Lu–Hf and Sm–Nd dating of metamorphic garnet: evidence for multiple accretion events during the Caledonian orogeny in Scotland

Caledonian orogenesis in Scotland is currently interpreted in terms of a Mid-Ordovician arc–continent collision (Grampian event) followed by the Silurian collision of Laurentia with Baltica (Scandian event). Lu–Hf and Sm–Nd garnet ages of c. 475–460 Ma obtained from prograde garnets in metasedimentary successions and metabasic intrusions within the Northern Highland and Grampian terranes confirm that the Mid-Ordovician Grampian orogenic event was approximately synchronous in the two terranes. Lu–Hf and Sm–Nd ages of c. 450 Ma obtained from prograde garnets within the Moine Nappe of the Northern Highland terrane provide evidence for a hitherto unrecognized Late Ordovician regional metamorphic event. The existing two-stage Grampian–Scandian model for Caledonian orogenesis in northern Scotland is thus an oversimplification, and the new ages imply a more complex structural evolution. The restriction of the Late Ordovician and Silurian events to the Northern Highland terrane reinforces the suggestion that it was far removed from the Grampian terrane until juxtaposition following major end-Caledonian (Devonian) sinistral displacement along the Great Glen Fault. A similar record of Mid- and Late Ordovician metamorphic events within the Laurentian-derived Uppermost Allochthons of Norway has been attributed to episodic accretion significantly prior to Silurian continent–continent collision and closure of the Iapetus Ocean. Supplementary materials: Results of trace element analysis of the garnets by laser ablation inductively coupled plasma mass spectrometry are available at www.geolsoc.org.uk/SUP18583.

Within many collisional orogens, the final suturing of continental blocks is preceded by accretionary events that commonly occur during initial stages of oceanic closure. These events typically result from the collisions of magmatic arcs with continental margins and examples have been documented in the Himalayas (Fraser et al. 2001), the Urals (Puchkov 2009), the Appalachians (van Staal et al. 2009), and the Caledonides of Norway (Roberts 2003). Recognition of early accretionary events is relatively straightforward where the rocks affected are overlain unconformably by successor basins of known age. These successor basins will in turn be deformed during later stages of ocean closure and continental collision. At deeper structural levels within orogens the recognition of, and distinction between, major tectonic events is more problematic and depends critically upon the isotopic dating of metamorphic minerals that can be linked to pressure-temperature paths (e.g. Vance et al. 1998;Cutts et al. 2010). However, it is often difficult to establish unambiguous temporal relationships between deformation fabrics and accessory minerals such as monazite, titanite and zircon. An alternative approach is to use Lu-Hf and/or Sm-Nd isotopic systems to date garnet, which is a common porphyroblast phase that can generally be related to deformation fabrics.
Caledonian orogenesis in the North Atlantic region resulted from the sinistrally oblique collision of three continental blocks, Laurentia, Baltica and Avalonia, and closure of the Iapetus Ocean ( Fig. 1; Soper & Hutton 1984;Pickering et al. 1988;Soper et al. 1992;Dewey & Strachan 2003). The final amalgamation of continental blocks in Silurian-Devonian time was preceded by a series of Cambrian-Ordovician arc-continent collisions that occurred on the Iapetan margins of both Laurentia and Baltica (Dewey & Ryan 1990;Dewey & Mange 1999;van Staal et al. 1999;Roberts 2003). One such event is the early Ordovician Grampian, which caused intense deformation and metamorphism of the Laurentian Neoproterozic-Cambrian rocks of the Dalradian Supergroup (Grampian terrane; Dewey & Shackleton 1984;Dewey & Ryan 1990;Chew et al. 2010). In Ireland and Scotland (Fig. 1), this is thought to be due to collision of the Lough Nafooey arc and its likely northeastward extension beneath the Late Palaeozoic cover of the Midland Valley in Scotland, and associated ophiolite obduction. In western Ireland, a Silurian successor basin was deposited unconformably on the eroded remnants of this orogenic tract and was later deformed during the culminating collision with Avalonia (Dewey & Ryan 1990).
In contrast, the tectonometamorphic evolution of the early Neoproterozoic Moine Supergroup of the Northern Highland terrane in Scotland has proved less tractable. The timing of tectonothermal events is not defined by any intra-orogenic unconformities and is almost entirely dependent on the isotopic dating of metamorphic mineral assemblages and igneous intrusions of known structural age. The results indicate that the Moine Supergroup was affected by Neoproterozoic and Early Palaeozoic (Ordovician and Silurian) orogenic events (Vance et al. 1998;Strachan et al. 2002, 2010a, andreferences therein;Cutts et al. 2010). Ordovician metamorphism has been correlated with the 'Grampian' accretionary event (Kinny et al. 1999), and the Silurian event attributed to the 'Scandian' collision between Baltica and the segment of the Laurentian margin that contained the Northern Highland terrane (Dewey & Strachan 2003;Kinny et al. 2003). In view of the relative lack of high-quality, modern isotopic data across significant tracts of the Scottish Caledonides, and to test and refine existing tectonic models, we have applied Lu-Hf and Sm-Nd geochronology to metamorphic garnets. The new data indicate that the existing two-stage model for Caledonian orogenesis in Scotland is an oversimplification, thus necessitating significant revision of published regional tectonic models.

