Open Recycling Argon through Metamorphic Reactions: the Record in Symplectites

The 40 Ar/ 39 Ar ages of metamorphic micas that crystallized at high temperatures are commonly interpreted as cooling ages, with grains considered to have lost 40 Ar via thermally-driven diffusion into the grain boundary network. Recently reported laser-ablation data suggest that the spatial distribution of Ar in metamorphic micas does notalwaysconformtothepatternspredictedbydiffusiontheoryandthatdespitehighmetamorphictemperatures, argon was not removed ef ﬁ ciently from the local system during metamorphic evolution. In the Western Gneiss Region (WGR), Norway, felsic gneisses preserve microtextural evidence for the breakdown of phengite to biotite and plagioclase symplectites during near isothermal decompression from c. 20 – 25 to c. 8 – 12 kbar at ~700 °C. These samples provide an ideal natural laboratory to assess whether the complete replacement of one K-bearing mineral by another at high temperatures completely ‘ resets ’ the Ar clock, or whether there is some inheritance of 40 Ar in the neocrystallized phase. The timing of the high-temperature portion of the WGR metamorphic cycle has been well constrained in previous studies. However, the timing of cooling following the overprint is still much debated. In-situ laser ablation spot dating in phengite, biotite-plagioclase symplectites and coarser, texturally later biotite yielded 40 Ar/ 39 Ar ages that span much of the metamorphic cycle. Together these data show that despite residence at temperatures of ~700 °C, Ar is not completely removed by diffusive loss or during metamorphic recrystallization. Instead, Ar released during phengite breakdown appears to be partially reincorporated into the newly crystallizing biotite and plagioclase (or is trapped in ﬂ uid inclusions in those phases) within a close system. Our data show that the microtextural and petrographic evolution of the sample being dated provides a critical framework in which local 40 Ar recycling can be tracked, thus potentially allowing 40 Ar/ 39 Ar dates to be linked more accurately to metamorphic history. © 2017 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).


40
Ar/ 39 Ar dating is commonly used to constrain the timing and the rate of cooling in high temperature metamorphic terranes via the closure temperature (T C ) formulation suggested by Dodson (1973). This formulation is underpinned by a number of assumptions: (1) that the mineral in question (re)crystallized with an insignificant initial 40 Ar concentration; (2) that thermally-activated volume diffusion, observing Fick's 2nd law, redistributed Ar within the mineral after its production, (3) that the grain boundary network surrounding the mineral had a sufficiently low Ar concentration and high connectivity to act as an infinite reservoir (i.e. an 'open' grain boundary system) during the time at which temperatures were high enough for efficient diffusion, and (4) that the grain crystallized at temperatures high enough for diffusion to initially be efficient. Furthermore, the formulation requires a cooling path of 1/T shape.
Many studies document equivocal metamorphic 40 Ar/ 39 Ar muscovite, phengite (high pressure (HP) muscovite) and biotite ages that are 'too old' to be interpreted as representing the timing of cooling when assessed within a chronological framework defined by higher closure-temperature chronometers such as U\ \Pb zircon (e.g. Baxter et al., 2002;Li et al., 1994;Warren et al., 2012b). These results are most commonly documented in HP metamorphic terranes, where the temperature was relatively low (b 600°C), the timescales of metamorphism commonly short (b 10 Ma), and fluid availability and permeability is low, thus limiting the efficiency of diffusion and efficiency of Ar removal from the grain boundary network. There has been much debate about whether the "too old" ages are the result of contamination with 'excess' or 'inherited' 40 Ar (decoupled from parent K or inherited from an earlier generation of mineral growth), whether they represent the timing of mineral crystallization, or a combination of the above (e.g. Allaz et al., 2011;Beltrando et al., 2013;de Sigoyer et al., 2000;Di Vincenzo, 2006;Hacker et al., 2003;Li et al., 1999;Putlitz et al., 2005;Sherlock and Kelley, 2002;Villa, 1997;Villa et al., 2014;Warren et al., 2011Warren et al., , 2012b. These older ages may also be the result of inefficient diffusive loss from the grain, due to slower diffusion than predicted by laboratory studies, or due to a grain boundary that does not allow argon to escape the grain efficiently. If diffusive loss is more inefficient than current theory suggests then it is much more difficult to link age to temperature, and the Dodson formulation will not provide the correct interpretation. HP metamorphic mineral assemblages commonly recrystallize to lower pressure assemblages during exhumation. Recrystallization during exhumation has been hypothesized to play a far greater role than diffusion in releasing/remobilizing Ar during metamorphism and deformation, and hence re-setting the 40 Ar/ 39 Ar 'clock' (Allaz et al., 2011;Kellett et al., 2016;Mulch and Cosca, 2004;Villa, 1997;Villa et al., 2014). The determination of when the target mineral grew during the metamorphic cycle provides a key relative chronological framework for Ar age interpretation, and provides an opportunity to test this hypothesis. This study aims to assess whether the complete replacement of one K-bearing mineral by another at high temperatures completely 'resets' the Ar clock, or whether there is some inheritance of 40 Ar in the neocrystallized phase.
