American Mineralogist, Volume 77, pages 545-557, 1992

Conversion of perovskite to anatase and TiO2 (B): A TEM study and the use offundamental building blocks for understanding relationships among the TiO, minerals

Jrr.r.r.lN F. BlNrrnr,Drt Dlvro R. VBnr,rNDepartment of Earth and Planetary Sciences, The Johns Hopkins University, Baltimore, Maryland 21218, U.S.A.

Ansrucr

TiO, crystallizes in at least seven modifications encompassing several major structuretypes including those of rutile, hollandite, and PbOr. We have adopted a fundamentalbuilding block (FBB) approach in order to understand better the relationships betweenthese structures. The identification of common structural elements provides insights intopossible phase transformation mechanisms as well as the weathering reactions that convertperovskite (CaTiOr) to TiO, anatase and TiO, (B).

We have studied the weathering of perovskite from the Salitre II carbonatite, MinasGerais, BraziI, by transmission electron microscopy. Although the dissolution of perov-skite, a proposed component of the high-level nuclear waste form Synroc, has been mod-eled thermodynamically and studied experimentally, its direct natural replacement by Tioxides during weathering has not been previously described. Our results show that theperovskite is replaced topotactically by both anatase and TiO, (B). The reaction appearsto involve an intermediate product that can be interpreted as a half-decalcified, collapsedderivative of perovskite. The second step, involving removal of the remaining Ca, canoccur in one of two ways, leading either to anatase or in rare cases to TiO, (B). Theorientations of the interfaces between both perovskite and anatase and perovskite andTiO, (B) are consistent with collapse of the corner-linked Ti-O framework to form edge-sharing chains. Despite the orthorhombic structure of perovskite, the results show that theanatase can develop with its c axis parallel to any of its three pseudocubic axes. Thecomplete destruction of perovskite produces arrays primarily of anatase that display avariety of textures, including porous aggregates and recrystallized needles. REE elementsreleased from perovskite are immobilized by precipitation of rhabdophane that is inti-mately intergrown with anatase.

INrnonucrroN

There are currently four naturally occurring TiO, poly-morphs [rutile, anatase, brookite, and TiO, (B)] and atleast three (plus some very high-pressure forms) poly-morphs that have been produced synthetically. Detailsfor six of the distinctive polymorph structures are listedin Table I . It is generally considered that at low pressuresonly rutile has a true field of stability; anatase and brook-ite form metastably (Dachille et al., 1968; Post and Burn-ham, 1986). Post and Burnham (1986) used ionic mod-eling to show that the electrostatic energies of rutile,anatase, and brookite are not greatly different. Navrotskyand Kleppa (1967) measured AH : -5.27 kJlmol for theanatase to rutile reaction.

The structure of the most commonly occurring and bestknown TiO, phase, rutile, is shared by a number of min-erals including pyrolusite, cassiterite, stishovite, and in adistorted form, marcasite. The CaCl, polymorph of TiO,

* Present address: Department of Geology and Geophysics,University of Wisconsin-Madison, 1215 West Dayton Street,Madison, Wisconsin 53706, U.S.A.

ooo3404x/ 9210506-0545$02.00

is related to the rutile structure by a second-order phasetransformation (discussed by Hyde, 1987). The high-pressure polymorph, TiO, II, has the a-PbO, structure;the highly metastable TiO, (H) polymorph has the hol-landite structure; and TiO, (B) has a structure that isclosely related to that of VO, (B). The similarities be-tween some of these polymorphs and the related struc-tures of anatase and brookite will be discussed below.

All of the polymorphs contain edge- and corner-linked,octahedrally coordinated Ti cations. The available struc-ture refinements (see asterisks in Table l) show relativelyconstant Ti-O distances and quite variable O-O distanc-es, particularly when the shared vs. unshared octahedraledges are compared. Pairs offace-sharing octahedra (con-taining trivalent Ti) are found only in crystallographicshear structures in reduced rutile derivatives (TiOr_.).

Previous discussions ofthe relationships between cer-tain of these structures have focused on the presence ofsheets of O atoms that approach a cubic closest-packingarrangement (such as in anatase), very distorted hexago-nal closest packing that approaches cubic (as in rutile;Hyde et al.,1974), or an approximately double hexagonal

545

546 BANFIELD AND VEBLEN: PEROVSKITE TO ANATASE AND TiO, (B)

TaBLE 1. Data for some TiO2 polymorphs and perovskite

DensityStructure Space group (g/cm")

Unit-cell data(nm) Reference

Rutile'Anatase*Brookite'Tio, (B)Tio, il'Tio, (H)Perovskite"

P4"lmnmAllamdPbcaC2lmPbcnAlmPcmn

4 .133.793.993.644.333.464.03

a : 0.459, c: 0.296a : 0.379, c: 0.951a : 0.917, b : 0.546, c : 0.514a : 1.2 '16, b : 0.374, c : 0.651, 0 : 1O7.2Va : 0.452. b : 0.550. c : 0.494a : 1.018, c: 0.297a: 0.537. b: 0.764, c: 0.544

Cromer and Herrington (1955)Cromer and Herrington (1955)Baur (1961)Marchand et al. (1980)Simons and Dachille (1967)Latroche et al. (1989)Kay and Bailey (1957)

* Structures refined by X-ray ditfraction.

closest-packing arrangement (as in brookite). These de-scriptions emphasize differences among the structures butfail to highlight their close relationships.

Andersson (1969) showed that rutile could convert tothe TiO, II structure by cation displacement. Simons andDachille (1970) noted that similar structural elements arepresent in anatase and TiO, II and suggested that a trans-formation between cubic closest-packed and hexagonalclosest-packed arrangements involves shear between OIayers.

