Chapter 6: Phase equilibria modelling of complex coronas in pelitic granulites from the Vredefort Dome

Transcription

1 Chapter 6: Phase equilibria modelling of complex coronas in pelitic granulites from the Vredefort Dome 6.1 Introduction The capacity of a rock to attain equilibrium is governed by complex interdependent relationships between evolving composition, grain size, temperature and heating and/or cooling rate (Carlson, 2002). The coronas of the Steynskraal traverse commonly exhibit evidence for temperature dependence of reaction extent and a strong sectoral development indicative of highly localized compositional domains. The complexity of the coronas provides a unique opportunity to evaluate controls on extent of corona development and degree of equilibration. The petrography of each coronal compositional domain has been rigorously described in Chapter 4 and diffusion modelling undertaken for the garnet-quartz compositional domain in a range of temperature and compositional scenarios in Chapter 5. Key petrographic features and findings of the diffusion modelling are summarised in Table 6.1. Embedded within the physico-chemical diffusion models and textural and modal configuration of each corona domain is information constraining the controls on equilibration extent, equilibration volume and evolution of bulk composition with prolonged reaction. Phase equilibria modelling in THERMOCALC may be used to extract this information and assess the dependence of equilibration and compositional evolution of coronal domains across the Steynskraal traverse on postshock cooling history, length-scales of diffusion, and evolving composition. Phase equilibria modelling of coronas is based on the premise that coronas form by steady-state, diffusion-controlled growth (Joesten, 1977; Mongkoltip and Ashworth, 1983; Foster, 1981; Grant, 1988; Johnson and Carlson, 1990; Carlson and Johnson, 1991; Ashworth and Birdi, 1990; Ashworth et al., 1992; Ashworth and Sheplev, 1997; Markl et al., 1998; Ashworth et al., 1998). It is instructive to review how corona formation occurs within the context of a pseudosection to ascertain what information may reliably be deduced from phase diagrams regarding reaction mechanism and equilibration volume (Figure 6.1). Changing pressure and temperature induces reaction between metastable reactants, e.g., garnet and quartz. 422

2 With incipient reaction, variable intergranular diffusivities of major components manifest as different length-scales of diffusion, such that the corona domain (e.g., garnet-quartz domain) is partitioned into a continuum of compositional subdomains or incipient effective bulk compositions (EBC1, 2 and 3 in Figure 6.1) in which local equilibrium is attained, each with unique chemical potentials (e.g., Figure 5.51 and Figure 6.1b). A layered corona assemblage develops across which a transient chemical potential gradient exists (Figure 6.1b, c). The nature of chemical potential differences between subdomains or incipient bulk compositions is demonstrated on a schematic chemical potential plot (Figure 6.1b). Prevailing chemical potential gradients across the corona drive diffusion through the layers. With prolonged reaction and enhanced intergranular diffusion, component flux through the corona layers equalises chemical potentials at all points in the corona. Local incipient bulk compositions of subdomains at layer boundaries gradually expand with mass transfer across layers and approach the final, steady-state, effective bulk composition for the corona as a whole (red star in Figure 6.1b and Figure 6.1d). Equilibrium is attained when no chemical potential gradients exist for any components, despite the spatial segregation of corona phases in layers as reaction is arrested at the solidus. Open-system diffusion modelling of the Steynskraal coronas (Chapter 5) suggests that at best the coronas only approach equilibrium, such that the equilibrium effective bulk composition for the entire corona domain is never attained, and the corona remains chemically partitioned into subdomains or effective bulk compositions (EBCs) in local equilibrium, between which chemical potential gradients exist. The variation in phase mode and composition across the corona reflects the prevailing chemical potential gradients. Mineral zonation and modal trends within the corona yield constraints on the degree of compositional partitioning into subdomains, which, in turn allows the extent of chemical communication between subdomains and equilibration within the corona domain as a whole to be assessed. This has been demonstrated for the garnet-quartz corona in Chapter

3 Table 6.1: Summary of coronal assemblage and textural trends Thickness Increase in layer thickness in all coronas from the NW group rocks to the SE group rocks Increase in layer thickness from core to rim coronas Layer thickness is greater in SK9-CLM than VT206A in the SE group. Locally, biotite-rich enclaves within VT206B are characterised by thicker corona layers than typical coronas in biotite-poor VT206A No systematic or consistent change in layer thickness with X Mg in the NW group Layer thickness is greatest in garnet-biotite coronas compared to all other domains for a given sample Corona layer thickness is greater where the peak assemblage is finer-grained compared to coarse-grained assemblages in the same sample Modal Proportions The mode of cordierite increases, orthopyroxene decreases and spinel decreases in all coronas from the NW group to the SE group. Plagioclase modes are higher in the SE group garnetbiotite and garnet-k-feldspar coronas compared to the NW group corona counterparts. In contrast, plagioclase modes decrease in garnet-quartz core and rim coronas from the NW group to the SE group. Results are equivocal in the garnet-plagioclase domain, due to difficulty in discerning product plagioclase from reactant plagioclase in the SE group. In all domains except garnet-plagioclase, the mode of cordierite increases, orthopyroxene decreases, plagioclase increases and spinel decreases from core to rim coronas in the NW group and the SE group. In garnet-plagioclase coronas, the mode of plagioclase decreases toward the garnet rim. Biotite and K-feldspar occur as subordinate phases in garnet-quartz and garnet-plagioclase domains in the SE group, indicative of open-system behaviour In general, the proportion of orthopyroxene relative to cordierite is greater in SK8C than SK6C from the NW group in all domains (except the garnet-biotite domain, where no difference was discernible) The proportion of orthopyroxene relative to cordierite is greater in VT206A than in SK9- CLM from the SE group rocks Cordierite modes are higher than predicted by pseudosections in all domains except the garnet-quartz domain where cordierite mode is lower than modelled and orthopyroxene and plagioclase modes are higher than expected Opx Vermicule /Grain Size Increase in vermicule size in all domains, core and rim coronas from the NW group to the SE group rocks No significant difference in vermicule size from core to rim coronas for the NW group rocks (with a marginal increase in SK6C garnet-k-feldspar and garnet-plagioclase domains) In the SE group rocks, vermicule size increases from core to rim coronas No significant difference in vermicule size with changing X Mg between SK6C and SK8C 424

4 Table 6.1 continued: Summary of coronal assemblage and textural trends Vermicule Spacing Increase in vermicule spacing in all domains, core and rim coronas from the NW group to the SE group rocks In the NW group rocks, vermicule spacing remains unchanged from core to rim coronas (with a marginal increase in SK6C garnet-k-feldspar and garnet-plagioclase domains) In the SE group rocks, vermicule spacing increases from core to rim coronas No significant difference in vermicule spacing with changing X Mg between SK6C and SK8C (~1 µm) in the NW group Distribution Tendency for reduced clustering/aggregation of vermicules in the SE group coronas vs. NW group coronas Shape Vermicules become more equant and idiomorphic in all domains from the NW group to the SE group coronas (especially in garnet-k-feldspar and garnet-plagioclase coronas) and from core to rim coronas in the NW group Orientation Alignment of vermicules perpendicular to layer boundaries is less pronounced in the SE group rocks compared to the NW group coronas for all domains. In the NW group, the preferred orientation of vermicules is diminished in rim coronas compared to core coronas. Diffusion Models Reaction affinity for the garnet-quartz corona domain decreases towards the garnet rim. Reaction affinity for the garnet-quartz coronas is reduced from the NW group to the SE group rocks. 425

5 Figure 6.1: Phase equilibria modelling of coronas. (a) With incipient reaction, different intergranular diffusion rates and hence length-scales of diffusion partition the corona compositional domain into subdomains in local equilibrium. (b) Each local equilibrium subdomain has a unique assemblage and associated chemical potentials. Chemical potential gradients exist between local equilibria in subdomains and drive diffusion through the corona as a whole. 426

6 Figure 6.1 continued: Phase equilibria modelling of coronas. (c) A layered corona assemblage develops, such that, each layer represents a sub-domain in local equilibrium with a distinct composition (X in the schematic T-X pseudosection). (d) With prolonged reaction, diffusion through the corona equalizes chemical potentials throughout the domain and the equilibration volume expands to include all layers. The final equilibrium assemblage includes all phases in the corona band as a whole, despite their segregation spatially. 427

7 Despite the fact that equilibrium for the corona as a whole is unlikely to be attained, it is still possible to identify final, steady-state effective bulk compositions or reaction domains in which equilibration is approached. The corona domains cannot be strictly defined as equilibration volumes but rather chemical communication volumes, i.e., portions of the rock mass that approach equilibrium through diffusion of components down prevailing chemical potential gradients. Given sufficient time and appropriate intergranular diffusion rates, the communication volume will become the equilibration volume. We have chosen to treat the communication volume for a corona domain as a proxy for the potential equilibration volume for the purposes of modelling. Comparison of observed modes and compositions of phases in coronas from across the Steynskraal traverse with those predicted by theoretical phase diagrams may thereby be used to identify petrographicallyconsistent compositional domains for the corona as a whole in which equilibrium is approached (Figure 6.1c). The degree of communication (or mixing) between corona compositional domains and the extent of equilibration is constrained by lengthscales of diffusion for major components across the traverse as a function of postshock temperature and local bulk compositional changes with prolonged reaction. Melt loss from the equilibration volume plays a crucial role in the preservation of the corona granulite facies assemblage (White and Powell, 2002). White and Powell (2002) quantitatively demonstrated that the loss of a silicate melt into which H 2 O has partitioned during dehydration melting effectively removes H 2 O from the equilibration volume, thereby preventing re-equilibration of the anhydrous, refractory granulite facies residuum during retrogression. As a consequence, the solidus is substantially elevated in the refractory bulk composition. Rocks will only re-equilibrate in younger metamorphic events if the solidus in the drier restitic composition is crossed under new metamorphic conditions (White and Powell, 2002). Since the Vredefort granulites have lost significant melt during an Archaean pre-impact metamorphic event (Chapter 2), the textural and compositional evolution of the coronas is assumed to entail partial equilibration in the phase fields immediately above the solidus. The textural evolution of the coronas ceases at or immediately below the solidus with only minor retrograde Fe-Mg compositional reequilibration or resetting at temperatures below this. An important corollary of this assertion is that layer thickness can be considered indicative of the extent of textural 428

8 evolution of the rock above the solidus and, as such, a proxy for reaction affinity. This is supported by diffusion models of the garnet-quartz corona domain in Chapter 5, in which thickest coronas (SK9A, B and C) exhibit lowest reaction affinities. 6.2 Bulk composition of corona domains The effective bulk composition of each domain is determined by combining peak assemblage phases for the bulk rock predicted in THERMOCALC, in an appropriate manner to yield an effective bulk composition of a compositional domain consistent with petrographic observations (Table 6.2). To simulate a garnet- matrix corona EBC, i.e., garnet in communication with all matrix phases, 50 mol% garnet is added to the matrix phases recalculated to make up the remaining portion of the EBC, whilst honouring their original proportions and compositions in the peak assemblage. A garnet-dominated end-member EBC was constructed using 95 mol% peak garnet with only 5 mol% remaining matrix phases to accommodate contribution of matrix phases included within garnet or ingression of components from the matrix along fractures in the garnet. End-member corona compositional domains were constructed with 50 mol% garnet, 49 mol% reactant and 1 mol% matrix to accommodate minor interaction with the matrix and allow modelling in NCKFMASHTO for all corona domains. Johnson et al. (2004) employed a similar technique to model melting relationships during cordierite-spinel symplectite replacement of andalusite with diapiric rise of the Phepane Dome beneath the Bushveld Complex. Compositions in Table 6.2 and on pseudosections throughout Chapter 6 are those normalised and reported by THERMOCALC at 2 decimal places. Rounding to two decimal places may misleadingly yield totals greater or less than 100 mol% by 0.01 mol%. Actual compositions modelled do total 100 mol% at 3 decimal places. 429

9 Table 6.2: End-member bulk compositions for phase equilibria modelling of corona domains (NCKFMASHTO) NW GROUP End-member bulk composition H 2O SiO 2 Al 2O 3 CaO MgO FeO K 2O Na 2O TiO 2 O X Mg SK6C garnet (95 mol% Grt, 5 mol% matrix) garnet-matrix (50 mol% Grt, 50 mol% matrix) garnet-quartz (50 mol% Grt, 49 mol% Qtz, 1 mol% matrix) garnet-biotite (50 mol% Grt, 49 mol% Bt, 1 mol% matrix) garnet-k-feldspar* (50 mol% Grt, 49 mol% Ksp, 1 mol% matrix) (mol%) unmodified garnet-plagioclase (50 mol% Grt, 49 mol% Pl, 1 mol% matrix) garnet-cordierite (50 mol% Grt, 49 mol% Crd, 1 mol% matrix) SK8C garnet (95 mol% Grt, 5 mol% matrix) garnet-matrix (50 mol% Grt, 50 mol% matrix) garnet-quartz (50 mol% Grt, 49 mol% Qtz, 1 mol% matrix) garnet-biotite (50 mol% Grt, 49 mol% Bt, 1 mol% matrix) garnet-k-feldspar* (50 mol% Grt, 49 mol% Ksp, 1 mol% matrix) unmodified garnet-plagioclase (50 mol% Grt, 49 mol% Pl 1 mol% matrix)

10 Table 6.2 continued: End-member bulk compositions for phase equilibria modelling of corona domains (NCKFMASHTO) SE GROUP End-member bulk composition H 2O SiO 2 Al 2O 3 CaO MgO FeO K 2O Na 2O TiO 2 O X Mg (mol%) SK9-CLM garnet (95 mol% Grt, 5 mol% matrix) garnet-matrix (50 mol% Grt, 50 mol% matrix) garnet-quartz (50 mol% Grt, 49 mol% Qtz, 1 mol% matrix) garnet-biotite (50 mol% Grt, 49 mol% Bt, 1 mol% matrix) garnet-k-feldspar* (50 mol% Grt, 49 mol% Ksp, 1 mol% matrix) unmodified garnet-plagioclase (50 mol% Grt, 49 mol% Pl 1 mol% matrix) garnet-cordierite (50 mol% Grt, 49 mol% Crd, 1 mol% matrix) VT206A garnet (95 mol% Grt, 5 mol% matrix) garnet-matrix (50 mol% Grt, 50 mol% matrix) garnet-quartz (50 mol% Grt, 49 mol% Qtz, 1 mol% matrix) garnet-biotite (50 mol% Grt, 49 mol% Bt, 1 mol% matrix) *Modified EBC composition from that predicted in THERMOCALC better accommodates observed plagioclase modes 431

