Journal of Petrology Advance Access originally published online on August 1, 2005
Journal of Petrology 2005 46(12):2593-2613; doi:10.1093/petrology/egi067
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Published by Oxford University Press 2005
Plastic Deformation and Recrystallization of Garnet: A Mechanism to Facilitate Diffusion Creep
1 DEPARTMENT OF GEOLOGY, UNIVERSITY OF LEICESTER, UNIVERSITY ROAD, LEICESTER LE1 7RH, UK
2 DEPARTMENT OF EARTH AND OCEAN SCIENCES, UNIVERSITY OF LIVERPOOL, LIVERPOOL L69 3GP, UK
RECEIVED SEPTEMBER 27, 2004; ACCEPTED JUNE 28, 2005
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONTECTONOMETAMORPHIC SETTINGANALYTICAL METHODSMICROSTRUCTURAL CONTEXT OF...CONCEPTUAL MODELCONCLUSIONSREFERENCES |
|---|
Elongate and deformed garnets from Glenelg, NW Scotland, occur within a thin shear zone transecting an eclogite body that has undergone partial retrogression to amphibolite facies at circa 700°C. Optical microscopy, back-scattered electron imaging, electron probe microanalysis and electron back-scatter diffraction reveal garnet sub-structures that are developed as a function of strain. Subgrains with low-angle misorientation boundaries occur at low strain and garnet orientations are dispersed, around rational crystallographic axes, across these boundaries. Towards high-strain areas, boundary misorientations increase and there is a loss of crystallographic control on misorientations, which tend towards random. In high-strain areas, a polygonal garnet microstructure is developed. The garnet orientations are randomly dispersed around the original single-crystal orientation. Some garnet grains are elongate and Ca-rich garnet occurs on the faces of elongate grains oriented normal to the foliation. Commonly, the garnet grains are admixed with matrix minerals, and, where in contact with other phases, garnet is well faceted. We suggest that individual garnet porphyroclasts record an evolution from low-strain conditions, where dislocation creep and recovery accommodated deformation, through increasing strain, where dynamic recrystallization occurred by subgrain rotation, to highest strains, where recrystallized grains were able to deform by diffusion creep assisted grain boundary sliding with associated rotations.
KEY WORDS: diffusion creep; EBSD; garnet; plastic deformation; recrystallization
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONTECTONOMETAMORPHIC SETTINGANALYTICAL METHODSMICROSTRUCTURAL CONTEXT OF...CONCEPTUAL MODELCONCLUSIONSREFERENCES |
|---|
Microstructures in deformed rock-forming minerals are crucial in aiding our understanding of the rheology of the bulk Earth. Garnet is a particularly important mineral, in both the crust and the mantle. It is a key constituent of eclogite at the base of thickened continental crust and within subducted crustal slabs, upper mantle peridotites and in the mantle Transition Zone (majoritic garnet) at depths of 400–650 km (Karato & Wu, 1993
; Vauchez & Mainprice, 1996) and may control mantle rheology (Karato et al., 1995). Additionally, garnet is a key mineral in many metamorphic rocks.
Garnet chemistry is used directly in many geothermobarometers (Pattison & Tracy, 1991; Spear, 1992) and in radiometric dating (Vance & Onions, 1988, 1990; Vance & Holland, 1993; Christensen et al., 1994; Vance, 1995; DeWolf et al., 1996; Vance et al., 1998; Scherer et al., 2000). Linking such data to the microstructural relationships of garnet porphyroblasts, their inclusions and external fabrics is often crucial to establishing pressure–temperature–time (P–T–t) paths of metamorphic terranes (e.g. Spear & Selverstone, 1983; Daly et al., 1989; Okudaira, 1996; Marshall et al., 1997; Vance & Mahar, 1998; Kohn et al., 2001). Hence, understanding the behaviour of garnet under high-strain conditions is critical to understanding both deformation and metamorphism in the mid to lower crust and in the upper mantle.
Until recently, our understanding of the deformation mechanisms of rock-forming minerals has been largely restricted to anisotropic crystallographic phases that can be studied using optical microscopy. Although the early Laue diffraction studies of flattened garnets in mylonites (Dalziel & Bailey, 1968; Ross, 1973), etching (Carstens, 1969, 1971) and transmission electron microscope (TEM) studies of natural garnet (Smith, 1982; Allen et al., 1987; Ando et al., 1993; Doukhan et al., 1994; Voegele et al., 1998a, 1998b, 1999, 2000) have suggested various mechanisms responsible for elongation, it is only recently, with the advent of scanning electron microscope (SEM)-based techniques such as electron channelling (Lloyd, 1987; Lloyd et al., 1991a, 1991b) and electron back-scatter diffraction (EBSD: see Venables & Harland, 1973; Dingley, 1984; Prior et al., 1999, 2002) that crystallographic data can be readily obtained from cubic (isotropic) minerals such as garnet. These new technologies have led to a recent resurgence of interest in garnet deformation mechanisms.
