American Mineralogist, Volume 71, pages 1441-1454, 1986

Transformation mechanisms between single-chain silicates

Ross JonN ANcnr-*Department of Earth Sciences, University of Cambridge, Downing Street, Cambridge CB2 3EQ, England

Arsrnlcr

The single-chain silicate structures of clinopyroxene, bustamite, and pyroxenoids arecompared. The possible types of transformations between these structures are then char-acterized in terms of the change in lattice type accompanying the structural transformation.Four types of transformation are shown to occur; these may be divided into two groups.Observations made by high-resolution transmission-electron microscopy on partially trans-formed samples show that the mechanism of transformation between clinopyroxene andthe pyroxenoids is one ofglide ofline defects on {001} ofpyroxenoid and on {lli} ofclinopyroxene. The structural reorganization associated with the passage of such a linedefect shows that it is not truly a partial dislocation, but that various different displacementsof atoms occur in the core of the defect.

Comparison of the structure of bustamite with that of the pyroxenoids shows that withthe cell setting in which the latter possess a C-face-centered lattice, bustamite has a latticethat is best described as an interleaved complex of two -F lattices. This indicates that therelative positioning of the silicate chains in bustamite is different from that in the pyrox-enoids or clinopyroxene. Consequently, inversions in the solid state between bustamiteand either clinopyroxene or pyroxenoid must proceed by mechanisms that carry out theinversions in two steps. Observations made by high-resolution transmission-electron mi-croscopy on partially inverted samples indicatethat the inversions from bustamite proceedby the propagation through the structure of line defects with Burgers vectors of 7z[010] onthe {102} planes (-41 cell) which create a wollastonite-like intermediate structure. Thisintermediate structure subsequently inverts to either clinopyroxene or pyroxenoid. Theinitial stage ofthe inversions in the opposite direction is shown also to be the creation ofthis wollastonitelike structure, probably by the glide of line defects on the {001} planesof pyroxenoid or the { I I I } planes of clinopyroxene.

INrnonucrroN

There are three main structure types for single-chainsilicates of formula MSiO3 (M: Ca,Mn,Fe,Mg): clino-pyroxene, pyroxenoid, and bustamite. They are distin-guished by the different relative positioning ofthe chainsin each structure, that is, by their lattice types. Transfor-mations between these phases have been the subject ofseveral single-crystal X-ray studies during which one phaseis inverted to another by heating. These studies have shownthat the inversions are topotactic and pseudomorphous(Dent-Glasser and Glasser , 196l: Morimoto eI al., 1966;Aikawa, 1979). The relative orientations of reactant andproduct phases have been used by these workers to deriveatomic-scale transformation mechanisms that involve thecooperative difiirsion of cations. This derivation of atom-ic-scale mechanisms from the relative orientation ofreac-tant and product phases in a transformation is clearly notsatisfactory. And although the pyroxenoids have been thesubject of many high-resolution transmission-electron

* Present address: Geophysical Laboratory, Carnegie Institu-tion of Washington, 2801 Upton Street, N.W., Washington, D.C.20008. u.s.A.

microscopy (unrnru) studies, there has yet to be an attemptto relate the observed faults to possible transformationmechanisms between the various structures. Indeed, sincemost of the high-resolution observations of pyroxenoidshave been made on synthetic material, the observed faultsare more likely growth features than the result of incipientphase transformation.

In this paper the three structure types of single-chainsilicates are compared, and the structural reorganizationassociated with each possible transformation is described.After a review of the possible types of transformationmechanism, the results of a nnrnlnl study of a number ofpartially inverted chain silicates are presented. The com-bination of these unit-cell-scale observations with datapreviously available from X-ray studies is shown to beconsistent with the suggestion ofAngel et al. (1984) thatsuch transformations may proceed by the propagation ofline defects through the structure of the starting material.

Stnucrunns

The crystal structures of chain silicates may be de-scribed as the stacking together of tetrahedral and octa-hedral layers (e.g., Pannhorst, 1979). In the case of py-

0003-004x/8 6 / | | Iz-t 44 tso2.oo 1441

1442 ANGEL: TRANSFORMATIONS BETWEEN SINGLE-CHAIN SILICATES

Fig. l. Idealized chain configurations ofpyroxenoids projected onto (100). (a) Wollastonite. (b) Rhodonite. (c) Pyroxmangite.(d) Ferrosilite III. (e) Clinopyroxene. (f) Bustamire.

roxenoids the oxygens coordinating these layersapproximate a close-packed array; this plane is chosen tobe (100). In this cell setting (due to Koto et al., 1976), thetetrahedral chains lie in the (100) planes and extend alongthe c axis. The b axis defines the relative displacementsofadjacent chains within the (100) layers, and the latticetype defines the relative positions ofchains in the adjacent(100) layers. The idealized chain configurations within(100) layers of the observed pyroxenoid structures areshown in Figures la-ld. Each chain repeat unit consistsofa number ofpyroxene-like pairs oftetrahedra, followedby three tetrahedra in a wollastonitelike configuration.Thus the pyroxenoid minerals have chain-length period-icities given by p : 2n + 3 (n integral): wollastonite (p :3), rhodonite (p : 5), pyroxmangite and pyroxferroite(isostructural with p : 7), and ferrosilite III (p : 9). Min-eral stmctures with periodicities greater than nine are notknown, but such repeats have been observed as shortsequences in partially disordered material.

Clinopyroxenes have often been described as pyrox-enoids with formal periodicity of p : *, i.e., with a chain

repeat unit with an infinite distance between consecutivewollastonite-like chain units. Unfortunately this has led,incorrectly, to the description of pyroxenoids as a struc-tural series made up of pyroxene and wollastonite slabs(modules), clinopyroxene thus being composed of an in-finite stack of pyroxene modules (e.g., Koto et al., 197 6;Czank and Liebau, 1980).

Although within a single (100) layer the relative posi-tioning ofthe tetrahedral chains in clinopyroxene and thepyroxeneJike portions of the pyroxenoid chains is thesame (Fig. l), the relative position ofthe tetrahedral chainsin adjacent (100) layers differs in clinopyroxene and thepyroxenoids (Fig. 2). In pyroxenoids the direction oftheb axis is defined by the relative positions of the offsettetrahedra of the wollastonite-chain units, and conse-quently a increases with increasing chain periodicity from99' in wollastonite to 114' in ferrosilite III (Koto et al.,1976). In order to properly compare the pyroxene struc-ture with that ofthe pyroxenoids, a ofthe clinopyroxenecell should be increased to I 20" by a new choice of b axis,as indicated in Figure 2b. With this cell orientation the

Fig.2. Tetrahedral chains in consecutive (100) layers in (a)pyroxenoid and (b) clinopyroxene. The shaded chains consist ofapex-down tetrahedra related to the left-hand chain by the C-lat-tice vectors in each structure. Note that when clinopyroxene isdescribed on axes parallel to those ofthe pyroxenoids, it has an1 lattice.

lattice ofclinopyroxene is body centered. Since the latticetype controls the relative positions ofthe tetrahedral chains,it is clear that the chains in clinopyroxene are arrangeddifferently from those in the pyroxeneJike modules of thepyroxenoids. Ifa structure was composed entirely ofthesemodules, alternate (100) layers of tetrahedral chains wouldhave to be translated by 7z[001] in order to reproduce theobserved clinopyroxene structure. This argument also ap-plies to the P2,/c clinopyroxenes since these have the samegeneral positioning of silicate chains as the C2/c struc-tures; the displacive transformation between the twostructures, when it occurs, consists mainly of tetrahedralfotatron.