Lu-Hf and Sm-Nd dating of metamorphic garnet: evidence for multiple accretion events during the Caledonian orogeny in Scotland
Regional geological setting The Northern Highland terrane lies between the Moine Thrust and the Great Glen Fault (Fig. 1). The intervening Moine Supergroup comprises thick successions of psammites, semi-pelites and pelites (Holdsworth et al. 1994). Amphibolite bodies are locally common and represent deformed and metamorphosed mafic intrusions. Units of Archaean orthogneiss (Friend et al. 2008) are thought to represent tectonically emplaced inliers of the basement on which the Moine Supergroup sediments were deposited (Peach et al. 1907;Tanner et al. 1970;Rathbone & Harris 1979). Deposition probably occurred in the early Neoproterozoic after c. 1000 Ma, the age of the youngest detrital zircons, and before c. 870 Ma, the age of the oldest intrusive igneous rocks (Friend et al. 1997. The Moine Supergroup is formed of three lithostratigraphical units (from west to east): the Morar, Glenfinnan and Loch Eil groups ( Fig. 1; Holdsworth et al. 1994). On the mainland, the Morar and Glenfinnan groups are separated by the Sgurr Beag Thrust (Tanner et al. 1970;Rathbone & Harris 1979). A possible transition between the two groups is preserved on the Ross of Mull ( Fig. 1; Holdsworth et al. 1987). However, the terms 'Morar Group' and 'Moine Nappe' are effectively equivalent. The Glenfinnan Group is overlain stratigraphically by the Loch Eil Group, these two units forming the Sgurr Beag Nappe. How these units link north of the Dornoch Firth with the structurally analogous metasedimentary rocks of the Naver and Skinsdale nappes ( Fig. 1) is not well understood (Kocks et al. 2006). There is widespread evidence that the Moine Supergroup has been affected by polyphase deformation (e.g. Ramsay 1957;Brown et al. 1970;Tobisch et al. 1970;Powell 1974;Strachan 1985). Complex porphyroblast growth histories also imply multiple episodes of low to upper amphibolite-facies metamorphism (e.g. MacQueen & Powell 1977;Zeh & Miller 2001). A range of U-Pb and Sm-Nd mineral ages, some linked to prograde pressure-temperature histories, indicate that the Moine Supergroup has been affected by three orogenic events. The earliest isoclinal folds and mineral assemblages formed during 'Knoydartian' events at c. 825-800 Ma and c. 725 Ma (Rogers et al. 1998;Vance et al. 1998;Cutts et al. 2010). The most recent syntheses view these as the result of accretionary tectonics along the margin of the early to mid-Neoproterozoic Rodinia supercontinent (Cawood et al. 2004(Cawood et al. , 2010Kirkland et al. 2011). Early Palaeozoic metamorphic episodes are assigned to the Caledonian orogeny. Ordovician 'Grampian' folding and metamorphism at c. 470-460 Ma has been demonstrated in the eastern part of the Sgurr Beag Nappe Cutts et al. 2010) and the Naver Nappe (Kinny et al. 1999). Regional NW-directed ductile thrusting, nappe stacking, and upright folding is thought to have occurred during the Silurian 'Scandian' event at c. 435-425 Ma Strachan & Evans 2008;Krabbendam et al. 2011). Scandian deformation culminated in the development of the Moine Thrust Zone that defines the western margin of the Scottish Caledonides ( Fig. 1; Goodenough et al. 2011).
East of the Great Glen Fault, the metasedimentary rocks of the Dava and Glen Banchor successions ( Fig. 1) are lithologically similar to the Moine Supergroup and have yielded evidence of highgrade ?Knoydartian metamorphism at c. 840 Ma (Piasecki 1980;Highton et al. 1999). Deposition of the younger Dalradian Supergroup (Fig. 1) from c. 750 Ma onwards was probably initiated by continental rifting that culminated in the development of the Iapetus Ocean (Anderton 1985;Leslie et al. 2008). Dalradian stratigraphy reflects multiple periods of lithospheric stretching, rifting and thermal subsidence, and sedimentation continued into the late Early Cambrian. Regional polyphase deformation and greenschistto amphibolite-facies metamorphism of the Dalradian Supergroup in Scotland and Ireland has been assigned entirely to the early Ordovician 'Grampian' event (e.g. Friedrich et al. 1999;Flowerdew et al. 2000;Oliver et al. 2000;Baxter et al. 2002;Chew et al. 2003). The absence of any evidence for Silurian deformation and metamorphism suggests that the Grampian terrane was located well away from the Laurentia-Baltica 'Scandian' collision that resulted in further reworking of the Northern Highland terrane. The two terranes were probably juxtaposed between c. 425 Ma and c. 390 Ma as a result of major sinistral strike-slip displacement along the Great Glen Fault (Dewey & Strachan 2003;Kinny et al. 2003).

Sample descriptions
Isotopic data from nine samples from the Moine Supergroup are reported here, which cover a wide geographical spread and include material from all major thrust nappes. Two samples were also collected east of the Great Glen Fault, from the Dava Succession and the Dalradian Supergroup (Fig. 1). Brief petrographic descriptions of the samples are provided below.
Moine Nappe, Northern Highland Terrane AB07-08: Morar Group Basal Pelite, Sandaig, Glenelg [NG 7689 1491]. The Basal Pelite occurs between Lewisianoid basement and the Lower Psammite of the Morar Group (Ramsay & Spring 1962). The sample is a biotite-muscovite-garnet-plagioclasequartz schist with accessory titanite, opaque minerals, apatite and zircon (Fig. 2a). Garnets are up to 10 mm in diameter and contain curved inclusion trails of fine-grained quartz that are continuous with the external mica schistosity. Quartz pressure tails are developed around some garnets. The cores of garnets are often extremely rich in inclusions of zircon, titanite, quartz and plagioclase. The margins of some garnets are retrogressed to chlorite. The dominant mica fabric is crenulated. Armadale,Skye [NG 6409 0353]. The Armadale Pelite sampled lies close to Lewisianoid basement and only a few hundred metres east above the Moine Thrust. The sample is a fine-grained, plagioclase-muscovite-biotite-garnetchlorite schist with accessory titanite (Fig. 2b). An intense, mylonitic fabric is defined by alternation of quartz-plagioclase and micaceous layers on a scale of 1-2 mm, and is tightly crenulated. The mylonite fabric wraps euhedral to subhedral garnets that are up to 10 mm in diameter and contain well-developed curved inclusion trails of quartz, biotite, zircon and titanite. However, it is not generally possible to demonstrate continuity between these inclusion trails and the external mylonitic fabric. The garnets are commonly fringed by quartz pressure tails.

AB07
AB07-15: Morar Group pelite, Talmine, west Sutherland [NC 5735 6324]. The Talmine Pelite is a discontinuous lens, c. 300 m long and a few tens of metres thick, within Morar Group psammites ( Fig. 1; British Geological Survey 1997). The sample is a biotite-muscovite-garnet-albite-quartz schist (Fig. 2c). Garnets are up to 10 mm in diameter and wrapped by a biotite-muscovite fabric. In thin section, garnets are generally euhedral to subhedral with sharp margins. There is no optical evidence for internal zoning. Inclusion trails within the garnets consist of fine-grained quartz, opaque minerals and biotite. The inclusion trails are curved and oblique to the external mica fabric. Some garnets contain small cracks that are filled with muscovite. The rims of many garnets are retrogressed to chlorite. Talmine,west Sutherland [NC 5535 5310]. The sample was obtained from part of the Ben Hope Sill, a pre-tectonic amphibolite intrusion, up to 500 m thick, within the Morar Group (Holdsworth 1989), 10 km away from AB07-15 (Fig. 1). The sample is dominated by hornblende-garnet-plagioclase with accessory biotite, muscovite, quartz and opaque minerals, and secondary chlorite (Fig. 2d). Euhedral to subhedral garnets are up to 20 mm in diameter and wrapped by a hornblende-biotite fabric. In thin section, garnets do not show any optical evidence for zoning. Prominent linear to curved inclusion trails are highly oblique to the enveloping fabric and consist of quartz, opaque minerals, biotite, chlorite and amphibole. Some garnets are extensively retrogressed to chlorite. [NM 3993 1880]. The Moine rocks of the Ross of Mull have been correlated with the upper Morar Group and the Glenfinnan Group (Holdsworth et al. 1987). However, it is uncertain whether these are part of the Moine Nappe or the Sgurr Beag Nappe; for simplicity they are here included in the former. The sample is from a 3 m thick sheet that cuts gently across lithological layering in host Glenfinnan Group psammites. The dominant mineral assemblage is hornblende-garnet-biotite-epidote-plagioclase-quartz with accessory rutile, zircon, opaque minerals, titanite and apatite (Fig. 3e). Subhedral garnets are 3-10 mm in diameter and rich in inclusions of quartz, plagioclase, rutile and apatite. The garnets are wrapped by a weak hornblende-biotite-plagioclase-quartz fabric.