Bursill and Hyde (1972) and Hyde et al. (1974) de-scribed a relationship between hollandite and rutile struc-tures involving rotation of adjacent rutile blocks byclockwise and counterclockwise rotation. Alternatively,Bursill (1979) demonstrated that the hollandite structurecan be generated from the rutile structure by intersectingantiphase boundaries or crystallographic shear. Latrocheet al. (1989) reported that the TiO, (H) polymorph con-verted to anatase rather than rutile with increasing tem-perature and noted that the structural relationship be-tween TiO, (H) and anatase was not clear.

The dissolution of perovskite and its replacement byTiO, has been studied experimentally under a range ofhydrothermal conditions (e.g., Myhra et al., 1984; Kas-trissios et al., 1986, 1987; Jostsons et al., 1990). The in-terest is partly due to the proposal that perovskite couldbe a major component of the synthetic assemblage oftitanate minerals known as Synroc, a potential nuclearwaste form. Thermodynamic calculations have clearlyshown that perovskite is unstable with respect to bothtitanite and rutile at low temperatures and in natural wa-ters (Nesbitt et al., l98l). Kastrissios et al. (1987) sug-gested that perovskite dissolved under hydrothermal con-ditions (between ll0 and 190 "C) and that euhedralanatase and brookite crystals grew epitaxially from anamorphous Ti-O layer. Jostsons et al. (1990) summarizedmuch of the experimental work on perovskite dissolutionincluding data ofPham et al. (unpublished data), whichsuggest that under epithermal conditions (

BANFIELD AND VEBLEN: PEROVSKITE TO ANATASE AND TiO, (B)

Fig. l. Representations of the fundamental building block.Iflight-stippled octahedra have a coordinate z : O,l (z perpen-dicular to the page), then dark-stippled octahedra are located atz : 0 . 5 .

Because the basic rutile structural elements can be createdby stacking ofthis unit and the rutile structure itselfcanbe generated from this assembly by shear, we have se-lected this four-octahedron unit to be the fundamentalbuilding block. The FBB is illustrated in four orientationsin Figure l.

Although we will not discuss it further, this FBB alsooccurs in many structures other than the TiO, poly-morphs. For example, it is the same as a four-octahedronsegment of the Ml cation chain found in the pyroxenes,and it occurs as a fragment of dioctahedral and triocta-hedral sheets in the layer silicates, oxides, and hydrox-ides.

Anatase, brookite, TiO, (B), and TiO, II and theirrelationships to rutile

There are a number of reasons to begin this discussionwith the anatase structure. Anatase is an important low-temperature alteration product of many Ti-bearing min-erals (biotite, pyroxene, perovskite), and a number ofthermally induced polymorphic transformations involv-ing this mineral have been reported. The anatase-rutilereaction that occurs during low-grade metamorphism wasstudied experimentally by Shannon and Pask (1964) andsubsequently (with conflicting results) by Kang and Bao(1986). The conversion of both TiO, (H) (Latroche et al.,1989) and TiO, (B) (Brohan et al.,1982) polymorphs toanatase has been reported. The structural rationalizationfor the TiO, (B) to anatase transformation is well under-stood (Brohan et al., 1982; Tournoux et al., 1986; Ban-field et al., l99l). However, an explanation for the TiO,(H) to anatase transformation has not been provided.

A representation of a slice of the anatase structure con-structed from the FBB (in the orientation shown on thefar right in Fig. l) is given in Figure 2. Consecutive layersstack directly above the layer shown, linked by corner

Fig.2. FBB representation ofanatase. The (l0l)- and (103)-bounded slabs are indicated. The FBB is shown by the solid darkline, and the octahedra included in the (103)-bounded slab aregrouped by the dashed line.

sharing, so that light-stippled octahedra at y : 0,1 definethe corners of the unit cell as shown. Dashed parallel linesisolate a (101)-bounded slab ofanatase referred to in thediscussion of the brookite structure, and those labeled(103) separate groups of octahedra (outlined) that formslabs common to the TiO, (B) structure.

Figure 3 shows a representation ofa slice through theTiO, (B) structure. The FBB is indicated by the heavysolid line, and the corner-linked FBB slab common to theanatase structure (Fig. 2) is shown by the dashed outline.The common structure parallel to (103) anatase rational-izes that TiO, (B) is converted to anatase by shear in-volving growth ofledges along the (103) planes ofanatase(Banfield et al., l99l).

The FBB representation of the brookite structure isshown in Figure 4a. This diagram illustrates that brookite

Fig. 3. FBB representation of TiO, (B). The FBB is shownby the dark solid line, and the (103)-bounded anatase slab ishighlighted by the dashed line.

548

Fig. 4. (a) FBB representation ofthe brookite structure. Thedark line shows the FBB, and the dashed line highlights the(101)-bounded anatase slab. Arrows indicate the brookite axes(bold) and the anatase directions (italics) in the plane ofthe slabs,(b) the FBB in the second orientation, (c) structural elementrelated to the rutile structure by shear, as indicated by the ar-rows.

can be constructed from structural slabs found in anatase(the {101}-bounded slabs). In brookite, these slabs occurin two alternating orientations, shown in Figure 4a asstippled and unstippled. Figure 4b shows the orientationof the FBB for the unstippled slab. The first anatase slab(stippled) is oriented with an anatase a axis coming outofthe page, the second with an anatase a axis vertical.

Figure 4c shows a strip ofbrookite structure that occursalong the plane where the stippled and unstippled ana-tase-type slabs are joined. This strip can be convertedinto the basic unit of the rutile structure (l l0 rutile slab)

Fig. 5. (a) Assembly of an FBB to form a chain, (b) arrange-ment ofchains to form sheets found in both anatase and brookite

[the (100) plane ofbrookite], (c) an FBB that forms sheets iden-tical to those in b but rotated 180o. The numbers indicate howthe sheet in c may be superimposed onto that in b.

