Mafic dyke remnants in the Lewisian Complex of the Outer Hebrides, NW scotland: a geochemical record of continental break-up and re-assembly

Precambrian Research 133 (2004) 121 141 Mafic dyke remnants in the Lewisian Complex of the Outer Hebrides, NW scotland: a geochemical record of continental break-up and re-assembly A.J. Mason, T.S. Brewer Department of Geology, University of Leicester, University Road, Leicester LE1 7RH, UK Received 7 April 2003; accepted 9 April 2004 Abstract In the predominantly late Archaean Lewisian Complex of NW Scotland, the Palaeoproterozoic (2.4 2.0 Ga) Scourie dykes have been used to discriminate between late Archaean events, and ca.1.7 Ga Laxfordian reworking. On the Outer Hebrides metabasite dykes intrude the Lewisian Complex, and were previously assumed to be correlatives of the Scourie dykes of the Scottish mainland. New geochemical data allow the Outer Hebrides metabasites to be divided into two groups. Dykes intruded into Archaean Gneisses north and south of the ca. 1.9 Ga South Harris Complex (SHC) are identical, and have depleted geochemical signatures. They are linked with an early episode of rifting (ca. 2 Ga) related to the fragmentation of the Lewisian crust and eventual formation of ocean crust. Later subduction of this ocean crust generated the ca. 1.9 Ga arc related rocks of the SHC, which contain a second group of subduction-related metabasites that are different from the dykes in the adjacent Archaean gneisses. Closure of this ocean and subsequent Laxfordian continent-arc-continent collision at ca. 1.7 Ga, sutured the rifted Archaean fragments back together. The northeast boundary of the SHC has previously been interpreted as a terrane boundary. However, the new dyke data demonstrates that any terrane boundary actually lies within the SHC. The northeast segment of the SHC, the Langavat Belt, contains dykes chemically identical to those intruding the Archaean gneisses to the northeast. The southwest segment of the SHC, the Harris Granulite Belt, contains arc related plutons and metabasites with arc signatures, which are distinct from mafic dykes in the adjacent Archaean gneisses and Langavat Belt. Consequently, much of the Langavat Belt is unrelated to the arc complex. Data from the Outer Hebrides dykes casts doubt on their correlation with the Scourie dykes, and demonstrate that the Outer Hebrides dykes have a closer affinity with the Palaeoproterozoic Kangâmiut dykes of Greenland, than to the Scourie dykes. 2004 Elsevier B.V. All rights reserved. Keywords: Lewisian Complex; Mafic dykes; Laxfordian; Outer Hebrides; Geochemistry 1. Introduction Corresponding author. Present address: NERC Isotope Geo- Science Laboratory (NIGL), Nottingham, NG12 5GG, UK. Fax: +44 115 936 3302. E-mail address: amason@bgs.ac.uk (A.J. Mason). Mafic to ultramafic Scourie dykes have been used to divide the history of the dominantly late Archaean Lewisian Complex of NW Scotland into two broad tectonothermal episodes (Sutton and Watson, 1951). 0301-9268/$ see front matter 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.precamres.2004.04.001

122 A.J. Mason, T.S. Brewer / Precambrian Research 133 (2004) 121 141 The older of these, the Scourian event is defined as pre-dating the Scourie dykes, while the younger Laxfordian event post-dates dyke emplacement. Based on the extent of Laxfordian reworking the mainland Lewisian was divided into a Northern, Southern and Central Region, the latter is largely unaffected by the reworking (Fig. 1A). Subsequent studies have demonstrated the existence of ca. 1.9 Ga juvenile Proterozoic arc/arc-accretion complexes within the Lewisian, in the form of the South Harris Complex (SHC), and Loch Maree Group and associated Ard Gneiss (Cliff et al., 1983; Baba, 1997, 1998; Park et al., 2001; Friend and Kinny, 2001; Whitehouse and Bridgwater, 2001; Mason et al., 2004). Geochronological studies have also detected fundamental differences in gneiss protolith age and metamorphic history between the different Archaean blocks (Friend and Kinny, 2001; Kinny and Friend, 1997). On this basis Friend and Kinny (2001, and references therein) have suggested the Lewisian was assembled from disparate terranes during the Laxfordian. For the mainland Northern and Central Regions (respectively renamed the Rhiconich and Assynt Terranes by Friend and Kinny (2001)) this occurred between 1854 and 1730 Ma, while on the Outer Hebrides the South Harris Complex (Roineabhal terrane) was supposedly juxtaposed against the Archaean gneisses to the NE (Tarbert terrane) after 1675 Ma (Friend and Kinny, 2001) (Fig. 1). A consequence of this model is that the Lewisian terranes would be expected to have independent histories, prior to amalgamation. Olivine gabbro and bronzite picrite members of the Scourie dykes from the Assynt terrane are dated at 1992 +3/ 2 Ma and 2418 +7/ 4 Ma, respectively (Heaman and Tarney, 1989), and post-date most metadolerite members of the Scourie dykes (Weaver and Tarney, 1981; Tarney, 1973). Hence, there should be little reason for the dykes in different terranes to be genetically related. In the Outer Hebrides, mafic dyke remnants are abundant, and have been regarded as a regional suite termed the Younger Basics, which have been correlated with the Scourie dykes of the mainland (Fettes et al., 1992; Dearnley, 1962). However, recent geochronological studies have failed to correlate the main outcrop of the Outer Hebrides below the Outer Isles Thrust (OIT), with the mainland outcrops above the OIT (Fig. 1) (Whitehouse and Bridgwater, 2001; Friend and Kinny, 2001; Mason et al., 2004). One particular problem is the presence of a 2480-Ma metamorphic event in the Assynt terrane, which is only 60 Ma older than the bronzite picrites, but is absent on the Outer Hebrides, implying these two areas were separate at 2480 Ma. Thus, the correlation of the pre-bronzite picrite (pre-2418 Ma) metadolerites of the Assynt terrane with metadolerites in the Outer Hebrides is suspect. Furthermore, the identification of a ca. 1890 Ma probable continental volcanic arc through South Harris (Baba, 1997; Friend and Kinny, 2001; Whitehouse and Bridgwater, 2001; Mason et al., 2004), casts doubt on the assertion of Fettes et al. (1992) that the dyke remnants in the southern Outer Hebrides are equivalent to those in the north. Despite these lines of evidence, it cannot be assumed that the metabasic intrusions in different terranes will be unrelated. For example, it is equally possible that parts of the Lewisian were rifted apart and then re-assembled in the Laxfordian, rather than being assembled from entirely independent blocks. In this case, pre- to syn-rift events, as well as syn- to post-assembly events could be shared. Alternatively, some of the dykes within the Archaean blocks of the Outer Hebrides could form part of the arc-related magmatic activity associated with the SHC. There are also problems with the terrane model with respect to the junction of the Tarbert and Roineabhal terranes. The boundary separating these terranes is regarded by Friend and Kinny (2001) to be marked by a horizon of ultramafic bodies within a 1 2 km wide zone known as the Langavat Belt (Fig. 1C). According to Friend and Kinny (2001) 1675 Ma granite pegmatites occur abundantly to the NE of this boundary, within the Tarbert terrane, but not to the SE. On this bases the terrane boundary is concluded to be younger than 1675 Ma. The problem is that the granite pegmatites do not stop at the horizon of ultramafic pods and occur throughout much of the Roineabhal Terrane (SHC), as is clearly demonstrated by Dearnley (1963) and Mendum (1982). Thus, any boundary is actually older than the 1675-Ma pegmatites. A second problem is that the Roineabhal terrane is supposed to be a ca. 1900-Ma granulite facies terrain (Friend and Kinny, 2001), but the main components of the Langavat Belt, both above and below the supposed boundary lack any evidence (see the following text) of having attained granulite facies conditions. Hence, both the age and

