The Crystallization of Anatase and Rutile from Amorphous Titanium

AmericanMineralogist, Volume61, pages419424, 1976

The crystallizationof anataseand rutile from amorphoustitanium dioxideunder hydrothermal conditions

Ar-eNMerrsEws Departmentof Geology,The Hebrew Uniuersity Jerusalem.Israel

Abstract

The hydrothermal crystallizationof amorphous titanium dioxide at 300-600'C, I kbar pressure,follows the reaction path: amorphous reactant - anatase- rutile. The transformation is acceleratedby increasingtemperature and decreasingsolution pH. Two mechanism stagescan be envisaged:(1) the formation of anatase by the dehydration and gradual structural rearrangementof the reactant titanium-oxygen lattice, and (2) solution of the anataseintermediate and direct precipitationof rutile. The metastabledevelopment of anatase is in accordancewith a modern formulation of the Ostwald Step Rule.

Introduction ride until the pH of the solution was in the range l-4. The kinetics and mechanismsof the solid state The resulting dense white precipitate was washed transformation anatase+ rutile have beenwell stud- thoroughly and dried under infrared heat. X-ray dif- ied (Czandernaet al., 1958;Rao, l96l; Shannonand fraction indicated the material to be completely Pask, 1965;Heald and Weiss,1972). However, less amorphous. Experimentswere performed at I kbar attention has been paid to the relationshipsexisting pressurein cold sealhydrothermal bombs, with reac- betweenthe two polymorphs in hydrothermal aque- tants sealedinside silver capsules.Charges consisted ous conditions.Osborn (1953)found that high tem- of 50 mg solid and 500-600 mg solution. Products perature, and to a lesser extent pressure increase, were examinedby X-ray diffraction, and where pos- favor the conversionof anataseinto rutile. Dachille sible (at 600"C only) anatase-rutileproportions were et al. (1968) hydrothermally crystallized titanium estimatedby the method of Spurr and Myers (1957), dioxide gel, obtaining anataseat lower temperatures Solutions used were sodium carbonate, water, and and rutile at higher temperatures. Mineralizing hydrochloric acid. Tests with pHydrion papers agents, such as alkali metal fluorides, catalyze the showedthat the solutionsunderwent a slight increase formation of rutile from gel or anatase (Osborn, in alkalinity during reaction.This presumablyarises 1953;Anikinand Rumyantseva,1964). Jamieson and from the releaseduring crystallization of ammonra Olinger (1969)showed that anataseis thermodynam- trapped in the amorphous reactant during its prepa- ically always unstablewith respectto rutile. Thus the ration. No evidencecould be seen of reaction be- persistenceor formation of anatase under hydro- tween the silver capsulesand the acid solutions.The thermal conditions should be regarded as a con- solution conditionsgiven in Table I detail both initial sequenceof sluggishreaction kinetics or metastable concentrationsand final pH values. crystallization.Correspondingly, the formation of rutile from gels or anataseis catalyzedby increasing Experimental results temperature, pressure,and the presenceof miner- The resultsof crystallizationexperiments are given alizing agents.The purpose of this paper is to in- in Table l. It can be seenthat neutral, alkaline, and vestigatethe conditions and mechanismsof anatase mildly acidic conditions favor anatase formation, and rutile formation during the hydrothermal crystal- whereasmore strongly acid environmentsfavor ru- lizationof amorphoustitanium dioxide. tile. This is clearly indicated by the results of long experimentsperformed at 500'C. In sodium carbon- Experimental methods ate, water, and 0.01 M HCl, anatase is formed, Amorphous titanium dioxide was preparedby the whereasin 0.1 M and 0.5M HCI the productis rutile. addition of 0.88 sp gr ammonia to titanium tetrachlo- Increasing temperature also enhancesrutile forma-