11 End-member EBCs for all corona domains were constructed for SK6C and SK8C from the NW group. Similarly, end member EBCs for corona domains were defined for SK9-CLM and VT206A from the SE group. Major compositional differences between end-member EBCs are demonstrated by Harker Plots in Figure 6.2 and a Si-Fe-Mg cation plot (Figure 6.3). The garnet-quartz end-member EBC domain in all samples is the richest in SiO 2 (~ 71 mol%) compared to other domains, followed by garnet-k-feldspar (~ 58 mol%), and then garnet-plagioclase (~ 56 mol%). End-member garnet and garnet-biotite domains are most depleted in silica (40-44 mol%), with the garnet-cordierite domain at intermediate silica contents (~ 48 mol%). There is a spread of SiO 2 contents ( mol%) for the garnet-matrix domains principally reflecting variation in the modal proportion of quartz in the peak assemblage for individual samples. Accordingly, sample SK9-CLM, with highest vol% quartz in the peak assemblage, has the most silica-rich garnet-matrix domain and VT206A, with the least quartz in the peak assemblage, has the most silica-depleted garnet-matrix domain (Chapter 4, Table 4.8). Alumina contents are highest in the garnet-cordierite domains ( mol%) and lowest in the garnet-quartz domains (< 7.3 mol%). The garnet-matrix domain contains Al 2 O 3 contents ranging between 9.4 mol% in VT206A and 11.9 mol% in SK6C. The garnet-biotite and garnet-cordierite domains are the most magnesian domains, reflecting the magnesian nature of cordierite and biotite relative to garnet. For individual samples, the X Mg of the garnet-quartz, garnet-plagioclase, garnet-kfeldspar and garnet-dominated domains are almost identical since the ferromagnesian component in the domain is inherited from garnet component alone (apart from a minor matrix contribution of 1 mol% in the quartz and feldspar domains and 5 mol% in the garnet-dominated domain). Bulk rock composition controls X Mg variation between samples for a specific corona domain, such that SK6C domains are most magnesian, followed by SK9-CLM, then SK8C and, finally, most Fe-rich in VT206A (Chapter 2, Tables 2.1 and 2.2). 432

12 Figure 6.2: Harker plots demonstrating compositional differences between corona end-member EBCs. a) Al 2 O 3 vs. SiO 2. b) CaO vs. SiO 2. c) Na 2 O vs. SiO 2. d) K 2 O vs. SiO 2. e) FeO+MgO vs. SiO 2. f) X Mg vs. SiO 2. g) H 2 O vs. SiO

13 Figure 6.3: Si Fe Mg cation plot comparing effective bulk compositions for corona endmember domains for the NW group rocks and the SE group rocks. 434

14 Variation in alkali contents between the domains reflects differences in the modal proportions and compositions of ternary plagioclase and ternary K-feldspar in the matrix of each sample (Figure 6.2 and 6.3). CaO contents are similarly a function of feldspar modes and compositions as well as garnet composition. The garnet-kfeldspar domain is the most potassic domain ( mol%), with lesser Na 2 O (~ 2.5 mol%) and CaO (< 1 mol%) contents. The CaO contents predicted from THERMOCALC were not quite calcic enough to reproduce the modes of plagioclase observed in garnet-k-feldspar coronas. The predicted compositon was perturbed slightly to better accommodate observed corona product modes, i.e., CaO was increased iteratively by < 0.5 mol% for each sample. Garnet-biotite domains are slightly less potassic ( mol%), with negligible Na 2 O and CaO contents of less than 0.15 mol%. Lowest K 2 O contents occur in the garnet-quartz, garnetdominated and garnet-cordierite domains (< 0.1 mol%). Garnet-plagioclase domains are most enriched in Na 2 O ( mol%), followed by garnet-k-feldspar domains ( mol%). Remaining domains comprise negligible Na 2 O contents. Garnet-plagioclase domains are richest in CaO ( mol%). Garnetmatrix and garnet-k-feldspar domains exhibit CaO molar amounts between 0.47 and 1.45 mol%. The garnet-quartz, garnet-biotite, garnet-cordierite and garnet domains are depleted in CaO compared to the garnet-feldspar domains, since CaO is contributed only from the garnet component in these domains. CaO content in garnet-quartz, garnet-biotite and garnet-cordierite domains for SK8C, SK6C and SK9-CLM is < 0.15 mol% and marginally higher in the garnet-dominated domain (< 0.3 mol%). This latter is due to the slightly higher CaO contribution from matrix plagioclase in the 5 mol% matrix component to the garnet-dominated domain, as opposed to 1 mol% in the garnet-biotite, garnet-cordierite and garnet-quartz domains. All VT206A domains are enriched in CaO, reflecting the higher bulk rock CaO contents compared to other samples. The most hydrous domain is the garnet-biotite domain (> 5 mol%). The garnetcordierite domain comprises slightly less water at 2.15 to 2.77 mol% H 2 O. The H 2 O content in the garnet-matrix domain is inherited only from biotite and, potentially, cordierite in the matrix. H 2 O content in this domain ranges from 0.24 to 1.29 mol%. The garnet-quartz, garnet-feldspar and garnet-dominated domains are essentially anhydrous. The garnet-dominated domain comprises slightly elevated H 2 O content 435

15 (~ 0.1 mol%) compared to other anhydrous domains owing to a marginally greater matrix component (5 mol%) than in other end-member EBCs. The most hydrous garnet-biotite domain occurs in VT206A (5.49 mol%), followed by SK8C (5.32 mol%), then SK6C (5.23 mol%) and, finally, SK9-CLM (5.24 mol%), reflecting variation in the amount of H 2 O in the biotite. Similarly, higher H 2 O content in the SK6C garnet-cordierite domain compared to SK9-CLM garnet-cordierite domain is attributed to relative differences in H 2 O content of peak cordierite between samples. Variation in H 2 O content for garnet-matrix domains between samples reflects variation in biotite and cordierite mode in the peak assemblage with only a minor contribution by variation in H 2 O content of peak biotite and cordierite between samples. Accordingly, the garnet-matrix domain is most hydrous in samples with highest modes of biotite and cordierite in the peak assemblage, i.e., SK6C (1.22 mol%) and SK9-CLM (1.29 mol%). Lowest H 2 O content in the garnet-matrix domain occurs in VT206A (0.24 mol %), where biotite mode in the peak assemblage is lowest. End-member bulk compositions are combined by means of a T-X pseudosection, thereby generating a range of compositions between the end-members, simulating greater or lesser communication between the end-member domains. T-X pseudosections were constructed in NCKFMASHTO and NCKFMASH between these model compositional domains using THERMOCALC 3.26 and the dataset of Holland and Powell (22 November 2003 update of Holland and Powell, 1998, dataset file tcds55.txt), to constrain the influence of post-shock temperature and cooling rate on the effective bulk composition of a corona domain, communication between domains and extent of equilibration. Post-impact metamorphic conditions are static and are unlikely to have involved a change in pressure, hence P-X pseudosections are not constructed for the coronas. T-X pseudosections were constructed for a pressure of 3 kbar, broadly consistent with non-equilibrium thermobarometric estimates (Section 5.3.9), conventional thermobarometric estimates (Stevens et al., 1997b) of corona formation, numerical modelling (Ivanov, 2005) and estimates of post-impact erosion of 8-10 km (McCarthy et al., 1990; Gibson et al., 1998). 436

16 6.3 Phase equilibria modelling of the matrix contribution to the garnet-core compositional domain Investigation of compositional controls on garnet breakdown in the core and at the rim of porphyroblasts with increased communication with matrix phases required the construction of a series of T-X pseudosections from the garnet-matrix endmember EBC (50 mol% garnet + 50 mol% matrix) to the garnet-dominated EBC (95 mol% garnet + 5 mol% matrix). T-X pseudosections were constructed for SK6C (bulk rock X Mg = 0.55) and SK8C (bulk rock X Mg = 0.49) from the NW group (Figure 6.4). In the SE group rocks, T-X pseudosections were constructed for SK9- CLM (bulk rock X Mg = 0.52) and VT206A (bulk rock X Mg = 0.37) (Figure 6.5). Each T-X pseudosection was constructed for a range of temperatures from 650 to 1100 C, with the compositional variable X ranging from a matrix-rich EBC (50 mol% garnet in communication with 50 mol% matrix) at X = 0 to garnet-dominated EBC (95 mol% garnet and 5 mol% matrix) at X = 1. In the NW group rocks (Figure 6.4), bulk compositions comprising the axes of SK8C are less aluminous, more Fe-rich and less hydrous than SK6C. These compositional differences manifest in some minor differences in the topology of the T-X pseudosections (Figure 6.4) and include a solidus in SK8C that is on average 5-10 C higher than in SK6C. The spinel-out line in SK6C also occurs at slightly lower values of X than in SK8C, implying marginally lower external fluxes of matrix components into the garnet core are required to destabilise spinel in SK8C. This is primarily attributed to the intrinsically more subaluminous nature of the SK8C whole-rock bulk composition. Similarly, at any intermediate composition in SK8C the mode of orthopyroxene relative to cordierite is higher than in SK6C. In both SK6C and SK8C, as X decreases from garnet-dominated EBCs at X = 1.0 to X = , i.e., the proportion of matrix in the effective bulk composition increases, the mode of cordierite increases relative to orthopyroxene and reaches a maximum just before the spinel-out line is crossed. With greater contribution of matrix to the effective bulk composition, approaching 50 mol% (X 0), the modes of cordierite and orthopyroxene decrease sympathetically, although orthopyroxene decreases marginally more rapidly (more closely spaced isopleths). Consequently, the cordierite:orthopyroxene ratio continues to increase toward X = 0.0. The near- 437

17 coincident solidi in T-X space for SK6C and SK8C imply reaction is arrested in both samples at similar temperatures. In the SE group rocks (Figure 6.5), the bulk composition comprising the matrix-rich effective bulk composition for VT206A is markedly less aluminous, more Fe-rich and less hydrous than SK9-CLM. The VT206A garnet-dominated end-member EBC (X = 1.0) is similar compositionally in all respects to SK9-CLM except for the significant Fe enrichment (29 mol% FeO vs. 23 mol% in SK9-CLM). The solidus in VT206A is elevated with respect to the solidus in SK9-CLM by 25 C at X = 1 and by as much as 105 C at X = 0. The elevation in the solidus is primarily due to reduced H 2 O content in all effective bulk compositions for VT206A. The spinel-out line occurs in VT206A effective bulk compositions at values of X ~ for a temperature range of C (Figure 6.5b). In contrast, the spinel-out line in SK9-CLM occurs at comparatively lower values of X ~ 0.38 at 900 C immediately above the solidus (Figure 6.5a), i.e., greater communication is required with the relatively more aluminous matrix of SK9-CLM to enrich the effective bulk compositions with enough silica to destabilise spinel. For the same X, the predicted mode of orthopyroxene relative to cordierite in VT206A is higher than that predicted and observed for SK9-CLM, indicative of more siliceous bulk compositions for VT206A compared to SK9-CLM. Within garnet-dominated effective bulk compositions (X > 0.5), the mode of orthopyroxene increases as cordierite decreases. As X approaches 0, to the left of the spinel-out line, the modes of cordierite and orthopyroxene decrease sympathetically in both SK9- CLM and VT206A, however the rate at which cordierite decreases relative to orthopyroxene is different for these samples. In SK9-CLM, cordierite modes decrease less rapidly with increasing matrix contribution compared to orthopyroxene since cordierite isopleths are not as steeply dipping in T-X space as orthopyroxene isopleths, such that the cordierite:orthopyroxene ratio increases. In contrast, predicted cordierite modes in VT206A decrease more rapidly than orthopyroxene as the garnet rim is approached and the effective bulk composition becomes more relatively depleted in Al 2 O 3 compared to SK9-CLM, reflecting the subaluminous matrix assemblage in VT206A. As a result, in VT206A the cordierite:orthopyroxene ratio decreases (Figure 6.5). 438

18 Petrographic observation of symplectite developed in the core and at the margin of garnet is in agreement with trends predicted by the T-X pseudosections for both the NW group and the SE group rocks (Chapter 4). In the NW group rocks, fracture symplectite in the core of garnet is characterised by an orthopyroxene:cordierite:spinel ratio approaching 0.67:0.22:0.11 (Section 4.2.7; Figure 4.23) consistent with an EBC range of 0.8 < X < 1.0 (Table 6.3). Solidus temperature ranges from C for SK6C and C for SK8C. Melt modes for a given suprasolidus temperature of 1000 C range from 0.07 (X = 0.8) to 0.02 (X = 1.0) for both SK6C and SK8C. No consistent difference in the ratio of cordierite relative to orthopyroxene between SK6C and SK8C fracture symplectite is observed. The spinel-out line, which marks the spatial limit of silica deficiency in garnet-dominated EBCs, manifests petrographically as a boundary between spinelpresent and spinel-absent cordierite-orthopyroxene symplectites. It also marks a distinct reduction in the mode of orthopyroxene and relative increase in the mode of cordierite towards the rim of the garnet. The spinel-out line occurs within µm of the inferred original garnet margin for the NW group rocks (Section 4.2.7; Figure 4.23). In SK9-CLM from the SE group rocks, fracture symplectite in the core of garnet is characterised by lower orthopyroxene:cordierite:spinel ratios approaching 0.54:0.36:0.1 (Section 4.3.7; Figure 4.47). Observed cordierite modes are anomalously high compared to the predicted modes. The best-fit petrographicallyconsistent EBC range is for SK9-CLM (Table 6.3). Orthopyroxene modes in VT206A are relatively higher than cordierite compared to SK9-CLM, in agreement with ratios predicted by the T-X pseudosections. In VT206A the orthopyroxene:cordierite ratio approaches 0.67:0.33 with minor spinel consistent with an EBC range of near-equilibration between 0.4 and 0.6 (Table 6.3). The range for solidus temperature in SK9-CLM is from 890 C (X = 0.4) to 910 C (X = 0.6). Melt modes in SK9-CLM range from 0.19 at X = 0.4 to 0.13 at X = 0.6 for a given post-shock temperature of 1000 C for the EBC range. Solidus temperatures in VT206A are much higher at C for X = Melt modes at 1000 C are lower than in SK9-CLM, ranging between 0.05 at X = 0.4 and 0.02 at X = 0.6. In SK9-CLM, the spinel-out line occurs up to 500 µm from the original garnet-matrix interface (Section 4.3.7; Figure 4.48), an order of magnitude greater than in the NW 439