Elongate garnets have been observed in a wide range of P–T environments from greenschist-facies (Azor et al., 1997) to granulite-facies (Ji & Martignole, 1994) and eclogite-facies (Ji et al., 2003). In granulite-facies garnets, deformation is interpreted to be due to plastic deformation, on the basis of TEM work that showed some subgrain structures (Ji & Martignole, 1994). Kleinschrodt and co-workers (Kleinschrodt & McGrew, 2000; Kleinschrodt & Duyster, 2002) have demonstrated that elongate garnets within granulite-facies quartzites from Sri Lanka have a weak crystallographic preferred orientation (CPO). However, these garnets exhibit little sub-structure and, as such, convincing deformation mechanism information derived from these samples is fairly restricted. Garnets in eclogites (Ji et al., 2003; Mainprice et al., 2004) also have weak CPOs—much weaker than the fabrics predicted by a visco-plastic self-consistent model that uses dislocation systems identified in the same samples using TEM. Mainprice et al. (2004) concluded that although garnet deformed by dislocation creep and recovery, the other minerals (omphacite and quartz) accommodated a significant proportion of the deformation. Prior et al. (2000) have shown that mantle garnets contain significant sub-structures that are best explained by high-temperature dislocation creep and recovery. Prior et al. (2002), during an earlier study on the Glenelg eclogite sample that will be examined further in this paper, suggested that elongate garnet within crustal rocks is also best explained by dislocation creep and recovery. These recent studies have served to convince us that crystal-plastic deformation can account for elongation of garnet under certain conditions, yet the precise deformation mechanisms and conditions are yet to be determined. Terry & Heidelbach (2004) have shown recently that fine-grained garnet can deform by grain boundary sliding with grain boundary migration as an associated mechanism. In their study, garnet grain sizes varied from about 30 µm at low strains to <10 µm at high strains and they showed evidence that garnet growth occurred during deformation. A key question that remains is whether garnet, when deformed to high strains, will recrystallize in a similar way to other rock-forming minerals (Urai et al., 1986; Urai & Jessell, 2001) and whether its mechanical properties will evolve with that recrystallization process.
We have, however, to be careful in interpreting garnet sub-structures, as they can be produced by mechanisms other than creep deformation. Fractures are common in garnet, and may be filled by garnet (precipitating from solution, for example) so that a sub-structure is imposed involving no other phases (Matthews et al., 1992; Prior, 1993; Whitney, 1996). A recent study of eclogitic garnet from the Sesia Zone of the Alps by Trepmann & Stöckhert (2002) is particularly important here as some of the microstructures they described are similar to the microstructures presented in this paper. They interpreted deformation within elongate, asymmetric and sub-structured garnet grains due to cataclasis during seismic loading (a more detailed discussion of this is presented later). It is also true that the presence of garnet sub-structure does not a priori necessitate that deformation has occurred. For example, garnet porphyroblasts can develop sub-structures from the amalgamation of many independently nucleated garnet grains (Spiess et al., 2001; Wheeler et al., 2001; Prior et al., 2002; Dobbs et al., 2003) and radial growth structures are observed in some undeformed garnets (Hirsch et al., 2003). Thus, it is important that each sample is assessed on the basis of as much microstructural and micro-chemical data as possible.
This paper advances our understanding of garnet deformation by presenting a study of deformed garnets from Glenelg, NW Scotland.
|
TECTONOMETAMORPHIC SETTING |
|---|
|
TOPABSTRACTINTRODUCTIONTECTONOMETAMORPHIC SETTINGANALYTICAL METHODSMICROSTRUCTURAL CONTEXT OF...CONCEPTUAL MODELCONCLUSIONSREFERENCES |
|---|
Within the Eastern Unit of the Glenelg–Attadale Inlier of NW Scotland (Fig. 1), eclogites and their retrogressed equivalents are ubiquitous. The eclogites were retrogressed to upper amphibolite-facies assemblages during a period of exhumation related to ductile shearing at this grade. The major shear zones are approximately N–S striking (where not refolded by later structures) and dip moderately to the east. A
1 km zone of mylonites and ultramylonites demarcates the main structural and metamorphic discontinuity between the Western and Eastern Units, herein termed the Barnhill Shear Zone (BSZ; Fig. 1). A few hundred metres structurally above this shear zone in the Eastern Unit, there is a large (100 m wide by several hundred metres long) N–S striking rib of eclogite. Within this rib at NG 8055 1570 (Fig. 1) is a thin 1 m wide, sub-vertical shear zone, striking approximately E–W. The shear zone represents a significant localization of strain as the surrounding eclogites do not contain a mesoscopic fabric. Within the shear zone, garnets form a sub-horizontal penetrative stretching lineation on the foliation surfaces. The matrix comprises very fine-grained amphibole and plagioclase demonstrating that it is a retrograde amphibolite-facies shear zone related to exhumation. This shear zone is cofacial with, and probably temporally related to the BSZ and may represent an anastomosed shear that was localized due to the rheological properties of the surrounding rigid eclogite body. Amphibole (pargasite) and the feldspar (oligoclase) are identical in composition to the matrix minerals in the major shear zones and to the static retrograde replacement of omphacite and garnet in undeformed eclogites (Storey, 2002).