Interleaved with the (100) tetrahedral layers are layersof large cation sites that are genorally of octahedral co-ordination. The cation sites in pyroxenoids form bandsthat run parallel to the tetrahedral chains, the bands beingtwo sites wide when associated with the pyroxene-likeportions of the tetrahedral chains and three sites widewhen associated with the wollastonite-like chain units (Fig.3). The structures of the pyroxenoid minerals may there-fore be considered as being formed from the stacking to-gether of two types of structure module. The wollastonitemodule, denoted W, is a one-unit-cell-thick layer of wol-lastonite cut parallel to (001), i.e., approximately perpen-dicular to the chain extension direction. The pyroxene-like module, P, may either be considered to be the residualstructure left after removal of the W modules from rho-donite, or be derived from clinopyroxene by the glide ofaltemate octahedral + tetrahedral layers by Yz[001] - one-half of the chain repeat distance. Thus the P module isalso a layer parallel to (00 I ) of pyroxenoid, but containsa pair oftetrahedra from each silicate chain. The observedstacking sequences are

Wollastonite (W) p:3Rhodonite (WP) p: 5Pyroxmangite (WPP) p :7Ferrosilite III (WPPP) p :9.

The cation sites associated with the W modules are

t443

Fig. 3. Single tetrahedral chains \ilith associated octahedralbands in (a) wollastonite and (b) a long-period pyroxenoid. Notethe association ofthe three-site-wide portions ofoctahedral bandwith the wollastonitelike units in the tetrahedral chains.

generally able to accommodate larger cations than thosein the P modules, because ofthe greater degree ofrotationpossible for the associated wollastonite tetrahedral-chainunits. The various stability fields ofthe stacking sequencescomposing the pyroxenoid structures are thus stronglydependent upon the relative sizes and proportions ofoc-tahedral cations. As the average cation size is increased,those phases with a larger proportion of wollastonite-typeslabs are stabilized with respect to those with less. Thus,in general, the chain-length periodicity ofthe stable py-roxenoid phase increases with increasing pressure, de-creasing temperature, and decreasing average radius ofnontetrahedral cations (Simons and Woermann, 1982).Consequently the chain periodicity of a structure maychange in response to a change in one ofthese variables,giving rise to transformations between pyroxenoids, andbetween pyroxenoid and clinopyroxene. The varioustransformations are best characteized by the change inlattice type that accompanies the change in structure:

Pyroxenoid - Pyroxenoid C - CClinopyroxene - Pyroxenoid I - CPyroxenoid - Bustamite C- "F-complex"Clinopyroxene- Bustamite 1 - "F-complex."

Complications are introduced into possible transfor-mation behavior by the fact that the bustamite structure,which has a chain periodicity of three, is more stable overmuch of the (Ca,Fe,Mn)SiO, system than the wollastonitestructure to which it is closely related. The structure ofbustamite, when described on axes that parallel those ofthe pyroxenoid CI cell settings, does not possess a simplelattice type. Koto et al. (1976) described the resultantlattice as a complex of two interleaved F-lattices.

TuNsronuATIoN MECHANTSMS

In most ofthe oriented transformations between silicateminerals, the oxygen atoms tend to preserve their posi-tions when possible, whereas the Si and other cations areredistributed within this anion frarnework (Taylor, 1960).The transformations considered here are all betweenstructures that possess approximately the same oxygen

ANGEL: TRANSFORMATIONS BETWEEN SINGLE-CHAIN SILICATES

ANGEL: TRANSFORMATIONS BETWEEN SINGLE-CHAIN SILICATES

packing, so that cation displacement is the only majorstructural change associated with each transformation. Thetwo types of mechanisms proposed for these transfor-mations achieve these changes in diferent ways.

Dent-Glasser and Glasser (1961) described the exper-imental inversion of rhodonite to what they termed "wol-lastonite." The mechanism proposed by Dent-Glasser andGlasser (1961) for this transformation requires the move-ment of Si atoms, with minor displacements of the oxygenatoms. A similar mechanism of cooperative cation dif-fusion was proposed by Morimoto et al. (1966) for theinversion of johannsenite (a Ca-Mn clinopyroxene iso-typic with diopside) to bustamite in which the relativeorientation ofjohannsenite and product bustamite is con-trolled by the preservation of the oxygen layers coordi-nating "dense-zones" of cations common to both struc-tures. This type ofmechanism has also been proposed forthe transformation of rhodonite to bustamite (Morimotoet al., 1966) and for that between rhodonite and pyrox-mangite (Aikawa, 1979).

The second type ofmechanism proposed for these trans-formations employs the passage of partial dislocationsthrough a structure to create an atomic configuration thatlocally resembles that of the product phase. Dislocationswere originally proposed as a microscopic explanation ofthe process of slip in metals. The Burgers vector of adislocation is defined as the displacement created by thedislocation between slipped and unslipped material. It isthus a measure of the displacement of all the atoms of thestructure caused by the passage of one dislocation. But inthe case of more complex materials such as the silicatesstudied here, the process of shear by a dislocation createsenergetically unfavorable environments for some atoms,and these undergo a further redistribution into more fa-vorable sites (synchroshear; Kronberg, 1957). The passageof line defects has already been shown to be the mecha-nism by which polytypically disordered bustamite elim-inates stacking mistakes (Angel, 1985). And the obser-vation of stacking faults in spinel from the Peace Rivermeteorite (Price et al., 1982) is consistent with the trans-formations between spinel and spinelloids [specifically thebeta phase of (Fe,Mg)rSiOo in this casel proceeding by thepassage of line defects. Electron microscopy of similarshocked material from the Tenham chondritic meteorite(Putnis and Price, 1979) and of synthetic material (Lacamet al., 1980) supports the proposal by Poirier (1981) thatthe olivine-to-spinel inversion also proceeds under certainconditions by a shear mechanism.

In the following sections the four types of transforma-tion (as characterized by the change in lattice type) be-tween the single-chain silicates are examined in more de-tail. Evidence from HRrEM observations of partiallyinverted samples suggests that these transformations pro-ceed by the passage ofline defects. The orientational re-lationship between the phases created by this mechanismis approximately the same as that predicted by the dense-zone theory.

PvnoxeNorn ro PYRoxENoro (C - C)

Aikawa (1979) studied the inversion between pyrox-mangite and rhodonite by single-crystal X-ray techniques.He proposed that the inversion proceeded by the coop-erative diffusion of two-fifths of the Si and other cationsinto sites that are vacant in the pyroxmangite structure,and that the relative orientation ofthe reactant and pro-duce phases is controlled by the preservation ofthe oxygenlayers coordinating "dense-zones" of cations common toboth structures.