419 420 ALAN MATTHEWS

Tesr-s l. Resultsof crystallizationexperiments ments.These observations are consistentwith the transformation: Solutions Toc** t. h"r* products Initial Final pH amorphoustitanium dioxide + anatase+ rutile The datagiven in TableI indicatethat therate of the 1.0M.HCI 0.0 304 68 Rutile transformationis acceleratedby increasingtemper- 406 0.4 Rutil e atureand decreasingsolution pH. 0.5 M.HCt 0.4 500 0.4 Rutile,smalIanatase The proposedreaction mechanism is analogousto 502 204 Ruti I e that observedfor the hydrothermalcrystallization of 0.I M.HC] 305 0.3 Anatasedevelopinq 295 1.5 Anatasedevelobin; amorphoussilica, in whichthe polymorphs cristobal- av2 3 Rut'ile,pooranitaie ite and occasionallysilica K precedethe formation z9a 14 Rutile,smallanatase? of 304 68 Rutil e the thermodynamicallystable phase, quartz. It is 404 0.3 Pooranatase,small ruti'le found that the rate of quartzformation is accelerated 398 72 RutiI e by increasingtemperature and pressure(Carr 500 0.4 Ruti1e,sma11anatase and 501 276 Ruti l e Fyfe,1958), but in contrastto theabove observations 597 264 Ruti l e the reactionis catalyzedby alkalinesolution species 0.01M.HCI 2.8 404 168 Anatase (Campbelland Fyfe, 1960;Fyfe and McKay, 1962; s05 235 Anatase Matthews, 1972).Campbell and Fyfe relatedthe al- 597 264 Anatase(75%),rutile(25%)kali catalysisto the natureof the silicatesolution Water 8.8 302 0.2 Anatasedevelopinq species,and presumablythe differencesnoted reflect 295 3 Anatasedevelobini 304 68 Anatase 410 0.3 AnatasedeveloDinq 401 188 Anatase A s02 0.4 Anatase 1o 501 216 Anatase 597 264 Anatase(90%),ruti'le(10%)

0.1 M.Na2C03 500 0.4 Anatase 505 235 Anatase

* Timeat temperature,ie. after completionof run up **Temperature t5oc

tion; crystallization of the amorphous reactant in water and 0.01 M HCI at 400.C and 500.C yields anataseonly, but at 600oCrutile appearsin the products. A third naturally occurring polymorph, brookite, is not found in any experimentalproducts. 1c R The mechanismof reaction is illustrated by X-ray R diffraction patterns obtained from the products of reactionsat 300'C using 0.1 M HCI (Fig. l). After l7z hours reaction anataseonly is developing (Fig. la), but after 3 hours reaction the poorly developed anatase has diminished and",rutile is forming (Fig. lb). Fourteenhours reactionyields rutile, with per- haps a small amount of anataseremaining (Fig. lc), and 68 hours reaction gives rutile only. Similarly, 10 35 32 28 21 other short experiments at 400'C and 500.C using 20 C" (Ka) 0.1 M and 0.5 M HCI indicate that anataseprecedes Ftc. I X-ray diffractionpatterns of products the formation of rutile. Reactionsin water at 300.C, from crystallizationexperiments at 300oCusing 0.1 M HCl. (a) 1.5hours reaction. 400oC,and 500oC,in which anatase is the only prod- (b) 3 hoursreaction. (c) l4 hoursreaction. Legend: A=Anatase, uct, also show anatase developing in short experi- R:Rutile CRYSTALLIZATION OF ANATASE AND RUTILE FROM AMORPHOUS TITANIUM DIOXIDE

