J. metamorphic Geol.

, 2007, 25, 439450

doi:10.1111/j.1525-1314.2007.00704.x

When epitaxy controls garnet growth

R. SPIESS,1 C. GROPPO2 AND R. COMPAGNONI2 1 Department of Geoscience, University of Padova, Via Giotto 1, I-35137 Padova, Italy (richard.spiess@unipd.it) 2 Department of Mineralogical and Petrological Sciences, via Valperga Caluso 35, I-10125, Torino, Italy

ABSTRACT

Within a mica schist from the coesite-bearing Brossasco-Isasca Unit (Western Alps), microstructural analysis shows that Alpine garnet grains are aligned with the crenulated foliation. Garnet crystallographic orientation was analysed with electron backscatter diffraction (EBSD): the obtained crystallographic dispersion patterns and distribution patterns of misorientation axes suggest a strong parallelism of {110} garnet planes with a 56W-dipping foliation. The data are interpreted as evidence for an epitaxial growth of garnet upon (001) biotite planes, sometime during and/or after dispersion of the biotite/garnet crystals from their initially foliation-parallel orientation by rotation about the Alpine crenulation axis. This interpretation is based on the comparison of the measured EBSD data with: (i) theoretical dispersion trajectories of garnet crystallographic data, (ii) numerically modelled pole gures, and (iii) numerically modelled misorientation axis distribution patterns. Our data suggest that epitaxial growth of garnet upon biotite is allowed by distortion of the pseudohexagonal basal oxygen ring structure on (001) biotite surfaces, and that distortion is driven by introduction of missing ions. Our data further suggest that the spatial distribution of precursor phases inuences the distribution patterns of garnet within mica schists. Key words: biotite; electron backscatter diffraction; epitaxial growth; garnet; misorientation analysis.

INTRODUCTION

Nucleation and growth processes of metamorphic phases can be described by physical laws involving measurable entities/parameters such as the spatial distribution and compositional zoning of grains, the radius of depletion zones surrounding them and their grain size. Systematic variations of these entities constrain the limits within which one nucleation and growth process prevails over another. Kretz (1969, 1974) with his pioneering work has attempted to understand garnet nucleation and growth mechanisms in such a context. Carlson (1989, 1991) set landmarks in the numerical analysis and comprehension of garnet nucleation and growth processes using the novel approach of computed X-ray tomography combined with appropriate statistical methods (Denison et al., 1997; Hirsch et al., 2000; Ketcham et al., 2005) to determine garnet crystal size and spatial distribution. In the nucleation and growth model of Carlson (1989), the spatial distribution of garnet crystals is ordered: the order is linked to intergranular diffusion as the rate limiting process in crystal growth. Two quantitatively perceptible microstructural features will characterize rocks where garnet growth is controlled by this process: growing porphyroblasts will be surrounded by depletion zones wherein further nucleation is depressed, and clustered grains will be small compared with isolated grains, because they have to compete for nutrients.
2007 Blackwell Publishing Ltd

A critical assumption in the model of Carlson (1989, 1991) and Denison et al. (1997) is that the spatial distribution of crystals is not inuenced by factors other than intergranular diffusion, excluding deformation or preferential growth from precursor phases as potential controls. It is therefore essential to assess whether factors other than intergranular diffusion have inuenced the spatial distribution of garnet. A powerful approach to reveal this is the analysis by electron backscatter diffraction (EBSD), because it has the potential to analyse at a statistically signicant level whether the crystallographic orientation of garnet grains relative to potential precursor phases is ordered, allowing the assumption that nucleation and growth of garnet was controlled by these precursor phases. This study aims to test this through the analysis of garnet grains nucleated within an ultrahigh pressure (UHP) mica schist from the Western Alps. The natural data and the numerical simulations suggest that epitaxial growth upon biotite has controlled the spatial distribution and crystallographic orientation of garnet in the studied sample.
GEOLOGICAL SETTING

The studied garnet mica schist was collected within the NW part of the coesite-bearing Brossasco-Isasca Unit (BIU) (Fig. 1), which is a coherent slice of Penninic continental crust (southern Dora-Maira Massif), tectonically sandwiched between two units which experienced a quartz-eclogite facies recrystallization
439

440 R. SPIESS ET AL.

(a)

(b)

Fig. 1. (a) Simplied tectonic sketch-map of the Western Alps. Helvetic-Dauphinois Domain: Mont Blanc-Aiguilles-Rouges (MB); Penninic Domain: Grand St Bernard Zone (SB), and Monte Rosa (MR), Gran Paradiso (GP), Dora-Maira (DM) and Valosio (V) Internal Crystalline Massifs; Piemonte Zone of Calcschists (light grey) with meta-Ophiolites (dark grey); Austroalpine Domain: Dent Blanche Nappe (DB), Monte Emilius Klippe (ME), and Sesia-Lanzo Zone (SZ); Southern Alps (SA); Embrunais-Ubaye Flysch Nappe (EU); Penninic Thrust Front (PF); Sestri-Voltaggio Line (SVL). (b) Enlargement of the southern Dora-Maira Massif. Tectonic units of the Massif are listed from the structurally lowermost Pinerolo Unit to the structurally highest Dronero-Sampeyre Unit.