19 group rocks. The more restricted occurrence of spinel in the SE group rocks toward the core of garnet and the more matrix-rich EBC ranges in which equilibration is approached are consistent with greater matrix contribution and greater length-scales of diffusion in the SE group rocks compared to the NW group rocks. Table 6.3: Observed and predicted corona product modes with attendant solidus temperatures for the fracture symplectite Observed Opx/Crd Best-fit Model Opx/Crd EBC Range Solidus Temperature Melt Modes (1000 C) SK6C C SK8C C SK9-CLM C VT206A C

20 Figure 6.4: Isobaric (3 kbar) T X pseudosections for the NW group rocks (a) SK6C and (b) SK8C. The bold black line represents the spinel-out line. With increasing garnet contribution to the EBC (X 1.0), the modal proportions of orthopyroxene and spinel increase whilst cordierite decreases. The black arrows represent petrographically-consistent EBC ranges in which equilibration is approached for SK6C and SK8C. The black circle on the solidus represents the predicted temperature at which reaction is arrested for the median of the EBC range. 441

21 Figure 6.5: Isobaric (3 kbar) T X pseudosections for the SE group rocks (a) SK9-CLM and (b) VT206A. The bold black line represents the spinel-out line. With increasing garnet contribution to the EBC (X 1.0), the modal proportions of orthopyroxene and spinel increase whilst cordierite decreases. The black arrows represent petrographically-consistent EBC ranges in which equilibration is approached for SK9-CLM and VT206A. The black circle on the solidus represents the predicted temperature at which reaction is arrested for the median of the EBC range. 442

22 6.4 Phase equilibria modelling of the garnet-reactant domains Two compositional ranges were defined to facilitate modelling of either increased matrix or garnet contribution to garnet-reactant corona domains, where the reactant is either quartz, biotite, plagioclase, K-feldspar or cordierite. The first of these compositional ranges models phase equilibria evolution in T-X space with increased matrix contribution to an end-member garnet-reactant EBC as the equilibration volume expands at higher post-shock temperatures and greater length-scales of diffusion to incorporate matrix phases. Additionally, it simulates increased matrix contribution to the garnet-reactant domain where the reactant occurs at the margin of the garnet and is, therefore, in closer proximity to matrix phases compared to inclusions in the core. T-X pseudosections were constructed from the garnet-matrix end-member EBC (50 mol% garnet + 50 mol% matrix) to the garnet-reactant endmember EBC (50 mol% garnet + 49 mol% reactant + 1 mol% matrix) for the NW group rocks (SK6C and SK8C) and the SE group rocks (SK9-CLM and VT206A). The second compositional range accounts for increasing garnet contribution to the garnet-reactant EBC. This compositional spectrum arises with increasing sequestration of a phase within the garnet core or as distance from an inclusion in garnet increases and the diffusion limits for components derived from the reactant are exceeded, such that garnet begins to dominate the effective bulk composition. A rim corona EBC may also become more garnet-rich as length scales of diffusion reduce with cooling, such that equilibration volumes shrink and partition downtemperature. T-X pseudosections were constructed from the garnet-reactant endmember EBC (50 mol% garnet + 49 mol% reactant + 1 mol% matrix) to the garnetdominated end-member EBC (95 mol% garnet + 5 mol% matrix) for the NW group rocks (SK6C and SK8C) and the SE group rocks (SK9-CLM and VT206A). The 5 mol% matrix accommodates the additional component flux from the matrix along fractures and other neighbouring inclusions in garnet. The garnet-cordierite compositional system was modelled only for SK6C and SK9- CLM, since cordierite is not present in the peak assemblage for SK8C and VT206A. T-X pseudosections for VT206A were only constructed for the garnet-quartz 443

23 compositional system since the feldspar and biotite domains are very poorly developed in this sample. For each garnet-reactant domain, the T-X pseudosections for both compositional ranges were merged to form a compound or fence T-X pseudosection, such that the mid-axis is the common garnet-reactant end-member EBC. The compositional range from X = -1.0 to 0.0 tracks from the garnet-matrix end-member EBC to the garnetreactant EBC. The range from X = 0.0 to 1.0 tracks from the end-member garnetreactant EBC to the garnet-dominated EBC. Presenting the T-X pseudosections in this way allows core and rim corona EBCs to be compared at the same time Garnet-quartz domain Fence T-X pseudosections for the SK6C and SK8C garnet-quartz corona domain are included in Figure 6.6. Fence T-X pseudosections for SK9-CLM and VT206A from the SE group rocks for the garnet-quartz corona domain are presented in Figure 6.7. In SK6C, and SK8C to a lesser extent, as the matrix contribution in an EBC increases (X -1.0) cordierite and orthopyroxene modes increase, although orthopyroxene increases less rapidly, manifesting as increased overall modes of cordierite relative to orthopyroxene. In the most matrix-dominated EBCs for SK8C, as X approaches -1.0, the cordierite mode increases less rapidly than orthopyroxene, resulting in increased predicted orthopyroxene relative to cordierite. Plagioclase and liquid modes increase in matrix-dominated domains for both SK6C and SK8C. Plagioclase modes are larger in SK6C with higher associated cordierite modes and lower orthopyroxene modes for all X compared to SK8C. The solidus is depressed substantially in both SK6C and SK8C with increasing matrix contribution to the EBC, from 967 C to 798 C in SK6C and 969 C to 803 C in SK8C (Figure 6.6). The solidus in SK6C is, at most, ~10 C lower than SK8C at X = For the entire range of X, the SK6C K-feldspar-out effective univariant occurs ~ 2 10 C below that for SK8C. The plagioclase-out effective univariant line occurs at a lower temperature in SK8C than SK6C for X = -1.0 to where molar content of CaO in SK8C EBCs is less than the corresponding EBC in SK6C. 444

24 At X > -0.22, the plagioclase-out line for SK6C occurs at lower temperatures than SK8C, where CaO content in SK8C EBCs exceeds that in SK6C. For a given temperature, higher molar H 2 O contents in SK6C compositions manifest as higher modelled melt proportions compared to SK8C, e.g., the SK8C 0.06 liquid isopleth is elevated by ~ 10 C for any X compared to SK6C. At X < in SK6C, biotite is stable just above the solidus in the coronal assemblage. The biotite univariant in SK8C does not intersect the solidus as in SK6C, implying no corona biotite is stable in the SK8C coronal assemblage, assuming that reaction was arrested at the solidus. In both SK8C and SK6C, as garnet contribution to the EBC increases (X 1.0), the modes of cordierite, orthopyroxene, plagioclase and liquid increase whilst quartz decreases rapidly (Figure 6.6). Orthopyroxene mode increases relatively more rapidly compared to cordierite and plagioclase, such that the proportion of orthopyroxene increases relative to other phases as the corona domain bulk composition becomes more garnet-dominated. For any X, the modes of orthopyroxene in SK8C exceed those in SK6C owing to the slightly more siliceous effective bulk compositions in SK8C for the same molar garnet contribution to the EBC. In both SK6C and SK8C, as garnet begins to dominate the effective bulk composition and spinel is stabilised in the assemblage (X > 0.74), the solidus is suddenly elevated to a higher temperature (> 975 C), with an associated sharp increase in orthopyroxene mode and decrease in cordierite mode. The respective T-X pseudosection topologies for SK9-CLM and VT206A are significantly different and reflect bulk-rock compositional differences (Figure 6.7). Increased matrix contribution manifests as an increase in cordierite mode relative to orthopyroxene in SK9-CLM. In contrast, enhanced matrix contribution to the EBC in VT206A is characterised by increased orthopyroxene mode relative to cordierite. Lower H 2 O contents in both garnet-quartz and garnet-matrix end-member EBCs result in a significantly elevated solidus in VT206A by as much as 105 C at X = -1.0 and ~ 15 C at X = 0.0, compared to SK9-CLM. Low H 2 O contents in VT206A EBCs arise because of the paucity of biotite in the matrix, such that the contribution of H 2 O to the EBC by the matrix is minimal. In VT206A, low K 2 O content means that K-feldspar is destabilised just above the solidus. In contrast, K-feldspar is stable over a broader temperature range above the solidus in SK9-CLM. Higher molar CaO 445

25 contents in VT206A ensure stability of plagioclase in the corona assemblage for the temperature range of interest (< 1100 C). The plagioclase-out line for SK9-CLM occurs at temperatures > 1010 C for all X. For the entire compositional range, orthopyroxene modes are higher and cordierite correspondingly lower in VT206A compared to SK9-CLM. Liquid modes at all temperatures may be up to an order of magnitude higher in SK9-CLM compared to VT206A as a result of a higher molar H 2 O content in SK9-CLM EBCs. With increasing garnet contribution to the EBC in the SE group samples (X 1.0), orthopyroxene modes increase relative to cordierite. Higher H 2 O contents in SK9- CLM flux the solidus by up to 100 C compared to VT206A for the compositional range 0.0 < X < 0.72 (Figure 6.7). Plagioclase is stable at all temperatures and all compositions in VT206A, owing to higher molar CaO for all X compared to SK9- CLM. As observed in the NW group rocks, there is an inflection in the solidus temperature from 958 C to 988 C where X > 0.72 in SK9-CLM, beyond the quartzout effective univariant. The quartz-out line is shifted to higher X values in VT206A owing to its less aluminous effective bulk compositions compared to SK9-CLM, with a consequently larger contribution of garnet required for destabilisation of quartz. The associated solidus inflection in VT206A is not as marked as in SK9- CLM ( T is ~ 13 C), such that for 0.74 < X < 0.92, the solidus is at a lower temperature than SK9-CLM. Detailed petrographic investigation of the garnet-quartz domain (Chapter 4, Sections and 4.3.1) confirms the trends predicted from phase equilibria modelling, although the exact mineral modes are not reproduced exactly in the pseudosections. Cordierite modes are marginally higher relative to orthopyroxene in SK6C and SK9- CLM compared to SK8C and VT206A, as predicted by the T-X pseudosections. All samples exhibit increased cordierite and plagioclase modes relative to orthopyroxene in rim coronas compared to core coronas, consistent with increased matrix contribution in rim coronas. In contrast, core coronas are characterised by higher orthopyroxene modes relative to cordierite and plagioclase, indicative of more garnet-dominated EBCs in the T-X pseudosections. Best-fit petrographicallyconsistent EBC ranges for all samples for both core and rim coronas are included in Table 6.4. In general, observed orthopyroxene and plagioclase modes are elevated compared to those predicted in the T-X pseudosection. Solidus temperatures and 446

26 melt modes at 1000 C for respective EBC ranges for each sample are also included in Table 6.4. Lowest solidus temperatures and highest melt modes at 1000 C are attained in rim coronas from SK9-CLM with greatest matrix contribution to the EBC. Table 6.4: Observed and predicted corona product modes with attendant solidus temperatures for the garnet-quartz domain CORE Observed Model Best-fit Opx/Crd Opx/Pl Opx/Crd Opx/Pl EBC Range Solidus Temperature Melt Modes (1000 C) SK6C C 0.02 SK8C C 0.02 SK9-CLM C RIM Observed Model Best-fit Opx/Crd Opx/Pl Opx/Crd Opx/Pl EBC Range Solidus Temperature Melt Modes (1000 C) SK6C C SK8C C SK9-CLM C VT206A C 0.01 <

27 Figure 6.6: Combined garnet-matrix, garnet-quartz and garnet T-X pseudosections for (a) SK6C (NW group), contoured for phase proportions. The black and grey arrows represent petrographically-consistent EBC ranges in which equilibration is approached for core and rim coronas respectively. The black and grey circles on the solidus represent the predicted temperatures at which reaction is arrested for the median of the core and rim EBC ranges. 448

28 Figure 6.6 continued: Combined garnet-matrix, garnet-quartz and garnet T-X pseudosections for (b) SK8C (NW group), contoured for phase proportions. The black and grey arrows represent petrographically-consistent EBC ranges in which equilibration is approached for core and rim coronas respectively. The black and grey circles on the solidus represent the predicted temperatures at which reaction is arrested for the median of the core and rim EBC ranges. 449

29 Figure 6.7: Combined garnet-matrix, garnet-quartz and garnet T-X pseudosections for (a) SK9-CLM (SE group), contoured for phase proportions. The black and grey arrows represent petrographically-consistent EBC ranges in which equilibration is approached for core and rim coronas respectively. The black and grey circles on the solidus represent the predicted temperatures at which reaction is arrested for the median of the core and rim EBC ranges. 450

30 Figure 6.7 continued: Combined garnet-matrix, garnet-quartz and garnet T-X pseudosections for (b) VT206A (SE group), contoured for phase proportions. The black and grey arrows represent petrographically-consistent EBC ranges in which equilibration is approached for core and rim coronas respectively. The grey circles on the solidus represent the predicted temperatures at which reaction is arrested for the median of the core and rim EBC ranges. 451

31 6.4.2 Garnet-biotite domain Fence T-X pseudosections for the garnet-biotite domain have been constructed for SK6C and SK8C from the NW group rocks (Figure 6.8) and SK9-CLM from the SE group rocks (Figure 6.9). Rim garnet-biotite domains are not developed in VT206A and, therefore, interaction between the garnet-biotite domain and the matrix was not modelled. For the compositional range -1.0 < X < 0.0, the immediate suprasolidus fields for SK6C, SK8C and SK9-CLM are contoured in detail for modes of cordierite, orthopyroxene, spinel, biotite, plagioclase and melt in Figures 6.10, 6.11 and 6.12, respectively. With increasing matrix contribution to the EBC, modes of melt, cordierite, orthopyroxene and plagioclase increase for any suprasolidus temperature relative to spinel and biotite (Figures 6.10, 6.11 and 6.12). The mode of cordierite increases more rapidly than orthopyroxene, such that the ratio of cordierite to orthopyroxene increases as matrix contribution to the EBC increases. For all samples, the solidus is depressed with increasing contribution of matrix components Na 2 O, CaO and SiO 2 to the EBC, demonstrating the dependence of melt fertility on these components in addition to ah 2 O. In the NW group rocks, the solidus is slightly depressed in SK8C compared to SK6C (maximum T = 20 C at X = -0.62). This is partly attributed to higher H 2 O contents in SK8C in EBCs approaching X = 0 but, as X approaches -1.0, where H 2 O in SK6C is higher than in SK8C, solidus depression in SK8C is likely to reflect lower closure temperatures in relatively more Fe-rich SK8C EBCs. Melt modes are higher in SK8C for all suprasolidus temperatures compared to SK6C. Cordierite modes in SK6C are higher for any given suprasolidus temperature, reflecting the more aluminous, Mg-rich EBCs in SK6C compared to SK8C. For all values of X, the plagioclase-out effective univariant in SK8C occurs at temperatures up to 20 C lower than in SK6C. Similarly, for X > -0.92, the K-feldspar-out univariant is depressed in SK8C compared to SK6C by up to 10 C. The occurrence of effective univariants for both plagioclase and K-feldspar at lower temperatures in SK8C may be attributed primarily to lower Al 2 O 3 and Na 2 O contents of SK8C EBCs for all X, and lower CaO contents for X < compared to SK6C. The biotite-out effective univariant for all X occurs at lower temperatures in SK8C than in SK6C, reflecting the less aluminous nature of SK8C EBCs compared to their SK6C 452

