The mantle of Scotland viewed through the Glen Gollaidh aillikite

The Glen Gollaidh aillikite dyke (58.36741°N 4.69751°W), N.W. Scotland, occurs within the Neoproterozoic sedimentary rocks of the Moine Supergroup ~4 km east of the Moine Thrust. Phlogopite 40Ar/36Ar measurements give a late Devonian maximum emplacement age of 360.3 ± 4.9 (2σ) Ma. This age occurs in a quiet period of Scottish magmatic history c. 30 Ma after the closure of the Iapetus and before the start of intra-plate alkali magmatism which affected southern Scotland for ~60 My from c. 350 Ma. Abundant chromites and Cr-diopsides and a few unaltered olivines, reflecting a mantle provenance, were recovered from heavy mineral concentrates. The North Atlantic Craton, exposed in Lewisian gneisses west of the Moine thrust, is therefore inferred to extend east at depth under Glen Gollaidh, presenting an opportunity to investigate the thickness and composition of the cratonic margin in the Devonian. The aillikite was found to be barren of diamond and no picro-ilmenites or garnets were definitively identified. However, mineral chemistry suggests that a proportion of Glen Gollaidh xenocrysts crystallised in equilibrium with garnet. Most spinels are Mg, Al chromites, with some Mg chromite present. All fall within the garnet peridotite field based on Ti and Cr but with insufficient Cr2O3 (up to 47.2 wt%) to be consistent with the diamond stability field. Amongst Cr-diopsides 30% of grains have Cr and Al contents consistent with derivation from garnet peridotite. The majority of clinopyroxenes also show a marked depletion in heavy compared to light rare-earth elements, again consistent with equilibration with garnet. The opx-cpx solvus thermometer demonstrates that average Cr-diopside compositions require at least 37 kbar to give a temperature (979 °C) lying even on a relatively warm 40 mWm−2 geotherm (Hasterok and Chapman Earth Planet Sc Lett 307:59–70, 2011). Large variations in the chemistry of mantle minerals reflect a complex history of metasomatism akin to constituents of alkali igneous rocks elsewhere in the Hebridean and Northern Highlands Terranes. Fertilised mantle provided the conditions for generation of aillikite melts, probably triggered by break-off of the advancing Avalonia slab. The cratonic root underlying the Glen Gollaidh aillikite during the late Devonian was apparently too thin to lie within the diamond stability field, consistent with xenoliths from alkali basalts further south. Nonetheless, sufficient geophysical and mineral chemical evidence supports Glen Gollaidh aillikite sitting close to the edge of diamond-prospective mantle therefore suggesting diamond potential a short distance to the west within the Lewisian and what is now East Greenland.


Introduction
and Faithfull (2012) describe a carbonaterich olivine macrocrystal dyke in Sutherland, Scotland originally reported as a monchiquite (Read 1931). More recent petrology correctly reclassifies the dyke as an aillikite (Hughes 2012). Aillikites have not otherwise been recorded in the UK and are particularly notable because in some locations they are hosts for diamonds (Hutchison and Frei 2009). As deep-sourced rocks with mantle components aillikites also inform us on the dynamics of the sub-lithospheric mantle which in Scotland likely differs from better exposed rocks sharing a common but more centrally located membership of the North Atlantic Craton (NAC). Here we determine emplacement age and report on the xenocryst component, thus allowing the Glen Gollaidh dyke to provide a time-specific window on the mantle lithosphere comprising the Scottish part of the NAC margin.

Geology
The 1 m wide dyke, named after its location in Glen Gollaidh (also termed Glen Golly), is exposed at 58.36741°N 4.69751°W, to the west of Ben Hope (Fig. 1). Stream-bed outcrops can be seen at two further locations <75 m west northwest along the Allt Strath Feinne-bheinne and a fourth lies 250 m east-southeast along the dyke's~100°strike within the Alltan Molach (Hughes 2012). Glen Gollaidh is situated within the N orthern Highl and Terrane amongst Neoproterozoic rocks of the Moine Supergroup. The Lewisian gneisses, the crustal expression of the margin of the North Atlantic Craton (NAC), occur a short distance (~4 km) to the west in the footwall of the Moine Thrust.