Fig. 1. Generalised geological map of the Lewisian Complex, compiled from Drury (1972), Fettes et al. (1981a,b), and Johnstone and Mykura (1989). Names in brackets are terrane names proposed by Friend and Kinny (2001). The Assynt terrane is only approximately equivalent to the Central Region as indicated in (A). Map coordinates on (B) and (C) correspond to Ordnance Survey (UK) grid. A.J. Mason, T.S. Brewer / Precambrian Research 133 (2004) 121 141 123

124 A.J. Mason, T.S. Brewer / Precambrian Research 133 (2004) 121 141 location of this boundary is uncertain. Because of this the terrane terminology proposed by Friend and Kinny (2001) for the Outer Hebrides is not adopted here. Comparison across the OIT is also valid. The OIT was probably initiated between 1.4 and 1.1 Ga during the Grenville event (Cliff and Rex, 1989; Magloughlin et al., 2001; Imber et al., 2002), and is younger than the assembly of the Lewisian, and associated Laxfordian tectonothermal activity, which based on metamorphic ages probably ceased between 1650 and 1625 Ma (Friend and Kinny, 2001; Cliff et al., 1998). It is therefore not unreasonable to assume that terranes in the footwall of the OIT on the Outer Hebrides have correlatives somewhere in the hanging wall, and these could occur on the mainland or Inner Hebrides. The purpose of this study is to present new geochemical data from the central portion of the Outer Hebrides. The objectives of this are as follows: (1) To determine if the dykes within the Archaean gneisses NE of the SHC are identical to those in the Archaean block to the SW of the SHC, and to determine if the dykes in either area are similar to the metabasites in the SHC itself. (2) To determine the likely origin of the dykes and their relationship to the ca. 1900-Ma arc rocks of the SHC. (3) To locate more precisely the NE boundary of the SHC arc rocks, and to determine whether the Langavat Belt has a closer affinity with these arc rocks, the Archaean block to the NE, or is distinct from both. (4) To compare the dykes on the Outer Hebrides with published data from the dykes above the OIT on the mainland and Inner Hebrides, to look for possible equivalents. (5) To compare the Outer Hebrides dykes with early Proterozoic mafic dykes from Greenland. 2. Geological setting The Outer Hebrides is composed almost exclusively of the Lewisian Complex, the vast majority of which consists of late Archaean tonalitic orthogneiss (grey gneisses) (Rollinson and Fowler, 1987; Park and Tarney, 1987; Fettes et al., 1992). In the footwall of the OIT, grey gneisses are divided into a northern and southern area by a belt of distinctive metasediments and meta-igneous rocks forming the SHC (Fettes et al., 1981a,b), and are here termed the Northern and Southern Gneisses (Fig. 1). 2.1. The Northern and Southern Gneisses The footwall grey gneisses typically consists of coarse-grained, crudely banded amphibolite facies biotite, or hornblende-biotite gneiss. Granulite facies pyroxene-bearing felsic gneisses have been reported locally below the OIT on South Uist and several small islands north of Barra (Fettes et al., 1992; Francis and Sibson, 1973; Francis, 1973) (Fig. 1). Within low strain zones, e.g. at Garry-a-Siar and Ardivachar (Fig. 1), two distinct occurrences of metabasic rocks can be identified. The older generation consists of intensely migmatised and dismembered masses that appear to be intruded by the youngest components of the gneisses, giving rise to agmatites. These are cross cut by unmigmatised, less deformed mafic dykes. A detailed study of these two areas by Dearnley and Dunning (1968) describes dykes in a range of deformation states from virtually undeformed and strongly discordant (Fig. 2A), through to relatively deformed and partly dismembered. In particular, Dearnley and Dunning (1968) describe a pair of dykes that are strongly discordant, but show local incipient disruption into blocks. Other dykes in the same area have been locally deformed into concordance with the gneisses. The lack of intense migmatisation, relative homogeneity, and the generally larger size of the concordant dykes allows them to be distinguished from the older mafic rocks integral to the grey gneisses. Outside areas of relatively low strain, the dykes are seldom well preserved, and take the form of either concordant pods, typically 1 20 m across, wrapped around by the gneissic foliation, or of thin (<0.5 m) concordant layers (Fig. 2B). Within high strain zones the metabasites are strongly deformed and have sharp, usually parallel contacts, and range from a few centimetres to >100 m in thickness (Fig. 2C). Angular discordances are occasionally still preserved indicating that many of these masses are deformed, fragmented mafic dykes, and just display more extreme forms of the relationships described by Dearnley and Dunning (1968).