AB07-30: metabasic intrusion, Ross of Mull
Sgurr Beag and Naver nappes, Northern Highland Terrane Loch Quoich,. The sample was obtained from one of a series of elongate pods, each several metres long. These probably resulted from the boudinage of formerly more continuous metabasic sheets. Contacts with host Glenfinnan Group gneisses are sharp and concordant. The sample comprises hornblende-biotite-garnet-plagioclase with accessory titanite, zircon, opaque minerals and apatite (Fig. 3a). Garnets are typically anhedral, 2-3 mm in diameter, and often surrounded by reaction rims of plagioclase and biotite. The garnets are wrapped by a weak planar fabric defined by aligned hornblende and biotite. There is no optical evidence for internal zoning of garnet; the interiors of grains are typically dusty with inclusions of titanite and quartz. Druimnadrochit [NH 4860 3255]. The sample was collected from a concordant amphibolite sheet, c. 10 m thick, within migmatitic gneisses assigned to the Glenfinnan Group. The sample is dominated by a granoblastic assemblage of garnet + hornblende (green-brown) + epidote + quartz + plagioclase with accessory opaque minerals and apatite (Fig. 3b). Euhedral garnets are up to 3 mm in diameter and contain many inclusions of quartz, hornblende, chlorite, feldspar, apatite and opaque minerals. Garnets do not show any optical evidence for zoning. [NC 7982 6590]. The sample was obtained from one of a series of late-tectonic intrusions within the meta-igneous Strathy Complex that may represent late Mesoproterozoic basement to the Moine Supergroup in north Sutherland (Burns et al. 2004). These intrusions are discrete bodies, up to 3 m thick, and are generally only weakly deformed. The sample comprises a largely granoblastic assemblage of garnet-gedrite-plagioclase-quartz with accessory opaque minerals, biotite and chlorite (Fig. 3c). Garnets are exceptionally coarse, up to 80 mm in diameter, and contain prominent  Kretz (1983). curved to straight ribbons that mainly comprise quartz together with gedrite, feldspar, opaque minerals and biotite. The sample is overall weakly foliated; the quartz ribbons identified within the garnets cannot be traced into the host matrix.

AB07-22: Moine psammitic migmatite, Kirtomy Point, north
Sutherland [NC 7511 6430]. The sample is a medium-grained, garnetiferous gneiss composed of quartz-plagioclase-garnetbiotite. The plagioclase has multiple twinning and often shows signs of recystallization. Minor amphibole and accessory titanite, apatite and zircon are also present (Fig. 3d). The garnets are c. 1 mm in diameter and typically subhedral to anhedral in form. They are often characterized by a reaction rim of fine-grained plagioclase. The garnets are relatively free of inclusions and cracks, and are wrapped by the dominant planar fabric defined by aligned biotite.
Grampian Terrane AB07-23: metabasic body, Loch Ruthven [NH 6283 2889]. The sample was obtained from a garnet amphibolite body that occurs within the Dava Succession, although contacts between the amphibolite and host metasedimentary rocks are unexposed. The sample is dominated by a granoblastic assemblage of garnetquartz-amphibole-epidote with accessory rutile, titanite and opaque minerals (Fig. 4a). Garnet is exceptionally coarse with some grains up to 80 mm in diameter. The garnets are typically euhedral to subhedral, often cracked and veined, and contain abundant inclusions of amphibole, quartz, titanite, rutile, epidote and opaque minerals. Some garnets show evidence for partial retrogression to chlorite. [NN 2611 8032]. The sample is a biotite-garnet-muscovite-quartz schist with accessory apatite, zircon and opaque minerals. The sequence of fabric development has been analysed by Phillips & Key (1992). An early (S 1 ) fine-grained (0.1 mm) quartz-mica fabric is weakly folded to form a crenulation fabric (S 2 ). Euhedral garnets, up to 5 mm in diameter, contain curved inclusion trails of quartz + mica and opaque minerals, which are continuous with the crenulation fabric in the matrix (Fig. 4b). The garnets therefore grew syn-to post-S 2 . Biotite poikiloblasts, rimmed by chlorite, are also present but were deformed during the development of S 2 and hence are older than the garnets.

Analytical techniques
All analyses were undertaken at Royal Holloway, University of London (RHUL). X-ray fluorescence (XRF) analysis of wholerock powders was used to establish concentrations of Nd, Y and Zr, which were used to estimate concentrations of Sm, Lu and Hf, to optimize the amount of spike added. The remaining crushed material was split into grain-size fractions of 0.25-0.5 mm and 0.5-1 mm by sieving, then the garnets were separated using a Frantz Isodynamic magnetic separator. The cleanest and most inclusionfree garnet fragments were then hand-picked under a binocular microscope. In some samples, it was possible to hand-pick separate core and rim material as these differed in colour; this was identified in thin section prior to picking.
Before commencing Lu-Hf and Sm-Nd analysis the garnets were subject to trace element analyses by laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS). The system used is a RESOlution M-50 Excimer 193 nm laser ablation system with a two volume laser ablation cell that is coupled to an Agilent 7500ce ICP-MS system (Müller et al. 2009). For initial analyses the garnets were mounted in epoxy and polished. For later analyses the garnets were analysed in situ in a thick thin section. For all, line traverses were carried out across the mineral grains. SiO 2 contents obtained by electron microprobe (Natural History Museum, London) were initially used as the internal standard, but as these are constant at 37.7 ± 0.2%, this value was used subsequently. External standardization was provided by two runs of the NIST SRM-612 glass standard at the beginning and end of each run. The spot size for data acquisition was 44 µm, the repetition rate was 15 Hz, the scan speed was 0.5 mm min −1 and the attenuator was taken out to give a stronger beam.
For Lu-Hf and Sm-Nd analyses, the procedures for sample leaching, spiking and dissolution followed the guidelines described by Anczkiewicz & Thirlwall (2003). Lu-Hf and Sm-Nd analyses were performed on a single total dissolution. The samples were first passed through AG50W-X8 cation resin to separate high field strength element (HFSE), light rare earth element (LREE) and heavy REE (HREE) fractions. The HFSE fraction required a second pass through these columns to minimize the HREE that may be in the fraction. The fractions were each passed through Eichrom LN resin to separate Hf, Sm and Nd, and Lu. Total procedure blanks were typically 24 pg for Hf and 23 pg for Nd. The lowest Hf mass analysed was 3.04 ng from sample AB07-17 Grt 1; a blank correction to this has no significant effect on the age obtained. This is also true for the sample with the lowest Nd mass (0.19 µg from AB07-08 Grt Core).
The analyses used the GV IsoProbe MC ICP-MS system at RHUL, using procedures of Thirlwall & Anczkiewicz (2004), except that static mode was used. Blank solutions analysed before each sample to provide on-peak zeroes have <0.07 mV 142 Nd and 0.08 mV 180 Hf respectively, less than 10 −3 times typical sample intensities. Drift commonly observed in static ratio analysis required frequent analysis of JMC475 Hf and Aldrich Nd standards. During the course of this study, JMC 475 yielded average static 176 Hf/ 177 Hf and 180 Hf/ 177 Hf of 0.282184 ± 0.000024 and 1.886724 ± 0.000162, 2sd, n = 48. Aldrich Nd and Aldrich mixed Nd-Ce solutions yielded 142 Nd/ 144 Nd and 143 Nd/ 144 Nd of 1.141397 ± 0.000276 and 0.511408 ± 0.000015, 2sd, n = 59, the latter after slope correction using the method of Thirlwall & Anczkiewicz (2004). Standard reproducibility on single days was significantly better than the external precisions quoted above on static 176 Hf/ 177 Hf and 143 Nd/ 144 Nd, and these values also include several days when there was temporal drift in the ratio measured (e.g. Thirlwall & Anczkiewicz 2004). In most cases, the samples used to calculate each age were analysed consecutively on a single day, minimizing uncertainty resulting from this drift, and a correction was made to the sample ratios based on the observed drift. The contribution of standard uncertainty to the ages calculated is thus much smaller (c. ±0.000010 on both Hf and Nd ratios) than indicated by the long-term external precision. Isochron ages and uncertainties were calculated using IsoPlot version 2.47 (Ludwig 2003) and a York II regression program (York 1966), with uncertainties multiplied by √MSWD for MSWD >1, using decay constants of 1.865 × 10 −11 a −1 for 176 Lu (Scherer et al. 2001) and 6.54 × 10 −12 a −1 for 147 Sm.