Angel et al. (1984) derived a possible shear mechanismprimarily for the inversion ofjohannsenite to bustamiteand then applied the same principles to inversions in-volving the pyroxenoid structures. It should be pointedout that, as a consequence ofa misunderstanding oftheprecise structural relationship between clinopyroxene andthe pyroxenoids, the Burgers vectors derived in the ap-pendix to Angel et al. (1984) for the pyroxenoid-to-py-roxenoid inversions are not vectors ofthe oxygen sublat-tices of the structures and are therefore incorrect. Theyare now derived correctly for the idealized structures.

The periodicity of pyroxenoids is defined by the fre-quency with which the wollastonite modules appear inthe stacking sequence. Each ofthese modules contributesone offset tetrahedron to each tetrahedral chain (Fig. l),and the principle ofthe derivation ofthe shear mechanismis to create one such offset tetrahedron in a pyroxeneJikeportion of one chain. As described in Angel et al. (1984),this tetrahedron is generated by the movement of a tet-rahedral chain into a position offset from that related bythe Clattice vector. The Burgers vector ofthe shear re-quired for such a movement is the sum ofthe same Clatticevector as for the clinopyroxene-to-bustamite inversion(Angel et al., 1984), Vdl7}l, and the equivalent of theoffset vector, Yr[011]"".. On axes parallel to those of thepyroxenoids (Fig. 2b) this offset vector in clinopyroxenebecomes V+l}l2l. The formal description of this offset vec-tor will vary not only from pyroxenoid to pyroxenoidbecause ofdiferences in chain periodicity, but also withina pyroxenoid because ofthe deviations ofthe tetrahedralchains away from the idealized structures considered here.Alternatively, these deviations may be considered to beaccommodated by the structural relaxation that followsthe shear. In the idealized structures the ofset may berewritten in terms of the chain periodicity p asthl0l4/pl,giving the Burgers vectors of the shears in each structuretype as

WollastoniteRhodonitePyroxmangiteFerrosilite III

Vep 3 a/31 (203)V+12 3 4/5) (205)rhl2 3 4/71 Q07)Vql2 3 4/91 (209).

The slip planes are derived by a similar transformationof the slip plane in clinopyroxene, which becomes { I 0l }on the /-centered cell, and by the constraint that the Bur-

gers vector of a glissile dislocation must lie in the slipplane. Note that the slip planes are slightly inclined to the(001) interfaces of the W and P modules of which thepyroxenoids are composed, and therefore shear on theseplanes would cut across the modules and convert onemodule type to the other. Calculation shows that if rho-donite and pyroxmangite are intergrown with these slipplanes and with their b axes parallel, the relative orien-tation of the two phases is the same as that predicted bythe dense-zone presewation theory.

Observations

The easiest transformation to study is that between rho-donite and pyroxmangite because of the low temperatureat which the inversion takes place. Aikawa (1979) carriedout annealing experiments to invert natural pyroxman-gites to rhodonite. Despite undergoing partial oxidationof the Mn2+ and Fe2+ in the pyroxmangite, the productrhodonite showed the orientation relationship to the py-roxmangite predicted by the dense-zone theory of Mori-moto et al. (1966). Although Aikawa (1979, 1984) wasable to demonstrate that many natural intergrowths ofpyroxmangite and rhodonite also exhibit this orientationrelationship, Ried and Korekawa (1980) reported fine-scale intergrowths and chain-periodicity faults in plrox-enoids that are oriented slightly differently. Unfortunatelythe other HRTEM studies of pyroxenoids reported in theliterature (Czank and Liebau, I 980; Jefferson et al., I 980;Jefferson and Pugh, 1981) do not give sufficient detailsfor the precise orientation relationships to be ascertained.These chain-periodicity faults and intergrowths generallyshow very little strain, but Ried and Korekawa (1980)noted that they are often terminatedby dislocations withinthe crystal, which suggested that transformations may pro-ceed by the passage ofpartial dislocations.

In order to ascertain the type of mechanism operativeduring pyroxenoid-to-pyroxenoid transformations, a se-ries of inversion experiments was carried out with syn-thetic material of MnSiO, composition. A glass was pre-pared from a mixture of silica and manganese sulfate bythe same method as described in Angel (1984). A seriesof hydrothermal devitrification runs was then carried out,which produced samples consisting of various mixturesof rhodonite and pyroxmangite, the starting material forthe inversion experiments coming from a two-week runat l.5 kbar and 650'C. Although these conditions are with-in the rhodonite stability field, Hnrenr of material fromthis run showed that it contained an appreciable propor-tion ofpyroxmangite and had a very low density ofstack-ing faults. This material was then annealed under pure Ar(to prevent oxidation) at higher temperatures to invert thepyroxmangite to rhodonite.

Observations made on the products ofseveral annealingruns show that the proportion of rhodonite increases rel-ative to pyroxmangite with increased annealing time with-in the rhodonite stability field. The residual pyroxmangitealso appears to have an increased density of chain peri-

1445

odicity faults, especially of those with the rhodonite struc-ture, but no direct evidence of the mechanism of for-mation of these faults was found. However, a possiblemechanism is suggested by micrographs of a rhodoniteproduced in a longer annealing run. Figure 4a shows anumber of line defects within a rhodonite grain that runparallel to the incident electron beam (along [10]) andthat terminate the (001) chain-periodicity faults. An en-largement of an area of this micrograph is provided inFigure 4b, and the structural interpretation based uponimage matching using the snnt-r (Simulation of High-Res-olution Lattice Images) program (O'Keefe and Buseck,1979) is drawn in Figure 4c. The rows of bright spots inthe micrograph are associated with the wollastonite-likeunits within each tetrahedral chain, so that the periodicityfaults consist of two W modules without an interveningP module, i.e., a short portion of two unit cells of wol-lastonite structure. For the MnSiO. composition, such astructure is not stable, and this sample appears to be inthe process of eliminating such faults by the passage ofline defects through the structure. At each defect it wouldappear that local atom displacements occur, possibly us-ing the empty core of the line defect as a fast diftrsionpath. The concept of a Burgers vector of a shear is notstrictly applicable to such a defect, but bearing in mindthat the line ofthe defect is parallel to a Clattice vector,the actual atomic reorganization may well be that whichwas described by the shear mechanism derived above.The only difference lies in the planes on which the defectstravel through the structure; the slip plane of the shearmechanism was (205) in rhodonite, but is obviously (001)in the micrographs, while consideration of the structuralrearrangement required at the defects of Figure 4 indicatesthat atom displacements with components normal to the(001) planes are necessary. Such atom rearrangementswithin the core of the line defect will also include themovement of the octahedral cations, which was previ-ously assigned to the final mechanistic step ofsyncroshearafter the passage ofa partial dislocation.