The dyke shows considerable internal variability in mineralogy and texture. Despite this, it is consistently porphyritic with larger grains (>3 mm and up to 8 mm) interpreted to be dominantly xenocrysts and smaller grains a mixture of phenocrysts and broken xenocrysts (Fig. 2a). Primary minerals consist of olivine (fresh and serpentinised), calcite, phlogopite (often longer than 1 mm, and with tetraferriphlogopite rims), clinopyroxene (cpx), apatite, chromite, magnetite, and accessory perovskite, rutile, Ca-Ti ± REE and Zr oxides, barite and pyrite. Spinels and perovskites commonly surround larger serpentinised olivines forming a necklace-like texture (Fig. 2b) and individual spinel grains occasionally show atollshaped textures typical of kimberlites (Fig. 2a;Mitchell 1986). Mineralogy and bulk chemical analyses (Hughes 2012; Supplementary Table 1) classify the rock as an aillikite (after Rock 1986;Tappe et al. 2005). While the presence of primary groundmass clinopyroxene precludes identification as kimberlite, the two rock-types, aillikite and kimberlite, are closely genetically and mineralogically related. Both generate interest for diamond potential and typically tap considerable depths within the mantle lithosphere. Larger phlogopites are commonly bent (Fig. 2b) implying emplacement as a crystal mush. The dyke grades into carbonatite (>50% carbonate by volume) along strike (Hughes 2012) due to fluctuating compositions being fed into the steadily dilating fissure. Similarly intimate aillikite / carbonatite associations have been documented in the NAC in West Greenland (Hutchison and Frei 2009) and at its southern margin in the Gardar assemblage (Upton et al. 2006) where textures are interpreted to have resulted from flowdifferentiation due to the melts being of contrasting ductility (Upton 2013). Bulk rock analysis of the Glen Gollaidh aillikite shows no strong compositional gap along the gradation between aillikite and carbonatite (Supplementary Table 1). Calcite veinlets up to 5 mm thick are particularly common at the northwestern extent of the dyke.
Xenocrysts and clasts consist of olivines and orthopyroxenes (almost exclusively pseudomorphed) and abundant serpentinised spinel lherzolite xenoliths (≤3 cm; Fig. 2c) and reflect a mantle component. Titanian and magnesian magnetites form pronounced reaction rims around isolated grains of brown spinel (Fig. 2a), indicating clearly that spinel-lherzolites have been disaggregated and reacted with the melt. Akin to other aillikites (Hutchison and Frei 2009), smaller olivine grains are expected to be a combination of phenocrysts and broken xenocrysts, emphasising the importance of mineral petrology rather than bulk chemical measurement in determination of the rock type. Weathered surfaces show the very competent, sandpaper-textured orange-brown knobbly appearance (Fig. 2d) ubiquitous in NAC aillikites (Hutchison and Frei 2009).

Samples and methodology
Hand samples were collected for the purposes of determination of emplacement age and characterisation of mantle minerals by means of disaggregation and mineral separation. Weathered surfaces were removed before further sample processing was carried out.  8°W   6°W  4°W  2°W  0°0 km 100

Age determination
Phlogopite was separated from the freshest and coarsest mica-rich material collected, a block of interbanded macrocrystal and carbonate-rich aillikite (sample 43A4). Each analysed aliquot represents 10 grains of phlogopite with an average diameter between 250 and 500 μm. Samples and neutron flux monitors were packaged in copper foil and stacked in quartz tubes with the relative positions of packets precisely measured for later reconstruction of neutron flux gradients.  Photomicrograph under plain polarised light showing deformed phlogopites (pl; 0.3 mm brown platy minerals) on the right edge. One large (0.6 mm) olivine macrocryst (ol) lies off centre to the right and another, highly serpentinised (serp) is surrounded by a necklace-texture of opaque chromites and perovskite (square cross-sections). c Hand sample of freshly exposed surface showing abundant carbonated olivine xenocrysts and spinel lherzolite xenolith with carbonate veining and segregation textures. The thickness (<5 mm) of the weathered layer is apparent in section. d Hand sample of distinctive orange-brown, nodular and sandpaper-textured weathered aillikite. The appearance is characteristic of aillikite and kimberlite samples in the field in the North Atlantic Craton (Hutchison and Frei 2009) isochron, is indistinguishable from the composition of air at the 2σ uncertainty level. 7) Age and uncertainty (SEM) were calculated using the mean weighted by the inverse variance of each step. 8) For samples with MSWD >95%, the uncertainty is expanded by the square root of MSWD and Student's t. Analytical data and results are presented in Table 1.
