Platinum-group element signatures in the North Atlantic Igneous Province: Implications for mantle controls on metal budgets during continental breakup

The North Atlantic Igneous Province (NAIP) is a large igneous province (LIP) that includes a series of lava suites eruptedfromtheearliestmanifestations ofthe(proto)-Icelandicplume,throughcontinentalriftingandultimate ocean opening. The lavas of one of these sub-provinces, the British Palaeogene Igneous Province (BPIP), were someofthe ﬁ rstlavastobeeruptedintheNAIPandoverlieathickcrustalbasementandsedimentarysuccession with abundant S-rich mudrocks. We present the ﬁ rst platinum-group element (PGE) and Au analyses of BPIP ﬂ ood basalts from the main lava ﬁ elds of the Isle of Mull and Morvern and the Isle of Skye, in addition to a suite of shallow crustal dolerite volcanic plugs on Mull, and other minor lavas suites. BPIP lavas display both S-saturated and S-undersaturated trends which, coupled with elevated PGE abundances ( N MORB), suggest that the BPIP is one of the most prospective areas of the NAIP to host Ni – Cu – PGE – (Au) mineralisation in conduit systems. Platinum-group element, Au and chalcophile element abundances in lavas from West and East Greenland, and Iceland, are directly comparable to BPIP lavas, but the relative abundances of Pt and Pd vary systematically between lavas suites of different ages. The oldest lavas (BPIP and West Greenland) have a broadly chondritic Pt/Pd ratio (~1.9). Lavas from East Greenland have a lower Pt/Pd ratio (~0.8) and the youngest lavas from Iceland have the lowest Pt/Pd ratio of the NAIP (~0.4). Hence, Pt/Pd ratio of otherwise equivalent ﬂ ood basalt lavas varies temporally across the NAIP and appears to be coincident with the changing geodynamic environment of the (proto)-Icelandic plume through time. We assess the possible causes for such systematic Pt/Pd variation inlightofmantleplumeandlithosphericcontrols,andsuggestthatthis re ﬂ ectsachangeintheavailability of lithospheric mantle Pt-rich sulphides for entrainment in ascending plume magmas. Hence the precious metal systematics and potential prospectivity of a LIP may be affected by contamination of plume-derived magmas by subcontinental lithospheric mantle at the margins of cratons that have been enriched by Palaeoproterozoic orogenesis. © 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).