|
Thermobarometry has constrained peak temperatures of
730°C for the eclogites (Sanders, 1989), with peak pressure of 20 kbar (Rawson et al., 2001; Storey, 2002; Storey et al., 2005). Sanders (1989) demonstrated an early peak isothermal decompression from 17 kbar to 15 kbar (i.e. within the eclogite-facies). More recent work has demonstrated that amphibolite-facies retrogression occurred at 650–700°C and 13 kbar (Storey, 2002; Storey et al., 2005). Therefore there is strong evidence that there was a pseudo-isothermal decompression of 20–25 km (7 kbar) that must have been rapid. Eclogite-facies metamorphism occurred close to 1100 Myr ago (Sanders et al., 1984), whereas the amphibolite-facies retrogression occurred at 1000 Myr ago (Brewer et al., 2003). Due to the temporal link between the BSZ and the shear zone discussed here, it is probable that similar P–T conditions of 650–700°C and 13 kbar were sustained during shearing. This is certainly compatible with upper amphibolite-facies conditions during shearing, on the basis of both petrographic and P–T data. |
ANALYTICAL METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONTECTONOMETAMORPHIC SETTINGANALYTICAL METHODSMICROSTRUCTURAL CONTEXT OF...CONCEPTUAL MODELCONCLUSIONSREFERENCES |
|---|
Electron back-scatter diffraction
The EBSD work presented here was completed during four sessions (1999, 2000, 2002 and 2004) using three different methods. SYTON-polished thin sections (Fynn & Powell, 1979
; Lloyd, 1987; Prior et al., 1996) were used throughout. The earliest work comprised manual EBSD analyses (Prior et al., 1999), located using forescatter orientation contrast images (Prior et al., 1996) on a Phillips XL30 SEM at the University of Liverpool. Analytical details are the same as those used by Spiess et al. (2001). All other data were collected using automated EBSD mapping (Adams et al., 1993; Prior et al., 2002) on a CamScan X500 crystal probe, at the University of Liverpool, fitted with a thermionic field emission gun running at 20 keV. Work in 2000 (everything except for the detailed maps of GTS202 and the data on phases other than garnet) was completed at the same time as work published by Prior et al. (2002) and uses the same methods. During the EBSD studies conducted from 2002 to 2004, we used a carbon-coated sample and a higher beam current (20 nA); additionally, EBSD patterns were indexed and EBSD maps processed using the Channel 5 software package from HKL Technology and the data were acquired by stitching together several EBSD beam scans. Data were acquired in 2002 (detailed maps of GTS202) at 0·67 s per point and in 2004 (amphibole, feldspar and quartz fabrics) at 0·12–0·18 s per point; increased data acquisition speeds were facilitated by new camera technology and faster computing systems. Error levels for the 2000 work were outlined by Prior et al. (2002); the later work has much lower error levels. Map processing procedures have been described by Prior et al. (2002).
Back-scattered electron images and X-ray maps
Most back-scattered electron images (BSE) were collected on a Phillips XL30 SEM, in the University of Liverpool. Further BSE images and X-ray maps were collected on a Cameca SX50 electron microprobe at the University of Leeds, using an accelerating voltage of 20 keV and a beam current of 50 nA. For high-resolution, small-scale X-ray maps we used a beam-scan with a grid spacing of 0·065 µm. Lower-resolution, larger maps used a stage-scan, with a grid spacing of 2 µm. In all cases, Ca, Mg and Fe X-rays were collected with a dwell time of 30 ms.
|
MICROSTRUCTURAL CONTEXT OF GARNET |
|---|
|
TOPABSTRACTINTRODUCTIONTECTONOMETAMORPHIC SETTINGANALYTICAL METHODSMICROSTRUCTURAL CONTEXT OF...CONCEPTUAL MODELCONCLUSIONSREFERENCES |
|---|
Thin sections of the shear zone, cut parallel to the lineation, reveal the intense degree of deformation (Figs 2–4). The sample comprises garnet porphyroclasts and quartz ribbons in a matrix of thoroughly mixed, fine-grained (
5–10 µm), plagioclase and amphibole (pargasite) with (1–10 µm) relict Na-clinopyroxenes and accessory rutile. Proportions of pyroxene and amphibole vary considerably within individual thin sections. Foliation is defined by garnet porphyroclast shapes (with axial ratios up to about 10:1), quartz ribbon shapes (also with axial ratios up to about 10:1) and millimetre-scale quartz–amphibole/clinopyroxene/plagioclase–garnet layering. Shear bands occur at about 20–30° to the main foliation and always have the same sense of asymmetry. Shear bands are 2–6 mm in length and up to a few hundred micrometres in width. They contain amphibole and/or plagioclase and often small garnet grains and fine quartz ribbons. The amphibole and plagioclase grains are sub-equant and, hence, do not define a tectonic fabric. Quartz ribbons comprise large (200 µm) granoblastic polygons that are almost entirely devoid of undulose extinction. Rutile occurs as stringers of small equant grains within the amphibole–plagioclase matrix.