Sample processing for mantle mineral chemistry A 9.8 kg aillikite sample was disaggregated to <1 mm grain size in two batches by high-voltage pulse fragmentation with all fines retained. All 6.05 kg of 0.25-1.0 mm grains, and 500 g of 0.18-0.25 mm grains were liquid density separated (S.G. 2.96) with a purpose-designed flowsheet built to retain any microdiamonds. Mantle-derived minerals were identified by optical microscopy, mounted in epoxy and polished to 0.25 μm for mineral chemical analyses. Subsequently 38 g of sample was archived and the total remaining sample was treated by caustic fusion for identification of diamond down to a threshold of 75 μm.

Mineral chemical determination
All analysed clinopyroxenes and olivines derive from the >250 μm size fraction whereas analysed spinels derive from both 180-250 μm and > 250 μm fractions. Major element chemistry was determined by Cameca SX100 electron probe microanalyser (EPMA; 15 keV, 10 and 100 nA, beam-size 5 μm). All elemental concentrations were determined from measurement of Kα lines with 20 to 60 s counting times on peaks and 10 to 60 s on background positions. Analyses were calibrated using internal well-characterised standards as follows: Al and Mg -Spinel-BL8; Ca and Si -Wollastonite-BL7; Cr -PuCr-BL7; Fe -Fayalite; Mn -PuMn-BL8; Na -Jadeite-BL7; Ni -PuNi-BL8; Ti -Rutile-BL8. Matrix corrections were made by the proprietary PAP software supplied by the manufacturer of the instrument. Spinels were classified by major element chemistry subdivision according to criteria modified from Ramsay (1992) and described in Hutchison (2018).
Determination of accurate concentrations of twenty-seven trace elements were made by laser ablation inductively coupled mass spectrometry (La-ICP-MS) using a Hewlett Packard Agilent 7700 mass spectrometer. The instrument employs an Excimer 193 nm laser and samples and standards were mounted in a custom built, dual volume sample cell. Instrument and data deconvolution methodologies employed follow Norman et al. (1996).
Olivine trace element analyses were performed with a 5 Hz laser repetition rate whereas clinopyroxenes and chromites were analysed at 10 Hz. Olivine and clinopyroxene grains were large enough to be analysed with a beam size of 137 μm. For the smaller chromite grains all 180-250 μm grains were analysed with a 47 μm beamsize whereas the  . Silicon (measured on mass 29) was used as an internal standard for olivine analyses, Ca (mass 43) for clinopyroxene and Al (mass 27) for chromite, based on EPMA analyses. Standardisation was achieved against NIST612 glass (Hollocher and Ruiz 1995) and with further quality assurance against garnet (Mongolian garnet MU53388; Norman et al. 1996) for clinopyroxene analyses and metallic CAN-MET (unpublished data from Jung-Woo Park, Canberra; written communication) and Ellendale chromite (P2; unpublished data from Wayne R Taylor, Perth; written communication) standards for chromite. Due to 53 Cr 40 Ar interference during analyses of chromites, measurement of 93 Nb, and hence Nb concentrations, are subject to an unquantified inaccuracy of up to~33% over the variance in instrument precision (2σ in Table 2). This value derives from the difference between the average ratio of Nb to Ta in modelled mantle reservoirs (18; Sun and McDonough 1989) and the measured ratios of Nb to Ta (median of 24). It is unclear whether further fractionation between Nb and Ta has a negative influence on this ratio. However, the uncertainty is likely an over-estimate as Dare et al. (2014) report a gas flow dependent 53 Cr 40 Ar interference of 9.6%. In any case, Nb values should be considered as maximums thus Ga/Nb ratios are minimums. Processed data for representative grains are presented in Table 2 with analyses of the NIST612 standards (Supplementary Table 2) and natural reference minerals (Supplementary Table 3) lodged in the supplementary data appendix. Analyses of standard materials showed good reproducibility and similarity to independently-derived data. Sinusoidal grain G13-6_75 is a small grain showing a porous or included texture on polishing, otherwise all grains featured in Table 2 presented a homogeneous texture and even polish.

