American Mineralogist, Volume 76, pages I l3-127, 1991

The structure and origin of Fe-bearing platelets in metamorphic rutile

Jrr-r.rru.q F. BaNrrnr,D, Dlvro R. Vnnr,rNDepartment of Earth and Planetary Sciences, Johns Hopkins University, Baltimore, Maryland 21218, U.S.A.

Ansrnacr

High-resolution transmission electron microscopy has been used to characterize thedefect microstructure of rutile from chlorite-, chloritoid-, garnet-, and staurolite-grademetapelites. Analytical electron microscopy (AEM) revealed that the rutile contained be-tween 0.5 and 3 wto/o FeO and that the Fe content generally increased with metamorphicgrade. High-resolution images, nanoprobe-AEM analyses, and electron diffraction patternsindicated that the Fe is contained within platelets generally less than I nm wide that havethe hematite structure. The regularly spaced platelets are coherently intergrown parallel to( 100) and (0 I 0) of rutile. They closely resemble platelets reported from experimental stud-ies of Ti3*- and Fe-bearing rutile. We propose that the platelets in natural rutile are he-matite and that apparent tripling of the { l0 I } spacings results from dynamical diffraction.The hematite (or TirOr) structure represents an end-member iurangement of pairs of face-sharing octahedra. At higher temperatures in synthetic rutile these pairs are ordered toform crystallographic shear planes. We interpret the hematite platelets to have formed bya precipitation mechanism. This origin is consistent with the defect distribution and ori-entation and is supported by related experimental studies.

INrnooucrroN

This study describes the defect microstructure of Fe-bearing rutile from a range of low- and medium-grademetamorphic rocks. The investigation follows about 40years ofintensive research by numerous workers on syn-thetic rutile, including reduced, deformed, and Me3*-sub-stituted materials. Despite the attention devoted to syn-thetic rutile, little is known about nonstoichiometry innatural samples.

Electrical conductivity measurements (Baumard et al.,1977) indicate that there are three regions that exhibitdifferent conductivity behavior in Ti3t-bearing rutile: (l)a solid solution of TiO, , containing isolated point de-fects; (2) a two-phase mixture of TiOr-, and Ti,O,,-,;and (3) a series of one- and two-phase regions of Ti,Or,_,(Smith et al., 1982). Marucco et al. (1981) and Gautronet al. (1981) demonstrate that over a range of tempera-tures and O partial pressures, regions exist in TiOr_, wherethe predominant defects are (l) interstitial Ti, (2) doublyionized O vacancies, and (3) O vacancies associated withcharge compensation of trivalent impurities.

Numerous studies of synthetic rutile have used trans-mission electron microscopy (TEM) and electron diffrac-tion to characteize a range of complex defect structures,including extended defects termed crystallographic shearplanes (CSP), point defects, and platelets that develop inorder to accommodate deviations from stoichiometry(e.g., Bursill and Hyde, 19721, Glbb and Anderson, 1972..

* Present address: Department of Geology and Geophysics,l215 W. Dayton Street, Madison, Wisconsin 53706, U.S.A.

0003{04xl9 1/01 02-o l l 3$02.00

Bursill, 1974; Blanchin et al., 1980; Smith et al., 1982;Bursill et al., 1982, 1983, 1984a, 1984b). Catlow andJames (1982) calculated that the most stable defects arevacancies, which exist in equilibrium with shear planes.They attribute the stability of CSP to large stabilizingrelaxations ofthe cations neighboring these planes. Blan-chin et al. (1984, p.36a466) reviewed the historical de-velopment of models for nucleation and growth of CSP.

Ordered structures with compositions Ti,Or,-' with 4< n < 36, termed Magn6li phases, or (Ti,Fe)"Or,-r, canbe derived from the rutile structure by crystallographicshear. These structures are interpreted as regular inter-growths of (l2l) CSP and (0ll) antiphase boundaries(APB) (Bursill and Hyde, 1972;Miyano et al., 1983), andthe CSP can adopt a variety oforientations and spacingsto accommodate variations in composition. High-reso-lution transmission electron microscopy (HRTEM) stud-ies at near-atomic resolution have made many contri-butions to our understanding of the roles of deformationand cooling rates on defect structures of nonstoichiomet-ric rutile and on the extent of disorder in the defect dis-tributions (Smith et al., 1982:' Bursill et al., 1982, 1983,1984a, 1984b; Blanchin et al., 1984:' Otero-Diaz andHyde, 1984). Smaller deviations from stoichiometry (t< -0.03) have been attributed to homogeneous solid so-lution of point defects in the rutile matrix (Baumard etal.,19771' Blanchin et al., 1980).

Bursill and Blanchin (1983) proposed a structural mod-el for interstitial defects in which nonstoichiometry is ac-commodated by incorporation of trivalent Ti atoms inpairs offace-sharing octahedra, rather than in interstitialdefects forming strings of three face-sharing octahedra.

l l 3

BANFIELD AND VEBLEN: Fe-BEARING PLATELETS IN RUTILE

Aggregation of pairs of face-shared octahedra producesCSP.

Although most experiments have been carried out onsamples cooled from high temperatures (e.g., I100 .C),

results obtained from TiOr_" samples annealed at lowertemperatures and then quenched revealed a new type ofplatelet defect parallel to { 100} (Smith et al., 1982: Bursillet al., 1983, 1984a, 1984b). Bursill et al. (1984a) suggestthat aggregation ofcation interstitial defects results in for-mation of either CSP or platelet defects, depending uponcooling history. The Ti3*-bearing defects described byBursill et al. (1984a) closely resemble the platelets de-scribed in our study of metamorphic rutile.