The Western Gneiss Region (WGR) of Norway provides the ideal natural laboratory for testing the relative importance of diffusion and recrystallization in resetting the Ar clock during a high temperature metamorphic cycle. The WGR is a~50,000 km 2 window into the lowest structural levels of the Scandinavian Caledonides, formed during the subduction of Baltica beneath Laurentia during the Caledonian Orogeny (Gee, 1975;Hacker, 2007;Roberts, 2003). Mafic rocks exposed within predominantly amphibolite-facies felsic gneiss preserve localized evidence for recrystallization under eclogite-facies conditions (Bryhni, 1966;Eskola, 1921;Smith, 1984). Coesite (the high pressure polymorph of quartz) has been documented in both mafic and felsic lithologies (e.g. Terry et al., 2000;Wain, 1997;Wain et al., 2000).
The UHP lithologies occur in three separate, antiformal domains which, from south to north, are the Nordfjord-Stadlandet, the Sorøyane, and the Nordøyane domains, together covering an area of 5000 km 2 (Root et al., 2005). This study focuses on the Nordfjord-Stadlandet domain, where the recorded/preserved P-T gradient ranges from the quartz eclogite-facies in the southeast (~23 kbar and~600°C) to coesite-stable eclogite-facies in the northwest (~30 kbar and~700°C) across 2500 km 2 (Root et al., 2005;Young et al., 2007). Peak conditions were reached between c. 410-400 Ma (Hacker et al., 2010;Kylander-Clark and Hacker, 2014;Root et al., 2004;Young and Kylander-Clark, 2015). Following (near) isothermal decompression from (U)HP conditions, the entire WGR experienced retrograde amphibolite-facies recrystallization at~700°C and 10 kbar by c. 399-379 Ma (Spencer et al., 2013;Walsh and Hacker, 2004). According to the results of simple open-system diffusion models, the diffusion of Ar in both white mica and biotite should be efficient at these conditions (Warren et al., 2012a(Warren et al., , 2012b. Published white mica and biotite 40 Ar/ 39 Ar multi-and single-grain step heating ages from the Outer Nordfjord area range from 389 to 374 Ma and 402-375 Ma, respectively, and interpreted as cooling ages (Berry et al., 1995;Hacker and Gans, 2005;Lux, 1985;Root et al., 2005;Walsh et al., 2007Walsh et al., , 2013Young et al., 2011). More recent single grain fusion and in-situ laser ablation techniques have yielded highly variable white mica 40 Ar/ 39 Ar ages, both within and between grains in the same sample, as well as between different samples (Warren et al., 2012b). The lack of core-rim profiles indicative of volume diffusion, as well as the inter-grain variability, suggests that a simple model of open-system diffusion does not fit the data and that more than one process may have been responsible for Ar redistribution during the metamorphic cycle.
Phengite in multiple felsic gneiss samples, analyzed using singlegrain fusion and in-situ techniques, yielded 40 Ar/ 39 Ar ages that were significantly older than the timing of peak metamorphism, and, with a range between 507 and 379 Ma, spanned the timing of the whole WGR metamorphic cycle (McDonald et al., 2016;Warren et al., 2012b). 40 Ar/ 39 Ar ages were heterogeneously distributed both within individual grains and between different grains within the same sample as well as across different samples. These results imply that micas in different lithologies within the WGR lose or gain 40 Ar differently during the metamorphic cycle and that the approximations that underpin the Dodson (1973) T C formulation may not be applicable in this terrane.
The common occurrence of symplectites in felsic gneisses from terranes that have decompressed from pressures N 20 kbar at temperatures N650°C ('diffusion-efficient' conditions) provides an opportunity to test the effects of high-temperature recrystallization on element remobilization (e.g. Heinrich, 1982;Wain et al., 2000;Gilotti and Krogh-Ravna, 2002;Anderson and Moecher, 2007). Of specific interest to metamorphic 40 Ar/ 39 Ar geochronology is the breakdown of phengite to symplectites of biotite plus plagioclase, which provides an excellent opportunity to assess the transport distances, pathways, sources and sinks of 40 Ar in K-rich assemblages as high temperature minerals break down and recrystallize during exhumation.
Garnet-bearing felsic gneisses preserving relict HP metamorphic assemblages are locally preserved in strain shadows around mafic boudins. Different samples preserve different stages of the mineralogical transition from phengite to biotite via symplectization . The preservation of these symplectites testifies to a lack of deformation during decompression (Hacker et al., 2010), thus allowing the effects of recrystallization at high temperature on Ar removal or redistribution to be investigated. This study focuses on a sample collected from Krokkenakken ( Fig. 1), on the northern shore of the Nordfjord, where garnet-bearing gneisses are preserved within the strain shadow of a N 200 m diamter mafic eclogite body.