Similar line defects were also observed in the productsof the very short hydrothermal annealing runs of theMnSiO, glass, in which longer chain periodicities are beingeliminated (Fig. 5). It therefore seems likely that the py-roxmangite-to-rhodonite inversion carried out in theseexperiments also proceeded by the passage of these linedefects. This would account for the increasing density ofchain-periodicity faults in the pyroxmangite, and if suchdefects nucleated at the wrong relative spacings, they wouldgenerate the WW-type (00 I ) chain-periodicity faults beingeliminated in Figure 4.

Cr-rNopvnoxENE To PYRoxENoID (I ' C)

The replacement of clinopyroxene by pyroxenoid wasdiscussed at length by Angel et al. (1984), who proposedthat the transformation mechanism was one of dislocationglide followed by cation synchroshear. The slip plane maybe derived by considering which lattice plane of clino-

ANGEL: TRANSFORMATIONS BETWEEN SINGLE-CHAIN SILICATES

1446 ANGEL: TRANSFORMATIONS BETWEEN SINGLE.CHAIN SILICATES

Fig. 4. Transformation ofstacking sequences in rhodonite by line defects that run parallel to the incident direction ofthe electronbeam, [110]. (a) Note the "en echelon" arrangement of the line defects that lie on every (001) plane of the rhodonite. (b) Enlargementofa portion ofthe micrograph in (a), together with a structural interpretation (c). Note that interpretation ofthe structure in thecore ofthe defect is not possible at this resolution.

pyroxene best matches that of(001) ofpyroxenoids andhas the same centering of lattice points. The latter con-dition ensures that the relative positioning ofthe chainsin a pyroxenoid structure formed by the inversion of cli-nopyroxene is correct. Figure 6 shows that the t I I I1 planesof clinopyroxene fulfill these conditions, this being theslip plane of the mechanism proposed by Angel et al.(1984), the Burgers vectors of the dislocations beingv4<23T).

Intergrowths of clinopyroxene with various pyroxe-

noids have been reported in Hnrnr'r studies of syntheticmaterial. In a study of ferrosilite III, Czank and Simons(1983) found long periods of pyroxene-like structure in-tergrown with the pyroxenoid with {lll}"o- parallel to{001}eyd. A more detailed study of similar intergrowthsby Ried ( I 984) confirmed the orientation relationship pre-dicted by the proposed shear mechanism and showed thatthe displacement vector associated with the chain-peri-odicity faults in the clinopyroxene was V+(01 l). Since theaddition ofa Clattice vector, 7z[I I0], and Yr[0I 1] is equal

Fig. 5. Line defects (some circled) running parallel to [1I0]in a disordered pyroxenoid.

to t/+l23ll, this also is consistent (see, for example, Ame-linckx and Van Landuyt,1976, p. 95) with the proposedmechanism. Even so, these observations were made onsynthetic material in which the microstructures describedare growth features, and do not necessarily constitute evi-dence for the transformation mechanism.

Veblen (1985) reported observations made by rrnreuon a natural johannsenite that clearly shows petrographicevidence of having undergone partial inversion to rho-donite. His observations of pyroxenoid lamellae parallelto the { I I I} planes of the clinopyroxene matrix are strongevidence in favor ofthe shear mechanism proposed byAngel et al. (198a). Similar chain-periodicity errors wereobserved occasionally in natural johannsenite (Fig. 7a) inwhich longer periodicities corresponding to {001} pyrox-enoid spacings occur parallel to the {l 111 planes of theclinopyroxene matrix. Some of the material was experi-mentally annealed in a Mn solution below the temperatureof inversion to bustamite in order to demonstrate thatsuch microstructures are the product of a partially com-pleted transformation. The increase in the Mn content ofthe structure has promoted the creation ofextensive stack-ing disorder on {llI} of the clinopyroxene matrix (Fig.7c). The associated selected-area electron-diffraction pat-tern (Fig. 7b) shows maxima corresponding to wollaston-ite and rhodonite, together with diffuse streaks parallel tothe (l I l) plane normal of the clinopyroxene. By contrast,difraction patterns from clinopyroxene annealed in CaCO,solution show strong maxima from wollastonite alone (Fig.7d), whereas the maxima corresponding to longer chainperiodicities are very difuse or absent. The orientationrelationship observed in these samples is (l0l) of cli-nopyroxene parallel to (100) ofpyroxenoid as expectedfrom the matching of the two lattice types. The disorderis therefore characterized by short sequences of pyrox-enoid mineral structures intergrown with clinopyroxene,with {00 I } ofthe pyroxenoids parallel to one ofthe { I I 1 }planes of the clinopyroxene.

Busrnvrrrn TRANSFoRMATIoNS

As was the case for the C - C and C - l transfor-mations discussed above, two types of mechanism havebeen proposed for the inversions ofchain silicates to bu-

1447

a t0101 b t0-111Fig. 6. (a) The (001) lattice section ofa pyroxenoid: lattice

parameters taken from pyroxmangite. O) A { 1l I} lattice sectionof johannsenite (clinopyroxene). Lattice spacings in Angstriimsare indicated.

stamite. Dent-Glasser and Glasser (1961) described theinversion ofrhodonite upon heating, but because oftheirbelief that the inversion product was wollastonite and notbustamite, their mechanism of cation difirsion partly gen-erates the wollastonite structure (Prewitt and Peacor, 1964).Morimoto et al. (1966) also studied the inversion of var-ious chain silicates to the bustamite structure. However,inconsistencies appear in the discussion given by Mori-moto et al. (1966); the diagrams from which the dense-zone preservation theory is derived are all ofwollastonite,rather than bustamite, while Figure l2b of their paper,which is labeled bustamite, is identical to Figure l5b,described as wollastonite. It would therefore appear thatinstead ofproducing bustamite, the transformation mech-anisms proposed by Morimoto et al. for both the clino-pyroxene-to-bustamite (1 * f'-complex) and the rhodo-nite-to-bustamite (C - .F-complex) inversions produce(incorrectly) the wollastonite structure.

Because of the differences in lattice types of bustamiteand clinopyroxene, the dislocation mechanism for the in-

Fig. 7. Stacking disorder in clinopyroxene parallel to {1lI}owing to partial inversion to pyroxenoid. (a) Naturaljohannsen-ite. (b) Selected-area electron-diftaction pattem of this materialafter annealing in Mn solution, and (c) the corresponding image.(d) Selected-area electron-diffraction pattern from a similar spec-imen annealed in a Ca solution. Note the predominance of wol-lastonite diffraction maxima (arrowed) over those from otherpyroxenoids.

ANGEL: TRANSFORMATIONS BETWEEN SINGLE-CHAIN SILICATES

t1001

t448 ANGEL: TRANSFORMATIONS BETWEEN SINGLE-CHAIN SILICATES

, \ QD

.:: :,L: :'!-.

.-.+a. .1.--a.