Putnis and Wilson (1978) and Putnis (1978) describedhydrothermal Fe-bearing rutile (0.55 wto/o FeO and 0.28wto/o VrOr) from quartz-rutile rocks. A TEM study of therutile at about l-nm resolution indicated that althoughthe rutile contained no separate Fe-bearing phase, an-nealing at 475 and 595 'C resulted in the precipitation oftwo transitional phases before formation of hematite(Putnis, 1978). The first-formed precipitate (Tr-l) wascoherent with the rutile matrix (no noticeable change inthe rutile diffraction pattern), and was inferred to be Fe-rich. Putnis described the domains of this phase as Gui-nier-Preston zones. The second transitional phase (Tr-2)shared the rutile subcell; based on several diffraction pat-terns, Putnis selected a monoclinic unit cell with a :0.546 nm, b : 0.459 nm, c : 0.887 nm, A : 122.8".Putnis proposed that Tr-2 contains an additional Fe atomin every third (0ll) layer. Bursill et al. (1984a) pointedout that unlike ilmenite and hematite, Putnis'model con-tains infinite chains of face-sharing octahedra and thateach (010) layer has the stoichiometry [FeTi.O.]3*, withno allowance for charge compensation. Bursill et al. con-structed alternative models that explain the repeat of 3x d,o, (rutile) observed in Tr-2 and Ti3*-bearing syntheticrutile.

In the present study, we have detailed the structuresand compositions of defects in rutile from a range ofmetamorphosed pelitic rocks. Our results from meta-morphic rutile also suggested a monoclinic unit cell basedon a rep€at of 3 x d,or (rutile), differing from the unit cellof Putnis (1978) only in the length of b due to splittingof the rutile subcell reflection. We show, however, thatthese results can be better interpreted as resulting fromdynamical diffraction involving finely intergrown hema-tite and rutile. Consequently, our interpretation of thediffraction characteristics and thus the structure of theplatelet mineral differs significantly from Putnis' modelfor defects in artificially annealed Fe-bearing rutile. Themodels of Bursill et al. (1984a) for platelet defects parallelto {100} rutile contain corner' and edge-linked pairs offace-sharing octahedra. Hematite also contains pairs offace-sharing octahedra, and as such represents an orderedend-member case of the structures proposed by Bursill etal. The similarity between the defects in metamorphicrutile and those produced in controlled laboratory studies

provides support for our interpretation of the lamellae inmetamorphic rutile as precipitation products.

ExpnnrprnNrAI, pRocEDURES

Crystals of rutile examined in this study occur in meta-pelite samples from two regions in Vermont. The first setof samples were chlorite- and chloritoid-grade slates fromthe Taconic Range, collected from roadcuts along Ver-mont Route 4 west of Rutland (sample numbers 2ll2and 2ll3). Electron microscopy of chloritoid in theserocks was described by Banfield et al. 1989). The rutilein the Taconic rocks occurs as extremely fine needles(generally no more than 4 pm in length and less than Ipm wide) subparallel to the foliation defined by musco-vite, paragonite, and chlorite. The petrology has been de-scribed by Znn (1960).

The second set of samples (Al, 18, FI65, 333, 561,565, 625; Karabinos, 1981, 1984; Banfield et al., 1989)were garnet- and staurolite-grade schists from the HoosacFormation on the east side of the Green Mountains an-ticlinorium, near Jamaica, Vermont. Rutile crystals (upto 30 pm long and 8 pm wide) occur as inclusions inchloritoid and coexist with ilmenite inclusions in garnet.In these metapelites rutile does not appear to be presentin the sheet-silicate dominated matrix.

Electron-transparent foils were prepared for electronmicroscopy by ion milling 2.3-mm slices of rock removedfrom petrographic thin sections. Specimens were coatedwith C and examined with a Philips 420T transmissionelectron microscope (TEM) operated at 120 keV and aJEM 4000EX TEM operated at 400 keV. In samples whererutile crystals were contained in chloritoid or micas, thesilicates thinned rapidly, leaving isolated peninsulas ofrutile that could be readily located and examined withoutinterference from the matrix (Fig. l). HRTEM imageswere obtained from thin edges of crystals at close toScherzer defocus (approximately 80 nm underfocus forthe 420T and 48 nm underfocus for the 4000EX). Typi-cally, images were recorded at magnifications between105000x and 490000x.

Both selected-area and convergent-beam electron dif-fraction (SAED and CBED) patterns were used in thisstudy to characterize the orientation, unit-cell parame-ters, and fine structure of intergrowths. Micro-CBED pat-terns were used to locate lamellae-rich regions for micro-analysis. The micro-CBED patterns and microanalyseswere obtained using the nanoprobe mode of the Philips420, wrth an effective spot size estimated to be between2 and 3 nm.

Dark-field images were obtained with the crystals ori-ented as closely as possible to two-beam diffracting con-ditions. A relatively small (10-pm) objective aperture wasused to isolate small areas of the SAED pattern, so thatimages could be formed from superlattice reflections,streaks, and split reflections. Typically, the dark-field im-ages were recorded at 60000x to 105000x, using longexposure times.

BANFIELD AND VEBLEN: Fe-BEARING PLATELETS IN RUTILE 1 1 5

Analytical electron microscope (AEM) X-ray analyseswere obtained with an EDAX energy dispersive spec-trometer (EDS) on the Philips 420 and reduced with aPrinceton Gamma-Tech System IV analyzer. AEM pro-cedures generally follow those described by Livi and Veb-len (1987), Appendix 2. Bulk analytical data were ob-tained from platelet-bearing rutile using a relatively largespot size (-60 nm). Duplicate analyses were obtainedfrom numerous crystals in random orientations, withcount rates between 900 and 1300 cps over the 20-keVspectrum and counting times of 100-200 s. Qualitativemicroanalyses were obtained at very low count rates(< 100 cps), over long counting times (up to 500 s), witha spot size of approximately 2-3 nm.

Rrsur,rsElectron diffraction and imaging

Figure I illustrates a low-magrification [00l]-zone im-age of a rutile crystal enclosed by phyllosilicates in a chlo-ritoid-grade metapelite (sample 2l l3). Dark contrast in-dicates the presence of platelets in two orthogonalorientations parallel to (100) and (010) of rutile.