BANFIELD AND VEBLEN: PEROVSKITE TO ANATASE AND TiO, (B)

(a) Brookite

A

BA

B

B

BB

A

ABB

A

(c) Anatase

Fig. 6. The stacking ofslabs shown in Figures 5a and 5c (Aand B) (a) to form brookite, (b) to form TiO, II. The samestacking of slabs containing straight rather than kinked chainsproduces rutile (c) to form anatase.

by shear, as indicated by the arrows. When viewed downc of brookite, shear of this type throughout the structureproduces infinite edge-linked chains (out ofthe page). Al-ternatively, conversion ofthese chains to infinite chainscould involve cation hopping (every second pair alongthe chains in Fig. 5 to form straight, edge-sharing hori-zontal chains) rather than shear. Similar cation displace-ments relating rutile and TiO, II are discussed by Simonsand Dachille (1970) and Hyde and Andersson (1988).

Figure 5 shows the arrangement of chains of FBBs (Fig.5a) that form the (100) layers of brookite, the (101)-bounded slabs ofanatase, and the (100) layers ofTiO, II.The stacking of layers as shown in Figure 5b (A) or iden-tical layers with an orientation as shown Figure 5c (B)generates the structural arrangements illustrated in Fig-ures 6a (brookite: AABBAABB . . .), 6b (TiO, II, and byrearrangement as discussed for Fig. 4c, rutile: ABAB . . .),and 6c (anatase: BBB . . .). Thus, ignoring the polyhedraldistortions, anatase, brookite, and TiO, II can be consid-ered to be simply polytypes of TiO, based on stacking oflayers illustrated in Figure 5b.

A 180'rotation ofoctahedra relates the A- and B-typelayers suggesting, as noted by Simons and Dachille (1970),that transformations between structures involve shear be-tween O planes. A similar mechanism involving shear ofthe O planes parallel to the planes ofoctahedra was pos-tulated for the enstatite to clinoenstatite transformation(Coe and Kirby, 1975). Although such mechanisms in-volving two sets of shear operations or cation hoppingplus shear would convert anatase or brookite to rutile,we currently have no direct evidence to indicate the ac-tual mechanisms by which the transformations occur, andother mechanisms may well operate.

The TiO, (H) structure and its relationship to theother polyrnorphs

An FBB representation of the TiO, (H) (hollandite)structure is given in Figure 7. Unlike the structures de-

BANFIELD AND VEBLEN: PEROVSKITE TO ANATASE AND TiO, (B)

(a) 11l

549

Fig. 7. FBB representation of the basic unit of the TiO, (H)structure. The structure contains two orientations of chains con-structed from the FBB highlighted by the dark lines (a) and in asecond orientation (b).

scribed above, in TiO, (H) the FBBs are stacked to forminfinite chains ofedge-linked octahedral these chains oc-cur in two orientations. The TiO, (H) polymorph seemsstructurally more closely related to rutile than to anatase,and it has been shown that electron-beam damage canresult in the conversion of synthetic hollandite com-pounds to the rutile structure (Bursill, 1979). Conse-quently, it is interesting that Latroche et al. (1989) reportthe conversion of TiO, (H) to anatase with increasingtemperature. A possible mechanism for this reaction in-volves displacement of cations from the chains in oneorientation into adjacent octahedral sites (Fig. 8a), fol-lowed by shear as indicated in Figure 8b. This producesthe TiO, (B) structure, which can in turn be converted toanatase by the previously reported shear mechanism. Ifthis type of structural transformation occurs, we predictthat TiO, (B) develops as an intermediate phase betweenTiO, (H) and anatase.

The structural relationships between perovskite andTiO, structures

It has been previously noted that the TiO, (B) structurecan be described as a derivative of the ReO. structure(Tournoux et al., 1986). Likewise, the perovskite struc-ture is closely related to the ReO, structure (Hyde andO'Keefe, 1973). In perovskite, Ca sits within a corner-linked Ti-O framework that is identical to that found inReO, (i.e., perovskite is a stuffed derivative of ReOr).Consequently, there are close similarities between the pe-rovskite and TiO, (B) structures. The relationships noted

Fig. 8. (a) Diagram illustrating the conversion oftwo chainorientations in TiO, (H) to one by cation hopping: [l] : TiO,(H) structure, [2] : cation displacement; [3] : resulting struc-ture. (b) Arrows indicate the shear needed to transform [3] toTio, (B).

in the preceding section thus imply that anatase, brookite,TiO, il, and in a more complex way, rutile are also re-lated to the perovskite structure.

ExpnnrprnNTAr, METHoDS AND SAMpLES usED

We studied naturally weathered perovskite crystals fromthe Salitre II carbonatites in Minas Gerais, Brazil. Thesamples,385-P4a, 385-143, and 385-179, were collectedby Anthony Mariano. Samples 385-179 and 385-143contain lamellar-twinned perovskite embayed and veinedby anatase, whereas 385-P4a is composed primarily ofanatase, exsolved magnetite, and only minor residual pe-rovskite.