|
10° misorientation), only for grains Quartz ribbons have weak CPOs, as do plagioclase, and amphibole in the matrix (Fig. 2). Clinopyroxene has a well defined non-random CPO, but the symmetry of the CPO does not correspond to the symmetry of the shear zone. Clinopyroxene
b
-axes are strongly clustered in an orientation significantly oblique to both the lineation and the pole to foliation, c-axes define a weak girdle normal to the b-axis cluster and a-axes are close to random. Although the quartz, amphibole and plagioclase CPOs are weak, they have non-random symmetry elements that are aligned to symmetry elements of the clinopyroxene CPO, but not to the symmetry elements of the shear zone (the foliation and lineation). Clusters of amphibole b-axes, quartz c-axes, and perhaps plagioclase a-axes, align approximately with the clinopyroxene b-axes. Garnet CPOs from the same area are clustered strongly around the orientation of a single porphyroclast that dominates the analysis area. Garnet CPOs will be discussed in more detail later.
Garnet porphyroclasts
Garnet porphyroclasts vary from equant to elongate. In three dimensions, elongate porphyroclasts are rods. The porphyroclast morphology varies from:
- equant large porphyroclasts (up to a few millimetres) that comprise single grains with little internal structure (Fig. 3a);
- elongate porphyroclasts (up to a few millimetres wide) with considerable internal distortion (between fractures), including low-angle boundaries (Fig. 3b); CPOs approximate a distorted single crystal (Fig. 3b);
- elongate porphyroclasts (up to 1 mm wide) that comprise aggregates of many smaller garnet grains (of about 50 µm diameter) that may have some internal deformation (Fig. 3c); these small grains are often equant but some are elongate, approximately parallel to the trace of the foliation, with axial ratios up to 2:1 (Fig. 3c); CPOs are scattered widely around a single crystal orientation (Fig. 3c);
- stringers (generally less than 0·5 mm wide) of closely spaced small idiomorphic garnet grains (of about 50 µm diameter) with intergranular amphibole, plagioclase, pyroxene and quartz (Fig. 3d); the garnet grains have some internal deformation; these grains are often equant but some are elongate, approximately parallel to the trace of the foliation, with axial ratios up to 2:1 (Fig. 3d); there are also numerous isolated small garnet grains (of about 50 µm diameter) within the amphibole/plagioclase matrix; CPOs are scattered widely around a single crystal orientation (Fig. 3d).
|
Most garnet porphyroclasts comprise a mixture of more than one of these morphologies (Figs 3e and 4). In a number of cases, porphyroclasts are affected by shear bands (Fig. 4) and there is a transition from morphology 1 to morphology 4 as the shear band is approached. These relationships suggest that the different morphologies are representative of different states of strain of garnet: single crystal porphyroclasts have been deformed to create polycrystalline porphyroclasts. The correspondence of the small grains to high-strain zones suggests that an alternative model in which the small grains amalgamate to create large grains (Spiess et al., 2001
) is not viable here.
|
Fractures of all orientations and varying width are recognized in many, but not all, garnet porphyroclasts. The fractures have similarities to those described from deformed eclogitic garnets from the Sesia zone (Trepmann & Stöckhert, 2002
), although fractures are not as extensive in the Glenelg garnets and most fractures show no shear offset (Fig. 3a). All garnet porphyroclasts of morphologies 1 and 2 have significant intragrain distortion between fractures and the range of shapes cannot be explained by fracture networks alone. Some arrays of spaced fractures have orientations and offsets consistent with their being shear fractures with similar kinematics to the overall shearing (Figs 3c and 4). Where garnet is cut by shear bands, these fractures are deflected into the shear band (Fig. 4) and, in these cases, arrays of more irregular fractures occur perpendicular to, and between, the shear fractures. Fractures are filled with fine-grained aggregates (5 µm) of amphibole, plagioclase and quartz. Irregular patches in which grains of these included minerals are slightly larger (5–20 µm) are located in the sites of larger fractures. Probable primary inclusions include rutile–ilmenite grains and relict clinopyroxene, usually surrounded by fine-grained aggregates of amphibole. Rutile inclusions extend along fractures and, in some cases, rutile grains can be traced along fractures to a matrix grain adjacent to the garnet grain boundary. Below, we present detailed microstructural data from one example (GTS202) of a porphyroclast affected by a shear band and use these data to infer the mechanisms by which the garnet has deformed.