Age determination
Sample characterisation is key to interpretation of phlogopite Ar/Ar data (Phillips 2012) and to that end the identification of analysed phlogopites as largely phenocrystic grains warrants caution in their age interpretation. Five reproducible stepheatings of phlogopite multi-grain aliquots yield a late Devonian composite plateau age of 360.3 ± 4.9 (2σ) Ma ( Fig. 3; Table 1). Due to the excessive scatter in the data, within and between aliquots, MSWD is high (3.5) and p is low. To account for this scatter, the uncertainty is expanded by the square root of MSWD and by Student's t, providing a more realistic measure of the age uncertainty. The relatively flat release spectra and level of reproducibility between aliquots suggest that the age records a simple rapid cooling event consistent with volcanic emplacement. The phlogopite argon data show no evidence for significant excess argon in the inverse isochron correlation diagrams. This is inconclusive however, since the points on the correlation diagrams are nearly all 100% radiogenic, i.e., little dispersal on the isochron and therefore the trapped component composition tends to be poorly defined. As phlogopite is notorious for containing excess argon (Phillips and Onstott 1988;Kelley and Wartho 2000), especially in phenocrysts and xenocrysts, a conservative interpretation would be that the 360 Ma age for these samples represents a maximum age constraint. Two of the release spectra (aliquots 2 and 5) show hints of a 'saddle' shape, potentially indicative of excess argon. Removal of these two aliquots from the composite yields an age of 360.7 ± 3.7 (2σ) Ma, indistguishable from the accepted age above, suggesting that if excess argon is present in some of the grains, it's effects are subtle and do not resolvably impact the age estimate. This leads us to suggest that the 360.3 Ma age reflects an emplacement age rather than a maximum age. The flatness of the age spectra also argue against reheating events at or below the closure temperature. If the deposit was reheated (post emplacement) at temperatures above the closure temperature, ca. 400-480°C, for a sufficient time, then the argon system could have been partially to completely reset. However, such an event would be recognizable in the petrology of the sample and it's setting, e.g., low to mid-grade metamorphic facies host rocks, for which there is no evidence.

Mantle mineral recovery
Inspection of heavy mineral concentrates recovered abundant chromites, Cr-diopsides (in both size fractions) and four olivine grains (>0.25 mm). No perovskite, picro-ilmenite or garnet was identified by this method (although a few 1-3 mm rounded brownish-red grains of what may be altered garnet occur on the weathered surface of one hand sample). The paucity of recovered olivine is interpreted to be due to serpentinisation, and recrystallisation of xenocrysts to fine grain sizes which disaggregate on sample preparation. While all fifteen qualitycontrol tracer diamonds were recovered during caustic fusion treatment, the aillikite was found to be barren of diamond to the minimum size of 75 μm tested.

Mineral chemistry
Olivine Olivines are Mg-rich (Fo 91 -Fo 89 ) with three of four grains over Fo 90 , typical of mantle-derived olivine. Ni contents are also high, in excess of 3000 ppm (Table 2), consistent with aillikites, and in particular Type Ia kimberlites from the NAC in West Greenland (Garnet Lake; Hutchison and Frei 2009). The mantle of Scotland viewed through the Glen Gollaidh aillikite S121     Nb (93)   The mantle of Scotland viewed through the Glen Gollaidh aillikite S123

Clinopyroxene
Clinopyroxene compositions range from 2.6-6.6 wt% Al 2 O 3 and 19.9-23.9 wt% CaO encompassing, at one place, most of the 'remarkable range' (Upton et al. 2011) exhibited by Scottish mantle-derived clinopyroxenes compared to world-wide xenolith localities. Based on Cr and Al content (Ramsay and Tompkins 1994), the majority of Cr-diopsides are consistent with spinel lherzolite compositions (CLS) although a significant proportion (30% of grains) also plots in the garnet peridotite (CGP) field with Cr 2 O 3 content up to 1.56 wt% (Fig. 4) Ramsay and Tompkins (1994). Rare-earth element (REE) trends refer to characteristics shown in Fig. 5 Al 2 O 3 depletion (<2.5 wt%) evident in samples derived from well within the diamond stability field (Hutchison and Frei 2009). However, the compositional range is similar to Loch Roag (Menzies et al. 1989;Upton et al. 1999). Loch Roag, further localities in the Northern Highlands Terrane (Upton et al. 1999) and Elie Ness in the Midland Valley Terrane (Upton et al. 1999) all include eclogitic / megacrystic / cognate clinopyroxenes which are not represented at Glen Gollaidh. Cr-diopsides express four distinct rare-earth element (REE) trends. 