Samples were examined with a Philips 420 ST trans-mission electron microscope (TEM) operated at 120 keV.Samples were prepared by dispersing crushed grains on aholey carbon grid and by Ar-ion milling specimens re-moved from petrographic thin sections. Minerals wereidentified by their selected-area electron diffraction(SAED) patterns and by their compositions. Analyticalelectron microscope (AEM) data were obtained from thethin edges of samples using an EDAX Lidrifted Si de-

BANFIELD AND VEBLEN: PEROVSKITE TO ANATASE AND TiO, G)

oooo

o o _@tt&4 l nI r v D

o oo.P,'.o2o.nOo.r9

nt

tdanatase

otobr ?"n","."0.19 nm

bperovskite

bperovskitea"n"r"a"f

a O2ooatlo

&?.s.3O. O. @^^rlO ro-+canatase

o ct%?oooo

ooA

danatasel

' Goo"n

oro ro

OOOO@/OOOO+canatase01 0p

ooo

o

Fig. 9. SAED patterns (a) down [001] ofanatase (a) and [001] ofperovskite (p; pseudocubic unit cell); (b), (c) down [010] ofanatase and [00 I ] of perovskite. In the SAED pattern in b the [00 I ] direction of anatase is perpendicular to and in c parallel to thedoubled axis (2x) ofperovskite (D ofthe orthorhombic cell). (d) Indexing ofa, b, c using orthorhombic perovskite cell.

tector and processed with a Princeton Gamma-Tech Sys-tem IV X-ray analyzer.

Anatase powder X-ray diffraction patterns were ob-tained using a Scintag automated powder diffractometerwith CuKa, radiation. Peak positions were corrected forinstrumental and physical aberrations through the use ofinternal standard lines from Si (Standard Reference Ma-terial 640b). Cell dimensions were determined using Lat-con, a Scintag least-squares lattice parameter refinementprogram.

ExprnrprnNTAl RESULTS

SAED patterns indicate that perovskite is largely re-placed by anatase and that the reciprocal lattices of theminerals are oriented preferentially with respect to eachother. When viewed with the beam parallel to [001] ofanatase, it is apparent that the a axes ofanatase are par-allel and approximately equal in length to the pseudo-cubic a axes of perovskite (Fig. 9a). We make referenceto the pseudocubic axes ofperovskite because this setting

BANFIELD AND VEBLEN: PEROVSKITE TO ANATASE AND TiO, (B) 5 5 1

0.381

0.380

0 379

0 378

o 377

0 3763

sample #

Fig. 10. Comparison between the lengths of the a axes (innm) of undoped anatase: I : Cromer and Herrington (1955);2 : JCPDS value; 3, 4 : values for undoped synthetic anatasedetermined using the procedures and equipment described here(Bischoft in preparation) and for anatase replacing perovskite,5. The 3o error bars for the standard deviations from the refineda-axis dimensions are shown.

permits an easier comparison with anatase. The relation-ship between the pseudocubic and orthorhombic cells isgiven by Megaw (1973) and White et al. (1985). For clar-ification, we have shown the orientation of the pseudo-cubic axes on the SAED patterns in Figures 9a,9b,9cand included the orthorhombic indices on the diagramsin Figure 9d. Although electron diffraction patterns showdoubling of one axis (b of the orthorhomic perovskitestructure), the c axis of anatase shows no tendency topreferential development parallel or perpendicular to thisdirection (Figs. 9b, 9c). Thus, as far as the weatheringreactions are concerned, perovskite behaves as ifit werecubic or pseudocubic. The origin of broad streaks andextra reflections in SAED patterns and the associatedmoir6 fringes in images will be discussed below.

Areas clearly showing the contact between anatase andperovskite were rare. In general, either unaltered perov-skite or porous aggregates of anatase were encountered.Where contacts were viewed with the beam parallel to

[010] of anatase and to a pseudocubic a axis of perov-skite, strips of anatase only a few tens of nanometers widewere preserved adjacent to the interface. No areas ofamorphous Ti oxide were detected along the boundaries.

AEM analyses ofthe anatase typically showed the pres-ence of 2-5o/oFeO (occasionally up to l0o/o FeO associ-ated with goethite-anatase intergrowths), l-2o/o CaO, and< lolo AlrO, and MgO. No other impurities were detected(e.g., Nb, REE). Figure l0 illustrates that the c axes ofanatase after perovskite are demonstrably larger than val-ues for pure anatase. The value determined for the c axiswas smaller at the lo level but not significantly differentat the 2o level. The larger a dimension compared with

Fig. I l. Low-magnification electron micrograph down [001]ofboth anatase (A) and perovskite (P) showing little evidencefor substantial volume change associated with the perovskite-anatase reactron.

undoped anatase may indicate cationic substitutions in-volving Fe3*, OH, and interstitial Ca within the anatase.

At low magnification with the beam parallel to [001]of anatase (Fig. I l), anatase and perovskite were inti-mately intergrown and the material exhibited very lowporosity. At higher magnification (Fig. l2), most inter-faces between perovskite and anatase were approximately

Fig. 12. High-resolution image down [001] of both anataseand perovskite showing the orientation ofthe interface betrweenthese minerals. Darker areas within the anatase (arrowed) may

be relict strips ofperovskite.

5s2 BANFIELD AND VEBLEN: PEROVSKITE TO ANATASE AND TiO, (B)

Fig. 13. (a) High-resolution image down [010] of anatase (A) and [001] of perovskite (P) showing the orientation of the interface.Strips ofa third phase with a lattice spacing larger than those found in anatase are indicated (I). (b) SAED pattern. (c) Enlargementof the image in a showing the perovskite margin.

parallel to the pseudocubic a axes of perovskite and aaxes of anatase. Lattice fringes are not present in the im-age of anatase because the spacings are so small that theywere excluded by the objective aperture (spacings smallerthan 0.22 nm were excluded). Lattice fringes associatedwith the irregular strips within anatase are probably as-sociated with relict perovskite.

Figure 13 illustrates a contact between perovskite andanatase viewed down [010] of anatase. The numerousmoir6 fringes inclined (two orientations) to the interfaceare a sample preparation artefacl (discussed below) andshould be ignored. In addition to these, there is a set offringes with a spacing of approximately 0.67 nm parallelto the interface and (001) ofanatase. The spacing ofthesefringes cannot be readily explained as a moir6 effect fromoverlapping perovskite and anatase, and they appear tobe developed in areas where perovskite is clearly notpresent (Fig. l3c). These fringes were not observed at allcontacts and are not present along interfaces other thanthose subparallel to (001) ofanatase. The zone betweenthe perovskite and all crystalline products is very narrow,generally < I nm wide. No amorphous Ti oxide was de-tected in this region.