that titanium-oxygen speciesare acid- rather than alkali- stabilized. The developmentof phasesas indicated by X-ray diffraction revealsa dissimilarityin the specificmech- anismsof anataseand rutile formation. The development of anataseis first indicatedby the appearanceof 'hump' a broad in the 20 region of the anatase l0l peak. For reactionsin water (and presumablyalso 0. I M Na,CO, and 0.01 M HCI) this hump gradually developsand sharpens(see Fig. la) untilfully crystalline anataseis indicated.For reactionsin 0.1 M and 0.5 M HCl, rutile growth commencesbefore full crystalline development of the anatasecan occur (Fig. lb). However, the rutile gives well defined X-ray peaks,even when presentin small amounts;i.e. in the earlystages of development(Figs. I b and lc). Similar observationscan be made for the silicic acid crystallization (cf. Carr and Fyfe, 1958,Fig. l). The appear- ance of well-orderedrutile is consistentwith its for- matlon occurring through a solution and Ftc. 3. SEM micrograph showing gross morphology of an precipitation mechanism.However, anataseforma- anatasegrain. Product from 235 hours reaction at 505'C using 0.01 M HCl. tlon may occur by the gradual structural rearrangement of the titanium-oxygenlattice of the amorphous vestigatethese mechanism proposals. Representative reactant, with accompanying dehydration. Dehy- micrographsare given in Figs.2-6, and the following dration may precede the structural rearrangement. observationscan be made: The isotopic studiesof Clayton et al. (1972)on silicic (l) The gross morphologiesof the reactantgrains acid crystallizationshowed that dehydration occurs are preservedin the product anatase(Figs. 2 and 3). before any loss of amorphous character can be de- (2) The large,fully crystalline,anatase grains con- tected in the reactant. sist of a mass of reasonablyequidimensional small Scanning electron microscopy was used to in- crystals(Fig. a). (3) The grossmorphologies of rutile grains do not possessthe similaritiesto the reactantthat are shown by anatasegrains (Figs. 2,3, and 5). (a) The rutile grains consist of a mass of small crystals.However, thesecrystals are much more in- tergrown than anatasecrystals (Figs. 4 and 6). Observations(l) and (2) are consistentwith the idea that anataseforms by the dehydration and gradual structural rearrangementof the reactant lattice. Small volumes of reactant appear to be forming into discrete fine anatasecrystals, with the gross morphology of the reactant being preservedin the product. The loss of water from the reactantduring dehy- 'space' dration potentially provides in which the lattice rearrangementscan occur. The role of the aqueous solution may be to catalyze the rearrangement by aiding the disproportionation of Ti-O bonds; or alternatively small scale dissolu- tion/precipitation processesare possible. The observationsfor rutile are seeminglyindicative Flc. 2. SEM micrograph showing gross morphology of the greaterrole that solution processesare playing reactant graln in its formation. They are consistentwith the pro- 422 ALAN MATTHEWS

Frc. 4 SEM micrograph of anatasecrystals. Detail from Fig. 3. FIc. 6. SEM micrograph of rutile crystals.Detail from Fig. 5 posal that rutile forms by solution of precursorana- the developinganatase and the amorphous reactant. taseand subsequentnucleation and crystalgrowth of This connection between amorphous reactant and rutile from the solution. product was noted by Carr and Fyfe (1958)for the amorphous silica crystallization(in this casecristo- Discussion balite shows order related to that of the amorphous The SEM micrographsgiven in Figs. 2-4 empha- reactant). Carr and Fyfe proposed that the overall sizethat a closestructural relationship existsbetween reaction mechanismis governedby a seriesof epitax- ial controls, in which products only form when suf- ficient surfaces of a structurally related precursor phaseare availablefor nucleation.In the crystallization of amorphoustitanium dioxidethis would mean that anatasenucleates first becauseit is structurally closer to the reactant, and rutile only forms later, when a sufficientsurface of anataseis available for nucleation to commence. It is also possibleto view the reaction sequencein an alternativemanner, in terms of the Ostwald Step Rule. One interpretationof this rule is that there is a tendency for reactionsto occur in which there is a minimum A,S,rather than reactionswhich yield the thermodynamicallystable phase (Turner, 1968,p.77). In the crystallizationreactions discussed there is an overall decreasein entropy, and the condition above would be satisfiedif the first-formed phasepossesses a higher entropy than subsequentphases. Such is the casefor cristobaliteand quartz, but not for anatase and rutile (Fig. 7). Anatase has a slightly lower entropy than rutile (-0.1 cal/deg/mole in the range 300-600"C.). The error limits for the Ftc 5. SEM micrograph showing gross morphology of a anataseand rutile grain. Product ftom 72 hours reactionsat 398"C using 0.1 rutile entropiesoverlap, and so it is possiblethat the M HCl. Scale Bar should read l0rr. entropy relationshipsgiven in Figure 7 can be re- CRYSTALLIZATION OF ANATASE AND RL]TILE FROM AMORPHOUS TITANIUM DIOXIDE 423