Fig. 8. Lattice sections perpendicular to the chain directionsof (a) wollastonite and O) bustamite in the cell setting of Kotoet al. (1976). The horizontal arrow indicates the relative offsetof the two -FJattices. (c and d) The same lattices with the cellsetting ofPeacor and Prewitt (1963). The open circles representlattice points displaced by one-half of the lattice repeat perpen-dicular to the section relative to those represented by filled circles.

version ofjohannsenite to bustamite proposed by Angelet al. (1984) necessarily consists oftwo stages. The firststage is the creation of a wollastonite intermediate phaseby the passage of line defects, with b : Vq(2311, throughthe structure on the { I I 11 planes of the clinopyroxene asfor the clinopyroxene-to-pyroxenoid (1 - C) transfor-mations described above, followed by the inversion ofthis intermediate to bustamite. The inversion of pyrox-enoid to bustamite would follow a similar mechanism;first, the creation of the wollastonite intermediate by themechanism for pyroxenoid-to-pyroxenoid (C - C) in-versions, followed by inversion to bustamite. In the re-verse direction the transformations require the inversionof bustamite to wollastonite as a first stage, followed bytransformation to the product structure. Thus both typesof inversion mechanism require the existence of an in-termediate phase with the wollastonite structure duringinversions between bustamite and either clinopyroxene orpyroxenoids.

Wollastonite and bustamite

It is clear from Figure I that although the individualtetrahedral chains in bustamite have the same configu-ration as those in wollastonite, their relative positionseven in a single (100) layer are different. Figures 8a and8b compare the lattices of wollastonite and bustamite inprojection down the chain-extension direction. This showsthat ifthe cell setting used for pyroxenoids is applied tobustamite, the lattice type is nonconventional. Koto et al.(1976) described it as an interleaved complex of twoF-lattices with a relative displacement of the origins oft/a(a + b). Although such a cell has advantages in com-paring bustamite with the pyroxenoids, it is inconvenientto use, so although the classification of the transformationsinvolving bustamite will retain the term "F-complex," analternative cell will be used to discuss structural detailsof the transformation.

This simpler cell setting for wollastonite and bustamite

Fig. 9. Faults in a bustamite sample annealed in the stabilityfield of wollastonite. The faults lie on the (102) planes of busta-mite and have a fault vector of y2[010].

was introduced by Peacor and Prewitt (1963) and is il-

lustrated in Figures 8c and 8d. The wollastonite cell is

now primitive, while bustamite has an l-face-centeredcell with a c parameter double that of wollastonite. It is

now clear that the bustamite structure may be derivedfrom that of wollastonite by the translation of alternate(001) layers by Yz[010], one halfofthe chain-repeat dis-tance along the chains. This relationship led Angel (1985)

to propose that wollastonite and bustamite were twomembers of a series of polytypes based on the stacking ofprismatic structural units along the a and c axes of wol-

lastonite. Polytypic stacking variation along a is well knownin wollastonite, and Angel (1985) found intergrowths ofwollastonite and bustamite with (001) common to bothphases. Transformations between these polytypes proceed

by the passage of 7d0l0l line defects on (ft0l) planes; in

Angel (1985) it was noted that, because of the pseudo-

monoclinic symmetry of the structures, glide of Yz[010]

defects on (l0l) of wollastonite and (102) of bustamitewould be an alternative to glide on (001) and would alsogenerate the other structure. Such defects were observedin an initially fault-free Fe-rich bustamite that was an-nealed in the stability field of wollastonite (Fig. 9). These

isolated (102) faults are the beginning ofthe inversion towollastonite, since in the region of a fault the relativepositions of the tetrahedral chains is that of the wollas-

tonite structure.The structural reason for the transformation between

wollastonite and bustamite proceeding by shear on the(102) planes ofbustamite, or (l0l) of wollastonite, is quite

10004 , ) *

l - f",t

*

500A

-t-F-a**

ANGEL: TRANSFORMATIONS BETWEEN SINGLE-CHAIN SILICATES 1449

Fig. 10. (a) High-resolution micrograph of a wollastonite lamella within a clinopyroxene matrix, viewed down [110] of theclinopyroxene. The offset ofthe (l l0) fringes across the fault is indicated. (b) Several wollastonite larnellae parallel to a single {1 I l}plane ofa clinopyroxene grain. (c) Inclined { I I I } lamellae ofwollastonite in a clinopyroxene gtain and (d) subgrains ofpolytypicallydisordered bustamite in the same grain. The bustamite is oriented with [031] approximately parallel to the electron beam, thesubgrains having mutual misorientations of up to 10o.

clear. In both structures these are the closest-packed planesand will thus be far more favorable for the passage ofdislocations than the (00 l) planes, which are the slip planessuggested by simple comparison of the two stacking se-quences. The observation of such faults in bustamite an-nealed in the stability fields of clinopyroxene or pyrox-enoid is therefore interpreted as the beginning of theformation of a wollastonite-like structure as an interme-diate step in the transformation from bustamite to cli-nopyroxene or pyroxenoid. It is to prove the existence ofsuch an intermediate structure that the following obser-vations are presented.

Clinopl'roxene-bustamite (1 - I-complex)

For the study ofthe inversion ofclinopyroxene to busta-mite, a number of annealing experiments were carried outusing natural johannsenite almost completely free ofstacking faults. The few found were parallel to { I 11} and

consisted of short sequences of longer chain periodicitiesofdisordered pyroxenoids (see Fig. 7a). The observationsmade on the products of a series of isothermal annealingof this material at 900"C (under Ar to prevent oxidation)afe now presented.

The first stage in the inversion is the appearance ofisolated faults parallel to the { I I 1 } planes ofjohannsenite,as predicted by the inversion mechanism. These faultscommonly run across the width of the flakes produced bycrushing the sample for nnrsM for distances of severalthousand ingstrtims. However, they are not usually singlelamella of wollastonite; for example the fault in FigurelOa appears to be 28 A wiOe, corresponding to the lengthof four chain-repeat units of a dreierketten single-chainsilicate. In view of the difficulties in interpreting imagecontrast around faults, it would be wrong to conclude thatthis was indeed a lamella of four chain-repeat units ofwollastonite without further evidence. In this case the

1450

evidence is provided by the displacement across the faultof the lattice fringes of the clinopyroxene matrix. Thesedisplacements can be shown to be those that would beproduced by the introduction of four wollastonite-likechain units into each chain, or overall by the introductionof four VqQ3T) shears into the structure.

In specimens that have undergone annealing for longerperiods of time, these { I I 11 lamellae become more nu-merous. As seen in Figure l0b, all of the faults within asingle grain lie parallel to one single {lll} plane oftheclinopyroxene, rather than on both of the symmetry-re-lated planes, a phenomenon first noted by Veblen (1985)in johannsenite being replaced by rhodonite and pyrox-mangite. Such orientational selectivity must be due to theloss of the symmetry elements in the point group of theclinopyroxene which relate the { I I I 1 planes to one another.This could occur throughout an entire grain of the cli-nopyroxene, which thereby makes growth processes non-equivalent on the two previously symmetrically equiva-lent planes, or it may occur at the point of nucleation ofthe line defects that give rise to the faults. In this case themost likely cause of this orientational selectivity is a phe-nomenon Veblen (1985) termed "templating." In a poly-crystalline specimen one of the adjacent grains may, be-cause ofthe relative orientation ofthe two grains, createa grain-boundary environment that favors fault nuclea-tion on just one of a set of symmetry-related planes. Tem-plating from adjacent grains is the favored explanation inthis case because orientational selectivity is observed inthe inversion to bustamite ofpolycrystallinejohannsenite,whereas in their single-crystal experiments where the only"neighboring grains" were small quantities of surface ox-ides, Morimoto et al. (1966) found both possible orien-tations of bustamite inverted from johannsenite.