The [001]-zone images of rutile (sample 2l l3) indicatethat the platelets are very narrow, generally only theequivalent of two rutile unit cells wide (Figs. 2a,2b,2c).Platelets are composed of discontinuous segments offsetlaterally by small distances, often one rutile unit cell (Fig.2b). Rutile (l l0) fringes are offset up to one halfofa unitcell across platelets (Fig. 2c). Distortion of the rutile dueto the presence of platelets is indicated by considerablestrain contrast(Fig.2a) and variations in the orientationof the rutile adjacent to platelet margins (Fig. 2b).

The selected area electron diffraction (SAED) patternfor the [001] zone (Fig. 2d) contains both rutile subcellreflections (open arrows) and additional reflections (filledarrows). The additional reflections are weaker and arepresent adjacent to the 200 and 020 but not the 100 and010 rutile reflections.

The SAED pattern from the I lI] zone (Fig. 3; sampleFl65) shows considerable streaking ofreflections that re-sults from subdivision of the rutile by the irregularlyspaced platelets. All SAED patterns from this zone (ex-cept one platelet-free crystal from a staurolite-grademetapelite discussed below) were characterized by thepresence of superlattice reflections with a repeat of 3 xd.,0,,. Apparent superperiodicities (moir6 fringes, see be-low) are present in images perpendicular to the [0] l] and[01] directions.

Images from the [010] zone (Fig. 4a, specimen Fl65)illustrate narrow platelets approximately parallel to (100).Superperiodicities parallel to (101) and (I0l) are clearlyvisible (Fig. 4a), associated with superlattice reflectionspresent in the SAED pattern (Fig. ab).

Figures 3 and 4 illustrate the presence of sharp super-lattice reflections corresponding to a repeat of 3 x 4,0,rdeveloped in two orientations. However, when the rutile

Fig. 1. Transmission electron micrograph of a rutile crystalfrom a chloritoid-grade metapelite (sample 2l I 3), viewed down[001]. Platelets can be seen parallel to (100) and (010).

is viewed down [012] as in Figure 5 (specimen 165), itcan be seen that the superlattice reflections actually havea pencil-like shape. This indicates that the mineral pro-ducing these reflections forms crystals that are well or-dered in two dimensions, but are probably no more thanthree unit cells wide in the third dimension. This is con-sistent with the high-resolution images of these platelets.

The SAED pattern from the [01] zone (Fig. 6a) showssplitting of all reflections except 010 and l0l. Dark-fieldimages were formed from a rutile substructure reflection(Fig. 6b) and from the additional reflectionjust inside it(Fig. 6c). The reflection used to form Figure 6c is illus-trated in Figure 6d. The dark strips parallel to {010} inFigure 6b (arrowed) correspond with the regions diffract-ing strongly in Figure 6c (arrowed) and the platelets in-dicated by arrows in the bright-field image @g. 6e). Theseresults indicate that splitting ofreflections in electron dif-fraction patterns is due to the presence of narrow lamellaeof a second mineral with a different reciprocal lattice.

A higher-magnification image of the [01] zone (Fig.7a; sample Fl65) illustrates discontinuous lamellae par-allel to (010). Microdiffraction (CBED) patterns were takenfrom the rutile matrix and from lamellae-rich regions(Figs. 7b, 7c). The 020 disks from rutile (Fig. 7b) indicatea spacing that is slightly smaller than the spacing fromthe lamella-rich region (Fig. 7c).

The [101] zone of rutile from sample 18 is illustratedin Figure 8. Bands of intensity in the lamellae can be seen

l l 6 BANFIELD AND VEBLEN: Fe-BEARING PLATELETS IN RUTILE

Fig.2. (a) High-resolution images viewed down [001] of rutile (sample 2113) illustrating the presence of platelets parallel to(100) and (010) ofrutile.

BANFIELD AND VEBLEN: Fe-BEARING PLATELETS IN RUTILE

(b) High-resolution images viewed down [00 1] of rutile (sample 2 l l 3) showing that platelets (P) are apparently - l nm

l t7

Fig.2.wide.

perpendicular to the (010) lattice fringes of the rutile, in-dicating a superlattice associated with the mineral form-ing these platelets. The weakly developed periodic con-trast inclined to (010) of rutile (a moir6 pattern) is due to

platelets parallel to (100) of rutile. Sample l8 containsmore platelets than any other sample examined and hasthe highest, and most variable, Fe content (in excess of 3wto/o as FeO in some regions). This sample is unique in

t 1 8 BANFIELD AND VEBLEN: Fe.BEARING PLATELETS IN RUTILE

Fig.2. (c) Image of rutile tilted fractionally from [001] show-ing a platelet (see arrows) that ofsets rutile (ll0) fringes (sightalong marked lines). (d) Part of the [00U SAED pattern showingsplitring ofthe (200) and (020) rutile reflections (reflections arisingfrom ruti-le shown by open arrows, additional reflections by whitearTowsl.

that it is characterized by the unequal development ofplatelets parallel to the crystallographically equivalent(100) and (010) planes. This is illustrated clearly by theimage down [001] of rutile (Fig. 9a) and by the SAEDpattern (Fig. 9b) showing extra diffraction spots (arrowed)associated with only one of the two { 100} directions. Theextra reflections form a reciprocal lattice that is rotatedrelative to the orientation observed in all other samples.

Fig. 3. SAED pattem for [ 1I] zone from rutile (sample 2112)illustrating the presence of a superlattice with a repeat of 3 xd,,0,,, splitting of I I0 and 110 rutile reflections, and pronouncedstreaking of subcell reflections.

Analytical electron microscopy

Average weight percent FeO determinations were ob-tained from the thin edges of rutile crystals from samplesFI65, 561, 18,2ll2,and2ll3. Results from each samplewere highly variable. This almost certainly reflects vari-ations in the abundance of lamellae in rutile. The esti-mated FeO contents for samples listed approximately inorder of increasing metamorphic grade were 0.7 wto/o forsample 2112,0.8 wto/o for sample 2113,1.4 wto/o for sam-ple F165, about 1.6 wto/o for sample 561, and very ap-proximately 3 wto/o for sample 18.