32 counterparts. The spinel-out line is very similar in T-X space for both SK6C and SK8C, apart from a 15 C elevation immediately above the solidus in SK6C from X = -0.9 to At temperatures greater than 875 C, the spinel-out line occurs at marginally less negative X values in SK6C than in SK8C. This is due to the more Fe-rich nature of SK8C EBCs for any X, i.e., marginally lower contribution of matrix biotite is required to stabilise Fe-rich hercynitic spinel in SK8C. With increasing contribution of garnet to the EBC (X = ), the proportion of orthopyroxene and spinel increase relative to cordierite in all samples with negligible plagioclase (Figures 6.8 and 6.9). The solidus for SK8C is relatively depressed by ~ 10 C for X = compared to SK6C. This is attributed to slightly higher H 2 O contents and higher X Fe for SK8C bulk compositions for X < 0.92, compared to SK6C. For all X, spinel and cordierite modes are higher in SK6C, reflecting more MgO-rich and Al 2 O 3 -rich bulk compositions compared to SK8C. Detailed petrographic investigation has been undertaken for the garnet-biotite domain (Chapter 4, Sections and 4.3.2). Observed cordierite modes are anomalously higher than those predicted, such that observed cordierite/orthopyroxene ratios are not attainable on the T-X pseudosection. Orthopyroxene/spinel and orthopyroxene/plagioclase ratios are more readily accommodated in the model. Best-fit petrographically-consistent EBCs are proposed based on the latter ratios and included in Table 6.5. The predicted relative differences in modal mineralogy of the coronas between SK8C and SK6C are not consistently observed petrographically. However, in all samples, the increase in the proportion of spinel and orthopyroxene relative to cordierite in core coronas is consistent with increasing garnet contribution to the EBCs as X 1.0. Conversely, higher cordierite and plagioclase modes relative to spinel and orthopyroxene suggest increasing matrix contribution in rim coronas. Similarly, higher cordierite modes in SK9-CLM core and rim coronas compared to their NW group counterparts, are also consistent with increased matrix contribution to both core and rim corona domains in the SE group. Matrix-rich EBCs are associated with lower attendant solidus temperatures and higher melt modes at 1000 C (Table 6.5). 453

33 Table 6.5: Observed and predicted corona product modes with attendant solidus temperatures for the garnet-biotite domain CORE Observed Model Best-fit Opx/Crd Opx/Spl Opx/Pl Opx/Crd Opx/Spl Opx/Pl EBC Range Solidus Temperature Melt Modes (1000 C) SK6C > C SK8C > C SK9-CLM C RIM Observed Model Best-fit Opx/Crd Opx/Spl Opx/Pl Opx/Crd Opx/Spl Opx/Pl EBC Range Solidus Temperature Melt Modes (1000 C) SK6C C SK8C C SK9-CLM C

34 Figure 6.8: Combined garnet-matrix, garnet-biotite and garnet T-X pseudosections for (a) SK6C from the NW group. The black and grey arrows represent petrographically-consistent EBC ranges for core and rim coronas respectively. The black and grey circles on the solidus represent the predicted temperatures at which reaction is arrested for the median of the core and rim EBC ranges. 455

35 Figure 6.8 continued: Combined garnet-matrix, garnet-biotite and garnet T-X pseudosections for (b) SK8C from the NW group. The black and grey arrows represent petrographically-consistent EBC ranges for core and rim coronas respectively. The black and grey circles on the solidus represent the predicted temperatures at which reaction is arrested for the median of the core and rim EBC ranges. 456

36 Figure 6.9: Combined garnet-matrix, garnet-biotite and garnet T-X pseudosections for SK9-CLM from the SE group. The black and grey arrows represent petrographically-consistent EBC ranges for core and rim coronas respectively. The black and grey circles on the solidus represent the predicted temperatures at which reaction is arrested for the median of the core and rim EBC ranges. 457

37 Figure 6.10: SK6C Suprasolidus fields for the garnet-matrix vs. garnet-biotite T-X pseudosection contoured for mineral modes. Biotite and spinel modes increase, whilst cordierite, orthopyroxene and plagioclase modes decrease toward the end-member garnet-biotite compositional domain (X = 0) from the matrix-rich domain (X = -1.0). 458

38 Figure 6.11: SK8C Suprasolidus fields for the garnet-matrix vs. garnet-biotite T-X pseudosection contoured for mineral modes. Biotite and spinel modes increase, whilst cordierite, orthopyroxene and plagioclase modes decrease toward the end-member garnet-biotite compositional domain (X = 0) from the matrix-rich domain (X = -1.0). 459

39 Figure 6.12: SK9-CLM Suprasolidus fields for the garnet-matrix vs. garnet-biotite T-X pseudosection contoured for mineral modes. Biotite and spinel modes increase, whilst cordierite, orthopyroxene and plagioclase modes decrease toward the end-member garnet-biotite compositional domain (X = 0) from the matrix-rich domain (X = -1.0). 460

40 6.4.3 Garnet-feldspar domain Fence T-X pseudosections for the garnet-plagioclase and garnet-k-feldspar domains for the NW group (SK6C, SK8C) are included in Figures 6.13 and 6.14, respectively. Fence T-X pseudosections for the garnet-plagioclase and garnet-kfeldspar domains from the SE group (SK9-CLM) are included in Figure 6.15a, b. Modes of melt and cordierite increase for any suprasolidus temperature with increasing matrix contribution to the EBC relative to spinel and orthopyroxene. With increasing garnet contribution to the garnet-feldspar EBC, orthopyroxene and spinel increase relative to cordierite. The bulk composition used for the pure end-member garnet-k-feldspar domain, is slightly modified from that produced in THERMOCALC (Table 6.2). Plagioclase modes are underestimated when using the THERMOCALC composition to model the observed corona product modes. The THERMOCALC composition was, thus, perturbed slightly by elevating molar CaO by 0.5 mol% and renormalising so as to better reproduce the plagioclase modes observed. The reason for this may stem from error in the bulk-rock analysis used to derive the peak mineral compositions, but it is more likely that Ca is preferentially introduced into the corona from the matrix through open-system metasomatic exchange with the matrix (Section 6.7). Topological differences between SK6C and SK8C for both garnet-k-feldspar and garnet-plagioclase fence T-X pseudosections are similar. Except for the interval between X = and -0.60, the solidus of SK6C is slightly depressed by up to ~ 10 C compared to SK8C for X = -1.0 to X = 0.0, reflecting higher H 2 O contents in SK6C domain compositions. Solidus temperatures for SK6C and SK8C approach each other at X = 0.0 where the difference in H 2 O contents is minimised (~ 0.01 mol%). The modal proportion of melt in SK8C and SK6C is similar for the compositional range < X < At X < -0.84, the mode of liquid in SK6C exceeds that in SK8C for the same temperature, reflecting the more hydrous SK6C bulk compositions. At X > -0.2 to X = 1.0, the mode of melt in SK8C is greater than that in SK6C for a given suprasolidus temperature, reflecting greater melt fertility in lower X Mg bulk compositions. Stevens et al. (1997a) similarly noted greater melt fertility in lower X Mg rocks in experimental investigation of partial melting of pelites and metagreywackes. The plagioclase-out line is up to 15 C lower in SK8C than in 461

41 SK6C in both garnet-k-feldspar and garnet-plagioclase domains. Increased cordierite, plagioclase and spinel modes in SK6C compared to SK8C reflect more aluminous effective bulk compositions for SK6C for the entire EBC range. Petrographic investigation of the garnet-feldspar domains from the NW group and the SE group (Chapter 4, Sections 4.2.3; 4.2.4; and 4.3.4) are in agreement with phase relationships in T-X space derived above. Observed orthopyroxene modes are higher relative to cordierite in SK8C compared to SK6C in garnetfeldspar coronas. Rim coronas are consistently characterised by higher observed cordierite modes relative to orthopyroxene modes in both the NW and SE group coronas. In contrast, core coronas are richer in orthopyroxene and spinel relative to cordierite in all samples. SK9-CLM core and rim coronas are similarly more cordierite-rich than their NW group counterparts. Petrographically-consistent EBCs are included in Table 6.6. Rim corona EBCs and all EBCs from the SE group coronas exhibit increased matrix contribution, with higher associated melt modes at 1000 C and lower solidus temperatures. 462

42 Table 6.6: Observed and predicted corona product modes with attendant solidus temperatures for the garnet-feldspar domains Garnet-Plagioclase CORE Observed Model Best-fit Opx/Crd Opx/Spl Opx/Crd Opx/ Spl EBC Range Solidus Temperature Melt Modes (1000 C) SK6C C SK8C C 0.02 SK9-CLM C RIM Observed Model Best-fit Opx/Crd Opx/ Spl Opx/Crd Opx/ Spl EBC Range Solidus Temperature Melt Modes (1000 C) SK6C C SK8C C SK9-CLM 0.38 negligible spinel 1.27 no spinel C Garnet-K-feldspar CORE Observed Model Best-fit Opx/Crd Opx/ Spl Opx/Crd Opx/ Spl EBC Range Solidus Temperature Melt Modes (1000 C) SK6C C 0.02 SK8C C SK9-CLM C RIM Observed Model Best-fit Opx/Crd Opx/ Spl Opx/Crd Opx/ Spl EBC Range Solidus Temperature Melt Modes (1000 C) SK6C C SK8C C SK9-CLM no spinel C

43 Figure 6.13: Combined garnet-matrix, garnet-plagioclase and garnet T-X pseudosections for (a) SK6C from the NW group. The black and grey arrows represent petrographically-consistent EBC ranges for core and rim coronas respectively. The black and grey circles on the solidus represent the predicted temperatures at which reaction is arrested for the median of the core and rim EBC ranges. 464

44 Figure 6.13 continued: Combined garnet-matrix, garnet-plagioclase and garnet T-X pseudosections for (b) SK8C from the NW group. The black and grey arrows represent petrographically-consistent EBC ranges for core and rim coronas respectively. The black and grey circles on the solidus represent the predicted temperatures at which reaction is arrested for the median of the core and rim EBC ranges. 465

45 Figure 6.14: Combined garnet-matrix, garnet-k-feldspar and garnet T-X pseudosections for (a) SK6C from the NW group. The black arrow grey arrows represent petrographically-consistent EBC ranges for core and rim coronas respectively. The black and grey circles on the solidus represent the predicted temperatures at which reaction is arrested for the median of the core and rim EBC ranges. 466

46 Figure 6.14 continued: Combined garnet-matrix, garnet-k-feldspar and garnet T-X pseudosections for (b) SK8C from the NW group. The black and grey arrows represent petrographically-consistent EBC ranges for core and rim coronas respectively. The black and grey circles on the solidus represent the predicted temperatures at which reaction is arrested for the median of the core and rim EBC ranges. 467

47 Figure 6.15: Fence T-X pseudosections for SK9-CLM (SE group). (a) Combined garnet-matrix, garnet-plagioclase and garnet T-X pseudosections. The black and grey arrows represent petrographically-consistent EBC ranges for core and rim coronas respectively. The black and grey circles on the solidus represent the predicted temperatures at which reaction is arrested for the median of the core and rim EBC ranges. 468

48 Figure 6.15 continued: Fence T-X pseudosections for SK9-CLM (SE group). (b) Combined garnet-matrix, garnet-k-feldspar and garnet T-X pseudosections. The black and grey arrows represent petrographically-consistent EBC ranges for core and rim coronas respectively. The black and grey circles on the solidus represent the predicted temperatures at which reaction is arrested for the median of the core and rim EBC ranges. 469

49 6.4.4 Garnet-cordierite domain Fence T-X pseudosections for the garnet-cordierite domain for the NW group (SK6C) and the SE group (SK9-CLM) are included in Figure As X decreases from 0.0 to -1.0, the proportion of matrix increases in the EBC, such that H 2 O, FeO, MgO, X Mg and Al 2 O 3 decrease, whilst SiO 2, CaO, Na 2 O, K 2 O and TiO 2 increase. Immediate suprasolidus fields for SK6C and SK9-CLM are contoured for modes of cordierite, orthopyroxene, spinel, and melt. Modes of melt increase for any suprasolidus temperature with increasing matrix contribution to the EBC, whilst spinel and cordierite decrease. Orthopyroxene mode increases slowly in SK6C ( ) and SK9-CLM ( ) for suprasolidus temperatures with increasing matrix contribution to the EBC. Consequently, the orthopyroxene:cordierite ratio increases as X approaches -1.0 and matrix contribution to the EBC is maximised. The solidus is most depressed at X = -1.0 (798 C) and increases in temperature as X approaches 0.0 to 851 C in SK6C and 888 C in SK9-CLM. Solidus elevation may be attributed to reduction in CaO and SiO 2 as the cordierite contribution to the EBC increases, despite the increase in ah 2 O in cordierite-rich domains (X = 0.0). With increasing contribution of garnet to the EBC (X 1.0), the ratio of orthopyroxene relative to cordierite increases from 0.6 to > 2.7 in both SK6C and SK9-CLM, whilst the modal proportion of spinel increases from 0.06 to The solidus is depressed in hydrous garnet-cordierite bulk compositions from 978 C in the garnet-dominated EBCs in SK6C and SK9-CLM. Progressively higher melt proportions for a given suprasolidus temperature are predicted for more cordieritedominated compositions. Petrographically-consistent EBC ranges for core and rim coronas from SK6C and SK9-CLM are included in Table 6.7. Matrix contribution increases in rim coronas with marginally lower attendant solidus temperatures and higher melt modes at 1000 C compared to garnet-dominated core corona EBC ranges in SK6C. EBC ranges for SK9-CLM core and rim coronas are very similar with associated solidus temperatures of 870 C in both. The relative matrix contribution in SK9-CLM core coronas is greater than core coronas in SK6C with marginally elevated associated melt modes at a given post-shock temperature of 1000 C. The solidus temperature 470

