternating nanometer up to a fewmicrometers thick bands of K-feldspar (or a meltwith composition close to that of K-feldspar) andmullite in pyrometamorphic muscovite which canexplain our EMPA results

The most significant textural and mineralogi-cal changes are observed in samples with carbon-ates when fired at T gt 800degC Also melting T andextent seems to be connected with the presence orabsence of carbonates In general melting starts atlower T (~ 800degC) when carbonates are present(Tite amp Maniatis 1975) Ca and Mg from carbon-ates may act as melting agents (Segnit ampAnderson 1972) but they are reported to some-how limit the extent of vitrification at T gt 1000degC(Everhart 1957 Nuacutentildeez et al 1992) In fact weobserved that vitrification is more extended athigher T when no carbonates exist This may beexplained considering that the phyllosilicate in thematrix of non-carbonate G samples not contribut-ing to the formation of high-T Ca (or Mg) silicatesand releasing significant amounts of H2O due todehydroxylation they can contribute to extensivemelting at high-T as demonstrated by Brearley ampRubie (1990) On the other hand the higher SiO2content in these samples provides a higher amountof potential silicate-rich melt

BSE images of reaction rims between carbon-ates and silicates show that formation of Ca-sili-cates starts at 800degC This is confirmed by XRD(Fig 3b) and EMPA analyses (Table 3) Gehlenitestarts to form at this T by grain-boundary reactionbetween CaO Al2O3 and SiO2 the first derivingfrom former carbonates and the later two fromalready dehydroxylated phyllosilicates (ie a

630

somehow amorphous mixture of 3SiO2Al2O3according to Peters amp Iberg 1978) as follows

illite calcite gehleniteKAl2(Si3Al)O10(OH)2 + 6CaCO3 reg 3Ca2Al2SiO7+ 6CO2 + 2H2O + K2O + 3SiO2 (7)

The formation of gehlenite is striking becauseone could expect the formation of an Al-richpyroxene such as fassaite (Dondi et al 1998)Nevertheless it should be indicated that gehleniteformation is a good example of the so-calledGoldsmithacutes ldquoease of crystallizationrdquo principlewhich states that a silicate phase with all Al in 4-coordination is easier to form (at least in the lab)than a phase that includes Al in both 4- and 6-coor-dination when (OH)- are not present (Goldsmith1953) This is due to the higher energy barrier nec-essary to overcome when Al has to enter in a 6-coordination if compared with a 4-coordination If(OH)- are present this energy barrier is reduceddue to the smaller electrostatic repulsion among(OH)- groups than among O2- in octahedral coor-dination Gehlenite has all Al in tetrahedral coor-dination (Warren 1930) therefore crystallizingpreferentially if compared with other anhydrousCa-Al silicates This simple principle also appliesto the high-T formation of other anhydrous Al-sil-icates as we will discuss later

Wollastonite formed at the carbonate-quartzinterface through the following reaction

wollastoniteCaO + SiO2 reg CaSiO3 (8)

Wollastonite and gehlenite appear at 800degC aT ~100degC lower than previously reported (Petersamp Iberg 1978 Dondi et al 1998) This is proba-bly due to a) the low resolution of the analyticaltechniques used in the past (ie XRD) which didnot allow the detection of the small initial reactionrim between carbonates and silicates as it is evi-denced by BSE images and EMPA analysesandor b) most previous works on high-T phaseformation in ceramic processing use single com-ponents as reactants without considering addition-al phases or previous reaction by-products (Petersamp Iberg 1978) We observed that in a closer-to-reality multicomponent system where interactionsbetween reactants and products occur differentformation T are obtained

At the dolomite-quartz interface diopsidestarts to form at 900degC by the following reaction

dolomite diopside2SiO2 + CaMg(CO3)2 reg CaMgSi2O6 + 2CO (9)

At this T anorthite starts to form at the expensesof calcite and illite (plus quartz)

illite calciteKAl2(Si3Al)O10(OH)2 + 2CaCO3 + 4SiO2 regK-feldspar anorthite

2KAlSi3O8 + 2Ca2Al2SiO8 + 2CO2 + H2O (10)

At higher temperatures (1000 to 1100degC)phyllosilicates have already disappeared in allsamples (V and G) being transformed into sani-dine and mullite plus a melt according to the fol-lowing reaction

illite (muscovite) sanidine 3KAl2(Si3Al)O10(OH)2 + 2SiO2 reg 3KAlSi3O8 +

mulliteAl6Si2O13 + 3H2O + melt (11)

Formation of a melt during the previous reaction(Brearley amp Rubie 1990) is consistent with theobserved ldquocellular structurerdquo of muscovite (Tite ampManiatis 1975)

Brearley (1986) proposed a somehow differentreaction for mullite formation at the expenses ofphengitic muscovite resulting in K-feldspar +mullite + biotite No biotite is detected in ourexperiments most probably due to the low Fe con-tent in V and G micas Brearley amp Rubie (1990)experimentally demonstrated that high-T mus-covite breakdown in the presence of quartz isinfluenced by the crystal size The smallest crys-tals have a strong tendency to react with H2Oresulting in enhanced melt formation while largemuscovite crystals form limited amounts of meltThis may account for the significant vitrificationof the matrix (specially in G samples) while largemuscovite grains (up to some mm in size) still donot show significant melting

Mullite which according to OM analysis firstappears at 800degC is the second most abundantphase (after quartz) at 1100degC in the richer inphyllosilicate G samples It is interesting to men-tion that K-feldspar progressively disappears as Tincreases This is not consistent with publisheddata (Maggetti 1982) It seems that low-T K-feldspar (microcline) is not stable at high-T there-fore it transforms into a high-T K-feldspar (iesanidine) or reacts with lime to form anorthiteBoth sanidine and anorthite are detected in thehigh-T bricks However a higher concentration ofsanidine should be expected to form speciallyconsidering reaction (11) A possible explanationfor the low K-feldspar content at high-T is that K-feldspar might not have enough time to develop itscrystalline structure (sanidine) therefore being

Carbonate and silicate phase reactions during ceramic firing 631

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

concentrated as an amorphous (sanidine-like)phase responsible in part for the rise in the back-ground noise in the XRD results Mullite forma-tion seems to be kinetically faster resulting in awell-developed crystalline phase which givessharp diffraction peaks Goldsmith (1953) indi-cates that mullite could be another example of hisldquosimplexity principlerdquo for the ldquoeaserdquo of crystal-lization if compared with other aluminum sili-cates In fact the mullite structure has a largeramount of Al in 4-coordination and less Al in 6-coordination than sillimanite (Rehak et al 1998)for instance which may account for the abundanceof the former and the non-existence of the latter inthe studied bricks

In carbonate-rich V samples fired at T gt1000degC a significant reduction of gehlenite maindiffraction peak is observed According to Petersamp Iberg (1978) this is due to the reaction of thisphase with quartz forming wollastonite and anor-thite This is consistent with the increase in themain Bragg peak intensity of these later mineralsat 1100degC

At the dolomite-silicate (ie Qtz and Phy)interface it is common to observe wollastoniteeither newly formed or resulting from the destruc-tion of gehlenite The texture of the interface israther complex Protrusion or fingers from thereaction zone penetrate into areas were completetransformation into wollastonite has occurred Theinterface geometry resembles the so-called ldquovis-cous fingersrdquo (Garcia-Ruiz 1992) occurringwhen two fluids or semi-plastic solids of differentcomposition and viscosity are placed in contactInitially the contact zone is a flat surface but lateron it reaches a fingered fractal geometry(Nittmamm et al 1985) In the carbonate-silicateinterface a similar process may have taken placewhere mass transport of the reactants may havebeen enhanced by partial melting The formationof a silicate-rich melt due to clay-minerals melting(Brearley amp Rubie 1990) in the external part ofthe ring in contact with a second inner melt formedby preexisting gehlenite or the eutectic CaO-SiO2would result in viscous finger development Thefractal geometry of the developing viscous fingersmay have enhanced diffusion of mobile ions suchas Ca toward silicate reactive sites Both EDXdata showing Mg enrichment in the dolomite core(Fig 5f) and the compositional trend towardsMgO-rich terms in the CMS diagram (Fig 6b)confirm that Ca is preferentially leached out of theformer dolomite grain Finger formation willresult in a significant increase in interface surfacearea resulting in higher reaction rates Another

alternative explanation for this fingered geometrydevelopment involves partial melting at the car-bonate-silicate interface and mass transport (ieviscous flow) through the clay-quartz porousmatrix resulting in the so-called ldquoscalloped struc-turerdquo proposed by Ortoleva et al (1987) for reac-tion-diffusion processes While the viscous fingermodel requires two melts the second model onlyrequires one melt It is not clear if at any momentconditions for two melt occurrence are reachedHence the second model seems more plausible

In any case the process which is originally dif-fusion driven is transformed into a mass (ie fin-gering) plus diffusion transport which greatlyenhances Ca-silicate formation This is consistentwith the thick reaction rims (up to 250 microm thicksee Fig 5f) that surround former carbonate grainsThese thick Ca-silicate shells would hardly devel-op just by diffusion in the short time-scale of theexperiment (3 hours at the peak T) In fact at800degC (conditions where melting would be verylimited) Ca (and Mg)-silicate rims formed aroundcalcite or dolomite are 2-5 microm thick At 1100degCwhen melting is extensive the reaction rate is sohigh that rims are two orders of magnitude thick-er Therefore mass transport (ie viscous flow)seems to have played a key role in high-T Ca-(Mg-) silicate formation Mass transport is also consis-tent with observation of some areas with differentcomposition (unreacted silicates) trapped betweenfingers resulting in the island geometry observedin Fig 5g It could be expected that without fin-gering the process of Ca- or (Ca-Mg)-silicate for-mation would be self-limiting ie once a rim ofthe newly formed silicate appears diffusion wouldbe jeopardized This shell would preclude or limitthe transport of Ca (andor Mg) from the carbon-ate or it would limit the transport of the lessmobile Si (and Al) from the surrounding silicatetowards the reaction front Therefore it is neces-sary that one of the above described transport-reaction mechanisms takes place This may alsoexplain why phase formation takes place at lowerT than reported previously The presence of multi-ple interfaces among different crystals mayenhance finger development resulting in kineticchanges favoring phase formation at lower T

5 Conclusions

It is observed that even though the overallcomposition of a ceramic material does not under-go major changes upon firing significant miner-alogical and textural modifications do occur

632

These changes are connected with two reactiontypes a) solid-state replacement of a phase by anew one with little or no compositional variation(ie the muscovite and illite case) and minimizingenergy changes by preserving the original textureand orientation of the replaced phase and b) reac-tion-diffusion at the interface between two miner-al grains with contrasting composition (iecarbonate-silicate interfaces) The latter reaction isenhanced if massviscous transport resulting frompartial melting and finger structure developmentoccurs Non-carbonated clay-rich materials (G)show phyllosilicate destruction at T ranging from700 to 900degC followed by vitrification which issignificant at T gt 1000degC Carbonate-rich materi-als (V) show major phase transformation uponphyllosillicates destruction including new Ca (orMg) silicates and aluminum-silicates formation Inthis later case the vitrification starts at lower Tand it is limited at higher T Therefore it is possi-ble to establish the firing T of ceramic materialswith composition broadly similar to that in ourwork based on the mineralogy Thus a) presenceof phyllosilicates indicates a low 700 to 900degC fir-ing T for the dehydroxylated illite (or muscovite)phase b) dolomite disappears at sup3 700degC whilecalcite is still present at 800degC c) gehleniteappears at 800degC reducing its concentration above1000degC d) diopside and wollastonite appear at800degC (in very low concentrations) as reactionrims between carbonates and silicates Formationof these minerals includes ldquofingeringrdquo develop-ment e) mullite appears at T gt 800degC while littleamounts of sanidine form as microcline disap-pears and f) anorthite develops at T gt1000degC incarbonate-rich bricks

When considering equilibrium thermodynamicdata and calculated reaction T for most of theabove mentioned phases and reactions (Holland ampPowell 1985 and references herein) it isobserved that rapid heating induces significantoverstepping Therefore phases formed are inmost cases metastable a fact consistent with anal-yses of pyrometamorphic assemblages (Brearleyamp Rubie 1990 and references herein)

The formation of non-equilibrium fractal mor-phologies at the carbonate-silicate reaction inter-faces may explain why high-T silicates incarbonate-rich ceramics develop very fast Thisgrain-to-grain reaction model may help explainhigh-T silicate formation in nature (eg contactmetamorphism pyrometamorphism in xenoliths)

The gained knowledge on mineral transforma-tion and the dynamics of interface mass and diffu-sion transport facilitated through ldquofingeringrdquo or

ldquoscallopingrdquo opens new ways to understand thecomplexities of ceramic processing in particularand high-T mineral transformation in general bothin natural (ie geologic settings) and artificialcontexts (eg archaeometric analysis of ancientceramics or the design of new bricks ndashreplicas ofancient onesndash for architectural heritage conserva-tion) However care must be taken when extrapo-lating these results to other raw materialcompositions andor ceramic processing methods

Acknowledgements This work has been finan-cially supported by the CGICYT projectMAT2000-1457 and the Research Group RNM0179 of the Junta de Andaluciacutea (Spain) We thankthe Centro de Instrumentacioacuten Cientifica (CIC) ofthe Granada University for allowing us to use theirSEM-EDX EMPA and XRF facilities We thankK Elert JM Garciacutea-Ruiz M Rodriguez-Gallego and A Sanchez-Navas for fruitful discus-sions and comments We also thank the in-depthreview by Dr F Goumltz-Neunhoeffer and an anony-mous referee

References

Barahona E Huertas F Pozzuoli A Linares J(1985) Firing properties of ceramic clays fromGranada province Spain Miner Petrogr Acta 29A577-590

Boynton RS (1980) Chemistry and technology oflime and limestone 2nd Ed Wiley New York

Brearley AJ (1986) An electron optical study of mus-covite breakdown in pelitic xenoliths duringpyrometamorphism Miner Mag 357 385-397

Brearley AJ amp Rubie DC (1990) Effects of H2O onthe disequilibrium breakdown of muscovite +quartz J Petrol 31 925-956

Butterworth B amp Honeyborne DB (1952) Bricksand clays of the hasting beds Trans Brit CeramSoc 51 211-259

Dondi M Ercolani G Fabbri B Marsigli M (1998)An approach to the chemistry of pyroxenes formedduring the firing of Ca-rich silicate ceramics ClayMinerals 33 443-452

Duminuco P Riccardi MP Messiga B Setti M(1996) Modificazioni tessiturali e mineralogichecome indicatori della dinamica del processo di cot-tura di manufatti ceramici Ceramurgia 26 5 281-288

Evans JL amp White J (1958) Further studies of thethermal decomposition of clays Trans Brit CeramSoc 57 6 298

Everhart JO (1957) Use of auxiliary fluxes to improvestructural clay bodies Bull Am Ceram Soc 36268-271

Carbonate and silicate phase reactions during ceramic firing 633

3 Results

31 Granulometry

V samples show a higher concentration of par-ticles lt 2 microm when compared with G samples(Fig 1) However the differences are not signifi-cant Both samples show a maximum in particlesize around 100 microm Carbonates are the mostabundant phase in the larger size fraction in Vsamples while schist pieces (with Ms plus Qtz andFs mineral symbols after Kretz (1983)) made upthe larger size fraction of G samples

Carbonate and silicate phase reactions during ceramic firing 623

Fig 2 Optical microscopy micrographs of a) calcite (Cal) grain fired at 800degC showing cracks which penetrates thebrick matrix (plane light) Reaction rims are observed at the carbonate-phyllosilicate (matrix) and carbonate-quartz(Qtz) interfaces and b) muscovite (Ms) crystal before (cross polars) and c) after heating up to 800degC (plane light)Mullite (Mul) observed in the latter

Fig 1 Grain-size distribution curves for Guadix andViznar raw material

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

32 Optical microscopy analysis

G and V samples fired at 700degC do not showsignificant mineralogical transformations Thetemper grain-size (ie phenocrysts) reaches up to15 mm No major textural changes are visible inthe carbonates at 700degC Phyllosilicates in thepaste are oriented due to compression duringmolding At 800degC carbonate decomposition in Vsamples is almost complete and some cracksappear surrounding the former carbonate grains(Fig 2a) The matrix in both G and V samplesturns dark with low birefringence as T increasesMuscovite crystals in both G and V samples startto transform into mullite at 800degC and some bub-bles develop within these crystals at higher T (Fig2b and 2c) Muscovite replacement by mullite iscompleted at 1000degC Mullite preserves the mor-phology of replaced muscovite In fact mullite(001) planes are ^ to (001) planes of former mus-covite grains as deduced by the optical axes ori-entation study At 1000-1100degC the matrix is fullyvitrified and brown colored (in both G and V sam-ples) Phyllosilicates in the matrix do not exist atthese T only some pseudomorphs after formercrystals can be observed An empty space withscattered grayish fragments at the pore edges canbe found where calcite was originally presentCracks observed in V samples at T lt 900degC disap-pear at 1100degC due to extended vitrification Poresbecome more abundant and they are no longerellipsoidal but spherical However an orientationof the remaining temper grains in planes parallel tothe brick largest faces (ie where compressionwas applied during molding) remains

33 XRD analysis

Raw G samples show significant amounts ofquartz and phyllosilicates with little feldspar Thelt 2 microm fraction includes illite and smectite withsmall amounts of kaolinite and paragonite (Table

1) There is no chlorite which could have beenmasked by the d(001) peaks of smectite (~ 14 Aring)and kaolinite (71 Aring) as evidenced by these peaksdisappearing after 90acute heating at 550degC Raw Vsamples have significant amounts of calcite anddolomite The clay fraction includes illite andsmectite with trace amounts of paragonite andkaolinite (Table 1) Therefore the main differenceamong the two raw materials lays in the miner-alogical composition of the gt 2 microm fraction

Regarding the phyllosilicates evolution uponfiring only the diffraction peak at 10 Aring corre-sponding to a dehydroxylated illite-like phaseremains at 700degC This peak intensity is reducedupon firing at higher T till it disappears at ~900degC (Fig 3a and 3b) Microcline main diffrac-tion peak reduces its intensity in G samples Thismineral disappears at T gt 1000degC when traceamounts of sanidine are formed Mullite is detect-ed in both G and V samples at T ~ 900degC increas-ing its main diffraction peak (d(210) = 339 Aring)intensity at higher T Mullite diffraction peaks aremore intense in phyllosilicate-rich G samples thanin carbonate-rich V samples Other phases under-go a significant increase in their diffraction peaksintensities this is the case of hematites in G sam-ples fired at T gt 1000degC New phases appear in Vsamples upon firing a) gehlenite (a melilite-groupphase) appears at 800degC increasing its mainBragg peak intensity at 900degC and reducing it athigher T b) wollastonite and diopside appear at1000degC and they show a significant increase intheir main diffraction peak at higher TDecomposition of calcite and dolomite begins at Tgt 700degC and both phases disappear at T gt 800degCQuartz remains as the most abundant phase at anyT in both V and G samples In V samples the pla-gioclase increases its Ca content (ie a more An-rich phase) as well as its main Bragg peakintensity as T increases This increase is more sig-nificant at T gt 1000degC The existence of an amor-phous phase (ie vitreous phase) is evidenced by

624

Table 1 Results of XRD analysis of the raw materials and the fraction with lt 2 m m grain size Mineral symbols afterKretz (1983) (in Table 3)

a rise of the background noise in the XRD patternat T gt 900degC in the G samples and at T gt 800degCin the V samples

34 SEM-EDX analysis

Significant textural changes most evident withrespect to pore morphology and volume weredetected upon firing An overall porosity reductionis observed at T gt 1000degC when the vitreous phasefills the pores

Samples fired at 700 and 800degC still preservethe laminar habit of phyllosilicates although mus-covite crystals clearly exfoliate along basal planes

most probably due to dehydroxylation (Fig 4a)The interlocking among particles is limited (Fig4d) V samples show some cracks At these T noclear evidence of sintering or partial melting isdetected However there is indirect evidence (ieincrease in background noise of XRD patterns)that some vitrification is reached at T lt 900degC incarbonate-rich V samples Vitrification is clearlyobserved in all samples at T gt 900degC ie phyl-losilicates appear deformed (their edges turnsmooth) andor aggregated (Fig 4b and 4e) andthe pores turn ellipsoidal with smooth surfaces AtT gt 1000degC these effects are less extended in thecarbonate-rich V samples (Fig 4c) than in the sil-

Carbonate and silicate phase reactions during ceramic firing 625

Fig 3 Guadix (a) and Viznar (b) samples powder X-ray difraction patterns Legend (mineral symbols after Kretz1983) Sm = smectite Ill = illite Pg = paragonite Qtz = quartz Cal = calcite Kln = kaolinite Phy = phyllosilicatesFs = feldspar Dol = dolomite Gh = gehlenite Hem = hematite Wo = wollastonite Di = diopside An = anorthiteMul = mullite Sa = sanidine CuKa X-ray radiation l = 15064 Aring

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

icate-rich G samples (Fig 4f) A higher degree ofparticle interlocking results in a porosity reduc-tion At 1100degC vitrification is significant in allsamples regardless of initial compositional differ-ences Pores coalesce as spherical cells due to par-tial or extended melting of clay particles in thematrix

BSE images of mineral and textural evolutionof samples with high carbonate content (ie Vsamples which undergo the most significant min-eralogical changes) give an interesting picture oflocal phase transformations taking place at grainboundaries No textural or mineralogical changescan be detected in brick fired at 700degC (Fig 5a)Carbonates comprise the larger-size fraction inthese samples which allows easy detection ofhigh-T reaction rims In fact at 800degC 2-5 micromthick reaction rims are observed developing at thecalcite-quartzphyllosilicate interface (Fig 5b)Rims composition corresponds to a calcium sili-cate phase presumably wollastonite (EDXresults) A dark rim surrounded by a light outershell is observed where dolomite was originallypresent (Fig 5c) However no compositional vari-ation among internal (greyish) and external(white) rims is detected We were not able tounambiguously detect the presence of gehlenite

within these reaction rims (ie some EDX analy-ses of 2-5 microm thick rims at the carbonate-silicateinterface showed Al together with Ca and Si butpossible contamination from underlying phyllosil-icates was not discarded) However the presenceof gehlenite was detected by XRD at this T(800degC) Muscovite phenocrysts (Fig 5d) show amorphology and composition typical for unalteredwhite mica At 1100degC muscovite crystals showsecondary bubbles development between layers(Fig 5e) The mica composition is homogeneous(no BSE image gray-level differences along thecrystals) and it is similar to that of an unchangedmuscovite but with slightly lower K concentra-tion At 1100degC carbonates (ie dolomite in Fig5f) shows an outer thick white Ca- and Si-rich rim(ie wollastonite) which includes small (micronsized) bright-white Ca- Si- and Al-rich particles(ie gehlenite or plagioclase) There is a strikingfeature of this reaction rim it shows a fingeredgeometry at the outer edge The darker dolomitecore shows a significant Mg enrichment

35 EMPA analysis

Due to the fact that only silicate phases in sam-ples with carbonates undergo mayor composition-

626

Fig 4 SEM secondary electron photomicrographs of Viznar samples fired at 800degC (a) 1000degC (b) and 1100degC(c) and Guadix samples fired at 700degC (d) 900degC (e) and 1000degC (f)

al changes upon firing only V samples were ana-lyzed using this technique In particular composi-tional changes of phyllosilicates and carbonateswere studied using polished thin sections of firedV bricks (one section per selected target T ie700 800 1100degC)

Phyllosilicates (Table 2) show a muscoviticcomposition at 700degC where some K is replaced byNa even though the overall composition does notseem to correspond to paragonite Neverthelessparagonite was detected in the raw clay (Fig 3b)

Fe Mg and Ti replace some Al in the octahedralsheets Replacement of OH- by F- is almost absentAt 800degC a slight increase in Fe content is detectedOnly analyses M9 and M10 show higher concen-trations of Na and Al At 1100degC the amount of Kis reduced while Ca and Si content is increasedThe total oxide content is close to 100 a valueconsistent with complete dehydroxylation of thephyllosilicates Analyses M11-M14 in Table 2 arein agreement with high-T phengite compositionaldata from Worden et al (1987)

Carbonate and silicate phase reactions during ceramic firing 627

Fig 5 BSE images and EDX analyses of carbonate-rich Viznar samples a) fired at 700degC b) fired at 800degC whenWo (EDX analysis) formed at the Cal- silicates interface c) detail of former carbonate grain showing a white outerreaction rim (Wo or Gh) a crack and an inner core with an outer darker shell (see text for details) d) Ms (see EDXanalysis) at 700degC e) former Ms after transformation into a Fs + Mul + melt mixture (see EDX analysis) f) DolPhy(+ Qtz) interface showing a reaction rim with fingered geometry formed at 1100degC g) detail of the outer reactionrim in previous figure EDX analysis shows the rim composition (ie Wo)

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

Calcite (Table 3) shows no chemical changesat 700degC The maximum MgO content is 104 wt Trace amounts of Fe Si Al and Mn are alsodetected in calcite (as well as in dolomite) At800degC both carbonates are fully transformed eitherinto burnt lime (CaO) or into a mixture of burtlime and periclase (MgO) (eg analysis C14)Gehlenite-like compositions are detected in someanalyses (eg C10) At 1100degC the presence ofnewly formed phases in particular wollastonite isclearly confirmed (eg C17) Ca content in thelime plus periclase mixture is reduced significant-ly due to the incorporation of this element intonewly formed high-T silicate phases As T increas-es a clear compositional trend toward Si- and Al-rich calcium phases (ie toward gehlenite andanorthite) is observed (Fig 6a) a fact consistentwith XRD and SEM-EDX results A clear trendtoward wollastonite-like compositions as well as a

trend toward MgO rich composition is alsoobserved as T increases (Fig 6b) Some EMPAanalyses suggest a larnite -like compositionHowever XRD results do not show the presenceof larnite

36 XRF analysis

According to the classification proposed forclay materials in ceramic industry by Fabbri ampFiori (1985) G raw material composition falls intothe quartz-feldspar sands category while V sam-ples fall out of any defined category due to theirlow silica and high calcium contents (Table 4) Asexpected firing does not affect major elementconcentration in the brick The only significantchange is detected in the loss on ignition values Itis evidenced that G samples lost almost all waterbetween 700 and 900degC while V samples show a

628

Table 2 EMPA analysis results of phyllosilicates composition (in wt )

Table 3 Composition of selected carbonate grains ( EMPA results in wt )

more gradual loss of water and CO2 as T increases(Table 4)

4 Discussion

Firing up to 700degC induce no significant min-eralogicaltextural transformations in both thecarbonate-rich V and the silicate-rich G sampleswith the exception of the clay minerals Clay min-erals structure is reported to collapse due to dehy-droxylation to resemble an illite-like structure(XRD data) when a T between 450 and 550degC isreached (Evans amp White 1958) A dehydroxylatedphyllosilicate phase structurally different to thehydrated one (Guggenheim et al 1987) is report-ed to exist up to 950degC when complete breakdownof the dehydroxylated illite occurs (Peters amp Iberg1978) Our data indicate that dehydroxylation isnot completed at T ~ 900degC which indicates thatthe kinetics of this process is slower than previ-ously estimated (Evans amp White 1958) (OH)-

released at T gt 700degC may play a significant rolein texture and mineral evolution at high-T as willbe discussed later On the other hand muscoviteshows a slight K deficit at 700degC probably due tothe high diffusion capacity of this element alongmuscovite (001) basal planes at T gt 500degC(Sanchez-Navas amp Galindo-Zaldivar 1993)Sanchez-Navas (1999) showed high-T K-diffu-sion-out of mica is induced by water adsorbedalong mica basal planes resulting from phyllosili-cate dehydroxylation

XRD data shows that dolomite most intenseBragg peak in V bricks significantly reduced itsintensity due to thermal decomposition at lower Tthan calcite main Bragg peak (Fig 3b) Dolomite

Carbonate and silicate phase reactions during ceramic firing 629

Table 4 Bulk composition of Guadix (G) and Viznar (V) samples (XRF results in wt )

LOI loss on ignition

Fig 6 a) ACS (Al2O3-CaO-SiO2) and b) CMS (CaO-MgO-SiO2) compositional diagrams of carbonates andreaction products (EMPA results)

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

starts to decompose according to (1) at ~ 700degC (1atm pressure) while calcite decomposes at higherT (830 ndash 870degC) according to (3) (Boynton 1980)Shoval (1988) proposed a two steps thermaldecomposition for dolomite involving formationof periclase and calcite (2) followed by calcitedecomposition at higher T (3)

dolomite lime periclaseCaMg(CO3)2 reg CaO + MgO + 2CO2 (1)

calcite CaMg(CO3)2 reg CaCO3 + MgO + CO2 (2)

CaCO3 reg CaO + CO2 (3)

However our XRD results showing no increase incalcite 104 Bragg peak intensity as dolomitedecomposes at 700degC (Fig 3b) are not consistentwith Shovalacutes two-steps process

Carbonate decomposition results in shrinkageof the brick (Barahona et al 1985) Upon storageat 22degC and 55 RH burnt lime (CaO) readilytransforms into portlandite according to the fol-lowing reaction (Boynton 1980)

portlanditeCaO + H2O reg Ca(OH)2 (4)

This process generates crystallization pressure inconfined spaces (ie the brick pores originallyoccupied by CaO) resulting in crack development(as observed in Fig 2a) The dark-color rimaround former carbonate grains (Fig 5c) is proba-bly due to limited portlandite carbonation in thepresence of CO2 also resulting in volume increase(Moorehead 1985) and crack development(Butterworth amp Honeyborne 1952) As a commonpractice to avoid these undesired side-effectsbricks that contain carbonates were traditionallyimmersed in water to dissolve and eliminate exist-ing CaO or Ca(OH)2

This effect is less evident in the case of MgOPericlase in the presence of H2O (room T) under-goes a slow transformation (it takes months oreven years Webb 1952) into brucite (Mg(OH)2)and may eventually transform into hydromagne-site (Mg5(CO3)4(OH)24H2O) according to thefollowing reactions (Garavelli et al 1990)