Fig. 2. Field relationships between metabasite dykes and the grey gneisses. (A) Well-preserved, weakly deformed mafic dyke, Garry-a-Siar, Benbecula (see Fig. 1 for location); pen length 150 mm. (B) Typical grey gneiss with boudinaged and dismembered dykes, South Harris. Height of rock face ca. 2.5 m. (C) Highly stretched and flattened mafic dykes in the Langavat Belt, South Harris; hammer ca. 0.45 m long. (D) Undeformed amphibolite with relict porphyritic texture from a ca. 0.5 m wide dyke, Ardivachar, South Uist; coin diameter 17.5 mm. A.J. Mason, T.S. Brewer / Precambrian Research 133 (2004) 121 141 125

126 A.J. Mason, T.S. Brewer / Precambrian Research 133 (2004) 121 141 The dyke-rocks consist mostly of fine to mediumcoarse grained, dark grey amphibolite. Some dykes contain garnet clinopyroxene-rich assemblages. Textures generally relate to deformation and metamorphism, but relict igneous textures are sometimes preserved (Fig. 2D). In addition to the mafic dykes, a suite of older banded metabasics was recognised by Coward et al. (1969) and Fettes et al. (1992). In areas of strong post-dyke deformation it is not always possible to differentiate between deformed dykes and metabasic rocks of other origins. 2.2. The South Harris complex The South Harris Complex is here subdivided into two units, a southwestern granulite facies portion (Harris Granulite Belt), and an amphibolite facies northeastern part (Langavat Belt; Fig. 1C). 2.2.1. The Harris Granulite belt The Harris Granulite Belt comprises metasediments (Leverburgh Belt), which contain three large meta-igneous masses, a metadiorite, a metanorite, and a meta-anorthosite (Dearnley, 1963; Baba, 1997). These have been subjected to granulite facies metamorphism, in part at least under high-pressure conditions, 800 ± 30 C and 13 14 kbar (Baba, 1998). This has resulted in dominantly anhydrous, usually garnet-rich mineral assemblages in both the metasediments and metabasic rocks, and the former are commonly migmatised. These assemblages have been variably re-hydrated during retrogression, and decompression symplectites commonly partially replace garnet in both the metasediments and metabasites (Baba, 1998). Metabasic rocks are abundant within the metasediments and meta-anorthosite, but are scarce within the metadiorite and metanorite. In the former, they occur as isolated concordant lenses and blocks ranging from a few centimetres to several tens of metres across. Those in the anorthosite occur as mafic spots, diffuse bands, and sharply differentiated and occasionally discordant bands. The first two types appear to be integral to the anorthosite itself, and are not considered further. The third type appears to represent deformed dykes. In addition, larger masses of banded metabasite up to about 200 m thick occur, and are mostly associated with the margins of the main meta-igneous masses, although not always in direct contact. Contacts are usually sheared, but one body is intruded by the metadiorite (Dearnley, 1963). Most metabasites consist of coarse garnet clinopyroxene rich granulites, but some amphibolites also occur. All three of the main meta-igneous bodies are locally intruded by mainly pyroxenitic net-vein systems (Fettes et al., 1992; Mendum, 1982; A. Mason, unpublished data). The contact between the granulites and the Southern Gneisses lies mainly under water (Mendum, 1982), but on Ensay (Fig. 1B) Graham (1980) describes it as a shear zone dipping NE at about 50. 2.2.2. The Langavat Belt The Langavat Belt dips steeply southwest beneath the Harris Granulite Belt and lies within a major shear zone, the Langavat Shear Zone (LSZ). Deformation associated with the LSZ occurs mainly within the Langavat Belt, but the adjacent margins of the Harris Granulite Belt and Northern Gneisses are also affected, although the boundaries of the LSZ are diffuse. The rocks that comprise the Langavat Belt are almost everywhere strongly deformed, and deformation is particularly intense adjacent to the metadiorite where ca. 200 m of thinly banded mylonites and mylonitic gneisses are developed (Fig. 3). The Langavat Belt comprises undoubted metasediments such as semi-pelite (garnet biotite schist), rare pelite (garnet biotite sillimanite schist), calc silicate and marble, as well as amphibolite, quartzo feldspathic and ultramafic rocks of probable meta-igneous parentage (Fig. 3). The latter mainly outcrop on a horizon near the centre of the belt. All of these lithologies are overwhelmingly dominated by hydrous mineral assemblages characteristic of the amphibolite facies. Unlike the Harris Granulite Belt, pelitic and semi-pelitic metasediments show an almost total lack of migmatisation. Garnet is also invariably fresh, commonly euhedral, and never shows the development of decompression symplectites. Furthermore, metabasic rocks are exclusively amphibolites or hornblende schists, that even where undeformed lack relics of garnet-rich granulite facies assemblages. In the rare cases where garnet is present, this forms small, fresh,

A.J. Mason, T.S. Brewer / Precambrian Research 133 (2004) 121 141 127 Fig. 3. New geological map of the Loch Langavat area simplified and redrawn from unpublished mapping by A. Mason. Map grid corresponds to square NG, Ordnance Survey (UK) grid. The upper (SW) contact of the Langavat Belt is here defined as the boundary between the metadiorite and the banded mylonitic gneiss. Note that the flaggy quartzo feldspathic rock forming the centre of the belt is locally continuous with the Northern Gneisses, and probably represents high-strain grey gneiss. Because of this, the belt has no overall clearly defined base, despite the individual metasediment-bearing packets being well defined. However, the lower (NE) contact is taken as the base of the lowest unit containing undoubted metasediment (garnet semi-pelite and calc silicate). Also note that the granite pegmatites occur throughout the belt, and that the ultramafic pods do not all fall on one horizon. The semi-pelites bands are not migmatised and have fresh garnets. The main amphibolite bodies are garnet-free, except for the amphibolite with ultramafic pods. This occasionally contains small, fresh, euhedral garnets. often euhedral crystals and not corroded relics. The occasional preservation of relict gabbroic textures is also significant, because it demonstrates that the very coarse garnet clinopyroxene dominated assemblages typical of the metabasites in the Harris Granulite Belt, were never developed. Metabasic rocks are abundant and can be divided into two types based on field relationships. The first comprises three large concordant slabs up to around 150 m thick that are traceable for many kilometres along strike, and form part of the stratigraphy of the belt (Fig. 3). These are associated with the main metasedimentary rocks in the belt. The origin of these bodies is uncertain. The second type comprises numerous smaller laterally discontinuous sheets that range from around 0.02 30 m in thickness. Contacts are mostly sharp and concordant, but where strain is relatively low, slightly discordant contacts and relict igneous textures are preserved. These bodies are interpreted as variably deformed mafic dykes. These dykes are most abundant within the quartzo feldspathic rocks of the belt, but are fairly common within the distinctive metasediments, and occur both above and below the main ultramafic pods. In the field, the dykes appear to be continuous from the highly deformed grey gneisses to the NE, into the Langavat Belt. Dykes have not been observed within the large concordant amphibolites, but this may well in part be due to the difficulty of recognising one amphibolite within another, rather than to their absence. Metabasic dykes are conspicuously absent from the metadiorite, and their disappearance apparently coinciding with the 200 m thick mylonitic zone at the contact (Fig. 3). Typically the dykes consist of fine to medium grained amphibolite, sometimes with clinopyroxene. The presence of clinopyroxene in