(550 C and 1.5 GPa, Chopin et al., 1991; Schertl et al., 1991; Sharp et al., 1993; Compagnoni et al., 1995; Matsumoto & Hirajima, 2000; Compagnoni & Rolfo, 2003; Hermann, 2003). The BIU, originally composed of a Variscan amphibolite facies metamorphic basement (Compagnoni et al., 1995), was intruded at 275 Ma by porphyritic granitoids (Gebauer et al., 1997), as shown by local preservation of relict pre-Alpine igneous and metamorphic structures (Compagnoni et al., 1995; Compagnoni & Rolfo, 2003). During the Alpine orogeny, the BIU experienced an early Alpine coesite-eclogite- (Gebauer et al., 1997; Rubatto & Hermann, 2001), and a late Alpine greenschist facies recrystallization so that the Variscan amphibolite facies basement plus the late-Variscan granitoids were converted to the present Polymetamorphic Complex and Monometamorphic Complex, respectively (Compagnoni et al., 1995). The Polymetamorphic Complex, from which the sample was collected, mainly consists of paragneiss and paraschist with marble and eclogite intercalations (Chopin et al., 1991; Compagnoni et al., 1995; Compagnoni & Rolfo, 2003). The paraschists contain rare

microstructural and/or mineralogical relics of the preAlpine amphibolite facies regional metamorphism, such as garnet, red-brown biotite partially replaced by phengite rutile, K-feldspar partly replaced by phengite-quartz intergrowths and prismatic sillimanite pseudomorphically replaced by kyanite aggregates (Compagnoni et al., 1995; Compagnoni & Rolfo, 2003). Ultrahigh pressure metamorphic peak conditions for the BIU were 750 30 C and 3.3 0.3 GPa, i.e. below the diamond stability eld (Chopin, 1984, 1987; Chopin et al., 1991; Kienast et al., 1991; Schertl et al., 1991; Sharp et al., 1993; Compagnoni et al., 1995). However, higher peak-pressure conditions have been obtained by the most recent thermodynamic modelling by Groppo et al. (2006b), which corroborate the hypothesis of Hermann (2003) based on experimental results that peak metamorphic conditions were well within the diamond stability eld at 730 C and 4.3 GPa. The retrogressive metamorphic evolution of the BIU is characterized by a substantial decompression coupled with a continuous cooling (Chopin et al., 1991;
2007 Blackwell Publishing Ltd

WHEN EPITAXY CONTROLS GARNET GROWTH 441

(a)

(b)

(c)

Fig. 2. (a) Photomicrograph (crossed nicols) of studied thin section showing a relict foliation preserved within the strain shadows of the large Variscan garnet porphyroblast, overprinted to an EW oriented Alpine differentiated crenulation cleavage outside. Open crenulations are preserved within the western strain shadow, whereas in the eastern strain shadow the foliation curves smoothly. Tails of blocky quartz have a left-hand asymmetry. Small Alpine garnet grains are aligned with the oblique to crenulated foliation in the strain shadows, whereas outside they may be enveloped by the differentiated crenulation cleavage. (b) Rectangle in (a) locates enlargement. Enlargement under plane light shows small Alpine garnet grains overgrowing the crenulations preserved within the strain shadow. The axial plane of the open crenulations is WNWSSE. (c) Sketch showing the main microstructural features of the mica schist. Crenulated relict foliation is SV. Alpine foliation is SA. Quartz is light grey, pre-Alpine garnet is grey, Alpine garnet overgrowing the large porphyroblast in a corona adjacent to the western strain shadow is dark grey. Black full dots show the analysed small Alpine garnet crystals (1 indexed EBSD pattern per grain). Matrix of the mica schist (white) mainly consists of quartz and phengite.

Schertl et al., 1991; Compagnoni et al., 1995; Compagnoni & Rolfo, 2003; Groppo et al., 2006, 2007). At relatively low pressures (0.5 GPa), this cooling path is followed by a moderate heating up to the boundary between upper greenschist- and amphibolite facies (Compagnoni et al., 1995; Rubatto & Hermann, 2001). Overall, the Alpine metamorphic evolution of the BIU is dened by a clockwise PT trajectory, similar to the PT path of other tectonic units from the Inner Western Alps.
MICROSTRUCTURE

The studied metapelite is a UHP garnet + kyanite + quartz/(coesite) + (jadeite) + high-Si phengite mica schist with accessory rutile and apatite (former occurrence of minerals inside brackets is inferred from microstructural and mineralogical evidence). In the discussed thin section, the microstructure is dominated by an EW trending, closely spaced Alpine crenulation
2007 Blackwell Publishing Ltd

cleavage that wraps around a large pre-Alpine garnet porphyroblast (Grt1) (Fig. 2a,c). Adjacent to the garnet porphyroblast, strain shadows with a left hand asymmetry contain tails of blocky quartz that are parallelized with the enveloping Alpine crenulation cleavage (SA in Fig. 2c). In the remaining parts of the strain shadows a relict, NS oriented foliation (SV in Fig. 2c) is preserved that rotates progressively towards the enveloping Alpine foliation in the strain shadow to the east, whereas it is deformed to a widely spaced differentiated crenulation cleavage (Bell & Rubenach, 1983) in the strain shadow to the west. Along the central left margin of the porphyroblast, a corona of small Alpine garnet grains (Grt2) overgrows a microstructure that resembles a strain cap of biotite (Passchier & Trouw, 1998, p. 148). It is important to stress that no biotite crystals are left in the whole analysed mica schist. However, many Alpine garnet grains are crowded with rutile needles that occasionally mimic the outlines of pseudo-hexagonal

442 R. SPIESS ET AL.

(a)

(b)

(c)

(d)

Fig. 3. (a) Central left margin of the pre-Alpine garnet porphyroblast (Grt1), 2 cm in diameter, with a corona of small Alpine garnet crystals (Grt2). (bd) Details of Alpine garnet crystals (Grt2), the cores of which are crowded with rutile needles, suggesting their growth at the expense of a former pre-Alpine Ti-bearing biotite. Rutile needles dene the outline of pseudo-hexagonal forms (b, c) and tabular forms (d) of earlier biotite crystals. Mineral abbreviations according to Bucher & Frey (2002).