MgO + H2O reg Mg(OH)2 (5)

hydromagnesite5Mg(OH)2 + 4CO2 reg Mg5(CO3)4(OH)24H2O (6)

EMPA results indicate the presence of Ca and Mgoxides however our XRD results do not clearlyshow the presence of Ca and Mg oxides or hydrox-ides mainly due to their low concentrations (lt 10wt ) and because their main diffraction peaks aremasked by other major phase peaks (eg quartzfeldspars hematites) The hydromagnesite mainBragg peak at 581 Aring is not present either

At T sup3 800degC and in the absence of carbonateschanges are constrained to textural modificationsand reactions where a low-T phase is decomposedand fully transformed into a high-T phase withoutmajor compositional changes (Riccardi et al1999) In particular muscovite undergoes a solid-state phase change into a mixture of mullite and K-feldspar (or a melt) at T ranging from 800 to1000degC The composition of high-T muscovite isclose to K-feldspar (EMPA results) although OManalysis points to the presence of mullite Brearley(1986) and Worden et al (1987) showed fine inter-growth of alternating nanometer up to a fewmicrometers thick bands of K-feldspar (or a meltwith composition close to that of K-feldspar) andmullite in pyrometamorphic muscovite which canexplain our EMPA results

The most significant textural and mineralogi-cal changes are observed in samples with carbon-ates when fired at T gt 800degC Also melting T andextent seems to be connected with the presence orabsence of carbonates In general melting starts atlower T (~ 800degC) when carbonates are present(Tite amp Maniatis 1975) Ca and Mg from carbon-ates may act as melting agents (Segnit ampAnderson 1972) but they are reported to some-how limit the extent of vitrification at T gt 1000degC(Everhart 1957 Nuacutentildeez et al 1992) In fact weobserved that vitrification is more extended athigher T when no carbonates exist This may beexplained considering that the phyllosilicate in thematrix of non-carbonate G samples not contribut-ing to the formation of high-T Ca (or Mg) silicatesand releasing significant amounts of H2O due todehydroxylation they can contribute to extensivemelting at high-T as demonstrated by Brearley ampRubie (1990) On the other hand the higher SiO2content in these samples provides a higher amountof potential silicate-rich melt

BSE images of reaction rims between carbon-ates and silicates show that formation of Ca-sili-cates starts at 800degC This is confirmed by XRD(Fig 3b) and EMPA analyses (Table 3) Gehlenitestarts to form at this T by grain-boundary reactionbetween CaO Al2O3 and SiO2 the first derivingfrom former carbonates and the later two fromalready dehydroxylated phyllosilicates (ie a

630

somehow amorphous mixture of 3SiO2Al2O3according to Peters amp Iberg 1978) as follows

illite calcite gehleniteKAl2(Si3Al)O10(OH)2 + 6CaCO3 reg 3Ca2Al2SiO7+ 6CO2 + 2H2O + K2O + 3SiO2 (7)

The formation of gehlenite is striking becauseone could expect the formation of an Al-richpyroxene such as fassaite (Dondi et al 1998)Nevertheless it should be indicated that gehleniteformation is a good example of the so-calledGoldsmithacutes ldquoease of crystallizationrdquo principlewhich states that a silicate phase with all Al in 4-coordination is easier to form (at least in the lab)than a phase that includes Al in both 4- and 6-coor-dination when (OH)- are not present (Goldsmith1953) This is due to the higher energy barrier nec-essary to overcome when Al has to enter in a 6-coordination if compared with a 4-coordination If(OH)- are present this energy barrier is reduceddue to the smaller electrostatic repulsion among(OH)- groups than among O2- in octahedral coor-dination Gehlenite has all Al in tetrahedral coor-dination (Warren 1930) therefore crystallizingpreferentially if compared with other anhydrousCa-Al silicates This simple principle also appliesto the high-T formation of other anhydrous Al-sil-icates as we will discuss later

Wollastonite formed at the carbonate-quartzinterface through the following reaction

wollastoniteCaO + SiO2 reg CaSiO3 (8)

Wollastonite and gehlenite appear at 800degC aT ~100degC lower than previously reported (Petersamp Iberg 1978 Dondi et al 1998) This is proba-bly due to a) the low resolution of the analyticaltechniques used in the past (ie XRD) which didnot allow the detection of the small initial reactionrim between carbonates and silicates as it is evi-denced by BSE images and EMPA analysesandor b) most previous works on high-T phaseformation in ceramic processing use single com-ponents as reactants without considering addition-al phases or previous reaction by-products (Petersamp Iberg 1978) We observed that in a closer-to-reality multicomponent system where interactionsbetween reactants and products occur differentformation T are obtained

At the dolomite-quartz interface diopsidestarts to form at 900degC by the following reaction

dolomite diopside2SiO2 + CaMg(CO3)2 reg CaMgSi2O6 + 2CO (9)

At this T anorthite starts to form at the expensesof calcite and illite (plus quartz)

illite calciteKAl2(Si3Al)O10(OH)2 + 2CaCO3 + 4SiO2 regK-feldspar anorthite

2KAlSi3O8 + 2Ca2Al2SiO8 + 2CO2 + H2O (10)

At higher temperatures (1000 to 1100degC)phyllosilicates have already disappeared in allsamples (V and G) being transformed into sani-dine and mullite plus a melt according to the fol-lowing reaction

illite (muscovite) sanidine 3KAl2(Si3Al)O10(OH)2 + 2SiO2 reg 3KAlSi3O8 +

mulliteAl6Si2O13 + 3H2O + melt (11)

Formation of a melt during the previous reaction(Brearley amp Rubie 1990) is consistent with theobserved ldquocellular structurerdquo of muscovite (Tite ampManiatis 1975)

Brearley (1986) proposed a somehow differentreaction for mullite formation at the expenses ofphengitic muscovite resulting in K-feldspar +mullite + biotite No biotite is detected in ourexperiments most probably due to the low Fe con-tent in V and G micas Brearley amp Rubie (1990)experimentally demonstrated that high-T mus-covite breakdown in the presence of quartz isinfluenced by the crystal size The smallest crys-tals have a strong tendency to react with H2Oresulting in enhanced melt formation while largemuscovite crystals form limited amounts of meltThis may account for the significant vitrificationof the matrix (specially in G samples) while largemuscovite grains (up to some mm in size) still donot show significant melting

Mullite which according to OM analysis firstappears at 800degC is the second most abundantphase (after quartz) at 1100degC in the richer inphyllosilicate G samples It is interesting to men-tion that K-feldspar progressively disappears as Tincreases This is not consistent with publisheddata (Maggetti 1982) It seems that low-T K-feldspar (microcline) is not stable at high-T there-fore it transforms into a high-T K-feldspar (iesanidine) or reacts with lime to form anorthiteBoth sanidine and anorthite are detected in thehigh-T bricks However a higher concentration ofsanidine should be expected to form speciallyconsidering reaction (11) A possible explanationfor the low K-feldspar content at high-T is that K-feldspar might not have enough time to develop itscrystalline structure (sanidine) therefore being

Carbonate and silicate phase reactions during ceramic firing 631

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

concentrated as an amorphous (sanidine-like)phase responsible in part for the rise in the back-ground noise in the XRD results Mullite forma-tion seems to be kinetically faster resulting in awell-developed crystalline phase which givessharp diffraction peaks Goldsmith (1953) indi-cates that mullite could be another example of hisldquosimplexity principlerdquo for the ldquoeaserdquo of crystal-lization if compared with other aluminum sili-cates In fact the mullite structure has a largeramount of Al in 4-coordination and less Al in 6-coordination than sillimanite (Rehak et al 1998)for instance which may account for the abundanceof the former and the non-existence of the latter inthe studied bricks

In carbonate-rich V samples fired at T gt1000degC a significant reduction of gehlenite maindiffraction peak is observed According to Petersamp Iberg (1978) this is due to the reaction of thisphase with quartz forming wollastonite and anor-thite This is consistent with the increase in themain Bragg peak intensity of these later mineralsat 1100degC

At the dolomite-silicate (ie Qtz and Phy)interface it is common to observe wollastoniteeither newly formed or resulting from the destruc-tion of gehlenite The texture of the interface israther complex Protrusion or fingers from thereaction zone penetrate into areas were completetransformation into wollastonite has occurred Theinterface geometry resembles the so-called ldquovis-cous fingersrdquo (Garcia-Ruiz 1992) occurringwhen two fluids or semi-plastic solids of differentcomposition and viscosity are placed in contactInitially the contact zone is a flat surface but lateron it reaches a fingered fractal geometry(Nittmamm et al 1985) In the carbonate-silicateinterface a similar process may have taken placewhere mass transport of the reactants may havebeen enhanced by partial melting The formationof a silicate-rich melt due to clay-minerals melting(Brearley amp Rubie 1990) in the external part ofthe ring in contact with a second inner melt formedby preexisting gehlenite or the eutectic CaO-SiO2would result in viscous finger development Thefractal geometry of the developing viscous fingersmay have enhanced diffusion of mobile ions suchas Ca toward silicate reactive sites Both EDXdata showing Mg enrichment in the dolomite core(Fig 5f) and the compositional trend towardsMgO-rich terms in the CMS diagram (Fig 6b)confirm that Ca is preferentially leached out of theformer dolomite grain Finger formation willresult in a significant increase in interface surfacearea resulting in higher reaction rates Another

alternative explanation for this fingered geometrydevelopment involves partial melting at the car-bonate-silicate interface and mass transport (ieviscous flow) through the clay-quartz porousmatrix resulting in the so-called ldquoscalloped struc-turerdquo proposed by Ortoleva et al (1987) for reac-tion-diffusion processes While the viscous fingermodel requires two melts the second model onlyrequires one melt It is not clear if at any momentconditions for two melt occurrence are reachedHence the second model seems more plausible

In any case the process which is originally dif-fusion driven is transformed into a mass (ie fin-gering) plus diffusion transport which greatlyenhances Ca-silicate formation This is consistentwith the thick reaction rims (up to 250 microm thicksee Fig 5f) that surround former carbonate grainsThese thick Ca-silicate shells would hardly devel-op just by diffusion in the short time-scale of theexperiment (3 hours at the peak T) In fact at800degC (conditions where melting would be verylimited) Ca (and Mg)-silicate rims formed aroundcalcite or dolomite are 2-5 microm thick At 1100degCwhen melting is extensive the reaction rate is sohigh that rims are two orders of magnitude thick-er Therefore mass transport (ie viscous flow)seems to have played a key role in high-T Ca-(Mg-) silicate formation Mass transport is also consis-tent with observation of some areas with differentcomposition (unreacted silicates) trapped betweenfingers resulting in the island geometry observedin Fig 5g It could be expected that without fin-gering the process of Ca- or (Ca-Mg)-silicate for-mation would be self-limiting ie once a rim ofthe newly formed silicate appears diffusion wouldbe jeopardized This shell would preclude or limitthe transport of Ca (andor Mg) from the carbon-ate or it would limit the transport of the lessmobile Si (and Al) from the surrounding silicatetowards the reaction front Therefore it is neces-sary that one of the above described transport-reaction mechanisms takes place This may alsoexplain why phase formation takes place at lowerT than reported previously The presence of multi-ple interfaces among different crystals mayenhance finger development resulting in kineticchanges favoring phase formation at lower T

5 Conclusions

It is observed that even though the overallcomposition of a ceramic material does not under-go major changes upon firing significant miner-alogical and textural modifications do occur

632

These changes are connected with two reactiontypes a) solid-state replacement of a phase by anew one with little or no compositional variation(ie the muscovite and illite case) and minimizingenergy changes by preserving the original textureand orientation of the replaced phase and b) reac-tion-diffusion at the interface between two miner-al grains with contrasting composition (iecarbonate-silicate interfaces) The latter reaction isenhanced if massviscous transport resulting frompartial melting and finger structure developmentoccurs Non-carbonated clay-rich materials (G)show phyllosilicate destruction at T ranging from700 to 900degC followed by vitrification which issignificant at T gt 1000degC Carbonate-rich materi-als (V) show major phase transformation uponphyllosillicates destruction including new Ca (orMg) silicates and aluminum-silicates formation Inthis later case the vitrification starts at lower Tand it is limited at higher T Therefore it is possi-ble to establish the firing T of ceramic materialswith composition broadly similar to that in ourwork based on the mineralogy Thus a) presenceof phyllosilicates indicates a low 700 to 900degC fir-ing T for the dehydroxylated illite (or muscovite)phase b) dolomite disappears at sup3 700degC whilecalcite is still present at 800degC c) gehleniteappears at 800degC reducing its concentration above1000degC d) diopside and wollastonite appear at800degC (in very low concentrations) as reactionrims between carbonates and silicates Formationof these minerals includes ldquofingeringrdquo develop-ment e) mullite appears at T gt 800degC while littleamounts of sanidine form as microcline disap-pears and f) anorthite develops at T gt1000degC incarbonate-rich bricks

When considering equilibrium thermodynamicdata and calculated reaction T for most of theabove mentioned phases and reactions (Holland ampPowell 1985 and references herein) it isobserved that rapid heating induces significantoverstepping Therefore phases formed are inmost cases metastable a fact consistent with anal-yses of pyrometamorphic assemblages (Brearleyamp Rubie 1990 and references herein)

The formation of non-equilibrium fractal mor-phologies at the carbonate-silicate reaction inter-faces may explain why high-T silicates incarbonate-rich ceramics develop very fast Thisgrain-to-grain reaction model may help explainhigh-T silicate formation in nature (eg contactmetamorphism pyrometamorphism in xenoliths)

The gained knowledge on mineral transforma-tion and the dynamics of interface mass and diffu-sion transport facilitated through ldquofingeringrdquo or

ldquoscallopingrdquo opens new ways to understand thecomplexities of ceramic processing in particularand high-T mineral transformation in general bothin natural (ie geologic settings) and artificialcontexts (eg archaeometric analysis of ancientceramics or the design of new bricks ndashreplicas ofancient onesndash for architectural heritage conserva-tion) However care must be taken when extrapo-lating these results to other raw materialcompositions andor ceramic processing methods

Acknowledgements This work has been finan-cially supported by the CGICYT projectMAT2000-1457 and the Research Group RNM0179 of the Junta de Andaluciacutea (Spain) We thankthe Centro de Instrumentacioacuten Cientifica (CIC) ofthe Granada University for allowing us to use theirSEM-EDX EMPA and XRF facilities We thankK Elert JM Garciacutea-Ruiz M Rodriguez-Gallego and A Sanchez-Navas for fruitful discus-sions and comments We also thank the in-depthreview by Dr F Goumltz-Neunhoeffer and an anony-mous referee

References

Barahona E Huertas F Pozzuoli A Linares J(1985) Firing properties of ceramic clays fromGranada province Spain Miner Petrogr Acta 29A577-590

Boynton RS (1980) Chemistry and technology oflime and limestone 2nd Ed Wiley New York

Brearley AJ (1986) An electron optical study of mus-covite breakdown in pelitic xenoliths duringpyrometamorphism Miner Mag 357 385-397

Brearley AJ amp Rubie DC (1990) Effects of H2O onthe disequilibrium breakdown of muscovite +quartz J Petrol 31 925-956

Butterworth B amp Honeyborne DB (1952) Bricksand clays of the hasting beds Trans Brit CeramSoc 51 211-259

Dondi M Ercolani G Fabbri B Marsigli M (1998)An approach to the chemistry of pyroxenes formedduring the firing of Ca-rich silicate ceramics ClayMinerals 33 443-452

Duminuco P Riccardi MP Messiga B Setti M(1996) Modificazioni tessiturali e mineralogichecome indicatori della dinamica del processo di cot-tura di manufatti ceramici Ceramurgia 26 5 281-288

Evans JL amp White J (1958) Further studies of thethermal decomposition of clays Trans Brit CeramSoc 57 6 298

Everhart JO (1957) Use of auxiliary fluxes to improvestructural clay bodies Bull Am Ceram Soc 36268-271

Carbonate and silicate phase reactions during ceramic firing 633

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

32 Optical microscopy analysis

G and V samples fired at 700degC do not showsignificant mineralogical transformations Thetemper grain-size (ie phenocrysts) reaches up to15 mm No major textural changes are visible inthe carbonates at 700degC Phyllosilicates in thepaste are oriented due to compression duringmolding At 800degC carbonate decomposition in Vsamples is almost complete and some cracksappear surrounding the former carbonate grains(Fig 2a) The matrix in both G and V samplesturns dark with low birefringence as T increasesMuscovite crystals in both G and V samples startto transform into mullite at 800degC and some bub-bles develop within these crystals at higher T (Fig2b and 2c) Muscovite replacement by mullite iscompleted at 1000degC Mullite preserves the mor-phology of replaced muscovite In fact mullite(001) planes are ^ to (001) planes of former mus-covite grains as deduced by the optical axes ori-entation study At 1000-1100degC the matrix is fullyvitrified and brown colored (in both G and V sam-ples) Phyllosilicates in the matrix do not exist atthese T only some pseudomorphs after formercrystals can be observed An empty space withscattered grayish fragments at the pore edges canbe found where calcite was originally presentCracks observed in V samples at T lt 900degC disap-pear at 1100degC due to extended vitrification Poresbecome more abundant and they are no longerellipsoidal but spherical However an orientationof the remaining temper grains in planes parallel tothe brick largest faces (ie where compressionwas applied during molding) remains

33 XRD analysis

Raw G samples show significant amounts ofquartz and phyllosilicates with little feldspar Thelt 2 microm fraction includes illite and smectite withsmall amounts of kaolinite and paragonite (Table

1) There is no chlorite which could have beenmasked by the d(001) peaks of smectite (~ 14 Aring)and kaolinite (71 Aring) as evidenced by these peaksdisappearing after 90acute heating at 550degC Raw Vsamples have significant amounts of calcite anddolomite The clay fraction includes illite andsmectite with trace amounts of paragonite andkaolinite (Table 1) Therefore the main differenceamong the two raw materials lays in the miner-alogical composition of the gt 2 microm fraction

Regarding the phyllosilicates evolution uponfiring only the diffraction peak at 10 Aring corre-sponding to a dehydroxylated illite-like phaseremains at 700degC This peak intensity is reducedupon firing at higher T till it disappears at ~900degC (Fig 3a and 3b) Microcline main diffrac-tion peak reduces its intensity in G samples Thismineral disappears at T gt 1000degC when traceamounts of sanidine are formed Mullite is detect-ed in both G and V samples at T ~ 900degC increas-ing its main diffraction peak (d(210) = 339 Aring)intensity at higher T Mullite diffraction peaks aremore intense in phyllosilicate-rich G samples thanin carbonate-rich V samples Other phases under-go a significant increase in their diffraction peaksintensities this is the case of hematites in G sam-ples fired at T gt 1000degC New phases appear in Vsamples upon firing a) gehlenite (a melilite-groupphase) appears at 800degC increasing its mainBragg peak intensity at 900degC and reducing it athigher T b) wollastonite and diopside appear at1000degC and they show a significant increase intheir main diffraction peak at higher TDecomposition of calcite and dolomite begins at Tgt 700degC and both phases disappear at T gt 800degCQuartz remains as the most abundant phase at anyT in both V and G samples In V samples the pla-gioclase increases its Ca content (ie a more An-rich phase) as well as its main Bragg peakintensity as T increases This increase is more sig-nificant at T gt 1000degC The existence of an amor-phous phase (ie vitreous phase) is evidenced by

624

Table 1 Results of XRD analysis of the raw materials and the fraction with lt 2 m m grain size Mineral symbols afterKretz (1983) (in Table 3)

a rise of the background noise in the XRD patternat T gt 900degC in the G samples and at T gt 800degCin the V samples

34 SEM-EDX analysis

Significant textural changes most evident withrespect to pore morphology and volume weredetected upon firing An overall porosity reductionis observed at T gt 1000degC when the vitreous phasefills the pores

Samples fired at 700 and 800degC still preservethe laminar habit of phyllosilicates although mus-covite crystals clearly exfoliate along basal planes

most probably due to dehydroxylation (Fig 4a)The interlocking among particles is limited (Fig4d) V samples show some cracks At these T noclear evidence of sintering or partial melting isdetected However there is indirect evidence (ieincrease in background noise of XRD patterns)that some vitrification is reached at T lt 900degC incarbonate-rich V samples Vitrification is clearlyobserved in all samples at T gt 900degC ie phyl-losilicates appear deformed (their edges turnsmooth) andor aggregated (Fig 4b and 4e) andthe pores turn ellipsoidal with smooth surfaces AtT gt 1000degC these effects are less extended in thecarbonate-rich V samples (Fig 4c) than in the sil-

Carbonate and silicate phase reactions during ceramic firing 625

Fig 3 Guadix (a) and Viznar (b) samples powder X-ray difraction patterns Legend (mineral symbols after Kretz1983) Sm = smectite Ill = illite Pg = paragonite Qtz = quartz Cal = calcite Kln = kaolinite Phy = phyllosilicatesFs = feldspar Dol = dolomite Gh = gehlenite Hem = hematite Wo = wollastonite Di = diopside An = anorthiteMul = mullite Sa = sanidine CuKa X-ray radiation l = 15064 Aring

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

icate-rich G samples (Fig 4f) A higher degree ofparticle interlocking results in a porosity reduc-tion At 1100degC vitrification is significant in allsamples regardless of initial compositional differ-ences Pores coalesce as spherical cells due to par-tial or extended melting of clay particles in thematrix

BSE images of mineral and textural evolutionof samples with high carbonate content (ie Vsamples which undergo the most significant min-eralogical changes) give an interesting picture oflocal phase transformations taking place at grainboundaries No textural or mineralogical changescan be detected in brick fired at 700degC (Fig 5a)Carbonates comprise the larger-size fraction inthese samples which allows easy detection ofhigh-T reaction rims In fact at 800degC 2-5 micromthick reaction rims are observed developing at thecalcite-quartzphyllosilicate interface (Fig 5b)Rims composition corresponds to a calcium sili-cate phase presumably wollastonite (EDXresults) A dark rim surrounded by a light outershell is observed where dolomite was originallypresent (Fig 5c) However no compositional vari-ation among internal (greyish) and external(white) rims is detected We were not able tounambiguously detect the presence of gehlenite

within these reaction rims (ie some EDX analy-ses of 2-5 microm thick rims at the carbonate-silicateinterface showed Al together with Ca and Si butpossible contamination from underlying phyllosil-icates was not discarded) However the presenceof gehlenite was detected by XRD at this T(800degC) Muscovite phenocrysts (Fig 5d) show amorphology and composition typical for unalteredwhite mica At 1100degC muscovite crystals showsecondary bubbles development between layers(Fig 5e) The mica composition is homogeneous(no BSE image gray-level differences along thecrystals) and it is similar to that of an unchangedmuscovite but with slightly lower K concentra-tion At 1100degC carbonates (ie dolomite in Fig5f) shows an outer thick white Ca- and Si-rich rim(ie wollastonite) which includes small (micronsized) bright-white Ca- Si- and Al-rich particles(ie gehlenite or plagioclase) There is a strikingfeature of this reaction rim it shows a fingeredgeometry at the outer edge The darker dolomitecore shows a significant Mg enrichment

35 EMPA analysis

Due to the fact that only silicate phases in sam-ples with carbonates undergo mayor composition-

626

Fig 4 SEM secondary electron photomicrographs of Viznar samples fired at 800degC (a) 1000degC (b) and 1100degC(c) and Guadix samples fired at 700degC (d) 900degC (e) and 1000degC (f)

al changes upon firing only V samples were ana-lyzed using this technique In particular composi-tional changes of phyllosilicates and carbonateswere studied using polished thin sections of firedV bricks (one section per selected target T ie700 800 1100degC)

Phyllosilicates (Table 2) show a muscoviticcomposition at 700degC where some K is replaced byNa even though the overall composition does notseem to correspond to paragonite Neverthelessparagonite was detected in the raw clay (Fig 3b)

Fe Mg and Ti replace some Al in the octahedralsheets Replacement of OH- by F- is almost absentAt 800degC a slight increase in Fe content is detectedOnly analyses M9 and M10 show higher concen-trations of Na and Al At 1100degC the amount of Kis reduced while Ca and Si content is increasedThe total oxide content is close to 100 a valueconsistent with complete dehydroxylation of thephyllosilicates Analyses M11-M14 in Table 2 arein agreement with high-T phengite compositionaldata from Worden et al (1987)

Carbonate and silicate phase reactions during ceramic firing 627

Fig 5 BSE images and EDX analyses of carbonate-rich Viznar samples a) fired at 700degC b) fired at 800degC whenWo (EDX analysis) formed at the Cal- silicates interface c) detail of former carbonate grain showing a white outerreaction rim (Wo or Gh) a crack and an inner core with an outer darker shell (see text for details) d) Ms (see EDXanalysis) at 700degC e) former Ms after transformation into a Fs + Mul + melt mixture (see EDX analysis) f) DolPhy(+ Qtz) interface showing a reaction rim with fingered geometry formed at 1100degC g) detail of the outer reactionrim in previous figure EDX analysis shows the rim composition (ie Wo)

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

Calcite (Table 3) shows no chemical changesat 700degC The maximum MgO content is 104 wt Trace amounts of Fe Si Al and Mn are alsodetected in calcite (as well as in dolomite) At800degC both carbonates are fully transformed eitherinto burnt lime (CaO) or into a mixture of burtlime and periclase (MgO) (eg analysis C14)Gehlenite-like compositions are detected in someanalyses (eg C10) At 1100degC the presence ofnewly formed phases in particular wollastonite isclearly confirmed (eg C17) Ca content in thelime plus periclase mixture is reduced significant-ly due to the incorporation of this element intonewly formed high-T silicate phases As T increas-es a clear compositional trend toward Si- and Al-rich calcium phases (ie toward gehlenite andanorthite) is observed (Fig 6a) a fact consistentwith XRD and SEM-EDX results A clear trendtoward wollastonite-like compositions as well as a

trend toward MgO rich composition is alsoobserved as T increases (Fig 6b) Some EMPAanalyses suggest a larnite -like compositionHowever XRD results do not show the presenceof larnite

36 XRF analysis

According to the classification proposed forclay materials in ceramic industry by Fabbri ampFiori (1985) G raw material composition falls intothe quartz-feldspar sands category while V sam-ples fall out of any defined category due to theirlow silica and high calcium contents (Table 4) Asexpected firing does not affect major elementconcentration in the brick The only significantchange is detected in the loss on ignition values Itis evidenced that G samples lost almost all waterbetween 700 and 900degC while V samples show a

628

Table 2 EMPA analysis results of phyllosilicates composition (in wt )

Table 3 Composition of selected carbonate grains ( EMPA results in wt )

more gradual loss of water and CO2 as T increases(Table 4)

4 Discussion

Firing up to 700degC induce no significant min-eralogicaltextural transformations in both thecarbonate-rich V and the silicate-rich G sampleswith the exception of the clay minerals Clay min-erals structure is reported to collapse due to dehy-droxylation to resemble an illite-like structure(XRD data) when a T between 450 and 550degC isreached (Evans amp White 1958) A dehydroxylatedphyllosilicate phase structurally different to thehydrated one (Guggenheim et al 1987) is report-ed to exist up to 950degC when complete breakdownof the dehydroxylated illite occurs (Peters amp Iberg1978) Our data indicate that dehydroxylation isnot completed at T ~ 900degC which indicates thatthe kinetics of this process is slower than previ-ously estimated (Evans amp White 1958) (OH)-

released at T gt 700degC may play a significant rolein texture and mineral evolution at high-T as willbe discussed later On the other hand muscoviteshows a slight K deficit at 700degC probably due tothe high diffusion capacity of this element alongmuscovite (001) basal planes at T gt 500degC(Sanchez-Navas amp Galindo-Zaldivar 1993)Sanchez-Navas (1999) showed high-T K-diffu-sion-out of mica is induced by water adsorbedalong mica basal planes resulting from phyllosili-cate dehydroxylation

XRD data shows that dolomite most intenseBragg peak in V bricks significantly reduced itsintensity due to thermal decomposition at lower Tthan calcite main Bragg peak (Fig 3b) Dolomite

Carbonate and silicate phase reactions during ceramic firing 629

Table 4 Bulk composition of Guadix (G) and Viznar (V) samples (XRF results in wt )

LOI loss on ignition

Fig 6 a) ACS (Al2O3-CaO-SiO2) and b) CMS (CaO-MgO-SiO2) compositional diagrams of carbonates andreaction products (EMPA results)

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

starts to decompose according to (1) at ~ 700degC (1atm pressure) while calcite decomposes at higherT (830 ndash 870degC) according to (3) (Boynton 1980)Shoval (1988) proposed a two steps thermaldecomposition for dolomite involving formationof periclase and calcite (2) followed by calcitedecomposition at higher T (3)

dolomite lime periclaseCaMg(CO3)2 reg CaO + MgO + 2CO2 (1)

calcite CaMg(CO3)2 reg CaCO3 + MgO + CO2 (2)

CaCO3 reg CaO + CO2 (3)

However our XRD results showing no increase incalcite 104 Bragg peak intensity as dolomitedecomposes at 700degC (Fig 3b) are not consistentwith Shovalacutes two-steps process

Carbonate decomposition results in shrinkageof the brick (Barahona et al 1985) Upon storageat 22degC and 55 RH burnt lime (CaO) readilytransforms into portlandite according to the fol-lowing reaction (Boynton 1980)

portlanditeCaO + H2O reg Ca(OH)2 (4)

This process generates crystallization pressure inconfined spaces (ie the brick pores originallyoccupied by CaO) resulting in crack development(as observed in Fig 2a) The dark-color rimaround former carbonate grains (Fig 5c) is proba-bly due to limited portlandite carbonation in thepresence of CO2 also resulting in volume increase(Moorehead 1985) and crack development(Butterworth amp Honeyborne 1952) As a commonpractice to avoid these undesired side-effectsbricks that contain carbonates were traditionallyimmersed in water to dissolve and eliminate exist-ing CaO or Ca(OH)2