Introduction
Some of the world's most important orthomagmatic Ni-Cu-PGE sulphide deposits reside in continental flood basalt provinces, or large igneous provinces (LIPs)-for example the Norilsk-Talnakh conduit-hosted mineralisation in the Siberian Traps (e.g., Arndt, 2011). Many factors contribute to the formation of such deposits, not the least the timing of S-saturation and segregation of immiscible sulphide liquids, triggered by various processes including contamination of magmas by crustal S-rich sediments (e.g., Ripley et al., 2003). Sulphursaturation may also be achieved by magma mixing or fractional crystallisation of ascending magmas (e.g., Irvine, 1975;Naldrett, 2011 and references therein). Once formed, immiscible sulphide liquids scavenge chalcophile elements from repeated influxes of silicate magma, and if collected, may form sulphide deposits and mineralisation within intrusions such as mid-to upper-crustal magma conduits (e.g., Naldrett, 2004Naldrett, , 2011. The North Atlantic Igneous Province (NAIP) is a LIP and certain portions are documented to host Au-Cu-PGE mineralisation, such as the Skaergaard Intrusion, Kap Edvard Holm Complex, and associated Macrodykes in East Greenland (Bird et al., 1991(Bird et al., , 1995Andersen et al., 1998;Arnason and Bird, 2000;Thomassen and Nielsen, 2006;Holwell et al., 2012). Continental flood basalts in West Greenland appear to have undergone assimilation-induced S-saturation, highlighted as potentially overlying significant orthomagmatic PGE mineralisation, Lithos 233 (2015)   Abbreviations: NAIP, North Atlantic Igneous Province; BPIP, British Palaeogene Igneous Province; PGE, platinum-group elements; REE, rare earth elements; NAC, North Atlantic Craton; MSS, monosulphide solution; SDRS, Seaward Dipping Reflector Series; LIP, large igneous province. and are currently under exploration (e.g., Keays and Lightfoot, 2007 and references therein). In contrast, onshore and offshore continental flood basalts in this region, along with basaltic lavas of the seaward dipping reflectors offshore of the Southeast Greenlandic coast, are predominantly S-undersaturated (Philipp et al., 2001;Momme et al., 2002). However, minor S-saturation has taken place in some flows here, following silicate magma differentiation and fractionation (Philipp et al., 2001).
Aside from S-saturation, the initial concentration of PGE, Ni and Cu in mantle-derived magmas is a key factor in the probability of their forming orthomagmatic mineralisation (e.g., Naldrett, 2004). Worldwide, such mineralisation appears to be correlated with the occurrence of mantle plume events impinging on thick cratonic (Archaean) lithospheric mantle (e.g., Groves and Bierlein, 2007;Maier and Groves, 2011 and references therein). The debate over the cause for this association remains unresolved (e.g., Arndt, 2013). This may imply some dependence on there being high-degree melting and/or deeper mantle (or even mantle-core) melting associated with mantle plumes, and hence important geochemical/geophysical constraints on mineralisation potential specific to plumes. Alternatively the association may suggest enrichment of PGE in cratonic subcontinental lithospheric mantle (SCLM) which could become incorporated in ascending LIP magmas. It may also reflect a biased preservation and/or discovery potential (e.g., Maier and Groves, 2011).
The magmatic centres and plateau lavas of the British Palaeogene Igneous Province (BPIP) formed prior to continental breakup and at a similar time to continental flood basalts in West Greenland (Storey et al., 2007 and references therein). However, no significant Ni-Cu-PGE exploration has taken place in the BPIP to date, despite favourable circumstances for crustal S contamination in this region (see Andersen et al., 2002). In this paper we report new platinum-group element (PGE) and Au analyses for a series of BPIP basaltic lavas to establish their S-saturation status, and to assess the fertility and prospectivity of the Eastern part of the BPIP.
PGE abundances in various Greenlandic and Icelandic lava suites will be compared with our new BPIP lava dataset in order to assess the implications for mineralisation potential and metal ratios, from early plume impingement beneath the continental lithosphere, throughout continental breakup and ultimately during oceanic rifting (± a nonplume mid-ocean ridge melting component).

Geology and formation of the North Atlantic Igneous Province (NAIP)
Impingement of the proto-Icelandic mantle plume under the continental lithosphere of Greenland, Canada and Europe (i.e., beneath the North Atlantic Craton; NAC) generatedtheNAIP.Priortothis,Greenland andEuropewerejoinedandseafloor spreading had previously initiated in the South Labrador Sea between Canada and Greenland (ca. 80 Ma; Roest and Srivistava, 1989) during a period of slow extension apparently devoid of extensive volcanism (Chambers and Pringle, 2001). However, ca. 62 Ma, significant volumes of mafic magmatism below what is now the UK, Greenland and BaffinIsland (Fig. 1), signified a dramatic change in thermal mantle conditions, which predated continental rifting (c. 55 Ma) but ultimately led to the opening of the Atlantic Ocean (Saunders et al., 1997). The main products of this prolonged period of magmatism were alkali and tholeiitic flood basalts. In total, Palaeocene and Eocene magmas of the NAIP likely ranged from 5 to 10 × 10 6 km 3 in volume (White and McKenzie, 1989;Holbrook et al., 2001 and references therein). The NAIP can be divided into several distinct igneous temporal intervals, as outlined below (Saunders et al., 1997;Chambers and Pringle, 2001;Jolley and Bell, 2002;Storey et al., 2007)-see Fig. 2.
Continent-based magmatism and stretching during impingement of the proto-Icelandic mantle plume : 1 Phase 1-The onset of widespread and voluminous volcanism initially in West Greenland and Baffin Island at 61.7 ± 0.5 Ma (Sinton et al., 1998) comprising basaltic and picritic lavas up to~7 km thick (Storey et al., 2007), and alkali basaltic lavas in the Faroe-Shetland Basin and Scottish (Hebridean) portion of the BPIP. At the end of this interval there was a switch to localised, and increasingly intermittent, volcanic activity (ca. 57-56 Ma). 2 Phase 2-Eruption of widespread volcanic sequences due to an increase in melt production rate at 56.1 ± 0.5 Ma which coincided with continental rifting (Storey et al., 2007). Tholeiitic basaltic sequences were erupted in West and East Greenland (e.g., Blosseville Kyst, Kangerlussuaq, and Hold with Hope regions), the Irmiger Basin, Faroe Islands, Vøring Plateau, Faroe-Shetland Basin, Rockall Basin, and the Inner Hebrides and Antrim (BPIP).
Syn-and post-breakup, plate separation at the Mid-Atlantic Ridge (56 Ma to present): 3 Phase 3-Voluminous volcanism continued, but was restricted to the margins of the proto-North Atlantic rift, forming abnormally thick oceanic crust offshore and along the line of plate separation, ca. 56-50 Ma (Philipp et al., 2001;Storey et al., 2007 and references therein). Basalts from this period are observed in the Irmiger Basin, Faroe Islands, East Greenland margin, and the Vøring Plateau and are seismically imaged as seaward-dipping reflector sequences. 4 Phase 4-Continuing volcanism of the Iceland plume diminished and narrowed along a restricted portion of the East Greenland margin along the Greenland-Iceland ridge (ca. 50 Ma) to present-day hotspot activity in Iceland (Storey et al., 2007).
The position of the (proto-) Icelandic mantle plume has moved through time relative to the overlying tectonic plates ( Fig. 1-e.g., Lawver and Muller, 1994). The various methods used to assess the centre of the plume (e.g., locus of highly magnesian lavas) suggest that the Palaeogene proto-Icelandic plume lay beneath central Greenland in the Palaeogene and Early Eocene, and gradually tracked SE-ESE beneath what is now the Greenland-Iceland ridge (delineated as a zone of thickened oceanic crust) to its present day position beneath Iceland at the Mid-Atlantic Ridge.