In addition to anatase, a second TiO, mineral was iden-tified in these samples. SAED patterns indicated that thisis TiO, (B) (Marchand et al., 1980), which has been re-ported by Banfield et al. (1991). TiO, (B) was present (in

both ion-milled and crushed grain-mount samples) as ho-mogeneous areas adjacent to perovskite surfaces (Fig. la)and intergrown with anatase.

SAED patterns from the coexisting minerals indicatedthat the [0 I I ] TiO, (B) and I I I 2] perovskite zones (pseu-docubic axes) were subparallel, and (200) of TiO, (B) and(l l0) of perovskite (pseudocubic axes) were inclined toeach other by about 20". The interface is subparallel to( I I 0) of perovskite (pseudocubic axes) and inclined to c*of TiO, (B) by about 45".

When regions composed entirely of anatase were vieweddown (100), the crystals were found almost always tocontain planar gaps parallel to (001). These gaps give riseto streaking in some electron diffraction patterns (Fig.l 5b).

The SAED patterns (Fig. 9) indicate that the c axis ofanatase develops parallel to the three pseudocubic perov-skite axes. Consequently, the weathered perovskite shouldcontain crystals of anatase in three orientations. How-ever, it is difficult to establish (statistically) whether canatase develops parallel to each ofthese axes with equalprobability.

Figure 16 illustrates an area composed of intergrownanatase crystals in both [001] and [00] orientations. TheSAED pattern (Fig. l6b) indicates a very slight misori-entation between the (001) and (100) planes of the twosets of crystals.

BANFIELD AND VEBLEN: PEROVSKITE TO ANATASE AND TiO, (B) 553

10 nm

Fig. 14. (a) Image showing an area of TiO, (B) exhibiting0.62-nm (001) fringes. TiO, (B) adjacent to perovskite has beendestroyed by ion milling. (b) SAED pattern from TiO, (B). Ar-row indicates the c* direction.

Althougl some areas of anatase viewed with the beamparallel to [001] show little porosity, some aggregates ofneedle-shaped anatase crystals with a high intercrystalporosity were occasionally noted (Fig. l7). The interiors

Fig. 15. (a) High-resolution image down [010] of anataseshowing regular gaps parallel to the (002) fringes. Moir6 fringesare inclined to (002) fringes. (b) SAED pattern. X indicates ad-ditional reflections referred to in the text.

of crystals show signs of recrystallization and eliminationof very fine-scale porosity.

The anatase is extensively intergrown with Ca-bearinglight rare earth element (LREE) phosphate minerals (Fig.l8). Previous work (Mariano. 1989) has shown that the

Fig. 16. (a) Low-magnification image showing an inter-gro\4'th of anatase crystals oriented with their [001] directionsboth perpendicular (as indexed) and parallel to the beam. (b)SAED Dattern lrom area shown in a.

554 BANFIELD AND VEBLEN: PEROVSKITE TO ANATASE AND TiO, (B)

Fig. 17. Aggregate of anatase needles with a high intercrystalporosrty.

LREE pattern of the minerals is identical to that of theperovskite, indicating that the REEs were released fromthe perovskite and reprecipitated without fractionation.SAED patterns from several zone axes were recorded fromneedles such as those shown in Figure 18. These SAEDpatterns were consistent with the identification of thismineral as rhabdophane. Sample 385-P4a also containsabundant, complexly exsolved magnetite that shows nosign ofalteration.

Sample preparation artefacts

Most of the specimens examined in this study wereprepared for transmission electron microscopy by Ar-ionmilling without liquid N, cooling of the specimen. Alldiffraction patterns from the anatase samples prepared inthis way showed extra reflections (particularly apparentin Figs. 9c, l3b, l5b) that could be indexed to a cubicunit cell v/rth d2oo: 0.21 nm that developed in two ori-entations with the ( 100) directions slightly rotated to ei-ther side of the tetragonal anatase axes. Moir6 fringes inhigh-resolution images (Figs. l3a, l5a) were pervasive,occurring in small patches in two symmetry-related ori-entations over the entire sample. Dark-field images sug-gest that the phase causing the moir6s was more abundantalong the sample margins. Diffraction patterns and im-ages from crushed grains of anatase samples showed noevidence of extra spots or moir6 fringes. Furthermore,specimens prepared by ion milling using a liquid-Nr-cooled stage did not show the extra reflections or moir6fringes. We conclude that a surface-coating phase formedduring non-cooled ion milling. Based on the size and shapeof the unit cell, we tentatively suggest that the material isa surface coating ofTiO. In any case, the extra spots andmoir6 fringes should be ignored since they are not relatedto the natural sample microstructure.

Fig. 18. (a) Iow-magnification image showing intimately in-

tergrown anatase (A) and rhabdophane (R). (b) SAED pattern

from rhabdophane.

DrscussronPerovskite alteration

Perovskite is a major igneous mineral in the Salitre IIcarbonatite. Some altered samples are almost completelycomposed of TiO, minerals that pseudomorph the perov-skite. The coexistence of anatase and TiO, (B) with rhab-dophane, calcite, and goethite supports the currently heldview (Mariano, 1989) that the replacement of perovskiteby Ti minerals occurred as the result of weathering.

