P–T constraints and timing of Barrovian metamorphism in the Shetland Islands, Scottish Caledonides: implications for the structural setting of the Unst ophiolite

Abstract: An integrated in situ monazite laser-ablation inductively coupled plasma mass spectrometry and metamorphic equilibria study is used to establish the P–T conditions and timing of Barrovian metamorphism in the Shetland Islands, Scottish Caledonides. The results have implications for the structural setting of the Unst ophiolite, which was obducted onto metasedimentary rocks of the Dalradian Supergroup. Metapelites in the footwall of the ophiolite yield U–Pb ages between 462 and 451 Ma with P–T conditions varying from c. 7.5 kbar and 550 °C directly below the ophiolite to c. 10 kbar and 775 °C at structurally deeper levels. The timing of peak metamorphism corresponds closely to that of Grampian (c. 450–470 Ma) metamorphism in mainland Scotland and Ireland, and Taconic (c. 450–460 Ma) metamorphism in the Appalachians, thus confirming the near-synchroneity of this important arc accretion event along the Laurentian margin. There is a significant metamorphic contrast between the low-grade rocks associated with the Unst ophiolite and the P–T conditions recorded in its footwall. If published K–Ar ages of c. 470 Ma broadly record obduction of the ophiolite, its present basal contact is probably a younger tectonic break that was associated with the excision of at least c. 10 km of crustal section. Supplementary material: Electron microporobe chemical analyses of garnet grains are available at www.geolsoc.org.uk/SUP18494

The origin and obduction history of ophiolites have been a source of significant debate since the 1970s (e.g. Dewey & Bird 1970;Moores 1970;Coleman 1971;Dewey 1976;Robinson et al. 1983;Moores et al. 2000;Wakabayashi & Dilek 2003;Metcalf & Shervais 2008;Dilek & Furnes 2009). Geochemical studies have demonstrated that most ophiolites do not originate at mid-ocean ridge systems but instead form in a 'suprasubduction zone' (SSZ) arc-forearc setting as a result of subduction rollback (Miyashiro 1973;Pearce & Cann 1973;Pearce et al. 1984;Dilek & Furnes 2009). Many SSZ ophiolites in orogenic belts were generated shortly prior to orogenesis during which they were thrust onto colliding passive continental margins (Dilek et al. 2007;Metcalf & Shervais 2008;Dilek & Furnes 2009). In a simple model, these ophiolites might therefore be expected to occupy the structurally highest parts of a forelandpropagating wedge of thickened strata, with Barrovian metamorphic grade increasing systematically downwards. However, considerable complications can arise from the potential existence within the orogenic wedge of out-of-sequence thrusts and extensional detachments. These can result from the tectonic extrusion of high-pressure rocks and obscure the relationship between ophiolites and their original footwall rocks (e.g. Maruyama et al. 1996). In view of the probable rheological contrasts between ophiolite sheets and their footwall metasedimentary rocks, it seems very likely that original obduction thrusts will be reactivated, possibly several times, to further complicate the tectonic evolution. To demonstrate an unambiguous link between the obduction of an ophiolite nappe and the tectonic evolution of rocks within its footwall, it is necessary to show (1) that obduction occurred shortly prior to peak Barrovian metamorphism (taking into account the time delay for heating of the footwall by conduction) and (2) that there is no significant contrast in peak metamorphic grade across the basal tectonic contact of the ophiolite.
The Unst ophiolite in the Shetland Islands, Scotland was obducted during the Ordovician-Silurian Caledonian orogeny. This is the northernmost and best exposed of a chain of SSZ ophiolites; other examples crop out intermittently between the Highland Boundary and Southern Upland faults in mainland Scotland and Ireland (Fig. 1a). These ophiolites are thought to have been obducted onto the margin of Laurentia during early Ordovician arc-continent collision that culminated in regional deformation and Barrovian metamorphism of footwall sedimentary successions (Lambert & McKerrow 1976;Dewey & Shackleton 1984;Dewey & Ryan 1990;Friedrich et al. 1999;Oliver et al. 2000;Chew et al. 2003Chew et al. , 2008Chew et al. , 2010. In Scotland and Ireland, the footwall successions include metasedimentary rocks of the Neoproterozoic to early Palaeozoic Dalradian Supergroup (Tanner & Sutherland 2007). The Unst ophiolite is unique within the Scottish Caledonides as it occupies a downfolded klippe that was emplaced onto stratigraphically low levels of the Dalradian Supergroup that record kyanite-grade metamorphism (Flinn 2000). In contrast, ophiolitic slices along the Highland Boundary Fault (e.g. Bute, Fig. 1a) occur in contact with stratigraphically high levels of the Dalradian Supergroup that record greenschistfacies metamorphism.
In this paper we assess the P-T conditions and timing of Barrovian metamorphism within the footwall metasedimentary successions of the Unst ophiolite and use these to place constraints on its structural setting. In situ U-Pb laser-ablation inductively coupled mass spectrometry (LA-ICPMS) data from monazite-bearing metapelitic assemblages and quantified P-T constraints demonstrate Ordovician-aged (Grampian) amphibolite-facies metamorphism. Peak pressures are estimated at 7-10 kbar and peak temperatures vary from 550 8C at the immediate footwall of the ophiolite to 775 8C within structurally deeper migmatitic metasediments. Existing geochronological constraints support a temporal link between ophiolite obduction and regional deformation and Barrovian metamorphism within the footwall. However, map relations and the significant metamorphic grade contrast between the ophiolite and its footwall rocks imply that the intervening contact is not the original obduction thrust but a younger tectonic break.