50 is not depressed in SK9-CLM from core to rim coronas and, in fact, is marginally elevated compared to SK6C from the NW group. Table 6.7: Observed and predicted corona product modes with attendant solidus temperatures for the garnet-cordierite domain CORE Observed Model Best-fit Opx/Crd Opx/Spl Opx/Crd Opx/Spl EBC Range Solidus Temperature Melt Modes (1000 C) SK6C C 0.04 SK9- CLM C RIM Observed Model Best-fit Opx/Crd Opx/Spl Opx/Crd Opx/Spl EBC Range Solidus Temperature Melt Modes (1000 C) SK6C C SK9- CLM ~

51 Figure 6.16: Combined garnet-matrix, garnet-cordierite and garnet T-X pseudosections for (a) SK6C (NW group). The black and grey arrows represent petrographically-consistent EBC ranges for core and rim coronas respectively. The black and grey circles on the solidus represent the predicted temperatures at which reaction is arrested for the median of the core and rim EBC ranges. 472

52 Figure 6.16 continued: Combined garnet-matrix, garnet-cordierite and garnet T-X pseudosections for (b) SK9-CLM (SE group). The black and grey arrows represent petrographically-consistent EBC ranges for core and rim coronas respectively. The black and grey circles on the solidus represent the predicted temperatures at which reaction is arrested for the median of the core and rim EBC ranges. 473

53 6.5 Phase equilibria modelling of communication between contiguous corona domains The degree of mixing between contiguous corona domains depends on the lengthscales of intergranular diffusion for major components. Intergranular diffusion is, in turn, dependent on temperature and volume of melt present. A coupled petrographic and phase equilibria modelling approach is used to assess the degree of communication between domains in the NW and SE groups, as well as any variation in diffusive mixing of corona domains between each group. Coronas in the coarsest portions of the peak assemblage most isolated from additional matrix phases were examined, so as to minimise matrix contribution to the juxtaposed corona domains. Contiguous corona domains in the NW group rocks are characterised by rapid morphological and modal transitions over the original inferred contact between domains (Figure 6.17). Garnet-biotite domains adjacent to garnet-quartz domains (Figure 6.17a) typically comprise a cordierite-orthopyroxene-spinel symplectite grading into a cordierite-biotite or plagioclase-biotite symplectite and, finally, reequilibrated biotite replacing reactant biotite. Corona biotite (delimited by the yellow border in Figure 6.17a) and spinel are present only within the garnet-biotite domain. Spinel and biotite do not occur within the garnet-quartz domain bordering the garnet-biotite domain in the NW group rocks. Neighbouring garnet-quartz and garnet-feldspar domains from SK8C in the NW group rocks are characterised by a marked change in the morphology and mode of orthopyroxene at the inferred original interface between reactants. In Figure 6.17b, blocky, equigranular, monomineralic orthopyroxene within the garnet-quartz domain changes abruptly to a lamellar orthopyroxene network intricately intergrown with product plagioclase in the garnet-plagioclase domain. Spinel is observed only within the garnet-plagioclase domain immediately adjacent to the inferred original interface between plagioclase and quartz reactants. The mode and morphology of orthopyroxene intergrown with plagioclase in the garnet-feldspar domains adjacent to garnet-biotite domains also changes sharply from a lamellar intergrowth to bleblike and granular in adjacent garnet-biotite domains (Figure 6.17c). Corona biotite is not present in the garnet-feldspar domains adjacent to the garnet-biotite domain. 474

54 Figure 6.17: Modal and textural relationships between contiguous corona domains in the NW group. (a) BSE image of adjacent garnet-quartz and garnet-biotite domains in sample SK8C. The absence of corona biotite (delineated by yellow border in the garnet-biotite corona domain) in the garnet-quartz domain suggests negligible communication between the domains in the NW group rocks. (b) BSE image of juxtaposed garnet-quartz and garnet-plagioclase corona domains from SK8C. Rapid inflection in orthopyroxene mode and morphology and the occurrence of spinel in the garnet-plagioclase domain immediately adjacent to the inferred original contact between reactants (yellow arrow) suggest marked compositional partitioning between coronas with limited or negligible mixing or communication between domains. 475

55 Figure 6.17 continued: Modal and textural relationships between contiguous corona domains in the NW group. (c) BSE image of juxtaposed garnet-plagioclase and garnet-biotite compositional domains in SK6C. There is a sharp change in orthopyroxene morphology across the contact between plagioclase and biotite. Corona biotite is absent from the garnet-plagioclase domain. (d) BSE image of juxtaposed garnet-k-feldspar and garnet-plagioclase compositional domains in SK6C. The sharp change in orthopyroxene morphology and modes across the contact between plagioclase and K-feldspar, and restriction of high orthopyroxene modes to the garnet-k-feldspar domain suggests negligible mixing or communication between adjacent domains. 476

56 In contiguous garnet-k-feldspar and garnet-plagioclase corona domains (Figure 6.17d), the mode of layer 2 orthopyroxene increases sharply over a distance of < 20 µm as the EBC shifts from plagioclase-dominated to K-feldspar-dominated. The morphology of orthopyroxene similarly changes from bleb-like anhedral laths in the plagioclase-dominated domain to a more cuneiform intergrowth of layer 2 orthopyroxene vermicules in the K-feldspar-dominated domain over the original reactant interface. Contiguous corona domains in the SE group are characterised by smoother modal and morphological transitions across domain boundaries compared to the NW group coronas (Figure 6.18). Indicator minerals (e.g., biotite or K-feldspar in the garnetquartz corona domain) that require open-system exchange with adjacent corona domains for stability may be used to track the length-scales of diffusion of components between adjacent domains. The ubiquitous presence of corona biotite at the contact between layer 2 plagioclase and the cordierite-orthopyroxene symplectite is characteristic of both rim and core garnet-quartz domains adjacent to garnetbiotite domains in SK9-CLM (Figure 6.18a). Where a garnet-quartz corona domain abuts a garnet-k-feldspar domain, K-feldspar occurs between blocky orthopyroxene and the plagioclase layers (Figure 6.18b). In garnet-feldspar and garnet-biotite domains adjacent to garnet-quartz corona domains, spinel is rare or absent whilst cordierite modes in the garnet-quartz domain are elevated compared to the NW group coronas. Corona biotite occurs intergrown with orthopyroxene and plagioclase in layer 2 orthopyroxene-plagioclase symplectite from both garnet-k-feldspar and garnet-plagioclase domains adjacent to garnet-biotite domains (Figure 6.18c). The boundary between garnet-k-feldspar and garnet-plagioclase domains is diffuse with a gradual increase in orthopyroxene mode toward the garnet-k-feldspar domain and a smooth change in vermicule morphology from euhedral rhombs in the garnetplagioclase domain to idiomorphic laths and blades in the garnet-k-feldspar domain (Figure 6.18c). K-feldspar blebs are commonly scattered through the plagioclaseorthopyroxene layer in the garnet-plagioclase domain whilst plagioclase modes in adjacent garnet-k-feldspar domains are elevated compared to the NW group corona domains. 477

57 Figure 6.18: Modal and textural relationships between contiguous corona domains in the SE group. (a) BSE image of adjacent garnet-quartz and garnetbiotite domains in sample SK9-CLM from the SE group rocks. Corona biotite in the garnet-quartz domain occurs intergrown with the cordieriteorthopyroxene symplectite or with plagioclase adjacent to blocky orthopyroxene. K-feldspar occurs as a discontinuous layer in the garnet-quartz domain between the plagioclase and blocky orthopyroxene layers. Spinel is rare or absent from the garnet-biotite corona assemblage. 478

58 Figure 6.18 continued: Modal and textural relationships between contiguous corona domains in the SE group. (b) BSE image of adjacent garnet-quartz and garnet-k-feldspar domains in SK9- CLM. In the garnet-quartz domain, K-feldspar occurs as a layer between the plagioclase and blocky orthopyroxene layers, decreasing in thickness as distance from the neighbouring garnet-kfeldspar domain increases. Spinel is absent from the garnet-k-feldspar domain assemblage. (c) BSE image of juxtaposed garnet-k-feldspar and garnet-plagioclase compositional domains from SK9-CLM. The garnet-plagioclase domain is characterised by comparatively low orthopyroxene modes, high plagioclase modes and isolated, discrete bleb-like orthopyroxene vermicules. In contrast, the K-feldspar-dominated domain comprises higher orthopyroxene proportions with tabular to lath-like layer 2 orthopyroxene vermicules. A transitional zone (~ 100 µm in extent) of diffusional mixing occurs between the domains in which orthopyroxene mode gradually increases. 479

59 A series of T-X pseudosections between corona end-member domains was constructed for SK6C and SK8C from the NW group rocks and SK9-CLM for the SE group rocks. Intermediate compositions between the extreme end-member EBCs comprising the axes of the T-X pseudosections represent subordinate EBCs of mixing between the adjacent domains. These hybrid domains are characterised by diagnostic mineral modes, allowing petrographically-consistent EBC ranges for adjacent domains to be constrained and the extent of mixing between domains to be assessed Garnet-quartz vs. garnet-biotite domains T-X pseudosections governing the compositional spectrum between end-member garnet-quartz (50 mol% garnet + 49 mol% quartz + 1 mol% matrix) and garnetbiotite (50 mol% garnet + 49 mol% biotite + 1 mol% matrix) corona EBCs from SK8C in the NW group and SK9-CLM from the SE group are included in Figure Intermediate, hybrid compositional domains track from a quartz-dominated EBC at X = 0.0 to more potassic, ferromagnesian and hydrous compositions toward X = 1.0 as the proportion of biotite relative to quartz increases in the EBC. At X > 0.18, corona biotite becomes a stable diagnostic product phase, which is absent in end-member garnet-quartz EBCs. The mode of biotite increases rapidly from 0 to 0.25 from X = 0.18 to X = At X > 0.8, spinel is stable above the solidus in the corona assemblage. Intermediate, hybrid compositions are characterised by solidus depression to 828 C at X = 0.5 in SK8C and 832 C in SK9-CLM. In comparison, the solidus is elevated in the garnet-quartz end-member EBC and end-member garnet-biotite EBC (Figure 6.19). Petrographically-consistent EBC ranges, solidus temperatures and attendant melt modes at a post-shock temperature of 1000 C for adjacent garnet-quartz and garnetbiotite domains are included in Table 6.8 (pg. 494). The spatial limits of spinel and biotite occurrence in contiguous garnet-quartz and garnet-biotite coronas from SK8C and SK9-CLM correspond to the spinel-in and biotite-in lines from the T-X pseudosections in Figure

60 Figure 6.19: T-X pseudosections for the compositional range between end-member garnet-quartz and adjacent end-member garnet-biotite EBCs for (a) SK8C (NW group) for a temperature range of C and (b) C. Intermediate compositions represent hybrid domains attained by diffusional mixing between end-member EBCs. The biotite-in line is delineated by the green line; the spinel-in line is delineated by the yellow line and the solidus by the blue line. The grey arrows represent petrographically-consistent EBC ranges for adjacent coronas. The grey circles on the solidus represent the predicted temperatures at which reaction is arrested for the median of modelled EBC ranges. 481

61 Figure 6.19 continued: T-X pseudosections for the compositional range between end-member garnet-quartz and adjacent end-member garnet-biotite EBCs for (c) SK9-CLM (SE group) for a temperature range of C and (d) C. Intermediate compositions represent hybrid domains attained by diffusional mixing between end-member EBCs. The biotite-in line is delineated by the green line; the spinel-in line is delineated by the yellow line and the solidus by the blue line. The grey arrows represent petrographically-consistent EBC ranges for adjacent corona domains. The grey circles on the solidus represent the predicted temperatures at which reaction is arrested for the median of modelled EBC ranges. 482

62 In the NW group coronas, the restriction of biotite- and spinel-bearing EBCs to the immediate vicinity of reactant biotite is indicative of limited mixing between domains and an attained EBC range of 0.8 < X < 1.0 for the garnet-biotite domain. The absence of biotite in the adjacent garnet-quartz domain suggests that EBCs with X > 0.18 required to stabilise corona biotite in the garnet-quartz domain are never realised and the EBC range is restricted to 0.0 < X < 0.18 (Figure 6.19). In SK9- CLM from the SE group, the absence of spinel in the garnet-biotite domain and the presence of corona biotite in the garnet-quartz corona domain are diagnostic of significant mixing between these adjacent domains (Figure 6.18a). Biotite modes approaching 0.12 in the garnet-quartz corona suggest EBC ranges of 0.18 < X < 0.5 are attained. Adjacent garnet-biotite domains are constrained to an EBC range in which spinel is absent in immediate suprasolidus fields, i.e., 0.6 < X < Garnet-quartz vs. garnet-k-feldspar/plagioclase domains T-X pseudosections for the compositional spectrum between adjacent end-member garnet-quartz and garnet-k-feldspar corona EBCs from SK6C in the NW group and SK9-CLM from the SE group are included in Figure The compositional range between garnet-quartz and garnet-plagioclase juxtaposed corona domains is modelled in T-X pseudosections for SK8C from the NW group and SK9-CLM from the SE group in Figure The model chemical system adopted was NCKFMASH. A modified garnet-k-feldspar end-member EBC was used in the T-X pseudosections (Section 6.4.3). The solidus is marginally elevated in garnet-k-feldspar end-member EBCs (978 C in SK6C and 977 C in SK9-CLM) and in garnet-plagioclase end-member EBCs (986 C in SK8C and 992 C in SK9-CLM) compared to the garnet-quartz endmember EBC (964 C in SK6C, 965 C in SK8C and 959 C in SK9-CLM). The mode of K-feldspar increases from < in the garnet-quartz EBC to > 0.40 in the neighbouring garnet-k-feldspar EBC. Similarly, plagioclase mode increases rapidly in the garnet-quartz domain toward the garnet-plagioclase end-member EBC from 0.05 to 0.5. In spinel-absent phase fields (X < 0.84), cordierite and orthopyroxene isopleths are isothermal, i.e., independent of composition. In more feldspardominant EBCs (X > 0.84), spinel is stable and cordierite mode decreases sharply 483