This effect is less evident in the case of MgOPericlase in the presence of H2O (room T) under-goes a slow transformation (it takes months oreven years Webb 1952) into brucite (Mg(OH)2)and may eventually transform into hydromagne-site (Mg5(CO3)4(OH)24H2O) according to thefollowing reactions (Garavelli et al 1990)

MgO + H2O reg Mg(OH)2 (5)

hydromagnesite5Mg(OH)2 + 4CO2 reg Mg5(CO3)4(OH)24H2O (6)

EMPA results indicate the presence of Ca and Mgoxides however our XRD results do not clearlyshow the presence of Ca and Mg oxides or hydrox-ides mainly due to their low concentrations (lt 10wt ) and because their main diffraction peaks aremasked by other major phase peaks (eg quartzfeldspars hematites) The hydromagnesite mainBragg peak at 581 Aring is not present either

At T sup3 800degC and in the absence of carbonateschanges are constrained to textural modificationsand reactions where a low-T phase is decomposedand fully transformed into a high-T phase withoutmajor compositional changes (Riccardi et al1999) In particular muscovite undergoes a solid-state phase change into a mixture of mullite and K-feldspar (or a melt) at T ranging from 800 to1000degC The composition of high-T muscovite isclose to K-feldspar (EMPA results) although OManalysis points to the presence of mullite Brearley(1986) and Worden et al (1987) showed fine inter-growth of alternating nanometer up to a fewmicrometers thick bands of K-feldspar (or a meltwith composition close to that of K-feldspar) andmullite in pyrometamorphic muscovite which canexplain our EMPA results

The most significant textural and mineralogi-cal changes are observed in samples with carbon-ates when fired at T gt 800degC Also melting T andextent seems to be connected with the presence orabsence of carbonates In general melting starts atlower T (~ 800degC) when carbonates are present(Tite amp Maniatis 1975) Ca and Mg from carbon-ates may act as melting agents (Segnit ampAnderson 1972) but they are reported to some-how limit the extent of vitrification at T gt 1000degC(Everhart 1957 Nuacutentildeez et al 1992) In fact weobserved that vitrification is more extended athigher T when no carbonates exist This may beexplained considering that the phyllosilicate in thematrix of non-carbonate G samples not contribut-ing to the formation of high-T Ca (or Mg) silicatesand releasing significant amounts of H2O due todehydroxylation they can contribute to extensivemelting at high-T as demonstrated by Brearley ampRubie (1990) On the other hand the higher SiO2content in these samples provides a higher amountof potential silicate-rich melt

BSE images of reaction rims between carbon-ates and silicates show that formation of Ca-sili-cates starts at 800degC This is confirmed by XRD(Fig 3b) and EMPA analyses (Table 3) Gehlenitestarts to form at this T by grain-boundary reactionbetween CaO Al2O3 and SiO2 the first derivingfrom former carbonates and the later two fromalready dehydroxylated phyllosilicates (ie a

630

somehow amorphous mixture of 3SiO2Al2O3according to Peters amp Iberg 1978) as follows

illite calcite gehleniteKAl2(Si3Al)O10(OH)2 + 6CaCO3 reg 3Ca2Al2SiO7+ 6CO2 + 2H2O + K2O + 3SiO2 (7)

The formation of gehlenite is striking becauseone could expect the formation of an Al-richpyroxene such as fassaite (Dondi et al 1998)Nevertheless it should be indicated that gehleniteformation is a good example of the so-calledGoldsmithacutes ldquoease of crystallizationrdquo principlewhich states that a silicate phase with all Al in 4-coordination is easier to form (at least in the lab)than a phase that includes Al in both 4- and 6-coor-dination when (OH)- are not present (Goldsmith1953) This is due to the higher energy barrier nec-essary to overcome when Al has to enter in a 6-coordination if compared with a 4-coordination If(OH)- are present this energy barrier is reduceddue to the smaller electrostatic repulsion among(OH)- groups than among O2- in octahedral coor-dination Gehlenite has all Al in tetrahedral coor-dination (Warren 1930) therefore crystallizingpreferentially if compared with other anhydrousCa-Al silicates This simple principle also appliesto the high-T formation of other anhydrous Al-sil-icates as we will discuss later

Wollastonite formed at the carbonate-quartzinterface through the following reaction

wollastoniteCaO + SiO2 reg CaSiO3 (8)

Wollastonite and gehlenite appear at 800degC aT ~100degC lower than previously reported (Petersamp Iberg 1978 Dondi et al 1998) This is proba-bly due to a) the low resolution of the analyticaltechniques used in the past (ie XRD) which didnot allow the detection of the small initial reactionrim between carbonates and silicates as it is evi-denced by BSE images and EMPA analysesandor b) most previous works on high-T phaseformation in ceramic processing use single com-ponents as reactants without considering addition-al phases or previous reaction by-products (Petersamp Iberg 1978) We observed that in a closer-to-reality multicomponent system where interactionsbetween reactants and products occur differentformation T are obtained

At the dolomite-quartz interface diopsidestarts to form at 900degC by the following reaction

dolomite diopside2SiO2 + CaMg(CO3)2 reg CaMgSi2O6 + 2CO (9)

At this T anorthite starts to form at the expensesof calcite and illite (plus quartz)

illite calciteKAl2(Si3Al)O10(OH)2 + 2CaCO3 + 4SiO2 regK-feldspar anorthite

2KAlSi3O8 + 2Ca2Al2SiO8 + 2CO2 + H2O (10)

At higher temperatures (1000 to 1100degC)phyllosilicates have already disappeared in allsamples (V and G) being transformed into sani-dine and mullite plus a melt according to the fol-lowing reaction

illite (muscovite) sanidine 3KAl2(Si3Al)O10(OH)2 + 2SiO2 reg 3KAlSi3O8 +

mulliteAl6Si2O13 + 3H2O + melt (11)

Formation of a melt during the previous reaction(Brearley amp Rubie 1990) is consistent with theobserved ldquocellular structurerdquo of muscovite (Tite ampManiatis 1975)

Brearley (1986) proposed a somehow differentreaction for mullite formation at the expenses ofphengitic muscovite resulting in K-feldspar +mullite + biotite No biotite is detected in ourexperiments most probably due to the low Fe con-tent in V and G micas Brearley amp Rubie (1990)experimentally demonstrated that high-T mus-covite breakdown in the presence of quartz isinfluenced by the crystal size The smallest crys-tals have a strong tendency to react with H2Oresulting in enhanced melt formation while largemuscovite crystals form limited amounts of meltThis may account for the significant vitrificationof the matrix (specially in G samples) while largemuscovite grains (up to some mm in size) still donot show significant melting

Mullite which according to OM analysis firstappears at 800degC is the second most abundantphase (after quartz) at 1100degC in the richer inphyllosilicate G samples It is interesting to men-tion that K-feldspar progressively disappears as Tincreases This is not consistent with publisheddata (Maggetti 1982) It seems that low-T K-feldspar (microcline) is not stable at high-T there-fore it transforms into a high-T K-feldspar (iesanidine) or reacts with lime to form anorthiteBoth sanidine and anorthite are detected in thehigh-T bricks However a higher concentration ofsanidine should be expected to form speciallyconsidering reaction (11) A possible explanationfor the low K-feldspar content at high-T is that K-feldspar might not have enough time to develop itscrystalline structure (sanidine) therefore being

Carbonate and silicate phase reactions during ceramic firing 631

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

concentrated as an amorphous (sanidine-like)phase responsible in part for the rise in the back-ground noise in the XRD results Mullite forma-tion seems to be kinetically faster resulting in awell-developed crystalline phase which givessharp diffraction peaks Goldsmith (1953) indi-cates that mullite could be another example of hisldquosimplexity principlerdquo for the ldquoeaserdquo of crystal-lization if compared with other aluminum sili-cates In fact the mullite structure has a largeramount of Al in 4-coordination and less Al in 6-coordination than sillimanite (Rehak et al 1998)for instance which may account for the abundanceof the former and the non-existence of the latter inthe studied bricks

In carbonate-rich V samples fired at T gt1000degC a significant reduction of gehlenite maindiffraction peak is observed According to Petersamp Iberg (1978) this is due to the reaction of thisphase with quartz forming wollastonite and anor-thite This is consistent with the increase in themain Bragg peak intensity of these later mineralsat 1100degC

At the dolomite-silicate (ie Qtz and Phy)interface it is common to observe wollastoniteeither newly formed or resulting from the destruc-tion of gehlenite The texture of the interface israther complex Protrusion or fingers from thereaction zone penetrate into areas were completetransformation into wollastonite has occurred Theinterface geometry resembles the so-called ldquovis-cous fingersrdquo (Garcia-Ruiz 1992) occurringwhen two fluids or semi-plastic solids of differentcomposition and viscosity are placed in contactInitially the contact zone is a flat surface but lateron it reaches a fingered fractal geometry(Nittmamm et al 1985) In the carbonate-silicateinterface a similar process may have taken placewhere mass transport of the reactants may havebeen enhanced by partial melting The formationof a silicate-rich melt due to clay-minerals melting(Brearley amp Rubie 1990) in the external part ofthe ring in contact with a second inner melt formedby preexisting gehlenite or the eutectic CaO-SiO2would result in viscous finger development Thefractal geometry of the developing viscous fingersmay have enhanced diffusion of mobile ions suchas Ca toward silicate reactive sites Both EDXdata showing Mg enrichment in the dolomite core(Fig 5f) and the compositional trend towardsMgO-rich terms in the CMS diagram (Fig 6b)confirm that Ca is preferentially leached out of theformer dolomite grain Finger formation willresult in a significant increase in interface surfacearea resulting in higher reaction rates Another

alternative explanation for this fingered geometrydevelopment involves partial melting at the car-bonate-silicate interface and mass transport (ieviscous flow) through the clay-quartz porousmatrix resulting in the so-called ldquoscalloped struc-turerdquo proposed by Ortoleva et al (1987) for reac-tion-diffusion processes While the viscous fingermodel requires two melts the second model onlyrequires one melt It is not clear if at any momentconditions for two melt occurrence are reachedHence the second model seems more plausible

In any case the process which is originally dif-fusion driven is transformed into a mass (ie fin-gering) plus diffusion transport which greatlyenhances Ca-silicate formation This is consistentwith the thick reaction rims (up to 250 microm thicksee Fig 5f) that surround former carbonate grainsThese thick Ca-silicate shells would hardly devel-op just by diffusion in the short time-scale of theexperiment (3 hours at the peak T) In fact at800degC (conditions where melting would be verylimited) Ca (and Mg)-silicate rims formed aroundcalcite or dolomite are 2-5 microm thick At 1100degCwhen melting is extensive the reaction rate is sohigh that rims are two orders of magnitude thick-er Therefore mass transport (ie viscous flow)seems to have played a key role in high-T Ca-(Mg-) silicate formation Mass transport is also consis-tent with observation of some areas with differentcomposition (unreacted silicates) trapped betweenfingers resulting in the island geometry observedin Fig 5g It could be expected that without fin-gering the process of Ca- or (Ca-Mg)-silicate for-mation would be self-limiting ie once a rim ofthe newly formed silicate appears diffusion wouldbe jeopardized This shell would preclude or limitthe transport of Ca (andor Mg) from the carbon-ate or it would limit the transport of the lessmobile Si (and Al) from the surrounding silicatetowards the reaction front Therefore it is neces-sary that one of the above described transport-reaction mechanisms takes place This may alsoexplain why phase formation takes place at lowerT than reported previously The presence of multi-ple interfaces among different crystals mayenhance finger development resulting in kineticchanges favoring phase formation at lower T

5 Conclusions

It is observed that even though the overallcomposition of a ceramic material does not under-go major changes upon firing significant miner-alogical and textural modifications do occur

632

These changes are connected with two reactiontypes a) solid-state replacement of a phase by anew one with little or no compositional variation(ie the muscovite and illite case) and minimizingenergy changes by preserving the original textureand orientation of the replaced phase and b) reac-tion-diffusion at the interface between two miner-al grains with contrasting composition (iecarbonate-silicate interfaces) The latter reaction isenhanced if massviscous transport resulting frompartial melting and finger structure developmentoccurs Non-carbonated clay-rich materials (G)show phyllosilicate destruction at T ranging from700 to 900degC followed by vitrification which issignificant at T gt 1000degC Carbonate-rich materi-als (V) show major phase transformation uponphyllosillicates destruction including new Ca (orMg) silicates and aluminum-silicates formation Inthis later case the vitrification starts at lower Tand it is limited at higher T Therefore it is possi-ble to establish the firing T of ceramic materialswith composition broadly similar to that in ourwork based on the mineralogy Thus a) presenceof phyllosilicates indicates a low 700 to 900degC fir-ing T for the dehydroxylated illite (or muscovite)phase b) dolomite disappears at sup3 700degC whilecalcite is still present at 800degC c) gehleniteappears at 800degC reducing its concentration above1000degC d) diopside and wollastonite appear at800degC (in very low concentrations) as reactionrims between carbonates and silicates Formationof these minerals includes ldquofingeringrdquo develop-ment e) mullite appears at T gt 800degC while littleamounts of sanidine form as microcline disap-pears and f) anorthite develops at T gt1000degC incarbonate-rich bricks

When considering equilibrium thermodynamicdata and calculated reaction T for most of theabove mentioned phases and reactions (Holland ampPowell 1985 and references herein) it isobserved that rapid heating induces significantoverstepping Therefore phases formed are inmost cases metastable a fact consistent with anal-yses of pyrometamorphic assemblages (Brearleyamp Rubie 1990 and references herein)

The formation of non-equilibrium fractal mor-phologies at the carbonate-silicate reaction inter-faces may explain why high-T silicates incarbonate-rich ceramics develop very fast Thisgrain-to-grain reaction model may help explainhigh-T silicate formation in nature (eg contactmetamorphism pyrometamorphism in xenoliths)

The gained knowledge on mineral transforma-tion and the dynamics of interface mass and diffu-sion transport facilitated through ldquofingeringrdquo or

ldquoscallopingrdquo opens new ways to understand thecomplexities of ceramic processing in particularand high-T mineral transformation in general bothin natural (ie geologic settings) and artificialcontexts (eg archaeometric analysis of ancientceramics or the design of new bricks ndashreplicas ofancient onesndash for architectural heritage conserva-tion) However care must be taken when extrapo-lating these results to other raw materialcompositions andor ceramic processing methods

Acknowledgements This work has been finan-cially supported by the CGICYT projectMAT2000-1457 and the Research Group RNM0179 of the Junta de Andaluciacutea (Spain) We thankthe Centro de Instrumentacioacuten Cientifica (CIC) ofthe Granada University for allowing us to use theirSEM-EDX EMPA and XRF facilities We thankK Elert JM Garciacutea-Ruiz M Rodriguez-Gallego and A Sanchez-Navas for fruitful discus-sions and comments We also thank the in-depthreview by Dr F Goumltz-Neunhoeffer and an anony-mous referee

References

Barahona E Huertas F Pozzuoli A Linares J(1985) Firing properties of ceramic clays fromGranada province Spain Miner Petrogr Acta 29A577-590

Boynton RS (1980) Chemistry and technology oflime and limestone 2nd Ed Wiley New York

Brearley AJ (1986) An electron optical study of mus-covite breakdown in pelitic xenoliths duringpyrometamorphism Miner Mag 357 385-397

Brearley AJ amp Rubie DC (1990) Effects of H2O onthe disequilibrium breakdown of muscovite +quartz J Petrol 31 925-956

Butterworth B amp Honeyborne DB (1952) Bricksand clays of the hasting beds Trans Brit CeramSoc 51 211-259

Dondi M Ercolani G Fabbri B Marsigli M (1998)An approach to the chemistry of pyroxenes formedduring the firing of Ca-rich silicate ceramics ClayMinerals 33 443-452

Duminuco P Riccardi MP Messiga B Setti M(1996) Modificazioni tessiturali e mineralogichecome indicatori della dinamica del processo di cot-tura di manufatti ceramici Ceramurgia 26 5 281-288

Evans JL amp White J (1958) Further studies of thethermal decomposition of clays Trans Brit CeramSoc 57 6 298

Everhart JO (1957) Use of auxiliary fluxes to improvestructural clay bodies Bull Am Ceram Soc 36268-271

Carbonate and silicate phase reactions during ceramic firing 633

a rise of the background noise in the XRD patternat T gt 900degC in the G samples and at T gt 800degCin the V samples

34 SEM-EDX analysis

Significant textural changes most evident withrespect to pore morphology and volume weredetected upon firing An overall porosity reductionis observed at T gt 1000degC when the vitreous phasefills the pores

Samples fired at 700 and 800degC still preservethe laminar habit of phyllosilicates although mus-covite crystals clearly exfoliate along basal planes

most probably due to dehydroxylation (Fig 4a)The interlocking among particles is limited (Fig4d) V samples show some cracks At these T noclear evidence of sintering or partial melting isdetected However there is indirect evidence (ieincrease in background noise of XRD patterns)that some vitrification is reached at T lt 900degC incarbonate-rich V samples Vitrification is clearlyobserved in all samples at T gt 900degC ie phyl-losilicates appear deformed (their edges turnsmooth) andor aggregated (Fig 4b and 4e) andthe pores turn ellipsoidal with smooth surfaces AtT gt 1000degC these effects are less extended in thecarbonate-rich V samples (Fig 4c) than in the sil-

Carbonate and silicate phase reactions during ceramic firing 625

Fig 3 Guadix (a) and Viznar (b) samples powder X-ray difraction patterns Legend (mineral symbols after Kretz1983) Sm = smectite Ill = illite Pg = paragonite Qtz = quartz Cal = calcite Kln = kaolinite Phy = phyllosilicatesFs = feldspar Dol = dolomite Gh = gehlenite Hem = hematite Wo = wollastonite Di = diopside An = anorthiteMul = mullite Sa = sanidine CuKa X-ray radiation l = 15064 Aring

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

icate-rich G samples (Fig 4f) A higher degree ofparticle interlocking results in a porosity reduc-tion At 1100degC vitrification is significant in allsamples regardless of initial compositional differ-ences Pores coalesce as spherical cells due to par-tial or extended melting of clay particles in thematrix

BSE images of mineral and textural evolutionof samples with high carbonate content (ie Vsamples which undergo the most significant min-eralogical changes) give an interesting picture oflocal phase transformations taking place at grainboundaries No textural or mineralogical changescan be detected in brick fired at 700degC (Fig 5a)Carbonates comprise the larger-size fraction inthese samples which allows easy detection ofhigh-T reaction rims In fact at 800degC 2-5 micromthick reaction rims are observed developing at thecalcite-quartzphyllosilicate interface (Fig 5b)Rims composition corresponds to a calcium sili-cate phase presumably wollastonite (EDXresults) A dark rim surrounded by a light outershell is observed where dolomite was originallypresent (Fig 5c) However no compositional vari-ation among internal (greyish) and external(white) rims is detected We were not able tounambiguously detect the presence of gehlenite

within these reaction rims (ie some EDX analy-ses of 2-5 microm thick rims at the carbonate-silicateinterface showed Al together with Ca and Si butpossible contamination from underlying phyllosil-icates was not discarded) However the presenceof gehlenite was detected by XRD at this T(800degC) Muscovite phenocrysts (Fig 5d) show amorphology and composition typical for unalteredwhite mica At 1100degC muscovite crystals showsecondary bubbles development between layers(Fig 5e) The mica composition is homogeneous(no BSE image gray-level differences along thecrystals) and it is similar to that of an unchangedmuscovite but with slightly lower K concentra-tion At 1100degC carbonates (ie dolomite in Fig5f) shows an outer thick white Ca- and Si-rich rim(ie wollastonite) which includes small (micronsized) bright-white Ca- Si- and Al-rich particles(ie gehlenite or plagioclase) There is a strikingfeature of this reaction rim it shows a fingeredgeometry at the outer edge The darker dolomitecore shows a significant Mg enrichment

35 EMPA analysis

Due to the fact that only silicate phases in sam-ples with carbonates undergo mayor composition-

626

Fig 4 SEM secondary electron photomicrographs of Viznar samples fired at 800degC (a) 1000degC (b) and 1100degC(c) and Guadix samples fired at 700degC (d) 900degC (e) and 1000degC (f)

al changes upon firing only V samples were ana-lyzed using this technique In particular composi-tional changes of phyllosilicates and carbonateswere studied using polished thin sections of firedV bricks (one section per selected target T ie700 800 1100degC)

Phyllosilicates (Table 2) show a muscoviticcomposition at 700degC where some K is replaced byNa even though the overall composition does notseem to correspond to paragonite Neverthelessparagonite was detected in the raw clay (Fig 3b)

Fe Mg and Ti replace some Al in the octahedralsheets Replacement of OH- by F- is almost absentAt 800degC a slight increase in Fe content is detectedOnly analyses M9 and M10 show higher concen-trations of Na and Al At 1100degC the amount of Kis reduced while Ca and Si content is increasedThe total oxide content is close to 100 a valueconsistent with complete dehydroxylation of thephyllosilicates Analyses M11-M14 in Table 2 arein agreement with high-T phengite compositionaldata from Worden et al (1987)

Carbonate and silicate phase reactions during ceramic firing 627

Fig 5 BSE images and EDX analyses of carbonate-rich Viznar samples a) fired at 700degC b) fired at 800degC whenWo (EDX analysis) formed at the Cal- silicates interface c) detail of former carbonate grain showing a white outerreaction rim (Wo or Gh) a crack and an inner core with an outer darker shell (see text for details) d) Ms (see EDXanalysis) at 700degC e) former Ms after transformation into a Fs + Mul + melt mixture (see EDX analysis) f) DolPhy(+ Qtz) interface showing a reaction rim with fingered geometry formed at 1100degC g) detail of the outer reactionrim in previous figure EDX analysis shows the rim composition (ie Wo)

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

Calcite (Table 3) shows no chemical changesat 700degC The maximum MgO content is 104 wt Trace amounts of Fe Si Al and Mn are alsodetected in calcite (as well as in dolomite) At800degC both carbonates are fully transformed eitherinto burnt lime (CaO) or into a mixture of burtlime and periclase (MgO) (eg analysis C14)Gehlenite-like compositions are detected in someanalyses (eg C10) At 1100degC the presence ofnewly formed phases in particular wollastonite isclearly confirmed (eg C17) Ca content in thelime plus periclase mixture is reduced significant-ly due to the incorporation of this element intonewly formed high-T silicate phases As T increas-es a clear compositional trend toward Si- and Al-rich calcium phases (ie toward gehlenite andanorthite) is observed (Fig 6a) a fact consistentwith XRD and SEM-EDX results A clear trendtoward wollastonite-like compositions as well as a

trend toward MgO rich composition is alsoobserved as T increases (Fig 6b) Some EMPAanalyses suggest a larnite -like compositionHowever XRD results do not show the presenceof larnite

36 XRF analysis

According to the classification proposed forclay materials in ceramic industry by Fabbri ampFiori (1985) G raw material composition falls intothe quartz-feldspar sands category while V sam-ples fall out of any defined category due to theirlow silica and high calcium contents (Table 4) Asexpected firing does not affect major elementconcentration in the brick The only significantchange is detected in the loss on ignition values Itis evidenced that G samples lost almost all waterbetween 700 and 900degC while V samples show a

628

Table 2 EMPA analysis results of phyllosilicates composition (in wt )

Table 3 Composition of selected carbonate grains ( EMPA results in wt )

more gradual loss of water and CO2 as T increases(Table 4)

4 Discussion

Firing up to 700degC induce no significant min-eralogicaltextural transformations in both thecarbonate-rich V and the silicate-rich G sampleswith the exception of the clay minerals Clay min-erals structure is reported to collapse due to dehy-droxylation to resemble an illite-like structure(XRD data) when a T between 450 and 550degC isreached (Evans amp White 1958) A dehydroxylatedphyllosilicate phase structurally different to thehydrated one (Guggenheim et al 1987) is report-ed to exist up to 950degC when complete breakdownof the dehydroxylated illite occurs (Peters amp Iberg1978) Our data indicate that dehydroxylation isnot completed at T ~ 900degC which indicates thatthe kinetics of this process is slower than previ-ously estimated (Evans amp White 1958) (OH)-

released at T gt 700degC may play a significant rolein texture and mineral evolution at high-T as willbe discussed later On the other hand muscoviteshows a slight K deficit at 700degC probably due tothe high diffusion capacity of this element alongmuscovite (001) basal planes at T gt 500degC(Sanchez-Navas amp Galindo-Zaldivar 1993)Sanchez-Navas (1999) showed high-T K-diffu-sion-out of mica is induced by water adsorbedalong mica basal planes resulting from phyllosili-cate dehydroxylation

XRD data shows that dolomite most intenseBragg peak in V bricks significantly reduced itsintensity due to thermal decomposition at lower Tthan calcite main Bragg peak (Fig 3b) Dolomite

Carbonate and silicate phase reactions during ceramic firing 629

Table 4 Bulk composition of Guadix (G) and Viznar (V) samples (XRF results in wt )

LOI loss on ignition

Fig 6 a) ACS (Al2O3-CaO-SiO2) and b) CMS (CaO-MgO-SiO2) compositional diagrams of carbonates andreaction products (EMPA results)

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

starts to decompose according to (1) at ~ 700degC (1atm pressure) while calcite decomposes at higherT (830 ndash 870degC) according to (3) (Boynton 1980)Shoval (1988) proposed a two steps thermaldecomposition for dolomite involving formationof periclase and calcite (2) followed by calcitedecomposition at higher T (3)

dolomite lime periclaseCaMg(CO3)2 reg CaO + MgO + 2CO2 (1)

calcite CaMg(CO3)2 reg CaCO3 + MgO + CO2 (2)

CaCO3 reg CaO + CO2 (3)

However our XRD results showing no increase incalcite 104 Bragg peak intensity as dolomitedecomposes at 700degC (Fig 3b) are not consistentwith Shovalacutes two-steps process

Carbonate decomposition results in shrinkageof the brick (Barahona et al 1985) Upon storageat 22degC and 55 RH burnt lime (CaO) readilytransforms into portlandite according to the fol-lowing reaction (Boynton 1980)

portlanditeCaO + H2O reg Ca(OH)2 (4)

This process generates crystallization pressure inconfined spaces (ie the brick pores originallyoccupied by CaO) resulting in crack development(as observed in Fig 2a) The dark-color rimaround former carbonate grains (Fig 5c) is proba-bly due to limited portlandite carbonation in thepresence of CO2 also resulting in volume increase(Moorehead 1985) and crack development(Butterworth amp Honeyborne 1952) As a commonpractice to avoid these undesired side-effectsbricks that contain carbonates were traditionallyimmersed in water to dissolve and eliminate exist-ing CaO or Ca(OH)2

This effect is less evident in the case of MgOPericlase in the presence of H2O (room T) under-goes a slow transformation (it takes months oreven years Webb 1952) into brucite (Mg(OH)2)and may eventually transform into hydromagne-site (Mg5(CO3)4(OH)24H2O) according to thefollowing reactions (Garavelli et al 1990)

MgO + H2O reg Mg(OH)2 (5)

hydromagnesite5Mg(OH)2 + 4CO2 reg Mg5(CO3)4(OH)24H2O (6)

EMPA results indicate the presence of Ca and Mgoxides however our XRD results do not clearlyshow the presence of Ca and Mg oxides or hydrox-ides mainly due to their low concentrations (lt 10wt ) and because their main diffraction peaks aremasked by other major phase peaks (eg quartzfeldspars hematites) The hydromagnesite mainBragg peak at 581 Aring is not present either

At T sup3 800degC and in the absence of carbonateschanges are constrained to textural modificationsand reactions where a low-T phase is decomposedand fully transformed into a high-T phase withoutmajor compositional changes (Riccardi et al1999) In particular muscovite undergoes a solid-state phase change into a mixture of mullite and K-feldspar (or a melt) at T ranging from 800 to1000degC The composition of high-T muscovite isclose to K-feldspar (EMPA results) although OManalysis points to the presence of mullite Brearley(1986) and Worden et al (1987) showed fine inter-growth of alternating nanometer up to a fewmicrometers thick bands of K-feldspar (or a meltwith composition close to that of K-feldspar) andmullite in pyrometamorphic muscovite which canexplain our EMPA results

The most significant textural and mineralogi-cal changes are observed in samples with carbon-ates when fired at T gt 800degC Also melting T andextent seems to be connected with the presence orabsence of carbonates In general melting starts atlower T (~ 800degC) when carbonates are present(Tite amp Maniatis 1975) Ca and Mg from carbon-ates may act as melting agents (Segnit ampAnderson 1972) but they are reported to some-how limit the extent of vitrification at T gt 1000degC(Everhart 1957 Nuacutentildeez et al 1992) In fact weobserved that vitrification is more extended athigher T when no carbonates exist This may beexplained considering that the phyllosilicate in thematrix of non-carbonate G samples not contribut-ing to the formation of high-T Ca (or Mg) silicatesand releasing significant amounts of H2O due todehydroxylation they can contribute to extensivemelting at high-T as demonstrated by Brearley ampRubie (1990) On the other hand the higher SiO2content in these samples provides a higher amountof potential silicate-rich melt

BSE images of reaction rims between carbon-ates and silicates show that formation of Ca-sili-cates starts at 800degC This is confirmed by XRD(Fig 3b) and EMPA analyses (Table 3) Gehlenitestarts to form at this T by grain-boundary reactionbetween CaO Al2O3 and SiO2 the first derivingfrom former carbonates and the later two fromalready dehydroxylated phyllosilicates (ie a

630

somehow amorphous mixture of 3SiO2Al2O3according to Peters amp Iberg 1978) as follows

illite calcite gehleniteKAl2(Si3Al)O10(OH)2 + 6CaCO3 reg 3Ca2Al2SiO7+ 6CO2 + 2H2O + K2O + 3SiO2 (7)