Results and interpretation
Total dissolution of garnet incorporates inclusions, which can affect Sm-Nd and Lu-Hf analyses (e.g. Prince et al. 2000;Scherer et al. 2000;Anczkiewicz & Thirlwall 2003). A benefit of determining both Sm-Nd and Lu-Hf ages is that they are affected by different inclusions: LREE-enriched inclusions such as monazite may affect Sm-Nd ages, whereas zircon inclusions have the potential to influence Lu-Hf ages. The majority of the zircons within garnets and pelite whole-rocks are likely to be detrital and thus substantially older than the time of garnet growth. If a garnet incorporated a similar detrital zircon population to that present in the host rock, this should not affect its calculated age. If a garnet incorporated disproportionally more ancient zircon, measured ages could be younger than true ages. Establishing the Hf ratios of the zircon inclusions within the garnets was not a part of this study, thus the effect of zircon inclusions on the ages achieved cannot be accurately modelled. However, comparison between Sm-Nd and Lu-Hf ages on the same samples allows the influence from inclusions to be estimated.
This study includes both two-point garnet-whole-rock ages and three-or four-point isochrons. Three-or four-point isochrons are more statistically meaningful than two-point garnet-whole-rock ages, as an MSWD can be derived. However, in a polymetamorphic terrain the potential exists for multiple garnet growth stages, and it may prove impossible to obtain multiple garnet separates that crystallized at a single age. The two-point garnet-whole-rock ages are statistically less reliable but most of the ages obtained correspond closely to major metamorphic events defined either by the three-or four-point isochrons, or by previously published datasets, thus demonstrating the utility of this reconnaissance approach. The results are summarized in Figures 5 and 6, and Tables 1 and 2.

Moine Nappe, Northern Highland Terrane
Garnet cores within the Morar Group Basal Pelite (AB07-08) gave a Lu-Hf two-point garnet-whole-rock age of 589 ± 17 Ma and rims yielded a Lu-Hf two-point garnet-whole-rock age of 458.7 ± 4.5 Ma (Table 1). These ages do not lie on a three-point isochron (Fig. 5a). The core and the rim were distinguished during picking based on colour, as the cores were orange and far more inclusion-rich than the purple garnet rims. Material from the core has a low 176 Lu/ 177 Hf ratio, owing to a high Hf content of 2.49 ppm and a particularly low Lu content of 1.95 ppm. The concentrations from solution analysis by isotope dilution (ID) can be compared with data from the LA-ICP-MS to assess the potential influence of inclusions. From the latter, the Hf concentration of the pure garnet was estimated to be c. 0.2 ppm based on parts of the LA-ICP-MS profile that were not affected by zircon inclusions. This is much lower than the ID concentration, implying that this is strongly influenced by zircon inclusions. In view of this and the low 176 Lu/ 177 Hf and 176 Hf/ 177 Hf ratios for the core, the 589 Ma age should be viewed with caution. The in situ Hf concentration from part of the pure garnet rim was c. 0.4 ppm; ID gave 1.86 ppm, implying that some zircon inclusions were also dissolved with the garnet rim, but as the 176 Lu/ 177 Hf and 176 Hf/ 177 Hf ratios are higher for the rim this should not affect the age greatly. The inclusion-rich nature of this sample is documented in Figure 7c.
The Morar Group Basal Pelite (AB07-08) gives a Sm-Nd twopoint age of 449.7 ± 2.3 Ma using the garnet core and the whole-rock fraction (Table 2). This is 9 ± 5 Ma younger than the rim Lu-Hf age.
The garnet has a high 147 Sm/ 144 Nd ratio, suggesting that the age is likely to be reliable. This is supported by a low ID Nd concentration of 0.234 ppm, which is similar to a pure garnet concentration (by LA-ICP-MS) of c. 0.3 ppm. The garnet rim does not yield an age, as its 143 Nd/ 144 Nd ratio is identical to that of the whole-rock (Fig. 5b). The garnet rim fraction analysed appears to have incorporated monazite inclusions, as it has an ID Nd concentration of 15.5 ppm whereas the pure garnet rim analysed by LA-ICP-MS gave a Nd concentration of c. 0.5 ppm. The monazite inclusions are very small and the garnet may not have been crushed finely enough to ensure full leaching.
The Armadale Pelite (AB07-11) gives a Lu-Hf two-point garnetwhole-rock age of 466.2 ± 2.9 Ma. Pure garnet Hf concentration is Lu-Hf and Sm-Nd isochron diagrams for samples from the Moine Nappe that have three or more points. Grt 1, garnet digestion 1; Grt 2, garnet digestion 2; Grt 3, garnet digestion 3; WR, whole-rock. Where error bars are not visible they are smaller than the symbols. estimated as c. 0.2 ppm by LA-ICP-MS, substantially less than the ID Hf concentration of 1.70 ppm indicating some zircon incorporation into the dissolved garnet fraction. However, as the 176 Lu/ 177 Hf and 176 Hf/ 177 Hf are high this age is likely to be reliable. This sample did not give a reliable Sm-Nd age (628 ± 260 Ma) as the ID Nd concentration of 19.0 ppm is much greater than the c. 0.4 ppm of pure garnet by LA-ICP-MS, suggesting substantial incorporation of Nd-rich inclusions.
Two garnet fractions were analysed for Lu-Hf from the Talmine Pelite (AB07-15) and yield two-point ages within error of 449.3 ± 2.7 Ma and 443.1 ± 5.3 Ma respectively. Calculation of a three-point age is compromised by a high MSWD of 8.3 using the York (1966) regression program, or 8.5 using Isoplot. However, the uncertainties derived using the two regression routines are very different, at 448 ± 7 Ma and 449 ± 55 Ma respectively. In view of the agreement of two-point ages within error, it would appear that the ±7 Ma uncertainty is a more plausible estimate of the uncertainty on this three-point age (Fig. 5c). Grt 1 has a higher 176 Lu/ 177 Hf ratio, owing to a much higher Lu concentration of 8.8 ppm compared with 6.0 ppm. The LA-ICP-MS gave a pure garnet Hf concentration of c. 0.3 ppm whereas ID gave a Hf concentration of 2.1 ppm for both garnet separates, which suggests that some zircon was dissolved.
A Sm-Nd four-point isochron has been obtained for the Talmine Pelite sample (AB07-15), using three garnet separates and the whole-rock, which gives an age of 456 ± 7 Ma (MSWD = 0.66), shown in Figure 5d. The LA-ICP-MS Nd concentration was c. 1.0 ppm, which is marginally more than that given by Nd ID data (0.701 ppm, 0.424 ppm and 0.451 ppm), indicating that there was no influence of Nd-rich inclusions on the age obtained from this sample. This age is within error of the Lu-Hf ages, implying that the zircon inclusions sampled in the Lu-Hf dissolution have had no significant effect on the measured age.
A Lu-Hf four-point isochron has been obtained for the Ben Hope Sill sample (AB07-17), using three garnet fractions and a whole-rock, and gives an age of 447.3 ± 1.7 Ma (MSWD = 0.92) (Fig. 5e). The LA-ICP-MS gave a Hf concentration of c. 0.1 ppm for pure garnet, whereas ID gave 0.133 for Grt 1, 0.094 ppm for Grt 2 and 0.246 ppm for Grt 3, which suggests little or no input from zircon inclusions. No useful Sm-Nd age was obtained from the Ben Hope Sill (AB07-17) as the garnets had similar Sm/Nd ratios to the whole-rock (Fig. 5f). The LA-ICP-MS gave a pure garnet Nd concentration of c. 0.2 ppm, whereas ID gave values ranging from 9.2 to 18.8 ppm, showing that many Nd-rich inclusions were dissolved with the garnet fraction. This could be because the sample was not crushed finely enough or more probably because minerals present such as epidote and titanite would not be affected by the extra leaching step. This is probably the case for many of the amphibolites, which did not produce a Sm-Nd age.
The metabasic intrusion from the Ross of Mull (AB07-30) yields a Lu-Hf three-point isochron that gives an age of 448.7 ± 5.0 Ma Fig. 6. Lu-Hf and Sm-Nd isochron diagrams for samples from the Sgurr Beag and Naver nappes that have three points or more. Grt 1, garnet digestion 1; Grt 2, garnet digestion 2; WR, whole-rock. Where error bars are not visible, they are smaller than the symbols.
(MSWD = 0.6, Fig. 5g). The ID garnet fractions have concentrations of 1.56 ppm and 1.49 ppm, higher than the c. 0.2 ppm from LA-ICP-MS, suggesting that there may be some influence from inclusions. This sample did not give a meaningful Sm-Nd age, as the 12-16 ppm Nd concentrations are much higher than the LA-ICP-MS c. 0.6 ppm pure garnet value, indicating a large influence from inclusions (Fig. 5h). Although there was a significant difference between the garnet fractions and the whole-rock 147 Sm/ 144 Nd ratios, the 143 Nd/ 144 Nd values for the garnet fractions were either almost within error or less than the whole-rock ratio.