1) Of 41 grains, the most common trend (63% of grains) shows a light rare-earth elements (LREE) enrichment up to 50 times chondritic concentrations, to chondritic values amongst the heavy rare-earth elements (HREE) (Fig. 5). This trend mirrors the composition of CGP from the Garnet Lake aillikite (Hutchison and Frei 2009) albeit with slightly smaller Nd/Yb ratios (Fig. 5). The LREE-enriched trend also mirrorsalthough with lower absolute REE concentrationseclogitic, megacrystic or cognate (CPP) clinopyroxene from further north in the Northern Highland Terrane at Rinibar (Bonadiman et al. 2008;Upton et al. 2011), further south at Gribun, and at Loch Roag, Isle of Lewis (Hebridean Terrance; Upton et al. 1999 (Fig. 4). The HREE-enriched trend can be further subdivided based on elevated or depleted chondritenormalised La and Ce compared to Pr. It is notable ( Fig. 4; filled black symbols) that the La and Ce depleted samples lie furthest into the spinel peridotite field. No similar clinopyroxene HREE-enriched distributions like those at Glen Gollaidh are seen in other Scottish mantle samplebearing rocks. 3) Five grains (12% of all grains; all with CLS compositions) show a flat REE trend (Fig. 5) consistently close to 10 times chondritic concentrations matching trends from the Midland Valley Terrane at Fidra, and in particular at Hawke's Nib (Downes et al. 2007) and at Streap Com'laidh in the Northern Highland Terrane (Upton et al. 2011). 4) A single CGP outlier (G13-6 75) shows a sinuous trend -LREE enriched and with a minimum at Dy (0.49 ppm)resembling the average clinopyroxenes from Rinibar (Hughes et al. 2015). Cryptic melt infiltration into mantle minerals has been reported from kimberlites (Aulbach and Viljoen 2015) and can have an influence on trace element chemistry by raising concentrations, particularly for incompatible elements. Similar infiltration into mantle minerals may in theory have occurred in Glen Gollaidh aillikites. High values for elements such as Ba, Nb and Rb in some analyses could be influenced by the presence of microscopic melt inclusions and the unusual sinuous characteristic of the single CGP grain outlier draws particular attention. However, Glen Gollaidh clinopyroxenes exhibit similar or lower REE concentrations compared to grains from other Scottish localities (Fig. 5). Furthermore, for all analyses (with the exception of the sinuous REE-pattern grain G13-6 75) Ba, Rb and Nb concentrations are consistent with published values from other localities, particularly for clinopyroxenes from Fidra (Upton et al. 1999) and Loch Roag (Hughes et al. 2015). In order to assess the effect of infiltration, Aulbach and Viljoen (2015) compare melt / clinopyroxene partition coefficients with compositions of bulk kimberlite and clinopyroxene grains. In the case of Glen Gollaidh, the large clinopyroxene grains are xenocrysts and thus aillikite / clinopyroxene partition coefficients are of little relevance. However, bulk compositional data provide knowledge of the composition of possible infiltrating fluid and are thus provided in Supplementary Table 1. Bulk rock compositions of ten Glen Gollaidh aillikite samples from three localities provide consistent REE compositions evenly declining from La at 500-times to Yb at 10-times CI chondrite concentrations. REE concentrations coincide with the more light REE-enriched samples from the Tugtutôq-Nûgârmiut melaaillikites at the Greenlandic margin of the NAC in the Gardar province (Upton et al. 2006). No combination of any Glen Gollaidh bulk chemical data with any other clinopyroxene type can reproduce the highly depleted mid-heavy REE trend seen in the sinuous clinopyroxene grain G13-6 75. While rare at Glen Gollaidh, given that similar compositions are common at Rinibar (including for incompatible elements Nb and Rb; Hughes et al. 2015) its composition is concluded to be genuine.