Electron microprobe analysis
Major phases were characterized by back-scattered electron images and electron-microprobe analyses of polished thin sections on the Cameca SX100 (5 spectrometer) Electron Microprobe (EMP) at the Open University. Operating conditions were set to 15 kV and 20 nA, with a defocused beam of 10 μm. A ZAF matrix correction routine was applied and a selection of natural standards was used for calibration. Unknowns were bracketed by analyses of internal secondary standards; uncertainty on major element concentrations is b 1%. X-ray element maps of Al, Ca, Mg, Si, and Ti in phengite, biotite, and symplectites; and Al, Ca, Fe, Mg, and Si in garnet were obtained from polished thin sections scanned in mosaic mode (4 × 7 grid) with operating conditions of 20 kV and 20 nA. Dwell time was set at 5 ms with a step size of 20 μm. Chemical formulae were calculated stoichiometrically based on 12 oxygens for garnet, 8 for plagioclase, 22 for phengite and biotite, and 23 for amphibole. Representative core-rim analyses of phengite, biotite and plagioclase are shown in Table 1

Bulk chemistry determination
Major elements were determined using sample-glass discs prepared by fusing a 5:1 ratio mix of lithium metaborate/tetraborate flux (Spectroflux 105) and powdered samples prepared by jaw crushing and tema mill. Aliquots of~3 g of sample powder were dried overnight in porcelain crucibles in an over at~100°C. 3.59 g of flux and 0.7 g of sample powder were weighted into platinum crucibles and stirred with a glass rod to ensure homogeneity, before being fired in a furnace at 1100°C for~20 min. Sample mixtures were stirred every 5 min to remove gas bubbles and promote homogeneity. The melt was poured into a pre-heated brass mold and formed using a sprung press. Discs were left to cool prior to analysis. 1.2 g of sample powder from the dried aliquots was weighed to four decimal places into pre-ignited alumina crucibles. Samples were ignited in the furnace at 1000°C for 1 h, cooled for 15 min and reweighed to four decimal places. The difference in weight before and after ignition was used to assess the mass loss on ignition for each sample.
Trace elements were determined from pressed sample powder pellets formed by thoroughly mixing approximately 9.5 g of sample powder with 0.9 ml of Polyvinylpyrollidone(PVP)-methyl cellulose binder using the 'bag technique' (Watson, 1996). The mixture was emptied into a hardened steel mold and formed into a pellet using a hydraulic  press. The sample pellets were dried overnight in the oven at~100°C prior to analysis. Results of XRF analyses are presented in Supplementary Material S.2.

Thermodynamic modelling
Symplectites form from the breakdown of unstable HP phases (e.g. phengite and pyroxene) into two or more minerals at lower P-T conditions. Symplectites provide evidence that textural and compositional equilibrium was attained only on a domainal scale. This lack of textural equilibrium represents a challenge when modelling the metamorphic evolution of the sample. In the present case, a simplified approach was used to constrain the P-T evolution of the major phases: (i) the whole rock composition, determined by XRF, was used to model the growth of the HP assemblage as well as the symplectites. Microstructural evidence suggests that amphibole, biotite, and plagioclase formed from the breakdown of phengite and pyroxene (most probably omphacite, e.g. Wain et al., 2000). Garnets were replaced by fine grained aggregates of biotite and plagioclase. Following previous work (e.g. Godard and Mabit, 1998;Groppo et al., 2015) we have assumed the whole rock volume was chemically involved and behaving as a closed-system during symplectite formation; (ii) the retrograde formation of amphibole-plagioclase symplectites was constrained by the plagioclase-hornblende thermobarometer (temperatures using the calibration of Blundy and Holland, 1990;Holland and Blundy, 1994 and pressures using the calibrations of Anderson and Smith, 1995;Schmidt, 1992); (iii) the reconstructed metamorphic evolution was compared with published P-T paths for the WGR.

Diffusion modelling
In order to determine the expected 40 Ar/ 39 Ar 'cooling' ages of phengite, biotite and plagioclase for a purely diffusive, open system given the best-fit post-crystallization P-T-t history, diffusion ages were modelled using DiffArg_P, which incorporates the pressure-dependance of Ar diffuion in muscovite (modified from DiffArg after Wheeler, 1996;Warren et al., 2012a). All models were calculated for cylindrical diffusion geometry, with a Crank-Nicholson time integration solver, and a time step of 10 (Wheeler, 1996).
On the basis of petrographic observation constraints, 1 mm radius phengite was modelled as crystallizing at peak pressure conditions of 23.5 kbar and 710°C. It then decompressed to 12 kbar (at 680°C) over 12 Ma (calculated from the difference between the youngest published U\ \Pb zircon, and the weighted average titanite age (Root et al., 2004;Spencer et al., 2013) followed by linear cooling at rates of 10, 25, 50, 70, and 100°CMa −1 . Current estimates for the cooling rates of the WGR range from 20 to 90°CMa − 1 based on the assumption that 40 Ar/ 39 Ar ages represent cooling through the muscovite closure temperature of~450°C (Hacker et al., 2010;Kylander-Clark et al., 2008;Root et al., 2004Root et al., , 2005Walsh et al., 2013). Decompression set so that P reached zero at the same time as T.