Sgurr Beag and Naver nappes, Northern Highland Terrane
The metabasic intrusion at Loch Quoich (AB07-05) gave a Lu-Hf two-point garnet-whole-rock age of 462.9 ± 1.7 Ma. The LA-ICP-MS data gave a Hf concentration of c. 0.2 ppm, which is similar to the ID value of 0.237 ppm, indicating no influence from zircon inclusions. This sample did not yield a meaningful Sm-Nd age, as the 147 Sm/ 144 Nd and 143 Nd/ 144 Nd ratios from the garnet are lower than that of the whole-rock. The garnet fraction had an ID Nd concentration of 7.0 ppm, whereas the LA-ICP-MS data gave a concentration of c. 0.5 ppm, suggesting that mixing with inclusions was the cause of the low 147 Sm/ 144 Nd and 143 Nd/ 144 Nd ratios.
The metabasic intrusion at Drumnadrochit (AB07-14) gave a Lu-Hf three-point isochron age of 474.8 ± 1.2 Ma with an MSWD of 0.9 (Fig. 6a). The garnet fractions have ID Hf concentrations of 0.464 ppm and 0.491 ppm, which are only a little higher than the value obtained from LA-ICP-MS of c. 0.2 ppm. This suggests that zircon inclusions have had little effect on the Lu-Hf age. The sample gave a Sm-Nd three-point isochron age of 470.9 ± 2.7 Ma (MSWD = 3, Fig. 6b), within error of the Lu-Hf age. The ID Nd concentrations of 0.555 ppm and 0.821 ppm are similar to the c. 0.7 ppm obtained by LA-ICP-MS, implying that this age has not been affected by Nd-rich inclusions. A three-point Lu-Hf isochron was obtained from the metabasic intrusion within the Strathy Complex (AB07-21) using two garnet fractions and a whole-rock fraction, giving 447 ± 15 Ma (MSWD = 1.4, Fig. 6c). Garnets from this sample had unusually low Lu ID concentrations of 0.59 and 0.62 ppm, which are similar to the LA-ICP-MS value of c. 0.9 ppm. This results in low 176 Lu/ 177 Hf ratios, accounting for the large age uncertainty. However, ID Hf contents (0.71-0.75 ppm) are similar to the 0.9 ppm estimated for pure garnet by LA-ICP-MS, suggesting little effect from zircon inclusions. This sample gave a threepoint Sm-Nd isochron age of 432.8 ± 2.5 Ma, MSWD of 1.9 (Fig. 6d), within error of the Lu-Hf age. The Nd concentrations from the garnet fractions are fairly low (1.223 ppm and 0.845 ppm) and very similar to the values from the LA-ICP-MS of c. 1.0 ppm, suggesting that this age has not been influenced by inclusions.
Two separate Lu-Hf two-point garnet-whole-rock ages were obtained for the Moine migmatite at Kirtomy Point (AB07-22) of 466.0 ± 2.1 Ma (AB07-22 Grt 1) and 453.9 ± 4.1 Ma (AB07-22 Grt 2), which are 7.9 Ma outside of error, and do not yield a threepoint isochron (Fig. 6e). The pure garnet has a Hf concentration of c. 0.2 ppm (LA-ICP-MS), whereas ID Hf on the garnets yields 3.1 and 3.9 ppm. AB07-22 gave a Sm-Nd age of 531 ± 7 Ma from Grt 2 (all three data points are shown in Fig. 6f). The pure garnet from this sample gave a Nd concentration by LA-ICP-MS of c. 0.5 ppm, which is lower than the Nd concentration obtained by ID from Grt 2 (4.1 ppm), suggesting that some Nd-rich inclusions were incorporated in the dissolved garnet fraction. Grt 1 from this sample gave a much higher Nd ID concentration of 24.9 ppm, which has clearly been affected by inclusions, giving a lower 147 Sm/ 144 Nd ratio than the whole-rock, and an apparent age of 1196 ± 180 Ma.

The Grampian Terrane
A Lu-Hf two-point garnet-whole-rock age of 463.2 ± 1.7 Ma was obtained from the metabasic intrusion within the Dava Succession (AB07-23). LA-ICP-MS analysis gave a Hf concentration of c. 0.8 ppm, which is substantially higher than the ID Hf value of 0.130 ppm, implying no effect from zircon inclusions. Sm-Nd from this sample gave a two-point garnet-whole-rock age of 431.2 ± 6.5 Ma. The LA-ICP-MS gave a pure garnet Nd concentration of c. 0.2 ppm, whereas ID Nd was 0.85 ppm, suggesting some influence from inclusions. The Leven Schist (AB07-27) gave a Lu-Hf two-point age of 470.8 ± 3.2 Ma. The pure garnet from this sample gave a Hf concentration of c. 0.5 ppm, whereas the ID concentration was 2.1 ppm, suggesting that the age may have been affected by inclusions. This sample did not yield a meaningful Sm-Nd age as the ID Nd concentration of 20.8 ppm is substantially higher than the LA-ICP-MS concentration of c. 0.2 ppm, implying a large effect from Nd-rich inclusions.