Microanalyses were obtained from very small regionsof sample Fl65 containing several closely spaced lamel-lae (Fig. l0). CBED pattems were observed during anal-ysis to ensure that the reciprocal lattice associated withthe platelet mineral was dominant (e.g., Fig. 7c). Due tothe narrow size of the platelets, it was not possible toanalyze the platelet mineral alone; consequently the datain Figure l0 reflect variable contamination by the essen-tially Fe-free rutile. Considerable concentration of Fe inthe lamellae is apparent, with enrichments commonly fourtimes that of the bulk rutile plus lamellae (Fie. l0).

Abundance of lamellae in the rutile samples

Relatively low-magnification images (0.46-nm fringesjust visible) were obtained with the electron beam parallelto [010] or [I0l] in order to estimate the abundance oflamellae in sample F165. This information, combinedwith the FeO content of the crystal, could indicate thecomposition of the lamellar mineral if the thickness ofthe Fe-enriched region was known accurately. As dis-

BANFIELD AND VEBLEN: FE-BEARING PLATELETS IN RUTILE l l 9

Fig.4. (a) Image viewed down [010] of rutile (sample Fl65)illustrating discontinuous platelets (P) parallel to (100) and aweak superstructure (S) in two orientations parallel to (l0l) and(I01) rutile; (b) SAED pattern for a showing very clear extrareflections.

cussed below, this approach did not produce an accuratelamellar composition. The results suggested sample Fl65contains 3 to 6 volo/o platelets (average of 4.5 volo/o) ineach orientation and thus approximately 9olo by volumeof the Fe-enriched mineral. It was estimated that the totalvolumetric concentration of platelets in sample 2ll2 was5o/o; and in specimen 2113, 60/o. The abundance of la-mellae in sample 18 was estimated to be somewhat lessthan 18 volo/o, but calculations were complicated by theunequal development of platelets parallel to (100) and(010) of rutile.

Only one of the metamorphic rutile crystals examinedin this study did not contain lamellae (Fig. 1l). This crys-tal was included in garnet from a staurolite-bearing sam-ple (Al). The rutile crystal partially encloses an ilmenitecrystal (close to end-member composition) and a mag-netite crystal (at the ilmenite-rutile interface). Ilmenite isa common inclusion in all garnet crystals from the Hoo-sac schist; it characteristically contains magnetite crystalsthat range in size from a few nanometers to hundreds ofnanometers.

Fie; 5. SAED pattern viewed down [012] (sample F165) il'lustrating elongate superlattice reflections.

DrscussroxElectron diffraction, bright- and dark-field imaging, and

X-ray microanalysis were used to study rutile crystals frommetapelites from a range of metamorphic grades. Theserutile crystals contain very nalTow discontinuous plate-lets parallel to (100) and (010) ofrutile (Fie. 2) that areFe-enriched compared to the rutile host (Fig. l0). Thespacing and widths of the lamellae vary to accommodatethe variations in Fe content.

Mineralogic and crystallographic characterization of theintergrowths

The [001] zone of rutile illustrated in Figure 2d revealsthe superposition onto the rutile diftaction pattern of twoadditional reciprocal lattices at 90'to each other, corre-sponding to the two platelet orientations. The lamellaeshare either a f or a f of rutile and have a second recip-rocal lattice vector that is shorter than a* of rutile and at90" to the first. Ifthese directions are selected as a* andb*, with c* defined by the apparent superlattice along thel0l reciprocal lattice row of rutile (Figs. 3, 4), then amonoclinic unit cell can be described with a : 0.54 nm,b : 0.50 nm, c : 0.89 nm, and B : 122.5, similar tothat selected by Putnis (1978) for his intermediate phase(Tr-2). However, it is also possible that the 0.75-nm su-perlattice reflection (3 x d,o, of rutile) is introduced bydynamical diffraction involving the 0.37-nm reflection(3/2 x j,o,1and the 0.25-nm l0l rutile reflection. In thiscase, we can interpret the diffraction patterns as resultingfrom the superposition of zones from a rhombohedral

120 BANFIELD AND VEBLEN: Fe.BEARING PLATELETS IN RUTILE

tJ,

\'td'

irltJt "rlf

oxide (ilmenite or hematite with do,, : 0.37 nm) andrutile.

The shape ofthe superlattice reflections in electron dif-fraction patterns (Fig. 5) and the dark-field images formedfrom split reflections clearly demonstrate that the super-lattice arises from the mineral contained within the nar-row platelets. Consequently, the [010] rutile electron dif-fraction pattern (Fig. a) is logically interpreted in termsof the superposition of diffraction effects from rutile, fromthe narrow lamellae oriented edge-on (giving rise tostreaky reflections), and from the lamellae that lie in theplane of the image (giving rise to the sharp reflectionsthat trisect [01] rutile).

tu

Figure l2 illustrates the individual diffraction pattems(Figs. l2a, l2b, l2c) that are superimposed to form theSAED pattern shown in Figure 4b and illustrated in Fig-ure l2d. These include [010] rutile (Fig. l2b), the (twinned)

[001] hematite zones (from lamellae parallel to (100) ru-tile; Fig. l2a), and (twinned) [00] hematite zones (fromthe lamellae parallel to (010) of rutile; Fig. l2c). Figure13 shows a schematic diagram of the intergrowth. Thediffraction pattem in Figure l2d results if the electronbeam is vertical.

Figure 2c shows an offset of the rutile (l l0) fringes byapproximately 0.5 x 4rror aaross the platelets. Figure 13demonstrates that this could result from the presence of

Fig. 6. (a) SAED pattern from [01] rutile (Fl65) showing splitting of subcell reflections; (b) dark-field image formed from rutilereflection; (c) dark-field image formed from additional reflection; (d) SAED pattern showing the reflection used to form the imagein c (marked by image of an objective aperture); (e) bright-field image down [I0l]. Platelets are indicated by arrows.