Orthopyroxene has been identified optically in lherzolite xenoliths but compositions have not been determined. In the absence of orthopyroxene recovery from mineral separates, calculations using the opx-cpx solvus thermometer (Bertrand and Mercier 1985) have been made assuming that equilibrium orthopyroxene is end-member MgSiO 3 enstatiteproviding minimum temperature estimates. Results show (Fig. 6) that Crdiopside compositions require at least 19 kbar (59 km) to give an average minimum temperature of 961°C. Given the spread of data, this is the lowest pressure at which the temperature was cool enough for some clinopyroxene compositions to coincide with a 40 mWm −2 geotherm (Hasterok and Chapman 2011). Grütter (2009) (Grütter 2009) and diamond (C d ) / graphite (C g ) stability (Kennedy and Kennedy 1976). Temperature calculations are modelled on 19, 27 and 37 kbar following the methodology of Bertrand and Mercier (1985). Isobaric temperatures calculated for individual grains are represented by crosses, averages for each pressure increment are shown by solid squares.
As equilibrium with end-member enstatite is assumed, all temperature estimates are minimums. Depth calculations are based on Dziewonski and Anderson (1981). Temperatures along the preferred 40 mWm −2 geotherm of Hasterok and Chapman (2011) Grütter et al. 2006). Identification of unaltered orthopyroxene in equilibrium with clinopyroxene in mantle xenoliths would allow more robust geothermometry. However, it can be confidently stated that 19 kbar is the minimum equilibrium pressure for any measured Glen Gollaidh Cr-diopside. An average Crdiopside (with composition equivalent to a temperature of 979°C) requires a pressure of 37 kbar (Fig. 6), equivalent to conditions at 110 km in depth (Dziewonski and Anderson 1981), to lie even on a locally warm (40 mWm −2 ) geotherm. Thus geothermometry places typical mantle components from the dyke close to the spinel-garnet transition (although average Cr-diopsides likely derive from firmly within the graphite stability field). This observation is consistent with independent conclusions based on clinopyroxene (and chromite) chemistry. Furthermore, and assuming a mantle geotherm of 40 mWm −2 , an iterative approach to calculating pressures and temperatures using the Bertrand and Mercier (1985) opx-cpx solvus thermometer also shows the Glen Gollaidh clinopyroxenes to have derived from a range of depths crossing the spinel to garnet lherzolite transition and with some rare grains extending into the diamond stability field (Fig. 6).

Chromite
All analysed spinels are chromites. Of 45 grains analysed for major element chemistry, a single Mg chromite (which dominate the population at Clare Island, Eire; Upton et al. 2001), two Ti-Fe-Mg-Al chromites and one Ti-Mg-Al chromite were identified (following the classification scheme described in Hutchison 2011). The remaining, and large majority of grains, are Mg-Al chromites. Ubiquitous Mg-Al chromite is otherwise seen further west at Loch Roag and north-east at Rinibar (Upton et al. 2011). In contrast, further south in Scotland, alkaline igneous rock-hosted spinel xenocrysts are dominated by Al spinel (Upton et al. 2011). One Mg-Al chromite was identified with a Ti-Mg-Al chromite composition rim and another has a Ti-Fe-Mg-Al chromite rim. This is the same trend described in the NAC at Garnet Lake (Hutchison and Frei 2009) and is consistent with cognate groundmass chromite crystallisation on xenocrystal seed grains. All Glen Gollaidh chromite compositions coincide with chromites in kimberlites and within the garnet peridotite field (based on Ti and Cr content; Grütter and Apter 1998). All chromites also describe the spinel Magmatic Trend 2 (Mitchell 1986) based on divalent / trivalent cations. Compositions are therefore consistent with chromites from world-wide kimberlites and lamproites and in particular spinel xenocrysts from the Garnet Lake aillikite (Hutchison and Frei 2009). In no cases are the Cr contents (<47.2 wt% Cr 2 O 3 ) abundant enough to be consistent with the diamond stability field (following the method of Grütter and Apter 1998). Major element chemistry therefore shows the Glen Gollaidh chromites to be consistent with mantle xenoliths and similar to aillikites and kimberlite with a shared geological history elsewhere on the NAC.