Discussion
Published pressure-temperature data for Caledonian metamorphism of the Moine and Dalradian supergroups give peak temperatures mainly in the region 600-700 °C (e.g. Friend et al. 2000;Baxter et al. 2002;Cutts et al. 2010). A sample collected near to AB07-14 has had detailed pressure and temperature modelling undertaken on it, based on preserved garnet compositions, which gives 7 kbar and 650 °C associated with Caledonian metamorphism dated at 463 ± 4 Ma . These temperatures do not exceed estimated closure temperatures for Nd diffusion in garnet, estimated at 700-750 °C (e.g. Ganguly et al. 1998). A similar or higher temperature range has been estimated for the isotopic closure of Lu-Hf systems (Scherer et al. 2001). Accordingly, the Sm-Nd and Lu-Hf ages obtained in this study are most easily interpreted as approximately dating garnet growth during prograde metamorphism. The LA-ICP-MS data can provide more information on whether the garnets are reflecting prograde growth or cooling, as samples with Lu enrichment towards the garnet cores (e.g. Fig. 7a) are more likely to provide ages that reflect garnet growth (Lapen et al. 2003;Skora et al. 2008). Rayleigh distillation describes the partitioning of a element between two reservoirs during metamorphism. If an element is very compatible within a mineral, the Rayleigh distribution will show a high concentration of the element in the core with a decrease towards the rim, whereas incompatible elements (e.g. Hf or Nd in garnet) will show the opposite trend. Such a distribution is commonly accepted as reflecting prograde garnet growth conditions (Skora et al. 2006;Dutch & Hand 2009). Core Lu enrichment is seen in the metabasic intrusions at Loch Quoich (AB07-05), Drumnadrochit (AB07-14) and on the Ross of Mull (AB07-30), as well as the Armadale Pelite (AB07-11), the Talmine Pelite (AB07-15), the Moine migmatite at Kirtomy Point (AB07-22) and the Leven Schist (AB07-27). Samples that show no Rayleigh zoning are the Morar Group Basal Pelite (AB07-08), the Ben Hope Sill (AB07-17), the Strathy Complex (AB07-21) and the Dava Succession (AB07-23). An example with no Lu enrichment preserved is shown from the Strathy Complex (AB07-21) in Figure 7b.
The ages obtained in this study are summarized in Table 3. Five of the samples provided both Lu-Hf and Sm-Nd ages (AB07-08, AB07-15, AB07-14, AB07-21 and AB07-23). Two samples give Sm-Nd ages that are substantially younger than the Lu-Hf ages: the Strathy Complex (AB07-21) and the Dava Succession (AB07-23); both have exceptionally coarse garnets of up to 80 mm contained within high-grade host rocks (sillimanite grade; Johnstone et al. 1969;Winchester 1974;Tanner 1976;Powell et al. 1981) that are relatively low in the local structural successions. Neither sample shows Lu enrichment towards the garnet cores, which suggests that these ages reflect cooling, and the difference in ages is likely to be related to the difference in closure temperature between the two isotopic systems (e.g. Anczkiewicz et al. 2007).

Timing of Mid-Ordovician metamorphism in the Northern Highland and Grampian terranes
The ages of 474.8 ± 1.2 Ma (Lu-Hf, three-point isochron) and 470.9 ± 2.7 Ma (Sm-Nd, three-point isochron) obtained from a metabasic intrusion near Druimnadrochit (AB07-14) (Fig. 8) are interpreted as dating peak metamorphism of the Sgurr Beag Nappe during the Grampian orogenic event. These ages are similar to, or slightly older than, previously published data relating to the timing of this event in the Moine Supergroup. The outer rims of garnets within nearby Moine gneisses in this area contain monazites that have yielded a LA-ICP-MS age of 464 ± 3 Ma ). This age is near identical to a U-Pb (secondary ion mass spectrometry; SIMS) zircon age of 463 ± 4 Ma obtained from a synkinematic pegmatite within the eastern Moine rocks of Glen Urquhart . A few kilometres further south, titanites from the Fort Augustus granite gneiss have yielded a U-Pb (thermal ionization mass spectrometry) age of 470 ± 2 Ma . The two-point garnet-whole-rock Lu-Hf age of 462.9 ± 1.8 Ma obtained from a metabasic intrusion at Loch Quoich (AB07-05) (Fig. 8) is within error of the monazite and zircon ages reported from Glen Urquhart. Similarly, the two-point garnet-whole-rock Lu-Hf age of 466.0 ± 2.1 Ma obtained from one garnet fraction from a migmatite at Kirtomy Point (AB07-22) (Fig. 8) falls within the range of ages provided by U-Pb (SIMS) dating of zircon from these migmatites (467 ± 10 Ma and 461 ± 13 Ma; Kinny et al. 1999). The new data reported here are therefore consistent with published data that indicate a widespread Mid-Ordovician (Grampian) metamorphic event in the Sgurr Beag and Naver nappes.
The two-point garnet-whole-rock Lu-Hf age of 466.2 ± 2.9 Ma obtained from the basal Morar Group at Armadale (AB07-11) (Fig. 8) is potentially significant in that it implies that the Moine Nappe was also affected by the Grampian orogenic event. The only previously published isotopic result that might correspond to an event of this age within the Moine Nappe is an imprecise Rb-Sr whole-rock isochron of 467 ± 20 Ma obtained from the Knoydart Pelite (Brewer et al. 1979). Further data are required to substantiate these indications that the Grampian orogenic event extended into the footwalls of the Sgurr Beag and Naver thrusts.
The two-point garnet-whole-rock Lu-Hf age of 470.8 ± 3.2 Ma obtained from the Leven Schist (AB07-27) (Fig. 8) falls within the broad range of published ages for the Grampian orogenic event east of the Great Glen Fault. Within the Dalradian rocks of NE Scotland, the Grampian event is constrained by a U-Pb zircon age of 472 ± 3 Ma for the syn-D 2 Portsoy gabbro (Oliver et al. 2000) and a U-Pb monazite age of 470 ± 1 Ma for the syn-to post-tectonic Aberdeen granite (Kneller & Aftalion 1987). Both are within error of a Sm-Nd age of 472.9 ± 2.9 Ma obtained from syntectonic garnets in Glen Clova (Baxter et al. 2002). The Grampian orogenic event in western Ireland also occurred at c. 470 Ma, as indicated by U-Pb zircon ages obtained from the syntectonic Currywongaun (475 ± 1 Ma) and Cashel (470 ± 1 Ma) gabbros in Connemara (Friedrich et al. 1999). The close agreement between our Leven Schist age and other Grampian terrane ages, despite the high Hf content of the dissolved garnet fraction, strongly suggests that incorporation of a substantial number of zircon inclusions into the dissolved garnet has little effect on the accuracy of the Lu-Hf age. The age of 463.2 ± 1.7 Ma from a metabasic intrusion within the Dava Succession (AB07-23) (Fig. 8) is younger than the published ages for the Grampian, which suggests that this age may reflect cooling. This is supported by the large size of the garnets (up to 80 mm), the high-grade nature of this sample and the lack of a Rayleigh profile for Lu in this garnet.
In summary, the new data reported here are entirely consistent with published isotopic data and indicate approximately synchronous Mid-Ordovician metamorphism in the Northern Highland and Grampian terranes.