Microstructure of GTS202
Figure 4 shows the microstructural setting of garnet porphyroclast GTS202; this is one of several garnet grains that are affected by a sinistral shear band oriented about 20° away from the main foliation orientation (Fig. 4a and c). GTS202 is a fragment at the end of a larger parent porphyroclast (Fig. 4b). A series of fractures break up the parent porphyroclast (Fig. 4). GTS202 is about 0·8 x 0·4 mm outside the shear band and is oriented obliquely to both the main foliation and the shear band, the orientation being controlled by the fractures. GTS202 is deflected through 45° and thins to <100 µm as it approaches the shear band. We suggest that this deflection and thinning correspond to an increase in shear strain associated with the shear band. A second porphyroclast, adjacent to the GTS202 parent, undergoes the same shear deformation (Fig. 4b). Stringers of garnet grains (morphology 4) extend into the shear band, from the end of GTS202 and the neighbouring sheared porphyroclast (Fig. 4c). GTS202 shows a transition from a deformed garnet (morphology 1–2) into an aggregate of many smaller garnet grains (morphology 3; Fig. 5) with increasing shear strain. There are considerable sub-structures within GTS202 (Fig. 5).
|
To facilitate more concise descriptions of the microstructure, we will define terms thus:
- low-angle boundaries have misorientations less than 10°, high-angle boundaries have misorientations greater than or equal to 10°; the choice of 10° is arbitrary; we do not think that a more sophisticated approach (e.g. Trimby et al., 1998) will change the conclusions significantly;
- the term ‘interface’ refers to the contact of unlike phases (Sutton & Balluffi, 1995);
- grains are regions surrounded entirely by high-angle boundaries and/or interfaces;
- grains may contain two or more subgrains; at least part of the boundary between subgrains must be low-angle;
- the GTS202 microstructure is divided into three broad areas, as defined in the grain map shown in Fig. 5d; these are called the low-strain area, the transitional area and the high-strain area.
Boundaries, grains and subgrains
Boundaries and interfaces can be constrained by comparing the BSE image (Fig. 5a), which shows where garnet is present, with the EBSD band contrast image (Fig. 5b), which shows where there are boundaries, and the corresponding boundary map (Fig. 6b) that quantifies misorientation angle across boundaries. The low-strain area of GTS202 approximates a single grain with lattice distortion (Figs 5b–d, 6b and 7a–c) and some low-angle boundaries. Most of the low-strain area of the grain (Fig. 4) is not shown in the detailed maps. EBSD data and orientation contrast images (not shown) indicate that the bulk of this grain (including the right-hand side of the low-strain area shown in the detailed maps Figs 5 and 6) contains no low-angle boundaries—just some discrete changes in orientation that correspond to fractures. The density of boundaries and the misorientations across boundaries increase with increasing strain within the low-strain area (Figs 6b, 7b and c). At the left-hand end of the low-strain area, more low-angle boundaries are developed and define subgrains of about 50 µm in size.
|
The transitional area comprises a few large 50 to
200 µm grains containing lattice distortion and 10–50 µm subgrains. The high-strain area of GTS202 comprises individual garnet grains (Figs 5c and 6a) of 10–50 µm diameter, which exhibit lattice distortion and some subgrains (Figs 5b and 6b). The proportion of second phases (quartz, feldspar and amphibole) is higher in the high-strain area (Fig. 5a). Garnet grains in the high-strain area are broadly polygonal. Some of these grains are elongate, approximately parallel to the trace of the foliation, with axial ratios up to about 2:1. High-angle boundaries tend to be straight to gently curved between triple junctions (Figs 5b and 6a) with some irregularity (amplitudes of a few micrometres over wavelengths of a few micrometres; Fig. 5d). Most of these boundaries do not correspond to rational crystallographic planes. In contrast, the garnet interfaces tend to be faceted (Fig. 5a and e). An analysis of the facet traces (Fig. 5e) suggests that most are consistent with being {112} faces, with the remainder {110} faces. {112} and {110} faces are the two most commonly expected in rhombohedral or icositetrahedral garnets (Read, 1970; Deer et al., 1992).
Crystallographic preferred orientations and misorientations
The pole figures for GTS202 (Fig. 6c) show a mixed signature. Grains of >200 µm in diameter correspond to the low-strain area and part of the transitional area and these give an orientation close to a single crystal, with a rotational dispersion of about 25° around an axis somewhere between the [110] and the [111] on the left side of the pole figure (Fig. 6e). As there are only a few grains in the low-strain and transitional areas, a pole figure in which each grain is represented by one point (Fig. 6d) is dominated by the small garnet grains in the high-strain area. The grains in the high-strain area have orientations (Fig. 6d) that cluster around the preferred orientation of the low-strain area, but contain an extra component of dispersion (25–30°) with no consistent rotation axis.