(Fig. 3b,c) or tabular-shaped crystal forms (Fig. 3d). As Ti-solubility in garnet at 4.3 GPa and 730 C is low (Zhang et al., 2003), we interpret these rutile needle inclusions as evidence for Grt2 formation at the expense of pre-Alpine Ti-rich biotite broken down during HP to UHP metamorphism. The differentiated Alpine foliation is made up by discontinuous layers of high-celadonitic phengite (partially replaced by ner grained, less-celadonitic phengite) alternating with medium-grained lens-like quartz-rich domains. Fine-grained staurolite and greenish-blue chloritoid idioblasts statically overgrow this main foliation. A peculiar feature of the mica schist is the nucleation and growth of hundreds of small Alpine garnet grains (Grt2) ranging in apparent (2D) diameter from some tens of microns up to some hundreds of microns within the matrix. In the western strain shadow many of these garnet grains are arranged according to the open crenulations (Fig. 2b), whereas in the matrix outside the strain shadow, garnet

grains may be enveloped by ne seamed foliation domains or overgrow them. Microstructural data suggest that Grt2 nucleated during the high pressure Alpine metamorphism preferentially after the mica domains of the crenulated relict foliation (Fig. 2a,b). Because of the crenulation development, mica akes and garnet grains nucleated upon them would have been rotated away from their original orientation. The microstructural analysis is unable to dene how much garnet has been rotated after overgrowing mica akes and how much mica akes have been rotated before garnet nucleation. The microstructural analysis cannot unequivocally reveal the timing of garnet growth upon mica akes relative to the Alpine crenulation cleavage development. The open crenulations within the strain shadow could have been overgrown by garnet when in the matrix a strongly differentiated crenulation cleavage already had formed. In fact, in the matrix, garnet becomes enveloped by, and overgrows, the differentiated
2007 Blackwell Publishing Ltd

WHEN EPITAXY CONTROLS GARNET GROWTH 443

crenulation cleavage, suggesting that nucleation and growth of garnet progressed through time.
EXPERIMENTAL PROCEDURES

For every Alpine garnet grain one EBSD pattern was collected and indexed. The total number of analyses was 666. Of these, 391 were from garnet grains of the western strain shadow, whereas 275 were from garnet crystals within the foliation domains outside the strain shadow. Fifty analyses of 391 came from the overgrown strain cap. Because the orientation of these last 50 garnet crystals may have been controlled by the crystallographic orientation of the large garnet porphyroblast (both types have similar orientations), they are omitted then from the pole gures. Therefore, the overall number of processed EBSD patterns is 616. The thin section was polished using Syton uid to remove the mechanical damage generated during previous mechanical polishing (Flynn & Powell, 1979; Prior et al., 1996). After polishing, the sample was coated with a thin carbon lm to prevent electrical charging problems. EBSD analysis was performed using a CamScan MX2500 SEM equipped with a tungsten lament at the Department of Mineralogy and Petrology, the University of Padova (Italy). An acceleration voltage of 25 kV, a lament emission current of 150 lA, and a working distance of 25 mm were used. EBSD patterns were indexed using CHANNEL 5.0 software from HKL-Technology. Indexing was accepted when at least ve detected Kikuchi bands corresponded with those contained in the standard reector le for garnet. Collected data were processed using the Mambo component of the CHANNEL 5.0 software. For the numerical simulation of garnet dispersion, data les were constructed and imported to the Project Manager component of CHANNEL 5.0 for plotting with Mambo. Forty degree concentric dispersion was simulated numerically, whereas 20 lateral dispersion within the contoured pole gures was simulated by means of the half width angle during contouring. For the positioning of the rotation axis the CS0 > CS1 facility of the Virtual Chamber component of CHANNEL 5.0 was used.
DESCRIPTION OF {100}, {111} AND {110} POLE FIGURES

(a)

(b)

(c)

Fig. 4. Distribution patterns of poles to the {100} (a), {111} (b) and {110} (c) crystallographic planes of Alpine garnet crystals. The pole gures are spatially oriented relative to the thin section shown in Fig. 2, with the relict foliation running NS and the Alpine crenulation cleavage running EW.

The {111} pole gure (Fig. 4b) is dened by two main characteristics: (i) a maximum of {111} poles coincides with the minimum of {100} poles, and (ii) a girdle of {111} poles extends from NNE towards SSW (within the stereonet frame of reference) and is closely centred on the {100} minimum. Outside these areas the clustering density of {111} poles varies constantly between 0.7 and 0.9 mud. Compared with the {100} and {111} pole gures, the {110} pole gure (Fig. 4c) has only very subtle density differences. The reason for this is the high number of equivalent faces belonging to this family (six faces plotting in the same hemisphere), so that the general tendency is a constant density of 0.9 mud throughout the pole gure. Nevertheless, a small density increase of {110} poles is observed in correspondence with the {100} minimum and within an irregularly shaped girdle centred slightly off the {100} minimum. Inside this girdle, a maximum with the highest density of 1.33 mud is positioned at 065/30.
INTERPRETATION OF POLE FIGURES

The {100} poles (Fig. 4a) show a weak but clearly perceptible preferred orientation, that denes a concentric girdle with a pole clustering density >1 mud (multiples of uniform density) centred on a minimum with a clustering density of 0.55 mud. Point maxima of up to 1.45 mud exist within the girdle, whereas outside the girdle the density is fairly constant at 0.80.9 mud. The two most signicant point maxima are positioned at 090/80 and 010/56, whereas the minimum, on which the girdle is centred, is placed at 280/56.
2007 Blackwell Publishing Ltd