cristobalite and anatase have considerably higher molar volumes than their respective polymorphs. Thus, this alternative formulation of the Ostwald StepRule providesa satisfactoryinterpretation of the observedproduct sequences. Anatase and rutile occur in a variety of natural hydrothermal environments;for example,in veins in granitic pegmatites.However, anataseformation is a metastablephenomenon, and it may be queried why U) anatasedoes not transform into rutile. It is possible E that the anataseis removed from contact with the oz4 U hydrothermal solution immediately after formation. a would then haveto J The anatase-rutiletransformation o take place by solid state reactions,and these have been shown to be extremelyslow below 600'C (Rao, d_ 22 o 196l). Alternatively, the metastable persistenceof t and pH condi- z anatasecould reflectthe temperature U tions of hydrothermal deposition. Low temperature/high solution pH could result in the anatase- rutile transformation being inhibited. Temperature

c CRIST0BALITE 300 100 500 600 OC TEMPERAIURE FIc. 7. Molar entropies at I kbar pressure.The curves were calculated from entropy data at I atm, using the relation Sr., xr., = Sr i aVLP, in which a = thermal expansioncoefficient, P : pressure in bars, and V = molar volume in cal/bar. Data sources:en- tropiesand molar volumes,Robie and Waldbaum (1968);thermal expansion coefficients for rutile, quartz, and cristobalite, Skinner (1966), and for anatase,Clarke (1928). 21

versed.Nevertheless, entropy criteria do not provide IU a satisfactoryinterpretation for anataseformation. 2 3zz An alternative interpretation of the Ostwald Step c) Rule has that the first phaseto form will be the one possessingthe least surface energy with respectto the G ANATASE reactant (Fyfe et al., 1958,p. 73). Surfaceenergy is inverselyrelated to molar volume (Fyfe e/ al., 1958, lzo p. 73), and thus the Step Rule may be alternatively formulated to state that a reaction is preferred which givesthe product with the highestmolar volume with respectto the reactant. In Figure 8, molar volumes have been plotted for the various phasesunder discussion.The molar volumes at I atm pressurehave 300 400 500 600 OC beenplotted, sinceinsufficient compressibility data is TEMPERATURE available for them to be computed at I kbar. How- FIc. 8. Molar volumes at I atm pressure.The curves were ever,the pressurecorrections to I kbar are very small calculated from molar volume data at 25"C (Robie and Wald- (-0.001 cc. for rutile) and will not affectthe relation- baum, 1968), and thermal expansion data taken from the sources ships depicted in Figure 8. It can be seenthat both referred to in Fig. 7. 424 ALAN MATTHEWS