The Hebridean portion of the BPIP
The BPIP is part of the earliest magmatic series of the NAIP, which includes Palaeogene rocks of the Hebridean Igneous Province (along the W coast of Scotland) and Northern Ireland (Fig. 3). In the Scottish Hebrides, the BPIP includes the Isles of Mull, Skye, the Small Isles, and the mainland igneous complex of Ardnamurchan and lava flows of Morvern. The crustal sequence into which the magmas of the Hebridean portion of the BPIP were intruded includes the Archaean basement rocks of the Lewisian Gneiss Complex (part of the NAC), Mesoproterozoic Stoer, Sleat and Torridon Group sediments (mostly coarse sandstones, with some shale and silt members), areas of Neoproterozoic Dalradian and Moine metasediments, Triassic limestones and a thick package (N1 km estimated total thickness) of Jurassic and minor Cretaceous sediments representing part of the Hebrides Basin (Fig. 3). Emeleus and Bell (2005) provide a comprehensive compendium of radiometric and palynological age determinations from various published sources. Broadly, the eruption of lavas in this province spanned approximately 3 Myr (61-58 Ma; White and Lovell, 1997). Detailed work by Single and Jerram (2004) has revealed volcanic disconformities within eroded lava fields. Such juxtaposition of volcanic sites and lava batches of the same geochemical type, common in the Hebrides, is a feature of continental flood basalt plateaus which were fed by multiple centres or fissure systems, and must be taken into account in regional lava suite correlation (Jerram and Widdowson, 2005).

The Isle of Mull and Morvern lavas and doleritic plugs
TheMulllavasuccession (Fig. 3) is estimated as being 1800 m thick and covering approximately 840 km 2 (Emeleus and Bell, 2005); the maximum thickness presently exposed in a single section is 1000 m on Ben More (Kerr, 1995a). These flat-lying lavas overlie thin Upper 90 H.S.R. Hughes et al. / Lithos 233 (2015) 89-110 Cretaceous rocks, Triassic sandstones, and Moine Supergroup lithologies which had significant topographic relief at the time of lava eruption. Jurassic sediments also outcrop around limited parts of the perimeter of the island, although the extent of these lithologies underlying the flood basalts is unknown. The Mull Lavas were divided into 3 main groups, originally identified by Bailey et al. (1924), correlated in detail and modified according to petrogenetic and geochemical characterisation by Kerr (1993Kerr ( , 1995a and in subsequent publications. In chronological order, these groups are the Staffa Lava Group, Mull Plateau Group, the Coire Gorm type and the Central Mull Tholeiites. The Mull Plateau Group is the most complete lava succession on Mull and includes a series of predominantly alkali to transitional alkali Mg-rich basalts and hawaiites with minor mugearites and trachytes. Due to their earlier eruption and the  Hughes et al. / Lithos 233 (2015)   relatively high abundance of high-MgO lavas, all Mull and Morvern lava samples analysed during this study are from the Mull Plateau Group.
A series of elongate volcanic plugs are aligned in the direction of the regional dyke swarm and are found throughout Mull and Morvern. These plugs span a similar range of compositions including the Mull Plateau Group lavas and so are thought to have been shallow feeder conduits (Kerr, 1997). A range of tholeiitic dolerite plugs sampled by Kerr (1997) with varying MgO contents were selected for analysis during this study.