For modelling purposes, biotite and plagioclase were assumed to have crystallized at conditions of 12 kbar and 680°C. Cooling initiated immediately in the models and multiple models were run with the same cooling rates as the white mica models. Idioblastic biotite was modelled using a grain radius of 0.5 mm while the symplectic biotite and plagioclase were modelled with a grain radius of 0.25 mm.
To determine the possible affects of variations in the model starting temperatures, pressures, grain sizes, and cooling rates, a sensitivity analysis was performed. Muscovite peak temperatures and pressures were modelled at ± 50°C and ± 1.5 kbar, to reflect uncertainties in the modelled peak metamorphic conditions. Amphibolite-facies conditions for muscovite, biotite, and plagioclase were also modelled at ± 50°C and ± 5 kbar. Grain size was modelled at 0.25-1.5 mm for each mineral at the above mentioned cooling rates.
Model results and sensitivity analysis (Table 3a, b, and Supplementary Tables S.4) are reported as bulk (area-integrated) age differences between the timing of cooling initiation based upon the U-Pb zircon and titanite ages, and the modelled grain age.

40 Ar/ 39 Ar analysis
Polished thick sections (~250 μm thick; Fig. 2D) were prepared by adhesion to glass slides with cyanoacrylate ("super glue"). Sections were removed from the glass by soaking in acetone for up to 24 h and then cut into 5 × 5 mm squares, washed in acetone or methanol and distilled water, and packed into aluminum foil packets, which were irradiated at McMaster University Reactor, Ontario.
High spatial resolution single spot and traverse analyses on polished thick sections were achieved using a New Wave Systems Nd-YAG 213 nm ultraviolet (UV) laser coupled to a Nu Instruments Noblesse gas mass spectrometer run in single-collector mode. Prior to analysis, samples were loaded into an ultra-high vacuum laser port, warmed under a heat lamp for 8 h to reduce atmospheric background levels and left to pump down for at least 48 h. Analyses consisted of ablating 50/30/15 μm diameter spots for 90 s followed by 90 s gettering time. Active gas species were gettered using two SAES AP10 Zr-Al getters (one at room temperature and the other at 450°C). Peaks of 40 Ar, 39 Ar, 38 Ar, 37 Ar and 36 Ar were scanned ten times each and the data extrapolated back to the inlet time (time zero).
Irradiation flux was monitored using the biotite standard GA-1550 with an age of 99.74 ± 0.10 Ma (Renne et al., 2011). Sample J values and uncertainties are given in Supplementary Table S.3 and were calculated by linear interpolation between two bracketing standards. The following corrections were applied to the standards: ( 39 Ar/ 37 Ar)Ca = 0.00065 ± 0.0000033, ( 36 Ar/ 37 Ar)Ca = 0.0002654 ± 0.0000013, ( 40 Ar/ 39 Ar)K = 0.0085 ± 0.0000425 based on analyses of Ca and K salts; only the K correction was applied to the analyzed samples due to negligible Ca in the sample micas. Background (blank) measurements bracketed every 1-2 sample measurements. Analyses were corrected for mass spectrometer discrimination using an assumed value of 295, determined from regular analysis of modern glass.
Data were reduced using the in-house software package ArMaDiLo using a decay constant of 5.530 × 10 − 10 ± 0.013 a − 1 (Renne et al., 2011). Age uncertainties are reported to 2σ; uncertainties on the isotopic measurements are reported to 1σ. The maximum and minimum phengite, biotite, and plagioclase 40 Ar/ 39 Ar ages are reported in Table 2. The full dataset is provided in Supplementary Table S.3.
Biotite crystallizes in two different microstructural, and textural domains: (i) in symplectites pseudomorphing phengite, with the grain size coarsening outwards from core to rim (40-200 μm laths, Fig. 3C), and (ii) randomly oriented aggregates of grains within the matrix (300-1200 μm, Fig. 3B). The biotite-plagioclase symplectites make up ap-proximately~50 vol% (Fig. 2B). Biotite in the symplectites is Mg-rich with X M g (Mg/(Mg + Fe) values of 0.65-0.70 and Ti contents ranging from 0.12-0.31 apfu (Table 1). Geochemical analysis and maps (Fig. 3C) show that each grain is homogeneous in geochemistry; and that the variability seen is between individual grains within the symplectite (detailed geochemical maps are supplied Supplementary Material S.5).
Idioblastic biotites form approximately 10 modal% of the sample (Fig. 2C) and are found in the matrix, occasionally adjacent to garnet and quartz, but never adjacent to phengite. They are similar in composition to the symplectitic biotites, with X Mg values of 0.67-0.73 and Ti contents ranging from 0.21-0.32 apfu (Table 1). Geochemical maps show that these grains are relatively homogeneous cores and narrow (b5 μm) rims (Fig. 3D). The idioblastic grains contain inclusions of plagioclase (Fig. 2C), suggesting that they formed by coarsening of the symplectite biotites.