Trace elements are particularly useful in assessment of the petrological affinities of xenocrystal chromites (in addition to those from unknown sources such as are encountered in mineral exploration). Nb stands out as being the most valuable element for this purpose. The large uncertainty (~33%) in Nb compositions may appear to pose a significant problem in the utility of this element. However, discrimination diagrams use log 10 ratios and hence are resistant to uncertainties of this magnitude. Figure 7 shows Glen Gollaidh chromites plotted in terms of log 10 (Ga/Nb) against log 10 (Co/V). Data are compared to the mantle field of Yaxley (2008; developed by Taylor WR, Australian National University's Research School of Earth Sciences). Compositions from regional samples from the North Australian Craton (Hutchison 2011) are also included. This data-set has been chosen due to a paucity of NAC chromite trace element data and because the North Australian Craton data express the range of compositions expected from a diamondiferous craton with chromite from a range of crustal and mantle sources (akin to the NAC). Given the high uncertainty of Nb concentrations, as a quality assessment, the same grain analyses have been plotted with Nb replaced with concentrations of Ta × 18. The same distribution of grains between the mantle field and magmatic array were identified. Figure 7 shows that the Mg chromite grain lies clearly within the mantle array compared to the Mg-Al chromites which describe a spread towards shallower magmatic compositions with the range in compositions likely reflecting both a range of depths of origin and the influence of mantle metasomatism (Hughes et al. 2015). The Ti-bearing chromites (whole grains and rims) lie within the mantle array as a consequence of elevated Nb (also exhibiting similarly elevated Ta).

Discussion
The Glen Gollaidh aillikite provides the opportunity to address where the locality lies in the context of the margin of the NAC mantle lithosphere. The aillikite also gives insights into the mineralogy of the NACincluding diamond prospectivityand hence its prior geological history.

Glen Gollaidh's place at the NAC margin
The Glen Gollaidh outcrop lies amongst the Neoproterozoic sedimentary rocks of the Moine Supergroup, the hanging-wall of the Moine Thrust which is located~4 km to the west (Fig. 1). However, the Archean and Proterozoic rocks of the Lewisian component of the NAC are exposed to the west of the Moine Thrust. The Lewisian rocks comprise part of the reworked NAC mobile belt, most of which is preserved to the north-west in the previously adjoining Greenlandic component, there termed the Nagssugtoqidian Orogen. Hughes et al. (2015) argued that the NAC lithosphere keel is evidenced by mantle xenoliths at Loch Roag, Isle of Lewis. Evidence for an unusually deep mantle source in the context of other Scottish localities is apparent at Glen Gollaidh, and some criteria draw closer comparisons with NAC mantle sources in Greenland. West Greenland comprises what were more central parts of the NAC (as evidenced by deep-sourced peridotitic xenoliths and diamonds; Frei 2009, Wittig et al. 2010). The northern part of the West Greenland NAC is particularly characterised by depleted mantle at shallow depths where clinopyroxene is rare (Sarfartoq, West Greenland, Bizzarro and Stevenson 2003;Ubekendt Ejland, West Greenland, Bernstein et al. 2006). Depletion is due to considerable melt extraction. However, deeper parts of the northern West Greenland NAC and rocks to the south, previously geographically closer to Glen Gollaidh, have seen less melt extraction and, like at Glen Gollaidh, are clinopyroxenebearing. The clinopyroxenes from the Garnet Lake aillikite (Sarfartoq, West Greenland NAC) occur with garnet and derive from 195 km depth (1258°C and 62 kbar assuming a 41 mWm −2 geotherm; Hutchison and Frei 2009). Chondrite normalised rare-earth patterns from these samples (Fig. 5) best match the dominant Glen Gollaidh clinopyroxene LREEenriched trend, the closest other Scottish analogue being at Loch Roag (Upton et al. 1999). Mantle components in the Glen Gollaidh aillikite are not as deep-sourced as at Garnet Lake, hence during the late Devonian the cratonic root underlying this part of Scotland was apparently too thin to lie within the diamond stability field. However, mineral chemistry still places the dyke as hosting some of the deepest mantle components recovered in the UK. Minimum pressure estimates for Glen Gollaidh are 22 kbar but more likely greater depths have been sampled, equivalent to 37 kbar and more, consistent with the range in compositions of clinopyroxenes and the (inconclusive) identification of garnet grains. The proximity of Glen Gollaidh to the Moine Thrust sheet and the considerable mantle