determination of TiOr-bearing assemblagesis pos- Fyns, W. S. eNo D. S. McKny (1962)Hydroxyl ion catalysisof the sible by oxygen isotope geothermometry(Addy and crystallizationof amorphous silica at 330"C and some observa- Garlick, 1974).Thus, accuratecalibration of the so- tions on the hydrolysis of albite solutions.Am. Mineral. 47, lution pH 83-89 and temperatureconditions under which -, F. J TURNERnNo J Vrnuooceu (1958) Metamorphic the anatase-rutiletransformation takes place may reactions and metamorphic facies. Geol Soc Am Mem 13, provide a valuable high temperaturepH indicator. 259 p. Hrnlo, E. F. nNn C. W. Wsrss (1972)Kinetics and mechanismof Acknowledgments the anatase/rutile transformation, as calalyzed by ferric oxide and reducingconditions. Am. Mineral 57,10-23. Dr. B. P. Cohn is thanked for criticallyreviewing the manuscript. Jnlrlrsou, J. C. nNo B. Or-rNcpn(1969) Pressure-temperature studies of anatase,brookite, rutile, and TiOdll): A discussion.,4rt References Mineral. 54, 1477-1481. Aooy,S. K. ero G. D. Glnlrcr (1974)Oxygen isotope fraction- Mnttnnws, A. (1972) The crystallizationof quartz from silicic acid: a kinetic investigationusing sodium carbonatesolutions ation between rutile and water. Contrib. Mineral. petol. 45, I t9-t2l . Progress in Experimental Petrology, Vol. 2, p 62-65, N.E.R.C., London. ANlKlN, I. N. eNo G. V. RurvyeNrslvn (1964) Solubility and OsnonN, E. F. (1953)Subsolidus relations in oxide systemsin the recrystallization of TiO, under hydrothermal conditions Iss/ed. presence of water at high pressures.J. Am. Ceram. Soc 36, Prir. Tekh. Mineraloobrazou. Mater. Soueshch. 7th. Luou. o. 147-15t. 277-279 (in Russian). ffiansl. Chem. Abst. 66-324951. Rlo, C. N. R. (1961)Kinetics and thermodynamicsof the crystal CnurnrLl, A. S. eNo W. S. FvFE (1960)Hydroxyl ion catalysisof structure transformation of spectroscopically pure anatase to the hydrothermalcrystallization of amorphoussilica. A possible rulile Can. J Chem.39, 498-500. high temperature pH indicator Am. Mineral. 45, 465-46g. Rosts, R. A. nNn D. R. Wer-oseuNt(1968) Thermodynamic prop- Cnnn, R. M. lNo W. S. Fvrr (1958) Some observationson the ertiesof mineralsand relatedsubstances at 298 l5'K (25.0"C) crystallization of amorphous silica. Am. Mineral. 43, 90g-916. and one atmosphere(1.013 bars) pressureand at highertemper- Clenxs, J. R. (1928) Density and thermal expansionof chemical atures. U.S. Geol. Suru. Bull 1259,256 p. compounds in the crystalline state, under atmospheric Dressure. SHnNNoN,R. D. eNn J. A. Pnsr (1965) Kinetics of the in Internationat Critical Tables of Numerical Data, physics, anarase- rutife transformation.,/. Am. Chem Soc. 48, 391-398. Chemistry and Technology, Vol. IIl. McGraw-Hill, New york. p. 43-45 SrtrNrn, B. J. (1966)Thermalexpansion. In, S P. Clarke,Jr., Ed., Handbook of Physical Constants, Geol. Soc. Am. Mem. 97. ClevroN, R. N., J. R. O'Nrn eNo T. K. Mryroe (1972)Oxygen isotope 7 5-96 exchangebetween quartz and water. J. Geophys. Res.77, 3057-3067 Spunn, R. A. eNo H. Mvrns (1957)Quantitative analysisof anatase-rutile mixtures with an X-ray diffractometer. Anal. Chem. CzeNornNe, A. W., C. N. R. Reo eNo J. M. HoNrc (195g)The 29.769-762. anatase-rutile transition. l. Kinetics of the transformation of TURNER, F J, (1968), Metamorphic petrology-mineralogical pure anatase. Trans. Farad. Soc 54,1069-1073. and feld aspects McGraw-Hill, New York, 403 p. DrcHrr-Lt, F., P. Y. SrlroNs eNo R. Rov (196g) pressure_remper_ ature studies of anatase, brookite, rutile, and TiO" (ll). Am. Manuscript receiued, July 7,1975; accepted Mineral. 53. 1929-1939. for publication, January 19, 1976.