The formation of gehlenite is striking becauseone could expect the formation of an Al-richpyroxene such as fassaite (Dondi et al 1998)Nevertheless it should be indicated that gehleniteformation is a good example of the so-calledGoldsmithacutes ldquoease of crystallizationrdquo principlewhich states that a silicate phase with all Al in 4-coordination is easier to form (at least in the lab)than a phase that includes Al in both 4- and 6-coor-dination when (OH)- are not present (Goldsmith1953) This is due to the higher energy barrier nec-essary to overcome when Al has to enter in a 6-coordination if compared with a 4-coordination If(OH)- are present this energy barrier is reduceddue to the smaller electrostatic repulsion among(OH)- groups than among O2- in octahedral coor-dination Gehlenite has all Al in tetrahedral coor-dination (Warren 1930) therefore crystallizingpreferentially if compared with other anhydrousCa-Al silicates This simple principle also appliesto the high-T formation of other anhydrous Al-sil-icates as we will discuss later

Wollastonite formed at the carbonate-quartzinterface through the following reaction

wollastoniteCaO + SiO2 reg CaSiO3 (8)

Wollastonite and gehlenite appear at 800degC aT ~100degC lower than previously reported (Petersamp Iberg 1978 Dondi et al 1998) This is proba-bly due to a) the low resolution of the analyticaltechniques used in the past (ie XRD) which didnot allow the detection of the small initial reactionrim between carbonates and silicates as it is evi-denced by BSE images and EMPA analysesandor b) most previous works on high-T phaseformation in ceramic processing use single com-ponents as reactants without considering addition-al phases or previous reaction by-products (Petersamp Iberg 1978) We observed that in a closer-to-reality multicomponent system where interactionsbetween reactants and products occur differentformation T are obtained

At the dolomite-quartz interface diopsidestarts to form at 900degC by the following reaction

dolomite diopside2SiO2 + CaMg(CO3)2 reg CaMgSi2O6 + 2CO (9)

At this T anorthite starts to form at the expensesof calcite and illite (plus quartz)

illite calciteKAl2(Si3Al)O10(OH)2 + 2CaCO3 + 4SiO2 regK-feldspar anorthite

2KAlSi3O8 + 2Ca2Al2SiO8 + 2CO2 + H2O (10)

At higher temperatures (1000 to 1100degC)phyllosilicates have already disappeared in allsamples (V and G) being transformed into sani-dine and mullite plus a melt according to the fol-lowing reaction

illite (muscovite) sanidine 3KAl2(Si3Al)O10(OH)2 + 2SiO2 reg 3KAlSi3O8 +

mulliteAl6Si2O13 + 3H2O + melt (11)

Formation of a melt during the previous reaction(Brearley amp Rubie 1990) is consistent with theobserved ldquocellular structurerdquo of muscovite (Tite ampManiatis 1975)

Brearley (1986) proposed a somehow differentreaction for mullite formation at the expenses ofphengitic muscovite resulting in K-feldspar +mullite + biotite No biotite is detected in ourexperiments most probably due to the low Fe con-tent in V and G micas Brearley amp Rubie (1990)experimentally demonstrated that high-T mus-covite breakdown in the presence of quartz isinfluenced by the crystal size The smallest crys-tals have a strong tendency to react with H2Oresulting in enhanced melt formation while largemuscovite crystals form limited amounts of meltThis may account for the significant vitrificationof the matrix (specially in G samples) while largemuscovite grains (up to some mm in size) still donot show significant melting

Mullite which according to OM analysis firstappears at 800degC is the second most abundantphase (after quartz) at 1100degC in the richer inphyllosilicate G samples It is interesting to men-tion that K-feldspar progressively disappears as Tincreases This is not consistent with publisheddata (Maggetti 1982) It seems that low-T K-feldspar (microcline) is not stable at high-T there-fore it transforms into a high-T K-feldspar (iesanidine) or reacts with lime to form anorthiteBoth sanidine and anorthite are detected in thehigh-T bricks However a higher concentration ofsanidine should be expected to form speciallyconsidering reaction (11) A possible explanationfor the low K-feldspar content at high-T is that K-feldspar might not have enough time to develop itscrystalline structure (sanidine) therefore being

Carbonate and silicate phase reactions during ceramic firing 631

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

concentrated as an amorphous (sanidine-like)phase responsible in part for the rise in the back-ground noise in the XRD results Mullite forma-tion seems to be kinetically faster resulting in awell-developed crystalline phase which givessharp diffraction peaks Goldsmith (1953) indi-cates that mullite could be another example of hisldquosimplexity principlerdquo for the ldquoeaserdquo of crystal-lization if compared with other aluminum sili-cates In fact the mullite structure has a largeramount of Al in 4-coordination and less Al in 6-coordination than sillimanite (Rehak et al 1998)for instance which may account for the abundanceof the former and the non-existence of the latter inthe studied bricks

In carbonate-rich V samples fired at T gt1000degC a significant reduction of gehlenite maindiffraction peak is observed According to Petersamp Iberg (1978) this is due to the reaction of thisphase with quartz forming wollastonite and anor-thite This is consistent with the increase in themain Bragg peak intensity of these later mineralsat 1100degC

At the dolomite-silicate (ie Qtz and Phy)interface it is common to observe wollastoniteeither newly formed or resulting from the destruc-tion of gehlenite The texture of the interface israther complex Protrusion or fingers from thereaction zone penetrate into areas were completetransformation into wollastonite has occurred Theinterface geometry resembles the so-called ldquovis-cous fingersrdquo (Garcia-Ruiz 1992) occurringwhen two fluids or semi-plastic solids of differentcomposition and viscosity are placed in contactInitially the contact zone is a flat surface but lateron it reaches a fingered fractal geometry(Nittmamm et al 1985) In the carbonate-silicateinterface a similar process may have taken placewhere mass transport of the reactants may havebeen enhanced by partial melting The formationof a silicate-rich melt due to clay-minerals melting(Brearley amp Rubie 1990) in the external part ofthe ring in contact with a second inner melt formedby preexisting gehlenite or the eutectic CaO-SiO2would result in viscous finger development Thefractal geometry of the developing viscous fingersmay have enhanced diffusion of mobile ions suchas Ca toward silicate reactive sites Both EDXdata showing Mg enrichment in the dolomite core(Fig 5f) and the compositional trend towardsMgO-rich terms in the CMS diagram (Fig 6b)confirm that Ca is preferentially leached out of theformer dolomite grain Finger formation willresult in a significant increase in interface surfacearea resulting in higher reaction rates Another

alternative explanation for this fingered geometrydevelopment involves partial melting at the car-bonate-silicate interface and mass transport (ieviscous flow) through the clay-quartz porousmatrix resulting in the so-called ldquoscalloped struc-turerdquo proposed by Ortoleva et al (1987) for reac-tion-diffusion processes While the viscous fingermodel requires two melts the second model onlyrequires one melt It is not clear if at any momentconditions for two melt occurrence are reachedHence the second model seems more plausible

In any case the process which is originally dif-fusion driven is transformed into a mass (ie fin-gering) plus diffusion transport which greatlyenhances Ca-silicate formation This is consistentwith the thick reaction rims (up to 250 microm thicksee Fig 5f) that surround former carbonate grainsThese thick Ca-silicate shells would hardly devel-op just by diffusion in the short time-scale of theexperiment (3 hours at the peak T) In fact at800degC (conditions where melting would be verylimited) Ca (and Mg)-silicate rims formed aroundcalcite or dolomite are 2-5 microm thick At 1100degCwhen melting is extensive the reaction rate is sohigh that rims are two orders of magnitude thick-er Therefore mass transport (ie viscous flow)seems to have played a key role in high-T Ca-(Mg-) silicate formation Mass transport is also consis-tent with observation of some areas with differentcomposition (unreacted silicates) trapped betweenfingers resulting in the island geometry observedin Fig 5g It could be expected that without fin-gering the process of Ca- or (Ca-Mg)-silicate for-mation would be self-limiting ie once a rim ofthe newly formed silicate appears diffusion wouldbe jeopardized This shell would preclude or limitthe transport of Ca (andor Mg) from the carbon-ate or it would limit the transport of the lessmobile Si (and Al) from the surrounding silicatetowards the reaction front Therefore it is neces-sary that one of the above described transport-reaction mechanisms takes place This may alsoexplain why phase formation takes place at lowerT than reported previously The presence of multi-ple interfaces among different crystals mayenhance finger development resulting in kineticchanges favoring phase formation at lower T

5 Conclusions

It is observed that even though the overallcomposition of a ceramic material does not under-go major changes upon firing significant miner-alogical and textural modifications do occur

632

These changes are connected with two reactiontypes a) solid-state replacement of a phase by anew one with little or no compositional variation(ie the muscovite and illite case) and minimizingenergy changes by preserving the original textureand orientation of the replaced phase and b) reac-tion-diffusion at the interface between two miner-al grains with contrasting composition (iecarbonate-silicate interfaces) The latter reaction isenhanced if massviscous transport resulting frompartial melting and finger structure developmentoccurs Non-carbonated clay-rich materials (G)show phyllosilicate destruction at T ranging from700 to 900degC followed by vitrification which issignificant at T gt 1000degC Carbonate-rich materi-als (V) show major phase transformation uponphyllosillicates destruction including new Ca (orMg) silicates and aluminum-silicates formation Inthis later case the vitrification starts at lower Tand it is limited at higher T Therefore it is possi-ble to establish the firing T of ceramic materialswith composition broadly similar to that in ourwork based on the mineralogy Thus a) presenceof phyllosilicates indicates a low 700 to 900degC fir-ing T for the dehydroxylated illite (or muscovite)phase b) dolomite disappears at sup3 700degC whilecalcite is still present at 800degC c) gehleniteappears at 800degC reducing its concentration above1000degC d) diopside and wollastonite appear at800degC (in very low concentrations) as reactionrims between carbonates and silicates Formationof these minerals includes ldquofingeringrdquo develop-ment e) mullite appears at T gt 800degC while littleamounts of sanidine form as microcline disap-pears and f) anorthite develops at T gt1000degC incarbonate-rich bricks

When considering equilibrium thermodynamicdata and calculated reaction T for most of theabove mentioned phases and reactions (Holland ampPowell 1985 and references herein) it isobserved that rapid heating induces significantoverstepping Therefore phases formed are inmost cases metastable a fact consistent with anal-yses of pyrometamorphic assemblages (Brearleyamp Rubie 1990 and references herein)

The formation of non-equilibrium fractal mor-phologies at the carbonate-silicate reaction inter-faces may explain why high-T silicates incarbonate-rich ceramics develop very fast Thisgrain-to-grain reaction model may help explainhigh-T silicate formation in nature (eg contactmetamorphism pyrometamorphism in xenoliths)

The gained knowledge on mineral transforma-tion and the dynamics of interface mass and diffu-sion transport facilitated through ldquofingeringrdquo or

ldquoscallopingrdquo opens new ways to understand thecomplexities of ceramic processing in particularand high-T mineral transformation in general bothin natural (ie geologic settings) and artificialcontexts (eg archaeometric analysis of ancientceramics or the design of new bricks ndashreplicas ofancient onesndash for architectural heritage conserva-tion) However care must be taken when extrapo-lating these results to other raw materialcompositions andor ceramic processing methods

Acknowledgements This work has been finan-cially supported by the CGICYT projectMAT2000-1457 and the Research Group RNM0179 of the Junta de Andaluciacutea (Spain) We thankthe Centro de Instrumentacioacuten Cientifica (CIC) ofthe Granada University for allowing us to use theirSEM-EDX EMPA and XRF facilities We thankK Elert JM Garciacutea-Ruiz M Rodriguez-Gallego and A Sanchez-Navas for fruitful discus-sions and comments We also thank the in-depthreview by Dr F Goumltz-Neunhoeffer and an anony-mous referee

References

Barahona E Huertas F Pozzuoli A Linares J(1985) Firing properties of ceramic clays fromGranada province Spain Miner Petrogr Acta 29A577-590

Boynton RS (1980) Chemistry and technology oflime and limestone 2nd Ed Wiley New York

Brearley AJ (1986) An electron optical study of mus-covite breakdown in pelitic xenoliths duringpyrometamorphism Miner Mag 357 385-397

Brearley AJ amp Rubie DC (1990) Effects of H2O onthe disequilibrium breakdown of muscovite +quartz J Petrol 31 925-956

Butterworth B amp Honeyborne DB (1952) Bricksand clays of the hasting beds Trans Brit CeramSoc 51 211-259

Dondi M Ercolani G Fabbri B Marsigli M (1998)An approach to the chemistry of pyroxenes formedduring the firing of Ca-rich silicate ceramics ClayMinerals 33 443-452

Duminuco P Riccardi MP Messiga B Setti M(1996) Modificazioni tessiturali e mineralogichecome indicatori della dinamica del processo di cot-tura di manufatti ceramici Ceramurgia 26 5 281-288

Evans JL amp White J (1958) Further studies of thethermal decomposition of clays Trans Brit CeramSoc 57 6 298

Everhart JO (1957) Use of auxiliary fluxes to improvestructural clay bodies Bull Am Ceram Soc 36268-271

Carbonate and silicate phase reactions during ceramic firing 633

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

icate-rich G samples (Fig 4f) A higher degree ofparticle interlocking results in a porosity reduc-tion At 1100degC vitrification is significant in allsamples regardless of initial compositional differ-ences Pores coalesce as spherical cells due to par-tial or extended melting of clay particles in thematrix

BSE images of mineral and textural evolutionof samples with high carbonate content (ie Vsamples which undergo the most significant min-eralogical changes) give an interesting picture oflocal phase transformations taking place at grainboundaries No textural or mineralogical changescan be detected in brick fired at 700degC (Fig 5a)Carbonates comprise the larger-size fraction inthese samples which allows easy detection ofhigh-T reaction rims In fact at 800degC 2-5 micromthick reaction rims are observed developing at thecalcite-quartzphyllosilicate interface (Fig 5b)Rims composition corresponds to a calcium sili-cate phase presumably wollastonite (EDXresults) A dark rim surrounded by a light outershell is observed where dolomite was originallypresent (Fig 5c) However no compositional vari-ation among internal (greyish) and external(white) rims is detected We were not able tounambiguously detect the presence of gehlenite

within these reaction rims (ie some EDX analy-ses of 2-5 microm thick rims at the carbonate-silicateinterface showed Al together with Ca and Si butpossible contamination from underlying phyllosil-icates was not discarded) However the presenceof gehlenite was detected by XRD at this T(800degC) Muscovite phenocrysts (Fig 5d) show amorphology and composition typical for unalteredwhite mica At 1100degC muscovite crystals showsecondary bubbles development between layers(Fig 5e) The mica composition is homogeneous(no BSE image gray-level differences along thecrystals) and it is similar to that of an unchangedmuscovite but with slightly lower K concentra-tion At 1100degC carbonates (ie dolomite in Fig5f) shows an outer thick white Ca- and Si-rich rim(ie wollastonite) which includes small (micronsized) bright-white Ca- Si- and Al-rich particles(ie gehlenite or plagioclase) There is a strikingfeature of this reaction rim it shows a fingeredgeometry at the outer edge The darker dolomitecore shows a significant Mg enrichment

35 EMPA analysis

Due to the fact that only silicate phases in sam-ples with carbonates undergo mayor composition-

626

Fig 4 SEM secondary electron photomicrographs of Viznar samples fired at 800degC (a) 1000degC (b) and 1100degC(c) and Guadix samples fired at 700degC (d) 900degC (e) and 1000degC (f)

al changes upon firing only V samples were ana-lyzed using this technique In particular composi-tional changes of phyllosilicates and carbonateswere studied using polished thin sections of firedV bricks (one section per selected target T ie700 800 1100degC)

Phyllosilicates (Table 2) show a muscoviticcomposition at 700degC where some K is replaced byNa even though the overall composition does notseem to correspond to paragonite Neverthelessparagonite was detected in the raw clay (Fig 3b)

Fe Mg and Ti replace some Al in the octahedralsheets Replacement of OH- by F- is almost absentAt 800degC a slight increase in Fe content is detectedOnly analyses M9 and M10 show higher concen-trations of Na and Al At 1100degC the amount of Kis reduced while Ca and Si content is increasedThe total oxide content is close to 100 a valueconsistent with complete dehydroxylation of thephyllosilicates Analyses M11-M14 in Table 2 arein agreement with high-T phengite compositionaldata from Worden et al (1987)

Carbonate and silicate phase reactions during ceramic firing 627

Fig 5 BSE images and EDX analyses of carbonate-rich Viznar samples a) fired at 700degC b) fired at 800degC whenWo (EDX analysis) formed at the Cal- silicates interface c) detail of former carbonate grain showing a white outerreaction rim (Wo or Gh) a crack and an inner core with an outer darker shell (see text for details) d) Ms (see EDXanalysis) at 700degC e) former Ms after transformation into a Fs + Mul + melt mixture (see EDX analysis) f) DolPhy(+ Qtz) interface showing a reaction rim with fingered geometry formed at 1100degC g) detail of the outer reactionrim in previous figure EDX analysis shows the rim composition (ie Wo)

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

Calcite (Table 3) shows no chemical changesat 700degC The maximum MgO content is 104 wt Trace amounts of Fe Si Al and Mn are alsodetected in calcite (as well as in dolomite) At800degC both carbonates are fully transformed eitherinto burnt lime (CaO) or into a mixture of burtlime and periclase (MgO) (eg analysis C14)Gehlenite-like compositions are detected in someanalyses (eg C10) At 1100degC the presence ofnewly formed phases in particular wollastonite isclearly confirmed (eg C17) Ca content in thelime plus periclase mixture is reduced significant-ly due to the incorporation of this element intonewly formed high-T silicate phases As T increas-es a clear compositional trend toward Si- and Al-rich calcium phases (ie toward gehlenite andanorthite) is observed (Fig 6a) a fact consistentwith XRD and SEM-EDX results A clear trendtoward wollastonite-like compositions as well as a

trend toward MgO rich composition is alsoobserved as T increases (Fig 6b) Some EMPAanalyses suggest a larnite -like compositionHowever XRD results do not show the presenceof larnite

36 XRF analysis

According to the classification proposed forclay materials in ceramic industry by Fabbri ampFiori (1985) G raw material composition falls intothe quartz-feldspar sands category while V sam-ples fall out of any defined category due to theirlow silica and high calcium contents (Table 4) Asexpected firing does not affect major elementconcentration in the brick The only significantchange is detected in the loss on ignition values Itis evidenced that G samples lost almost all waterbetween 700 and 900degC while V samples show a

628

Table 2 EMPA analysis results of phyllosilicates composition (in wt )

Table 3 Composition of selected carbonate grains ( EMPA results in wt )

more gradual loss of water and CO2 as T increases(Table 4)

4 Discussion

Firing up to 700degC induce no significant min-eralogicaltextural transformations in both thecarbonate-rich V and the silicate-rich G sampleswith the exception of the clay minerals Clay min-erals structure is reported to collapse due to dehy-droxylation to resemble an illite-like structure(XRD data) when a T between 450 and 550degC isreached (Evans amp White 1958) A dehydroxylatedphyllosilicate phase structurally different to thehydrated one (Guggenheim et al 1987) is report-ed to exist up to 950degC when complete breakdownof the dehydroxylated illite occurs (Peters amp Iberg1978) Our data indicate that dehydroxylation isnot completed at T ~ 900degC which indicates thatthe kinetics of this process is slower than previ-ously estimated (Evans amp White 1958) (OH)-

released at T gt 700degC may play a significant rolein texture and mineral evolution at high-T as willbe discussed later On the other hand muscoviteshows a slight K deficit at 700degC probably due tothe high diffusion capacity of this element alongmuscovite (001) basal planes at T gt 500degC(Sanchez-Navas amp Galindo-Zaldivar 1993)Sanchez-Navas (1999) showed high-T K-diffu-sion-out of mica is induced by water adsorbedalong mica basal planes resulting from phyllosili-cate dehydroxylation

XRD data shows that dolomite most intenseBragg peak in V bricks significantly reduced itsintensity due to thermal decomposition at lower Tthan calcite main Bragg peak (Fig 3b) Dolomite

Carbonate and silicate phase reactions during ceramic firing 629

Table 4 Bulk composition of Guadix (G) and Viznar (V) samples (XRF results in wt )

LOI loss on ignition

Fig 6 a) ACS (Al2O3-CaO-SiO2) and b) CMS (CaO-MgO-SiO2) compositional diagrams of carbonates andreaction products (EMPA results)

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

starts to decompose according to (1) at ~ 700degC (1atm pressure) while calcite decomposes at higherT (830 ndash 870degC) according to (3) (Boynton 1980)Shoval (1988) proposed a two steps thermaldecomposition for dolomite involving formationof periclase and calcite (2) followed by calcitedecomposition at higher T (3)

dolomite lime periclaseCaMg(CO3)2 reg CaO + MgO + 2CO2 (1)

calcite CaMg(CO3)2 reg CaCO3 + MgO + CO2 (2)

CaCO3 reg CaO + CO2 (3)

However our XRD results showing no increase incalcite 104 Bragg peak intensity as dolomitedecomposes at 700degC (Fig 3b) are not consistentwith Shovalacutes two-steps process

Carbonate decomposition results in shrinkageof the brick (Barahona et al 1985) Upon storageat 22degC and 55 RH burnt lime (CaO) readilytransforms into portlandite according to the fol-lowing reaction (Boynton 1980)

portlanditeCaO + H2O reg Ca(OH)2 (4)

This process generates crystallization pressure inconfined spaces (ie the brick pores originallyoccupied by CaO) resulting in crack development(as observed in Fig 2a) The dark-color rimaround former carbonate grains (Fig 5c) is proba-bly due to limited portlandite carbonation in thepresence of CO2 also resulting in volume increase(Moorehead 1985) and crack development(Butterworth amp Honeyborne 1952) As a commonpractice to avoid these undesired side-effectsbricks that contain carbonates were traditionallyimmersed in water to dissolve and eliminate exist-ing CaO or Ca(OH)2

This effect is less evident in the case of MgOPericlase in the presence of H2O (room T) under-goes a slow transformation (it takes months oreven years Webb 1952) into brucite (Mg(OH)2)and may eventually transform into hydromagne-site (Mg5(CO3)4(OH)24H2O) according to thefollowing reactions (Garavelli et al 1990)

MgO + H2O reg Mg(OH)2 (5)

hydromagnesite5Mg(OH)2 + 4CO2 reg Mg5(CO3)4(OH)24H2O (6)

EMPA results indicate the presence of Ca and Mgoxides however our XRD results do not clearlyshow the presence of Ca and Mg oxides or hydrox-ides mainly due to their low concentrations (lt 10wt ) and because their main diffraction peaks aremasked by other major phase peaks (eg quartzfeldspars hematites) The hydromagnesite mainBragg peak at 581 Aring is not present either

At T sup3 800degC and in the absence of carbonateschanges are constrained to textural modificationsand reactions where a low-T phase is decomposedand fully transformed into a high-T phase withoutmajor compositional changes (Riccardi et al1999) In particular muscovite undergoes a solid-state phase change into a mixture of mullite and K-feldspar (or a melt) at T ranging from 800 to1000degC The composition of high-T muscovite isclose to K-feldspar (EMPA results) although OManalysis points to the presence of mullite Brearley(1986) and Worden et al (1987) showed fine inter-growth of alternating nanometer up to a fewmicrometers thick bands of K-feldspar (or a meltwith composition close to that of K-feldspar) andmullite in pyrometamorphic muscovite which canexplain our EMPA results

The most significant textural and mineralogi-cal changes are observed in samples with carbon-ates when fired at T gt 800degC Also melting T andextent seems to be connected with the presence orabsence of carbonates In general melting starts atlower T (~ 800degC) when carbonates are present(Tite amp Maniatis 1975) Ca and Mg from carbon-ates may act as melting agents (Segnit ampAnderson 1972) but they are reported to some-how limit the extent of vitrification at T gt 1000degC(Everhart 1957 Nuacutentildeez et al 1992) In fact weobserved that vitrification is more extended athigher T when no carbonates exist This may beexplained considering that the phyllosilicate in thematrix of non-carbonate G samples not contribut-ing to the formation of high-T Ca (or Mg) silicatesand releasing significant amounts of H2O due todehydroxylation they can contribute to extensivemelting at high-T as demonstrated by Brearley ampRubie (1990) On the other hand the higher SiO2content in these samples provides a higher amountof potential silicate-rich melt

BSE images of reaction rims between carbon-ates and silicates show that formation of Ca-sili-cates starts at 800degC This is confirmed by XRD(Fig 3b) and EMPA analyses (Table 3) Gehlenitestarts to form at this T by grain-boundary reactionbetween CaO Al2O3 and SiO2 the first derivingfrom former carbonates and the later two fromalready dehydroxylated phyllosilicates (ie a

630

somehow amorphous mixture of 3SiO2Al2O3according to Peters amp Iberg 1978) as follows

illite calcite gehleniteKAl2(Si3Al)O10(OH)2 + 6CaCO3 reg 3Ca2Al2SiO7+ 6CO2 + 2H2O + K2O + 3SiO2 (7)

The formation of gehlenite is striking becauseone could expect the formation of an Al-richpyroxene such as fassaite (Dondi et al 1998)Nevertheless it should be indicated that gehleniteformation is a good example of the so-calledGoldsmithacutes ldquoease of crystallizationrdquo principlewhich states that a silicate phase with all Al in 4-coordination is easier to form (at least in the lab)than a phase that includes Al in both 4- and 6-coor-dination when (OH)- are not present (Goldsmith1953) This is due to the higher energy barrier nec-essary to overcome when Al has to enter in a 6-coordination if compared with a 4-coordination If(OH)- are present this energy barrier is reduceddue to the smaller electrostatic repulsion among(OH)- groups than among O2- in octahedral coor-dination Gehlenite has all Al in tetrahedral coor-dination (Warren 1930) therefore crystallizingpreferentially if compared with other anhydrousCa-Al silicates This simple principle also appliesto the high-T formation of other anhydrous Al-sil-icates as we will discuss later

Wollastonite formed at the carbonate-quartzinterface through the following reaction

wollastoniteCaO + SiO2 reg CaSiO3 (8)

Wollastonite and gehlenite appear at 800degC aT ~100degC lower than previously reported (Petersamp Iberg 1978 Dondi et al 1998) This is proba-bly due to a) the low resolution of the analyticaltechniques used in the past (ie XRD) which didnot allow the detection of the small initial reactionrim between carbonates and silicates as it is evi-denced by BSE images and EMPA analysesandor b) most previous works on high-T phaseformation in ceramic processing use single com-ponents as reactants without considering addition-al phases or previous reaction by-products (Petersamp Iberg 1978) We observed that in a closer-to-reality multicomponent system where interactionsbetween reactants and products occur differentformation T are obtained

At the dolomite-quartz interface diopsidestarts to form at 900degC by the following reaction

dolomite diopside2SiO2 + CaMg(CO3)2 reg CaMgSi2O6 + 2CO (9)

At this T anorthite starts to form at the expensesof calcite and illite (plus quartz)

illite calciteKAl2(Si3Al)O10(OH)2 + 2CaCO3 + 4SiO2 regK-feldspar anorthite

2KAlSi3O8 + 2Ca2Al2SiO8 + 2CO2 + H2O (10)

At higher temperatures (1000 to 1100degC)phyllosilicates have already disappeared in allsamples (V and G) being transformed into sani-dine and mullite plus a melt according to the fol-lowing reaction

illite (muscovite) sanidine 3KAl2(Si3Al)O10(OH)2 + 2SiO2 reg 3KAlSi3O8 +

mulliteAl6Si2O13 + 3H2O + melt (11)

Formation of a melt during the previous reaction(Brearley amp Rubie 1990) is consistent with theobserved ldquocellular structurerdquo of muscovite (Tite ampManiatis 1975)

Brearley (1986) proposed a somehow differentreaction for mullite formation at the expenses ofphengitic muscovite resulting in K-feldspar +mullite + biotite No biotite is detected in ourexperiments most probably due to the low Fe con-tent in V and G micas Brearley amp Rubie (1990)experimentally demonstrated that high-T mus-covite breakdown in the presence of quartz isinfluenced by the crystal size The smallest crys-tals have a strong tendency to react with H2Oresulting in enhanced melt formation while largemuscovite crystals form limited amounts of meltThis may account for the significant vitrificationof the matrix (specially in G samples) while largemuscovite grains (up to some mm in size) still donot show significant melting

Mullite which according to OM analysis firstappears at 800degC is the second most abundantphase (after quartz) at 1100degC in the richer inphyllosilicate G samples It is interesting to men-tion that K-feldspar progressively disappears as Tincreases This is not consistent with publisheddata (Maggetti 1982) It seems that low-T K-feldspar (microcline) is not stable at high-T there-fore it transforms into a high-T K-feldspar (iesanidine) or reacts with lime to form anorthiteBoth sanidine and anorthite are detected in thehigh-T bricks However a higher concentration ofsanidine should be expected to form speciallyconsidering reaction (11) A possible explanationfor the low K-feldspar content at high-T is that K-feldspar might not have enough time to develop itscrystalline structure (sanidine) therefore being

Carbonate and silicate phase reactions during ceramic firing 631

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

concentrated as an amorphous (sanidine-like)phase responsible in part for the rise in the back-ground noise in the XRD results Mullite forma-tion seems to be kinetically faster resulting in awell-developed crystalline phase which givessharp diffraction peaks Goldsmith (1953) indi-cates that mullite could be another example of hisldquosimplexity principlerdquo for the ldquoeaserdquo of crystal-lization if compared with other aluminum sili-cates In fact the mullite structure has a largeramount of Al in 4-coordination and less Al in 6-coordination than sillimanite (Rehak et al 1998)for instance which may account for the abundanceof the former and the non-existence of the latter inthe studied bricks

In carbonate-rich V samples fired at T gt1000degC a significant reduction of gehlenite maindiffraction peak is observed According to Petersamp Iberg (1978) this is due to the reaction of thisphase with quartz forming wollastonite and anor-thite This is consistent with the increase in themain Bragg peak intensity of these later mineralsat 1100degC