lithosphere thickness sampled at Glen Gollaidh suggests that the NAC extends at depth to the east of its Lewisian surface expression. A similar conclusion is made by Hughes et al. (2015) for Rinibar, Orkney (Fig. 1). Hughes et al. (2015) also comment that the underlying keel of NAC is not, however, present~175 km south of Glen Gollaidh at Streap Com'laidh. Estimates for equilibrium conditions for Streap Com'laidh peridotites are 1100-1200°C at 14-23 kbar (Praegel 1981). A thinning of the NAC margin south and east is also borne out in the dominance of various mantle mineral classes. Anomalous samples occur as garnets reported in Elie Ness, Fife and in Dunaskin Glen and the Heads of Ayr (Upton et al. 2003) with inferred pressures up to 20 kbar. However, more typically, there is also a progression from dominant, deep-sourced Mg chromite (Clare Island; Upton et al. 2001), to Mg-Al chromite (Loch Roag, Upton et al. 2011; with minor Mg chromite, Glen Gollaidh), to Mg-Al chromite with Mg-Cr-Al spinel (Rinibar) to shallow-sourced Mg-Cr-Al spinel (off-craton at Streap Com'laidh and Fidra). Mantle field and magmatic array from Yaxley (2008). Glen Gollaidh Ga/Nb values are minimum ratios as Nb data are maximum values. Chromites from regional surface-sediment samples from Northern Territory, Australia are shown to demonstrate the range of compositions expected from a diamond-bearing cratonic region with both kimberlite-hosted and crustal-derived (basaltic) chromites While the variability in chemistry of the Glen Gollaidh mantle component has a partial explanation in a range of sampling depths, the majority of the variability is explained by a complex history of metasomatism and partial melting; a consequence of the upheavals associated with the closure of the Iapetus and accumulation of Laurussia (Macdonald and Fettes 2007). It is concluded that this same complex mantle chemistry created suitable conditions to allow melting of the mantle below Glen Gollaidh at this late stage in the deformations of the margins of the NAC. Furthermore, it appears that at the end of the Devonian, a global change in the physical properties of continental lithosphere was beginning to occur. In the UK and Eire, basaltic-hosted mantle xenoliths started to occur early compared to world-wide where they are much more prevalent from the Cenozoic. The Glen Gollaidh and Clare Island (Upton et al. 2001) volcanics are the last mantle xenolith-hosting melts derived from small, volatile-rich melt-fractions before a change to a proliferation of large melt fraction (basaltic) melts. It is proposed that this point in the Devonian represents a time of significant change in the thickness, temperature or mechanical properties of continental lithosphere laterally allowing the large scale of fracturing required for penetration of basaltic melts.

Diamond prospectivity
The characteristics of the Glen Gollaidh aillikites of absence of definitively identified garnet grains, some relatively low Cr-content chromite and some spinel lherzolite Cr-diopside compositions are consistent with the failure of the sample tested to yield diamonds. The small sample size tested does not definitively prove that the body is non-diamondiferous but the negative result for both garnet and diamond makes the commercial prospects of the dyke itself very unlikely. While not derived from mantle thick enough to lie within the diamond stability field at this point, the Glen Gollaidh aillikite provides evidence for a craton margin with enough inferred thickness to suggest that it sits at the edge of diamondprospective mantle lithosphere. Sufficient cratonic keel thickness to lie within the diamond stability field may occur only a short distance to the west. This may occur under the Lewisian west at Clare Island, Eire, as evidenced by spinel populations dominated by Mg chromite over Mg-Al chromite (Upton et al. 2001; the converse being the case at Glen Gollaidh). Certainly, if not under the Lewisian, then most likely the NAC keel lies within the diamond stability field under East Greenland. Currently East Greenland is not known as a diamond-bearing province, unlike its counterpart on West Greenland (e.g. Hutchison and Frei 2009) but it is very much under-explored.
In conclusion, study of the Glen Gollaidh aillikite considerably advances understanding of the limits of diamond prospective areas at the peripheries of the NAC, likely a relatively short distance to the north and west. Magmatism in the late Devonian was likely initiated as part of the dying stages of subduction of Avalonia below Laurentia as a consequence of slab break-off. Furthermore, the considerable variability of mantle geochemistry at Glen Gollaidh represents a microcosm of North Atlantic Craton evolution evidencing a complex multi-component system.