Lavas of the Ardnamurchan Peninsula
The Ardnamurchan Peninsula hosts three gabbroic intrusive centres, emplaced concentrically within country rocks of Moine psammites and pelites, Triassic and Jurassic sediments, and partially intruding into thin outliers of the Mull Lava Field (Bell and Williamson, 2002). A selection of these lavas from the lava field east of Beinn an Leathaid were analysed for PGE and Au in the present study.

The Small Isles lavas
Lavas of the Small Isles comprise predominantly olivine and tholeiitic basalts, with minor more evolved compositions (Emeleus and Bell, 2005). Lavas on the Isles of Canna and Sanday are at least 200 m thick, predominantly of alkali olivine basalt, hawaiite, and mugearite. Lavas of the Canna Lava Formation are preserved on NW Rum (Emeleus and Bell, 2005) and a series of samples from this area were analysed during this study (Fig. 3).

Isle of Skye lavas
The lava field on Skye (predominantly the 'Skype Main Lava Series'; 1.2-1.7 km thick transitional (tholeiitic-alkali) flood basalts) is mostly preserved to the NW of the Cuillin Central Complex (Fig. 3)a n d comprises alkali olivine basalts, hawaiites, mugearites, benmoreites, and trachytes (Thompson, 1982;Bell and Williamson, 2002). Lavas were extruded onto Mesoproterozoic Torridonian sediments of the  H.S.R. Hughes et al. / Lithos 233 (2015) 89-110 Torridon and Sleat Groups, Cretaceous and Jurassic sediments of the Mesozoic Hebrides Basin and parts of the Cuillin Central Complex. The absence of volcanic plugs associated with the lava fields on Skye and the more common observation of inter-lava sediments, may be an artefact of the relatively intermittent nature of Palaeogene volcanic activity in this area (Kerr, 1997). Bell and Williamson (2002), amongst others, have divided the Skye Main Lava Series into formations (10 to 760 m in thickness) and members based on flow lithological associations in each region of the island (north, north-central, west-central and south-central). However, extensive block faulting across the lava field, together with common inter-lava sedimentary units, makes full correlation across Skye difficult. All samples analysed during this study were sampled from the west-central and south-central sectors of the Skye Main Lava Series.

Analytical methods
Extensive major and trace element geochemical data and curated samples, exist for the Isle of Mull and Morvern basaltic lavas and doleritic plugs (Kerr, 1993(Kerr, , 1995aKerr et al., 1998Kerr et al., , 1999. In this investigation, we analyse new samples along with a representative selection of curated BPIP lava and plug samples (Supplementary material Table A) for the PGE and Au.
Weathered material was removed from each sample before it was crushed to fine gravel. If present, amygdales and filled vesicles were hand-picked and removed from crushed samples before the remainder was milled to a fine powder in an agate planetary ball mill. Loss on Ignition (LOI) was determined gravimetrically by heating at 900°C for two hours. Major and trace elements were measured by inductively coupled plasma optical emission spectrometry (ICP-OES) and inductively coupled plasma mass spectrometry (ICP-MS) respectively at Cardiff University following fusion using methods and instrumentation described by McDonald and Viljoen (2006). PGE and Au analysis was carried out by Ni-sulphide fire assay (sample weight 15 g) followed by Te co-precipitation and ICP-MS (Huber et al., 2001;McDonald and Viljoen, 2006). Accuracy for whole-rock elemental geochemistry was constrained by analysis of the certified international reference materials TDB1 and WMG1 for PGE and Au, and JB1a, NIM-P and NIM-N for all other trace and major elements (see Supplementary material Tables B and Ca-b). Precision for fire assay was estimated by repeat analysis of a sub-set of samples (Supplementary material Table D).
In total, 58 basalt lava samples (including 3 from Ardnamurchan, 1 from Rum, 40 from the Mull Plateau Group and 14 from the Skye Main Lava Series) and 11 dolerite plugs from Mull and Morvern were analysed by fire assay for PGE and Au during this study. All PGE and Au results are presented in Table 1. For major and trace element analyses 63 new lava samples were analysed from Skye, Rum and Ardnamurchan, along with 11 curated Mull dolerite plug samples. A full table of major and trace element data analysed during this study is available in Supplementary material Table E.