Regional geology
The Unst ophiolite crops out on the islands of Unst and Fetlar (Fig. 1b) in northern Shetland. It consists of two thrust sheets of ultrabasic and basic rocks (variably serpentinized peridotite with pyroxenite, dunite and gabbro) and associated low-grade metasedimentary rocks and mélanges that represent an original structural thickness of c. 8-10 km (Flinn 1985, 2000, andreferences therein). A U-Pb zircon age obtained from a plagiogranite within a gabbro unit of the ophiolite dates crystallization at 492 AE 3 Ma (Spray & Dunning 1991). K-Ar ages of c. 470 Ma that have been obtained from hornblende in the metamorphic sole of the ophiolite are interpreted to broadly date its Ordovician obduction (Spray 1988; see also Flinn et al. 1991).
The Unst ophiolite structurally overlies metapsammites and metapelites that have been correlated with the mid-Neoproterozoic to Cambrian Dalradian Supergroup in mainland Scotland and Ireland (Fig. 1b;Flinn et al. 1972;Flinn 1985;Prave et al. 2009). Structurally deeper metasediments of the Yell Sound Division to the west and SW (Fig. 1b) are correlated with the early Neoproterozoic Moine Supergroup of mainland Scotland (Flinn 1988). Read (1934) divided the metamorphic rocks of Unst into two units: a western Valla Field Block (including the pre-Dalradian Westing Group) and an eastern Saxa Vord Block, separated by a tectonic break located along Burra Firth (Fig. 1b). At least two sets of structural fabrics are present within the Dalradian rocks. An early ESE-WSW-trending mineral lineation associated with high-grade metamorphism is present throughout much of the Valla Field Block and Cannat (1989) suggested that this might correspond to the transport direction during ophiolite obduction. A younger NE-trending mineral lineation present in much of the Saxa Vord Block and the ophiolite nappes is associated with retrogressive metamorphism. Cannat (1989) identified abundant shear criteria that indicated a top-to-the-NE sense of shear parallel to this lineation, and suggested that final emplacement of the ophiolite resulted from transcurrent shearing in the terminal stages of the Caledonian orogeny.
Published geochronological data from the Dalradian rocks in Shetland include a wide range of Cambrian to Devonian K-Ar and 40 Ar/ 39 Ar mineral ages and some Rb-Sr whole-rock ages (Miller & Flinn 1966;Flinn & Pringle 1976), all of which are of uncertain significance. 40 Ar/ 39 Ar ages of c. 420 Ma obtained from Dalradian rocks and a metagranite in NE Unst appear to date end-Caledonian retrogressive metamorphism (Flinn & Oglethorpe 2005). Published P-T data from the Shetlands are limited to the metamorphic sole of the ophiolite, and the Yell Sound Division   (Spray 1988;Flinn et al. 1991;Flinn 1994). P-T estimates using garnet-clinopyroxene thermobarometry from assemblages in the ophiolite sole give c. 750 8C and a pressure of 3 kbar (the latter based on the estimated thickness of the overlying ophiolite; Spray 1988). Metamorphic conditions in the thrust sheets never exceeded those of the greenschist facies (Flinn 2001). Metagabbro bodies contain a heterogeneously developed greenschist-facies assemblage that includes albite, actinolite and epidote (Flinn 2001). The ophiolite itself contains extensive lizardite serpentinization (Flinn 2001). Phyllites associated with the lower thrust sheet carry a planar, bedding-parallel fabric defined by chlorite and sericite, with no relics of higher-grade minerals (Read 1934;Flinn 1958). On Yell, conventional thermobarometry from metapelitic assemblages and interlayered garnet-bearing mafic rocks of the Yell Sound Division gives conditions between 620 and 680 8C and 7 and 10 kbar (Flinn 1994).

Geological setting and sample petrography
Unst Sample FRQ-1 [UK National Grid Reference HP 63127 16454] was collected north of the Shetland ophiolite from a muscovitechlorite-garnet schist sequence that is part of the Saxa Vord Block of the Dalradian Supergroup in Shetland (Figs 1b and 2a; Read 1934). The sample contains garnet, staurolite, chlorite, chloritoid, muscovite, paragonite, plagioclase, ilmenite and quartz (Fig. 3a). Garnet grains up to 1 cm in diameter contain inclusions of staurolite, chlorite, chloritoid, paragonite, muscovite, quartz and plagioclase. Plagioclase occurs only as an inclusion in garnet. Garnet and staurolite porphyroblasts are wrapped by a finer-grained fabric of aligned muscovite, chlorite, paragonite, chloritoid and quartz (Fig. 3a). This fabric defines a penetrative foliation that dips east and is parallel to lithological layering. It is associated with a shallow-plunging, NNE-SSW-trending stretching lineation. On the basis of published K-Ar and 40 Ar/ 39 Ar ages obtained from NE Unst, this fabric is plausibly Silurian in age (Cannat 1989;Flinn & Oglethorpe 2005).
Sample KSH07-12 [HP 58781 11042] was collected on the western side of Unst from the Valla Field Block of the Dalradian Supergroup in Shetland (Read 1934). This sequence contains a pervasive foliation that dips gently to moderately to the east and is parallel to the lithological layering. The fabric contains a mineral stretching lineation that plunges toward the east. The sample contains kyanite, garnet, biotite, quartz, staurolite, muscovite, plagioclase and ilmenite. Kyanite occurs in two textures, as porphyroblasts up to 4 mm in size and in the matrix where it is finegrained and appears to partially replace staurolite. Garnet grains also occur as porphyroblasts up to 2 mm in size (Fig. 3b). The porphyroblasts of both kyanite and garnet contain inclusions of biotite, quartz and ilmenite that define a fabric that is not consistently oriented with respect to the matrix foliation. Both kyanite and garnet porphyroblasts are wrapped by the matrix fabric, which is defined by biotite, muscovite and finer-grained kyanite, and also contains staurolite, plagioclase, quartz and ilmenite. Many of the garnet porphyroblasts have rims that contain significantly less inclusions than the cores. These rims have an embayed appearance and the inclusions that they do have indicate a fabric parallel to the matrix foliation of the sample. Staurolite grains are small (200-500 ìm), and in some places have been partially replaced by finer-grained kyanite. Cannat (1989) suggested that the dominant high-grade mineral assemblages and deformation fabrics in these rocks developed during the Ordovician.