63 toward zero. In contrast, orthopyroxene mode remains roughly constant across the suprasolidus phase fields for all X. Higher modes of orthopyroxene and plagioclase/k-feldspar relative to cordierite in the presence of spinel thus characterise feldspar-dominated EBCs with limited diffusional communication with garnet-quartz domains. Petrographically-consistent EBC ranges, solidus temperatures and melt modes at 1000 C for contiguous garnet-quartz and garnet-feldspar domains are included in Table 6.8. In the NW group, the occurrence of spinel in garnet-feldspar domains immediately adjacent to the inferred original interface between reactants is consistent with an EBC range of near-equilibration approaching X > 0.84 in garnet- K-feldspar domains (Figure 6.20) and X > 0.86 in garnet-plagioclase domains (Figure 6.21). In contiguous garnet-k-feldspar and garnet-quartz domains, the absence of K-feldspar in the garnet-quartz domain constrains the attained EBC of the neighbouring garnet-quartz domain to X < 0.1 (Figure 6.20). Modes of plagioclase approaching 0.05 in garnet-quartz domains adjacent to garnetplagioclase domains similarly indicate an EBC range of near-equilibration of X < 0.1 (Figure 6.21). In the SE group coronas, the absence of spinel in the garnet-feldspar domains suggests pervasive silica diffusion from the neighbouring garnet-quartz domain, such that silica-undersaturated, feldspar-dominated EBCs are not realised. Plagioclase modes in excess of 0.05 in the garnet-k-feldspar domain indicate an approximate EBC range of near-equilibration for garnet-k-feldspar EBCs in the SE group of 0.6 < X < 0.84 (Figure 6.20). Reduction in plagioclase mode from 0.5 to 0.3 in garnet-plagioclase domains adjacent to garnet-quartz domains is consistent with an EBC range of 0.6 < X < 0.86 (Figure 6.21). The presence of K-feldspar (< 0.15) in the garnet-quartz domain adjacent to a garnet-k-feldspar domain invokes EBCs between 0.0 < X < 0.3 (Figure 6.20). Similarly, elevated modes of plagioclase exceeding 0.1 constrain an EBC of near-equilibration to 0 < X < 0.3 in the garnetquartz domains adjacent to garnet-plagioclase domains (Figure 6.21). 484

64 Figure 6.20: T-X pseudosections for the compositional range between end-member garnet-quartz and adjacent end-member garnet-k-feldspar EBCs for (a) SK6C (NW group) contoured for cordierite and orthopyroxene; and (b) contoured for K-feldspar and plagioclase. Intermediate compositions represent hybrid domains attained by diffusional mixing between end-member EBCs. The grey arrows represent petrographically-consistent EBC ranges for adjacent coronas. The grey circles on the solidus represent the predicted temperatures at which reaction is arrested for the median of modelled EBC ranges. 485

65 Figure 6.20 continued: T-X pseudosections for the compositional range between end-member garnet-quartz and adjacent end-member garnet-k-feldspar EBCs for (c) SK9-CLM (SE group) contoured for cordierite and orthopyroxene; and (d) contoured for K-feldspar and plagioclase. Intermediate compositions represent hybrid domains attained by diffusional mixing between endmember EBCs. The grey arrows represent petrographically-consistent EBC ranges for adjacent coronas. The grey circles on the solidus represent the predicted temperatures at which reaction is arrested for the median of modelled EBC ranges. 486

66 Figure 6.21: T-X pseudosections for the compositional range between end-member garnet-quartz and adjacent garnet-plagioclase EBCs for (a) SK8C (NW group) and (b) SK9-CLM (SE group). Intermediate compositions represent hybrid domains attained by diffusional mixing between endmember EBCs. The grey arrows represent petrographically-consistent EBC ranges for adjacent coronas. The grey circles on the solidus represent the predicted temperatures at which reaction is arrested for the median of modelled EBC ranges. 487

67 6.5.3 Garnet-K-Feldspar/plagioclase vs. garnet-biotite domains T-X pseudosections have been constructed for the compositional range between adjacent garnet-k-feldspar and garnet-biotite domains for SK6C from the NW group rocks and SK9-CLM from the SE group rocks (Figure 6.22). T-X pseudosections for the garnet-plagioclase to garnet-biotite domain from SK6C and SK9-CLM from the NW group and the SE group rocks, respectively, are included in Figure The model chemical system is NCKFMASHTO. The unmodified THERMOCALC garnet-k-feldspar EBC was used in modelling (Table 6.2). With increasing biotite contribution to the EBC, for the compositional range 0.0 < X < 1.0, i.e., from the garnet-plagioclase/k-feldspar domain to the garnet-biotite domain, the proportion of feldspar decreases in suprasolidus fields from 0.45 to ~ with increasing biotite mode, whilst orthopyroxene and cordierite modes remain essentially unchanged with shifting composition (Figures 6.22 and 6.23). Petrographically-consistent EBC ranges, solidus temperatures and melt modes at 1000 C for contiguous garnet-feldspar and garnet-biotite domains are included in Table 6.8. In the NW group coronas, the biotite-out line (bold line in Figures 6.22 and 6.23) restricts biotite-absent, feldspar-dominated corona bulk compositions from SK6C in the NW group rocks to 0.0 < X < 0.1. Trace feldspar in the garnet-biotite domain constrains a potential EBC range to 0.9 < X < 1.0 (Figures 6.22b and 6.23a). In SK9-CLM from the SE group rocks, the presence of corona biotite within the boundaries of the garnet-plagioclase or garnet-k-feldspar domain and elevated feldspar modes approaching within the garnet-biotite domain indicate EBC ranges of near-equilibration of 0.1 < X < 0.4 for the garnet-feldspar domain and 0.7 < X < 0.9 for the contiguous garnet-biotite domain in SK9-CLM (Figures 6.22d and 6.23b). 488

68 Figure 6.22: T-X pseudosections for the compositional range between end-member garnet-kfeldspar and adjacent end-member garnet-biotite EBCs for (a) SK6C (NW group), contoured for cordierite, orthopyroxene and spinel; and (b) contoured for K-feldspar and biotite. Intermediate compositions represent hybrid domains attained by diffusional mixing between end-member EBCs. The grey arrows represent petrographically-consistent EBC ranges for adjacent coronas. The grey circles on the solidus represent the predicted temperatures at which reaction is arrested for the median of modelled EBC ranges. 489

69 Figure 6.22 continued: T-X pseudosections for the compositional range between end-member garnet-k-feldspar and adjacent end-member garnet-biotite EBCs for (c) SK9-CLM (SE group), contoured for cordierite, orthopyroxene and spinel; and (d) contoured for K-feldspar and biotite. Intermediate compositions represent hybrid domains attained by diffusional mixing between endmember EBCs. The grey arrows represent petrographically-consistent EBC ranges for adjacent coronas. The grey circles on the solidus represent the predicted temperatures at which reaction is arrested for the median of modelled EBC ranges. 490

70 Figure 6.23: T-X pseudosection for the compositional range between end-member garnetplagioclase and adjacent garnet-biotite EBCs for (a) SK6C (NW group) and (b) SK9-CLM (SE group). Intermediate compositions represent hybrid domains attained by diffusional mixing between end-member EBCs. The grey arrows represent petrographically-consistent EBC ranges for adjacent coronas. The grey circles on the solidus represent the predicted temperatures at which reaction is arrested for the median of modelled EBC ranges. 491

71 6.5.4 Garnet-K-Feldspar vs. garnet-plagioclase domains T-X pseudosections constructed for the compositional range between garnet-kfeldspar and garnet-plagioclase domains for SK6C from the NW group rocks and SK9-CLM from the SE group rocks are presented in Figure T-X pseudosections were modelled in NCKFMASH. Modified garnet-k-feldspar EBCs were used in the model (Table 6.2). The compositional range tracks from 0.0 (garnet-k-feldspar domain) to +1.0 (garnet-plagioclase domain). Toward plagioclase-rich domains, the proportion of plagioclase increases in suprasolidus fields from 0.04 to > 0.48, K-feldspar decreases from > 0.4 to zero and cordierite mode increases slightly from 0.06 to Petrographically-consistent EBC ranges, solidus temperatures and melt modes at 1000 C for contiguous garnet-k-feldspar and garnet-plagioclase domains are included in Table 6.8. In garnet-k-feldspar domains abutting garnet-plagioclase domains from SK6C, plagioclase modes < 0.1 restrict the EBC range of nearequilibration to X < 0.2. Negligible K-feldspar in contiguous garnet-plagioclase domains constrains the EBC range to X > 0.8. In SK9-CLM from the SE group rocks, a comparatively lower mode of orthopyroxene and higher mode of plagioclase (~ ) relative to cordierite in the garnet-k-feldspar domain is consistent with diffusional communication with an adjacent neighbouring garnetplagioclase domain, such that a hybrid composition is attained (0.1 < X < 0.4). The occurrence of K-feldspar modes approaching 0.15 in garnet-plagioclase domains from SK9-CLM suggests the composition of the EBC tracks away from the endmember composition toward a composition richer in components derived from K- feldspar (0.6 < X < 1.0). 492

72 Figure 6.24: T-X pseudosections for the compositional range between end-member garnet-kfeldspar and adjacent garnet-plagioclase EBCs for samples (a) SK8C (NW group) and (b) SK9- CLM (SE group). Intermediate compositions represent hybrid domains attained by diffusional mixing between end-member EBCs. The grey arrows represent petrographically-consistent EBC ranges for adjacent coronas. The grey circles on the solidus represent the predicted temperatures at which reaction is arrested for the median of modelled EBC ranges. 493

73 Table 6.8: Best-fit model EBC ranges for contiguous corona domains Group Diagnostic modes Garnet-Quartz Best-fit EBC Solidus T C Melt modes (1000 C) Group NW Biotite absent < NW SE Biotite mode SE Garnet-Biotite Diagnostic Best-fit Solidus Melt modes modes EBC T C (1000 C) Spinel present, biotite mode > Spinel absent, biotite mode approaches 0.2 Group NW SE Diagnostic modes K-feldspar absent Ksp mode 0.15 Garnet-Quartz Best-fit EBC Solidus T C Melt modes (1000 C) Group Diagnostic modes Garnet-K-feldspar Best-fit EBC Solidus T C Melt modes (1000 C) < NW Spinel present > SE Spinel absent, Pl mode Garnet-Quartz Garnet-Plagioclase Group Diagnostic Best-fit Solidus Melt modes Diagnostic Best-fit Solidus Melt modes Group modes EBC T C (1000 C) modes EBC T C (1000 C) NW Pl mode <0.1 < NW Spinel present > SE Pl mode > SE Spinel absent, Pl mode Group Diagnostic modes Garnet-K-feldspar Best-fit EBC Solidus T C Melt modes (1000 C) NW No biotite < SE Biotite mode Group NW group SE Group Diagnostic modes Garnet-Biotite Best-fit EBC Solidus T C Melt modes (1000 C) Ksp mode < Ksp mode Garnet-Plagioclase Garnet-Biotite Group Diagnostic Best-fit Solidus Melt modes Diagnostic Best-fit Solidus Melt modes Group modes EBC T C (1000 C) modes EBC T C (1000 C) NW No biotite < NW Pl mode < SE Biotite mode SE Pl mode > Group Diagnostic modes Garnet-K-feldspar Best-fit EBC Solidus T C Melt modes (1000 C) Group NW Pl mode < 0.10 < NW SE Pl mode < SE Diagnostic modes Pl mode < 0.5, Ksp < 0.05 Ksp mode 0.15 Garnet-Plagioclase Best-fit EBC Solidus T C Melt modes (1000 C) > > For all contiguous corona domains, in the NW and SE groups, SE group corona domains typically exhibit greater mixing between adjacent domains, enhanced solidus depression and higher melt modes at a post-shock temperature of 1000 C. 494

74 6.6 Phase equilibria modelling of cordierite-biotite-sillimanite compositional domains In metapelitic samples SK6C and SK9-CLM, where cordierite comprises part of the Archaean equilibrium peak assemblage, a reaction texture between cordierite and included biotite and sillimanite is observable (Chapter 4; Sections and 4.3.6). The reaction texture is characterised by a spinel-cordierite symplectite after sillimanite grading sharply into a cordierite moat, fringed by a symplectitic intergrowth of biotite and cordierite replacing the original biotite reactant. In the NW group rocks, orthopyroxene may be intergrown with the cordierite-biotite symplectite after the biotite reactant. In contrast, orthopyroxene is notably absent or rare in the cordierite-biotite symplectite in SK9-CLM from the SE group rocks. Vermicule size and spacing in the symplectite from the NW group pseudomorphs are greater than in the SE group pseudomorphs (Sections and 4.3.6). T-X pseudosections for SK6C and SK9-CLM have been constructed for the compositional range between sillimanite-dominated, biotite+cordierite-bearing domains (35 mol% sillimanite + 30 mol% cordierite + 30 mol% biotite + 5 mol% remaining matrix phases) and biotite+cordierite-dominated domains (47.5 mol% cordierite mol% biotite + 5 mol% remaining matrix phases) in Figure In both SK6C and SK9-CLM, the immediate suprasolidus assemblage in sillimanitedominated EBCs (X < 0.74 in SK6C and X < 0.6 in SK9-CLM) is characterised by corundum stability. In the most biotite+cordierite-dominated EBCs, orthopyroxene becomes stable above the solidus (X > 0.96 in SK6C and X > 0.82 in SK9-CLM). The observed orthopyroxene-bearing, corundum-absent, spinel-rich assemblage of SK6C is stable at the most biotite+cordierite-rich compositions (X > 0.96; Table 6.9). In contrast, the absence of orthopyroxene and corundum in SK9-CLM assemblages constrains the EBC of near-equilibration to a more aluminous 0.6 < X < 0.82 (Table 6.9). 495

75 Table 6.9: Observed and predicted corona product modes with attendant solidus temperatures and melt modes for the cordierite-biotite-sillimanite domain SAMPLE Observed Model Best-fit Opx/Crd Crd/Bt Opx/Crd Crd/Bt EBC RANGE SOLIDUS TEMP MELT MODES (1000 C) SK6C (NW GROUP) SK9-CLM (SE GROUP) > 0.01 < 1.09 > C C The dominance of cordierite+biotite EBCs over the more aluminous sillimanitebearing EBCs in the near-equilibration volume at lower post-shock temperatures in the NW group (Table 6.9, Figure 6.25), indicates limited or restricted diffusion of components derived from sillimanite compared to more pervasive length-scales of diffusion for components from the more hydrous, melt-rich cordierite+biotitedominated domains. At higher post-shock temperatures in SK9-CLM, the equilibration volume for the pseudomorph becomes gradually more enriched in the components derived from the sillimanite end-member EBC, indicative of enhanced diffusion of Al 2 O 3 toward the centre of the Dome. 496