The need for cation synchroshear and relaxation ofthestructure created by the passage of a single line defectappears to favor the formation of lamellae with widthscorresponding to a few chain repeats ofwollastonite, rath-er than single faults. These multiple width faults will de-velop because the existence of the initial single fault willcreate a favorable environment for the nucleation of sub-sequent line defects. However, because wide wollastonitelamellae are not found, there must be a critical widthbeyond which it is unfavorable for the lamellae to growby further passage ofline defects.

Figure lOc shows a set of {ll1} faults in a thinnedspecimen ofjohannsenite; a few thousand trngstrdms awaythe same original grain of clinopyroxene has inverted tobustamite (Fig. l0d). The bustamite has formed assubgrains that show relative misorientations of up to 10".Each subgrain shows extensive polytypic stacking disorderon (100), which is eliminated during further annealing bythe passage of further line defects through the structureas described by Angel (1985). These observations indicatethat there then follows a period of rapid gowth of thewollastonite nucleated on the lamellae, together with analmost simultaneous inversion to bustamite. That thisinversion is rapid is hardly surprising when one recalls

ANGEL: TRANSFORMATIONS BETWEEN SINGLE-CHAIN SILICATES

that a wollastonite structure with the johannsenite com-position (ideally CaMnSirOu) will be highly thermody-namically unstable with respect to bustamite, owing tothe presence of the relatively small Mn cation. The slowrate of nucleation and growth of the lamellae may also berelated to this instability, or alternatively to the need todiffuse Mn away from, and Ca to, the lamellae. A furtherlimitation on the coherent growth of the lamellae withinthe clinopyroxene matrix is the volume difference betweenthe two phases, and the reliefofthe strains introduced bythe volume increase on inversion may account for therelative misorientations of the subgrains seen in Figurel0d.

Samples of Ca-rich bustamites from Broken Hill, NewSouth Wales, Australia, provided an opportunity to studythe initial stages of the reverse transformation, from bu-stamite to clinopyroxene. These bustamites have exsolvedclinopyroxene by a mechanism that involves solution andreprecipitation via fluid inclusions (Wilkins and Sverjen-sky, 1977); HRrEM shows that the bustamite matrix hasnumerous faults parallel to the (102) planes. Figure 1laclearly shows that these (102) faults have an associateddisplacement vector of Vz[0 l0], and low-resolution images(Fig. I lb) indicate that they are produced by the passageof dislocations through the bustamite matrix. The pres-ence of (102) faults, which are far more common in thesethan in any other natural samples studied, appears to berelated to their thermal history. Experimental annealingof these bustamites increased the fault density to the pointwhere streaking was apparent parallel to the (102) planenormal on the k-odd rows of electron-diffraction patterns(Fig. I lc). The faults therefore represent the first stage ofthe solid state rnechanism of the exsolution of clinopy-roxene from bustamite, being the creation of the wollas-tonite-like intermediate structure.

Pyroxenoid-bustamite (C - I-conplex)

Apart from the special case of the transformation be-tween wollastonite and bustamite, the only other trans-formation of this type that occurs is that between rho-donite and bustamite. The inversion of rhodonite tobustamite is extremely rare in natural material, the onlyevidence for this being a few occurrences of a two-rho-donite assemblage described by Abrecht et al. (1978). Innatural samples the transformation occurs far more com-monly in the opposite direction as the result of metaso-matic alteration ofbustamite during which Ca in the struc-ture is replaced by Mn, and more rarely bustamite mayunmix at low temperatures into an assemblage of clino-pyroxene plus rhodonite (Abrecht et al., 1978). However,the easiest way to study the transformation experimentallyis to invert rhodonites with compositions close to the Ca-rich limit of its stability field by annealing at temperaturesabove the miscibility gap between the two phases (Abrechtand Peters, 1980). Such experiments have been carriedout twice before in order to study the transformation byX-ray diffraction techniques and have been repeated

ANGEL: TRANSFORMATIONS BETWEEN SINGLE-CHAIN SILICATES 145 1

Fig. ll. (a) High-resolution rnicrograph of two (102) faults in a bustamite sample that has exsolved clinopyroxene and (b) asimilar set offaults at lower resolution, with terminating dislocations. (c) Selected-area electron-difraction pattem ofa [2] Il sectionof a bustamite heated in the stability field of clinopyroxene. Note the streaking of the k-odd layers parallel to t* (l 02).

here in order to study the transformation mechanism byHRTEM.

The study by Dent-Glasser and Glasser (1961) and theinterpretation of their results by Morimoto et al. (1966)have been shown to be mistaken, because of the conclu-sion of Dent-Glasser and Glasser ( 196 I ) that the inversionproduct was wollastonite rather than bustamite. Yaman-aka and Tak6uchi (1981) also studied this inversion byheating natural rhodonite, but their primary aim was toinvestigate the structure of the Mn-rich bustamite pro-duced by the inversion. Unfortunately they did not reportthe orientation of the product bustamite relative to therhodonite, but did note that the morphologies of the prod-uct compared to that of the starting material suggestedthat an oriented transformation had taken place. Yaman-aka and Tak6uchi (1981) proposed an inversion mecha-nism of cation movement within the (100) planes similarto that proposed by Morimoto et al. (1966), but as in thatpaper the figures ofYamanaka and Tak6uchi (1981) areof wollastonite rather than bustamite.

Similar annealing experiments were carried out in orderto study the inversion by nnrrr"r. The material used forthese experiments was a well-crystallized natural rhodo-nite that HRrEM showed to be free of chain-periodicityfaults, but that did contain a few (100) twins. A series ofannealing runs was carried out on this material, under Arto prevent oxidation, at temperatures believed to bewithin the bustamite stability field. The first stage of theinversion is the creation ofchain-periodicity faults in rho-donite that consist of three tetrahedra. i.e.. wollastonite-like W modules. Figure 12 shows a single chain-perio-dicity fault in a specimen slightly inclined from the I l0]axis, so that the fault lying parallel to (001) is slightlyinclined. Unfortunately no partially formed chain-period-icity faults were observed in these samples so the defects

that generated them can only be assumed to be identicalto those involved in the pyroxenoid-to-pyroxenoid in-versions. However, it should be noted that a single faultsuch as this cannot be produced by a purely diffirsivemechanism because the number of tetrahedra in each chainhas been increased within the stacking fault, whereas thisis possible by a shear-type mechanism. In longer annealingruns and those at higher temperatures, the quantity ofbustamite is greater, but no increase in the density of faultsin the residual rhodonite grains is observed. From this itmay be concluded that, as in the transformation fromclinopyroxene to bustamite, the rate-determining step isthe formation of narrow lamellae of wollastonite inter-mediate within the host phase. Once a small critical vol-ume of the intermediate structure has been formed, it actsas a nucleus for the rapid inversion of the remaining rho-donite matrix 10 bustamite.