Plagioclase is found within the symplectic domains associated with biotite or amphibole and forms anhedral grains with a typical grain   (Fig. 2B). X-ray compositional maps (Fig. 3B) show that the grains are compositionally zoned, with low-Ca cores (An 11 Ab 89 Or 0 ) and higher-Ca rims (An 45 Ab 55 Or 0 ). Garnet crystals form 10-15% of the sample, are sub-to anhedral in shape and range from 0.5-15 mm in diameter. Garnets contain abundant, randomly-oriented inclusions of quartz, rutile, white mica, and apatite. Minor replacement of garnet by biotite and plagioclase ( Fig. 2A-C) is visible at grain rims. Grains are zoned in X Fe (0.66 in the core to 0.50 in the rim) and X Mg (0.14 in the core to 0.31 at the rim), whilst X Ca is more homogeneous, varying from 0.18 in the core to 0.19 at the rim. Some grains show a high-Ca annulus between the core and rim at X Ca~0 .23. X Mn is low throughout garnet at 0.01-0.02 (Supplementary Material S.5).
Fine-grained amphibole is associated with plagioclase in symplectites that form~10 modal % of the sample. Amphiboles are magnesiohornblende in composition with X Mg = 0.68-0.74 (Table 1).
Rare small kyanite crystals (50-100 μm) are present in the matrix (mostly at the contact with garnet rims) whilst rutile, zircon, and apatite are present in both the matrix and included in garnet (Supplementary Material S.5).
During decompression from the HP stage, at an approximate T of 700°C, the pseudosection suggests that feldspar becomes stable at b16 kbar and biotite at b12 kbar. This marks the beginning of symplectite formation; phengite is no longer stable at conditions below 7-11 kbar and b660°C. The X Mg and X Na compositions of biotite and plagioclase help to constrain the conditions of phengite replacement by biotite-plagioclase symplectites. The modelled isopleths in biotite and plagioclase suggest that the symplectites formed at T b 720°C and P b 12 kbar following (near) isothermal decompression from peak pressures (Fig. 4b).
The compositions of amphibole and plagioclase in equilibrium in symplectic domains constrain the PT conditions of replacement of earlier pyroxene. The amphibole-plagioclase thermobarometer (Anderson and Smith, 1995;Blundy and Holland, 1990;Holland and Blundy, 1994;Schmidt, 1992) suggests crystallization at 660 ± 70°C and 7.9 ± 1.1 kbar.

40 Ar/ 39 Ar laser spot dating
Forty-one spots on 5 phengite grains yielded ages varying from 517.7 ± 5.6 Ma to 388.8 ± 4.6 Ma (Fig. 5A). Age distributions are patchy, with clusters of 'older' ages in regions that are neither associated spatially with the geographical nor the geochemical cores of the phengite (Fig. 5A). Ten spots on 3 different symplectite biotite grains yielded ages from 488 ± 36 Ma to 329.8 ± 9.2 Ma (Fig. 5 B1, B2). Twelve spots on 12 different symplectite plagioclase grains yielded highly variable 40 Ar/ 39 Ar ages from 687 ± 264 to 93 ± 1050 Ma (Fig. 5 B1, B2). Three analyses had 39 Ar measured/ 39 Ar background measurements b 2 (analyses GM7,GM11, and GM12). The calculated ages have large errors, are therefore not considered robust and are ignored for the purposes of further discussion. Ignoring these three analyses, the plagioclase yielded ages ranging from 434 ± 84 Ma to 378 ± 130 Ma. Twenty-six spots on 6 grains of idioblastic biotite yielded a range of ages from 448 ± 5.8 Ma to 376.6 ± 6.4 Ma. There is no systematic distribution of age within the grains (Fig. 5C).

Metamorphic evolution
Three main metamorphic stages have been recognized in the Krokkenakken gneisses based on microtextural relationships, crystal chemistry, conventional thermobarometry and pseudosection modelling. The first stage occurs under prograde conditions with the growth of garnet. The major element zoning in the garnets core and rim are interpreted as reflecting this prograde growth. Peterman et al. (2009) have also documented prograde-zoned garnet at this locality. Eclogite-facies conditions followed the prograde growth of garnet and are characterized by the assemblage phengite + garnet + omphacite + quartz + rutile ± amphibole ± kyanite. Because of the poor preservation of eclogitic mineral relics such as omphacite, this stage can only be loosely constrained at 20-25 kbar and 670-770°C (Fig. 4b).The white mica solution model  correctly predicts the stability field of Krich white micas but does not incorporate the right amount of Na. The main consequence of this is the modelled appearance of paragonite (Pg in Fig. 4a).