At the dolomite-silicate (ie Qtz and Phy)interface it is common to observe wollastoniteeither newly formed or resulting from the destruc-tion of gehlenite The texture of the interface israther complex Protrusion or fingers from thereaction zone penetrate into areas were completetransformation into wollastonite has occurred Theinterface geometry resembles the so-called ldquovis-cous fingersrdquo (Garcia-Ruiz 1992) occurringwhen two fluids or semi-plastic solids of differentcomposition and viscosity are placed in contactInitially the contact zone is a flat surface but lateron it reaches a fingered fractal geometry(Nittmamm et al 1985) In the carbonate-silicateinterface a similar process may have taken placewhere mass transport of the reactants may havebeen enhanced by partial melting The formationof a silicate-rich melt due to clay-minerals melting(Brearley amp Rubie 1990) in the external part ofthe ring in contact with a second inner melt formedby preexisting gehlenite or the eutectic CaO-SiO2would result in viscous finger development Thefractal geometry of the developing viscous fingersmay have enhanced diffusion of mobile ions suchas Ca toward silicate reactive sites Both EDXdata showing Mg enrichment in the dolomite core(Fig 5f) and the compositional trend towardsMgO-rich terms in the CMS diagram (Fig 6b)confirm that Ca is preferentially leached out of theformer dolomite grain Finger formation willresult in a significant increase in interface surfacearea resulting in higher reaction rates Another

alternative explanation for this fingered geometrydevelopment involves partial melting at the car-bonate-silicate interface and mass transport (ieviscous flow) through the clay-quartz porousmatrix resulting in the so-called ldquoscalloped struc-turerdquo proposed by Ortoleva et al (1987) for reac-tion-diffusion processes While the viscous fingermodel requires two melts the second model onlyrequires one melt It is not clear if at any momentconditions for two melt occurrence are reachedHence the second model seems more plausible

In any case the process which is originally dif-fusion driven is transformed into a mass (ie fin-gering) plus diffusion transport which greatlyenhances Ca-silicate formation This is consistentwith the thick reaction rims (up to 250 microm thicksee Fig 5f) that surround former carbonate grainsThese thick Ca-silicate shells would hardly devel-op just by diffusion in the short time-scale of theexperiment (3 hours at the peak T) In fact at800degC (conditions where melting would be verylimited) Ca (and Mg)-silicate rims formed aroundcalcite or dolomite are 2-5 microm thick At 1100degCwhen melting is extensive the reaction rate is sohigh that rims are two orders of magnitude thick-er Therefore mass transport (ie viscous flow)seems to have played a key role in high-T Ca-(Mg-) silicate formation Mass transport is also consis-tent with observation of some areas with differentcomposition (unreacted silicates) trapped betweenfingers resulting in the island geometry observedin Fig 5g It could be expected that without fin-gering the process of Ca- or (Ca-Mg)-silicate for-mation would be self-limiting ie once a rim ofthe newly formed silicate appears diffusion wouldbe jeopardized This shell would preclude or limitthe transport of Ca (andor Mg) from the carbon-ate or it would limit the transport of the lessmobile Si (and Al) from the surrounding silicatetowards the reaction front Therefore it is neces-sary that one of the above described transport-reaction mechanisms takes place This may alsoexplain why phase formation takes place at lowerT than reported previously The presence of multi-ple interfaces among different crystals mayenhance finger development resulting in kineticchanges favoring phase formation at lower T

5 Conclusions

It is observed that even though the overallcomposition of a ceramic material does not under-go major changes upon firing significant miner-alogical and textural modifications do occur

632

These changes are connected with two reactiontypes a) solid-state replacement of a phase by anew one with little or no compositional variation(ie the muscovite and illite case) and minimizingenergy changes by preserving the original textureand orientation of the replaced phase and b) reac-tion-diffusion at the interface between two miner-al grains with contrasting composition (iecarbonate-silicate interfaces) The latter reaction isenhanced if massviscous transport resulting frompartial melting and finger structure developmentoccurs Non-carbonated clay-rich materials (G)show phyllosilicate destruction at T ranging from700 to 900degC followed by vitrification which issignificant at T gt 1000degC Carbonate-rich materi-als (V) show major phase transformation uponphyllosillicates destruction including new Ca (orMg) silicates and aluminum-silicates formation Inthis later case the vitrification starts at lower Tand it is limited at higher T Therefore it is possi-ble to establish the firing T of ceramic materialswith composition broadly similar to that in ourwork based on the mineralogy Thus a) presenceof phyllosilicates indicates a low 700 to 900degC fir-ing T for the dehydroxylated illite (or muscovite)phase b) dolomite disappears at sup3 700degC whilecalcite is still present at 800degC c) gehleniteappears at 800degC reducing its concentration above1000degC d) diopside and wollastonite appear at800degC (in very low concentrations) as reactionrims between carbonates and silicates Formationof these minerals includes ldquofingeringrdquo develop-ment e) mullite appears at T gt 800degC while littleamounts of sanidine form as microcline disap-pears and f) anorthite develops at T gt1000degC incarbonate-rich bricks

When considering equilibrium thermodynamicdata and calculated reaction T for most of theabove mentioned phases and reactions (Holland ampPowell 1985 and references herein) it isobserved that rapid heating induces significantoverstepping Therefore phases formed are inmost cases metastable a fact consistent with anal-yses of pyrometamorphic assemblages (Brearleyamp Rubie 1990 and references herein)

The formation of non-equilibrium fractal mor-phologies at the carbonate-silicate reaction inter-faces may explain why high-T silicates incarbonate-rich ceramics develop very fast Thisgrain-to-grain reaction model may help explainhigh-T silicate formation in nature (eg contactmetamorphism pyrometamorphism in xenoliths)

The gained knowledge on mineral transforma-tion and the dynamics of interface mass and diffu-sion transport facilitated through ldquofingeringrdquo or

ldquoscallopingrdquo opens new ways to understand thecomplexities of ceramic processing in particularand high-T mineral transformation in general bothin natural (ie geologic settings) and artificialcontexts (eg archaeometric analysis of ancientceramics or the design of new bricks ndashreplicas ofancient onesndash for architectural heritage conserva-tion) However care must be taken when extrapo-lating these results to other raw materialcompositions andor ceramic processing methods

Acknowledgements This work has been finan-cially supported by the CGICYT projectMAT2000-1457 and the Research Group RNM0179 of the Junta de Andaluciacutea (Spain) We thankthe Centro de Instrumentacioacuten Cientifica (CIC) ofthe Granada University for allowing us to use theirSEM-EDX EMPA and XRF facilities We thankK Elert JM Garciacutea-Ruiz M Rodriguez-Gallego and A Sanchez-Navas for fruitful discus-sions and comments We also thank the in-depthreview by Dr F Goumltz-Neunhoeffer and an anony-mous referee

References

Barahona E Huertas F Pozzuoli A Linares J(1985) Firing properties of ceramic clays fromGranada province Spain Miner Petrogr Acta 29A577-590

Boynton RS (1980) Chemistry and technology oflime and limestone 2nd Ed Wiley New York

Brearley AJ (1986) An electron optical study of mus-covite breakdown in pelitic xenoliths duringpyrometamorphism Miner Mag 357 385-397

Brearley AJ amp Rubie DC (1990) Effects of H2O onthe disequilibrium breakdown of muscovite +quartz J Petrol 31 925-956

Butterworth B amp Honeyborne DB (1952) Bricksand clays of the hasting beds Trans Brit CeramSoc 51 211-259

Dondi M Ercolani G Fabbri B Marsigli M (1998)An approach to the chemistry of pyroxenes formedduring the firing of Ca-rich silicate ceramics ClayMinerals 33 443-452

Duminuco P Riccardi MP Messiga B Setti M(1996) Modificazioni tessiturali e mineralogichecome indicatori della dinamica del processo di cot-tura di manufatti ceramici Ceramurgia 26 5 281-288

Evans JL amp White J (1958) Further studies of thethermal decomposition of clays Trans Brit CeramSoc 57 6 298

Everhart JO (1957) Use of auxiliary fluxes to improvestructural clay bodies Bull Am Ceram Soc 36268-271

Carbonate and silicate phase reactions during ceramic firing 633

al changes upon firing only V samples were ana-lyzed using this technique In particular composi-tional changes of phyllosilicates and carbonateswere studied using polished thin sections of firedV bricks (one section per selected target T ie700 800 1100degC)

Phyllosilicates (Table 2) show a muscoviticcomposition at 700degC where some K is replaced byNa even though the overall composition does notseem to correspond to paragonite Neverthelessparagonite was detected in the raw clay (Fig 3b)

Fe Mg and Ti replace some Al in the octahedralsheets Replacement of OH- by F- is almost absentAt 800degC a slight increase in Fe content is detectedOnly analyses M9 and M10 show higher concen-trations of Na and Al At 1100degC the amount of Kis reduced while Ca and Si content is increasedThe total oxide content is close to 100 a valueconsistent with complete dehydroxylation of thephyllosilicates Analyses M11-M14 in Table 2 arein agreement with high-T phengite compositionaldata from Worden et al (1987)

Carbonate and silicate phase reactions during ceramic firing 627

Fig 5 BSE images and EDX analyses of carbonate-rich Viznar samples a) fired at 700degC b) fired at 800degC whenWo (EDX analysis) formed at the Cal- silicates interface c) detail of former carbonate grain showing a white outerreaction rim (Wo or Gh) a crack and an inner core with an outer darker shell (see text for details) d) Ms (see EDXanalysis) at 700degC e) former Ms after transformation into a Fs + Mul + melt mixture (see EDX analysis) f) DolPhy(+ Qtz) interface showing a reaction rim with fingered geometry formed at 1100degC g) detail of the outer reactionrim in previous figure EDX analysis shows the rim composition (ie Wo)

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

Calcite (Table 3) shows no chemical changesat 700degC The maximum MgO content is 104 wt Trace amounts of Fe Si Al and Mn are alsodetected in calcite (as well as in dolomite) At800degC both carbonates are fully transformed eitherinto burnt lime (CaO) or into a mixture of burtlime and periclase (MgO) (eg analysis C14)Gehlenite-like compositions are detected in someanalyses (eg C10) At 1100degC the presence ofnewly formed phases in particular wollastonite isclearly confirmed (eg C17) Ca content in thelime plus periclase mixture is reduced significant-ly due to the incorporation of this element intonewly formed high-T silicate phases As T increas-es a clear compositional trend toward Si- and Al-rich calcium phases (ie toward gehlenite andanorthite) is observed (Fig 6a) a fact consistentwith XRD and SEM-EDX results A clear trendtoward wollastonite-like compositions as well as a

trend toward MgO rich composition is alsoobserved as T increases (Fig 6b) Some EMPAanalyses suggest a larnite -like compositionHowever XRD results do not show the presenceof larnite

36 XRF analysis

According to the classification proposed forclay materials in ceramic industry by Fabbri ampFiori (1985) G raw material composition falls intothe quartz-feldspar sands category while V sam-ples fall out of any defined category due to theirlow silica and high calcium contents (Table 4) Asexpected firing does not affect major elementconcentration in the brick The only significantchange is detected in the loss on ignition values Itis evidenced that G samples lost almost all waterbetween 700 and 900degC while V samples show a

628

Table 2 EMPA analysis results of phyllosilicates composition (in wt )

Table 3 Composition of selected carbonate grains ( EMPA results in wt )

more gradual loss of water and CO2 as T increases(Table 4)

4 Discussion

Firing up to 700degC induce no significant min-eralogicaltextural transformations in both thecarbonate-rich V and the silicate-rich G sampleswith the exception of the clay minerals Clay min-erals structure is reported to collapse due to dehy-droxylation to resemble an illite-like structure(XRD data) when a T between 450 and 550degC isreached (Evans amp White 1958) A dehydroxylatedphyllosilicate phase structurally different to thehydrated one (Guggenheim et al 1987) is report-ed to exist up to 950degC when complete breakdownof the dehydroxylated illite occurs (Peters amp Iberg1978) Our data indicate that dehydroxylation isnot completed at T ~ 900degC which indicates thatthe kinetics of this process is slower than previ-ously estimated (Evans amp White 1958) (OH)-

released at T gt 700degC may play a significant rolein texture and mineral evolution at high-T as willbe discussed later On the other hand muscoviteshows a slight K deficit at 700degC probably due tothe high diffusion capacity of this element alongmuscovite (001) basal planes at T gt 500degC(Sanchez-Navas amp Galindo-Zaldivar 1993)Sanchez-Navas (1999) showed high-T K-diffu-sion-out of mica is induced by water adsorbedalong mica basal planes resulting from phyllosili-cate dehydroxylation

XRD data shows that dolomite most intenseBragg peak in V bricks significantly reduced itsintensity due to thermal decomposition at lower Tthan calcite main Bragg peak (Fig 3b) Dolomite

Carbonate and silicate phase reactions during ceramic firing 629

Table 4 Bulk composition of Guadix (G) and Viznar (V) samples (XRF results in wt )

LOI loss on ignition

Fig 6 a) ACS (Al2O3-CaO-SiO2) and b) CMS (CaO-MgO-SiO2) compositional diagrams of carbonates andreaction products (EMPA results)

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

starts to decompose according to (1) at ~ 700degC (1atm pressure) while calcite decomposes at higherT (830 ndash 870degC) according to (3) (Boynton 1980)Shoval (1988) proposed a two steps thermaldecomposition for dolomite involving formationof periclase and calcite (2) followed by calcitedecomposition at higher T (3)

dolomite lime periclaseCaMg(CO3)2 reg CaO + MgO + 2CO2 (1)

calcite CaMg(CO3)2 reg CaCO3 + MgO + CO2 (2)

CaCO3 reg CaO + CO2 (3)

However our XRD results showing no increase incalcite 104 Bragg peak intensity as dolomitedecomposes at 700degC (Fig 3b) are not consistentwith Shovalacutes two-steps process

Carbonate decomposition results in shrinkageof the brick (Barahona et al 1985) Upon storageat 22degC and 55 RH burnt lime (CaO) readilytransforms into portlandite according to the fol-lowing reaction (Boynton 1980)

portlanditeCaO + H2O reg Ca(OH)2 (4)

This process generates crystallization pressure inconfined spaces (ie the brick pores originallyoccupied by CaO) resulting in crack development(as observed in Fig 2a) The dark-color rimaround former carbonate grains (Fig 5c) is proba-bly due to limited portlandite carbonation in thepresence of CO2 also resulting in volume increase(Moorehead 1985) and crack development(Butterworth amp Honeyborne 1952) As a commonpractice to avoid these undesired side-effectsbricks that contain carbonates were traditionallyimmersed in water to dissolve and eliminate exist-ing CaO or Ca(OH)2

This effect is less evident in the case of MgOPericlase in the presence of H2O (room T) under-goes a slow transformation (it takes months oreven years Webb 1952) into brucite (Mg(OH)2)and may eventually transform into hydromagne-site (Mg5(CO3)4(OH)24H2O) according to thefollowing reactions (Garavelli et al 1990)

MgO + H2O reg Mg(OH)2 (5)

hydromagnesite5Mg(OH)2 + 4CO2 reg Mg5(CO3)4(OH)24H2O (6)

EMPA results indicate the presence of Ca and Mgoxides however our XRD results do not clearlyshow the presence of Ca and Mg oxides or hydrox-ides mainly due to their low concentrations (lt 10wt ) and because their main diffraction peaks aremasked by other major phase peaks (eg quartzfeldspars hematites) The hydromagnesite mainBragg peak at 581 Aring is not present either

At T sup3 800degC and in the absence of carbonateschanges are constrained to textural modificationsand reactions where a low-T phase is decomposedand fully transformed into a high-T phase withoutmajor compositional changes (Riccardi et al1999) In particular muscovite undergoes a solid-state phase change into a mixture of mullite and K-feldspar (or a melt) at T ranging from 800 to1000degC The composition of high-T muscovite isclose to K-feldspar (EMPA results) although OManalysis points to the presence of mullite Brearley(1986) and Worden et al (1987) showed fine inter-growth of alternating nanometer up to a fewmicrometers thick bands of K-feldspar (or a meltwith composition close to that of K-feldspar) andmullite in pyrometamorphic muscovite which canexplain our EMPA results

The most significant textural and mineralogi-cal changes are observed in samples with carbon-ates when fired at T gt 800degC Also melting T andextent seems to be connected with the presence orabsence of carbonates In general melting starts atlower T (~ 800degC) when carbonates are present(Tite amp Maniatis 1975) Ca and Mg from carbon-ates may act as melting agents (Segnit ampAnderson 1972) but they are reported to some-how limit the extent of vitrification at T gt 1000degC(Everhart 1957 Nuacutentildeez et al 1992) In fact weobserved that vitrification is more extended athigher T when no carbonates exist This may beexplained considering that the phyllosilicate in thematrix of non-carbonate G samples not contribut-ing to the formation of high-T Ca (or Mg) silicatesand releasing significant amounts of H2O due todehydroxylation they can contribute to extensivemelting at high-T as demonstrated by Brearley ampRubie (1990) On the other hand the higher SiO2content in these samples provides a higher amountof potential silicate-rich melt

BSE images of reaction rims between carbon-ates and silicates show that formation of Ca-sili-cates starts at 800degC This is confirmed by XRD(Fig 3b) and EMPA analyses (Table 3) Gehlenitestarts to form at this T by grain-boundary reactionbetween CaO Al2O3 and SiO2 the first derivingfrom former carbonates and the later two fromalready dehydroxylated phyllosilicates (ie a

630

somehow amorphous mixture of 3SiO2Al2O3according to Peters amp Iberg 1978) as follows

illite calcite gehleniteKAl2(Si3Al)O10(OH)2 + 6CaCO3 reg 3Ca2Al2SiO7+ 6CO2 + 2H2O + K2O + 3SiO2 (7)

The formation of gehlenite is striking becauseone could expect the formation of an Al-richpyroxene such as fassaite (Dondi et al 1998)Nevertheless it should be indicated that gehleniteformation is a good example of the so-calledGoldsmithacutes ldquoease of crystallizationrdquo principlewhich states that a silicate phase with all Al in 4-coordination is easier to form (at least in the lab)than a phase that includes Al in both 4- and 6-coor-dination when (OH)- are not present (Goldsmith1953) This is due to the higher energy barrier nec-essary to overcome when Al has to enter in a 6-coordination if compared with a 4-coordination If(OH)- are present this energy barrier is reduceddue to the smaller electrostatic repulsion among(OH)- groups than among O2- in octahedral coor-dination Gehlenite has all Al in tetrahedral coor-dination (Warren 1930) therefore crystallizingpreferentially if compared with other anhydrousCa-Al silicates This simple principle also appliesto the high-T formation of other anhydrous Al-sil-icates as we will discuss later

Wollastonite formed at the carbonate-quartzinterface through the following reaction

wollastoniteCaO + SiO2 reg CaSiO3 (8)

Wollastonite and gehlenite appear at 800degC aT ~100degC lower than previously reported (Petersamp Iberg 1978 Dondi et al 1998) This is proba-bly due to a) the low resolution of the analyticaltechniques used in the past (ie XRD) which didnot allow the detection of the small initial reactionrim between carbonates and silicates as it is evi-denced by BSE images and EMPA analysesandor b) most previous works on high-T phaseformation in ceramic processing use single com-ponents as reactants without considering addition-al phases or previous reaction by-products (Petersamp Iberg 1978) We observed that in a closer-to-reality multicomponent system where interactionsbetween reactants and products occur differentformation T are obtained

At the dolomite-quartz interface diopsidestarts to form at 900degC by the following reaction

dolomite diopside2SiO2 + CaMg(CO3)2 reg CaMgSi2O6 + 2CO (9)

At this T anorthite starts to form at the expensesof calcite and illite (plus quartz)

illite calciteKAl2(Si3Al)O10(OH)2 + 2CaCO3 + 4SiO2 regK-feldspar anorthite

2KAlSi3O8 + 2Ca2Al2SiO8 + 2CO2 + H2O (10)

At higher temperatures (1000 to 1100degC)phyllosilicates have already disappeared in allsamples (V and G) being transformed into sani-dine and mullite plus a melt according to the fol-lowing reaction

illite (muscovite) sanidine 3KAl2(Si3Al)O10(OH)2 + 2SiO2 reg 3KAlSi3O8 +

mulliteAl6Si2O13 + 3H2O + melt (11)

Formation of a melt during the previous reaction(Brearley amp Rubie 1990) is consistent with theobserved ldquocellular structurerdquo of muscovite (Tite ampManiatis 1975)

Brearley (1986) proposed a somehow differentreaction for mullite formation at the expenses ofphengitic muscovite resulting in K-feldspar +mullite + biotite No biotite is detected in ourexperiments most probably due to the low Fe con-tent in V and G micas Brearley amp Rubie (1990)experimentally demonstrated that high-T mus-covite breakdown in the presence of quartz isinfluenced by the crystal size The smallest crys-tals have a strong tendency to react with H2Oresulting in enhanced melt formation while largemuscovite crystals form limited amounts of meltThis may account for the significant vitrificationof the matrix (specially in G samples) while largemuscovite grains (up to some mm in size) still donot show significant melting

Mullite which according to OM analysis firstappears at 800degC is the second most abundantphase (after quartz) at 1100degC in the richer inphyllosilicate G samples It is interesting to men-tion that K-feldspar progressively disappears as Tincreases This is not consistent with publisheddata (Maggetti 1982) It seems that low-T K-feldspar (microcline) is not stable at high-T there-fore it transforms into a high-T K-feldspar (iesanidine) or reacts with lime to form anorthiteBoth sanidine and anorthite are detected in thehigh-T bricks However a higher concentration ofsanidine should be expected to form speciallyconsidering reaction (11) A possible explanationfor the low K-feldspar content at high-T is that K-feldspar might not have enough time to develop itscrystalline structure (sanidine) therefore being

Carbonate and silicate phase reactions during ceramic firing 631

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

concentrated as an amorphous (sanidine-like)phase responsible in part for the rise in the back-ground noise in the XRD results Mullite forma-tion seems to be kinetically faster resulting in awell-developed crystalline phase which givessharp diffraction peaks Goldsmith (1953) indi-cates that mullite could be another example of hisldquosimplexity principlerdquo for the ldquoeaserdquo of crystal-lization if compared with other aluminum sili-cates In fact the mullite structure has a largeramount of Al in 4-coordination and less Al in 6-coordination than sillimanite (Rehak et al 1998)for instance which may account for the abundanceof the former and the non-existence of the latter inthe studied bricks

In carbonate-rich V samples fired at T gt1000degC a significant reduction of gehlenite maindiffraction peak is observed According to Petersamp Iberg (1978) this is due to the reaction of thisphase with quartz forming wollastonite and anor-thite This is consistent with the increase in themain Bragg peak intensity of these later mineralsat 1100degC

At the dolomite-silicate (ie Qtz and Phy)interface it is common to observe wollastoniteeither newly formed or resulting from the destruc-tion of gehlenite The texture of the interface israther complex Protrusion or fingers from thereaction zone penetrate into areas were completetransformation into wollastonite has occurred Theinterface geometry resembles the so-called ldquovis-cous fingersrdquo (Garcia-Ruiz 1992) occurringwhen two fluids or semi-plastic solids of differentcomposition and viscosity are placed in contactInitially the contact zone is a flat surface but lateron it reaches a fingered fractal geometry(Nittmamm et al 1985) In the carbonate-silicateinterface a similar process may have taken placewhere mass transport of the reactants may havebeen enhanced by partial melting The formationof a silicate-rich melt due to clay-minerals melting(Brearley amp Rubie 1990) in the external part ofthe ring in contact with a second inner melt formedby preexisting gehlenite or the eutectic CaO-SiO2would result in viscous finger development Thefractal geometry of the developing viscous fingersmay have enhanced diffusion of mobile ions suchas Ca toward silicate reactive sites Both EDXdata showing Mg enrichment in the dolomite core(Fig 5f) and the compositional trend towardsMgO-rich terms in the CMS diagram (Fig 6b)confirm that Ca is preferentially leached out of theformer dolomite grain Finger formation willresult in a significant increase in interface surfacearea resulting in higher reaction rates Another

alternative explanation for this fingered geometrydevelopment involves partial melting at the car-bonate-silicate interface and mass transport (ieviscous flow) through the clay-quartz porousmatrix resulting in the so-called ldquoscalloped struc-turerdquo proposed by Ortoleva et al (1987) for reac-tion-diffusion processes While the viscous fingermodel requires two melts the second model onlyrequires one melt It is not clear if at any momentconditions for two melt occurrence are reachedHence the second model seems more plausible

In any case the process which is originally dif-fusion driven is transformed into a mass (ie fin-gering) plus diffusion transport which greatlyenhances Ca-silicate formation This is consistentwith the thick reaction rims (up to 250 microm thicksee Fig 5f) that surround former carbonate grainsThese thick Ca-silicate shells would hardly devel-op just by diffusion in the short time-scale of theexperiment (3 hours at the peak T) In fact at800degC (conditions where melting would be verylimited) Ca (and Mg)-silicate rims formed aroundcalcite or dolomite are 2-5 microm thick At 1100degCwhen melting is extensive the reaction rate is sohigh that rims are two orders of magnitude thick-er Therefore mass transport (ie viscous flow)seems to have played a key role in high-T Ca-(Mg-) silicate formation Mass transport is also consis-tent with observation of some areas with differentcomposition (unreacted silicates) trapped betweenfingers resulting in the island geometry observedin Fig 5g It could be expected that without fin-gering the process of Ca- or (Ca-Mg)-silicate for-mation would be self-limiting ie once a rim ofthe newly formed silicate appears diffusion wouldbe jeopardized This shell would preclude or limitthe transport of Ca (andor Mg) from the carbon-ate or it would limit the transport of the lessmobile Si (and Al) from the surrounding silicatetowards the reaction front Therefore it is neces-sary that one of the above described transport-reaction mechanisms takes place This may alsoexplain why phase formation takes place at lowerT than reported previously The presence of multi-ple interfaces among different crystals mayenhance finger development resulting in kineticchanges favoring phase formation at lower T

5 Conclusions

It is observed that even though the overallcomposition of a ceramic material does not under-go major changes upon firing significant miner-alogical and textural modifications do occur

632

These changes are connected with two reactiontypes a) solid-state replacement of a phase by anew one with little or no compositional variation(ie the muscovite and illite case) and minimizingenergy changes by preserving the original textureand orientation of the replaced phase and b) reac-tion-diffusion at the interface between two miner-al grains with contrasting composition (iecarbonate-silicate interfaces) The latter reaction isenhanced if massviscous transport resulting frompartial melting and finger structure developmentoccurs Non-carbonated clay-rich materials (G)show phyllosilicate destruction at T ranging from700 to 900degC followed by vitrification which issignificant at T gt 1000degC Carbonate-rich materi-als (V) show major phase transformation uponphyllosillicates destruction including new Ca (orMg) silicates and aluminum-silicates formation Inthis later case the vitrification starts at lower Tand it is limited at higher T Therefore it is possi-ble to establish the firing T of ceramic materialswith composition broadly similar to that in ourwork based on the mineralogy Thus a) presenceof phyllosilicates indicates a low 700 to 900degC fir-ing T for the dehydroxylated illite (or muscovite)phase b) dolomite disappears at sup3 700degC whilecalcite is still present at 800degC c) gehleniteappears at 800degC reducing its concentration above1000degC d) diopside and wollastonite appear at800degC (in very low concentrations) as reactionrims between carbonates and silicates Formationof these minerals includes ldquofingeringrdquo develop-ment e) mullite appears at T gt 800degC while littleamounts of sanidine form as microcline disap-pears and f) anorthite develops at T gt1000degC incarbonate-rich bricks

When considering equilibrium thermodynamicdata and calculated reaction T for most of theabove mentioned phases and reactions (Holland ampPowell 1985 and references herein) it isobserved that rapid heating induces significantoverstepping Therefore phases formed are inmost cases metastable a fact consistent with anal-yses of pyrometamorphic assemblages (Brearleyamp Rubie 1990 and references herein)

The formation of non-equilibrium fractal mor-phologies at the carbonate-silicate reaction inter-faces may explain why high-T silicates incarbonate-rich ceramics develop very fast Thisgrain-to-grain reaction model may help explainhigh-T silicate formation in nature (eg contactmetamorphism pyrometamorphism in xenoliths)

The gained knowledge on mineral transforma-tion and the dynamics of interface mass and diffu-sion transport facilitated through ldquofingeringrdquo or

ldquoscallopingrdquo opens new ways to understand thecomplexities of ceramic processing in particularand high-T mineral transformation in general bothin natural (ie geologic settings) and artificialcontexts (eg archaeometric analysis of ancientceramics or the design of new bricks ndashreplicas ofancient onesndash for architectural heritage conserva-tion) However care must be taken when extrapo-lating these results to other raw materialcompositions andor ceramic processing methods

Acknowledgements This work has been finan-cially supported by the CGICYT projectMAT2000-1457 and the Research Group RNM0179 of the Junta de Andaluciacutea (Spain) We thankthe Centro de Instrumentacioacuten Cientifica (CIC) ofthe Granada University for allowing us to use theirSEM-EDX EMPA and XRF facilities We thankK Elert JM Garciacutea-Ruiz M Rodriguez-Gallego and A Sanchez-Navas for fruitful discus-sions and comments We also thank the in-depthreview by Dr F Goumltz-Neunhoeffer and an anony-mous referee

References

Barahona E Huertas F Pozzuoli A Linares J(1985) Firing properties of ceramic clays fromGranada province Spain Miner Petrogr Acta 29A577-590

Boynton RS (1980) Chemistry and technology oflime and limestone 2nd Ed Wiley New York

Brearley AJ (1986) An electron optical study of mus-covite breakdown in pelitic xenoliths duringpyrometamorphism Miner Mag 357 385-397

Brearley AJ amp Rubie DC (1990) Effects of H2O onthe disequilibrium breakdown of muscovite +quartz J Petrol 31 925-956

Butterworth B amp Honeyborne DB (1952) Bricksand clays of the hasting beds Trans Brit CeramSoc 51 211-259

Dondi M Ercolani G Fabbri B Marsigli M (1998)An approach to the chemistry of pyroxenes formedduring the firing of Ca-rich silicate ceramics ClayMinerals 33 443-452

Duminuco P Riccardi MP Messiga B Setti M(1996) Modificazioni tessiturali e mineralogichecome indicatori della dinamica del processo di cot-tura di manufatti ceramici Ceramurgia 26 5 281-288