Yell
Samples SH-9 [HP 54080 98765] and SH-24 [HP 54996 96973] were both collected in eastern Yell from the Yell Sound Division (Fig. 1b). The samples were collected from sequences of kyanitegarnet-staurolite-quartz-muscovite-biotite AE plagioclase-bearing metapelitic schist (Fig. 2b) that were interlayered with quartzite. The fabric in both samples is dominated by muscovite with some biotite and is steep, striking roughly north-south. In both samples, kyanite is weakly aligned. With the exception of plagioclase and ilmenite, which occur only in sample SH-9, both samples contain garnet, quartz, muscovite, biotite, kyanite, rutile (f) Peak mineral assemblage of sample SH-31c. (g) Replacement of sillimanite by muscovite in sample FRQ-20a. The inset shows a muscovite halo around a garnet grain separating garnet and biotite, indicating that muscovite has also grown at the expense of garnet and biotite. (h) Peak mineral assemblage from sample FRQ-20b. and relict staurolite armoured in coarse-grained quartz (SH-24, Fig. 3c) or plagioclase grains (SH-9, Fig. 3d). Garnet grains are up to 1 cm in diameter in SH-24 and poikiloblastic, containing inclusions of rutile, staurolite, biotite, quartz and chlorite. Garnet grains in sample SH-9 are about 3 mm in diameter and contain inclusions of rutile, ilmenite, chlorite and quartz.

Lunna Ness
The rocks of Lunna Ness (Fig. 1b) consist of migmatitic metapsammites and metapelites of the Yell Sound Division where melting has produced garnet (Fig. 2c). These rocks are intruded by amphibolite sheets and a prominent metagranite, the Valayre Gneiss ( Fig. 2d; Flinn 1988).
Sample SH-29 [HP 51620 74106] was obtained from the Valayre Gneiss (Fig. 2d). The sample is a foliated granite gneiss that contains garnet, biotite, plagioclase, muscovite, quartz and large (up to 3 cm) ovoid K-feldspar porphyroclasts (Fig. 3e). Garnet grains are up to 2 mm in size and contain inclusions of muscovite, biotite and quartz. Fine-grained muscovite partially replaces K-feldspar in places (Fig. 3e). The granite gneiss contains abundant metre-to centimetre-scale boudins and lenses of mafic garnet-amphibolite. These lenses contain a variably developed foliation that is misaligned in adjacent mafic bodies. This would suggest that these lenses record a phase of deformation that predated their incorporation in the protolith of the granitic gneiss. The regional fabric within the Valayre Gneiss dips moderately to steeply to the NW to west but is locally variable in orientation in zones where there are a high proportion of mafic schollen.
Sample SH-31c [HP 51851 74117] was obtained from the Yell Sound Division. The locality sampled is dominated by highly migmatized meta-semipelites that contain concordant metre-to centimetre-scale boudins of garnet-amphibolite (Fig. 2e). The meta-semipelites are characterized by an undulating to subhorizontal foliation that is locally cross-cut by centimetre-scale layers of leucosome. Sample SH-31c is a migmatized metasemipelite originating from within a north-dipping high-strain zone. It contains garnet, biotite, plagioclase, K-feldspar, quartz and ilmenite with the foliation defined by biotite and ilmenite. Garnet grains are up to 3 mm in size, and contain inclusions of ilmenite, quartz and biotite. The larger garnet grains are wrapped by the matrix foliation, and contain inclusions that define an earlier fabric. This earlier fabric has no systematic angular relation with the external foliation of the sample. The smaller garnet grains overgrow the foliation. Some garnets are mantled by a 'corona' of biotite (Fig. 3f).
Sample FRQ-20 (HP 52301 73112) is a migmatized garnetkyanite-sillimanite-bearing metapelite from the Yell Sound Division (Fig. 2f). The foliation is defined by sillimanite or muscovite ( Fig. 3g and h); it dips moderately to the west and contains a westerly plunging mineral stretching lineation. The metasediments at and around the sample locality contain abundant lenses of deformed and migmatized mafic rocks. Two samples were obtained from this location. The first, FRQ-20a, contains garnet, biotite, muscovite, plagioclase, rutile, ilmenite and rare sillimanite (Fig. 3g). Muscovite defines the foliation and also occurs as large unfoliated aggregates. Garnet grains in this sample are up to 3 mm in size and are also rimmed by muscovite (Fig. 3g), and plagioclase grains contain unfoliated muscovite. Garnet grains contain inclusions of kyanite, biotite, rutile and quartz. Rutile is present in the matrix, and is commonly rimmed by ilmenite. The second sample, FRQ-20b, was collected less than 1 m from FRQ-20a. This sample contains similar minerals to FRQ-20a with the addition of K-feldspar and rare kyanite (Fig. 3h). Muscovite is not present in this sample and the abundance of sillimanite and rutile is higher than in sample FRQ-20a whereas the abundance of ilmenite is lower (Fig. 3h). Garnet is up to 2 mm in size and contains inclusions of kyanite, biotite, rutile, plagioclase and quartz. Both kyanite and sillimanite are present in the matrix; however, sillimanite is more common and defines the foliation, wrapping garnet grains (Fig. 3h). In places, especially near sillimanite, rutile is rimmed by ilmenite.

Analytical methods
Not all samples used in this study were suitable for all the applied techniques. Table 1 presents a summary of the techniques that were applied to each of the samples.

Mineral chemistry
Compositional traverses of garnet grains for selected samples (FRQ-1, KSH07-12, SH-9, SH-24 and FRQ-20; Fig. 4; supplementary table) were obtained using a Cameca SX51 Electron Microprobe at Adelaide Microscopy, located at the University of Adelaide. Quantitative analyses were run at an accelerating voltage of 15 kV and a beam current of 20 nA, with a beam diameter of 2-3 ìm. Analyses were collected on wavelengthdispersive spectrometers and all data are included in a supplementary table.
Compositional mapping was conducted on garnet grains from samples FRQ-1, KSH07-12, SH-9, SH-24 and FRQ-20 using the same instrumentation but an accelerating voltage of 15 kV and a beam current of 110 nA (Fig. 4).