{

'v

*

BANFIELD AND VEBLEN: Fe-BEARING PLATELETS IN RUTILE

Fig. 7. (a) The [01] zone of rutile containing lamellae (P) parallel to (010). Micro-CBED pattern from rutile host O) and fromlamellae (c).

t2l

lamellae two unit cells wide within the rutile. The ofsetarises because the dimension perpendicular to the lamel-lae is larger than a of rutile, so that the ofset of (l l0)fringes will vary w'ith the width of the lamellae. Similaroffsets were noted in some regions of high-resolution im-ages of Ti3*-bearing rutile by Bursill et al. (1984a), whoproposed an alternative precipitation mechanism to ex-plain platelets showing this phenomenon.

In his study of the phases produced by annealing Fe-bearing rutile, Putnis (1978) described twinning in an in-termediate phase (Tr-2) and suggested that twinning re-duces the overall strain energy of the precipitate-matrixinterface by reducing the degree of misfit between the

rutile and Tr-2. Electron diffraction patterns in this studyare interpreted to indicate twinning of the platelets. Thetwin operation is a mirror parallel to (001) of the rhom-bohedral oxide; the labels on the a axes related by themirror have been reversed on Figure 13 to satisfy theright-hand rule. Putnis (1978) noted that contrast in im-ages of the intermediate phase is structure factor contrast,not strain contrast. He suggested a lamellar compositionsimilar to that inferred in this study from the lamellarabundance and FeO content described below. Our imagesof lamellae in metamorphic rutile suggest that coherencystrain is associated with the presence ofthe platelets, andthis presumably contributes to the variable appearance

r22 BANFIELD AND VEBLEN: Fe-BEARING PLATELETS IN RUTILE

Fig. 8. (a) The [l0l] zone of rutile (sample 18) illustrating a superlattice (S) associated with the abundant lamellae. Moir6 fringes(M) arise from the presence of inclined platelets; (b) corresponding SAED pattern.

of high-resolution images (e.9., in Fig. 2), as well as to thetotal contrast.

The composition of the lamellae

Using the rutile as an internal standard, we determinedthe c-dimension of the rhombohedral oxide to be ap-proximately 1.38 nm. Plots of hexagonal cell dimensionsvs. composition for ilmenite-hematite solid solutions(Lindsley, 1965) suggest a composition close to that ofhematite, although this method of determination of com-position may not be accurate for coherently intergrownlamellae that are generally no more than three unit cells

wide. The absence ofthe 003 reflection in the I l0] zoneprovides strong evidence for the identification of the min-eral in the lamellae as hematite, or disordered ilmenite,rather than ilmenite. The 003 reflection (hexagonal cellsetting) is absent in space group R3c but present in R3(ilmenite) due to ordering of Fe and Ti. As the stablerhombohedral oxides at low temperature are close to theend-member compositions, we suggest that the compo-sition of the lamellae is probably close to FerOr. Thepresence of hematite rather than ilmenite, which is theoxide predicted to be stable on the basis of the inferredf o, of Ihe metapelites containing rutile, chloritoid, and,

BANFIELD AND VEBLEN: Fe-BEARING PLATELETS IN RUTILE

Fig. 9. (a) The [001] zone of rutile (sample l8) showing unequal development of lamellae (arrowed) parallel to (010) and (100)of rutile; (b) SAED pattern for a illustrating a slight rotation of the lattice of the lamellae (arrowed) relative to that of the rutile.

123

in the case of the Hoosac schists, ilmenite, suggests thatthe lamellae may have formed as a result of a late-stageoxidation event. As ilmenite has a closely related struc-ture and a unit cell very similar to that of hematite, it islikely that under different conditions ilmenite could alsoform narrow lamellae in rutile.

The volumetric data and bulk FeO contents suggestthat the platelet mineral has an Fe:Ti ratio of approxi-mately l:4. We have included this incorrect observation,even though we know that the maximum Fe:Ti ratio isl: l, because it highlights the discrepancy that almost cer-tainly arises in the estimates of platelet volumes. Thesevolumes were greatly overestimated, partly because of thedifficulty in determining the positions of the lamellarboundaries, i.e., in making the platelet vs. matrix dis-tinction. Careful examination of the highest-resolutionimages of the intergrowths (Figs. 2b, 2c) illustrates this.In Figure 2c the platelet could be considered to be definedby the tips of the arrows, or to consist, as it apparentlydoes, ofonly the narrow strip in the interior ofthis zone.In the lower-magnification images used in our volumemeasurements, strain contrast further increases the ap-parent widths of the lamellae.

Dependence of bulk composition on metamorphic grade

The rutile crystals from a variety of metapelites haveFe contents that show a positive correlation with meta-

morphic grade. This may reflect increased solid solutionof Fe in the rutile structure, or possibly an increase in theinterstitial defect concentration, with increase in temper-ature.

0123456Wt Vo FeO

Fig. 10. Wt% FeO content vs. number of analyses for sampleF165. Bulk lamellae + rutile has an average FeO content of 1.4wto/o, while microanalyses from lamellae-rich regions show con-siderable Fe enrichment.

( t AC)U)

hsr-'!

t24 BANFIELD AND VEBLEN: FC-BEARING PLATELETS IN RUTILE

Fig. I l. A platelet-free rutile crystal (Al).

Further research is needed to determine whether thecorrelation between the Fe content of rutile and meta-morphic grade is independent of whole-rock compositionand other aspects of metamorphic history. The potentialof such a correlation may be in its application as a geo-thermometer in rocks that have experienced tempera-tures less than 500'C, where conventional geothennome-ters (e.9., Spencer and Lindsley, 1981) cannot be used.