Evans JL amp White J (1958) Further studies of thethermal decomposition of clays Trans Brit CeramSoc 57 6 298

Everhart JO (1957) Use of auxiliary fluxes to improvestructural clay bodies Bull Am Ceram Soc 36268-271

Carbonate and silicate phase reactions during ceramic firing 633

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

Calcite (Table 3) shows no chemical changesat 700degC The maximum MgO content is 104 wt Trace amounts of Fe Si Al and Mn are alsodetected in calcite (as well as in dolomite) At800degC both carbonates are fully transformed eitherinto burnt lime (CaO) or into a mixture of burtlime and periclase (MgO) (eg analysis C14)Gehlenite-like compositions are detected in someanalyses (eg C10) At 1100degC the presence ofnewly formed phases in particular wollastonite isclearly confirmed (eg C17) Ca content in thelime plus periclase mixture is reduced significant-ly due to the incorporation of this element intonewly formed high-T silicate phases As T increas-es a clear compositional trend toward Si- and Al-rich calcium phases (ie toward gehlenite andanorthite) is observed (Fig 6a) a fact consistentwith XRD and SEM-EDX results A clear trendtoward wollastonite-like compositions as well as a

trend toward MgO rich composition is alsoobserved as T increases (Fig 6b) Some EMPAanalyses suggest a larnite -like compositionHowever XRD results do not show the presenceof larnite

36 XRF analysis

According to the classification proposed forclay materials in ceramic industry by Fabbri ampFiori (1985) G raw material composition falls intothe quartz-feldspar sands category while V sam-ples fall out of any defined category due to theirlow silica and high calcium contents (Table 4) Asexpected firing does not affect major elementconcentration in the brick The only significantchange is detected in the loss on ignition values Itis evidenced that G samples lost almost all waterbetween 700 and 900degC while V samples show a

628

Table 2 EMPA analysis results of phyllosilicates composition (in wt )

Table 3 Composition of selected carbonate grains ( EMPA results in wt )

more gradual loss of water and CO2 as T increases(Table 4)

4 Discussion

Firing up to 700degC induce no significant min-eralogicaltextural transformations in both thecarbonate-rich V and the silicate-rich G sampleswith the exception of the clay minerals Clay min-erals structure is reported to collapse due to dehy-droxylation to resemble an illite-like structure(XRD data) when a T between 450 and 550degC isreached (Evans amp White 1958) A dehydroxylatedphyllosilicate phase structurally different to thehydrated one (Guggenheim et al 1987) is report-ed to exist up to 950degC when complete breakdownof the dehydroxylated illite occurs (Peters amp Iberg1978) Our data indicate that dehydroxylation isnot completed at T ~ 900degC which indicates thatthe kinetics of this process is slower than previ-ously estimated (Evans amp White 1958) (OH)-

released at T gt 700degC may play a significant rolein texture and mineral evolution at high-T as willbe discussed later On the other hand muscoviteshows a slight K deficit at 700degC probably due tothe high diffusion capacity of this element alongmuscovite (001) basal planes at T gt 500degC(Sanchez-Navas amp Galindo-Zaldivar 1993)Sanchez-Navas (1999) showed high-T K-diffu-sion-out of mica is induced by water adsorbedalong mica basal planes resulting from phyllosili-cate dehydroxylation

XRD data shows that dolomite most intenseBragg peak in V bricks significantly reduced itsintensity due to thermal decomposition at lower Tthan calcite main Bragg peak (Fig 3b) Dolomite

Carbonate and silicate phase reactions during ceramic firing 629

Table 4 Bulk composition of Guadix (G) and Viznar (V) samples (XRF results in wt )

LOI loss on ignition

Fig 6 a) ACS (Al2O3-CaO-SiO2) and b) CMS (CaO-MgO-SiO2) compositional diagrams of carbonates andreaction products (EMPA results)

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

starts to decompose according to (1) at ~ 700degC (1atm pressure) while calcite decomposes at higherT (830 ndash 870degC) according to (3) (Boynton 1980)Shoval (1988) proposed a two steps thermaldecomposition for dolomite involving formationof periclase and calcite (2) followed by calcitedecomposition at higher T (3)

dolomite lime periclaseCaMg(CO3)2 reg CaO + MgO + 2CO2 (1)

calcite CaMg(CO3)2 reg CaCO3 + MgO + CO2 (2)

CaCO3 reg CaO + CO2 (3)

However our XRD results showing no increase incalcite 104 Bragg peak intensity as dolomitedecomposes at 700degC (Fig 3b) are not consistentwith Shovalacutes two-steps process

Carbonate decomposition results in shrinkageof the brick (Barahona et al 1985) Upon storageat 22degC and 55 RH burnt lime (CaO) readilytransforms into portlandite according to the fol-lowing reaction (Boynton 1980)

portlanditeCaO + H2O reg Ca(OH)2 (4)

This process generates crystallization pressure inconfined spaces (ie the brick pores originallyoccupied by CaO) resulting in crack development(as observed in Fig 2a) The dark-color rimaround former carbonate grains (Fig 5c) is proba-bly due to limited portlandite carbonation in thepresence of CO2 also resulting in volume increase(Moorehead 1985) and crack development(Butterworth amp Honeyborne 1952) As a commonpractice to avoid these undesired side-effectsbricks that contain carbonates were traditionallyimmersed in water to dissolve and eliminate exist-ing CaO or Ca(OH)2

This effect is less evident in the case of MgOPericlase in the presence of H2O (room T) under-goes a slow transformation (it takes months oreven years Webb 1952) into brucite (Mg(OH)2)and may eventually transform into hydromagne-site (Mg5(CO3)4(OH)24H2O) according to thefollowing reactions (Garavelli et al 1990)

MgO + H2O reg Mg(OH)2 (5)

hydromagnesite5Mg(OH)2 + 4CO2 reg Mg5(CO3)4(OH)24H2O (6)

EMPA results indicate the presence of Ca and Mgoxides however our XRD results do not clearlyshow the presence of Ca and Mg oxides or hydrox-ides mainly due to their low concentrations (lt 10wt ) and because their main diffraction peaks aremasked by other major phase peaks (eg quartzfeldspars hematites) The hydromagnesite mainBragg peak at 581 Aring is not present either

At T sup3 800degC and in the absence of carbonateschanges are constrained to textural modificationsand reactions where a low-T phase is decomposedand fully transformed into a high-T phase withoutmajor compositional changes (Riccardi et al1999) In particular muscovite undergoes a solid-state phase change into a mixture of mullite and K-feldspar (or a melt) at T ranging from 800 to1000degC The composition of high-T muscovite isclose to K-feldspar (EMPA results) although OManalysis points to the presence of mullite Brearley(1986) and Worden et al (1987) showed fine inter-growth of alternating nanometer up to a fewmicrometers thick bands of K-feldspar (or a meltwith composition close to that of K-feldspar) andmullite in pyrometamorphic muscovite which canexplain our EMPA results

The most significant textural and mineralogi-cal changes are observed in samples with carbon-ates when fired at T gt 800degC Also melting T andextent seems to be connected with the presence orabsence of carbonates In general melting starts atlower T (~ 800degC) when carbonates are present(Tite amp Maniatis 1975) Ca and Mg from carbon-ates may act as melting agents (Segnit ampAnderson 1972) but they are reported to some-how limit the extent of vitrification at T gt 1000degC(Everhart 1957 Nuacutentildeez et al 1992) In fact weobserved that vitrification is more extended athigher T when no carbonates exist This may beexplained considering that the phyllosilicate in thematrix of non-carbonate G samples not contribut-ing to the formation of high-T Ca (or Mg) silicatesand releasing significant amounts of H2O due todehydroxylation they can contribute to extensivemelting at high-T as demonstrated by Brearley ampRubie (1990) On the other hand the higher SiO2content in these samples provides a higher amountof potential silicate-rich melt

BSE images of reaction rims between carbon-ates and silicates show that formation of Ca-sili-cates starts at 800degC This is confirmed by XRD(Fig 3b) and EMPA analyses (Table 3) Gehlenitestarts to form at this T by grain-boundary reactionbetween CaO Al2O3 and SiO2 the first derivingfrom former carbonates and the later two fromalready dehydroxylated phyllosilicates (ie a

630

somehow amorphous mixture of 3SiO2Al2O3according to Peters amp Iberg 1978) as follows

illite calcite gehleniteKAl2(Si3Al)O10(OH)2 + 6CaCO3 reg 3Ca2Al2SiO7+ 6CO2 + 2H2O + K2O + 3SiO2 (7)

The formation of gehlenite is striking becauseone could expect the formation of an Al-richpyroxene such as fassaite (Dondi et al 1998)Nevertheless it should be indicated that gehleniteformation is a good example of the so-calledGoldsmithacutes ldquoease of crystallizationrdquo principlewhich states that a silicate phase with all Al in 4-coordination is easier to form (at least in the lab)than a phase that includes Al in both 4- and 6-coor-dination when (OH)- are not present (Goldsmith1953) This is due to the higher energy barrier nec-essary to overcome when Al has to enter in a 6-coordination if compared with a 4-coordination If(OH)- are present this energy barrier is reduceddue to the smaller electrostatic repulsion among(OH)- groups than among O2- in octahedral coor-dination Gehlenite has all Al in tetrahedral coor-dination (Warren 1930) therefore crystallizingpreferentially if compared with other anhydrousCa-Al silicates This simple principle also appliesto the high-T formation of other anhydrous Al-sil-icates as we will discuss later

Wollastonite formed at the carbonate-quartzinterface through the following reaction

wollastoniteCaO + SiO2 reg CaSiO3 (8)

Wollastonite and gehlenite appear at 800degC aT ~100degC lower than previously reported (Petersamp Iberg 1978 Dondi et al 1998) This is proba-bly due to a) the low resolution of the analyticaltechniques used in the past (ie XRD) which didnot allow the detection of the small initial reactionrim between carbonates and silicates as it is evi-denced by BSE images and EMPA analysesandor b) most previous works on high-T phaseformation in ceramic processing use single com-ponents as reactants without considering addition-al phases or previous reaction by-products (Petersamp Iberg 1978) We observed that in a closer-to-reality multicomponent system where interactionsbetween reactants and products occur differentformation T are obtained

At the dolomite-quartz interface diopsidestarts to form at 900degC by the following reaction

dolomite diopside2SiO2 + CaMg(CO3)2 reg CaMgSi2O6 + 2CO (9)

At this T anorthite starts to form at the expensesof calcite and illite (plus quartz)

illite calciteKAl2(Si3Al)O10(OH)2 + 2CaCO3 + 4SiO2 regK-feldspar anorthite

2KAlSi3O8 + 2Ca2Al2SiO8 + 2CO2 + H2O (10)

At higher temperatures (1000 to 1100degC)phyllosilicates have already disappeared in allsamples (V and G) being transformed into sani-dine and mullite plus a melt according to the fol-lowing reaction

illite (muscovite) sanidine 3KAl2(Si3Al)O10(OH)2 + 2SiO2 reg 3KAlSi3O8 +

mulliteAl6Si2O13 + 3H2O + melt (11)

Formation of a melt during the previous reaction(Brearley amp Rubie 1990) is consistent with theobserved ldquocellular structurerdquo of muscovite (Tite ampManiatis 1975)

Brearley (1986) proposed a somehow differentreaction for mullite formation at the expenses ofphengitic muscovite resulting in K-feldspar +mullite + biotite No biotite is detected in ourexperiments most probably due to the low Fe con-tent in V and G micas Brearley amp Rubie (1990)experimentally demonstrated that high-T mus-covite breakdown in the presence of quartz isinfluenced by the crystal size The smallest crys-tals have a strong tendency to react with H2Oresulting in enhanced melt formation while largemuscovite crystals form limited amounts of meltThis may account for the significant vitrificationof the matrix (specially in G samples) while largemuscovite grains (up to some mm in size) still donot show significant melting

Mullite which according to OM analysis firstappears at 800degC is the second most abundantphase (after quartz) at 1100degC in the richer inphyllosilicate G samples It is interesting to men-tion that K-feldspar progressively disappears as Tincreases This is not consistent with publisheddata (Maggetti 1982) It seems that low-T K-feldspar (microcline) is not stable at high-T there-fore it transforms into a high-T K-feldspar (iesanidine) or reacts with lime to form anorthiteBoth sanidine and anorthite are detected in thehigh-T bricks However a higher concentration ofsanidine should be expected to form speciallyconsidering reaction (11) A possible explanationfor the low K-feldspar content at high-T is that K-feldspar might not have enough time to develop itscrystalline structure (sanidine) therefore being

Carbonate and silicate phase reactions during ceramic firing 631

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

concentrated as an amorphous (sanidine-like)phase responsible in part for the rise in the back-ground noise in the XRD results Mullite forma-tion seems to be kinetically faster resulting in awell-developed crystalline phase which givessharp diffraction peaks Goldsmith (1953) indi-cates that mullite could be another example of hisldquosimplexity principlerdquo for the ldquoeaserdquo of crystal-lization if compared with other aluminum sili-cates In fact the mullite structure has a largeramount of Al in 4-coordination and less Al in 6-coordination than sillimanite (Rehak et al 1998)for instance which may account for the abundanceof the former and the non-existence of the latter inthe studied bricks

In carbonate-rich V samples fired at T gt1000degC a significant reduction of gehlenite maindiffraction peak is observed According to Petersamp Iberg (1978) this is due to the reaction of thisphase with quartz forming wollastonite and anor-thite This is consistent with the increase in themain Bragg peak intensity of these later mineralsat 1100degC

At the dolomite-silicate (ie Qtz and Phy)interface it is common to observe wollastoniteeither newly formed or resulting from the destruc-tion of gehlenite The texture of the interface israther complex Protrusion or fingers from thereaction zone penetrate into areas were completetransformation into wollastonite has occurred Theinterface geometry resembles the so-called ldquovis-cous fingersrdquo (Garcia-Ruiz 1992) occurringwhen two fluids or semi-plastic solids of differentcomposition and viscosity are placed in contactInitially the contact zone is a flat surface but lateron it reaches a fingered fractal geometry(Nittmamm et al 1985) In the carbonate-silicateinterface a similar process may have taken placewhere mass transport of the reactants may havebeen enhanced by partial melting The formationof a silicate-rich melt due to clay-minerals melting(Brearley amp Rubie 1990) in the external part ofthe ring in contact with a second inner melt formedby preexisting gehlenite or the eutectic CaO-SiO2would result in viscous finger development Thefractal geometry of the developing viscous fingersmay have enhanced diffusion of mobile ions suchas Ca toward silicate reactive sites Both EDXdata showing Mg enrichment in the dolomite core(Fig 5f) and the compositional trend towardsMgO-rich terms in the CMS diagram (Fig 6b)confirm that Ca is preferentially leached out of theformer dolomite grain Finger formation willresult in a significant increase in interface surfacearea resulting in higher reaction rates Another

alternative explanation for this fingered geometrydevelopment involves partial melting at the car-bonate-silicate interface and mass transport (ieviscous flow) through the clay-quartz porousmatrix resulting in the so-called ldquoscalloped struc-turerdquo proposed by Ortoleva et al (1987) for reac-tion-diffusion processes While the viscous fingermodel requires two melts the second model onlyrequires one melt It is not clear if at any momentconditions for two melt occurrence are reachedHence the second model seems more plausible

In any case the process which is originally dif-fusion driven is transformed into a mass (ie fin-gering) plus diffusion transport which greatlyenhances Ca-silicate formation This is consistentwith the thick reaction rims (up to 250 microm thicksee Fig 5f) that surround former carbonate grainsThese thick Ca-silicate shells would hardly devel-op just by diffusion in the short time-scale of theexperiment (3 hours at the peak T) In fact at800degC (conditions where melting would be verylimited) Ca (and Mg)-silicate rims formed aroundcalcite or dolomite are 2-5 microm thick At 1100degCwhen melting is extensive the reaction rate is sohigh that rims are two orders of magnitude thick-er Therefore mass transport (ie viscous flow)seems to have played a key role in high-T Ca-(Mg-) silicate formation Mass transport is also consis-tent with observation of some areas with differentcomposition (unreacted silicates) trapped betweenfingers resulting in the island geometry observedin Fig 5g It could be expected that without fin-gering the process of Ca- or (Ca-Mg)-silicate for-mation would be self-limiting ie once a rim ofthe newly formed silicate appears diffusion wouldbe jeopardized This shell would preclude or limitthe transport of Ca (andor Mg) from the carbon-ate or it would limit the transport of the lessmobile Si (and Al) from the surrounding silicatetowards the reaction front Therefore it is neces-sary that one of the above described transport-reaction mechanisms takes place This may alsoexplain why phase formation takes place at lowerT than reported previously The presence of multi-ple interfaces among different crystals mayenhance finger development resulting in kineticchanges favoring phase formation at lower T

5 Conclusions

It is observed that even though the overallcomposition of a ceramic material does not under-go major changes upon firing significant miner-alogical and textural modifications do occur

632

These changes are connected with two reactiontypes a) solid-state replacement of a phase by anew one with little or no compositional variation(ie the muscovite and illite case) and minimizingenergy changes by preserving the original textureand orientation of the replaced phase and b) reac-tion-diffusion at the interface between two miner-al grains with contrasting composition (iecarbonate-silicate interfaces) The latter reaction isenhanced if massviscous transport resulting frompartial melting and finger structure developmentoccurs Non-carbonated clay-rich materials (G)show phyllosilicate destruction at T ranging from700 to 900degC followed by vitrification which issignificant at T gt 1000degC Carbonate-rich materi-als (V) show major phase transformation uponphyllosillicates destruction including new Ca (orMg) silicates and aluminum-silicates formation Inthis later case the vitrification starts at lower Tand it is limited at higher T Therefore it is possi-ble to establish the firing T of ceramic materialswith composition broadly similar to that in ourwork based on the mineralogy Thus a) presenceof phyllosilicates indicates a low 700 to 900degC fir-ing T for the dehydroxylated illite (or muscovite)phase b) dolomite disappears at sup3 700degC whilecalcite is still present at 800degC c) gehleniteappears at 800degC reducing its concentration above1000degC d) diopside and wollastonite appear at800degC (in very low concentrations) as reactionrims between carbonates and silicates Formationof these minerals includes ldquofingeringrdquo develop-ment e) mullite appears at T gt 800degC while littleamounts of sanidine form as microcline disap-pears and f) anorthite develops at T gt1000degC incarbonate-rich bricks

When considering equilibrium thermodynamicdata and calculated reaction T for most of theabove mentioned phases and reactions (Holland ampPowell 1985 and references herein) it isobserved that rapid heating induces significantoverstepping Therefore phases formed are inmost cases metastable a fact consistent with anal-yses of pyrometamorphic assemblages (Brearleyamp Rubie 1990 and references herein)

The formation of non-equilibrium fractal mor-phologies at the carbonate-silicate reaction inter-faces may explain why high-T silicates incarbonate-rich ceramics develop very fast Thisgrain-to-grain reaction model may help explainhigh-T silicate formation in nature (eg contactmetamorphism pyrometamorphism in xenoliths)

The gained knowledge on mineral transforma-tion and the dynamics of interface mass and diffu-sion transport facilitated through ldquofingeringrdquo or

ldquoscallopingrdquo opens new ways to understand thecomplexities of ceramic processing in particularand high-T mineral transformation in general bothin natural (ie geologic settings) and artificialcontexts (eg archaeometric analysis of ancientceramics or the design of new bricks ndashreplicas ofancient onesndash for architectural heritage conserva-tion) However care must be taken when extrapo-lating these results to other raw materialcompositions andor ceramic processing methods

Acknowledgements This work has been finan-cially supported by the CGICYT projectMAT2000-1457 and the Research Group RNM0179 of the Junta de Andaluciacutea (Spain) We thankthe Centro de Instrumentacioacuten Cientifica (CIC) ofthe Granada University for allowing us to use theirSEM-EDX EMPA and XRF facilities We thankK Elert JM Garciacutea-Ruiz M Rodriguez-Gallego and A Sanchez-Navas for fruitful discus-sions and comments We also thank the in-depthreview by Dr F Goumltz-Neunhoeffer and an anony-mous referee

References

Barahona E Huertas F Pozzuoli A Linares J(1985) Firing properties of ceramic clays fromGranada province Spain Miner Petrogr Acta 29A577-590

Boynton RS (1980) Chemistry and technology oflime and limestone 2nd Ed Wiley New York

Brearley AJ (1986) An electron optical study of mus-covite breakdown in pelitic xenoliths duringpyrometamorphism Miner Mag 357 385-397

Brearley AJ amp Rubie DC (1990) Effects of H2O onthe disequilibrium breakdown of muscovite +quartz J Petrol 31 925-956

Butterworth B amp Honeyborne DB (1952) Bricksand clays of the hasting beds Trans Brit CeramSoc 51 211-259

Dondi M Ercolani G Fabbri B Marsigli M (1998)An approach to the chemistry of pyroxenes formedduring the firing of Ca-rich silicate ceramics ClayMinerals 33 443-452

Duminuco P Riccardi MP Messiga B Setti M(1996) Modificazioni tessiturali e mineralogichecome indicatori della dinamica del processo di cot-tura di manufatti ceramici Ceramurgia 26 5 281-288

Evans JL amp White J (1958) Further studies of thethermal decomposition of clays Trans Brit CeramSoc 57 6 298

Everhart JO (1957) Use of auxiliary fluxes to improvestructural clay bodies Bull Am Ceram Soc 36268-271

Carbonate and silicate phase reactions during ceramic firing 633

more gradual loss of water and CO2 as T increases(Table 4)

4 Discussion

Firing up to 700degC induce no significant min-eralogicaltextural transformations in both thecarbonate-rich V and the silicate-rich G sampleswith the exception of the clay minerals Clay min-erals structure is reported to collapse due to dehy-droxylation to resemble an illite-like structure(XRD data) when a T between 450 and 550degC isreached (Evans amp White 1958) A dehydroxylatedphyllosilicate phase structurally different to thehydrated one (Guggenheim et al 1987) is report-ed to exist up to 950degC when complete breakdownof the dehydroxylated illite occurs (Peters amp Iberg1978) Our data indicate that dehydroxylation isnot completed at T ~ 900degC which indicates thatthe kinetics of this process is slower than previ-ously estimated (Evans amp White 1958) (OH)-

released at T gt 700degC may play a significant rolein texture and mineral evolution at high-T as willbe discussed later On the other hand muscoviteshows a slight K deficit at 700degC probably due tothe high diffusion capacity of this element alongmuscovite (001) basal planes at T gt 500degC(Sanchez-Navas amp Galindo-Zaldivar 1993)Sanchez-Navas (1999) showed high-T K-diffu-sion-out of mica is induced by water adsorbedalong mica basal planes resulting from phyllosili-cate dehydroxylation

XRD data shows that dolomite most intenseBragg peak in V bricks significantly reduced itsintensity due to thermal decomposition at lower Tthan calcite main Bragg peak (Fig 3b) Dolomite

Carbonate and silicate phase reactions during ceramic firing 629

Table 4 Bulk composition of Guadix (G) and Viznar (V) samples (XRF results in wt )

LOI loss on ignition

Fig 6 a) ACS (Al2O3-CaO-SiO2) and b) CMS (CaO-MgO-SiO2) compositional diagrams of carbonates andreaction products (EMPA results)

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

starts to decompose according to (1) at ~ 700degC (1atm pressure) while calcite decomposes at higherT (830 ndash 870degC) according to (3) (Boynton 1980)Shoval (1988) proposed a two steps thermaldecomposition for dolomite involving formationof periclase and calcite (2) followed by calcitedecomposition at higher T (3)

dolomite lime periclaseCaMg(CO3)2 reg CaO + MgO + 2CO2 (1)

calcite CaMg(CO3)2 reg CaCO3 + MgO + CO2 (2)

CaCO3 reg CaO + CO2 (3)

However our XRD results showing no increase incalcite 104 Bragg peak intensity as dolomitedecomposes at 700degC (Fig 3b) are not consistentwith Shovalacutes two-steps process

Carbonate decomposition results in shrinkageof the brick (Barahona et al 1985) Upon storageat 22degC and 55 RH burnt lime (CaO) readilytransforms into portlandite according to the fol-lowing reaction (Boynton 1980)

portlanditeCaO + H2O reg Ca(OH)2 (4)

This process generates crystallization pressure inconfined spaces (ie the brick pores originallyoccupied by CaO) resulting in crack development(as observed in Fig 2a) The dark-color rimaround former carbonate grains (Fig 5c) is proba-bly due to limited portlandite carbonation in thepresence of CO2 also resulting in volume increase(Moorehead 1985) and crack development(Butterworth amp Honeyborne 1952) As a commonpractice to avoid these undesired side-effectsbricks that contain carbonates were traditionallyimmersed in water to dissolve and eliminate exist-ing CaO or Ca(OH)2

This effect is less evident in the case of MgOPericlase in the presence of H2O (room T) under-goes a slow transformation (it takes months oreven years Webb 1952) into brucite (Mg(OH)2)and may eventually transform into hydromagne-site (Mg5(CO3)4(OH)24H2O) according to thefollowing reactions (Garavelli et al 1990)

MgO + H2O reg Mg(OH)2 (5)

hydromagnesite5Mg(OH)2 + 4CO2 reg Mg5(CO3)4(OH)24H2O (6)

EMPA results indicate the presence of Ca and Mgoxides however our XRD results do not clearlyshow the presence of Ca and Mg oxides or hydrox-ides mainly due to their low concentrations (lt 10wt ) and because their main diffraction peaks aremasked by other major phase peaks (eg quartzfeldspars hematites) The hydromagnesite mainBragg peak at 581 Aring is not present either

At T sup3 800degC and in the absence of carbonateschanges are constrained to textural modificationsand reactions where a low-T phase is decomposedand fully transformed into a high-T phase withoutmajor compositional changes (Riccardi et al1999) In particular muscovite undergoes a solid-state phase change into a mixture of mullite and K-feldspar (or a melt) at T ranging from 800 to1000degC The composition of high-T muscovite isclose to K-feldspar (EMPA results) although OManalysis points to the presence of mullite Brearley(1986) and Worden et al (1987) showed fine inter-growth of alternating nanometer up to a fewmicrometers thick bands of K-feldspar (or a meltwith composition close to that of K-feldspar) andmullite in pyrometamorphic muscovite which canexplain our EMPA results

The most significant textural and mineralogi-cal changes are observed in samples with carbon-ates when fired at T gt 800degC Also melting T andextent seems to be connected with the presence orabsence of carbonates In general melting starts atlower T (~ 800degC) when carbonates are present(Tite amp Maniatis 1975) Ca and Mg from carbon-ates may act as melting agents (Segnit ampAnderson 1972) but they are reported to some-how limit the extent of vitrification at T gt 1000degC(Everhart 1957 Nuacutentildeez et al 1992) In fact weobserved that vitrification is more extended athigher T when no carbonates exist This may beexplained considering that the phyllosilicate in thematrix of non-carbonate G samples not contribut-ing to the formation of high-T Ca (or Mg) silicatesand releasing significant amounts of H2O due todehydroxylation they can contribute to extensivemelting at high-T as demonstrated by Brearley ampRubie (1990) On the other hand the higher SiO2content in these samples provides a higher amountof potential silicate-rich melt

BSE images of reaction rims between carbon-ates and silicates show that formation of Ca-sili-cates starts at 800degC This is confirmed by XRD(Fig 3b) and EMPA analyses (Table 3) Gehlenitestarts to form at this T by grain-boundary reactionbetween CaO Al2O3 and SiO2 the first derivingfrom former carbonates and the later two fromalready dehydroxylated phyllosilicates (ie a

630

somehow amorphous mixture of 3SiO2Al2O3according to Peters amp Iberg 1978) as follows

illite calcite gehleniteKAl2(Si3Al)O10(OH)2 + 6CaCO3 reg 3Ca2Al2SiO7+ 6CO2 + 2H2O + K2O + 3SiO2 (7)

The formation of gehlenite is striking becauseone could expect the formation of an Al-richpyroxene such as fassaite (Dondi et al 1998)Nevertheless it should be indicated that gehleniteformation is a good example of the so-calledGoldsmithacutes ldquoease of crystallizationrdquo principlewhich states that a silicate phase with all Al in 4-coordination is easier to form (at least in the lab)than a phase that includes Al in both 4- and 6-coor-dination when (OH)- are not present (Goldsmith1953) This is due to the higher energy barrier nec-essary to overcome when Al has to enter in a 6-coordination if compared with a 4-coordination If(OH)- are present this energy barrier is reduceddue to the smaller electrostatic repulsion among(OH)- groups than among O2- in octahedral coor-dination Gehlenite has all Al in tetrahedral coor-dination (Warren 1930) therefore crystallizingpreferentially if compared with other anhydrousCa-Al silicates This simple principle also appliesto the high-T formation of other anhydrous Al-sil-icates as we will discuss later

Wollastonite formed at the carbonate-quartzinterface through the following reaction

wollastoniteCaO + SiO2 reg CaSiO3 (8)

Wollastonite and gehlenite appear at 800degC aT ~100degC lower than previously reported (Petersamp Iberg 1978 Dondi et al 1998) This is proba-bly due to a) the low resolution of the analyticaltechniques used in the past (ie XRD) which didnot allow the detection of the small initial reactionrim between carbonates and silicates as it is evi-denced by BSE images and EMPA analysesandor b) most previous works on high-T phaseformation in ceramic processing use single com-ponents as reactants without considering addition-al phases or previous reaction by-products (Petersamp Iberg 1978) We observed that in a closer-to-reality multicomponent system where interactionsbetween reactants and products occur differentformation T are obtained

At the dolomite-quartz interface diopsidestarts to form at 900degC by the following reaction

dolomite diopside2SiO2 + CaMg(CO3)2 reg CaMgSi2O6 + 2CO (9)

At this T anorthite starts to form at the expensesof calcite and illite (plus quartz)

illite calciteKAl2(Si3Al)O10(OH)2 + 2CaCO3 + 4SiO2 regK-feldspar anorthite

2KAlSi3O8 + 2Ca2Al2SiO8 + 2CO2 + H2O (10)