LA-ICPMS monazite geochronology
In situ U-Pb monazite dating was undertaken by LA-ICPMS at the University of Adelaide following the method of Cutts et al. (2010). Thin sections from samples SH-9, SH-29, SH-31c, KSH07-12 and FRQ-20 were imaged using a back-scattered electron (BSE) detector on a Phillips XL30 SEM to determine the textural location of monazite and its compositional variability. U-Pb isotopic analyses were acquired using a New Wave 213 nm Nd-YAG laser in a He ablation atmosphere fixed to an Agilent 7500cs ICP-MS system. A spot size of 12 ìm was used for all analysed samples. A 50 s gas blank was collected followed by 40 s of sample analysis. Prior to each ablation, the laser was fired for 10 s with the shutter closed to allow crystal and beam stabilization. Isotopes measured were 204 Pb, 206 Pb, 207 Pb and 238 U for 20, 30, 60 and 30 ms, respectively. Common lead was not corrected for in the age calculations owing to unresolvable interference of 204 Hg on the 204 Pb isotope peak. To compensate  for this, the 204 mass peak was monitored during analysis and analyses were omitted if appreciable common lead was observed. Monazite age calculations were determined using the real-time correction program Glitter, which was developed at Macquarie University, Sydney (Jackson et al. 2004). U-Pb fractionation was corrected using the Madel monazite standard (thermal ionization mass spectrometry (TIMS) normalization data: 207 Pb/ 206 Pb age ¼ 491.7 Ma; 206 Pb/ 238 U age ¼ 514.8 Ma; 207 Pb/ 235 U age ¼ 510.4 Ma; Payne et al. 2008) and accuracy was verified using an in-house standard of known age (94-222/Bruna-NW, c. 450 Ma; Payne et al. 2008) for samples SH-9, SH-29, SH-31c and KSH07-12. The Madel standard contains two age domains separated by c. 4 Ma, with the younger domain ( 207 Pb/ 235 U age of 510 Ma) being the dominant one used in this study. This age variation is well within the uncertainty of the LA-ICPMS monazite method and to further ensure against inaccurate age reporting, an overestimated uncertainty of 1% relative was assigned to the age of the standard for the propagated sample age error calculations. Samples and standards were also analysed over multiple analytical sessions. Over the duration of this study the average normalization ages for the Madel standard were 207 Alienikoff et al. 2006) and accuracy was verified using the Madel monazite standard as well as the 94-222/Bruna-NW in-house standard. Over the duration of this study, the average normalization ages for the 44069 standard were 207 Pb/ 206 Pb ¼ 421 AE 12 Ma, 206 Pb/ 238 U ¼ 420 AE 3 Ma and 207 Pb/ 235 U ¼ 420 AE 3 Ma (n ¼ 24), whereas the weighted average 206 Pb/ 238 U of the Madel standard was 514 AE 9 Ma (n ¼ 8) and the weighted average 206 Pb/ 238 U age of 94-222/Bruna-NW was 460 AE 5 Ma (n ¼ 8). All concordia and weighted mean age calculations were conducted using Isoplot v.3.0 (Ludwig 2003). All analyses are given in Table 2.
Pseudosections were calculated for the geologically realistic system MnNCKFMASHTO (MnO- The modelling for this system uses the a-x relationships of White et al. (2007) for silicate melt; a combination of Mahar et al. (1997) and Holland & Powell (1998) for cordierite and staurolite; White et al. (2005) for garnet, biotite, ilmenite and hematite; White et al. (2002) for orthopyroxene, spinel and magnetite; a combination of Mahar et al. (1997) and White et al. (2000) for chloritoid; Coggon & Holland (2002) for muscovite and paragonite; and Holland & Powell (2003) for plagioclase and alkali feldspar.
The bulk composition of each sample, obtained by whole-rock X-ray fluorescence (XRF) analysis, was used to model the expected mineral assemblages and the mineral compositions on calculated P-T pseudosections. The constraint on maximum H 2 O content is taken as equivalent to the 'loss on ignition' from the XRF analyses. The proportion of Fe 2 O 3 to FeO has been estimated by considering the abundance of Fe 3þ -bearing minerals and modal constraints in the context of recalculated electron microprobe analyses (using the method of Droop 1987). Compositional isopleths for garnet were calculated and have been plotted onto the peak field of the pseudosections to aid with interpretation of the P-T path.

Results
Garnet mineral chemistry FRQ-1. Garnet in sample FRQ-1 ( Fig. 4a and b) shows X alm increasing from core to rim. X prp and X grs also increase slightly, whereas X sps decreases (Fig. 4a). The increase in X alm and X prp and decease in X sps are consistent with growth zoning during prograde metamorphism Spear 1993;Baxter et al. 2002). At the very edge of the garnet there is an increase in X sps and decrease in X grs , suggesting minor garnet resorption by Mn-poor, Ca-rich matrix minerals ( Fig. 4a and b).
KSH07-12. Garnet in sample KSH07-12 has slightly more complex chemical zoning trends than sample FRQ-1 ( Fig. 4c-f). From the garnet core, X sps increases toward the rim, X Fe decreases slightly toward the rim and X grs is flat. However, at c. 500 ìm from the edge of the garnet, the cations change their trend. X sps decreases significantly whereas X grs and X Fe increase. This corresponds approximately to the change in the abundance of inclusions within the garnet and is more pronounced on the right-hand side of the traverse (Fig. 4c). Compositional mapping of this garnet grain revealed that the asymmetric zonation was due to subsequent rim garnet growth, which contains inclusions defining a foliation parallel to the foliation in the matrix of the sample (Fig. 4d-f).
SH-24. Garnet in sample SH-24 has a homogeneous composition for X prp and X grs ; however, X sps and X alm appear to preserve prograde zoning with X alm increasing toward the rim and X sps decreasing toward the rim (Fig. 4g). The same pattern was seen in the compositional maps for this sample for X prp , X sps and X alm ; however, X grs appears to increase from the core until about 800 ìm from the rim and then decrease to the rim (Fig. 4h).
SH-9. Garnet in sample SH-9 has a similar zoning profile to sample SH-24 with homogeneous X prp and X grs profiles, but in addition has a homogeneous composition for X alm with only X sps preserving prograde zoning and decreasing toward the rim of the grain ( Fig. 4i and j).
FRQ-20. Sample FRQ-20 displays homogeneous zoning profiles across traverse 1 for X alm , X prp and X sps : However, traverse 2, which is on the rim of the garnet grain adjacent to biotite, has increasing X alm and X sps whereas X prp decreases close to the edge of the grain. Along traverse 1, X grs decreases from core to rim with a slight rise again on the right-hand side ( Fig. 4k and l). Along traverse 2, X grs appears to be homogeneous. The compositional map for this sample also indicates that X sps is homogeneous for this garnet grain whereas X alm and X pyr rise and fall respectively where the grain boundary is in contact with biotite (Fig. 4l).