128 A.J. Mason, T.S. Brewer / Precambrian Research 133 (2004) 121 141 metabasic rocks, and sillimanite in pelitic lithologies indicates metamorphism to upper amphibolite facies. The lack of migmatisation in the metasediments, or any other textural evidence (e.g. decompression symplectites), implies that the Langavat Belt has never attained granulite facies conditions, unlike the Harris Granulite Belt. 2.3. Age of the metabasites The age of the dominant metadolerite/metagabbro dyke remnants in the grey gneisses is not known with certainty. They must be younger than the country rock gneisses, dated at 2.8 Ga (Pidgeon and Aftalion, 1972; Friend and Kinny, 2001; Whitehouse and Bridgwater, 2001; Mason et al., 2004), and contain Laxfordian metamorphic assemblages dated at 1625 1660 Ma (Cliff et al., 1998). U Pb zircon lower intercept ages of 2049 +78/ 80 Ma and 2039 +99/ 100 Ma from Northern Gneiss samples from adjacent to mafic dykes, provide the best available estimate of the emplacement age for the dykes (Mason et al., 2004). The scarcity of dykes intruding the 1888 ± 2 Ma SHC metadiorite and 1890 +2/ 1Ma metanorite (Mason et al., 2004) is consistent with this. Cliff et al. (1998) has dated two mafic bodies from the Northern Gneisses at 2.54 ± 0.08 Ga and 2.140 ± 0.038 Ga, with noritic and picritic affinities (Fettes et al., 1992), which may or may not be related to the metadolerite/metagabbro dykes considered here. The metabasites within the Harris Granulite Belt generally lack primary intrusive field relations and occur as dismembered pods. Hence, little can be said about their age of emplacement, except that most pre-date 1827 ± 16 Ma (Cliff et al., 1998) granulite facies metamorphism. Cliff et al. (1983) obtained Nd-model ages of 2.64 1.49 Ga from the metabasites. 3. Sampling A total of 105 metabasites have been collected from across the central portion of the Outer Hebrides covering the four areas described above (Fig. 1). Most material has been obtained from the coast, or along blasted road sections. Within the grey gneisses, bodies that retain some evidence of cross cutting relationships have been preferentially collected. Although, in order to provide reasonable coverage this has not been possible over much of the area. In this case, samples were preferentially obtained from relatively homogeneous layers or pods that lack intense migmatisation, and are on a scale similar to the well-preserved dykes in the low strain zones. The intention was to maximise the number of sampled bodies representing dyke fragments. Samples from the Langavat Belt originate from relatively small (<30 m thick) mostly concordant bodies, all of which are considered to be heavily deformed mafic dykes; samples were collected from both above and below the ultramafics. The large concordant masses of uncertain origin within the belt are not considered. All the types of occurrence of mafic rocks within the Harris Granulite Belt have been collected (except the net-veins), in order to look for possible equivalents of the dykes in the adjacent areas. In all cases samples originate from structurally below the OIT (Fettes et al., 1992). Material with obvious signs of low-grade alteration, such as excessive amounts of biotite or epidote has largely been avoided. Most samples have suffered some degree of penetrative shearing at either amphibolite or granulite facies. 4. Petrography Based on their petrography the samples have been divided into three groups (Table 1). Textures in most samples relate to metamorphism and deformation, and relict igneous textures are rare; pristine igneous rocks have not been encountered. The growth of epidote and chlorite, in some samples appears to be relatively late stage, the former often occurring as post-tectonic grains, and the latter sometimes replaces biotite. In the granulite facies samples, biotite also appears to be a late retrogressive phase mantling and partly replacing pyroxene and sometimes hornblende. A large proportion of biotite is considered to indicate K metasomatism. Thus five (group 3) samples containing significant amounts of biotite, epidote, or chlorite are regarded as altered, and have been excluded. 5. Analytical method All samples were milled in agate and analysed for major and trace elements by X-ray fluorescence

A.J. Mason, T.S. Brewer / Precambrian Research 133 (2004) 121 141 129 Table 1 Summary of petrography of analysed samples Group Typical mineral assemblages Southern Gneisses Harris Granulite Belt Langavat Belt Northern Gneisses Hornblende plagioclase quartz [clinopyroxene] [titanite] [opaques] Hornblende plagioclase quartz [clinopyroxene] [titanite] [opaques] Garnet clinopyroxene hornblende [plagioclase] [quartz] [scapolite] [biotite] [orthopyroxene] [rutile] opaques 1 Relatively fresh metabasites Hornblende plagioclase quartz [garnet] [clinopyroxene] [titanite] [opaques] As above with biotite, epidote or partly As above with biotite, chlorite, epidote, or As above with biotite or partly altered partly altered plagioclase As above with extensive 2 Moderately altered metabasites As above with biotite, epidote, or partly altered plagioclase decomposed plagioclase As above with extensively altered plagioclase plagioclase As above with extensive development (> 10%) of biotite 3 Highly altered metabasites As above but with large amounts (> 10%) development of biotite (> 10%) of biotite, epidote, or extensively altered plagioclase Phases in brackets not present in all samples. spectrometry following the procedures by Brewer et al. (1998). To improve precision for Zr, Y and Nb these elements were determined by XRF using long counting times; this procedure both lowered the detection limits (Nb 0.4 ppm, Y 0.5 ppm, Zr 0.5 ppm) and resulted in a precision better than 1% compared to the normal counting statistics. Rare-earth elements (REE) were determined by inductively coupled plasma emission spectrometry following the procedures in Harvey et al. (1996) using a JY-Ultima-2 spectrometer. Representative data are presented in Table 2. 6. Results The metabasites have SiO 2 concentrations from 42.9 to 54.8%, TiO 2 from 0.2 to 3.2%, Al 2 O 3 from 7.5 to 19.0%, total iron (as Fe 2 O 3 ) from 8.6 to 18.5%, MgO from 3.5 to 14.0%, CaO from 6.0 to 17.0%, and P 2 O 5 from 0.03 to 0.45%. For most major elements moderately altered (group 2) and relatively fresh samples (group 1) generally plot together, although K 2 O is elevated in the more altered samples. Similarly, MgO reaches a maximum of about 11% in relatively fresh samples, but is about 14% in one moderately altered sample. Consequently, moderately altered samples are included in the following discussion, except where they lie outside the range defined by the fresher samples. Metabasites from the Northern Gneisses, Southern Gneisses and Langavat Belt consistently plot together on major element diagrams (Fig. 4). Slight differences are noted when these data are compared with samples from the Harris Granulite Belt. In the latter, SiO 2, K 2 O and TiO 2 are generally lower in concentration, while CaO and MgO on average are slightly enriched. Furthermore, data from the former areas produce fairly well-defined trends for most elements, while the data from the Harris Granulite Belt is rather more scattered (Fig. 4). The majority of trace element data for fresh (group 1) and moderately altered (group 2) samples plot together, although in the Langavat Belt and Northern Gneisses some altered (group 2) samples contain elevated Rb, and to a lesser extent Ba. Trace element data for the Langavat Belt, Northern Gneisses, and Southern Gneisses plot consistently together and the three