Orientations along linear transects across the low-strain area (Fig. 7a–c) show a progressive change that corresponds to a dispersion with an anticlockwise rotation around a 110, or perhaps a 111, direction. A transect across the lowest-strain part of the garnet (Fig. 7a), outside the mapped area, shows dispersion around the [110] direction in the bottom left of the pole figure. In transects closer to the transitional area (Fig. 7b–c); the dispersion is around an axis near the [110] and [111] directions that lie closest to the centre of the pole figure (the centre being the kinematic rotation axis of the shear band). Approximately 15° of anticlockwise rotation is accommodated in the low-strain area and a further 10° in the transitional area. Neighbour-pair misorientations (of adjacent pixels) within and between grains of >200 µm in diameter (these are restricted to the low-strain and transitional areas) (Fig. 6f) are dominated by low angles, to the exclusion of angles of >20°. These low angles are represented in the random-pair distribution but are not as significant, suggesting that neighbouring pixels have a tendency towards low-angle misorientations that is not merely a function of the strong preferred orientation (Wheeler et al., 2001). Neighbour-pair misorientation axes in the low-strain and transitional areas show a preferred orientation parallel to the 110 direction (Fig. 6f) for misorientations <10°, consistent with the transect data shown in Fig. 7.
|
Linear transects that cross from the transitional area into the high-strain area (Fig. 7d) show discrete changes in orientation corresponding to boundaries and a loss of the crystallographic control on dispersion as the high-strain area is entered. High-angle grain boundaries (dominantly in the high-strain area) have random misorientation axes (Fig. 8). Neighbour-pair misorientations for the whole dataset show a tail of high-angle misorientations superposed on the low-angle misorientations (Fig. 6g). As no misorientations of >20° are apparent in lower-strain areas (Fig. 6f), the tail of high-angle misorientations must relate to the high-strain area. Neighbour-pair misorientation axes for misorientations of <10° in the whole sample are generally parallel to
100 directions (Fig. 6g). As the low-strain and transitional area misorientation axes are 110 dominated, the 100 misorientation axes must be developed in the high-strain area: grains in the high-strain area have internal distortions and subgrains with 100 misorientation axes.
|
Compositional data
Back-scattered electron images do not display any clear chemical heterogeneities within the garnet (Fig. 5), so X-ray maps of Ca, Mg and Fe were taken from critical areas to investigate this further. In all cases, the chemical variation is most prominent for Ca. Mg and Fe display antithetic relationships to Ca but with a weaker X-ray signal; hence, all X-ray maps discussed are for Ca (Fig. 9).
|
Chemical zoning of garnet is only significant in the transitional and high-strain area. A clear chemical sub-structure correlates strongly with the locations of garnet boundaries (Fig. 9). Ca-lows occur along most of the boundaries and are equally well developed on low-angle (e.g. between subdomains 5 and 6, and 7 and 8; Fig. 9) and high-angle (e.g. between subdomains 1 and 4, 2 and 3, and 9 and 10; Fig. 9) boundaries. However, the most spectacular features of the X-ray maps are the Ca-highs that occur along some high-angle boundaries, as discrete patches adjacent to the boundaries. In the transitional area, at least two linear tracts of Ca-rich garnet (one including the boundary of subdomains 10 and 11 and the other including the boundary of subdomains 12 and 13) are oriented sub-perpendicular to the overall garnet long axis (Fig. 9f). Neither extends across the entire width of the garnet and both are associated with high-angle boundaries and/or interfaces. The boundaries are located within the high-Ca tracts rather than next to them. In the high-strain area, individual, elongate garnet grains (e.g. subdomains 1 and 2; Fig. 9) are zoned so that Ca-highs are developed preferentially on the boundaries or interfaces parallel to the short axis (Fig. 9c). A thin Ca-low, similar to that on most boundaries, occurs along relatively sharp lines that juxtapose cores of these grains and their Ca-rich ends. Generally, there is no crystallographic boundary corresponding to these changes in chemistry and where there is a boundary, it is low-angle. The Ca-highs are developed best, although not exclusively, at interfaces (most noticeably against quartz) and the Ca-rich garnet has well-developed facets at the interface. In another high-strain area, not covered by the detailed EBSD maps, discrete garnet grains have Ca-low rims and a ‘pore-filling’ network of Ca-rich garnet (Fig. 10). Garnet grains with Ca-highs are elongate. The Ca-low parts of these grains (e.g. subdomains 1 and 2 in Fig. 9) are also commonly elongate.
|
Bulk CPO
Measuring the bulk CPO of garnet in this sample by EBSD is not easy. A single thin section contains no more than 50 or so porphyroclasts. These show no preferred orientation; the small number of porphyroclasts analysed renders the result statistically insignificant. Measuring the bulk CPO of small garnet grains, developed in morphologies 3 and 4 (e.g. the high-strain area of GTS202) is more feasible. Figure 11 shows the CPOs of 17 individual porphyroclasts from one thin section. The data from each porphyroclast are plotted as one point per grain, so that a small grain is given as much weighting as a large one; the effect of this can been seen in data for GTS202, which are plotted as one point per grain in Fig. 6d. The datasets from all the porphyroclasts have been added together and the overall result (Fig. 11b) is a weaker CPO than that shown by any individual porphyroclast. The bulk CPO is still not random, however—each porphyroclast dataset clusters around a single crystal orientation and with only 17 porphyroclasts in the bulk CPO, there is still a weak single crystal signature that is biased by the most significant individual porphyroclast dataset. The bulk CPO does have some similarities to measured and modelled simple shear fabrics (Mainprice et al., 2004
). However, it is dominantly a single crystal signature and it seems reasonable that the weak CPO is an artefact of poor statistics. We believe that a statistically more significant dataset is likely to show a random CPO.