LA-ICPMS monazite geochronology
KSH07-12. Monazite grains (Table 2) are subhedral and up to 100 ìm in size, with most around 20 ìm. Under BSE imaging,  grains are generally unzoned (Fig. 5a). Monazite grains occur within the matrix of the sample as well as in the rims of garnet grains; however, no difference in age was detected. Several matrix monazite grains are surrounded by a region of alteration, and all monazite grains contain common lead. Owing to the presence of common lead, a Yorkfit regression was applied to the data to determine the initial Pb composition of the sample. This regression gave a calculated 207 Pb/ 206 Pb intercept of 0.544412 with an MSWD of 1.5. The analyses from this sample were plotted on a Terra-Wasserburg concordia diagram with the Terra-Wasserburg regression anchored to the common lead composition estimated from the Yorkfit intercept. This produced an intercept age of 462 AE 10 Ma (MSWD 1.5; Fig. 5b). High positive weighted residuals (.2.5) were not recorded for any analysis. However, one analysis had a high negative weighted residual indicating inherited Pb. Removing this analysis does not alter the age except to reduce the errors so it has been left in the dataset.
SH-9. Monazite grains are subhedral, up to 100 ìm in size and display no zoning under BSE imaging (Fig. 5c). There is no detectable difference in age based on the microstructural location of grains, with monazites in the matrix and in cracks in garnet producing ages within error. Two analyses produced older ages (Table 2; Fig. 5d). These analyses both occurred on matrix grains that had no discernible atomic number contrast (BSE contrast) with grains that gave younger analyses. In fact, analysis 9a, which gives an older age, originates from the same grain as analysis 9b and there is no obvious difference in the BSE response in this grain (Fig. 5c). The weighted average 206 Pb/ 238 U age of the younger monazite population from this sample is 451 AE 4 Ma (n ¼ 18; Fig. 5d).