Geochemistry of NAIP and BPIP volcanics
Our results are compared with basaltic lavas collected and analysed for PGE by similar fire assay methods, from onshore Greenlandic plateau suites: West Greenland, Disko Island (Keays and Lightfoot, 2007); East Greenland, the Sortbre Profile in the Kangerlussuaq and Blosseville Kyst areas (Momme et al., , 2006. Further comparison is made with PGE abundances in the seaward dipping reflector sequence offshore of Southeast Greenland (Philipp et al., 2001) and basaltic lavas from Iceland (Momme et al., 2003). Further Au data for Iceland and the Reykjanes Ridge was used for comparison (Webber et al., 2013).

Major elements
A selection of anhydrous major element variation diagrams (vs. MgO) for suites of lavas and plugs from the BPIP (Mull, Skye, Rum and Ardnamurchan) are shown in Fig. 4a-d. Data from Greenlandic lavas (onshore plateau basalts and offshore seaward dipping reflector sequences) and Icelandic lavas are also shown for comparison. These major element diagrams broadly define fractionation trends indicating fractionation of olivine and plagioclase, as has previously been detailed for these sequences (Kerr, 1995a;Fram and Lesher, 1997;Lightfoot et al., 1997;Saunders et al., 1998;Tegner et al., 1998a, b;Kerr et al., Fig. 3. Simplified timeline of lava sequences from phases 1 and 2 (pre-and syn-breakup; ca. 62-56 Ma) in the formation of the NAIP. See text for details (Chambers and Pringle, 2001 and references therein). Note that * indicates that major and trace elements were not analysed during this study but can be found in Kerr (1993). Similarly for samples marked as ** major and trace element data can be found in Jones (2005). Samples with code 'P' (Isle of Mull and Morvern plugs) were re-analysed for major and trace elements as part of this study (see Supplementary Table A) and the original analyses are available in Kerr (1997).

Chalcophile elements
All chalcophile elements in Fig. 5a-d have been plotted relative to Mg# (i.e., [MgO]/([MgO] + [FeO])) as a proxy for magma fractionation. 1 High Mg# lavas have high Ni, Cr and Co concentrations, but some of the lowest Cu abundances. Overall Ni, Cr and Co concentrations in BPIP magmas (according to Mg#) overlap those of Greenlandic and Icelandic lavas. Some West Greenlandic lavas with high Mg# (N 50) have anomalously low Ni concentrations (Fig. 5a). In contrast, Cu abundances in a significant number of BPIP lavas and plugs do not fully overlap other NAIP compositions (Fig. 5b), and some high-Mg# West Greenlandic lavas also have anomalously low Cu concentrations. All onshore East Greenlandic lavas follow an exponential negative trend of increasing Cu content with decreasing Mg#. The majority of offshore East Greenlandic lavas and Icelandic lavas follow a similar trend to onshore East Greenlandic lavas, although these are more scattered. However, the BPIP lavas and plugs appear to fall into one of two categories in Fig. 5b-a cluster that overlaps NAIP lavas for higher Mg# samples vs. a cluster that have lower Cu concentrations (either below this trend at higher Mg# or for samples with Mg# b 50). This division in BPIP samples does not fall rigorously within any one lava suite (e.g., Mull vs. Skye) although most Mull lavas overlap with NAIP.