Bustamites from the bustamite-rhodonite suite de-scribed by Abrecht et al. (1978) were examined by nnreiurfor evidence of the solid-state inversion mechanism ofbustamite to rhodonite, or of the unmixing of bustamiteto rhodonite plus clinopyroxene. In [211] sections ofthebustamite coherent lamellae were found that lie parallelto the (102) planes of the matrix (Fig. l3a). The associatedselected-area electron-diffraction pattern (Fig. I 3b) showsmaxima corresponding to bustamite, together with streak-ing parallel to the (102) plane normals on the ft-odd rows,indicating random Yz[010] displacements in the structure.In addition there are also weak maxima on these k-oddrows that cannot be attributed to bustamite, even intwinned orientations, but which are indexable in terms ofa wollastonite unit cell (Fig. l3c). The lamellae are there-fore wollastonite oriented with (101) parallel to (102) ofthe bustamite matrix, i.e. with the closest-packed planesof the two structures parallel. Note that the relative ori-

r452

Fig. 12. A single wollastonite-like chain-periodicity fault (W)parallel to (001) in a rhodonite specimen slightly inclined froml l 101.

entation is such that the wollastonite lamellae are in atwin orientation with respect to a wollastonite structurewith the same stacking sequence along [00] as the busta-mite matrix. In the T,G stacking vector notation oftenused to describe the wollastonite polytypes (Henmi et al.,1978), and which was extended to bustamite by Angel(1985), a (T) polytype of bustamite has (102) lamellaeconsisting of the (G) polytype of wollastonite. As a con-sequence of this twin relation, the b axis of the productwollastonite is parallel to -b of the matrix bustamite (Fig.l 3d).

Discussion

In the analysis of possible mechanisms for the trans-formations involving the bustamite structure, the exis-tence of an intermediate phase was postulated for struc-tural reasons. However, the need for an intermediatestructure can also be interpreted in kinetic terms. If thetransformation from bustamite to either clinopyroxene orrhodonite were to be carried out by a single-step mech-anism in the solid state, then this step would have to bethe direct nucleation and growth of the new phase withinthe bustamite matrix. The different arrangement of thetetrahedral chains within the structures would result in anincoherent, or at best semicoherent, interface between thematrix and the nuclei. This in turn would lead to a highactivation energy for the formation of such nuclei andthus a low overall rate for the transformation. By contrastthe two-step mechanism described in this paper has muchlower activation barriers and hence results in a faster rateof transformation. In forming the intermediate structurein a bustamite matrix, the first-neighbor coordinations ofall the atoms remain essentially the same, with only sec-ond- and higher-neighbor coordinations being changed.The interface between wollastonite lamellae and busta-mite matrix is therefore fully coherent, and the creationof this surface does not make any significant contributionto the activation energy of the transformation. Similarly,the second stage ofinversion from intermediate to productphase may proceed in a fully coherent manner. This is

ANGEL: TRANSFORMATIONS BETWEEN SINGLE-CHAIN SILICATES

-.,jLo-JLoo_loJ!_1_

.:::, .;il oo*L

1:; 1::, .,T :;

-

Fig. 13. (a) Wollastonite lamellae lying on the (102) planesof bustamite. (b) Selected-area electron-diffraction pattern of thespecimen in (a) and (c) the indexing of this pattern. Upper indicescorrespond to wollastonite, lower ones to bustamite. (d) Therelative orientations of the wollastonite and bustamite in realspace in projection on (001).

also the first stage of the reverse transformation (i.e., tobustamite) in which coherent lamellae of wollastonite areformed within either clinopyroxene (Fig. 8) or rhodonite(Fig. l2). The formation of the wollastonite intermediateis therefore an example of the operation of the Ostwaldstep rule: In any phase transformation or reaction thekinetically most favorable sequence of structures will form,rather than those involving the greatest possible reductionin free energy. In the case of solid-state transformations,this is equivalent to the statement that transformationsproceed via the structures that are most easily formedwithin the matrix of the starting material. Quite clearlythis principle could be employed to suggest possible in-termediate phases in other mineral transformations.

CoNcr,usroNs

In this paper the possible transformations between thesingle-chain silicates in the (Ca,Mn,Fe,Mg)SiO3 systemhave been distinguished by the change in lattice type ac-companying each transformation. This change corre-sponds to the re-arrangement of the main structural com-ponent of these structures, the silicate chains. Theobservations made by HRTEM show that the transforma-tions classifiedas C - Cand 1 - C(pyroxenoid-pyrox-enoid and clinopyroxene-pyroxenoid) proceed by the pas-sage of line defects through the structure of the startingmaterial to create the product phase. When a similarmechanism was considered for the two transformationsinvolving the bustamite structure, it was shown that asingle-step mechanism of this kind was not capable ofreproducing the required lattice type, i.e., the correct dis-tribution of the silicate chains. Careful re-examination ofthe alternative diffusive topotactic mechanisms previ-ously proposed for these transformations (Morimoto et

al., 1966) showed that these too create the wollastonitestructure rather than that ofbustamite. Both ofthese typesof mechanism were shown to therefore require an inter-mediate phase in order to carry out the transformations.This phase was confirmed to be wollastonite by unrervrobservations. Its formation can be interpreted as a con-sequence of the Ostwald step rule, which indicates thatkinetic factors may be important in determining the mech-anism of these transformations.

This study has also demonstrated the difficulties in-volved in characterizing solid-state transformation mech-anisms in minerals, even with the use of electron mi-croscopy. These problems mainly arise from the transientnature of the partially transformed states and the need todistinguish such microstructures from other artifacts, suchas growh faults in the starting material and faults inducedby specimen preparation. The latter two problems wereavoided in this study by the careful characteization ofstarting materials by Hnrer'r and by preparing specimensfor electron microscopy by both crushing and ion-beamthinning. It is unlikely that these two very different prep-aration methods would introduce the same types of faultsinto the specimens, and comparison of observations madeon samples prepared by both methods did allow somespecific artifacts of the crushing of samples to be identifiedas such.

However, the assumption remains that the quenched-in faults observed in samples recovered from a series ofisothermal annealing experiments do represent the com-plete sequence oftransformation states. Indeed this studyprovides examples in that although the wollastoniteJikeintermediate structure has been observed, the rnechanismby which it is transformed to and from clinopyroxene andrhodonite was not directly observed. Rather it was de-duced that these transitions were likely to proceed by thepropagation of line defects. Similarly, because of the ap-parently rapid transformation rates, no transition was ob-served between the microstructures shown in Figures l0cand l0d. Until it is possible to study such transformationsin situ in the electron microscope at high resolution, theresults of a study such as this may be subject to reinter-pretation.