Although amphibole, interpreted as part of the HP assemblage, has not been observed in the sample, the model shows that the amount of it stable in these fields corresponds to b2 vol%. Therefore, we cannot exclude that amphibole was part of the high-pressure assemblage and it was subsequently replaced by later phases. Kyanite probably once included in garnet rims later replaced by biotite + plagioclase aggregates ( Supplementary Fig. S.5), has been interpreted as part of the same assemblage. The calculated HP conditions overlap with the data presented for the same area (e.g. Cuthbert et al., 2000;Wain, 1997).
Peak pressure conditions were followed by (near)-isothermal decompression to amphibolite-facies conditions of T~700°C and P b 12 kbar as documented by the growth of biotite-plagioclase and amphiboleplagioclase symplectites developed after phengite and pyroxene respectively. The association of amphibole and plagioclase in fine-grained symplectites has been documented in other HP gneisses from the WGR (e.g. Wain et al., 2000) and has been interpreted as forming from the breakdown of an HP pyroxene (e.g. omphacite). The pseudosection constraints presented here provide similar PT estimates to those previously presented for the WGR gneisses (e.g. Cuthbert et al., 2000;Engvik et al., 2007;Hacker, 2007;Johnson et al., 2007;Labrousse et al., 2004;Root et al., 2005;Terry et al., 2000;Terry and Robinson, 2003;Wain, 1997;Wain et al., 2000;Walsh and Hacker, 2004). These calculated conditions indicate that the amphibole-plagioclase symplectites formed after the biotite-plagioclase symplectites; however, the pseudosection suggests that omphacite is unstable below 15 kbar. Therefore, we suggest that the amphibole-plagioclase symplectites most probably re-equilibrated at lower P-T conditions after formation.
The timing of titanite crystallization plays an important role in the interpretation of the 40 Ar/ 39 Ar age patterns, as the titanite U\ \Pb age has been used to constrain the amphibolite-facies recrystallization across the WGR (Krogh et al., 2011;Schärer and Labrousse, 2003;Spencer et al., 2013;Tucker et al., 1987). The conditions of titanite crystallization are controlled primarily by the bulk rock Ca/Al ratio (Frost et al., 2001). In rocks with high Ca/Al ratios, titanite can be stable up to granulitefacies conditions (Frost et al., 2001). However, in orthogneisses, such as those from the WGR, titanite is only stable at much lower P-T conditions (Frost et al., 2001). A large dataset of titanite U\ \Pb ages from across the WGR accompanied by pseudosections of different titanite-bearing rocks showed both that the ages are widely variable and that titanite is stable across a broad range of PT conditions depending on the bulk composition (Spencer et al., 2013). The titanite stability field in the pseudosection presented here (Fig. 4b) overlaps with the titanite-in reactions published by Spencer et al. (2013) for comparable bulk compositions at T b 500°C and P b 12 kbar. Therefore, we suggest that titanite crystallization occurred syn-to post-symplectite formation.
The model results are variably sensitive to a number of uncertainties, many of which have previously been discussed in detail (Warren et al., 2012b). Most of the model input parameters affect the whole popuation, and would not generate a spread of ages, rather they would just shift the model age to older or younger values. For example uncertainties in the activation energy (E a ) and the diffusion coefficient (D) lead to an uncertainty of ±4 Ma on the resulting model age.
In the phengite models, varying the starting temperature and pressure by ± 50°C and ± 1.5 kbar, respesctively, showed no change in resutling modelled ages. Varying the amphibolite-facies temperature and pressure conditions by ±50°C and ± 5 kbar, respesctively, leads  Whitney and Evans (2011). b) P-T evolution for the symplectic gneiss showing the modelled P-T path experienced by the sample (grey arrow) with the approximate positions of the HP assemblage, the area Bt-Pl symplectite stability, the Hbl \ \Pl thermobarometry results, the titanite stability field; and biotite-in and plagioclase feldspar-in reactions. The location of the core and rim compositions of garnet is taken from the intersection of the X Fe -, X Mg -, and X Ca -isopleths. P-T path for Nordfjord from Kylander-Clark and Hacker, (2014) shown for comparison.
For biotite and plagioclase models, varying the model starting temperature by ±50°C results in uncertainities on the model ages of 5.0-0.5 Ma for cooling rates from 10°CMa − 1 to 100°CMa − 1 for both minerals (Table 3b).
For all other parameters being equal, the variation between the modelled age of grains between 1.5 mm to 0.25 mm yields a ± 3.0-0.4 Ma uncertainty on the final age of the phengite models, ±2.5-0.3 Ma uncertainy on the biotite models, and a ±2.9-0.4 Ma uncertainty on the plagioclase models for cooling rates between 10°CMa − 1 and 100°CMa −1 (Table 3b).