SAED patterns and images from intergrown perov-skite, anatase, and TiO, (B) (Figs. 9, l3) clearly demon-strate that the TiO, minerals replacing perovskite are to-potactically oriented. The anatase intergrowths composedof crystals in two or more orientations (Fig. 16) may re-flect initiation ofthe anatase-forming reaction at separatesites in the perovskite. The direct overgrowth of TiO,minerals in specific orientations onto the perovskite sur-face is consistent with the possibility that the reactionsinvolve direct structural inheritance ofparts ofthe Ti-Oframework of perovskite.

Alternative mechanisms, (l) dissolution followed byreprecipitation or (2) dissolution followed by formationof an amorphous Ti intermediate and crystallization ofanatase, are not inconsistent with the topotactic relation-ship described above. However, given the extremely lowsolubility of Ti under weathering conditions, a mecha-nism involving complete congruent dissolution of perov-skite seems unlikely. Furthermore, Jostsons et al. (1990)report no Ti in the solutions associated with hydrother-mal perovskite dissolution experiments. The perovskite-TiO, mechanism suggested by Kastrissios et al. (1986,1987) and Myhra et al. (1984) emphasizes the importanceof Ca leaching and development of a disordered or amor-phous Ti-rich surface layer. Because we find no evidencefor an amorphous Ti phase along the interface between

BANFIELD AND VEBLEN: PEROVSKITE TO ANATASE AND TiO, (B) 555

anatase and perovskite (e.g., Fig. l3), we suggest that thismechanism does not predominate during surficial weath-enng.

High-resolution images indicate that the interface re-gion between perovskite and anatase is very narrow (< Inm wide; Fig. l3c). If this remaining perovskite is to beconverted to anatase without large-scale solution trans-port of Ti, Ca leaching and structural reorganization mustoccur along this planar region. In a zone as narrow asthis, distinguishing among transport of componentsthrough a solution layer, surface restructuring, growth ofan amorphous intermediate film, and direct structural in-heritance is extremely difrcult.

The presence of two metastable polymorphs in placeof the thermodynamically stable polymorph rutile sug-gests the importance of kinetic and structural factors indetermining the reaction products. The formation of thesepolymorphs may be controlled by the specific structuralinheritance involved in the reaction mechanisms or bythe structure of the surface onto which epitactic growthoccurs.

The growth of two TiO, minerals suggests that decal-cification ofperovskite can occur by alternate routes withsimilar energetic requirements. Although anatase is by farthe predominant Ti-bearing alteration product, the pres-ence of TiO, (B) is of significance. TiO, (B) has beenreported in low-grade metamorphic rocks (Banfield et al.,1991), but it has not been recognized previously as aweathering product.

Images recorded with the beam parallel to [001] of an-atase (e.9., Fig. I l) srrggested that the areas occupied byreactant and product structures is approximately the same,as observed in this orientation. Contacts between perov-skite and anatase showing the third dimension of the in-tergowths (beam parallel to ( 100) anatase) and preserv-ing strips of anatase greater than a few nanometers widewere extremely rare. In regions where perovskite had beencompletely destroyed, the anatase contained abundantgaps (Fig. l5), suggesting that a substantial reduction involume occurs parallel to [001] of anatase.

Recrystallization of the anatase to form arrays of elon-gate crystals results in the appearance of very large holes.The very significant volume change associated with theperovskite-anatase reaction is particularly apparent wherethis recrystallization has occurred.

The perovskite to anatase and TiO, (B) reactions

It is not possible to use textural arguments to provethat transformations involved leaching of Ca and recon-struction of the existing Ti-O framework rather than oc-curring through a transient amorphous intermediate. Wefavor the former mechanism because (l) an amorphousphase was not detected and (2) the direct mechanismsdescribed below, which are a consequence of the closestructural relationships between perovskite, anatase, andTiO, (B), provide reasonable reaction pathways consis-tent with the orientation relationships between reactantand products.

Figure l9a shows an idealized (undistorted) represen-tation of the perovskite structure viewed down (100)(pseudocubic axes). Orthorhombic distortions from thisidealized cubic arrangement involve displacement of theO atoms with minimal disruption of the cation array(O'Keefe and Hyde, 1977; White et al., 1985).

The conversion of perovskite to anatase and TiO, (B)requires elimination of Ca, which is transported awayfrom the reaction site and reprecipitated as calcite in oth-er areas. The reaction also requires conversion of somecorner-linked to edgeJinked pairs of octahedra. The re-moval of every second plane of Ca atoms by leaching andthe collapse of the Ti-O framework, as indicated by thearrows in Figure l9a, generate a hypothetical structure(Figs. l9b, l9c) with the composition Ca.rTiOrr. Thisstructure is closely related to sheared perovskite deriva-tive structures in the (Na,K,Ca,Nb)TiO, system (Hydeand Andersson, 1988). The arrangement of more darklystippled octahedra in Figures I 9b and I 9c is found in theTiO, (B) structure.

The micrographs shown in Figures l3a and 13c illus-trate a phase developed at the interface between anataseand perovskite that is characteized by a 0.67-nm spacingperpendicular to the interface, between that of droo of an-atase (0.475 nm) and the 0.77-nm spacing equivalent to[020] perovskite (pseudocubic axes). Because this mate-rial has the characteristics predicted for the intermediate,partly decalcifred weathering product (Figs. l9b, 19c), itmay have the structure shown in Figure l9c.

The second step to form anatase from the intermediatephase may occur by removal of the remaining Ca andcollapse of the Ti-O framework, as indicated by the ar-rows in Figure l9c. The experimentally observed orien-tations of the intergrown perovskite and anatase, thepresence of what is interpreted to be the predictedintermediate phase, the orientations of the interfaces be-tween the three phases, the seemingly small volume changeassociated with the intergrowths in two dimensions,and the significant volume change associated with thethird dimension parallel to [001] of anatase are all con-sistent with the view that perovskite is converted to an-atase by leaching and structural collapse as illustratedin Figure 19. Consequently, the reaction can be writ-ten 2CaTiOr(pro") + 2CaorTiOrnr,, + Cafrd, + Oe) -2TiO2("n","".) + 2Cai^[, + 2Ol;or. The actual aqueous spe-cies are not known.