Comparison with studies of synthetic rutile

The similarity in appearance of Ti3*-bearing plateletdefects (Bursill et al., 1984a) and those described herehas already been noted. Bursill et al. recognized lineardefect contrast parallel to {101} that is very similar tothe linear contrast labeled "s" in Figure 4a.

Bursill et al. (1984a) proposed that models advancedto explain the precipitation of pairs of crystallographicshear planes (CSP) (Bursill and Blanchin, 1983; Blanchinet al., 1984) could be adapted to provide a more reason-able structure for the Tr-2 phase than that proposed byPutnis (1978), as well as the features observed in reducedrutile. The model of Bursill et al. requires that three Ti4*are replaced by four Ti3t cations, with the developmentof pairs of face-sharing octahedra, as are found in he-matite (Bursill et al., 1984a).

Arrangements such as those described by Bursill et al.(1984a, their Figs. 13, 15, and 16) were designed to re-

produce the superlattice with a repeat 6f I x d,,o',. If thetripling of {l0l} arises from dynamical diffraction in-volving the 012 hematite and l0l rutile reflection as webelieve, the requirement for the complex arrangementsshown by Bursill et al. is removed. The (l2I)-CSP models(Fig. 16 ofBursill et al., 1984a) can then be extended inan orderly manner to generate the hematite (or TirOr)structure, consistent with the identification of plateletswith this structure in metamorphic rutile. Thus, extend-ing the work of Bursill and coworkers, we can visualizetrivalent cation substitution in the rutile structure in bothnatural and synthetic samples as resulting in the devel-opment of either strips (discontinuous in some cases) ofthe rhombohedral oxide structure (CSP or Wadsley de-fects) or planes of this structure (platelet defects).

The mechanism by which synthetic rutile accommo-dates Al substitution has been recently reported by Blan-chin and Bursill (1989) and Bursill and Blanchin (1989).In addition to the presence of a-AlrO. precipitates theynote regions displaying a rutile superlattice with a repeatof 3 x 4,0,,. They interpret this superperiodicity as evi-dence for a transitional structure intermediate betweenrutile and alumina (Blanchin and Bursill, 1989; Bursilland Blanchin, 1989). This tripling, which is apparent atthe interface between the rutile and alumina precipitates,may be alternatively explained as a moir6 effect analo-gous to that observed in this study.

BANFIELD AND VEBLEN: Fe-BEARING PLATELETS IN RUTILE 125

1 2 0n

o orD

@ o^ n

tr

a 3 0 1 r

006h+ 006h

o @ o @o i 2 h o i a h o i b n o i l o h

@oizo o2'q n @ozen@ozi6n

h @ 0 3 6 h + o b 6 h @

tr

300 h

"1" € 1 l o h € t

o @

G

o o

@

Fig. I 2. Diagram illustrating the electron diffraction patternsfrom rutile and twinned platelets in two orientations: (a) [001]and [001] of hematite; (b) t0l0l of rutile; (c) tl00l and [I00] ofhematite. The diffraction pattern resulting from the superposi-tion of a, b, and c is illustrated in (d). Open circles indicateadditional reflections resulting from dynamical diffraction.

Nonstoichiometry in other minerals with the rutilestructure

There are a number of other minerals that possess therutile structure or its simple derivatives, including sti-shovite, cassiterite, pyrolusite, and marcasite. It will beinteresting to discover whether lamellae with the rhom-bohedral oxide structure can accommodate nonstoi-chiometry in these oxides. Some evidence exists to indi-cate that similar defects do occur in pyrolusite. Rask andBuseck (1986) describe a phase M, which is characterizedby a tripling along (l0l) in their [010] pyrolusite diffrac-tion patterns (their Fig. 6). Rask and Buseck report thatthe sample contains no cations other than Mn, O, (andH), and they interpret the extra diffraction spots to indi-cate the presence ofa new manganese oxide (hydroxide)mineral. A possible interpretation is that their pyrolusitecontains platelets analogous to those reported here.

The origin of rutile in metapelites

Rutile typically occurs as needle-shaped crystals in thesamples examined in this study. Within each sample,crystals appear to be fairly uniform in size, but they aresmaller in the samples from the chlorite and chloritoidgrades than in samples from the garnet and staurolitegrades. Although the size variation may be explained inother ways, we interpret the needlelike, euhedral mor-phology and composition of the crystals to indicate thatthe rutile grew during metamorphism, rather than beingdetrital in origin.

It cannot be assumed that rutile in the Hoosac For-mation garnet-, staurolite-, and kyanite-bearing samplesgrew by the same mechanism as rutile in the greenschist-

Fig. 13. Block diagram illustrating the intergrowth of twinnedhematite platelets parallel to (100) and (010) of rutile. Solid lineslabeled (l l0). and their dotted extrapolations illustrate the otrsetof these planes evident in Figure 2c.

facies rocks. In the higher-grade samples, rutile and Fe-Ti oxide crystals exist within garnet, but rutile is not pres-ent in the sheet-silicate-dominated groundmass. Rutile ispresent in the chlorite- and chloritoid-grade samples, al-though Goldsmith and Force (1978) suggest that it is typ-ically not reported until much higher grades (e.g., kyanitegrade). Goldsmith and Force (1978) suggest that the ap-parently uncharacteristic early appearance of rutile in theTaconic samples may be due to the unusual compositionof these rocks. Alternatively, their view that rutile is re-stricted to higher-grade samples may simply reflect thefact that very small crystals would be overlooked by tech-niques other than TEM.