130 A.J. Mason, T.S. Brewer / Precambrian Research 133 (2004) 121 141 Table 2 Representative data from analysed metabasite samples from the Outer Hebrides Spl. AM221 AM209 AM160 AM151 AM219 AM180 AM003 AM230 AM156 AM193 AM205 AM163 AM126 Lctn. SG SG SG NG NG NG LB LB LB LB HGB HGB HGB SiO 2 49.7 49.4 51.0 51.0 49.3 48.7 49.2 48.7 48.9 53.1 52.5 42.9 47.8 TiO 2 1.70 1.36 2.03 1.42 1.24 1.77 1.49 1.48 1.75 0.27 0.75 0.59 1.05 Al 2 O 3 12.9 13.9 13.5 12.8 13.6 13.2 13.7 13.7 13.0 17.4 13.2 16.2 13.1 Fe 2 O 3 15.3 13.9 15.6 15.3 14.8 14.9 16.0 14.3 15.3 8.6 10.6 15.1 14.9 MnO 0.22 0.20 0.24 0.21 0.23 0.23 0.24 0.21 0.22 0.14 0.10 0.18 0.21 MgO 5.4 6.4 4.7 4.9 5.5 5.8 5.8 6.5 6.2 6.2 6.1 8.5 8.3 CaO 9.3 10.4 9.1 8.9 10.1 10.1 10.4 10.8 10.1 9.3 11.5 12.1 11.0 Na 2 O 2.5 2.7 2.1 2.4 2.9 2.8 2.6 2.4 2.3 4.6 3.6 1.0 2.0 K 2 O 1.13 0.85 1.09 1.25 0.95 1.50 0.57 0.44 0.64 0.66 0.34 0.29 0.28 P 2 O 5 0.21 0.15 0.19 0.16 0.14 0.18 0.14 0.15 0.11 0.06 0.16 0.05 0.06 LOI 0.34 0.32 0.23 0.37 0.26 0.40 0.00 0.38 0.52 0.45 0.91 0.74 0.39 Total 98.62 99.65 99.69 98.81 99.09 99.54 100.45 99.06 99.01 100.79 99.93 97.72 99.11 Sc 47 47 48 43 48 45 45 47 45 38 32 64 49 V 347 319 361 449 307 335 322 321 369 180 196 724 408 Cr 36 98 61 0 72 54 86 132 91 29 293 N.D. 180 Co 64 59 63 64 63 64 64 60 68 42 45 68 65 Ni 49 70 40 47 46 56 67 79 75 61 38 30 117 Cu 240 78 217 55 123 38 212 70 479 80 122 52 94 Zn 132 107 133 114 126 134 143 112 127 64 36 92 115 Ga 19 18 20 21 17 18 18 18 18 19 17 19 18 Rb 25 8 25 17 17 13 7 5 11 17 2 1 4 Sr 199 164 189 188 150 146 170 236 169 544 127 327 170 Y 36.1 29.3 41.2 33.9 28.1 37.1 32.3 28.9 29.2 8.7 20.4 6.4 26.3 Zr 143.7 99.9 130.7 104.6 87.0 119.4 119.5 106.7 98.9 44.2 81.5 27.6 40.7 Nb 13.2 9.6 8.6 11.2 8.5 11.4 10.0 9.0 8.7 1.2 4.2 1.2 1.0 Ba 275 122 219 227 86 157 99 56 116 221 67 95 129 La 13 11.0 11.4 5 8.2 11 8.3 5 7 13 13 5 7 Ce 27.2 30.3 21.8 23.0 Pr 4.0 4.2 3.1 4.3 Nd 19 14.6 18.4 8 12.6 17 13.9 18 13 8 16 12 11 Sm 3.9 5.1 3.5 3.4 Eu 1.2 1.5 1.1 1.2 Gd 4.0 5.5 3.8 4.4 Dy 4.1 5.7 3.9 5.0 Er 2.4 3.4 2.3 3.1 Yb 2.3 3.3 2.3 2.7 Lu 0.3 0.5 0.4 0.4 SG, Southern Gneisses; NG, Northern Gneisses; LB, Langavat Belt; HGB, Harris Granulite Belt. areas are virtually inseparable (Fig. 4). The only slight differences noted are that the samples from the Langavat Belt are very slightly poorer in Sc, and slightly richer in Sr. More pronounced differences are seen in the samples from the Harris Granulite Belt. In general these metabasites have lower incompatible element concentrations, and more specifically have a higher P/Nb and Nd/Nb ratios (Fig. 4). 7. Discussion 7.1. Effects of alteration and metamorphism The data suggest that most elements are largely unaffected by relatively late stage alteration, as evidenced by the fact that relatively fresh (group 1) and moderately altered (group 2) samples generally plot together. Only the more mobile elements such

A.J. Mason, T.S. Brewer / Precambrian Research 133 (2004) 121 141 131 Fig. 4. Bivariate plots for selected elements for metabasites from the Outer Hebrides. Crosses: Southern Gneiss dykes; open stars: Northern Gneiss dykes; open triangles: Langavat Belt dykes; filled circles: Harris Granulite Belt metabasites.