At higher temperatures (1000 to 1100degC)phyllosilicates have already disappeared in allsamples (V and G) being transformed into sani-dine and mullite plus a melt according to the fol-lowing reaction

illite (muscovite) sanidine 3KAl2(Si3Al)O10(OH)2 + 2SiO2 reg 3KAlSi3O8 +

mulliteAl6Si2O13 + 3H2O + melt (11)

Formation of a melt during the previous reaction(Brearley amp Rubie 1990) is consistent with theobserved ldquocellular structurerdquo of muscovite (Tite ampManiatis 1975)

Brearley (1986) proposed a somehow differentreaction for mullite formation at the expenses ofphengitic muscovite resulting in K-feldspar +mullite + biotite No biotite is detected in ourexperiments most probably due to the low Fe con-tent in V and G micas Brearley amp Rubie (1990)experimentally demonstrated that high-T mus-covite breakdown in the presence of quartz isinfluenced by the crystal size The smallest crys-tals have a strong tendency to react with H2Oresulting in enhanced melt formation while largemuscovite crystals form limited amounts of meltThis may account for the significant vitrificationof the matrix (specially in G samples) while largemuscovite grains (up to some mm in size) still donot show significant melting

Mullite which according to OM analysis firstappears at 800degC is the second most abundantphase (after quartz) at 1100degC in the richer inphyllosilicate G samples It is interesting to men-tion that K-feldspar progressively disappears as Tincreases This is not consistent with publisheddata (Maggetti 1982) It seems that low-T K-feldspar (microcline) is not stable at high-T there-fore it transforms into a high-T K-feldspar (iesanidine) or reacts with lime to form anorthiteBoth sanidine and anorthite are detected in thehigh-T bricks However a higher concentration ofsanidine should be expected to form speciallyconsidering reaction (11) A possible explanationfor the low K-feldspar content at high-T is that K-feldspar might not have enough time to develop itscrystalline structure (sanidine) therefore being

Carbonate and silicate phase reactions during ceramic firing 631

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

concentrated as an amorphous (sanidine-like)phase responsible in part for the rise in the back-ground noise in the XRD results Mullite forma-tion seems to be kinetically faster resulting in awell-developed crystalline phase which givessharp diffraction peaks Goldsmith (1953) indi-cates that mullite could be another example of hisldquosimplexity principlerdquo for the ldquoeaserdquo of crystal-lization if compared with other aluminum sili-cates In fact the mullite structure has a largeramount of Al in 4-coordination and less Al in 6-coordination than sillimanite (Rehak et al 1998)for instance which may account for the abundanceof the former and the non-existence of the latter inthe studied bricks

In carbonate-rich V samples fired at T gt1000degC a significant reduction of gehlenite maindiffraction peak is observed According to Petersamp Iberg (1978) this is due to the reaction of thisphase with quartz forming wollastonite and anor-thite This is consistent with the increase in themain Bragg peak intensity of these later mineralsat 1100degC

At the dolomite-silicate (ie Qtz and Phy)interface it is common to observe wollastoniteeither newly formed or resulting from the destruc-tion of gehlenite The texture of the interface israther complex Protrusion or fingers from thereaction zone penetrate into areas were completetransformation into wollastonite has occurred Theinterface geometry resembles the so-called ldquovis-cous fingersrdquo (Garcia-Ruiz 1992) occurringwhen two fluids or semi-plastic solids of differentcomposition and viscosity are placed in contactInitially the contact zone is a flat surface but lateron it reaches a fingered fractal geometry(Nittmamm et al 1985) In the carbonate-silicateinterface a similar process may have taken placewhere mass transport of the reactants may havebeen enhanced by partial melting The formationof a silicate-rich melt due to clay-minerals melting(Brearley amp Rubie 1990) in the external part ofthe ring in contact with a second inner melt formedby preexisting gehlenite or the eutectic CaO-SiO2would result in viscous finger development Thefractal geometry of the developing viscous fingersmay have enhanced diffusion of mobile ions suchas Ca toward silicate reactive sites Both EDXdata showing Mg enrichment in the dolomite core(Fig 5f) and the compositional trend towardsMgO-rich terms in the CMS diagram (Fig 6b)confirm that Ca is preferentially leached out of theformer dolomite grain Finger formation willresult in a significant increase in interface surfacearea resulting in higher reaction rates Another

alternative explanation for this fingered geometrydevelopment involves partial melting at the car-bonate-silicate interface and mass transport (ieviscous flow) through the clay-quartz porousmatrix resulting in the so-called ldquoscalloped struc-turerdquo proposed by Ortoleva et al (1987) for reac-tion-diffusion processes While the viscous fingermodel requires two melts the second model onlyrequires one melt It is not clear if at any momentconditions for two melt occurrence are reachedHence the second model seems more plausible

In any case the process which is originally dif-fusion driven is transformed into a mass (ie fin-gering) plus diffusion transport which greatlyenhances Ca-silicate formation This is consistentwith the thick reaction rims (up to 250 microm thicksee Fig 5f) that surround former carbonate grainsThese thick Ca-silicate shells would hardly devel-op just by diffusion in the short time-scale of theexperiment (3 hours at the peak T) In fact at800degC (conditions where melting would be verylimited) Ca (and Mg)-silicate rims formed aroundcalcite or dolomite are 2-5 microm thick At 1100degCwhen melting is extensive the reaction rate is sohigh that rims are two orders of magnitude thick-er Therefore mass transport (ie viscous flow)seems to have played a key role in high-T Ca-(Mg-) silicate formation Mass transport is also consis-tent with observation of some areas with differentcomposition (unreacted silicates) trapped betweenfingers resulting in the island geometry observedin Fig 5g It could be expected that without fin-gering the process of Ca- or (Ca-Mg)-silicate for-mation would be self-limiting ie once a rim ofthe newly formed silicate appears diffusion wouldbe jeopardized This shell would preclude or limitthe transport of Ca (andor Mg) from the carbon-ate or it would limit the transport of the lessmobile Si (and Al) from the surrounding silicatetowards the reaction front Therefore it is neces-sary that one of the above described transport-reaction mechanisms takes place This may alsoexplain why phase formation takes place at lowerT than reported previously The presence of multi-ple interfaces among different crystals mayenhance finger development resulting in kineticchanges favoring phase formation at lower T

5 Conclusions

It is observed that even though the overallcomposition of a ceramic material does not under-go major changes upon firing significant miner-alogical and textural modifications do occur

632

These changes are connected with two reactiontypes a) solid-state replacement of a phase by anew one with little or no compositional variation(ie the muscovite and illite case) and minimizingenergy changes by preserving the original textureand orientation of the replaced phase and b) reac-tion-diffusion at the interface between two miner-al grains with contrasting composition (iecarbonate-silicate interfaces) The latter reaction isenhanced if massviscous transport resulting frompartial melting and finger structure developmentoccurs Non-carbonated clay-rich materials (G)show phyllosilicate destruction at T ranging from700 to 900degC followed by vitrification which issignificant at T gt 1000degC Carbonate-rich materi-als (V) show major phase transformation uponphyllosillicates destruction including new Ca (orMg) silicates and aluminum-silicates formation Inthis later case the vitrification starts at lower Tand it is limited at higher T Therefore it is possi-ble to establish the firing T of ceramic materialswith composition broadly similar to that in ourwork based on the mineralogy Thus a) presenceof phyllosilicates indicates a low 700 to 900degC fir-ing T for the dehydroxylated illite (or muscovite)phase b) dolomite disappears at sup3 700degC whilecalcite is still present at 800degC c) gehleniteappears at 800degC reducing its concentration above1000degC d) diopside and wollastonite appear at800degC (in very low concentrations) as reactionrims between carbonates and silicates Formationof these minerals includes ldquofingeringrdquo develop-ment e) mullite appears at T gt 800degC while littleamounts of sanidine form as microcline disap-pears and f) anorthite develops at T gt1000degC incarbonate-rich bricks

When considering equilibrium thermodynamicdata and calculated reaction T for most of theabove mentioned phases and reactions (Holland ampPowell 1985 and references herein) it isobserved that rapid heating induces significantoverstepping Therefore phases formed are inmost cases metastable a fact consistent with anal-yses of pyrometamorphic assemblages (Brearleyamp Rubie 1990 and references herein)

The formation of non-equilibrium fractal mor-phologies at the carbonate-silicate reaction inter-faces may explain why high-T silicates incarbonate-rich ceramics develop very fast Thisgrain-to-grain reaction model may help explainhigh-T silicate formation in nature (eg contactmetamorphism pyrometamorphism in xenoliths)

The gained knowledge on mineral transforma-tion and the dynamics of interface mass and diffu-sion transport facilitated through ldquofingeringrdquo or

ldquoscallopingrdquo opens new ways to understand thecomplexities of ceramic processing in particularand high-T mineral transformation in general bothin natural (ie geologic settings) and artificialcontexts (eg archaeometric analysis of ancientceramics or the design of new bricks ndashreplicas ofancient onesndash for architectural heritage conserva-tion) However care must be taken when extrapo-lating these results to other raw materialcompositions andor ceramic processing methods

Acknowledgements This work has been finan-cially supported by the CGICYT projectMAT2000-1457 and the Research Group RNM0179 of the Junta de Andaluciacutea (Spain) We thankthe Centro de Instrumentacioacuten Cientifica (CIC) ofthe Granada University for allowing us to use theirSEM-EDX EMPA and XRF facilities We thankK Elert JM Garciacutea-Ruiz M Rodriguez-Gallego and A Sanchez-Navas for fruitful discus-sions and comments We also thank the in-depthreview by Dr F Goumltz-Neunhoeffer and an anony-mous referee

References

Barahona E Huertas F Pozzuoli A Linares J(1985) Firing properties of ceramic clays fromGranada province Spain Miner Petrogr Acta 29A577-590

Boynton RS (1980) Chemistry and technology oflime and limestone 2nd Ed Wiley New York

Brearley AJ (1986) An electron optical study of mus-covite breakdown in pelitic xenoliths duringpyrometamorphism Miner Mag 357 385-397

Brearley AJ amp Rubie DC (1990) Effects of H2O onthe disequilibrium breakdown of muscovite +quartz J Petrol 31 925-956

Butterworth B amp Honeyborne DB (1952) Bricksand clays of the hasting beds Trans Brit CeramSoc 51 211-259

Dondi M Ercolani G Fabbri B Marsigli M (1998)An approach to the chemistry of pyroxenes formedduring the firing of Ca-rich silicate ceramics ClayMinerals 33 443-452

Duminuco P Riccardi MP Messiga B Setti M(1996) Modificazioni tessiturali e mineralogichecome indicatori della dinamica del processo di cot-tura di manufatti ceramici Ceramurgia 26 5 281-288

Evans JL amp White J (1958) Further studies of thethermal decomposition of clays Trans Brit CeramSoc 57 6 298

Everhart JO (1957) Use of auxiliary fluxes to improvestructural clay bodies Bull Am Ceram Soc 36268-271

Carbonate and silicate phase reactions during ceramic firing 633

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

starts to decompose according to (1) at ~ 700degC (1atm pressure) while calcite decomposes at higherT (830 ndash 870degC) according to (3) (Boynton 1980)Shoval (1988) proposed a two steps thermaldecomposition for dolomite involving formationof periclase and calcite (2) followed by calcitedecomposition at higher T (3)

dolomite lime periclaseCaMg(CO3)2 reg CaO + MgO + 2CO2 (1)

calcite CaMg(CO3)2 reg CaCO3 + MgO + CO2 (2)

CaCO3 reg CaO + CO2 (3)

However our XRD results showing no increase incalcite 104 Bragg peak intensity as dolomitedecomposes at 700degC (Fig 3b) are not consistentwith Shovalacutes two-steps process

Carbonate decomposition results in shrinkageof the brick (Barahona et al 1985) Upon storageat 22degC and 55 RH burnt lime (CaO) readilytransforms into portlandite according to the fol-lowing reaction (Boynton 1980)

portlanditeCaO + H2O reg Ca(OH)2 (4)

This process generates crystallization pressure inconfined spaces (ie the brick pores originallyoccupied by CaO) resulting in crack development(as observed in Fig 2a) The dark-color rimaround former carbonate grains (Fig 5c) is proba-bly due to limited portlandite carbonation in thepresence of CO2 also resulting in volume increase(Moorehead 1985) and crack development(Butterworth amp Honeyborne 1952) As a commonpractice to avoid these undesired side-effectsbricks that contain carbonates were traditionallyimmersed in water to dissolve and eliminate exist-ing CaO or Ca(OH)2

This effect is less evident in the case of MgOPericlase in the presence of H2O (room T) under-goes a slow transformation (it takes months oreven years Webb 1952) into brucite (Mg(OH)2)and may eventually transform into hydromagne-site (Mg5(CO3)4(OH)24H2O) according to thefollowing reactions (Garavelli et al 1990)

MgO + H2O reg Mg(OH)2 (5)

hydromagnesite5Mg(OH)2 + 4CO2 reg Mg5(CO3)4(OH)24H2O (6)

EMPA results indicate the presence of Ca and Mgoxides however our XRD results do not clearlyshow the presence of Ca and Mg oxides or hydrox-ides mainly due to their low concentrations (lt 10wt ) and because their main diffraction peaks aremasked by other major phase peaks (eg quartzfeldspars hematites) The hydromagnesite mainBragg peak at 581 Aring is not present either

At T sup3 800degC and in the absence of carbonateschanges are constrained to textural modificationsand reactions where a low-T phase is decomposedand fully transformed into a high-T phase withoutmajor compositional changes (Riccardi et al1999) In particular muscovite undergoes a solid-state phase change into a mixture of mullite and K-feldspar (or a melt) at T ranging from 800 to1000degC The composition of high-T muscovite isclose to K-feldspar (EMPA results) although OManalysis points to the presence of mullite Brearley(1986) and Worden et al (1987) showed fine inter-growth of alternating nanometer up to a fewmicrometers thick bands of K-feldspar (or a meltwith composition close to that of K-feldspar) andmullite in pyrometamorphic muscovite which canexplain our EMPA results

The most significant textural and mineralogi-cal changes are observed in samples with carbon-ates when fired at T gt 800degC Also melting T andextent seems to be connected with the presence orabsence of carbonates In general melting starts atlower T (~ 800degC) when carbonates are present(Tite amp Maniatis 1975) Ca and Mg from carbon-ates may act as melting agents (Segnit ampAnderson 1972) but they are reported to some-how limit the extent of vitrification at T gt 1000degC(Everhart 1957 Nuacutentildeez et al 1992) In fact weobserved that vitrification is more extended athigher T when no carbonates exist This may beexplained considering that the phyllosilicate in thematrix of non-carbonate G samples not contribut-ing to the formation of high-T Ca (or Mg) silicatesand releasing significant amounts of H2O due todehydroxylation they can contribute to extensivemelting at high-T as demonstrated by Brearley ampRubie (1990) On the other hand the higher SiO2content in these samples provides a higher amountof potential silicate-rich melt

BSE images of reaction rims between carbon-ates and silicates show that formation of Ca-sili-cates starts at 800degC This is confirmed by XRD(Fig 3b) and EMPA analyses (Table 3) Gehlenitestarts to form at this T by grain-boundary reactionbetween CaO Al2O3 and SiO2 the first derivingfrom former carbonates and the later two fromalready dehydroxylated phyllosilicates (ie a

630

somehow amorphous mixture of 3SiO2Al2O3according to Peters amp Iberg 1978) as follows

illite calcite gehleniteKAl2(Si3Al)O10(OH)2 + 6CaCO3 reg 3Ca2Al2SiO7+ 6CO2 + 2H2O + K2O + 3SiO2 (7)

The formation of gehlenite is striking becauseone could expect the formation of an Al-richpyroxene such as fassaite (Dondi et al 1998)Nevertheless it should be indicated that gehleniteformation is a good example of the so-calledGoldsmithacutes ldquoease of crystallizationrdquo principlewhich states that a silicate phase with all Al in 4-coordination is easier to form (at least in the lab)than a phase that includes Al in both 4- and 6-coor-dination when (OH)- are not present (Goldsmith1953) This is due to the higher energy barrier nec-essary to overcome when Al has to enter in a 6-coordination if compared with a 4-coordination If(OH)- are present this energy barrier is reduceddue to the smaller electrostatic repulsion among(OH)- groups than among O2- in octahedral coor-dination Gehlenite has all Al in tetrahedral coor-dination (Warren 1930) therefore crystallizingpreferentially if compared with other anhydrousCa-Al silicates This simple principle also appliesto the high-T formation of other anhydrous Al-sil-icates as we will discuss later

Wollastonite formed at the carbonate-quartzinterface through the following reaction

wollastoniteCaO + SiO2 reg CaSiO3 (8)

Wollastonite and gehlenite appear at 800degC aT ~100degC lower than previously reported (Petersamp Iberg 1978 Dondi et al 1998) This is proba-bly due to a) the low resolution of the analyticaltechniques used in the past (ie XRD) which didnot allow the detection of the small initial reactionrim between carbonates and silicates as it is evi-denced by BSE images and EMPA analysesandor b) most previous works on high-T phaseformation in ceramic processing use single com-ponents as reactants without considering addition-al phases or previous reaction by-products (Petersamp Iberg 1978) We observed that in a closer-to-reality multicomponent system where interactionsbetween reactants and products occur differentformation T are obtained

At the dolomite-quartz interface diopsidestarts to form at 900degC by the following reaction

dolomite diopside2SiO2 + CaMg(CO3)2 reg CaMgSi2O6 + 2CO (9)

At this T anorthite starts to form at the expensesof calcite and illite (plus quartz)

illite calciteKAl2(Si3Al)O10(OH)2 + 2CaCO3 + 4SiO2 regK-feldspar anorthite

2KAlSi3O8 + 2Ca2Al2SiO8 + 2CO2 + H2O (10)

At higher temperatures (1000 to 1100degC)phyllosilicates have already disappeared in allsamples (V and G) being transformed into sani-dine and mullite plus a melt according to the fol-lowing reaction

illite (muscovite) sanidine 3KAl2(Si3Al)O10(OH)2 + 2SiO2 reg 3KAlSi3O8 +

mulliteAl6Si2O13 + 3H2O + melt (11)

Formation of a melt during the previous reaction(Brearley amp Rubie 1990) is consistent with theobserved ldquocellular structurerdquo of muscovite (Tite ampManiatis 1975)

Brearley (1986) proposed a somehow differentreaction for mullite formation at the expenses ofphengitic muscovite resulting in K-feldspar +mullite + biotite No biotite is detected in ourexperiments most probably due to the low Fe con-tent in V and G micas Brearley amp Rubie (1990)experimentally demonstrated that high-T mus-covite breakdown in the presence of quartz isinfluenced by the crystal size The smallest crys-tals have a strong tendency to react with H2Oresulting in enhanced melt formation while largemuscovite crystals form limited amounts of meltThis may account for the significant vitrificationof the matrix (specially in G samples) while largemuscovite grains (up to some mm in size) still donot show significant melting

Mullite which according to OM analysis firstappears at 800degC is the second most abundantphase (after quartz) at 1100degC in the richer inphyllosilicate G samples It is interesting to men-tion that K-feldspar progressively disappears as Tincreases This is not consistent with publisheddata (Maggetti 1982) It seems that low-T K-feldspar (microcline) is not stable at high-T there-fore it transforms into a high-T K-feldspar (iesanidine) or reacts with lime to form anorthiteBoth sanidine and anorthite are detected in thehigh-T bricks However a higher concentration ofsanidine should be expected to form speciallyconsidering reaction (11) A possible explanationfor the low K-feldspar content at high-T is that K-feldspar might not have enough time to develop itscrystalline structure (sanidine) therefore being

Carbonate and silicate phase reactions during ceramic firing 631

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

concentrated as an amorphous (sanidine-like)phase responsible in part for the rise in the back-ground noise in the XRD results Mullite forma-tion seems to be kinetically faster resulting in awell-developed crystalline phase which givessharp diffraction peaks Goldsmith (1953) indi-cates that mullite could be another example of hisldquosimplexity principlerdquo for the ldquoeaserdquo of crystal-lization if compared with other aluminum sili-cates In fact the mullite structure has a largeramount of Al in 4-coordination and less Al in 6-coordination than sillimanite (Rehak et al 1998)for instance which may account for the abundanceof the former and the non-existence of the latter inthe studied bricks

In carbonate-rich V samples fired at T gt1000degC a significant reduction of gehlenite maindiffraction peak is observed According to Petersamp Iberg (1978) this is due to the reaction of thisphase with quartz forming wollastonite and anor-thite This is consistent with the increase in themain Bragg peak intensity of these later mineralsat 1100degC

At the dolomite-silicate (ie Qtz and Phy)interface it is common to observe wollastoniteeither newly formed or resulting from the destruc-tion of gehlenite The texture of the interface israther complex Protrusion or fingers from thereaction zone penetrate into areas were completetransformation into wollastonite has occurred Theinterface geometry resembles the so-called ldquovis-cous fingersrdquo (Garcia-Ruiz 1992) occurringwhen two fluids or semi-plastic solids of differentcomposition and viscosity are placed in contactInitially the contact zone is a flat surface but lateron it reaches a fingered fractal geometry(Nittmamm et al 1985) In the carbonate-silicateinterface a similar process may have taken placewhere mass transport of the reactants may havebeen enhanced by partial melting The formationof a silicate-rich melt due to clay-minerals melting(Brearley amp Rubie 1990) in the external part ofthe ring in contact with a second inner melt formedby preexisting gehlenite or the eutectic CaO-SiO2would result in viscous finger development Thefractal geometry of the developing viscous fingersmay have enhanced diffusion of mobile ions suchas Ca toward silicate reactive sites Both EDXdata showing Mg enrichment in the dolomite core(Fig 5f) and the compositional trend towardsMgO-rich terms in the CMS diagram (Fig 6b)confirm that Ca is preferentially leached out of theformer dolomite grain Finger formation willresult in a significant increase in interface surfacearea resulting in higher reaction rates Another

alternative explanation for this fingered geometrydevelopment involves partial melting at the car-bonate-silicate interface and mass transport (ieviscous flow) through the clay-quartz porousmatrix resulting in the so-called ldquoscalloped struc-turerdquo proposed by Ortoleva et al (1987) for reac-tion-diffusion processes While the viscous fingermodel requires two melts the second model onlyrequires one melt It is not clear if at any momentconditions for two melt occurrence are reachedHence the second model seems more plausible

In any case the process which is originally dif-fusion driven is transformed into a mass (ie fin-gering) plus diffusion transport which greatlyenhances Ca-silicate formation This is consistentwith the thick reaction rims (up to 250 microm thicksee Fig 5f) that surround former carbonate grainsThese thick Ca-silicate shells would hardly devel-op just by diffusion in the short time-scale of theexperiment (3 hours at the peak T) In fact at800degC (conditions where melting would be verylimited) Ca (and Mg)-silicate rims formed aroundcalcite or dolomite are 2-5 microm thick At 1100degCwhen melting is extensive the reaction rate is sohigh that rims are two orders of magnitude thick-er Therefore mass transport (ie viscous flow)seems to have played a key role in high-T Ca-(Mg-) silicate formation Mass transport is also consis-tent with observation of some areas with differentcomposition (unreacted silicates) trapped betweenfingers resulting in the island geometry observedin Fig 5g It could be expected that without fin-gering the process of Ca- or (Ca-Mg)-silicate for-mation would be self-limiting ie once a rim ofthe newly formed silicate appears diffusion wouldbe jeopardized This shell would preclude or limitthe transport of Ca (andor Mg) from the carbon-ate or it would limit the transport of the lessmobile Si (and Al) from the surrounding silicatetowards the reaction front Therefore it is neces-sary that one of the above described transport-reaction mechanisms takes place This may alsoexplain why phase formation takes place at lowerT than reported previously The presence of multi-ple interfaces among different crystals mayenhance finger development resulting in kineticchanges favoring phase formation at lower T

5 Conclusions

It is observed that even though the overallcomposition of a ceramic material does not under-go major changes upon firing significant miner-alogical and textural modifications do occur

632

These changes are connected with two reactiontypes a) solid-state replacement of a phase by anew one with little or no compositional variation(ie the muscovite and illite case) and minimizingenergy changes by preserving the original textureand orientation of the replaced phase and b) reac-tion-diffusion at the interface between two miner-al grains with contrasting composition (iecarbonate-silicate interfaces) The latter reaction isenhanced if massviscous transport resulting frompartial melting and finger structure developmentoccurs Non-carbonated clay-rich materials (G)show phyllosilicate destruction at T ranging from700 to 900degC followed by vitrification which issignificant at T gt 1000degC Carbonate-rich materi-als (V) show major phase transformation uponphyllosillicates destruction including new Ca (orMg) silicates and aluminum-silicates formation Inthis later case the vitrification starts at lower Tand it is limited at higher T Therefore it is possi-ble to establish the firing T of ceramic materialswith composition broadly similar to that in ourwork based on the mineralogy Thus a) presenceof phyllosilicates indicates a low 700 to 900degC fir-ing T for the dehydroxylated illite (or muscovite)phase b) dolomite disappears at sup3 700degC whilecalcite is still present at 800degC c) gehleniteappears at 800degC reducing its concentration above1000degC d) diopside and wollastonite appear at800degC (in very low concentrations) as reactionrims between carbonates and silicates Formationof these minerals includes ldquofingeringrdquo develop-ment e) mullite appears at T gt 800degC while littleamounts of sanidine form as microcline disap-pears and f) anorthite develops at T gt1000degC incarbonate-rich bricks

When considering equilibrium thermodynamicdata and calculated reaction T for most of theabove mentioned phases and reactions (Holland ampPowell 1985 and references herein) it isobserved that rapid heating induces significantoverstepping Therefore phases formed are inmost cases metastable a fact consistent with anal-yses of pyrometamorphic assemblages (Brearleyamp Rubie 1990 and references herein)

The formation of non-equilibrium fractal mor-phologies at the carbonate-silicate reaction inter-faces may explain why high-T silicates incarbonate-rich ceramics develop very fast Thisgrain-to-grain reaction model may help explainhigh-T silicate formation in nature (eg contactmetamorphism pyrometamorphism in xenoliths)

The gained knowledge on mineral transforma-tion and the dynamics of interface mass and diffu-sion transport facilitated through ldquofingeringrdquo or

ldquoscallopingrdquo opens new ways to understand thecomplexities of ceramic processing in particularand high-T mineral transformation in general bothin natural (ie geologic settings) and artificialcontexts (eg archaeometric analysis of ancientceramics or the design of new bricks ndashreplicas ofancient onesndash for architectural heritage conserva-tion) However care must be taken when extrapo-lating these results to other raw materialcompositions andor ceramic processing methods

Acknowledgements This work has been finan-cially supported by the CGICYT projectMAT2000-1457 and the Research Group RNM0179 of the Junta de Andaluciacutea (Spain) We thankthe Centro de Instrumentacioacuten Cientifica (CIC) ofthe Granada University for allowing us to use theirSEM-EDX EMPA and XRF facilities We thankK Elert JM Garciacutea-Ruiz M Rodriguez-Gallego and A Sanchez-Navas for fruitful discus-sions and comments We also thank the in-depthreview by Dr F Goumltz-Neunhoeffer and an anony-mous referee

References

Barahona E Huertas F Pozzuoli A Linares J(1985) Firing properties of ceramic clays fromGranada province Spain Miner Petrogr Acta 29A577-590

Boynton RS (1980) Chemistry and technology oflime and limestone 2nd Ed Wiley New York

Brearley AJ (1986) An electron optical study of mus-covite breakdown in pelitic xenoliths duringpyrometamorphism Miner Mag 357 385-397

Brearley AJ amp Rubie DC (1990) Effects of H2O onthe disequilibrium breakdown of muscovite +quartz J Petrol 31 925-956

Butterworth B amp Honeyborne DB (1952) Bricksand clays of the hasting beds Trans Brit CeramSoc 51 211-259

Dondi M Ercolani G Fabbri B Marsigli M (1998)An approach to the chemistry of pyroxenes formedduring the firing of Ca-rich silicate ceramics ClayMinerals 33 443-452

Duminuco P Riccardi MP Messiga B Setti M(1996) Modificazioni tessiturali e mineralogichecome indicatori della dinamica del processo di cot-tura di manufatti ceramici Ceramurgia 26 5 281-288

Evans JL amp White J (1958) Further studies of thethermal decomposition of clays Trans Brit CeramSoc 57 6 298

Everhart JO (1957) Use of auxiliary fluxes to improvestructural clay bodies Bull Am Ceram Soc 36268-271

Carbonate and silicate phase reactions during ceramic firing 633

somehow amorphous mixture of 3SiO2Al2O3according to Peters amp Iberg 1978) as follows

illite calcite gehleniteKAl2(Si3Al)O10(OH)2 + 6CaCO3 reg 3Ca2Al2SiO7+ 6CO2 + 2H2O + K2O + 3SiO2 (7)

The formation of gehlenite is striking becauseone could expect the formation of an Al-richpyroxene such as fassaite (Dondi et al 1998)Nevertheless it should be indicated that gehleniteformation is a good example of the so-calledGoldsmithacutes ldquoease of crystallizationrdquo principlewhich states that a silicate phase with all Al in 4-coordination is easier to form (at least in the lab)than a phase that includes Al in both 4- and 6-coor-dination when (OH)- are not present (Goldsmith1953) This is due to the higher energy barrier nec-essary to overcome when Al has to enter in a 6-coordination if compared with a 4-coordination If(OH)- are present this energy barrier is reduceddue to the smaller electrostatic repulsion among(OH)- groups than among O2- in octahedral coor-dination Gehlenite has all Al in tetrahedral coor-dination (Warren 1930) therefore crystallizingpreferentially if compared with other anhydrousCa-Al silicates This simple principle also appliesto the high-T formation of other anhydrous Al-sil-icates as we will discuss later

Wollastonite formed at the carbonate-quartzinterface through the following reaction

wollastoniteCaO + SiO2 reg CaSiO3 (8)