Thompson (1972) proposed that in metapelites the sta-bility field for rutile increases at the expense of ilmenitewith increasing grade. Blattner (1976) suggested that ru-tile crystals within garnet form from Ti released as garnetgrows by consuming ilmenite. These considerations sug-gest that the coexistence of ilmenite and rutile in garnetreflects the incomplete consumption of ilmenite duringprograde metamorphism. As no ilmenite or garnet ispresent in the greenschist-facies rocks, we presume thatthe rutile formed via another mechanism. A precursormineral has not yet been identified, but it seems likelythat finely crystalline TiO, (e.g., anatase or brookite) mayhave been present.

r26 BANFIELD AND VEBLEN: Fe-BEARING PLATELETS IN RUTILE

Rutile is the stable polymorph of TiO, over most or allof the range of geologically important conditions (Da-chille et al., 1962; Post and Burnham, 1986) and is foundin igneous, metamorphic, and sedimentary rocks. Theabundance of precipitated oxides in rutile crystals maytherefore be useful in determining the provenance ofde-trital rutile crystals in sedimentary rocks and placer de-posits, for example.

The origin of platelets in rutile

There are two mechanisms that might cause the for-mation of the regularly spaced, Fe-rich lamellae in rutile:(l) periodic growth of hematite during the growth of therutile, and (2) precipitation, or oxidation ofFe and pre-cipitation within the rutile. This second mechanism mustinvolve diffusion ofFe to fairly regularly spaced zones inthe structure, where ordering of cations produces the nar-row strips of rhombohedral oxide structure. We stronglyargue for an origin due to precipitation from Fe-rich ru-tile for the platelets, rather than interpreting them as pri-mary growth features. The reasoning is as follows:

1. Unusual conditions resulting in the periodic copre-cipitation of Fe-enriched lamellae could be invoked toexplain the development of platelets in rutile crystals fromone locality, but it is difficult to believe that such condi-tions are normally encountered during crystallization ofrutile by different reaction mechanisms, in a range ofmetamorphic grades, and from separate localities.

2. The correlation between Fe content and metamor-phic grade suggests, not surprisingly, that Fe solubility inrutile is a function of temperature. It is difficult to explainthis observation if Fe is never in solid solution in therutile.

3. Experimental work of Bursill and coworkers de-scribed above has established that platelet defects withcharacteristics very closely resemblfurg those describedhere form by aggregation ofpairs ofcations that are ran-domly distributed throughout the rutile at very high tem-peratures.

4. The platelets adopt distinct, symmetrically relatedcrystallographic orientations that correspond to the low-temperature minima in Young's modulus (Bursill et al.,1978; Bursill et al., 1984a). As noted by Bursill et al.,(1978) the (100) habit is favored for precipitarion at thesetemperatures.

CoNcr.usroNs

l. We have documented the presence of narrow la-mellae in metamorphic rutile from metapelites that ex-perienced a wide range of peak metamorphic tempera-tures. The data are consistent with the identification ofthe lamellar mineral as hematite (or ilmenite). These la-mellae are generally one or two unit cells wide, and maythus be viewed as extended defects of rhombohedral ox-ide structure in rutile. Variations in Fe content of the bulkrutile are accommodated by increased thickness andabundance of the lamellae.

Our observations are similar in many respects to those

reported previously for experimentally annealed Fe-bear-ing rutile (Putnis, 1978) and for synthetic, Ti3*-bearingrutile (Bursill et al., 1984a). However, subtle but impor-tant differences were noted in the diffraction patterns inthis study, most importantly, that the rutile subcell is notshared. The results suggest that nonstoichiometry in bothnatural and synthetic rutile may be explained by the pres-ence of defects with the hematite structure.

2. The development of the lamellae throughout the ru-tile crystals in structurally reasonable orientations, thecorrelation between Fe content and metamorphic grade,and the regularity in size and shape ofthe platelets strong-ly suggest that the rutile is metamorphic in origin andthat the lamellae are formed by expulsion of Fe originallypresent in the rutile structure. The similarities betweenthese defects and Ti3*-bearing platelets produced undercontrolled experimental conditions by previous workerssupport a retrograde precipitation origin for the platelets.The results from a range of metapelites suggest that pre-cipitation of hematite is extremely common in meta-morphic rutile.

AcxNowr-nocMENTs

We rhank David Smith, Dimiri Sverjenski, Donald Miser, and JamesRask for helpful discussions and reviews. The present research was sup-ported by NSF grants EAR-8609277 and EAR-8903630. Electron mi-croscopy was performed at the Johns Hopkins EM laboratory, which wasestablished with partial funding from NSF grant EAR-8300365, and atthe National Facility for High Resolution Electron Microscopy at ArizonaState University, which is supported by NSF Grant DMR-8611609. Weare grateful to David Smith for assistance with recording high-resolutionelectron micrographs at the ASU facility.

RrrnnnNcns crrnoBanfield, J.F., Karabinos, P., and Veblen, D.R. (1989) Transmission elec-

tron microscopy of chloritoid: Intergowth with sheet silicates and re-actions in metapelites. American Mineralogist, 7 4, 549-564.

Baumard, J.F., Panis, D., and Anthony, A.M. (1977) Study of Ti-O sys-tem and TiO, at high-temperature by means of electrical-resistivity.Joumal of Solid State Chemistry, 20,43-51.

Blanchin, M.G., and Bursill, L.A. (1989) Non-classical twinning of alu-mina precipitates in rutile. Philosophical Magazine A 60, 619-630.

Blanchin, M.G., Faisant, P., Picard, C., Exxo, M., and Fontaine, G. (1980)Transmission electron microscope observations of slightly reduced ru-tile. Physica Status Solidi, 460,357-364.

Blanchin, M.G., Bunill, L.A., and Smith, D.J. (1984) Precipitation phe-nomena in non-stoichiometric oxides I. Pairs of crystallographic shearplanes in reduced rutiles. Proceedings of the Royal Society of London,A 39r, 35r-372.

Blattner, P. (1976) Replacement of homblende by gamet in granulite fa-cies assemblages near Milford Sound, New Znaland- Contributions toMineralogy and Petrology, 55, l8l-190.

Bursill, L.A. (1974) An electron microscope study of the FeO-FerO3-TiO,system and of the nature of iron-doped rutile. Journal of Solid StateChemistry, 1O,72-94.