132 A.J. Mason, T.S. Brewer / Precambrian Research 133 (2004) 121 141 Fig. 4. (Continued ). as Rb, and to a lesser extent K 2 O, appear to have been significantly affected. The highest values occur in samples containing modal biotite, with concentrations exceeding 2% and 80 ppm, respectively for Rb and K 2 O. The influence of the amphibolite/granulite facies metamorphism is more difficult to assess, since no pristine material is available for comparison. However, the increase of K 2 O, TiO 2, and P 2 O 5, and the decrease of Al 2 O 3, and CaO with decreasing MgO content in the Langavat Belt, Northern Gneisses and Southern Gneisses data are consistent with fractional crystallisation (Fig. 4). The generally strong positive correlation between highly incompatible elements such as the LREE, Nb, Zr, Y, Ti, and between highly compatible elements such as MgO, Cr and Ni (Fig. 4) is also consistent with igneous processes. Furthermore, Rb and Ba still retain a fairly good positive correlation with other incompatible components such as P 2 O 5 and Nb (in group 1 samples), suggesting only limited element mobility. Consequently, it appears that metamorphism in the Langavat Belt, Northern Gneisses and Southern Gneisses was relatively isochemical. The chemistry of the metabasic rocks of the Harris Granulite Belt appears to have been more strongly affected by metamorphism, as suggested by the general lack of well-preserved major element trends (Fig. 4). Some incompatible trace elements (Nb, Zr, P 2 O 5, Nd, TiO 2, and to a lesser extent Y and La) still show reasonable positive correlations consistent with crystal fractionation. 7.2. Igneous evolution trends and contamination Due to the differences between the data from the Harris Granulite Belt and the other areas, and the close similarity of the data from the Langavat Belt, Northern Gneisses and Southern Gneisses, it is appropriate to consider the data from the former area separately. The remaining data, when plotted on the AFM diagram (Irvine and Baragar, 1971), fall mainly within the tholeiitic field and define a Fe-enrichment trend. The trends displayed on major element plots (Fig. 4), namely decreasing CaO and Al 2 O 3, with decreasing MgO are consistent with the fractionation of plagioclase and one or more mafic phases (Mg-rich, Ca-poor pyroxene, or olivine). Plagioclase fractionation is supported by the rare occurrence of relict plagioclase phenocrysts (Fig. 2D). Trace element data, particularly the removal of Ni with decreasing MgO is consistent with the fractionation of olivine (Fig. 4). A similar pattern is displayed by Cr and implies the involvement of an additional mafic phase, possibly a pyroxene or opaque phase. The Harris Granulite Belt metabasites mainly fall within the tholeiitic field on the AFM diagram (Irvine and Baragar, 1971). But it should be noted that use of this diagram might be inappropriate for these samples, because of the apparent disturbance to the major elements. Since no coherent major element trends are preserved little can be said about the fractionation history. However, the generally lower SiO 2 content, less concentrated incompatible trace elements and greater concentration of CaO and MgO implies the samples from the Harris Granulite Belt

A.J. Mason, T.S. Brewer / Precambrian Research 133 (2004) 121 141 133 Fig. 5. Primitive mantle normalised plots. Scourie dyke data from Weaver and Tarney (1981), Muir et al. (1993), MD dykes from Hall et al. (1985), Kangâmiut dykes from Cadman et al. (2001). Geochemical data for Lewisian felsic gneiss from the Outer Hebrides (Mason, unpublished data). Primitive mantle normalisation values after Wood et al. (1979), except Ti from Wood et al. (1981). are generally less evolved than those from the adjacent areas. This does not account for the different incompatible element ratios P/Nb and Nd/Nb, which should not be significantly modified by fractionation. Metabasic rocks from the Northern Gneisses have slightly incompatible element enriched normalised plots, with many samples displaying negative P and Sr anomalies, and positive Nd anomalies (Fig. 5A). Nb-anomalies are mostly absent, but where present

134 A.J. Mason, T.S. Brewer / Precambrian Research 133 (2004) 121 141 they are slightly negative. Metabasic rocks from the Southern Gneisses show a very similar pattern (Fig. 5B), although they are less enriched in the most incompatible elements with slightly flatter profiles. Samples from the Langavat belt are similar, except that some (about one third) have distinct negative Nb anomalies (Fig. 5C). The samples with the largest Nb anomalies (e.g. AM193) develop primitive mantle normalised patterns similar to those displayed by the grey gneisses and Langavat quartzo feldspathic rocks (Fig. 5D). Consequently, it is probable that those metabasic rocks with prominent Nb anomalies have been significantly contaminated by grey gneiss-like material. 7.3. Tectonic setting A crucial constraint on the setting of the mafic dykes in the Northern and Southern Gneisses is provided by their cross cutting relationships with continental Archaean orthogneiss. In the Langavat Belt too, the dykes cross cut quartzo feldspathic rocks, which probably represent highly deformed equivalents of the grey gneisses. Thus all oceanic settings can be eliminated. Little can be said about the samples from the Harris Granulite Belt samples because of their lack of primary intrusive relationships. To further constrain the tectonic setting, geochemical discrimination diagrams have been employed, the results are summarised in Table 3, and Fig. 6. On four out of the six of these diagrams (see Table 3 for references) the majority of the metabasites from the Northern Gneisses, Southern Gneisses, and Langavat Belt fall in the MORB field. Clearly, in view of the field relationships, these rocks are not MORB. The remaining diagrams suggest an island arc and/or plate margin setting. Samples from the Harris Granulite Belt predominantly plot in fields for arc-related rocks (Table 3). The distribution of the Harris Granulite Belt data on the Zr Ti diagram (Fig. 6) is different to that from the adjacent areas. The lack of a marked Nb anomaly in most samples from the Northern Gneisses, Southern Gneisses, and Langavat Belt suggest these dykes were not generated in a subduction-influenced environment, or from a lithospheric mantle source. By contrast, samples from the Harris Granulite Belt commonly show pronounced Fig. 6. Zr Ti discrimination diagram (Pearce and Cann, 1973) showing metabasite data from the Outer Hebrides. Field A: island arc tholeiites; B: MORB, calc alkali basalt and island arc tholeiites; C: calc alkali basalt; D: MORB. negative Nb anomalies (Fig. 5E). Unlike samples with negative Nb anomalies from the Langavat Belt, the Harris Granulite Belt samples show patterns dissimilar to the tonalitic grey gneisses, and contamination by these rocks cannot be invoked to explain the anomalies. Thus a subduction related setting for the Harris Granulite Belt samples is implied, which is consistent with previous geochemical data from this belt (Baba, 1997). 7.4. Implications for the Outer Hebrides Little regional geochemical variation is seen, and the data from the Langavat Belt, Northern Gneisses and Southern Gneisses are virtually inseparable. Furthermore, the majority of samples fall on a relatively well-defined fractionation trend. Both these facts imply that the majority of these dyke remnants probably belong to a single related swarm, which we collectively term the OH dykes. If this is the case, it implies the Langavat Belt, Northern Gneisses, and Southern Gneisses probably have shared much of their history. This is consistent with protolith ages of ca. 2.8 Ga having been obtained from the Northern and Southern Gneisses (Mason et al., 2004; Whitehouse and Bridgwater, 2001; Friend and Kinny, 2001). Friend and Kinny (2001) have demonstrated that components as old as 3125 Ma are present in the Northern Gneisses, but this in no way precludes the possibility of the