The concentric small girdle dispersion of {100} poles about a {100} pole-minimum, the coincidence of this minimum with a {111} pole-maximum, plus the girdlelike distribution of {111} poles (and less clearly of the {110} poles) suggest that these crystallographic distribution patterns have not formed coincidentally, but reect the interaction of specic processes that have operated during garnet nucleation and growth. It can be inferred that the {100} minimum/{111} maximum coincides with a rotation axis, and that the arrangement of {100} poles around this minimum inside a broad small girdle with an opening angle of 70110 reects the dispersion of {100} poles that initially had a crystallographic preferred orientation. If this preferred orientation were a unique {100} orientation, because of the 90 interrelationship between (100), (010) and (001) poles, the opening angle of the small girdle would be 90. Therefore, more than just a single {100} orientation is required to explain the actual {100} pole arrangement. These orientations

444 R. SPIESS ET AL.

(a)

(c)

(b)
Fig. 5. Drawing of the garnet-bearing mica schist, showing the spatial arrangement of the relict foliation (NS), Alpine crenulation cleavage (EW) and trend of crenulation axis. Arrow points to the N, crenulation axis plunges 280/56. Rectangle shows studied thin section.

(d)

must have been related to each other in such a way that part of the {111} poles were allowed to align closely with the rotation axis, whereas the rest fall in distinct positions within the broad great girdle closely centred on the {100}-minimum. In order to understand the evolution of this dispersion pattern it is important to assess the signicance of the rotation axis. As the crenulation axis (Fig. 5) of the analysed mica schist is spatially correlated with this rotation axis, it is most reasonable that it is a kinematic axis. The observation that in some cases garnet preserves rutile inclusions tracking the outlines of pseudohexagonal biotite planes (Fig. 3bd) could suggest that the preferred crystallographic orientations of garnet before dispersion have been controlled by the nucleation site, i.e. the crystallography of biotite (and possibly white mica) contained within the differentiated crenulation cleavage. Considering the spatial coincidence of the crenulation and rotation axes, we hypothesize that the observed dispersion pattern of crystallographic orientations results from rotation of foliation-parallel mica planes upon which garnet was nucleating. Frondel (1940) observed that garnet commonly grows with one {110} face parallel to the (001) of white mica, forming angles of 0, 30, 60 or 90 between the {001} garnet traces and the (100) muscovite trace. Analogous observations have been reported by Powell (1966). He studied inclusion bands and trails of micas in Moinian garnet crystals that were approximately parallel to a dodecahedral garnet plane. Based on his own and Frondels observations, Powell (1966) con-

Fig. 6. Comparison of the oxygen conguration of garnet adjacent to {110} planes (a, b), and of biotite on (001) planes (c, d). (a, b) Oxygen is red and Al is green, (c) open green circles are basal oxygen atoms, (d) open red circles are apical oxygen atoms. Blue full dots in panels c and d are K+ ions. The [001] direction of garnet is NS in panel a and EW in panel b. The (100) trace of mica is NS in panel c and d. Arrows in the central pseudohexagonal oxygen rings in panels c and d show in-plane sixfold symmetry directions. After rotation and distortion, hexagons in panel c formed by O2) and K+ ions could match the oxygen structure of garnet in panel a (supposed that introduction of O2) in positions of K+ is possible), whereas hexagons designed by apical oxygen in panel d could potentially t the oxygen structure of garnet in panel b.

cluded that garnet may grow epitaxially upon mica. He attributed this to a close relationship between the oxygen atom arrangement adjacent to the K-interlayer in mica (considering the conguration of apical oxygen) with that of garnet on {110} planes, so that Al atoms could possibly be accommodated in the potassium positions in a pattern very similar to that shown by Al on the {110} of garnet. We have compared the oxygen distributions in biotite adjacent to the K-interlayer with that of garnet adjacent to {110} planes using CRYSTALMAKER 1.3.5 Demonstration version plus the crystal structure data les of Brigatti & Davoli (1990) and Quartieri et al. (1995) available in the American Mineralogist Crystal Structure Database (Downs & Hall-Wallace, 2003). From this comparison shown in Fig. 6 we deduce that: 1 the biotite oxygen ring structure on the (001) nucleation surface needs rotation and distortion to t the oxygen structure on the {110} garnet plane; 2 a systematic variation of the [001] garnet directions by multiples of 30 relative to the (100) mica trace requires nucleation to occur with a similar probability in all six-fold directions on the pseudohexagonal basal and the pseudohexagonal apical oxygen rings, as they are rotated relative to each other by 30 (Fig. 6c,d).
2007 Blackwell Publishing Ltd

WHEN EPITAXY CONTROLS GARNET GROWTH 445

(a)

clockwise and/or an anticlockwise sense about the crenulation axis. If the crenulations have a cylindrical geometry, then the crystallographic orientations of the garnet crystals will describe concentric dispersion paths about the rotation axis. The opening angles of the dispersion paths (Fig. 8) will vary as a function of the crystallographic directions taken into consideration, i.e. {001}, {111}, {110}, and as a function of the original garnet crystal orientations, i.e. 0, 30, 60, or 90 relative to the (100) trace.
GEOMETRICAL SIMULATION OF DISPERSION TRAJECTORIES

(b)

(c)

(d)

(e)

(f)

(g)

Fig. 7. (a) Garnet growing with a {110} plane upon (001) planes of biotite. The [001] traces of garnet, shown as dashed lines, describe angles of 0, 30, 60 and 90 degree with the NS oriented (100) trace of biotite. The rotation axis relating the different garnet growth positions to another is [110]. The {110} poles of the variously rotated garnet crystals have different colours. The same colours are used in the pole gures shown in panels bg for the crystallographic orientations of garnet crystals in different growth positions. (b) Poles to {100} planes of garnet (equal area projection in lower hemisphere) grown with a {110} plane onto a (001) of biotite. The orientation of (001) biotite plane is vertical and NS oriented. (c) Poles to {111} planes of garnet. (d) Poles to {110} planes of garnet. (eg) Same projections as panels bd but with the (001) biotite plane dipping 56 towards W (asterisk is pole to biotite plane). The trace of the (100) biotite plane onto the (001) plane is NS oriented.