Platinum-group elements and Au
Chondrite-normalised (McDonough and Sun, 1995)multi-element patterns for PGE and Au are shown in Fig. 6a-d for the BPIP lavas and plugs along with comparative patterns for Icelandic and Greenlandic lavas ( Fig. 6e-h). It is clear that NAIP lava PGE abundances are typically elevated above MORB (Bézos et al., 2005)-see Table 1.NotethatallBPIP lavas analysed for PGE and plotted in Fig. 6 onwards, are transitional alkali-tholeiitic and tholeiitic, to allow for a fair comparison with wider NAIP lavas available in the literature.
All NAIP samples have fractionated PGE such that Rh, Pt and Pd ( P d -g r o u pP G E ;P P G E )a r ee n r i c h e dr e l a t i v et oO s ,I ra n dR u( I rgroup PGE; IPGE), but the relative PGE abundances in each sample suite vary considerably. With the exception of two Skye lava samples, chondrite-normalised Pd is similarly slight enriched over Pt in Fig. 6b. Lava samples from Mull fall into two categories-some have negative Pd anomalies relative to Pt, while the rest have nearly flat Pt to Pd normalised trends (Fig. 6c) (Irvine, 1975).  Kerr, 1993), and east Greenlandic, SDRS and Icelandic lavas (Lightfoot et al., 1997;Philipp et al., 2001 and references therein; Momme et al., 2002Momme et al., , 2003Momme et al., , 2006 and references therein). McDonough and Sun, 1995) whereas lavas from other areas fall consistently below this. In all BPIP lavas, normalised IPGE patterns are relatively flat (with (Ru/Os) N ranging 0.49 to 6.8), although there is variability in Os vs. Ir. For the BPIP lava samples, normalised Au content is highly variable-most lavas have slight positive Au anomalies (Fig. 6e). The normalised PGE patterns for the Mull doleritic plugs broadly overlap that of the Mull lavas, although the plugs lack any negative Pd anomalies and generally have higher Au. Overall, with the exception of Au, Pd and Pt, all BPIP lava compositions overlap Icelandic and Greenlandic (continental) lavas, although the systematic negative Ru anomaly of the Greenlandic ODP lavas (Fig. 6g) has not been found in any of BPIP samples.
The variation in PGE and Au abundances with Mg# are shown in Fig. 7a-f. For low Mg# (b65) BPIP samples, Ir and Ru concentrations are consistently low (b0.25 ppb for Ir, b 0.5 ppb), but at Mg# N 65, Ir and Ru abundances sharply increase in line with existing data from Greenland. By contrast, Rh, Pt and Pd abundances are more variable and scattered when plotted relative to Mg# (Fig. 7c-e). In Fig. 7c-d, there is a broad positive correlation between increasing Mg# and increasing Rh and Pt concentrations, but this does not appear to be the case for Pd. Overall, BPIP lavas and plugs have lower absolute total PGE concentrations at Mg# N 50, in comparison to Greenlandic and Icelandic equivalents. Au abundances in BPIP and W Greenland samples are highly variable and do not correlate with Mg#, unlike the broad positive correlation for Icelandic lavas. However, there is no correlation for Icelandic-Reykjanes Ridge lavas (Fig. 7f).
We find no correlation between Pt/Pd ratio with Sm/Yb or Zr/Y (as proxies for mantle source region fertility-see Fig. 8a-b) nor with Dy/Yb (as a proxy for the presence of garnet in the mantle source region- Fig. 8c). There is also no correlation between Pt/Pd and Nb/La or Nb/Th (Fig. 8d-e) as proxies for crustal contamination across NAIP lavas.
Partitioning of PGE into sulphide liquid and silicate minerals can be identified by bivariant correlations between PGE and other chalcophile elements, as shown in Fig. 9a-d. For example, Ru is incompatible into olivine whereas Ir is weakly compatible (Capobianco et al., 1991;Brenan et al., 2002;Righter et al., 2004) and there is a positive correlation between in Ir and Ni in Fig. 9a. Rh and Ru have been experimentally shown to be compatible in Cr-spinel (e.g., Pagé et al., 2012 and references therein) and there is a positive correlation between Ru and Cr in Fig. 9b.
Pd and Cu have a positive correlation for most BPIP samples, although it is a broad correlation (Fig. 9c). By contrast, Greenlandic lavas are highly variable and scattered, and Icelandic lavas have a range of Pd concentrations but restricted Cu abundances. A very broad positive correlation is observed between Au and Pd (Fig. 9d). However, most BPIP lavas have lower Au abundances than Icelandic lavas. Cu and Pd are incompatible in silicate minerals but partition strongly into immiscible sulphide liquids formed during the S-saturation of silicate magmas. Estimates of partition coefficients range widely according to the experimental conditions (e.g., oxygen fugacity) used to establish them, and this is especially the case for PGE. In summary D Cu~1 000 and D Pd~1 00,000 (Mungall and Brenan, 2014)b u t partitioning estimates for Pd can be particularly variable (57,000 to 580,000 in the experiments of Mungall and Brenan (2014) for example). Nonetheless, this order of magnitude difference for the partitioning behaviour of Cu in comparison to Pd allows ratio plots of these elements to identify S-saturation and sulphide depletion (as originally described by Barnes et al., 1993). If an immiscible sulphide liquid forms, both Cu and Pd will become concentrated in the sulphide and depleted in any residual silicate magma, but Pd will partition more strongly than Cu, sharply increasing the Cu/Pd ratio of the residual silicate magma. Hence if S-saturation was achieved at some prior stage of silicate magma fractionation, the Cu/Pd ratio of the silicate magma will be significantly elevated above primitive upper mantle (cf. Barnes et al., 1993).
The Cu/Pd ratio of BPIP lavas ranges 8000 to 1,000,000 and when plotted against Pd concentration these values overlap with East Greenlandic (offshore), West Greenlandic and Icelandic lavas (Fig. 10a). However, East Greenlandic (onshore) lavas form a tighter cluster of Cu/Pd ratios (~10,000 to 40,000). Although both elements are compatible into sulphide liquid, Pd partitions more strongly than Cu (Peach et al., 1990;Stone et al., 1990;Fleet et al., 1999;Mungall and Brenan, 2014 and references therein). Hence a sulphide accumulation trend vs. sulphide depletion magma trend is defined by Cu/Pd ratio. Lavas have been further subdivided according to Mg# in Fig. 9b-d.  Momme et al. (2002Momme et al. ( , 2003, and Keays and Lightfoot (2007). Two distinct spidergram patterns are defined for Icelandic lavas, according to Mg# (e).
The clustered East Greenlandic (onshore) lavas with limited Cu/Pd ratio indicate that for this suite, S-saturation did not play a major role in changing chalcophile element concentration of ascending silicate magmas prior to eruption as lavas. Thus Cu/Pd ratio is 'unmodified' in East Greenlandic (onshore) lavas, irrespective of the degree of silicate magma fractionation. However, evidence of chalcophile element depletion can be identified in most BPIP samples (particularly for lavas with Mg# b 70) and West Greenlandic lavas, as well as some East Greenlandic (offshore seaward dipping reflector sequence) lavas. Additionally some East Greenlandic (offshore) and Icelandic lavas