A.J. Mason, T.S. Brewer / Precambrian Research 133 (2004) 121 141 135 Table 3 Results from geochemical discrimination diagrams for data from the Outer Hebrides (highly altered samples excluded) Diagram Field Number of Data LB, NG, SG (n = 81) HGB (n = 18) Zr Ti Y (Pearce and Cann, 1973) Island arc tholeiite 3 4 MORB 65 5 Calc alkaline basalt 3 1 Within-plate basalt 6 2 Not discriminated 4 6 Zr Ti (Pearce and Cann, 1973) Island arc tholeiite 9 6 MORB, calc alkaline 15 7 basalts, and island arc tholeiites Calc alkali basalts 3 1 MORB 37 0 Not discriminated 17 4 Zr Zr/Y (Pearce and Norry, 1979) Volcanic arc basalt 9 6 MORB 26 2 Within-plate basalt 7 1 MORB and volcanic arc basalts 32 6 MORB and within-plate basalt 1 0 Not discriminated 6 3 Ti/Y Zr/Y (Pearce and Gale, 1977) Within-plate basalts 6 3 Plate margin basalts 75 14 (other environments) Not discriminated 0 1 MnO TiO 2 P 2 O 5 (Mullen, 1983) Ocean Island tholeiite 3 0 Ocean island alkali basalt 4 0 MORB 19 3 Island arc tholeiite 48 10 Calc alkaline basalt 7 5 Zr Nb Y (Meschede, 1986) Within-plate alkali basalt 0 0 Within-plate alkali 2 0 basalt and tholeiitic basalt E-MORB 19 1 Within-plate tholeiites 15 4 and volcanic arc basalts N-MORB and volcanic 44 12 arc basalt Not discriminated 1 1 LB, Langavat Belt dykes; NG, Northern Gneiss dykes; SG, Southern Gneiss dykes; HGB, Harris Granulite Belt metabasites. ca. 2.8 Ga components of the Northern and Southern Gneisses being equivalent. The metabasites from the Harris Granulite Belt are distinct from those in the adjacent areas. On South Harris, the change in metabasite chemistry, as well as metamorphic grade appears to coincide with the contact between the Langavat Belt and Harris Granulite Belt, implying a major structural break at this boundary. Conversely, no geochemical evidence is seen for a major structural break between the Northern Gneisses and Langavat Belt; both contain identical dykes. This would suggest that the focus of deformation of the LSZ lies above (to the SW of) the main mass of the Langavat Belt, and is probably associated with the 200 m

136 A.J. Mason, T.S. Brewer / Precambrian Research 133 (2004) 121 141 wide zone of thinly banded mylonites and mylonitic gneisses flanking the metadiorite, and not the train of ultramafic pods (Fig. 3). Consequently, the Langavat Belt on geochemical and metamorphic grounds is more closely allied to the Northern Gneisses than to the Harris Granulite Belt. This conclusion contradicts the terrane model proposed by Friend and Kinny (2001). They treat the Harris Granulite Belt and Langavat Belt as a single entity (their Roineabhal terrane), with the LSZ forming the northeastern terrane boundary, this being based on a similarity of zircon provenance in the Langavat and Leverburgh Belts. However, their sample OH9811 from the Langavat supracrustal Belt used in this comparison does not actually originate from within the Langavat Belt as defined here. Instead, it comes from a thin strip of partially retrogressed granulite facies metasediments within the SHC metadiorite, i.e. within the Harris Granulite Belt (Fig. 1). This strip (Bay Steinigie metasediments, Fig. 1) although traditionally included in the Langavat Belt (e.g. Dearnley, 1963) has never been demonstrated to be equivalent to the main mass of the Langavat Belt. Consequently, the data presented by Friend and Kinny (2001) only demonstrate the probable equivalence of the Leverburgh Belt and Bay Steinigie metasediments, not the equivalence of the Langavat Belt (as defined here) and the Leverburgh Belt. Furthermore, unpublished U Pb TIMS ages obtained by the author (AM) from detrital zircons extracted from a graphitic garnet biotite schist from the main mass of the Langavat Belt, suggest an exclusively late Archaean provenance, as opposed to a dominantly Palaeoproterozoic provenance for the Leverburgh and Bay Steinigie metasediments. This further supports the assertion that the Langavat Belt (as defined here) is not part of the Roineabhal terrane. Evidence for contamination, and field relationships indicates the OH dykes were emplaced into late Archaean continental crust. Furthermore, their geochemistry suggests they are unrelated to the probable arc-related metabasites of the Harris Granulite Belt. The geochemical similarity of the dykes in the Northern and Southern Gneisses implies dyke emplacement while these areas were juxtaposed. Similar dykes are not found within the ca. 1.9 Ga arc-related plutons of the Harris Granulite Belt, suggesting the OH dykes may relate to a pre-arc phase of rifting and intracontinental magmatism. Thus a simple model can be constructed as follows: (1) Formation of local ca. 3.1 Ga grey gneiss components. (2) Formation of the ca. 2.8 Ga grey gneiss protoliths. (3) Mafic magmatism in response to rifting, and ultimately the formation of an ocean of unknown width ( 2.04 Ga). (4) Subduction at one or both margins of this ocean during closure (1.89 Ga early Laxfordian ), resulting in the formation of the Harris Granulite Belt protoliths. (5) Collision and re-assembly during the Laxfordian event (prior to 1675 Ma) of originally contiguous grey gneisses blocks, hence the similarity of dyke chemistry and gneiss protolith age in the Northern and Southern Gneisses. 7.5. Comparison with the Scourie dykes and implications for the Lewisian Four petrographic types of Scourie dyke have been recognised from the Assynt terrane, bronzite picrites, norites, olivine gabbros, and metadolerites (Tarney, 1973), and based on field evidence the majority of the dolerites pre-date the bronzite picrites (Tarney, 1973; Weaver and Tarney, 1981). The first three groups show significant differences to the OH dykes, in that they are typically richer in MgO (see published data of Weaver and Tarney, 1981). In contrast, metadolerites reported from the mainland (Weaver and Tarney, 1981), and Inner Hebrides (Muir et al., 1993) are comparable to the OH dykes, and define similar fractionation trends, but subtle differences are present in the trace element chemistry. Most significantly the Nb/Zr (Fig. 7) and Nd/Nb ratios are slightly lower in the OH dykes. These ratios should not be strongly modified during crystal fractionation, implying the differences are primary. The Zr/Y Nb/Y diagram (Fitton et al., 1997) efficiently separates most of the Scourie and OH dyke data (Fig. 8). The former fall predominantly below the Iceland array, suggesting derivation from a relatively depleted mantle source, while the latter lie mostly on the Iceland array, suggesting derivation from a less depleted mantle source. Furthermore, primitive mantle normalised plots demonstrate that the mainland metadolerites have marked negative Nb anoma-