If garnet nuclei can arise by distortion of the biotite lattice in this way, then it is possible to predict that garnet would show preferred crystallographic orientations relative to the (100) mica trace, with the [001] garnet direction forming angles of 0, 30, 60, or 90. The above hypothesis forms the background for the geometrical and the numerical simulations we present in this paper to interpret the formation of the garnet crystallographic dispersion pattern observed in the analysed mica schist. For the scope of simulation it is assumed that (001) mica planes with (100) traces running NS are aligned parallel to the 56W dipping relict foliation (Fig. 7bg). During the Alpine deformation, this foliation becomes crenulated about a 280/ 56 plunging axis, and the mica akes together with the grown garnet crystals become rotated up to 40 in a
2007 Blackwell Publishing Ltd

Let us consider that garnet grew with a {110} face upon the (001) face of biotite (parallel to the paper sheet in Fig. 7a). Rotation of garnet by 30 or multiples of it about the horizontal [110] axis (normal to the paper sheet in Fig. 7a) results in angles of 0, 30, 60, 90 between the [001] directions of garnet (dashed lines in Fig. 7a) and the (100) traces of biotite (vertical in Fig. 7a). Accordingly, plotting the poles to the {001}, {111} and the {110} planes of garnet in equal area projections give the pole-gures shown in Fig. 7bd, respectively [primitive great circle in equal area plots is (100) of biotite, NS great circle is (001) of biotite, garnet [110] rotation axis is EW]. Tilting the {001} biotite plane by 56 towards W will cause this plane to intersect with the Alpine foliation (dipping 80 towards N) along the crenulation axis preserved within the analysed mica schist (see Fig. 7eg and simulation in Fig. 8). After tilting, if we consider only garnet positions of 0, 30 and +60, then no {100} poles will coincide with the {100}-minimum, whereas all other {100} poles occupy distinct positions within the broad small girdle (Fig. 8a) along which natural {100}-data disperse (Fig. 8a). Additionally, part of the {111} poles overlap with, or fall close to, the crenulation axis (Fig. 8b), while the rest occupies distinct positions within the large great girdle in Fig. 8b along which natural {111}-data disperse. Poles to {110} planes (Fig. 8c) are arranged along a small girdle and a great girdle centred on the crenulation axis. A maximum of {110} poles coincides obviously with the tilted [110] rotation axis that plunges 34/090 (Fig. 8c). In the geometrical simulations of Fig. 8 we show the results of folding the foliation parallel biotite akes (upon which garnet grew) around the observed crenulation axis from the postulated starting position (Fig. 7fh). If the crenulation geometry is cylindrical, then with increasing deformation, the {100}, {111} and {110} poles must disperse away from the starting position along the concentric dispersion trajectories centred on the rotation/crenulation axis. Asymmetric crenulations will result in a preferential clockwise or anticlockwise dispersion depending on the sense of shear: crenulations of coaxial domains will disperse symmetrically, and increasing strain will result in larger rotations.

446 R. SPIESS ET AL.

(a)

(b)

(c)

Fig. 8. Theoretical dispersion trajectories of {100} (a), {111} (b) and {110} (c) poles about a rotation axis plunging 280/56 (equal area projection, lower hemisphere). Trajectories are designed for 40 clockwise and 40 anticlockwise rotation. Dispersion of poles increases from hypothetical original growth positions (0,+30, +60, )30, compare Fig. 7) with increasing distance from the rotation axis position. Great circles are relict and Alpine foliations, intersection is rotation axis. Light grey colour denes areas with a pole clustering density >1 mud, white colour denes areas with a pole clustering density <1 mud (see Fig. 4 for comparison).

Our geometrical simulation shows clockwise and anticlockwise trajectories of rotations up to 40. These are probably maximum values for coaxial and reasonable values for non-coaxial deformation domains. A comparison of the hypothesized dispersion patterns in Fig. 8 with those observed in nature shows that many of the characteristics of the natural pole gure data are predicted by the geometrical simulation. We have therefore decided to simulate numerically pole gures reecting a situation where garnet crystals nucleate on biotite according to the orientations dened above, and become variously dispersed from their original crystallographic position by rotation about a crenulation axis that plunges 280/56.
NUMERICAL SIMULATION OF POLE FIGURES

The numerical simulation is based on the assumption that garnet nucleates with a {110} plane upon (001) of biotite, and the [001] garnet direction forms angles of )30, 0, +30, and +60 with the (100) trace of biotite. Garnet grains become rotated about a rotation axis plunging 280/56, corresponding to the observed orientation of the crenulation axis. In order to simulate symmetrical crenulations, garnet crystals are allowed to rotate both clockwise and anticlockwise up to 40. It is assumed that folding was cylindrical, so that there is a constant pole density along any 040 dispersion range. Higher densities in the simulations are consequently either due to a concentration of poles that were initially situated closer to the rotation axis (and became therefore less dispersed), or because dispersion has caused poles to concentrate in certain areas. A comparison between the natural pole gure (Fig. 9ac) and the simulated pole gure (Fig. 9df) shows that most of the characteristic features of both pole-gures coincide. The concentric small girdle dispersion of {100} poles together with the concentration of poles

along the primitive circle in the SE-sector is faithfully reproduced (Fig. 9a,d). Equally well replicated is the characteristic {111} pole maximum about the rotation axis together with the broad great girdle centred on this maximum (Fig. 9b,e). Even the {110} pole gure shows some of the characteristic features, as the discontinuous great girdle crossing the NE- and SE-sectors (Fig. 9c,f). In the natural sample only the maximum in the NE-sector is evident (Fig. 9c), perhaps reecting the sinistral shear sense in the foliation domain outside the strain shadow. This would also explain the asymmetric distribution of poles about the rotation axis, which is symmetric in the numerical simulation. However, relying on the natural {110} pole gure will be dangerous, as the high number of equivalent planes (six in one hemisphere) results in small density differences and in potentially ambiguous patterns. Nevertheless, it is surprising how faithfully the simulation reproduces the natural pole gures. Therefore, we believe that these data strongly support a model that explains the actual arrangement of crystallographic orientations as the product of an epitaxial growth of garnet upon biotite followed by rotation during a crenulation deformation event. In order to additionally constrain this model we have simulated the misorientation axes distribution inherent to this model and have compared the data with those of the natural sample.
DISTRIBUTION OF MISORIENTATION AXES: SIMULATION V. NATURAL DATA