98
H.S.R. Hughes et al. / Lithos 233 (2015) 89-110 have Cu/Pd~6000 or within range of primitive mantle (McDonough and Sun, 1995)-see Fig. 9a-d. Pd/Ir ratio can be used as a proxy for the degree of partial melting i nt h em a n t l es o u r c er e g i o no fm a g m a s( c f . Barnes and Lightfoot, 2005 and references therein), indicating whether melting has exhausted Cu-rich (PPGE-bearing) sulphides and begun to incorporate higher temperature IPGE-bearing MSS or metal alloys. During magma ascent through the crust, and if S-saturation is obtained   Hughes et al. / Lithos 233 (2015)   (forming an immiscible sulphide liquid) the residual silicate magma will become depleted in both Pd and Ir alike, according to their strong partitioning into sulphide (D Ir~4 60,000; Mungall and Brenan, 2014 and references therein). Thus Pd/Ir ratios may remain broadly the same as prior to S-saturation, but absolute Pd and Ir concentrations will have decreased in the silicate magma. With very few exceptions, NAIP and BPIP samples have Pd/Ir ratios N primitive mantle (McDonough and Sun, 1995). However, BPIP samples do not overlap with the majority of other NAIP lavas, and sit at systematically lower Ir concentrations at equivalent Pd/Ir ratios (Fig. 11).