A.J. Mason, T.S. Brewer / Precambrian Research 133 (2004) 121 141 137 Fig. 7. Bivariate plots of selected trace elements comparing data from the Outer Hebrides (excluding Harris Granulite Belt), Scourie dykes from the Scottish mainland and Inner Hebrides, and the Kangâmiut and MD dykes of Greenland. Data sources for dykes as in Fig. 5. lies (Fig. 5E), a feature not shown by the majority of the samples from the Outer Hebrides (excluding the Harris Granulite Belt). Two possible explanations for the geochemistry can be suggested: (1) The OH dykes and Scourie metadolerites are similar but distinct swarms, the differences being a function of their independent origin. (2) The two dyke swarms are genetically linked but sampled different mantle sources. Available geochronology pertaining to the metadolerite/metagabbro dykes favours the first possibility, since many of the Assynt Terrane metadolerites appear to predate the ca. 2.4 Ga bronzite picrites (Heaman and Tarney, 1989; Tarney, 1973), while the Northern Gneiss dykes were intruded around 2 Ga. Thus, the geochemistry supports the growing body of geochronological evidence (Mason et al., 2004; Whitehouse and Bridgwater, 2001; Friend and Kinny, 2001; Corfu et al., 1994) suggesting that the Outer

138 A.J. Mason, T.S. Brewer / Precambrian Research 133 (2004) 121 141 Fig. 8. Zr/Y Nb/Y diagrams comparing data from the Outer Hebrides (excluding Harris Granulite Belt), Scourie dykes from the Scottish mainland and Inner Hebrides, and the Kangâmiut and MD dykes of Greenland. Data sources for dykes as in Fig. 5. Fields for the Iceland array from Fitton et al. (1997). Hebridean rocks below the OIT are not directly equivalent to those of the adjacent mainland. Therefore, the previous correlation of the OH dykes with the Scourie dykes (Fettes et al., 1992; Dearnley, 1962) is at best dubious. 7.6. Comparison with Greenland The Palaeoproterozoic Nagssugtoqidian orogen of southern Greenland comprises a ca. 200 km wide, approximately E W striking belt of reworked Archaean gneisses, with minor ca. 1.9 Ga juvenile arc terranes (van Gool et al., 2002, and references therein; Bridgwater et al., 1990, Kalsbeek et al., 1993). The recognition of similar arc terranes in the Lewisian Complex (Park et al., 2001; Friend and Kinny, 2001; Whitehouse and Bridgwater, 2001; Mason et al., 2004; Baba, 1997, 1998), and palaeomagnetic data (Buchan et al., 2000), suggests that the Lewisian Complex represents a southeastwards extension of the Nagssugtoqidian orogen, consistent with previous correlations (e.g. Park, 1994). In SW Greenland three main dyke swarms have been identified (Hall et al., 1990, and references therein). Kangâmiut dykes comprise gabbro and dolerite dykes intruded into the Archaean gneisses in the southern part of the Nagssugtoqidian orogen, and adjacent southern foreland, and are affected by Nagssugtoqidian deformation (Escher et al., 1975; Bridgwater et al., 1995; Windley, 1970). MD dykes comprise dolerites mainly intruded into the cratonic Archaean foreland further to the south (Hall et al., 1990). BN dykes comprise more magnesian noritic dykes, also within the southern cratonic block (Hall and Hughes, 1987). Limited published geochemical data is available for the Nagssugtoqidian orogen of SE Greenland (also termed the Ammassalik mobile belt). However, preliminary geochemical data (Hall et al., 1989) suggests the presence of MD and BN dyke equivalents within the southern part of the orogen, while unpublished data (summarised in van Gool et al., 2002) suggests the presence of Kangâmiut dyke equivalents in the northern part of the orogen. It is therefore possible that the OH dykes could correlate with either the MD or Kangâmiut dykes. Compared to published data, the OH dykes, in terms of their major element chemistry are similar to both members of the MD swarm, and the Kangâmiut dykes, but have distinctly lower MgO contents than the BN dykes (Cadman et al., 2001; Hall et al., 1985; Hall and Hughes, 1987). In contrast, significant differences in trace element chemistry occur between the MD and OH dykes. In particular, the former fall below the Iceland array on the Nb/Y Zr/Y diagram, whereas the latter mainly fall on the Iceland array (Fig. 8). In this respect the MD dykes more closely resemble the Scourie dykes of the Scottish mainland. The MD