Implications of Late Ordovician garnet growth within the Northern Highland terrane
Evidence for a hitherto unrecognized Late Ordovician metamorphic event at c. 450 Ma is provided by: (1) Lu-Hf isochrons obtained from the Ben Hope Sill (AB07-17) (447.3 ± 1.7 Ma, MSWD = 0.92) and a metabasic intrusion on the Ross of Mull (AB07-30) (448.7 ± 5.0 Ma, MSWD = 0.6); (2) two separate garnet-whole-rock Lu-Hf ages (449.3 ± 1.9 Ma and 443.1 ± 4.8 Ma) obtained from a Morar Group pelite at Talmine (AB07-15) and a Sm-Nd isochron (456 ± 7 Ma, MSWD = 0.68) for this sample that is within error; (3) a garnet-whole-rock Sm-Nd age obtained from the Morar Group Basal Pelite at Glenelg (AB07-08) (449.7 ± 2.3 Ma) (Fig. 8). The significance of the Lu-Hf isochron of 447 ± 15 Ma obtained from a metabasic intrusion within the Strathy Complex (AB07-21) (Fig. 8) is uncertain, as it could be recording metamorphism in any of the Caledonian events under discussion. Late Ordovician metamorphism at 450 Ma postdates peak Grampian metamorphism in the Sgurr Beag and Naver nappes by at least 14 Ma, and predates Scandian nappe stacking and associated deformation by at least 15 Ma. Discussion centres on the likely geotectonic significance of this metamorphic event and its structural expression within the Moine Supergroup.
It might be argued that the Late Ordovician ages are the result of a prolonged Grampian event, or cooling from the Grampian event. However, all available evidence indicates that the Grampian event was relatively short-lived (<7-8 Ma?) and followed rapidly by a change from ocean-ward to continent-directed subduction and development during the middle Ordovician of an accretionary prism in the Southern Upland terrane ( Fig. 1; Dewey & Ryan 1990;Soper et al. 1999). General considerations also support the notion that arccontinent collisions tend to be relatively short-lived events (Dewey 2006). Accordingly, it seems improbable that these Late Ordovician ages can be attributed to the Grampian arc-continent collision. The possibility that they represent slow cooling following the Grampian event is also highly unlikely owing to the similarity of the Lu-Hf and Sm-Nd ages along a considerable strike distance (Fig. 1). Further, most garnets show Lu enrichment in the core, indicating prograde growth. Most importantly, the Moine Nappe is in general lower grade than the Sgurr Beag and the Naver nappes, and therefore garnets would be expected to cool more quickly.
An alternative solution is that Late Ordovician metamorphism corresponds to an entirely younger accretion event. However, this is unlikely to have occurred with the Grampian and Northern Highland terranes in their present relative positions. The almost continuous record of sedimentation within the Southern Uplands accretionary prism from the Mid-Ordovician to the Wenlock precludes any further accretionary event along this part of the Laurentian margin until the collision with Avalonia in the Silurian-Devonian. Dewey & Strachan (2003) suggested that the restriction of Silurian (Scandian) orogenic activity to the Northern Highland terrane resulted from its location well away from the Grampian terrane until the two were juxtaposed following major end-Caledonian Devonian sinistral displacement along the Great Glen Fault. This scenario could similarly account for the restriction of regionally significant Late Ordovician metamorphism to the Northern Highland terrane. Comparisons can be drawn between the Moine Supergroup and the Laurentian-derived Uppermost Allochthon in west Norway (Fig. 1 inset), which contains evidence for a complex series of Ordovician accretionary events that occurred over a much longer time span than the Grampian event but significantly predated the final Scandian collision of Baltica and Laurentia (Roberts 2003;Roberts et al. 2007). These events culminated in eclogitefacies metamorphism at c. 450 Ma (Corfu et al. 2003), followed by rapid exhumation and development of a Late Ordovician to Silurian successor basin in the northern part of the Uppermost Allochthon (Roberts 2003). The ages that Corfu et al. (2003) obtained from zircons and titanites from an eclogite within the Uppermost Allochthon are within error of the Late Ordovician metamorphic event now recognized in the Moine Supergroup, thus suggesting that the Northern Highland terrane records a Caledonian history that has more in common with the Laurentian-derived rocks of Baltica than with the Grampian terrane of Scotland.
The dominant planar (S 2 ) and (L 2 ) fabrics developed throughout the northern part of the Morar Group wrap the Late Ordovician garnets dated in the present study. These fabrics have been assigned to the Scandian orogenic event on the basis that they are present within syn-D 2 metagranites that have yielded Silurian U-Pb zircon ages of c. 435-420 Ma Alsop et al. 2010). What, therefore, was the structural expression of the Late Ordovician event in the northern Morar Group? One solution is that it resulted in the D 1 folds and S 1 schistosity. Against this is the absence of these structures in a late Neoproterozoic metagabbro that intrudes the Morar Group of the Moine Nappe east of the Kyle of Tongue (Strachan et al. 2010b). In this light, D 1 is probably assignable to the mid-Neoproterozoic Knoydartian orogenic event that has been documented elsewhere within the southern part of the Morar Group (Rogers et al. 1998;Vance et al. 1998). An alternative solution that deserves consideration is that the D 2 folds within the northern Morar Group are composite in origin. These structures may have initially developed as tight to open structures during the Late Ordovician event, and were later strongly modified into their present tight to isoclinal, sheath-like geometry during intense shear associated with Scandian nappe stacking. Late Ordovician deformation and fabric development is certainly indicated by the curved inclusion trails within the dated garnets. These fabrics have presumably been completely transposed in the northern Morar Group, but are probably preserved on outcrop scale, at least locally, at the Basal Pelite (AB07-08) sample site where there is still continuity between inclusion trails and the external mica schistosity. Recognition of Late Ordovician metamorphism prompts the reexamination of published isotopic data from elsewhere in the Moine Supergroup. Swarms of variably deformed trondjhemitic pegmatites intrude the high-grade Moine rocks of the Glenfinnan Group in Inverness-shire and Ross-shire. These were once thought to be linked to regional migmatization but it now seems clear on structural grounds that the migmatites and the pegmatites are of different ages. Regional migmatization most probably occurred during the Neoproterozoic (Knoydartian) and the pegmatites cut folds and fabrics that plausibly formed during either the Knoydartian or Grampian events . Two pegmatites near Glenfinnan have yielded ages of 445 ± 10 Ma (Rb-Sr muscovite) and 450 ± 10 Ma (monazite bulk fractions) (van Breemen et al. 1974), although given the analytical errors these could have been intruded during either Grampian or Scandian events. A more precise age comes from the Glen Dessary syenite pluton, which was intruded into the Moine Nappe of Inverness-shire at 447.9 ± 2.9 Ma (U-Pb zircon, Goodenough et al. 2011), which is within error of the Late Ordovician ages of this study. The wider significance of these intrusions has been difficult to understand in the context of a two-phase Grampian-Scandian tectonic model. However, it now seems entirely possible that they are genetically related to the Late Ordovician event, which may therefore affect both the Moine and Sgurr Beag nappes.