SH-29.
Monazite grains are subhedral and up to 200 ìm in size. Several grains are surrounded by an irregular hole or region of altered sample (Fig. 5e). There is a textural control on age, with monazite grains occurring within garnet generally having an older age than those occurring in the matrix. Additionally, some analyses of the cores of zoned matrix monazite grains give older ages (Fig. 5e). Analyses define a discordia with intercepts at 913 AE 37 and 458 AE 16 Ma (Fig. 5f). The majority of analyses form a group at the younger intercept, and this has a weighted average 206 Pb/ 238 U age of 459 AE 4 Ma (n ¼ 13; Fig. 5f). The upper intercept is tied by a c. 920 Ma concordant analysis as well as a second c. 900 Ma analysis that is less than 10% discordant (Fig. 5f). These analyses were obtained from monazite grains within garnet.
SH-31c. Monazite grains are subhedral, up to 150 ìm in size and display no zoning under BSE imaging (Fig. 5g). There is no textural control on age, with matrix monazite and those grains located along cracks in garnet grains giving ages within error. The weighted average 206 Pb/ 238 U age of this sample is 458 AE 3 Ma (n ¼ 20; Fig. 5h).
FRQ-20. Monazite grains are subhedral, up to 100 ìm in size and pale yellow to green in colour. The monazite grains are zoned under BSE imaging with some grains appearing to contain distinct cores (Fig. 5i). These cores generally give older ages; however, the age difference between the cores and the rims is within analytical error. The difference in age between cores and rims is best demonstrated by the grain depicted in Figure 5i. As this age difference is within analytical error, it is difficult to resolve. For this reason, the age presented for this sample is a total (n ¼ 29) weighted average 206 Pb/ 238 U age of 467 AE 4 Ma     20k, 20l, 20m, 20n, 20o, 20p, 20c2 and 20d2. (j) Concordia plot of all monazite data from sample FRQ-20. A weighted mean 206 Pb/ 238 U age is given for analyses from monazite cores (with a grey border). The analyses from interpreted monazite cores are displayed in grey on the concordia plot. A weighted mean 206 Pb/ 238 U age is also given for interpreted monazite rims and unzoned grains (with a black border). These analyses are displayed in black on the concordia. (Fig. 5j). However, if the analyses are divided between those from the cores identified by BSE imaging and those remaining, the cores give a weighted average 206 Pb/ 238 U age of 474 AE 6 Ma (n ¼ 12, MSWD ¼ 0.77), whereas the rims and apparently unzoned grains give a weighted average 206 Pb/ 238 U age of 462 AE 3 Ma (n ¼ 17, MSWD ¼ 0.89; Fig. 5j).
Pressure-temperature conditions FRQ-1. The P-T path can be defined by the change in mineral assemblage in this sample and also by the garnet chemical zoning (Fig. 6a). Within garnet grains in this sample, there are numerous inclusions of chloritoid, staurolite, muscovite, ilmenite, quartz, chlorite and paragonite as well as some minor plagioclase inclusions. These suggest that the earlier (prograde) P-T evolution of the sample involved passing through mineral assemblage stability fields that contained some combination of the included minerals. It is improbable that the up-temperature part of the P-T path passed through a single mineral assemblage field in which all of the included minerals were simultaneously in equilibrium with garnet. However, the pseudosection for this sample (Fig. 6a) indicates that a mineral assemblage stability field involving garnet and all of its inclusions occurs at 525-550 8C and 4-7 kbar. Therefore the prograde path may have passed through this field. Because garnet grains are prograde zoned, the garnet composition can be used to indicate the prograde path that the sample experienced (e.g. Cutts et al. 2009a). For this reason compositional isopleths have been calculated and plotted in fields of interest. These isopleths suggest that garnet growth initiated at c. 5.5 kbar and 540 8C within the garnet-chloritoid-staurolite-muscovite-paragonite-ilmenitechlorite-plagioclase-quartz field. Based on the absence of plagioclase within the matrix of the sample, as well as the abundance of chloritoid and staurolite, the peak assemblage for this sample is interpreted to be garnet + chloritoid + staurolite + muscovite + paragonite + ilmenite + chlorite + quartz. This gives peak P-T conditions for this sample of c. 7.5 kbar and 550 8C.
KSH07-12. Based on the textures preserved in this sample, it is possible that it records two separate metamorphic events. The first is represented by the porphyroblastic garnet and kyanite grains and the mineral inclusions present within these grains. This reflects an assemblage of garnet + kyanite + biotite + muscovite + quartz + plagioclase. The flat zonation of the X grs and X Fe isopleths ( Fig. 4c-e) indicates that the sample reached a high temperature and on the P-T pseudosection (Fig. 6b) these isopleths intersect at P-T conditions of 7 kbar and c. 650 8C.
The second metamorphic event recorded in this sample is represented by the matrix assemblage that comprises muscovite, biotite, quartz, plagioclase, staurolite, second generation kyanite and second generation garnet. The composition isopleths of the second generation garnet intersect within the kyanite-bearing field (Fig. 6b). Considering the errors on the isopleths and the stable matrix mineral assemblage, this implies a peak metamorphic assemblage of garnet + kyanite + staurolite + biotite + muscovite + quartz + plagioclase, indicating P-T conditions of c. 7.5 kbar and 630 8C (Fig. 6b).
SH-24. Biotite, staurolite, chlorite, rutile and quartz occur as inclusions within garnet. These suggest that the early (prograde) P-T evolution of the sample involved passing through mineral assemblage stability fields that contained some combination of the included minerals. It is also possible that the up-temperature part of the P-T path passed through a single mineral assemblage field in which all of the included minerals were simultaneously in equilibrium with garnet. Based on the P-T pseudosection, it is also likely that the sample contained a small amount of plagioclase; however, the calculated modal proportion of plagioclase for the fields of interest in the diagram is small (around 3%), which means that it may not be readily observed in the rock (Fig. 6c). Although muscovite is not found as an inclusion in garnet, its prevalence in the matrix suggests that it may have formed part of an early assemblage along the prograde P-T path.
This suggests that the sample may have passed through fields containing muscovite in addition to the minerals included in garnet listed above during its prograde evolution. Staurolite within garnet provides a key constraint on the locus of the prograde P-T path. From staurolite-to garnet-bearing mineral assemblage fields in Figure 6c the prograde P-T path may have passed through P-T conditions of 6 kbar and c. 600 8C (Fig. 6c).
As both chlorite and staurolite are absent from the matrix the P-T evolution requires the loss of both of these minerals, moving up pressure and temperature. As the matrix contains muscovite and ilmenite, these are inferred to be part of the peak assemblage, comprising garnet + kyanite + biotite + muscovite + rutile + quartz AE plagioclase. Based on the above, the peak metamorphic conditions for this sample are 9-10 kbar and c. 650 8C (Fig. 6c). Garnet isopleths were calculated for the field that contains the peak metamorphic conditions to provide further constraints. However, the garnet grains have been heated to sufficiently high temperatures that diffusional re-equilibration has occurred, producing garnet grains with homogeneous or mixed zoning profiles ( Fig. 4f and g; e.g. Spear et al. 1991;Vance 1995). The potential significance of these compositionally homogeneous garnets will be explored in the discussion.
SH-9. Biotite, staurolite, rutile, ilmenite, chlorite, plagioclase and quartz occur as inclusions within garnet. These suggest  Kretz (1983). (a) Sample FRQ-1. The dashed, dotted and straight lines in the grt-chl-pl-pg-ms-ilm-cld-st-qtz field are garnet compositional isopleths, with those in bold corresponding to the composition of the garnet core in this sample. Fe isopleths for all samples correspond to X alm values. The dashed black arrow indicates the interpreted prograde P-T path for this sample. The black ellipse represents the conditions of initial garnet growth based on the intersection of the garnet compositional isopleths from the cores of garnets from this sample. The field of interpreted peak conditions has bold edges. (b) Sample KSH07-12. Within fields of interest, dashed, dotted and straight lines represent garnet compositional isopleths. The white ellipse indicates initial P-T conditions for the early assemblage of the sample based on the intersection of the isopleths for the garnet core (in bold). Isopleths for the garnet rim are given in double bold. The dashed white arrow indicates the interpreted P-T path for this earlier event based on mineral textural relations. The field with the black edges indicates the position of the matrix assemblage and interpreted peak Caledonian conditions. (c) Sample SH-24. The field with the black edges indicates the interpreted peak P-T conditions of the sample; the dashed black arrow represents the interpreted prograde P-T path. Dashed, dotted and straight lines represent garnet compositional isopleths, with those in bold corresponding to the core composition of the garnets from this sample. (d) Sample SH-9. The field with the black edges indicates peak P-T conditions of the sample; the dashed black arrow represents the prograde P-T path. Within the peak field, dashed, dotted and straight lines represent garnet compositional isopleths with those in bold corresponding to the core composition of garnets from this sample. that the early (prograde) P-T evolution of the sample involved passing through mineral assemblage stability fields that contained some combination of the included minerals. It is improbable that the up-temperature part of the P-T path passed through a single mineral assemblage field in which all of the included minerals were simultaneously in equilibrium with garnet. However, the variety of mineral inclusions places some constraints on the locus of the prograde P-T path and suggests that it passed through c. 5 kbar and 600 8C (Fig. 6d). The P-T evolution is defined by the loss of chlorite, which is not present in the matrix, and the addition of both muscovite and kyanite, which are part of the matrix assemblage of the sample. Staurolite, which only occurs shielded within large plagioclase or quartz grains, is also absent from the peak mineral assemblage. These observations constrain a P-T path that involves a near-isothermally up-pressure increase to the interpreted peak assemblage of garnet + biotite + muscovite + kyanite + plagioclase + quartz + rutile + ilmenite. This assemblage equates to peak P-T conditions of 7-10 kbar and c. 650 8C (Fig. 6d).
Garnet compositional isopleths have been calculated and plotted for the peak assemblage of the sample. However, as in sample SH-24, the garnet grains have been heated to sufficiently high temperatures to result in homogenization of compositions ( Fig. 4h and i). Based on the measured garnet composition and the isopleth calculations, we suggest that the homogeneous garnet may have been in equilibrium with the matrix assemblage at conditions of c. 8 kbar and 650 8C (Fig. 6d; e.g. Spear et al. 1991).
FRQ-20. Sample FRQ-20 experienced the highest metamorphic grade of all the samples examined in this study. Potentially as a consequence of this, there is little prograde metamorphic evolution preserved. In this sample, the peak conditions are given by the inclusion assemblage of the garnet, giving a mineral assemblage of garnet-plagioclase-kyanite-biotite-K-feldsparquartz-ilmenite-rutile. Once again, garnet grains are extensively compositionally homogenized. Therefore, the garnet zoning cannot be relied upon to indicate peak conditions ( Fig. 4j and k; e.g. Spear et al. 1991;Vance 1995). Based on the peak mineral assemblage, peak P-T conditions are c. 10 kbar and 750-800 8C (Fig. 7).
The matrix of sample FRQ-20b contains sillimanite, which is interpreted to overprint the kyanite-bearing peak assemblage (Fig. 3h). In the vicinity of the sillimanite-rich domains, rutile is partially or completely replaced by ilmenite (Fig. 3h). This suggests that the retrograde P-T path goes down-pressure into the sillimanite stable field, and then down P and T away from rutile stability (Fig. 7).
In sample FRQ-20a, K-feldspar is absent from the assemblage, and there is extensive muscovite present in the matrix, including distinct layered domains of fine-grained muscovite that are interpreted to overprint the sillimanite-rich domains that are present in sample FRQ-20b (Fig. 3g). This suggests that Kfeldspar and sillimanite were removed from the assemblage by the growth of muscovite (Fig. 3g). Muscovite coronas also enclose garnet and biotite, indicating that the P-T evolution must follow a path where garnet abundance decreases (Fig. 3g). This results in a retrograde assemblage of garnet-plagioclase-muscovite-biotite-ilmenite giving P-T conditions of c. 6-7 kbar and 650 8C.