The TiO, (B) structure can be generated from the in-termediate, partly decalcified structure (Fig. l9c) by analternative second step that involves Ca removal and col-lapse of the Ti-O framework as indicated in Figures 19eand l9f. This reaction can be written 2CaTiO, (prov.) +2CaorTiOrr,,",, + Cafg + Oi; - zTiO, @, + 2Ct(^ +2Offi. Consequently, we suggest that the formation of TiO,minerals from perovskite during weathering can be at-tributed to stepwise decalcification, with alternative di-rections of second-stage collapse of the Ti-O frameworkresulting in either anatase of TiO, (B). The conversion ofperovskite to TiO, minerals by a mechanism involving

556

(a) Perovskite:

BANFIELD AND VEBLEN: PEROVSKITE TO ANATASE AND TiO, (B)

structural collapse eliminates the requirement for large-scale transport of Ti in solution.

For the reaction to involve inheritance of slabs of theperovskite Ti-O framework, it is necessary to remove Ca.The presence of abundant gaps perpendicular to the di-rections of volume reduction and the proposed structuralcollapse suggest that relatively narrow slabs of perovskitereact at one time. A possible mechanism involves leach-ing of Ca from the surface layer (as occurs to depths ofat least 3 nm in labradorite dissolution; Hochella et al.,1988), with charge balance maintained by bonding of twoH* to two O atoms. This leaching opens pathways (viathe vacant Ca sites) to the interior of the crystal, throughwhich Ca could be transported to the surface. Collapse asindicated by Figure l9a superimposes the two OH groups,so that the excess structural O can be eliminated as HrO.This could be written CaTiO. + 2H* + (s.z+ * HrTiOr;H , T i O 3 - T i O r + H , O .

In this study we have not detected amorphous Ti oxideor brookite in the alteration assemblage, which distin-guishes the weathering reaction from the reaction studiedexperimentally at higher temperatures (e.g., Kastrissios etal., 1987). The absence ofbrookite is not surprising if, aswe suggest, the natural weathering reactions involve adirect structural transformation. As shown in Figure 4,the brookite structure contains anatase-type slabs in asecond orientation that cannot be generated from the pe-rovskite or intermediate structure by simple collapse ofthe Ti-O framework.

CoNcr,usroNs

We have presented a way of viewing the TiO, struc-tures that highlights their similarities and provides cluesto the mechanisms by which the polymorphic transfor-mations between them occur. Considering these struc-tures to be constructed from slabs or blocks related byshear or cation displacement simplifies the study of theperovskite to anatase and TiO, (B) weathering reactions(this study) and the TiO, (B) to anatase (Brohan et al.,1982; Banfield et al., l99l) and rutile-TiO' II transfor-mation mechanisms (Hyde and Andersson, 1988).

The FBB approach to interpreting other transforma-tion mechanisms, such as brookite or anatase to rutile orTiO, (H) to anatase, cannot yet be assessed. However, wehave made predictions about these reactions that can betested with future experimental work. If an FBB descrip-tion appropriate to all the observed transformations couldbe found, it would further enhance understanding oftherelationships between these minerals.

The results ofthe perovskite weathering study suggestthat the reactions involve direct structural inheritance ofplanes ofcorner-linked Ti octahedra and proceed by step-

F

Fig. 19. Diagram illustrating the stepwise conversion ofpe-rovskite (viewed down one pseudocubic axis) to anatase (a), (b),(c), (d); and Tio, (B) (a), (b), (e), (f).

wise removal of Ca and structural collapse to form edge-sharing octahedral pairs. Although some redistribution ofTi by dissolution and reprecipitation may occur (e.9.,during recrystallization), the mechanism described hereeliminates the need for large-scale solution transport ofthis very insoluble element. The formation of anatase andTiO, (B), rather than other TiO, polymorphs, emphasizesthe importance of the structure of the precursor mineralin determining the weathering product formed.

AcxNowr,BocMENTS

We thank Anthony Mariano for providing the weathered Salitre II car-bonatite samples and initial information about their mineralogy. TimothyWhite, George Guthrie, Jeffrey Post, and Dimitri Sverjensky providedhelpful comments that geatly improved this manuscript. This work wassupporred by NSF grant EAR-8903630.

RnrnnrNcns crrnoAndersson, S. (1969) Mechanism of transformation of rutile - a-PbO:.

A simple cation rearrang,ement. Bulletin de la Societ6 Chimique deFrance.4. 1065-1066.

Banfield, J.F., Veblen, D.R., and Smith, D.J. (1991) The identification ofnaturally occurring TiO, (B) by structure determination using high-resolution electron microscopy, image simulation, and distance-least-squares refinement. American Mineralogist, 7 6, 343-353.

Baur, W.H. (1961) Atomabst?inde und bindungswinkel im brookit, TiOr.Acta Crystallographica, 14, 214-216.

Brohan, L., Verbaere, A., and Tournoux, M. (1982) La transformationTiO, (B) - anatase. Materials Research Bulletin, 17, 355-361.

Bursill, L.A. (1979) Structural relationships between p-gallia, rutile, hol-landite, psilomelane, ramsdellite and gallium titanite type structures.Acta Crystallographica, B35, 530-538.

Bursill, L.A., and Hyde, B.G. (1972) Rotation faults in crystals. Nature-Physical Science, 240, 122.

Coe, R.S., and Kirby, S.H. (1975) The orthoenstatite to clinoenstatitetransformation by shearing and reversal by annealing: Mechanism andpotential applications. Contributions to Mineralogy and Petrology, 52,29-55.