|
Deformation, recovery and recrystallization mechanisms
Dislocation creep and recovery
Dispersions with rational crystallographic rotation axes, as seen in porphyroclast morphologies 1–2, are a clear indicator of plastic deformation by dislocation creep (Lloyd & Freeman, 1991
; Lloyd et al., 1997; Boyle et al., 1998; Prior et al., 2002; Bestmann & Prior, 2003). In GTS202, the sense of dispersion (anticlockwise rotation), with increasing strain, is consistent with the sinistral kinematics of the shear band and the amount of rotation in the low-strain and transitional areas is of the same order as the amount of bending of the porphyroclast margin in the same areas. Thus, it seems reasonable that dislocation creep deformation of garnet was due to larger-scale shearing.
Low-angle boundaries that share the same dispersion geometry suggest that recovery has accompanied plastic deformation. Some, but not all, low-angle boundaries correspond to fractures, indicating that fracture and plastic deformation maybe closely linked (see Trepmann & Stöckhert, 2002). However, the dominance of non-fracture low-angle boundaries in GTS202 and the intra-subgrain distortion suggest that dislocation creep and recovery are the most important processes in the development of garnet subgrains. Similar microstructures have been described recently in other garnet samples (Prior et al., 1999, 2000, 2002).
The dispersion geometry can be inverted to constrain the dislocation systems that accommodate plastic deformation and make up subgrain walls: in one garnet from this sample, dispersions around 110 are best explained by slip on dislocations with Burgers vectors 
111 (Prior et al., 2002). The low-strain and transitional areas of GTS202 are also dominated by dispersions around 110, with the most probable boundary planes {111}. These data are also consistent with slip on 111. The larger dataset shows a range of dispersion axes, dependent presumably upon the grain orientation and the potential for slip on different dislocation systems. The development of significant 100 misorientations on low-angle boundaries in the high-strain area of GTS202 may reflect the activation of different slip systems as the grain orientation changes. This aspect warrants further work but will not be explored further here.
Subgrain/grain boundary segregation
Ca-lows are observed on many boundaries. We will argue later that the Ca-highs represent the stable garnet composition that grew during deformation. Although it is tempting to suggest that the Ca-lows relate to diffusion along grain and subgrain boundaries, this idea is inconsistent with the high-Ca garnet being the stable composition. An alternative explanation is that the Ca-lows represent segregation along subgrain or grain boundaries such as that commonly observed in metallic alloys, semiconductors and ceramics (Bernardini, 1998; Ross et al., 2001; Wynblatt et al., 2003). Individual dislocations distort the lattice. In a solid solution such as garnet, the distorted lattice around the dislocation may have dimensions that match one solid-solution end-member better than the others. Short-range vacancy diffusion, to increase the concentration of that end-member in the distorted region, will reduce the stresses related to these dislocations. The process does not require long-range lattice diffusion to concentrate the favoured end-member along a subgrain boundary, as the diffusion will be facilitated by dislocation climb during recovery. As we suggest that subgrain boundaries develop into grain boundaries by rotation recrystallization (see below), the grain boundaries will inherit the subgrain boundary composition. These ideas must remain speculative at the moment—the level of detail needed to firm up these ideas is well beyond the scope of this work.
Generation of new grains: subgrain rotation recrystallization
The increase in low-angle boundary density and misorientation as the strain increases in the low-strain area suggests that subgrain rotation (White, 1977; Poirier & Guillope, 1979) has operated in conjunction with recovery (see definitions in Trimby et al., 1998). The observation that garnet grains in the high-strain area have similar sizes to the subgrains in the neighbouring transitional area is consistent with the grains having developed by subgrain rotation recrystallization (Poirier & Guillope, 1979).
Grain boundary sliding
The random dispersion of the orientations of grains in high-strain areas around the orientation of garnet in the low-strain area of the same porphyroclast can be achieved by grain boundary sliding and concomitant grain rotations that have no crystallographic control (Jiang et al., 2000; Bestmann & Prior, 2003). Ultimately, this would lead to the randomization of the garnet bulk CPO as the data presented in Fig. 11 suggest.
Grain boundary sliding would be consistent with the increasing proportion of second phases, incorporated into garnet porphyroclasts, with increasing strain as grain boundary sliding necessitates neighbour switching (Ashby & Verrall, 1973; Boullier & Gueguen, 1975; Drury & Humphreys, 1988; Fliervoet et al., 1997) and can facilitate mechanical mixing of garnet grains with amphibole, pyroxene, quartz and feldspar in neighbouring layers. This process requires that the other phases also deform by a sliding mechanism. It is notable that the other phases do not have CPOs that correspond to the shear zone kinematics. We suggest that the CPOs that are observed in pyroxene, quartz, amphibole and plagioclase pre-date shear zone deformation. These CPOs have been weakened but not destroyed by a randomization process associated with grain boundary sliding (see Jiang et al., 2000).