Misorientation analysis (Wheeler et al., 2001) deals with the crystallographic orientation differences between couples of grains of a crystal population (in our case garnet crystals), and expresses these differences with an angle of rotation about a specic axis
2007 Blackwell Publishing Ltd

WHEN EPITAXY CONTROLS GARNET GROWTH 447

(a)

(b)

(c)

Fig. 9. Comparison of pole gures of natural data (ac) with pole gures obtained by numerical simulation (df). The concentric small girdle dispersion of {100} poles together with the concentration of poles along the primitive circle in the SE-sector (a, d) is equally well replicated as the characteristic {111} pole maximum about the rotation axis together with the broad great girdle centred on this maximum (b, e), and the {110} discontinuous great girdle crossing the NE- and SE-sectors (c, f).

(d)

(e)

(f)

(angle-axis pair) that brings the two grains to a perfect crystallographic match on each other. Misorientation data can be plotted either within a crystal reference frame (inverse pole gure) or within a sample reference frame (pole gure). Plotting the data may result in preferred orientation patterns, in which case the microstructure of the analysed sample is the result of processes that have been controlled either kinematically or crystallographically. Considering that it is assumed in our model that deformation has rotated garnet crystals about a specic kinematic axis (the 280/ 56 plunging crenulation axis), we should be able to test the model by the use of misorientation analysis. The basic assumption in our model is that garnet crystals grew epitaxially upon 56W dipping (001) biotite planes with N-S running (100) biotite traces (i.e. the pre-Alpine foliation). Relative to these (100) traces, the [001] directions of garnet crystals formed angles of 0, )30, +30 or +60. The crenulation deformation event has dispersed the crystallographic data from these original orientations. If it is assumed that the crenulation geometry is cylindrical, then misorientation axes calculated between the actual garnet orientations and the hypothetical original epitaxial positions should plot preferentially in the position of the crenulation axis. The above assumption was checked by calculating the misorientation axis distribution for our numerical simulation (Fig. 10a), and the result is exactly what is predicted. Figure 10a shows two maxima: the prominent one coincides with the crenulation axis position, whereas the second coincides with the tilted [110] axis. Rotation by 30 (or multiples of this value) about this
2007 Blackwell Publishing Ltd

second axis brings all the garnet grains which are still in the hypothetical original epitaxial growth position to a perfect overmatch. If the simulated data are compared with those established for the natural sample (Fig. 10b), then we recognize exactly the same features. Obviously, the density in the simulated sample is much higher, because the simulation reproduces garnet rotation that obeys our model, whereas in the natural sample not all garnet crystals do this.
DISCUSSION

The indication obtained from all data can be synthesized as follows: back rotation of garnet grains by 040 about the 280/56 plunging crenulation axis results in a high number of garnet grains having one {110} plane parallel to 56W dipping (001) biotite planes aligned within the relict foliation. Relative to a hypothetical NS oriented, horizontal (100) biotite trace, [001] garnet directions form angles of )30, 0, +30 and +60. It is important to note that the orientation of the (100) biotite trace is hypothetical and has not been measured, rst because biotite has broken down to garnet during high pressure metamorphism, and second because EBSD analysis on mica is problematic because of severe polishing problems. Because it is unlikely that such a systematic relationship between the crystallographic orientation of garnet and biotite is coincidental (despite the high number of garnet planes belonging to the {110} family), our data support a crystallographic control of biotite on garnet nucleation. Inspired by the idea of Powell (1966), we have tested a model wherein garnet

448 R. SPIESS ET AL.

(a)

(b)

Fig. 10. Misorientation axis distribution. (a) The numerically simulated data set shows that rotation of garnet grains around the crenulation axis between 0 and 40 produces a strong maximum of misorientation axes coinciding with the crenulation axis (280/56). (b) The same maximum as in the numerical simulation is recognized in the natural sample.

grows epitaxially upon crystallographically isoriented (001) biotite planes, with the basal and the apical pseudohexagonal oxygen ring structures operating as nucleation sites. In our model the pseodohexagonal symmetry of the ring structures allows [001] garnet directions to orient in all six-fold directions with the same probability. However, our data do not show [001] directions forming angles of )60 and +90, suggesting one or both of the following two assumptions: 1 traces of (100) biotite planes on 56W dipping (001) biotite planes were not exclusively horizontal and NS directed; 2 epitaxial growth of garnet upon biotite does not occur in all six-fold directions. A likely explanation for the latter assumption is that distortion of the pseudohexagonal oxygen rings reduces the symmetry, so that distortion energy differences favour epitaxial nucleation just in one specic direction. A major issue in an epitaxial growth model is the force that drives distortion of the pseudohexagonal oxygen structure to t that of {110} garnet planes. Distortion of the pseudohexagonal ring structure is well known within different mica types (Ferraris & Ivaldi, 2002), and is essentially controlled by the sub-