Cromer, D.T., and Herrington, K. (1955) The structures of anatase andrutife. Journal ofthe American Chemical Society, 77, 4708-4709.

Dachille, F., Simons, P.Y., and Roy, R. (1968) Pressure-temperature stud-ies of anatase, brookite, rutile, and TiO, (ID. American Mineralogist,53, 1929-1939.

Hochella, M.F., Ponader, H.B., Turner, A.M., and Harris, D.W. (1988)The complexity of mineral dissolution as viewed by high resolutionscanning Auger microscopy: Labradorite under hydrothermal condi-tions. Geochimica et Cosmochirnica Acta, 52,385-f94.

Hyde, B.G. (1987) The effect ofnon-bonded, anion-anion interactions onthe Caclrlrutile transformation and on the bond lengths in the rutiletype. Zeitschrift fiir Kristallographie, 179, 205-213.

Hyde, B.G., and Andersson, S. (1988) Inorganic crystal structures,430 p.Wiley Interscience, New York.

Hyde, B.G., and O'Keefe, M. (1973) Relations between the DO, (ReO)structure type and some 'bronze' and 'tunne[' structures. Acta Crystal-lographica, 429, 243-248.

Hyde, B.G., Bagshaw, A.N., O'Keefe, M., and Andersson, S. (1974) De-

55',7

fect structures in crystalline solids. Annual Review of Material Science,4,43-92.

Jostsons, A., Smith, K.L., Blackford, M.G., Hart, trLP., Lumpkin, G.R.,McGlinn, P., Myhra, S., Netting, A., Pham, D.K., Smart, R.SI.C., andTurner, P.S. (1990) Description of Synroc durability: Kinetics andmechanisms of reaction. NERDDP Report no. 1319,261 p. ANSTOLucas Heights Research l-aboratories, New South Wales, Australia.

Kang,Z., and Bao, Q.X. (1986) A study of the interaction of VrO'/TiO,(anatase) by high-resolution electron-microscopy (HREM). AppliedCatalysis, 26, 25 l-263.

Kastrissios, T.K., Stephenson, M., Tumer, P.S., and White, T.J. (1986)

Epitaxial growth ofTiO, polymorphs on perovskite. Proceedings ofthellth International Congress on Electron Microscopy, Kyoto, 1697-l 698.

Kastrissios, T.K., Stephenson, M., and Turner, P.S. (1987) Hydrothermaldissolution of perovskite: Implications for Synroc formulation. Journalof the American Ceramic Society,70, Cl44-C146.

Kay, H.F., and Bailey, P.C. (1957) Structure and properties of CaTiOr.Acta Crystallographica, 10, 219-226.

Latroche, M., Brohan, L., Marchand, R., and Toumoux, M. (1989) Newhollandite oxides; TiO, (H) and IQouTiO'. Joumal of Solid State Chem-istry, 81, 78-82.

Marchand, R., Brohan, L., and Tournoux, M. (1980) TiO, (B), a newform of titanium dioxide and the potassium octatitanate K,TisOr?. Ma-teriafs Research Bulletin, 15, 1129-1133.

Mariano, A.N. (1989) Economic geology of rare earth elements. In Min-eralogical Society of America Reviews in Mineralogy, 21, 309-337.

Megaw, H.D. (1973) Crystal structures: A working approach, 563 p. W.B.Saunders, Philadelphia.

Moore, P.B. (1986) Quartz: Variations on a theme. American Mineralo-gist, 7 1 , 540-546.

Myhra, S., Savage, D., Atkinson, A., and Riviere, J.C. (1984) Surfacemodification of some titanate minerals subjected to hydrothermalchemical attack. American Mineralogist, 69, 9O2-9O9.

Navrotsky, A., and Kleppa, O.J. (1967) Enthalpy of the anatase-rutiletransformation. Journal ofthe American Ceramic Society, 50, 626.

Nesbitt, H.W., Bancroft, G.M., Fyfe, W.S., Karkhanis, S.N., and Nishi-jima, A. (1981) Thermodynamic stability and kinetics of perovskite

dissofution. Nature. 289. 358-362.O'Keefe, M., and Hyde, B.G. (1977) Some structures topologically related

to cubic perovskite (E2,), ReO3 (DOJ and Cudu (LlJ. Acta Crystal-lographica. B33, 3802-38 I 3.

Post, J.E., and Burnham, C.w. (1986) Ionic modeling of mineral struc-tures and energies in the electron gas approximation: TiOr polymorphs,quartz, forsterite, and diopside. American Mineralogist, 7 1, 1 42-1 50.

Shannon, R.D., and Pask, J.A. (1964) Topotaxy in the anatase-rutiletransformation. American Mineralogist, 49, 17 07 - 17 17 .

Simons, P.Y., and Dachille, F. (1967) The structure of TiO'-II, a highpressure phase of TiOr. Acta Crystallographica, 23, 334-335

- (1970) Possible topotaxy in the TiO, system. American Mineral-ogist,55, 403-415.

Tournoux, M., Marchand, R., and Brohan, L. (1986) I:yered K,TioOtand the open metastable TiO, (B) structure. Progress in Solid StateChemistry, 17,33-52.

Veblen, D.R. (1991) Polysomatism and polysomatic series: A review andapplications. American Mineralogist, 7 6, 80l-826.

White, T.J., Seeall, R.L., Barry, J.C., and Hutchison, J.L. (1985) Twinboundaries in perovskite. Acta Crystallographica, 84l, 93-98.

M.c.NuscRrrr RrcErvED Apnrl 22, l99lMeNuscnrrr AccEprED DrcBMger 18, l99l

BANFIELD AND VEBLEN: PEROVSKITE TO ANATASE AND TiO, (B)