Record of Silurian metamorphism in the Northern Highland Terrane
The lack of Silurian Lu-Hf isochrons or garnet-whole-rock ages in this study should not be taken as an indication of low metamorphic grade during the Scandian event in the Northern Highland Terrane. This in part reflects the sampling strategy which was to obtain garnets from samples that were for the most part relatively coarsegrained and texturally simple. There is abundant textural evidence in the northern Morar Group for amphibolite-facies metamorphism during the Scandian event. For example, at Port Vasgo, only a few kilometres away from the Talmine Pelite (AB07-15), the narrow euhedral rims of zoned garnets overgrow schistose fabrics assigned to the Scandian event (Burns et al. 2004). Furthermore, both staurolite and kyanite are observed to overgrow the main Scandian schistosity and lineation (Holdsworth et al. 2001). Feldspar within syntectonic metagranites has undergone widespread ductile recrystallization . At higher structural levels within the Naver Nappe, a U-Pb monazite age of 431 ± 10 Ma obtained from Moine migmatites at Kirtomy Point (Kinny et al. 1999) near the sample site of AB07-22 implies Scandian temperatures of c. 600 °C. The monazite age is close to the Sm-Nd garnet-whole-rock age of 433 ± 2 Ma reported in this study from a metabasic intrusion within the Strathy Complex (AB07-21). The lack of any Lu enrichment towards the rims of the garnets in sample AB07-21 indicates that no prograde Rayleigh zoning has been retained: the Sm-Nd age could thus record Scandian resetting of older (Mid-or Late Ordovician?) garnets. Johnson & Strachan (2006) drew attention to the synchroneity of Scandian thrusting in Sutherland with Barrovian metamorphism to amphibolite facies. This is difficult to reconcile with theoretical studies that indicate that there would not have been sufficient time to generate the high metamorphic temperatures during a thrusting event that lasted no longer than 18 Ma and probably rather less. Accordingly, they suggested that a plausible explanation for the heat source was that it resulted from a 20-25 Ma period of crustal heating in a back-arc setting between the Grampian and Scandian orogenic events. Thus most of the orogenic heat came from the preexisting back-arc rather than the orogenic process itself. The new evidence reported here rather suggests a different conclusion: the Scandian metamorphic temperatures were in part 'inherited' from the Late Ordovician event.

A revised plate-tectonic model for the Northern Highland Terrane
The results of the new mapping and geochronology summarized above prompt revision of Caledonian plate-tectonic models for the Northern Highland Terrane (Dallmeyer et al. 2001;Dewey & Strachan 2003;Kinny et al. 2003). A revised model can be divided into four main stages (Fig. 8).
(1) The Caledonian orogenic cycle commences with the convergence during the Late Cambrian and Early Ordovician of the rifted margin of Laurentia with an intra-oceanic subduction zone and associated Midland Valley-Lough Nafooey magmatic arc (Dewey & Ryan 1990) (Fig. 8a).
(2) Arc-continent collision at c. 490 Ma was associated with the obduction of ophiolite nappes (Dewey & Ryan 1990;Chew et al. 2010) and widespread 'Grampian I' deformation and metamorphism of the eastern part of the Moine Supergroup at 475-465 Ma (Fig. 8b). As yet, there is no unequivocal evidence that the westerly Moine Nappe was affected by the 'Grampian I' event. This was presumably followed by development of a subduction zone on the oceanic side of the accreted arc to accommodate continued closure of the Iapetus Ocean. Whether subduction was oceanward or continentward is uncertain; the latter is favoured here (Fig. 8b), to account for intrusion of the Glen Dessary syenite. (3) Accretion of an outboard microcontinental-arc fragment at c. 450 Ma resulted in 'Grampian II' deformation and metamorphism (Fig. 8c). Overthrusting of a proto-Sgurr Beag-Naver Nappe is invoked to explain garnet-grade metamorphism of the Morar Group-Moine Nappe. The structural expression of the 'Grampian II' event in the Moine Nappe is uncertain but some folds in the Morar Group may have been initiated at this stage as cylindrical and westfacing recumbent folds (see above). (4) Laurentia-Baltica collision at c. 435-425 Ma resulted in widespread Scandian deformation and metamorphism (Fig. 8d). Intense simple shear within the Moine Nappe and lower parts of the Sgurr Beag and Naver nappes was associated with regional ductile thrusting and rotation of fold axes to their present reclined attitude. Collision was followed from c. 425 Ma onwards by sinistral displacement along the Great Glen Fault, juxtaposing the Northern Highland Terrane against the Grampian Terrane to the SE, which was unaffected by the Scandian event (Dewey & Strachan 2003;Kinny et al. 2003).
The arc-microcontinental fragments that collided in the Ordovician with the Northern Highland Terrane may be preserved in the Laurentian-derived Uppermost Allochthons of Scandinavia ( Fig. 8d; see Roberts 2003). The use of the terms 'Grampian I' and 'Grampian II' is analogous to the use of 'Taconic I' and 'Taconic II' for multiple accretion events in the Newfoundland Appalachians (van Staal et al. 2009), although precise contemporaneity and lateral continuity is not implied. An implication of the above model is that the Scandian displacements on the Moine and Sgurr Beag-Naver thrusts were associated with the reworking of, respectively, the 'Grampian I' and 'Grampian II' orogenic fronts inferred in Figure 8b and c.

Conclusions
(1) Lu-Hf and Sm-Nd isochrons and two-point garnet-wholerock ages of c. 475-460 Ma have been obtained from prograde garnets in Moine gneisses, Dalradian schists and metabasic intrusions within the Northern Highland and Grampian terranes. The new dataset is consistent with previously published isotopic data and confirms that the Mid-Ordovician Grampian arc-accretion event was approximately synchronous in both terranes.
(2) Distinctively younger Lu-Hf and Sm-Nd isochrons and two-point garnet-whole-rock ages of c. 450 Ma have been obtained from prograde garnets within the structurally lowest unit of the Northern Highland terrane, the Moine Nappe, and provide evidence for a hitherto unrecognized Late Ordovician regional metamorphic event. This event postdates peak metamorphism in the Naver Nappe by c. 17 Ma, and predates Silurian (Scandian) nappe stacking and associated deformation by at least 13 Ma. The existing twostage Grampian-Scandian model for Caledonian orogenesis in northern Scotland is thus an oversimplification, and a significantly more complex structural evolution is required than considered previously. An additional episode of accretion of an arc-microcontinental fragment to the Laurentian margin is invoked to account for this Late Ordovician metamorphic event.
(3) The apparent restriction of the Late Ordovician and Silurian events to the Northern Highland terrane reinforces the suggestion that it was far removed from the Grampian terrane until the two were juxtaposed following major end-Caledonian sinistral displacement along the Great Glen Fault. The Laurentian-derived Uppermost Allochthons of Norway contain a very similar, protracted record of Ordovician accretion-related metamorphism that significantly predates Silurian continent-continent collision and closure of the Iapetus Ocean. The Caledonian history of the Northern Highland terrane thus has more in common with the Laurentian-derived rocks of Baltica than with the Grampian terrane of Scotland. (4) In many parts of the Moine Supergroup, traditional structural analysis using 'D-numbers' has identified only two or three phases of folding and fabric development. An implication of the emerging geochronological dataset is that this method of structural analysis provides only a partial record of a complex and protracted tectonothermal history. Multiple episodes of ductile reworking and recrystallization at amphibolite facies have presumably resulted in widespread modification and variable transposition of early folds and fabrics.
Journal reviewers S. Daly, I. Millar and R. Holdsworth are thanked for their detailed comments that resulted in substantial improvements to earlier versions of the paper.