Grain boundary sliding, accommodated by diffusion, has been suggested as an explanation of garnet microstructures in eclogite-facies shear zones (Terry & Heidelbach, 2004). In our study, it is clear that grain boundary sliding did not operate to the exclusion of other mechanisms. There is evidence that grains continued to deform internally by dislocation creep.
Diffusion creep
Grain boundary sliding needs an accommodation mechanism. Mechanisms include frictional sliding and associated dilatation (Sammis, 2001; Bocquet et al., 2002; Shodja & Nezami, 2003), diffusion creep (Ashby & Verrall, 1973; Gifkins, 1973, 1976) and grain boundary dislocation creep (Gifkins, 1991, 1994). Randomization of garnet orientations around porphyroclasts from the Sesia zone (Trepmann & Stöckhert, 2002) is postulated to occur by a cataclastic (frictional) process: Trepmann & Stöckhert (2002) show good evidence (their figs 4a and 7) that the shapes and size distributions of small garnets relate to fractures. Such evidence is lacking here and, as discussed above, the generation of small garnet grains in high-strain areas is best explained by subgrain rotation recrystallization. Ca-rich garnet provides the best evidence to suggest that diffusion was an important process during garnet deformation. The interfaces of Ca-rich garnet are well faceted. Dissolution can create facets, but on a smaller scale than the grains being dissolved (Prior, 1993). Idiomorphic crystal forms observed here, where facet dimensions are the same order as grain-size, are best related to growth of these surfaces (see discussion by Spiess et al., 2001). Garnet growth requires diffusion. Ca-rich garnet overgrowths are developed adjacent to some high-angle boundaries—most specifically, high-angle boundaries that would have been in an extensional orientation with respect to the shear zone kinematics. One way to explain this would be the growth of garnet in opening void spaces during deformation.
There is good evidence for diffusion. The relationship of Ca-rich garnet to extensional sites suggests that this diffusion operated during deformation and that diffusion creep provides the most probable mechanism to control the grain boundary sliding process. Furthermore, the preservation of zoning profiles within individual garnet grains and the correspondence of chemical heterogeneities to boundaries suggest that grain boundary diffusion (Coble creep, pressure solution) was more likely than volume diffusion. Volume diffusion would be unlikely to be a kinetically favourable process at the temperatures of deformation here (Yardley, 1977; Ayres & Vance, 1997).
The morphology of garnet triple junctions that contain second phases (Fig. 5e) is reminiscent of microstructures that indicate the presence of melt (D. Kohlstedt, personal communication; see Mei et al., 2002). Whether melt is present or not, a mechanism is still required for boundary sliding. The presence of melt has not been suggested by any of the detailed petrological or geochemical studies of the Glenelg eclogite described here (Sanders, 1988, 1989; Storey, 2002). Triple junctions generally contain single-phase inclusions that have similar dimensions to the grains of the same phase in layers outside the garnet. Triple junction geometries may be due simply to the impingement of faceted garnet interfaces around second-phase inclusions (Elliott et al., 1997; Holness et al., 2005) and do not necessitate that these inclusions were melt.
The role of fracture
The garnet porphyroclasts are clearly fractured and some fractures are kinematically linked to the shearing. They contain minerals that suggest that they formed at the same conditions as the plastic deformation of the garnet. Trepmann & Stöckhert (2002) suggested, for Sesia Zone garnets, that some fractures are developed along pre-existing low-angle grain boundaries and we cannot rule this out in the Glenelg garnets. However, the bending of fractures into shear zones, coupled with the plastic deformation of the garnet that forms the fracture walls, suggests that some fractures pre-date some of the plastic shearing. One suggestion is that fractures relate to decompression from eclogite to lower pressures (G. Cressey, personal communication). Given the kinematic links, it seems probable that fracturing and plastic deformation were broadly synchronous. This is an issue that warrants further investigation and is beyond the scope of this paper. Trepmann & Stöckhert (2002) have suggested that such fractures could represent a response to deep crustal seismic events. If this is true here, then the plastic deformation may correspond to pre-seismic stress build-up and post-seismic relaxation. Alternatively, the fractures may be sub-critical (Prior, 1993), and may accommodate deformation aseismically, essentially by processes that involve diffusion.
|
CONCEPTUAL MODEL |
|---|
|
TOPABSTRACTINTRODUCTIONTECTONOMETAMORPHIC SETTINGANALYTICAL METHODSMICROSTRUCTURAL CONTEXT OF...CONCEPTUAL MODELCONCLUSIONSREFERENCES |
|---|
We present here a conceptual model that explains best the data presented in this paper. The original garnet porphyroclasts are relics of the earlier eclogite-facies metamorphism. Deformation occurs in the amphibolite facies.