Dependent upon the bulk rock composition, titanite either crystallizes after the symplectite growth (as documented here), or may be contemporaneous with symplectite growth (Spencer et al., 2013). The reported U\ \Pb titanite ages for the Nordfjord area range from 399 to 379 Ma, with a weighted mean of 393 ± 4 Ma (Spencer et al., 2013). Since the diffusion models are calculated relative to the U\ \Pb titanite age, the 20 Ma spread in the titanite U\ \Pb data results in a 20 Ma spread in expected 40 Ar/ 39 Ar ages. Cooling may have begun prior to 393 Ma but the ±4 (2σ) uncertainties on the U\ \Pb and average ± 8 Ma (2σ) uncertainties on the 40 Ar/ 39 Ar ages means that any variation on the timing of amphibolite-facies crystallization is lost within the uncertainties of the datasets. Given the uncertainties in the U\ \Pb ages, diffusion model results, and the pseudosection calculations, we have taken the weighted mean of the U\ \Pb titanite ages (393 Ma) to represent the timing of cooling initiation for the purposes of this discussion.
Measured age ranges for phengite, symplectite biotite, symplectite plagioclase and idioblastic biotite are c.518-389 Ma, c.488-330 Ma, c.434-378 Ma, and c.449-377 Ma, respectively. These age ranges are much greater than the age range predicted by simple diffusion modelling despite uncertainties in the cooling rate and effective diffusive grain size (Fig. 6). The phengite 40 Ar/ 39 Ar ages are mostly older than those predicted by diffusion models, regardless of whether the weighted mean or the youngest age is compared with the model results. The phengite ages are also generally older than the reference titanite age, suggesting that phengite displays little diffusive resetting despite the high temperatures (700-750°C). Overall the phengite 40 Ar/ 39 Ar ages are not therefore interpretable as cooling ages, a conclusion also reached by previous studies (McDonald et al., 2016;Warren et al., 2012a).
Biotite ages are consistently younger, with a smaller spread than phengite, though the ages are still overall older than expected from the diffusion models (Fig. 6). There is little consistent difference in age between the smaller, earlier-crystallized biotite in the symplectite and the larger, later, idioblastic grains, although the idioblastic grains yield a wider range but with overlapping errors. The similar ages yielded by biotites across a range of grain sizes either suggests that diffusive loss of Ar was inefficient, even at temperatures N 680°C (and therefore Table 3 A) Diffusion modelling results. Shown are the predicted model 40 Ar/ 39 Ar ages for phengite, biotite, and plagioclase for various cooling rates given various starting ages for the timing of peak metamorphism (U-Pb zircon) and amphibolite-facies cooling (U-Pb titanite). The modelled time difference between the start of the model and the resulting age (Δt) for each of the minerals is shown at varying cooling rates. B) Results of sensitivity testing showing the standard deviation of the model age output by varying the peak temperature and pressure (Peak T, Peak P) of the phengite models, and the temperature and pressure of the the amphbolite-facies conditions (Amp T, Amp P) for phengite, biotite, and plagioclase.  that most of the ages cannot be considered true "Dodson" cooling ages), or that the rate of cooling experienced by this sample was fast enough to yield little difference between the biotite age of differing grain sizes. The plagioclase ages also span a considerable range (434-378 Ma). The plagioclase has low K concentrations (average 0.1 wt%), and consequently some of the variability may be due to low measured 39 Ar concentrations and the presence of fluid inclusions. It is therefore plausible that the elevated plagioclase 40 Ar/ 39 Ar ages represent either crystal-lattice-hosted or fluid-inclusion-hosted contamination by excess 40 Ar.
Despite the uncertainty in the timing of initiation of cooling of the WGR, there is clearly a disparity between both the observed 40 Ar/ 39 Ar ages (and age ranges) and the modelled ages derived from DiffArg_P. Overall the data suggests that a diffusion model for Ar distribution that includes a fully open grain boundary network and a zeroconcentration crystallization state is inadequate for the interpretation of the phengite, biotite, and plagioclase ages as cooling ages.

Implications: crystallization vs. contamination
As the 40 Ar/ 39 Ar phengite, biotite, and plagioclase ages measured in the WGR cannot be interpreted as simple thermally-activated diffusive loss in an open system, they may instead reflect either the timing of their crystallization, 'contamination' with excess 40 Ar, inefficient diffusive loss or a combination.

Phengite
The pseudosection in Fig. 4 and petrographic textures ( Fig. 2A) indicate that phengite crystallized at P b 25 kbar and T b 770°C. The timing of peak metamorphism has been constrained by U\ \Pb zircon at c. 410-400 Ma (Kylander-Clark and Hacker, 2014;Root et al., 2004;Young and Kylander-Clark, 2015). Therefore, if the 40 Ar/ 39 Ar phengite ages represent the timing of their crystallization, they should be similar.  cooling, the phengite ages should be 21-3 Ma younger than the timing of crystallization, depending upon the cooling rate. The observed phengite 40 Ar/ 39 Ar ages span a range of c.518-389 Ma, far greater than the spread expected for crystallization ages. The ages between c. 400-389 Ma imply that these grains lost some 40 Ar, but not enough compared to the loss expected for a fully open system. The ages N410 Ma suggest contamination that occurred during decompression, with Ar having diffused into the phengite from either the grain boundary fluid, adjacent grains, or incorporated into the phengite crystal lattice during crystallization. The data do not allow these two scenarios to be separated. The 'old' ages within the phengite crystals are concentrated in patches that are not associated with the geographic cores of the grains. Chemical maps (Fig. 3A) do not indicate the presence of zoning within the grains. The contaminating 40 Ar may therefore be held within microstructural defects in the phengite.