stitution processes occurring in the octahedral coordinated sites. This leads to an in-plane rotation of the Si(Al)-O4 tetrahedra, eventually distorting the pseudohexagonal oxygen ring to a perfect ditrigonal ring when rotation reaches the maximum value of 30 (Ferraris & Ivaldi, 2002). Such substitution processes in the octahedral layer can therefore not be invoked as the driving force for the distortion of the oxygen structure. Experiments in materials science suggest that distortion could be conned to the environment of broken symmetry at the surface (Matzdorf et al., 2000), i.e. the surface of (001) biotite planes upon which garnet nucleates. AFM images show that on (001) mica surfaces the basal oxygen are present, whereas K+ ions usually are not visible (Drake et al., 1989; Tsujimichi et al., 1997). Nucleation of epitaxial garnet should therefore be controlled by distortion of the basal oxygen conguration. Figure 6c shows that nucleation of garnet onto the basal oxygen conguration of (001) biotite planes needs introduction of Al and additional oxygen in the ring structure. We imagine that the introduction of these ions is the driving force for distortion of the pseudohexagonal ring structure. Distortion must not result in a perfect t between biotite and garnet,
2007 Blackwell Publishing Ltd

WHEN EPITAXY CONTROLS GARNET GROWTH 449

because it is well known from materials science that epitaxial layers relative to the substrates show mist strain (e.g. Rocher et al., 2002). We have suggested that epitaxial growth will probably occur in one specic direction on the (001) biotite plane, i.e. that with the lowest distortion energy. How can the high frequency of angles of )30, 0, +30 and + 60 between the [001] garnet direction and the hypothetically N-S trending, horizontal (100) biotite trace be interpreted? The explanation is straightforward: it implies that a signicant number of biotite crystals had their (001) planes parallel to the NS running relict foliation, whereas the (100) traces were not horizontal, but preferentially inclined between )30 and +60 from the horizontal. Figures 8 and 9ac show that the systematic dispersion trend of garnet crystallographic orientations supporting epitaxial growth is superimposed on a distribution pattern characterized by a signicant degree of randomness. One explanation for this is that deformation rotated several biotite/garnet crystals in a more arbitrary fashion than that assumed in our simplistic model.
CONCLUSIONS

Dispersion patterns of garnet crystallographic orientations together with misorientation axis distribution patterns support a model for the analysed mica schist wherein Alpine garnet grew epitaxially upon biotite, and became rotated about the crenulation axis during the Alpine deformation. Consequently, nucleation and growth sites of garnet were constrained by the spatial distribution of biotite (or mica in general) within the mica schist. The indication that nucleation of garnet may be controlled by precursor phases is an aspect that has to be addressed when the spatial distribution patterns of garnet crystals are contemplated in garnet nucleation and growth models (Carlson, 1989, 1991; Denison et al., 1997). EBSD analysis will potentially unravel such a control.
ACKNOWLEDGEMENTS

This research was funded by Progetto Ateneo 2003CPDA031431 (University of Padova), by P.R.I.N. Basamenti 2005, and by P.R.I.N. 2005-047810. We appreciate the reviews of D. Hirsch and D. Withney which helped to improve the nal version of the manuscript signicantly. Discussion on mica structures and epitaxy problems with M. Franca Brigatti and L. Secco are greatly acknowledged.
REFERENCES
Bell, T. H. & Rubenach, M. J., 1983. Sequential porphyroblast growth and crenulation cleavage developement during progressive deformation. Tectonophysics, 92, 171194.
2007 Blackwell Publishing Ltd

Brigatti, M. F. & Davoli, P., 1990. Crystal-structure renement of 1 M plutonic biotites sample M14 from a monzonite in the Valle del Cervo pluton. American Mineralogist, 75, 305313. Bucher, M. & Frey, K., 2002. Petrogenesis of Metamorphic Rocks, 7th edn. Springer-Verlag, Berlin, 341 pp. Carlson, W. D., 1989. The signicance of intergranular diffusion to the mechanisms and kinetics of porphyroblast crystallization. Contributions to Mineralogy and Petrology, 103, 124. Carlson, W. D., 1991. Competitive diffusion-controlled growth of porphyroblasts. Mineralogical Magazine, 55, 317330. Chopin, C., 1984. Coesite and pure pyrope in high-grade blueschists of the Western Alps: a rst record and some consequences. Contributions to Mineralogy and Petrology, 86, 107118. Chopin, C., 1987. Very high-pressure metamorphism in the Western Alps: new petrological and eld data. Terra Cognita, 7, 94. Chopin, C., Henry, C. & Michard, A., 1991. Geology and petrology of coesite bearing terrane, Dora-Maira Massif, Western Alps. European Journal of Mineralogy, 3, 263291. Compagnoni, R. & Rolfo, F., 2003. UHPM units in the Western Alps. In: Ultrahigh pressure metamorphism, EMU Notes in Mineralogy, Vol. 5 (eds Carwell, D.A. & Compagnoni, R.), pp. 1349. Eo tvo s University Press, Budapest. Compagnoni, R., Hirajima, T. & Chopin, C., 1995. Ultra-highpressure metamorphic rocks in the Western Alps. In: Ultrahigh Pressure Metamorphism (eds Coleman, R.G. & Wang, X.), pp. 206243. Cambridge University Press, Cambridge. Denison, C., Carlson, W. D. & Ketcham, R. A., 1997. Threedimensional quantitative textural analysis of metamorphic rocks using high-resolution computed X-ray tomography; Part I. Methods and techniques. Journal of Metamorphic Geology, 15, 2944. Downs, R.T. & Hall-Wallace, M., 2003. The American Mineralogist Crystal Structure Database. American Mineralogist, 88, 247250. Drake, B., Prater, C. B., Weisenhorn, A. L. et al. 1989. Imaging crystals, polymers, and processes in water with the atomic force microscope. Science, 243, 4898, 15861589. Ferraris, G. & Ivaldi, G., 2002. Structural features of mica. In: Micas: Crystal Chemistry & Metamorphic Petrology (eds Mottana, A., Sassi, F.P., Thompson, J.B. & Guggenheim, S.), Reviews in Mineralogy & Geochemistry, 46, 117153. Flynn, G. W. & Powell, W. J. A., 1979. The Cutting and Polishing of Electro-optic Materials. Adams Hilger, London. Frondel, C., 1940. Oriented inclusions of staurolite, zircon and garnet in muscovite. Skating crystals and their signicance. American Mineralogist, 25, 6987. Gebauer, D., Schertl, H-P., Brix, M. & Schreyer, W., 1997. 35 Ma old ultra-high-pressure metamorphism and evidence for very rapid exhumation in the Dora-Maira Massif, Western Alps. Lithos, 41, 524. Groppo, C., Castelli, D. & Compagnoni, R., 2006. Late chloritoid-staurolite assemblage in a garnet-kyanite bearing metapelite from the UHP Brossasco-Isasca Unit (Dora-Maira Massif, Western Alps): new petrological constraints for a portion of the P-T decompressional path. In: Ultrahigh-Pressure Metamorphism: Deep Continental Subduction, Vol. 403 (eds Hacker, B.H., Mc Clelland, W.C. & Liou, J.G.), pp. 127 138. GSA Special Paper. Groppo, C., Lombardo, B., Castelli, D. & Compagnoni, R., 2007. Exhumation history of the UHPM Brossasco-Isasca Unit, Dora-Maira Massif, as inferred from a phengiteamphibole eclogite. International Geology Review, 49, 142168. Hermann, J., 2003. Experimental evidence for diamond-facies metamorphism in the Dora-Maira massif. Lithos, 70, 163182. Hirsch, D. M., Ketcham, R. A. & Carlson, W. D., 2000. An evaluation of spatial correlation functions in textural analysis of metamorphic rocks. Geological Material Research, 2, 142. Ketcham, R. A., Meth, C., Hirsch, D. M. & Carlson, W. D., 2005. Improved methods for quantitative analysis of