Wollastonite and gehlenite appear at 800degC aT ~100degC lower than previously reported (Petersamp Iberg 1978 Dondi et al 1998) This is proba-bly due to a) the low resolution of the analyticaltechniques used in the past (ie XRD) which didnot allow the detection of the small initial reactionrim between carbonates and silicates as it is evi-denced by BSE images and EMPA analysesandor b) most previous works on high-T phaseformation in ceramic processing use single com-ponents as reactants without considering addition-al phases or previous reaction by-products (Petersamp Iberg 1978) We observed that in a closer-to-reality multicomponent system where interactionsbetween reactants and products occur differentformation T are obtained

At the dolomite-quartz interface diopsidestarts to form at 900degC by the following reaction

dolomite diopside2SiO2 + CaMg(CO3)2 reg CaMgSi2O6 + 2CO (9)

At this T anorthite starts to form at the expensesof calcite and illite (plus quartz)

illite calciteKAl2(Si3Al)O10(OH)2 + 2CaCO3 + 4SiO2 regK-feldspar anorthite

2KAlSi3O8 + 2Ca2Al2SiO8 + 2CO2 + H2O (10)

At higher temperatures (1000 to 1100degC)phyllosilicates have already disappeared in allsamples (V and G) being transformed into sani-dine and mullite plus a melt according to the fol-lowing reaction

illite (muscovite) sanidine 3KAl2(Si3Al)O10(OH)2 + 2SiO2 reg 3KAlSi3O8 +

mulliteAl6Si2O13 + 3H2O + melt (11)

Formation of a melt during the previous reaction(Brearley amp Rubie 1990) is consistent with theobserved ldquocellular structurerdquo of muscovite (Tite ampManiatis 1975)

Brearley (1986) proposed a somehow differentreaction for mullite formation at the expenses ofphengitic muscovite resulting in K-feldspar +mullite + biotite No biotite is detected in ourexperiments most probably due to the low Fe con-tent in V and G micas Brearley amp Rubie (1990)experimentally demonstrated that high-T mus-covite breakdown in the presence of quartz isinfluenced by the crystal size The smallest crys-tals have a strong tendency to react with H2Oresulting in enhanced melt formation while largemuscovite crystals form limited amounts of meltThis may account for the significant vitrificationof the matrix (specially in G samples) while largemuscovite grains (up to some mm in size) still donot show significant melting

Mullite which according to OM analysis firstappears at 800degC is the second most abundantphase (after quartz) at 1100degC in the richer inphyllosilicate G samples It is interesting to men-tion that K-feldspar progressively disappears as Tincreases This is not consistent with publisheddata (Maggetti 1982) It seems that low-T K-feldspar (microcline) is not stable at high-T there-fore it transforms into a high-T K-feldspar (iesanidine) or reacts with lime to form anorthiteBoth sanidine and anorthite are detected in thehigh-T bricks However a higher concentration ofsanidine should be expected to form speciallyconsidering reaction (11) A possible explanationfor the low K-feldspar content at high-T is that K-feldspar might not have enough time to develop itscrystalline structure (sanidine) therefore being

Carbonate and silicate phase reactions during ceramic firing 631

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

concentrated as an amorphous (sanidine-like)phase responsible in part for the rise in the back-ground noise in the XRD results Mullite forma-tion seems to be kinetically faster resulting in awell-developed crystalline phase which givessharp diffraction peaks Goldsmith (1953) indi-cates that mullite could be another example of hisldquosimplexity principlerdquo for the ldquoeaserdquo of crystal-lization if compared with other aluminum sili-cates In fact the mullite structure has a largeramount of Al in 4-coordination and less Al in 6-coordination than sillimanite (Rehak et al 1998)for instance which may account for the abundanceof the former and the non-existence of the latter inthe studied bricks

In carbonate-rich V samples fired at T gt1000degC a significant reduction of gehlenite maindiffraction peak is observed According to Petersamp Iberg (1978) this is due to the reaction of thisphase with quartz forming wollastonite and anor-thite This is consistent with the increase in themain Bragg peak intensity of these later mineralsat 1100degC

At the dolomite-silicate (ie Qtz and Phy)interface it is common to observe wollastoniteeither newly formed or resulting from the destruc-tion of gehlenite The texture of the interface israther complex Protrusion or fingers from thereaction zone penetrate into areas were completetransformation into wollastonite has occurred Theinterface geometry resembles the so-called ldquovis-cous fingersrdquo (Garcia-Ruiz 1992) occurringwhen two fluids or semi-plastic solids of differentcomposition and viscosity are placed in contactInitially the contact zone is a flat surface but lateron it reaches a fingered fractal geometry(Nittmamm et al 1985) In the carbonate-silicateinterface a similar process may have taken placewhere mass transport of the reactants may havebeen enhanced by partial melting The formationof a silicate-rich melt due to clay-minerals melting(Brearley amp Rubie 1990) in the external part ofthe ring in contact with a second inner melt formedby preexisting gehlenite or the eutectic CaO-SiO2would result in viscous finger development Thefractal geometry of the developing viscous fingersmay have enhanced diffusion of mobile ions suchas Ca toward silicate reactive sites Both EDXdata showing Mg enrichment in the dolomite core(Fig 5f) and the compositional trend towardsMgO-rich terms in the CMS diagram (Fig 6b)confirm that Ca is preferentially leached out of theformer dolomite grain Finger formation willresult in a significant increase in interface surfacearea resulting in higher reaction rates Another

alternative explanation for this fingered geometrydevelopment involves partial melting at the car-bonate-silicate interface and mass transport (ieviscous flow) through the clay-quartz porousmatrix resulting in the so-called ldquoscalloped struc-turerdquo proposed by Ortoleva et al (1987) for reac-tion-diffusion processes While the viscous fingermodel requires two melts the second model onlyrequires one melt It is not clear if at any momentconditions for two melt occurrence are reachedHence the second model seems more plausible

In any case the process which is originally dif-fusion driven is transformed into a mass (ie fin-gering) plus diffusion transport which greatlyenhances Ca-silicate formation This is consistentwith the thick reaction rims (up to 250 microm thicksee Fig 5f) that surround former carbonate grainsThese thick Ca-silicate shells would hardly devel-op just by diffusion in the short time-scale of theexperiment (3 hours at the peak T) In fact at800degC (conditions where melting would be verylimited) Ca (and Mg)-silicate rims formed aroundcalcite or dolomite are 2-5 microm thick At 1100degCwhen melting is extensive the reaction rate is sohigh that rims are two orders of magnitude thick-er Therefore mass transport (ie viscous flow)seems to have played a key role in high-T Ca-(Mg-) silicate formation Mass transport is also consis-tent with observation of some areas with differentcomposition (unreacted silicates) trapped betweenfingers resulting in the island geometry observedin Fig 5g It could be expected that without fin-gering the process of Ca- or (Ca-Mg)-silicate for-mation would be self-limiting ie once a rim ofthe newly formed silicate appears diffusion wouldbe jeopardized This shell would preclude or limitthe transport of Ca (andor Mg) from the carbon-ate or it would limit the transport of the lessmobile Si (and Al) from the surrounding silicatetowards the reaction front Therefore it is neces-sary that one of the above described transport-reaction mechanisms takes place This may alsoexplain why phase formation takes place at lowerT than reported previously The presence of multi-ple interfaces among different crystals mayenhance finger development resulting in kineticchanges favoring phase formation at lower T

5 Conclusions

It is observed that even though the overallcomposition of a ceramic material does not under-go major changes upon firing significant miner-alogical and textural modifications do occur

632

These changes are connected with two reactiontypes a) solid-state replacement of a phase by anew one with little or no compositional variation(ie the muscovite and illite case) and minimizingenergy changes by preserving the original textureand orientation of the replaced phase and b) reac-tion-diffusion at the interface between two miner-al grains with contrasting composition (iecarbonate-silicate interfaces) The latter reaction isenhanced if massviscous transport resulting frompartial melting and finger structure developmentoccurs Non-carbonated clay-rich materials (G)show phyllosilicate destruction at T ranging from700 to 900degC followed by vitrification which issignificant at T gt 1000degC Carbonate-rich materi-als (V) show major phase transformation uponphyllosillicates destruction including new Ca (orMg) silicates and aluminum-silicates formation Inthis later case the vitrification starts at lower Tand it is limited at higher T Therefore it is possi-ble to establish the firing T of ceramic materialswith composition broadly similar to that in ourwork based on the mineralogy Thus a) presenceof phyllosilicates indicates a low 700 to 900degC fir-ing T for the dehydroxylated illite (or muscovite)phase b) dolomite disappears at sup3 700degC whilecalcite is still present at 800degC c) gehleniteappears at 800degC reducing its concentration above1000degC d) diopside and wollastonite appear at800degC (in very low concentrations) as reactionrims between carbonates and silicates Formationof these minerals includes ldquofingeringrdquo develop-ment e) mullite appears at T gt 800degC while littleamounts of sanidine form as microcline disap-pears and f) anorthite develops at T gt1000degC incarbonate-rich bricks

When considering equilibrium thermodynamicdata and calculated reaction T for most of theabove mentioned phases and reactions (Holland ampPowell 1985 and references herein) it isobserved that rapid heating induces significantoverstepping Therefore phases formed are inmost cases metastable a fact consistent with anal-yses of pyrometamorphic assemblages (Brearleyamp Rubie 1990 and references herein)

The formation of non-equilibrium fractal mor-phologies at the carbonate-silicate reaction inter-faces may explain why high-T silicates incarbonate-rich ceramics develop very fast Thisgrain-to-grain reaction model may help explainhigh-T silicate formation in nature (eg contactmetamorphism pyrometamorphism in xenoliths)

The gained knowledge on mineral transforma-tion and the dynamics of interface mass and diffu-sion transport facilitated through ldquofingeringrdquo or

ldquoscallopingrdquo opens new ways to understand thecomplexities of ceramic processing in particularand high-T mineral transformation in general bothin natural (ie geologic settings) and artificialcontexts (eg archaeometric analysis of ancientceramics or the design of new bricks ndashreplicas ofancient onesndash for architectural heritage conserva-tion) However care must be taken when extrapo-lating these results to other raw materialcompositions andor ceramic processing methods

Acknowledgements This work has been finan-cially supported by the CGICYT projectMAT2000-1457 and the Research Group RNM0179 of the Junta de Andaluciacutea (Spain) We thankthe Centro de Instrumentacioacuten Cientifica (CIC) ofthe Granada University for allowing us to use theirSEM-EDX EMPA and XRF facilities We thankK Elert JM Garciacutea-Ruiz M Rodriguez-Gallego and A Sanchez-Navas for fruitful discus-sions and comments We also thank the in-depthreview by Dr F Goumltz-Neunhoeffer and an anony-mous referee

References

Barahona E Huertas F Pozzuoli A Linares J(1985) Firing properties of ceramic clays fromGranada province Spain Miner Petrogr Acta 29A577-590

Boynton RS (1980) Chemistry and technology oflime and limestone 2nd Ed Wiley New York

Brearley AJ (1986) An electron optical study of mus-covite breakdown in pelitic xenoliths duringpyrometamorphism Miner Mag 357 385-397

Brearley AJ amp Rubie DC (1990) Effects of H2O onthe disequilibrium breakdown of muscovite +quartz J Petrol 31 925-956

Butterworth B amp Honeyborne DB (1952) Bricksand clays of the hasting beds Trans Brit CeramSoc 51 211-259

Dondi M Ercolani G Fabbri B Marsigli M (1998)An approach to the chemistry of pyroxenes formedduring the firing of Ca-rich silicate ceramics ClayMinerals 33 443-452

Duminuco P Riccardi MP Messiga B Setti M(1996) Modificazioni tessiturali e mineralogichecome indicatori della dinamica del processo di cot-tura di manufatti ceramici Ceramurgia 26 5 281-288

Evans JL amp White J (1958) Further studies of thethermal decomposition of clays Trans Brit CeramSoc 57 6 298

Everhart JO (1957) Use of auxiliary fluxes to improvestructural clay bodies Bull Am Ceram Soc 36268-271

Carbonate and silicate phase reactions during ceramic firing 633

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

concentrated as an amorphous (sanidine-like)phase responsible in part for the rise in the back-ground noise in the XRD results Mullite forma-tion seems to be kinetically faster resulting in awell-developed crystalline phase which givessharp diffraction peaks Goldsmith (1953) indi-cates that mullite could be another example of hisldquosimplexity principlerdquo for the ldquoeaserdquo of crystal-lization if compared with other aluminum sili-cates In fact the mullite structure has a largeramount of Al in 4-coordination and less Al in 6-coordination than sillimanite (Rehak et al 1998)for instance which may account for the abundanceof the former and the non-existence of the latter inthe studied bricks

In carbonate-rich V samples fired at T gt1000degC a significant reduction of gehlenite maindiffraction peak is observed According to Petersamp Iberg (1978) this is due to the reaction of thisphase with quartz forming wollastonite and anor-thite This is consistent with the increase in themain Bragg peak intensity of these later mineralsat 1100degC

At the dolomite-silicate (ie Qtz and Phy)interface it is common to observe wollastoniteeither newly formed or resulting from the destruc-tion of gehlenite The texture of the interface israther complex Protrusion or fingers from thereaction zone penetrate into areas were completetransformation into wollastonite has occurred Theinterface geometry resembles the so-called ldquovis-cous fingersrdquo (Garcia-Ruiz 1992) occurringwhen two fluids or semi-plastic solids of differentcomposition and viscosity are placed in contactInitially the contact zone is a flat surface but lateron it reaches a fingered fractal geometry(Nittmamm et al 1985) In the carbonate-silicateinterface a similar process may have taken placewhere mass transport of the reactants may havebeen enhanced by partial melting The formationof a silicate-rich melt due to clay-minerals melting(Brearley amp Rubie 1990) in the external part ofthe ring in contact with a second inner melt formedby preexisting gehlenite or the eutectic CaO-SiO2would result in viscous finger development Thefractal geometry of the developing viscous fingersmay have enhanced diffusion of mobile ions suchas Ca toward silicate reactive sites Both EDXdata showing Mg enrichment in the dolomite core(Fig 5f) and the compositional trend towardsMgO-rich terms in the CMS diagram (Fig 6b)confirm that Ca is preferentially leached out of theformer dolomite grain Finger formation willresult in a significant increase in interface surfacearea resulting in higher reaction rates Another

alternative explanation for this fingered geometrydevelopment involves partial melting at the car-bonate-silicate interface and mass transport (ieviscous flow) through the clay-quartz porousmatrix resulting in the so-called ldquoscalloped struc-turerdquo proposed by Ortoleva et al (1987) for reac-tion-diffusion processes While the viscous fingermodel requires two melts the second model onlyrequires one melt It is not clear if at any momentconditions for two melt occurrence are reachedHence the second model seems more plausible

In any case the process which is originally dif-fusion driven is transformed into a mass (ie fin-gering) plus diffusion transport which greatlyenhances Ca-silicate formation This is consistentwith the thick reaction rims (up to 250 microm thicksee Fig 5f) that surround former carbonate grainsThese thick Ca-silicate shells would hardly devel-op just by diffusion in the short time-scale of theexperiment (3 hours at the peak T) In fact at800degC (conditions where melting would be verylimited) Ca (and Mg)-silicate rims formed aroundcalcite or dolomite are 2-5 microm thick At 1100degCwhen melting is extensive the reaction rate is sohigh that rims are two orders of magnitude thick-er Therefore mass transport (ie viscous flow)seems to have played a key role in high-T Ca-(Mg-) silicate formation Mass transport is also consis-tent with observation of some areas with differentcomposition (unreacted silicates) trapped betweenfingers resulting in the island geometry observedin Fig 5g It could be expected that without fin-gering the process of Ca- or (Ca-Mg)-silicate for-mation would be self-limiting ie once a rim ofthe newly formed silicate appears diffusion wouldbe jeopardized This shell would preclude or limitthe transport of Ca (andor Mg) from the carbon-ate or it would limit the transport of the lessmobile Si (and Al) from the surrounding silicatetowards the reaction front Therefore it is neces-sary that one of the above described transport-reaction mechanisms takes place This may alsoexplain why phase formation takes place at lowerT than reported previously The presence of multi-ple interfaces among different crystals mayenhance finger development resulting in kineticchanges favoring phase formation at lower T

5 Conclusions

It is observed that even though the overallcomposition of a ceramic material does not under-go major changes upon firing significant miner-alogical and textural modifications do occur

632

These changes are connected with two reactiontypes a) solid-state replacement of a phase by anew one with little or no compositional variation(ie the muscovite and illite case) and minimizingenergy changes by preserving the original textureand orientation of the replaced phase and b) reac-tion-diffusion at the interface between two miner-al grains with contrasting composition (iecarbonate-silicate interfaces) The latter reaction isenhanced if massviscous transport resulting frompartial melting and finger structure developmentoccurs Non-carbonated clay-rich materials (G)show phyllosilicate destruction at T ranging from700 to 900degC followed by vitrification which issignificant at T gt 1000degC Carbonate-rich materi-als (V) show major phase transformation uponphyllosillicates destruction including new Ca (orMg) silicates and aluminum-silicates formation Inthis later case the vitrification starts at lower Tand it is limited at higher T Therefore it is possi-ble to establish the firing T of ceramic materialswith composition broadly similar to that in ourwork based on the mineralogy Thus a) presenceof phyllosilicates indicates a low 700 to 900degC fir-ing T for the dehydroxylated illite (or muscovite)phase b) dolomite disappears at sup3 700degC whilecalcite is still present at 800degC c) gehleniteappears at 800degC reducing its concentration above1000degC d) diopside and wollastonite appear at800degC (in very low concentrations) as reactionrims between carbonates and silicates Formationof these minerals includes ldquofingeringrdquo develop-ment e) mullite appears at T gt 800degC while littleamounts of sanidine form as microcline disap-pears and f) anorthite develops at T gt1000degC incarbonate-rich bricks

When considering equilibrium thermodynamicdata and calculated reaction T for most of theabove mentioned phases and reactions (Holland ampPowell 1985 and references herein) it isobserved that rapid heating induces significantoverstepping Therefore phases formed are inmost cases metastable a fact consistent with anal-yses of pyrometamorphic assemblages (Brearleyamp Rubie 1990 and references herein)

The formation of non-equilibrium fractal mor-phologies at the carbonate-silicate reaction inter-faces may explain why high-T silicates incarbonate-rich ceramics develop very fast Thisgrain-to-grain reaction model may help explainhigh-T silicate formation in nature (eg contactmetamorphism pyrometamorphism in xenoliths)

The gained knowledge on mineral transforma-tion and the dynamics of interface mass and diffu-sion transport facilitated through ldquofingeringrdquo or

ldquoscallopingrdquo opens new ways to understand thecomplexities of ceramic processing in particularand high-T mineral transformation in general bothin natural (ie geologic settings) and artificialcontexts (eg archaeometric analysis of ancientceramics or the design of new bricks ndashreplicas ofancient onesndash for architectural heritage conserva-tion) However care must be taken when extrapo-lating these results to other raw materialcompositions andor ceramic processing methods

Acknowledgements This work has been finan-cially supported by the CGICYT projectMAT2000-1457 and the Research Group RNM0179 of the Junta de Andaluciacutea (Spain) We thankthe Centro de Instrumentacioacuten Cientifica (CIC) ofthe Granada University for allowing us to use theirSEM-EDX EMPA and XRF facilities We thankK Elert JM Garciacutea-Ruiz M Rodriguez-Gallego and A Sanchez-Navas for fruitful discus-sions and comments We also thank the in-depthreview by Dr F Goumltz-Neunhoeffer and an anony-mous referee

References

Barahona E Huertas F Pozzuoli A Linares J(1985) Firing properties of ceramic clays fromGranada province Spain Miner Petrogr Acta 29A577-590

Boynton RS (1980) Chemistry and technology oflime and limestone 2nd Ed Wiley New York

Brearley AJ (1986) An electron optical study of mus-covite breakdown in pelitic xenoliths duringpyrometamorphism Miner Mag 357 385-397

Brearley AJ amp Rubie DC (1990) Effects of H2O onthe disequilibrium breakdown of muscovite +quartz J Petrol 31 925-956

Butterworth B amp Honeyborne DB (1952) Bricksand clays of the hasting beds Trans Brit CeramSoc 51 211-259

Dondi M Ercolani G Fabbri B Marsigli M (1998)An approach to the chemistry of pyroxenes formedduring the firing of Ca-rich silicate ceramics ClayMinerals 33 443-452

Duminuco P Riccardi MP Messiga B Setti M(1996) Modificazioni tessiturali e mineralogichecome indicatori della dinamica del processo di cot-tura di manufatti ceramici Ceramurgia 26 5 281-288

Evans JL amp White J (1958) Further studies of thethermal decomposition of clays Trans Brit CeramSoc 57 6 298

Everhart JO (1957) Use of auxiliary fluxes to improvestructural clay bodies Bull Am Ceram Soc 36268-271

Carbonate and silicate phase reactions during ceramic firing 633

These changes are connected with two reactiontypes a) solid-state replacement of a phase by anew one with little or no compositional variation(ie the muscovite and illite case) and minimizingenergy changes by preserving the original textureand orientation of the replaced phase and b) reac-tion-diffusion at the interface between two miner-al grains with contrasting composition (iecarbonate-silicate interfaces) The latter reaction isenhanced if massviscous transport resulting frompartial melting and finger structure developmentoccurs Non-carbonated clay-rich materials (G)show phyllosilicate destruction at T ranging from700 to 900degC followed by vitrification which issignificant at T gt 1000degC Carbonate-rich materi-als (V) show major phase transformation uponphyllosillicates destruction including new Ca (orMg) silicates and aluminum-silicates formation Inthis later case the vitrification starts at lower Tand it is limited at higher T Therefore it is possi-ble to establish the firing T of ceramic materialswith composition broadly similar to that in ourwork based on the mineralogy Thus a) presenceof phyllosilicates indicates a low 700 to 900degC fir-ing T for the dehydroxylated illite (or muscovite)phase b) dolomite disappears at sup3 700degC whilecalcite is still present at 800degC c) gehleniteappears at 800degC reducing its concentration above1000degC d) diopside and wollastonite appear at800degC (in very low concentrations) as reactionrims between carbonates and silicates Formationof these minerals includes ldquofingeringrdquo develop-ment e) mullite appears at T gt 800degC while littleamounts of sanidine form as microcline disap-pears and f) anorthite develops at T gt1000degC incarbonate-rich bricks

When considering equilibrium thermodynamicdata and calculated reaction T for most of theabove mentioned phases and reactions (Holland ampPowell 1985 and references herein) it isobserved that rapid heating induces significantoverstepping Therefore phases formed are inmost cases metastable a fact consistent with anal-yses of pyrometamorphic assemblages (Brearleyamp Rubie 1990 and references herein)

The formation of non-equilibrium fractal mor-phologies at the carbonate-silicate reaction inter-faces may explain why high-T silicates incarbonate-rich ceramics develop very fast Thisgrain-to-grain reaction model may help explainhigh-T silicate formation in nature (eg contactmetamorphism pyrometamorphism in xenoliths)

The gained knowledge on mineral transforma-tion and the dynamics of interface mass and diffu-sion transport facilitated through ldquofingeringrdquo or

ldquoscallopingrdquo opens new ways to understand thecomplexities of ceramic processing in particularand high-T mineral transformation in general bothin natural (ie geologic settings) and artificialcontexts (eg archaeometric analysis of ancientceramics or the design of new bricks ndashreplicas ofancient onesndash for architectural heritage conserva-tion) However care must be taken when extrapo-lating these results to other raw materialcompositions andor ceramic processing methods

Acknowledgements This work has been finan-cially supported by the CGICYT projectMAT2000-1457 and the Research Group RNM0179 of the Junta de Andaluciacutea (Spain) We thankthe Centro de Instrumentacioacuten Cientifica (CIC) ofthe Granada University for allowing us to use theirSEM-EDX EMPA and XRF facilities We thankK Elert JM Garciacutea-Ruiz M Rodriguez-Gallego and A Sanchez-Navas for fruitful discus-sions and comments We also thank the in-depthreview by Dr F Goumltz-Neunhoeffer and an anony-mous referee

References

Barahona E Huertas F Pozzuoli A Linares J(1985) Firing properties of ceramic clays fromGranada province Spain Miner Petrogr Acta 29A577-590

Boynton RS (1980) Chemistry and technology oflime and limestone 2nd Ed Wiley New York

Brearley AJ (1986) An electron optical study of mus-covite breakdown in pelitic xenoliths duringpyrometamorphism Miner Mag 357 385-397

Brearley AJ amp Rubie DC (1990) Effects of H2O onthe disequilibrium breakdown of muscovite +quartz J Petrol 31 925-956

Butterworth B amp Honeyborne DB (1952) Bricksand clays of the hasting beds Trans Brit CeramSoc 51 211-259

Dondi M Ercolani G Fabbri B Marsigli M (1998)An approach to the chemistry of pyroxenes formedduring the firing of Ca-rich silicate ceramics ClayMinerals 33 443-452

Duminuco P Riccardi MP Messiga B Setti M(1996) Modificazioni tessiturali e mineralogichecome indicatori della dinamica del processo di cot-tura di manufatti ceramici Ceramurgia 26 5 281-288

Evans JL amp White J (1958) Further studies of thethermal decomposition of clays Trans Brit CeramSoc 57 6 298

Everhart JO (1957) Use of auxiliary fluxes to improvestructural clay bodies Bull Am Ceram Soc 36268-271

Carbonate and silicate phase reactions during ceramic firing 633

G Cultrone C Rodriguez-Navarro E Sebastian O Cazalla MJ De la Torre

Fabbri B amp Fiori C (1985) Clays and complementaryraw materials for stoneware tiles Miner PetrogrActa 29-A 535-545

Freestone IC amp Middleton AP (1987) Mineralogicalapplications of the analytical SEM in archaeologyMiner Mag 51 21-31

Garavelli CL Liviano R Vurro F Zinco M (1990)Idromagnesite nei materiali di rivestimento dellachiesa ipogea di S Maria della Grazia (LaterzaPuglia) in ldquoSuperfici dellacuteArchitetture LaFiniturerdquo Graffo Bressansone 189-197

Garcia-Ruiz JM (1992) ldquoPeacockrdquo viscous fingersNature 356 113

Goldsmith JR (1953) A ldquosimplexity principlerdquo and itsrelation to ldquoeaserdquo of crystallization Bull Geol SocAm 64 439-451

Govindaraju K (1989) 1989 compilation of workingvalues and sample description for 272 geostandardsGeostandards Newsletter Special issue 13 1-113

Guggenheim S Chang Y van Groos AFK (1987)Muscovite dehydroxylation high-temperature stud-ies Am Mineral 72 537-550

Holland TJB amp Powell R (1985) An internally con-sistent thermodynamic dataset with uncertaintiesand correlations 2 Data and results J metamor-phic Geol 3 343-370

Kretz R (1983) Symbols for rock-forming mineralsAm Mineral 68 277-279

Maggetti M (1982) Phase analysis and its significancefor technology and origin in ldquoArchaeologicalCeramicsrdquo Olin JS ed Smithsonian InstitutionPress Boston 121-133

Moorehead DR (1985) Cementation by the carbona-tion of hydrated lime Cement Concr Res 16 700-708

Nittmamm J Daccord G Stanley HE (1985)Fractal growth of viscous fingers quantitative char-acterization of a fluid instability phenomenonNature 314 141-144

Nuacutentildeez R Delgado A Delgado R (1992) The sinter-ing of calcareous illitic ceramics Application inarchaeological research in ldquoElectron MicroscopyEUREM 92rdquo A Galindo ed University ofGranada Granada 795-796

Ortoleva P Chadam J Merino E Sen A (1987)Geochemical self-organization II The reactive-infil-tration instability Am J Sci 287 1008-1040

Peters T amp Iberg R (1978) Mineralogical changesduring firing of calcium-rich brick clays CeramBull 57 5 503-509

Preston BJ Dempster TJ Bell BR Rogers G(1999) The petrology of mullite-bearing peralumi-nous xenoliths Implications for contamination pro-cesses in basaltic magmas J Petrol 40 349-473

Rehak P Kunath-Fandrel G Losso P Hildmann BSchneider H Jaumlger C (1998) Study of the Alcoordination in mullites with varying AlSi ratio by27Al NMR spectroscopy and X-ray diffraction AmMineral 83 1266-1276

Riccardi MP Messiga B Duminuco P (1999) Anapproach to the dynamics of clay firing Appl ClaySci 15 393-409

Rye OS (1976) Keeping your temper under controlMaterials and Manufacture of Papuan PotteryArcheology and Phys Anthropology in Oceania 1112 106-137

Sanchez-Navas A (1999) Sequential kinetics of a mus-covite-out reaction A natural example AmMineral 84 1270-1286

Sanchez-Navas A amp Galindo-Zaldivar J (1993)Alteration and deformation microstructures inbiotite from plagioclase-rich dykes (Ronda MassifS Spain) Eur J Mineral 5 245-256

Scott VD amp Love G (1983) Quantitative Electron-Prove Microanalysis John Wiley amp Sons NewYork

Segnit ER amp Anderson CA (1972) Scanning elec-tron microscopy of fired illite Trans Brit CeramSoc 71 85-88

Shoval S (1988) Mineralogical changes upon heatingcalcitic and dolomitic marl rocks ThermochimActa 135 243-252

Tite MS amp Maniatis Y (1975) Examination ofancient pottery using the scanning electron micro-scope Nature 257 122-123

Veniale F (1990) Modern techniques of analysisapplied to ancient ceramics in ldquoAdvancedWorkshop on Analytical Methodologies for theInvestigation of Damaged Stonesrdquo F Veniale and UZezza eds University of Pavia Pavia 1-45

Warren BE (1930) The structure of melilite(CaNa)2(MgAl)1(SiAl)2O7 Zeits Kristall 74131-138

Webb TL (1952) Chemical aspects of the unsoundnessand plasticity in building limes The South AfricanIndustrial Chemist 290-294

Worden RH Champness PE Droop GTR (1987)Transmission electron microscopy of pyrometamor-phic breakdown of phengite and chlorite MineralMag 51 107-121

Received 20 February 2000Modified version received 12 July 2000Accepted 8 January 2001

634