76 Figure 6.25: T-X pseudosections for the compositional range between sillimanite-dominated, biotite+cordierite and a sillimanite-absent, biotite+cordierite EBCs for (a) SK6C (NW group) and (b) SK9-CLM (SE group). In SK6C, the stability of orthopyroxene in the assemblage constrains an EBC range of near-equilibration of X > The absence of orthopyroxene and corundum from the SK9-CLM reaction texture suggests an EBC range of near-equilibration richer in sillimanite, i.e., 0.6 < X <

77 6.7 Non-linear, metasomatic modification and evolution of corona bulk composition with prolonged reaction A limitation of the phase equilibria modelling technique employed to simulate corona evolution is the inherent assumption that the composition of the effective bulk composition of the corona domain evolves in a linear manner with open-system communication with the surrounding system. All components are assumed to be perfectly mobile throughout the suprasolidus post-shock cooling history. In this chapter, the observed corona modes are compared with those predicted in T-X pseudosections assuming each corona domain evolves linearly and systematically through communication with the matrix or immediately contiguous domains as all components diffuse at the same rate, i.e., the exchange vector across the compositional spectrum is constant. In reality, open-system diffusion modelling of the garnet-quartz corona suggests some components (e.g., Fe, Mg, Ca) are markedly more mobile than others, resulting in relative enrichment or depletion in corona domains with prolonged reaction. This is evident in the need to selectively perturb the CaO molar content of the garnet-k-feldspar domain when modelling observed corona modes. The effective bulk composition of the corona domain evolves through a more complex, non-linear, open-system interaction with the surrounding rock than accommodated by a simple T-X pseudosection. Accordingly, the phase modes and compositions predicted by the T-X pseudosections may not accurately reflect those preserved in the natural corona and limit the confidence ascribed to proposed EBC ranges of near-equilibration. In the T-X pseudosections modelled previously, there is consistent mismatch between observed cordierite modes and those predicted in the T-X pseudosections. In general, cordierite modes in the garnet-feldspar and garnet-biotite domains tend to be higher than those accommodated in the T-X pseudosections (Figure 6.26). In the garnet-quartz domain, cordierite modes are anomalously lower whilst orthopyroxene and plagioclase modes are higher than those predicted in the model system (Figure 6.27). The degree of mismatch decreases in the SE group coronas relative to the NW group coronas. A reasonable explanation for mismatch, which is borne out in opensystem diffusion modelling in Chapter 5, is that, with prolonged reaction, it becomes less viable to assume each EBC behaves as a closed system, such that non-linear 498

78 open-system metasomatic exchange with respect to all components may occur between contiguous domains either due to preferential loss of components through diffusion, melt loss or gain. Figure 6.26: Comparison of observed and predicted Opx/Crd ratios for corona products from aluminous corona compositional domains (a) garnet-biotite, (b) garnet-plagioclase, (c) garnet-k- feldspar, (d) garnet-cordierite and (e) garnet fracture symplectite. In general, the model Opx/Crd ratios exceed the observed Opx/Crd ratios. The Opx/Crd ratio is lowest in all the SE group coronas for each domain. Deviation between model and observed Opx/Crd ratios is greatest in core coronas compared to rim coronas for the same corona domain. The difference between model and observed Opx/Crd ratios is lowest in the SE group coronas except for fracture symplectite, where deviation between observed and predicted is marginally higher in the SE group coronas (SK9-CLM, VT206A). 499

79 Figure 6.27: Comparison of observed and predicted Opx/Crd and Pl/Crd ratios for corona products for the garnet-q quartz corona compositional domain. (a) Orthopyroxene modes relative to cordierite are anomalously high in core coronas from the NW group and to a lesser extent in rim coronas from the same group. In core coronas from the SE group (SK9-CLM) Opx/Crd ratios still marginally exceed those predicted by the model, however this trend is reversed in rim coronas from the SE group, where Opx/Crd predicted d modes exceed those observed (as for aluminous domains). (b) Model Pl/Crd ratios are up to an order of magnitude greater than those observed, implying anomalously high plagioclase modes are observed in natural coronas. 500

80 6.7.1 Modal mismatch in garnet-feldspar and garnet-biotite domains Open-system diffusion modelling (Chapter 5) indicated that FeO is the most mobile component in corona domains, followed by MgO, CaO (and alkalis), SiO 2 and, lastly, Al 2 O 3. It is possible to explain the anomalously high cordierite modes if each aluminous EBC is considered as an open system with respect to rapidly diffusing components, such as FeO, and closed with respect to more sluggish components, i.e., Al 2 O 3. With prolonged reaction and metasomatic exchange with the surrounding rock, an EBC may become depleted in mobile FeO, and MgO to a lesser extent, whilst becoming relatively enriched in Al 2 O 3 and SiO 2. The T-X pseudosection in Figure 6.28 demonstrates the effect of FeO and MgO loss from a fracture symplectite EBC (95 mol% garnet + 5 mol% matrix) though opensystem metasomatic exchange of these components with neighbouring domains. Loss of ~ 2 mol% FeO and ~ 1 mol% MgO preferentially from the closed system EBC results in the relative enrichment of the residual EBC in less mobile components, such that both SiO 2 and Al 2 O 3 content increases. The EBC evolves from its initial closed system composition (X = 0.0, Figure 6.28) with prolonged reaction to relatively more aluminous, slightly magnesian compositions (X = 1.0, Figure 6.28). With increasing open-system behaviour with respect to FeO and MgO, X tracks from 0.0 to 1.0 and the predicted cordierite mode increases as orthopyroxene mode reduces. Anomalous cordierite modes relative to orthopyroxene are, thus, readily accounted for by invoking metasomatic exchange of FeO preferentially, and MgO to a lesser extent, with a sink in the surrounding rock. A potential sink for these components may be the garnet-quartz domain and this possibility is discussed in Section

81 Figure 6.28: T-X pseudosection simulating the effect of open-system metasomatic loss of FeO and MgO from a garnet-dominated EBC characteristic of that attained in fractures of garnet with only 5 mol% matrix input. Loss of FeO and MgO, such that X Mg increases marginally toward the right axis (with relative enrichment in SiO 2 and Al 2 O 3 ), results in an increase in cordierite relative to orthopyroxene. More enhanced open-system metasomatic exchange is expected in the SE group rocks where reaction is more prolonged at higher post-shock temperatures and lower fluxed solidi. 502

82 6.7.2 Modal mismatch in garnet-quartz domains In general, the layer sequence for a garnet-quartz compositional domain is garnet cordierite-orthopyroxene symplectite plagioclase blocky orthopyroxene quartz. Locally in both the NW group and the SE group, the cordierite-orthopyroxene symplectite is absent and a monomineralic plagioclase layer is succeeded by blocky orthopyroxene adjacent to quartz (Figure 6.29; Sections and 4.3.1). The proportion of plagioclase to orthopyroxene approaches 1:1. This layer sequence typically grades laterally into an adjacent aluminous domain comprising a thick cordierite and orthopyroxene symplectite that penetrates into the garnet core (Figure 6.29). A similar association of contiguous symplectite and monomineralic plagioclase and blocky orthopyroxene is observed in SK9B 3 between adjacent garnet-quartz and garnet-plagioclase compositional domains (Figure 6.30). The garnet-plagioclase domain at the margin of garnet is characterised by a thick, welldeveloped cordierite-orthopyroxene symplectite grading into a garnet-quartz domain comprising monomineralic plagioclase and blocky orthopyroxene. Garnet-quartz domains within the core of garnet are similarly characterised by the absence of a cordierite-bearing symplectite. Microprobe mapping of Fe, Mg, Na and Ca is included in Figure A T-X pseudosection for the garnet-quartz domain from SK9B 3 was constructed for a range of compositions from 90 mol % garnet + 9 mol % quartz + 1 mol% matrix to 90 mol % quartz + 9 mol % garnet + 1 mol% matrix, to test if cordierite is destabilised for any range of effective bulk composition assuming the garnet-quartz system is closed to adjacent compositional domains (Figure 6.31). At no intermediate composition is it possible to destabilise cordierite and produce plagioclase in the modes observed. Similarly, a T-X pseudosection constructed for garnet-plagioclase and garnet-quartz compositional end-members does not possess a cordierite-absent phase field (Figures 6.21, 6.32). The only way in which cordierite might be destabilised preferentially relative to plagioclase in the corona would be to increase CaO, FeO and MgO and decrease SiO 2 in the garnet-quartz EBC (Figure 6.33) relative to other components, requiring non-linear open-system interaction between adjacent domains with respect to mobile components CaO, FeO, MgO and SiO 2 with effective closure for Al 2 O 3. This is consistent with higher modelled 503

83 diffusion coefficients for these components in open-system diffusion simulations (Chapter 5). Figure 6.29: (a) PPL photomicrograph showing sectoral development of a garnet-quartz corona in SK9A comprising monomineralic plagioclase and orthopyroxene adjacent to an aluminous Crd- Opx symplectite domain penetrating into garnet along fractures. (b) BSE image showing sectoral development of monomineralic plagioclase and blocky orthopyroxene (cordierite-absent or present in trace amounts) in a garnet-quartz domain from SK9B 3, associated with a cordieriteorthopyroxene symplectite which penetrates into the core of garnet. 504

84 Figure 6.30: Electron microprobe element mapping for garnet-quartz and garnet-plagioclase domains in SK9B 3. Garnet-quartz domains in the core of garnet comprise blocky orthopyroxene and plagioclase only, juxtaposed with an aluminous garnet-plagioclase domain in which the cordierite-orthopyroxene symplectite is well-developed. (a) Schematic cartoon demonstrating phase relationships in the mapped area. (b) Fe EMPA element distribution map (c) Mg EMPA element distribution map (d) Ca EMPA element distribution map (e) Na EMPA element distribution map. 505

85 Figure 6.31: T-X pseudosection for the compositional range between garnet-dominated, garnetquartz EBCs and quartz-dominated, garnet-quartz EBCs. At no intermediate composition is it possible to destabilise cordierite to produce the monomineralic plagioclase and blocky orthopyroxene observed in SK9B

86 Figure 6.32: T-X pseudosection for the compositional range between garnet-dominated, garnetquartz EBCs and quartz-plagioclase EBCs. At no intermediate composition is it possible to destabilise cordierite to produce the monomineralic plagioclase and blocky orthopyroxene observed in SK9B 3 (cf. Figure 6.21). 507

87 Figure 6.33: T-X pseudosection accommodating open-system metasomatic exchange of FeO, MgO, and CaO with adjacent domains. Cordierite mode (red isopleths) approaches zero just above the solidus with increasing influx of MgO, FeO and CaO into the system and loss of SiO 2 from the system, whilst plagioclase (blue isopleths) and orthopyroxene modes (green isopleths) increase such that the Opx/Pl ratio approaches 2:1. Melt volumes (white isopleths) increase marginally as the solidus is depressed toward highest external fluxes at X =

88 The sectoral partitioning of the garnet-quartz corona into a calcic, ferromagnesian siliceous domain coupled with an adjacent aluminous domain across which chemical potential gradients exist is analogous to mesoscopic chemical partitioning of the bulk rock into leucosome and mesosome modelled by White et al. (2004). According to White et al. (2004) the partitioning of a rock into leucosome and mesosome is a result of spatial focusing of melting around peritectic garnet (Figure 6.34a). As temperature increases on the prograde path, garnet becomes stable and nucleates only in places where kinetic constraints to nucleation are locally overcome (proto-leucosome). In portions of the rock where garnet is not able to nucleate (proto-mesosome), metastable phase equilibria persist, with different chemical potentials to those where garnet is present. Chemical potential gradients are established between the garnet-bearing domains and the garnet-absent domains. In an attempt to equalise chemical potentials, Fe diffuses from the garnet-absent parts of the rock toward a nearby garnet, stabilising a K-feldspar- and melt-bearing assemblage enclosing garnet in the proto-leucosome. Mg diffuses away from the garnet-bearing areas to restitic biotite-sillimanite-quartz assemblages in the protomesosome. Compositional partitioning and spatial segregation of leucosome and mesosome assemblages will continue as long as chemical potential gradients exist between them. The model of garnet-quartz corona development in this section also invokes kinetic constraints in compositional partitioning between two domains, except that, in this case, garnet nucleation is not rate-limiting but rather different length-scales of diffusion for immobile vs. mobile components (Figure 6.34b). Mobile Ca, and lesser Fe and Mg diffuse from the garnet-, feldspar- and/or biotite-dominated compositional domains towards the quartz sink (black arrows in Figure 6.34b), yielding elevated orthopyroxene and plagioclase modes with melt, analogous to the formation of garnet enclosed by K-feldspar and melt in the proto-leucosome of White et al. (2004). Residual adjacent aluminous domains are relatively enriched in Al 2 O 3 and MgO, stabilising higher cordierite modes (as in Figure 6.33). SiO 2 diffuses away from the garnet-quartz corona, destabilising spinel in immediately contiguous symplectite. The plagioclase-orthopyroxene corona between quartz and garnet is, thus, the equivalent of the proto-leucosome of White et al. (2004) and the 509

89 residual cordierite-spinel-orthopyroxene symplectite is the equivalent of the protomesosome. Figure 6.34: Metasomatic exchange between compositional domains during a) leucosome development with the spatial focussing of melting around peritectic garnet (after White et al.., 2004) and b) development of cordierite-absent, garnet-quartz coronas adjacent to cordieriteenriched aluminous domains. 510