Biotite
The pseudosection suggests that the biotite crystallized during near isothermal decompression at P b 12 kbar, T~700°C and pre-dates titanite formation. Therefore, the timing of biotite crystallization is constrained by the youngest U\ \Pb zircon age (400 Ma) and the weighted mean titanite age (393 Ma).
The symplectite biotite ages lie between c.488-330 Ma, a greaterthan-expected spread given the grain size variation for a fixed regional cooling rate (Fig. 6). The larger idioblastic biotites range in age from c. 449-377 Ma; which is also greater than that predicted from simple open system cooling models (Fig. 6). Furthermore, the symplectite biotite is texturally older than the idioblastic biotite, yet yields ages that are less dispersed but lie within the same range. We therefore suggest that the idioblastic biotite ages show patchy evidence for 40 Ar contamination. This contamination mostly likely occurred during crystallization and growth, but may also have occurred during exhumation and cooling.

Plagioclase
Evidence for the presence of fluid that is enriched in 40 Ar comes from the dates yielded by the plagioclase that co-crystallized with the biotite during symplectization of HP phengite. The plagioclase yields very scattered and generally "old" ages that are considerably older than the ages yielded by the co-crystallizing biotite (Fig. 6). The source of the contaminating 40 Ar was most probably the reacting phengite. The 'excess' 40 Ar measured in the plagioclase implies that the grain boundary network was poorly interconnected, at least on the local scale.

Conclusions
Textural evidence for the reactions between different K-bearing minerals within the same sample combined with spatially-detailed insitu 40 Ar/ 39 Ar analyses can offer insights into how efficiently Ar is removed or recycled during high temperature metamorphism. In the WGR, gneisses that reached eclogite facies (~20-25 kbar) conditions   6. A plot of the UV laser ablation data vs. the predicted modelled age curves for A) Phengite, B) idioblastic bioite, C) symplectite biotite, and D) symplectite plagioclase. The predicted model age curves are shown with their greatest uncertainty as defined by the sensitivity modelling in Table 3b. at c. 410-400 Ma followed by isothermal decompression to b 12 kbar, record mineralogical evidence for all these stages. In-situ laser ablation 40 Ar/ 39 Ar data from phengite and the biotite-plagioclase symplectites that replace it yield wide age ranges that help to constrain how Ar behaves during recrystallization.
Phengite which grew at 20-25 kbar and 670-770°C and which subsequently partially broke down during decompression yields ages ranging from 518 to 389 Ma. Heterogeneous Ar distributions within and between grains suggest little diffusive resetting despite the high temperatures experienced after crystallization. Ages older than the estimated timing of crystallization suggest 40 Ar contamination during crystallization or decompression.
Symplectites that grew from phengite during decompression at P b 12 kbar document the partial redistribution of phengite-hosted Ar. 40 Ar/ 39 Ar ages of symplectite biotite grains that are "older" than those predicted from diffusion modelling suggest that either they incorporated Ar during their growth, or that they degassed inefficiently. The plagioclase that co-crystallized with the symplectite biotite yields ages that reflect contamination and incorporation of 40 Ar, possibly in fluid inclusions and not related to K in the plagioclse lattice. These data point towards plagioclase acting as a sink for the 40 Ar released by phengite during breakdown. 40 Ar/ 39 Ar ages of idioblastic biotite that grew at b12 kbar, show that as biotite coarsened following initial crystallization, Ar released during the breakdown of phengite was still present in the rock volume and that the grain boundary network within the rock was partially closed to Ar loss. These lines of evidence combine to indicate that the grain boundary network was partially closed during exhumation and symplectite-formation.
None of the measured ages can be reconciled with models of simple thermally-activated diffusion during cooling in a purely open system. Instead the ages of the phengite reflect contamination (which may have occurred during either crystallization or along the decompression path) and inefficient Ar loss. The plagioclase and biotite ages reflect continuing contamination by Ar released during the breakdown of phengite, within a grain boundary network with very restricted advection of either Ar or hydrous fluids.
Overall, these data demonstrate that even at high temperature metamorphic conditions, Ar may not diffuse efficiently out of micas unless there is sufficient advection of the grain boundary fluids. These datasets show that 40 Ar/ 39 Ar ages, and their interpretation may vary from sample to sample and mineral to mineral, even in the same terrain. The petrographic evolution of each rock should be carefully documented in order to provide a (relative) temporal framework in which to interpret the 40 Ar/ 39 Ar data. Furthermore, the analysis of multiple minerals aids the determination of how Ar has behaved during metamorphic recrystallization. Overall this approach allows 40 Ar mobility and distribution to be more precisely linked to a samples metamorphic evolution.