Bursill, L.A., and Blanchin, M.G. (1983) Structure of interstitial defecrsin non-stoichiometric rutile. Joumau de Physique Irtters, 44, L-165-L-170.

-(1989) Interphase structures observed as alumina precipitates inrutile. Philosophical Magazine A 60,631-642.

Bursill, L.A., and Hyde, B.G. (1972) Crystallographic shear in the higherTi oxides: Structue, texture, mechanisms and thermodynamics. In H.Reiss and J.O. McCaldin, Eds., Progress in Solid State Chemistry, 7,p. 177-253. Pergamon Press, Oxford.

Bursill, L.A., Netherway, D.J., and Grey, LE. (1978) Composition waves

in iron-doped rutile and the relationship between Young's modulusminima and crystallographic shear orientations. Nature, 27 2, 4O5-410-

Bursill, L.A., Blanchin, M.G., and Smirh, D.J. (1982) The nature andextent of disorder within rapidly cooled TiOr ee85. Proceedings of theRoyal Society ofI-ondon, A 384, 135-155.

Bursill, L.A., Blanchin, M.G., and AMelmalek Mebarek (1983) poinr,linear and extended defect structures in nonstoichiometric rutile. Ra-diation Effecrs. 7 4. 253-265.

Bursill, L.A., Blanchin, M.G., and Smith, D.J. (1984a) Precipiration phe-nomena in non-stoichiometric oxides II. {100} Platelet defects in re-duced rutiles. Proceedings ofthe Royal Society oflondon, A 391,37 3-3 9 1 .

Bursill, L.A., Shen, G.J., Smith, D.J., and Blanchin, M.G. (1984b) Emer-gence of small defect contrast within HREM images of nonstoichio-metric rutile. Ultramicroscopy, 13, l9l-204.

Catlow, C.R.A., and Jarnes, R. (1982) Disorder in TiOr_.. proceedings ofthe Royal Society of London, A 384, 157 -17 3.

Dachille, F., Simons, P.Y., and Roy, R. (1962) Pressure-temperature stud-ies of anatase, brookite, rutile, and TiO, (ID. American Mineralogist,53,1929-t9t9.

Gautron, J., Marucco, J.F., and kmasson, P. (1981) Reduction and dop-ing of semiconducting rutile (TiO,). Materials Research Bulletin, 16,575-580.

Gibb, R.M., and Anderson, J.S. (1972) Electron microscopy of solid so-lutions and crystallographic shear structures II. FerOr-TiO, and GarOr-TiO, systems. Journal of Solid State Chemistry, 5, 212-225.

Goldsmith, R., and Force, E.R. (197E) Distribution of rutile in meta-morphic rocks and implications for placer deposits. Mineralum De-posita, 13,329-343.

Karabinos, P. (1981) Deformation and metamorphism of Cambrian andPrecambrian rocks on the east limb of the Green Mountains anticli-norium near Jamaica, Yermont. Ph.D. thesis, Johns Hopkins Univer-sity, Baltimore, Maryland.

-(1984) Deformation and metamorphism on the east side of theGreen Mountain massif in southern Vermont. Geological Society ofAmerica Bulletin, 95, 584-593.

Lindsley, D.H. (1965) Iron-titanium oxides. Carnegie Institution ofWashington Year Book, 64, 144-148.

t27

Livi, K.J.T., and Veblen, D.R. (1987) *Eastonite" from Easton, Pennsyl-vania: A mixture ofphlogopite and a new form ofserpentine. AmericanMinerafogist, 72, lll-125.

Marucco, J.F., Gautron, J., and Lemasson, P. (1981) Thermogravimetricand electrical study of nonstoichiometric titanium dioxide TiOr-, be-tween 800 and I 100 qC. Journal of the Physical Chemistry of Solids,42, 363-367.

Miyano, T., Iwanishi, M., Harada, T., Kaito, C., and Shiojiri, M. (1983)A (l l0) CS structure in reduced rutile crystals. Philosophical Magazine,A,48, 163-167.

Otero-Diaz, L.C., and Hyde, B.G. (1984) A high-resolution electron mi-croscopy study ofdisorder in two types ofrutile-related crystallogra.ph-ic-shear phases. Acta Crystallographica, B 40, 237-244.

Post, J.E., and Bumham, C.W. (1986) Ionic modelling of mineral struc-tures and energies in the electron gas approximation: TiO, polymorphs,quartz, forsterite, diopside. American Mineralogist, 7 1, 142-1 50.

Putnis, A. (1978) The mechanism ofexsolution ofhematite from naturaliron-bearing rutiles. Physics and Chemistry of Minerals, 3, 183-197.

Putnis, A., and Wilson, M.M. (1978) A study of iron-bearing rutiles inthe paragenesis TiOr-AlrO3-PrO5-SiOr. Mineralogical Magazine, 42, 255-263.

Rask, J.H., and Buseck, P.R. (1986) Topotactic relations among pyrolu-site, manganite, and Mnsos: A high resolution transmission electronmicroscopy investigation. American Mineralogist, 7 l, 805-8 I 4.

Smith, D.J., Blanchin, M.G., and Bursill, L.A. (1982) High resolutionimaging of reduced rutiles. Micron, 13, 245-246.

Spencer, ICJ., and Lindsley, D-H. (1981) A solution model for coexistingiron-titanium oxides. American Mineralogist, 66, | 189-1202.

Thompson, J.B., Jr. (1972) Oxides and sulfides in regional metamorphismin pelitic schists. 24th Int€rnational Geological Congress, 10, 27-35.

Znn, E-an (1960) Metamorphism of the l,ower Paleozoic rocks in thevicinity of the Taconic Range in west-central Vermont. American Min-eralogist, 45, 129-17 5.

MeNuscmpr REcETvED JeNuanv 10, 1990Mm;scrum AccEmD NovsMBen 10. 1990

BANFIELD AND VEBLEN: Fe-BEARING PLATELETS IN RUTILE