90 6.7.3 Melt loss and corona textural and compositional evolution Melt loss from the corona domain is another potential mechanism for the complex non-linear, open-system evolution of the EBC with protracted reaction. In this section, the implications of melt loss on the observed corona modes and compositions are modelled and the potential extent of deviation between observed and modelled corona product modes is constrained. To demonstrate the effect of melt loss on corona assemblage evolution, two T-X pseudosections have been constructed for the garnet-quartz and garnet-plagioclase corona domains with different degrees of melt loss at two different temperatures. In Figure 6.35, the shift in modes and phase compositions are modelled for a garnetquartz corona in partial communication with matrix phases (28 mol % matrix) with 80% melt loss, 65 C above the solidus. The left axis of the T-X pseudosection corresponds to a garnet-quartz corona effective bulk composition with no melt loss. Removal of 80% of the melt in equilibrium with the corona assemblage for the latter bulk composition at 908 C from the equilibration volume yields the melt-depleted, restitic end-member bulk composition comprising the right axis of the T-X pseudosection. Assuming corona reaction is arrested at the solidus, the corona assemblage predicted immediately above the solidus for both end-member compositions (hereafter referred to as the solidus assemblage) should closely approximate the corona assemblage preserved in the rock in reality, apart from minor subsolidus compositional resetting. Corona product modes for the suprasolidus assemblage prior to melt loss and the corona solidus assemblages with no melt loss to 100% melt loss are compared in Table The modes of the melt-depleted solidus corona assemblage are very similar to that in the suprasolidus assemblage prior to melt loss. This preservation of the suprasolidus assemblage with melt loss is in agreement with White and Powell (2002). With increasing melt loss, the proportions of cordierite and orthopyroxene in the solidus assemblage both increase marginally (Figure 6.35, Table 6.10). Cordierite modal proportion increases from 0.30 to 0.31 and orthopyroxene increases from 0.36 to The cordierite:orthopyroxene ratio decreases from 0.83 to 0.79, reflecting a slightly more anhydrous, ferromagnesian bulk composition for the melt-depleted composition. Compositions of major phases 511

91 do not shift significantly with melt loss (Table 6.10). There is no change in X Mg number for cordierite or orthopyroxene, and plagioclase becomes very slightly more calcic (ca(pl) = 0.95 at X = 0.0; ca(pl) = 0.97 at X = 1.0). Figure 6.35: T-X pseudosection for a SK9-CLM garnet-quartz domain with a matrix component (28 mol%) showing the effect of 80% melt loss at 908 C on corona mineral modes and solidus elevation. 512

92 Table 6.10: Changes in the garnet-quartz corona assemblage with melt loss Modes* Suprasolidus corona assemblage 908 C - no melt loss Solidus corona assemblage at 843 C - no melt loss Solidus corona assemblage at 893 C - 80% melt loss Cordierite Orthopyroxene K-feldspar Plagioclase Ilmenite Quartz Crd/Opx ratio Compositions x(crd) h(crd) x(opx) y(opx) N(Opx) f(opx) na(ksp) ca(ksp) ca(pl) k(pl) x(ilm) Q(ilm) *renormalized to 1 with no liq present The effect of more extreme melt loss at higher temperatures is examined for the garnet-plagioclase domain in Figure This T-X pseudosection models the effect of 100% melt loss from the domain equilibration volume at 1050 C, i.e., 156 C above the modelled solidus. Corona product modes for the suprasolidus assemblage prior to melt loss and the corona solidus assemblages with no melt loss to 100% melt loss are compared in Table The solidus is elevated to 1049 C with melt loss and the peak modes and compositions of phases prior to melt loss are preserved in the melt-depleted solidus corona assemblage. In contrast, the modes and compositions of the corona phases in the solidus assemblage with no melt loss are 513

93 significantly different from those preserved in the solidus assemblage with 100% melt loss. Cordierite mode increases from 0.25 to 0.27 and orthopyroxene mode increases from 0.36 to 0.43 with melt loss. Plagioclase modes are lower with melt loss, and spinel modes are correspondingly higher. K-feldspar is present in the solidus assemblage with no melt loss but is absent in the melt-depleted corona composition. Cordierite and orthopyroxene compositions are more Mg-rich and plagioclase is more calcic with melt loss. Figure 6.36: T-X pseudosection for a SK9-CLM garnet-plagioclase domain (20 mol% matrix) showing the effect of 100% melt loss at 1050 C on corona mineral modes and solidus elevation. 514

94 Table 6.11: Changes in the garnet-plagioclase corona assemblage with melt loss Modes* Suprasolidus corona assemblage 1050 C - no melt loss Solidus corona assemblage at 894 C - no melt loss Solidus corona assemblage at 1049 C - 100% melt loss Cordierite Orthopyroxene K-feldspar Plagioclase Spinel Ilmenite Crd/Opx ratio Compositions x(crd) h(crd) x(opx) y(opx) N(Opx) f(opx) na(ksp) ca(ksp) ca(pl) k(pl) x(he) y(he) z(he) Phase equilibria modelling indicates that melt loss preserves the corona suprasolidus assemblage at the time of melt loss. The melt-depleted solidus assemblage is more restitic and anhydrous compared to the solidus assemblage with no melt loss and exhibits higher modelled orthopyroxene and spinel modes and lower plagioclase modes. The difference in modes and compositions between the two end-member solidus assemblages depends on the temperature and volume of melt loss. More melt loss at higher temperatures yields a greater deviation between the solidus assemblage modes. The influence of melt loss must be considered when evaluating the accuracy of inferred EBC ranges of near-equilibration in Chapter 6, in that potentially melt-depleted solidus assemblage corona modes with higher orthopyroxene-to-cordierite ratios may be used to infer a more restitic EBC range than may have prevailed in reality. The end result is that we are more likely to 515

95 underestimate matrix contribution to the corona domain owing to higher orthopyroxene modes in restitic melt-depleted solidus assemblages and, thus, be more conservative in assessing the extent of communication between different domains. In reality, it is not anomalously high orthopyroxene modes that make constraining the EBC range of equilibration difficult in the coronas, but rather anomalously high cordierite modes. The latter has been attributed to open-system loss of FeO, MgO and CaO preferentially to adjacent garnet-quartz domain sinks, thereby relatively enriching the depleted EBC in cordierite (Section 6.7.2). Opensystem modification of the domain in this manner skews the observed from modelled corona modes and renders the influence by melt loss negligible. Melt loss would however, significantly reduce the reaction time available and, thus, limit the capacity of a compositional domain to attain equilibration, since the solidus is elevated. Solidus elevation means that arrested reaction in a melt-depleted corona occurs at a higher temperature with reduced melt modes, restricting length-scales of diffusion and suppressing the equalisation of chemical potential gradients. This is the opposite of what is suggested by the textural, diffusion modelling and modal analysis between the NW group and SE group coronas (Chapters 4 and 5). Having quantified the potential influence of melt loss on corona assemblage evolution, it is important to constrain the probability or likelihood of melt loss from the granulites during post-shock metamorphism. In THERMOCALC, the molar volume of a pre-impact (Grt-Bt-Crd-Kfs-Pl-Qtz-Ilm) peak assemblage at 7.5 kbar and 870 C for SK9-CLM is calculated at cm 3. The molar volume for a postimpact assemblage (Crd-Opx-Kfs-Pl-Liq) in SK9-CLM at 3 kbar and 900 C is cm 3. This positive V should lead to overpressurization of the reaction domain and fracturing followed by rapid extraction and loss of the melt from site of generation (Clemens and Droop, 1998). However, Rushmer (2001) demonstrated that melt-induced fracturing occurs only during muscovite-melting with large dilational strains and positive volume changes an order of magnitude larger than that for biotite-dehydration melting. In contrast, negligible dilational strains with comparatively minor positive volume changes with biotite-dehydration melting will not induce melt-fracturing under static conditions without the application of rim applied stress. Melt loss under static post-impact conditions is, thus, unlikely. This is supported by the absence of any micro- or macro-scale evidence for melt extraction 516

96 (Sawyer, 1999). The K-feldspar moat between ternary plagioclase and quartz (e.g., Chapter 4, Figure 4.42) is a simple monomineralic corona between quartz and ternary feldspar and not crystallised melt along grain boundaries as suggested by Gibson (2002). The pseudotachylitic breccias are shock melts which have been siphoned into dilational structures immediately post-impact and are unrelated to generation of anatectic melts through corona reaction during longer-duration postimpact cooling and should not be confused with networks draining anatectic melts from coronas. An important consequence of melt retention in the Steynskraal granulites is that the impact event has not significantly altered the bulk rock compositon of the granulites through melt loss from that after the peak event in the Archaean (M 1 ; Chapter 2). This means that the XRF composition used to construct the P-T pseudosections for the Archaean event and the calculation of corona endmember bulk compositions in THERMOCALC from predicted peak phase compositions and modes, accurately reflects the peak M 1 bulk rock composition. 6.8 Extreme melt loss and the origin of the aluminous granofelses from the Inlandsee Pan Highly aluminous granofelses occur to the northeast of the Inlandsee Pan, 4 km from the inferred centre of the Dome (Stepto, 1979; Gibson, 2002). The mineralogy and mineral chemistry of these rocks have been described in Gibson (2002). The original Archaean fabric present in the migmatitic granulites from the Steynskraal traverse is absent in these rocks. VT484 comprises intricately intergrown euhedral to subhedral cordierite and K-feldspar oikokrysts enclosing spinel, corundum and rutile (Figure 6.37). This texture is reminiscent of igneous textures formed during crystallisation of peritectic phases from a melt. The absence of a fabric and unusual igneous textures have led workers to describe these rocks as granofelses (Stepto, 1979; Gibson, 2002). Spatially associated with the granofelses are a number of undeformed granite bodies yielding a 2.02 Ga age from unzoned and undeformed magmatic zircons (Gibson et al., 1997). 517

97 Figure 6.37: Photomicrographs showing textures in VT484. (a) A skeletal cordierite oikocryst encloses spinel, rutile, plagioclase and K-feldspar (plane polarised light). Adjacent K-feldspar laths comprise cores rich in corundum and spinel inclusions and rims with lesser rutile inclusions. (b) Approximately the same field of view as in (a) but with crossed polarised light. 518

98 Stepto (1979) described a sample of granite in contact with the granofels, in which the biotite in the granite increases in grain size toward the contact with the granofels. In a thin section of the granofels from the latter sample, Gibson (2002) noted granophyric patches within the aluminous granofelsic host. Cordierite oikokrysts abutting these granophyric patches display euhedral forms and are partially replaced by biotite. Stevens et al. (1997b) proposed that the granofelses are restitic residues of extreme anatexis during the Archaean mid-crustal granulite event (M 1 ), however, Gibson (2002) pointed out that the impact age for the associated granites, absence of fabric and low-pressure assemblage of the granofelses are more consistent with fractional crystallisation of a high-proportion partial melt from a garnet-granulite precursor under post-shock conditions that produced a more evolved granite melt. In an effort to elucidate the origin of the granofelses within the post-impact metamorphic context, a P-T pseudosection was constructed with THERMOCALC for sample VT484 (Figure 6.38) and contoured for mineral modal proportions for the P-T range from 1 to 5.5 kbar and from 875 to 1075 C (Figure 6.39). The bulk composition used to generate the pseudosection is a melt-depleted bulk rock composition (adjusted slightly from XRF analysis for VT484 to accommodate melt loss) and corresponds to that attained immediately before cessation of reaction just above the solidus below which no further melt loss from the rock is possible. The peak post-impact assemblage observed in the rock correlates with that predicted immediately above the solidus (hatched area, Figure 6.39). The presence of oikocrystic cordierite with inclusions of spinel, K-feldspar, rutile and plagioclase invokes an initially high-temperature cooling path through a cordierite-absent phase field comprising K-feldspar, spinel, plagioclase and rutile, followed by peritectic crystallisation of cordierite with cooling, such that the early crystallised phases were enclosed within the cordierite oikocrysts. The cordierite-in line has a very steep dp/dt slope, implying temperatures in excess of 1000 C must have been achieved in order for oikocrystic cordierite to form. 519

99 . Figure 6.38: P-T pseudosection for VT484. The peak phase field observed in the rock is delimited by bold phase field boundaries, i.e., cordierite, K-feldspar, plagioclase, spinel, corundum, rutile and melt. The composition differs slightly from the XRF composition- melt had to be removed from the XRF composition to yield an assemblage observed in the thin sections for VT

100 Figure 6.39: P-T pseudosection for VT484, a pelitic cordierite-k-feldspar-bearing granofels from the Inlandsee terrane. The peak phase field has been contoured for cordierite, K-feldspar and spinel mode. The peak post-impact assemblage observed in the rock correlates with that predicted immediately above the solidus (hatched area). 521

101 An investigation was conducted with the aid of THERMOCALC to simulate the process proposed by Gibson (2002) for granofels generation, i.e., the granofelses represent cumulates formed by varying degrees of fractional crystallisation of a high-proportion anatectic melt of a garnet granulite under post-impact conditions. A P-T pseudosection constructed for the metapelitic garnet granulite (SK9-CLM) from the Steynskraal traverse (Chapter 2) was used to predict the compositon of a highproportion partial melt at 3 kbar and 1100 C. This melt was extracted from the restite and the effect of varying degrees of fractional crystallisation on the composition of the restite and evolving granite melt was simulated. In a T-X pseudosection (Figure 6.40), the residual cumulate compositions and associated liquid compositions are modelled for progressive melt loss across a compositional spectrum from no melt loss to that producing a residuum composition corresponding to VT484. A mineral assemblage corresponding to that of a second sample described by Gibson (2002) - VT624 - is attained at intermediate magnitudes of melt loss (X ~ 0.5) from the residuum. In bulk compositions approaching that of VT484, a rutilebearing assemblage, as opposed to an ilmenite-bearing assemblage, implies greater extent of melt loss from the residuum, such that the solidus is elevated into the rutile-bearing field. The model compositions generated for the T-X pseudosection (Figure 6.41) are compared to those observed in reality (Figure 6.42). The cumulate compositions (VT624 to VT484,1) reflect the trend predicted by the model, i.e., the granofelses may reasonably be explained as the cumulates produced by progressive crystal fractionation of a liquid generated by extreme melting of a garnet granulite precursor. The deviation of cumulate compositions from the model residuum compositions may reflect compositional differences of the garnet granulite precursor from that modelled. The marginal enrichment of Na+K in VT624 and VT486 may be caused by only partial loss of associated granitic liquid from the residuum, thereby shifting the bulk rock for VT624 to more sodic-potassic compositions. The Inlandsee granites (IL5, VT359.1, and VT356.2) are more enriched in Al 2 O 3 and Na+Ca compared to the model compositions this may reflect assimilation of surrounding partially molten TTGs (Gibson and Reimold, 2008) and/or entrainment of residuum. 522

102 Figure 6.40: T-X pseudosection simulating residual cumulate and associated liquid compositions with progressive melt loss across a compositional spectrum ranging from a melt extract from garnet granulite SK9-CLM and no melt loss to that producing a residuum composition corresponding to VT484. The mineral assemblage correlating with VT624 is produced at intermediate melt loss. Ilmenite-absent assemblages in cumulate compositions approaching VT484 require the highest degrees of melt loss in order for the solidus to be elevated into the rutilebearing field. 523