450 R. SPIESS ET AL.

three-dimensional porphyroblastic textures. Geosphere, 1, 4259. Kienast, J. R., Lombardo, B., Biino, G. & Pinardon, G., 1991. Petrology of very high pressure eclogitic rocks from the Brossasco-Isasca Complex, Dora-Maira massif, Italian Western Alps. Journal of Metamorphic Geology, 9, 1934. Kretz, R., 1969. On the spatial distribution of crystals in rocks. Lithos, 2, 3966. Kretz, R., 1974. Some models for the rate of crystallization of garnet in metamorphic rocks. Lithos, 7, 123131. Matsumoto, N. & Hirajima, T., 2000. Garnet in pelitic schists from a quartz-eclogite unit of the southern Dora-Maira massif, Western Alps. Schweizerische Mineralogische Petrographische Mitteilungen, 80, 5362. Matzdorf, R., Fang, Z., Zhang, I. J. et al., 2000. Ferromagnetism stabilized by lattice distortion at the surface of the P-wave superconductor Sr2RuO4. Science, 289, 5480, 746748. Passchier, C. W. & Trouw, R. A. J., 1998. Microtectonics. Springer-Verlag, Berlin, Heidelberg. Powell, D., 1966. On the preferred crystallographic orientation of garnet in some metamorphic rocks. Mineralogical Magazine, 35, 10941109. Prior, D. J., Trimby, P. W., Weber, U. D. & Dingley, D. J., 1996. Orientation contrast imaging of microstructures in rocks using forescatter detectors in the scanning electron microscope. Mineralogical Magazine, 60, 859869. Quartieri, S., Chaboy, J., Merli, M., Oberti, R. & Ungaretti, L., 1995. Local structural environment of calcium in garnets: a combined structure-renement and XANES investigation. Physics and Chemistry of Minerals, 22, 159169.

Rocher, A., Ponchet, A., Blanc, S. & Fontaine, Ch., 2002. TEM evaluation of epitaxial strain in III-V semi-conductors: evidence of coherent and incoherent stress relaxation. Applied Surface Science, 188, 12, Rubatto, D. & Hermann, J., 2001. Exhumation as fast as subduction? Geology, 16, 577588. Schertl, H-P., Schreyer, W. & Chopin, C. (1991) The pyropecoesite rocks and their country rocks at Parigi, Dora-Maira Massif, Western Alps: detailed petrography, mineral chemistry and P-T path. Contributions to Mineralogy and Petrology, 108, 121. Sharp, Z. D., Essene, E. J. & Hunziker, J. C., 1993. Stable isotope geochemistry and phase equilibria of coesite-bearing whiteschists, Dora Maira Massif, Western Alps. Contributions to Mineralogy and Petrology, 114, 112. Tsujimichi, K., Tamura, H., Hirotani, A., Kubo, M., Komiyama, M. & Miyamoto, A., 1997. Simulation of atomic force microscopy images of cleaved mica surfaces. Journal of Physical Chemistry B, 101, 42604264. Wheeler, J., Prior, D. J., Jiang, Z., Spiess, R. & Trimby, P. J., 2001. The petrological signicance of misorientations between grains. Contributions to Mineralogy and Petrology, 141, 109 124. Zhang, R. Y., Zhai, S. M., Fei, Y. W. & Liou, J. G., 2003. Titanium solubility in coexisting garnet and clinopyroxene at very high pressure: the signifcance of exsolved rutile in garnet. Earth and Planetary Science Letters, 216, 591601. Received 31 May 2006; revision accepted 1 February 2007.

2007 Blackwell Publishing Ltd