Discussion and conclusions
Metamorphic evolution and garnet zonation Figure 8 shows a compilation of the P-T and age constraints obtained in this study. Also shown in Figure 8 is an additional data point of previously published conventional P-T data (Cutts et al. 2009b).
The P-T-t conditions determined in this study suggest a general southwestward increase in metamorphic grade and depth of exhumation ranging from c. 7 kbar in the immediate footwall of the ophiolite to .10 kbar in the Lunna Ness area. This increase in metamorphic grade is also associated with a loss of prograde zonation in garnet (compare Saxa Vord sample FRQ-1 with Lunna Ness sample FRQ-20). In sample SH-24, garnet grains range up to 1000 ìm in size. Assuming that the garnet originally grew with a prograde compositional zoning profile, it is possible to infer what the ideal prograde zoning would have looked like by utilizing the texturally constrained P-T path of SH-24 and a consideration of the modelled garnet abundance and compositional isopleths (Fig. 9). Based on diffusion data (e.g. Ganguly et al. 1998;Dutch & Hand 2010), a garnet with such a The field with the black edges indicates peak P-T conditions of the sample; the dashed black arrow represents the interpreted retrograde P-T path for this sample. Within the peak field, dashed, dotted and straight lines correspond to garnet compositional isopleths, with those in bold corresponding to the core composition of the garnets from this sample. Mineral abbreviations are from Kretz (1983). zoning profile would require a minimum of c. 20 Ma to compositionally homogenize at 650 8C. This suggests that if diffusional re-equilibration is the reason that the garnet compositions are homogeneous, the rocks must have been at depth for significant periods of time. In FRQ-1, with an inferred peak temperature of c. 550 8C, a similar residence time within the orogen would have led to relatively minor compositional modification. The apparent lack of retrograde zoning in garnet grains from most of the samples may imply that exhumation from peak conditions was rapid or that fluid was lost or absent such that further reaction could not occur.
Garnet grains from sample FRQ-20b appear to preserve diffusive cooling profiles. This is evidenced by an abrupt drop in X Mg and increase in X Fe at the edge of the grain where the rims of the grains are in contact with biotite (e.g. Anderson & Buckley 1973;Hauzenberger et al. 2005;Robl et al. 2007). From these profiles it is possible to determine the rate of cooling using garnet-biotite diffusion (e.g. Anderson & Buckley 1973;Robl et al. 2007). Figure 10 shows a comparison between the actual profile preserved in FRQ-20 and profiles generated by the modelling code of Robl et al. (2007) by varying the rate of cooling. The modelling suggests that cooling rates between 12 and 40 8C Ma À1 from 775 to 625 8C produce similar X Mg profiles to what is seen in sample FRQ-20b (Fig.  10). Based on the P-T pseudosection generated for FRQ-20b, cooling from 775 to 625 8C corresponds to exhumation from c. 10 to 5 kbar. This implies that sample FRQ-20b returned to pressures of 5 kbar within 4-13 Ma of peak metamorphism and suggests an exhumation rate of 1-4 mm a À1 : The 206 Pb/ 238 U monazite ages for samples in this study vary between 462 and 451 Ma. Given the metamorphic grade of the samples it is likely that these ages are metamorphic rather than  detrital (e.g. Smith & Barreiro 1990;Spear & Pyle 2002;Fitzsimons et al. 2005). With the exception of sample SH-29, there was no discernible age difference between matrix monazite and monazite included in garnet. This could be due to monazite grains being located along cracks in the garnet and thus not being isolated from chemical communication with the external or matrix part of the rock. Alternatively, garnet growth may have occurred sufficiently fast such that the ages of monazite included in the garnet are within error of the matrix monazite age. Samples SH-9, SH-29 and FRQ-20 all have evidence of older monazite populations as either cores in matrix grains  or a separate monazite population included in garnet (SH-29). In the case of SH-9, the older monazite population is not visibly distinguishable from the younger population and the difference in age is not significant. This minor age difference could be the result of partial resetting of older grains, perhaps sourced from older metamorphism in Shetland (e.g. Cutts et al. 2009b) or detrital grains. For sample FRQ-20, the older population (c. 474 Ma) was found in cores of large grains and could plausibly have grown at an early stage of Grampian metamorphism in response to a prograde silicate (garnet-bearing) reaction that produced monazite (Hermann & Rubatto 2003;Spear & Pyle 2010; see also Kelsey et al. 2008). Monazite from sample SH-29 gives ages that fall on a discordia between c. 460 Ma and c. 920 Ma, where most of the older ages originate from monazite grains shielded within garnet grains. Although we cannot preclude the possibility that these grains represent a detrital population in the sedimentary source for the Valayre gneiss, a similar age (c. 930 Ma) for metamorphism has been established for the Westing Group of western Unst (Cutts et al. 2009b). We therefore suggest that the Valayre Gneiss and the host Yell Sound Division may have been previously metamorphosed at c. 920 Ma, although this clearly requires further investigation. It is also plausible that the early metamorphism recorded in sample KSH07-12 is part of this early Neoproterozoic event. This sample was obtained a few kilometres north of the locality where evidence for the early Neoproterozoic metamorphism (c. 930 Ma) was found by Cutts et al. (2009b). Unfortunately, monazite inclusions within the cores of garnet grains are rare in sample KSH07-12 so the age of this earlier metamorphism is undefined.