Acrhlowr,nDcMENTS

Thanks are due to Dr. Andrew Putnis for his supervision ofthis project, and to him, Dr. Mike Bown, and Dr. David Pricefor their helpful comments on draft versions of this paper. Thereview by Dr. Ian McKinnon contributed much to the organi-zation of the material presented here. Natural specimens werekindly donated by Professor T. G. Vallance of the University ofSydney Geology Department, Dr. J. Abrecht ofthe MineralogicalInstitute of Basel University, and Dr. G. Chinner, curator of theHarker collection of the Earth Sciences Department, Universityof Cambridge. This work was carried out with the support of aresearch studentship from the Natural Environment ResearchCouncil of Great Britain.

RnrnnnNcns

Abrecht, J., and Peters, Tj. (1980) The miscibility gap betweenrhodonite and bustamite along the join MnSiOr-CaouMnooSiO,.Contributions to Mineralogy and Petrology, 74,261-269.

I 453

Abrecht, J., Peters, Tj., and Sommerauer, J. (1978) Manganifer-ous mineral assemblages of Ravinella di Sotto, Valle Strona,Italy. Memorie di Scienze Geologische, 33,215-221.

Aikawa, N. (1979) Oriented intergrowth of rhodonite and py-roxmangite and their transformation mechanism. Mineralog-ical Journal ofJapan, 9,255-269.

- (1984) Lamellar structure of rhodonite and pyroxmangiteintergrowths. American Mineralogist, 69, 27 0-27 6.

Amelinckx, S., and Van Landuyt, J. (1976) Contrast effects atplanar interfaces. In H.-R. Wenk, Ed. Electron microscopy inmineralogy. 68-1 12. Springer-Verlag, Berlin.

Angel, R.J. (1984) The determination of the johannsenite-busta-mite equilibrium inversion boundary. Contributions to Min-eralogy and Petrology, 85, 272-27 8.

- (1985) Structural variation in wollastonite and bustamite.Mineralogical Magazine, 49, 37-48.

Angel, R.J., Price, G.D., and Putnis, A. (1984) A mechanism forpyroxene-pyroxenoid and pyroxenoid-pyroxenoid transfor-mations. Physics and Chemistry of Minerals, 10,236-243.

Czank, M., and Liebau, F. (1980) Periodicity faults in chainsilicates: A new type ofplanar lattice fault observed with nnrrvr.Physics and Chemistry of Minerals, 6, 85-93.

Czank, M., and Simons, B. (1983) High resolution microscopestudies on ferrosilite III. Physics and Chemistry of Minerals,9 ,229-234.

Dent-Glasser, L.S., and Glasser, F.P. (1961) Silicate transfor-mations: Rhodonite-wollastonite. Acta Crystallographica, 14,8 l 8-822.

Henmi, C., Kusachi, I., Kawahara, A., and Henmi, K. (1978) 7Twollastonite from Fuka, Okayama Prefecture. MineralogicalJournal ofJapan, 9, 169-181.

Jeferson, D.A., and Pugh, N.J. (1981) The ultrastructure of py-roxenoid chain silicates, IlL lntersecting defects in a syntheticiron-manganese pyroxenoid. Acta Crystallographica, A37, 28 l-286.

Jefferson, D.A., Pugh, N.J., Alario-Franco, M., Mallinson, L.G.,Millward, G.R., and Thomas, J.M. (1980) Ultrastructure ofpyroxenoid chain silicates, I. Variation ofchain configurationin rhodonite. Acta Crystallographica, A36, 1058-1065.

Koto, K., Morimoto, N., and Narita, H. (1976) Crystallographicrelationships ofthe pyroxenes and pyroxenoids. Japanese As-sociation of Mineralogists, Petrologists and Economic Geol-oglsts Journal, 7 l, 248J54.

Kronberg, M.L. (1957) Plastic deformation of single crystals ofsapphire: Basal slip and twinning. Acta Metallographica, 5,507-524.

Lacam, A., Madon, M., and Poirier, J.P. (1980) Olivine glassand spinel formed in a laser heated, diamond anvil high pres-sure cell: An investigation by transmission electron micros-copy. Nature, 288, 155-157.

Morimoto, N., Koto, K., and Shinohara, T. (1966) Orientedtransformation of johannsenite to bustamite. MineralogicalJournal of Japan, 5, 44-64.

O'Keefe, M.A., and Buseck, P.R. (1979) Computation of highresolution rru images of minerals. American CrystallographicAssociation Transactions, 15, 27 -46.

Pannhorst, W. (1979) Structural relationships between pyrox-enes. Neues Jahrbuch fiir Mineralogie Abhandlungen, 135, l-1 7 .

Peacor, D.R., and Prewitt, C.T. (1963) Comparison of the crystalstructures of bustamite and wollastonite. American Mineral-ogist, 48, 588-596.

Poirier, J.P. (1981) On the kinetics ofthe olivine-spinel transi-tion. Physics of the Earth and Planetary Interiors, 26, 179-l 87 .

Prewitt, C.T., and Peacor, D.R. (1964) Crystal chemistry of thepyroxenes and pyroxenoids. Arnerican Mineralogist, 49, 1527 -

r542.Price, G.D., Purnis, A., and Smith, D.G.w. (1982) A spinel to

g-phase transformation mechanism in (Mg,Fe)rSiOn. Nature,296,729-73r.

ANGEL: TRANSFORMATIONS BETWEEN SINGLE-CHAIN SILICATES

t454

Putnis, A., and Price, G.D. (1979) High pressure (Mg,Fe).SiO.phases in the Tenham chondritic meteorite. Nature, 280,217-2r8.

Ried, H. (1984) Intergrowth ofpyroxene and pyroxenoid: Chainperiodicity faults in pyroxene. Physics and Chemistry of Min-erals. 10.230-235.

Ried, H., and Korekawa, M. (1980) Transmission electron mi-croscopy of synthetic and natural funferketten and siebener-ketten pyroxenoids. Physics and Chemistry of Minerals, 5,35 l-365.

Simons, B., and Woermann, E. (1982) Pyroxene-pyroxenoidtransformations under high pressure. In W. Schreyer, Ed. Highpressure researches in geoscience, 529-536. E. Schweizer-bart'sche Verlagbuchhandlung, Stuttgart.

Taylor, H.F.W. (1960) Aspects of the crystal structures ofcalcium

ANGEL: TRANSFORMATIONS BETWEEN SINGLE-CHAIN SILICATES

silicates and aluminates. Journal of Applied Chemistry, 10,3r7-323.

Veblen, D.R. (1985) rEM study ofa pyroxene-to-pyroxenoid re-action. American Mineralogist, 70, 885-901.

Wilkins, R.W.T., and Sverjensky, D.A. (1977) The role of fluidinclusions in the exsolution ofclinopyroxene in bustamite fromBroken Hill, New South Wales, Australia. American Miner-alogyst, 62, 465-47 4.

Yamanaka, T., and Takduchi, Y. (1981) X-ray study of the rho-donite-bustamite transformation. Zeitshrift fiir Kristallogra-ph ie , 157, 131-145.

Mexuscnrsr RECETvED SEprnMsen 20, 1985MnNuscnrsr AccEPTED Jurv 8, 1986