Comparison with similar aged events in the Caledonide-Appalachian belt
The 206 Pb/ 238 U monazite ages of all the samples vary between 462 and 451 Ma, indicating that peak metamorphism occurred in this age range. This is essentially within error of a number of wellconstrained isotopic ages that have been obtained for the peak of Grampian orogenic activity within the Dalradian Supergroup on mainland Scotland. These include Sm-Nd ages for prograde garnet growth (Oliver et al. 2000;Baxter et al. 2002) and U-Pb zircon ages obtained from granites and gabbros that were emplaced close to peak metamorphism at c. 465 Ma (Oliver et al. 2000;Dempster et al. 2002). Further west in western Ireland, Friedrich et al. (1999) concluded that peak metamorphism and anatexis of the Dalradian Supergroup in Connemara occurred at c. 468 Ma based on the U-Pb dating of zircon and titanite from various metasedimentary and meta-igneous lithologies. Pressure estimates of 7.5-10 kbar for Grampian metamorphism in Shetland are similar to those reported from the North Mayo Inlier in western Ireland (Yardley et al. 1987) and the SW Scottish Highlands (Graham 1985). From a wider perspective, the ages presented here are also similar to granulites-facies metamorphism of the Taconic arc-accretion event in the Appalachians (Karabinos et al. 1998;van Staal et al. 1998) dated at c. 450-460 Ma (Moecher et al. 2004;Corrie & Kohn 2007). The new data reported here from Shetland thus confirm the near-synchroneity of Grampian-Taconic metamorphism along the eastern margin of Laurentia.

Metamorphic perspective on the emplacement of the Unst ophiolite
Although ophiolites along the Highland Boundary Fault in Scotland and Ireland were probably emplaced onto the Laurentian passive margin early in the Grampian orogenic event, they are of insufficient structural thickness to have led directly to the deep burial and Barrovian metamorphism of Dalradian rocks in their Fig. 10. X Mg core to rim profile from sample FRQ-20b with outputs from Thermal History (Robl et al. 2007) for cooling rates of 0.5, 5, 12, 20 and 40 8C Ma À1 : The conditions used to calculate these rates were: biotite X Mg 0.59, temperature step of 1, 30 nodes, and 0.3:0.7 biotite to garnet ratio. Linear cooling was used and the garnet radius was set at 873 ìm. Thermal History uses the diffusion coefficients of Ganguly et al. (1998). footwall (Chew et al. 2010). This is reinforced by the metamorphic data reported here. Samples FRQ-1 and KSH07-12 were collected close to the base of the ophiolite and their metamorphic assemblages indicate c. 22-24 km of burial. The estimated c. 8-10 km total thickness of the ophiolite (Flinn 2000) is insufficient to produce the overburden required by the metamorphic data. Conceivably the ophiolite was imbricated at higher structural levels to create a much larger structural thickness that has subsequently been removed by erosion. However, the ophiolite thrust sheets are separated by sub-biotite grade (c. 300-400 8C; Dempster & Tanner 1997) metasedimentary rocks (Flinn 1985(Flinn , 2000. If the metamorphic overburden had been generated by imbrications of the ophiolite, these metasedimentary rocks between ophiolite thrust sheets would be expected to contain assemblages of similar metamorphic grade to those observed beneath the basal thrust sheet. In addition, the ophiolite itself contains extensive lizardite serpentinization (Flinn 2001). Above 300 8C, lizardite is no longer the most stable form of serpentinite (O'Hanley 1996;Mével 2003;Evans 2004). If we assume that the ophiolite was at the top of the structural pile and assign it the average geothermal gradient experienced by samples in this study (27.2 8C km À1 ), the ophiolite must have experienced pressures below 4 kbar for lizardite to remain stable. Thus the presence of such low-grade metamorphic rocks within the ophiolite complex and the low grade of the complex itself suggest that it was never sufficiently thick enough to generate the metamorphism recorded in its footwall. Barrovian metamorphism is more likely to have resulted from the crustal thickening induced by regional-scale thrusting and folding in the footwall of the ophiolite nappes (Chew et al. 2010).
The structural base of the lower ophiolite sheet on Unst cuts discordantly across the tectonic boundary between the Valla Field and Saxa Vord blocks (Read 1934). Even if obduction and Barrovian metamorphism of the Dalradian rocks occurred during the same tectonic event the basal tectonic contact of the ophiolite cannot represent the original obduction thrust. However, this need not preclude the presence of parts of the metamorphic sole being preserved as isolated tectonic slices (Spray 1988). The metamorphic contrast between the c. 300-400 8C sub-biotite grade metasedimentary rocks associated with the Unst ophiolite and the P-T conditions recorded in its footwall suggest that at least c. 10 km of crustal section is missing. This estimate of missing crustal section is calculated based on pressure estimates by previous workers for the metamorphic sole of the ophiolite (3 kbar; Spray 1988). In addition, for lizardite stability within the ophiolite, it must have been at pressures below 4 kbar, indicating a pressure difference of at least 3.5 kbar (c. 10 km) between the ophiolite and sample FRQ-1 (Flinn 1985(Flinn , 2000(Flinn , 2001Spray & Dunning 1991). If published K-Ar ages of c. 470 Ma broadly record obduction at an early stage in the Grampian orogenic event (Spray 1988), then the present basal contact of the ophiolite is probably a younger tectonic break that developed during or after peak metamorphism.
To account for the metamorphic contrast across the basal (tectonic) contact of the ophiolite, this structure could be an outof-sequence thrust that cut obliquely across more steeply dipping isograds in its footwall. Alternatively, the structure could have developed with extensional geometry. The fast exhumation evidenced by garnet chemical zoning profiles in FRQ-20 could correspond to this extension. If this is the case then the ophiolite could have been obducted, then juxtaposed against 8-10 kbar footwall rocks within 20-30 Ma. Direct structural evidence that might allow us to discriminate between these possibilities is, however, lacking because the basal tectonic boundary and under-lying Saxa Vord lithologies were strongly reworked during the Silurian (Cannat 1989;Flinn & Oglethorpe 2005).