Discussion
The partitioning behaviour of chalcophile elements can be used as further evidence of fractional crystallisation throughout magma ascent, in a similar fashion to major elements. For example, the decrease in Ni with decreasing Mg# in Fig. 5areflects incremental removal of Ni during olivine fractionation, although there are anomalously elevated Ni contents in four Skye lava samples and one Ardnamurchan lava, which do not fit this fractionation pattern. The anomalously low Ni concentrations in some high-Mg# West Greenlandic lavas suggest that Ni was depleted in these due to prior S-saturation. Hence sulphide depletion must have occurred early in the fractionation of these lavas (Fig. 5a). Co weakly partitions into olivine such that the positive correlation between Co and Mg# broadly fits olivine fractionation, although this is more scattered than for Ni. Cr partitions into spinel and hence the decrease in Cr abundance with decreasing Mg# reflects spinel fractionation. This is supported by the presence of negative Ru anomalies in Fig. 6a-d (BPIP) and 6 g (East Greenland) which suggest that chromite had crystallised earlier during silicate magma fractionation, and prior to S-saturation.
Cu is incompatible in silicate minerals so that as a silicate magma fractionates, the concentration of Cu in the remaining silicate magma increases sharply. As labelled in Fig. 5b, a significant proportion of BPIP lavas display low Cu concentrations (at lower Mg#) that does not fit with typical silicate fractionation, and instead indicates sulphide depletion during fractional crystallisation. If a silicate magma achieves S-saturation and exsolves an immiscible sulphide liquid, Cu partitions strongly into the sulphide (D sul/sil~1 000), leaving the remaining silicate magma (which ascends and erupts to form the lavas) depleted in Cu (e.g., Andersen et al., 2002). We identify this feature in most BPIP lavas with Mg# b 60, as well as minor Icelandic and East Greenlandic SDRS lavas. In addition some mafic lavas from Skye (Mg# 60-70) and West Greenland (Mg# 55-75) have anomalously low Cu contents and we suggest that these also signify S-saturation prior to eruption.
Various studies have shown that IPGE (such as Ir and Ru) are high-temperature PGE and mainly hosted by Fe-rich monosulphide solid solution (MSS) commonly included in silicate minerals in the mantle, or as refractory metal alloys of Os-Ir-Ru-Pt (Brenan and  H.S.R. Hughes et al. / Lithos 233 (2015)   Andrews, 2001; Ahmed and Arai, 2002;Sattari et al., 2002;Bockrath et al., 2004b). By contrast PPGE (such as Pd) are predominantly hosted by interstitial Cu-rich mantle sulphides (Sattari et al., 2002;Bockrath et al., 2004a;Lorand et al., 2008). During mantle melting, PPGE are preferentially incorporated into magmas over IPGE (which are retained in unmelted monosulphide solution (MSS) and metal alloy residues), so that the initial Pd/Ir ratio is high and PGE spectra of magmas are positively fractionated (Keays, 1982;Barnes et al., 1988Barnes et al., , 1990McDonald et al., 1995;Maier and Barnes, 2004). With increasing degrees of partial melting, the Pd/Ir ratio of partial melts increases until mantle sulphides become exhausted in the source region (Naldrett, 2011). If partial melting exceeds this point of sulphide exhaustion, IPGE-bearing alloys enter the melt while Pd becomes diluted, thereby lowering the Pd/Ir ratio of the magma again. This effect is evident in Fig. 7a-b, where high Mg# numbers reflect a higher degree of mantle partial melting, enough to begin to significantly melt and incorporate IPGE (at Mg# N 65). Additionally some Greenlandic (offshore) and Icelandic lavas have Cu/Pd~6000 (Fig. 10a) within range of primitive mantle (McDonough and Sun, 1995) and coupled with their high concentrations of Cu and Pd, this may suggest that sulphides were largely exhausted from the mantle source region during partial melting. Palladium partitions very strongly into immiscible sulphide liquid and can be identified as being in solid solution within sulphide minerals, particularly pentlandite and pyrrhotite (e.g., Holwell et al., 2012 and references therein). The majority of BPIP samples have significantly lower Pd concentrations than wider NAIP lavas of equivalent Mg# (Fig. 7e) and this may be a reflection of a previous S-saturation event in BPIP lavas, such that residual silicate magmas which erupted as lavas were depleted in Pd during ascent. This feature can also be identified in some West Greenlandic lavas. Evidence of a systematic S-saturation event prior to BPIP lava eruption is also evident in Cu/Pd, particularly in lavas with Mg# b 70 (e.g., Fig. 10c). Pd/Ir vs. Ir (Fig. 11) shows a systematic offset between BPIP lavas and Greenlandic/Icelandic lavas, such that BPIP lavas have lower Ir concentrations at Pd/Ir ratios equivalent to Greenlandic/Icelandic lavas. As demonstrated by the model S-depletion arrows in Fig. 11, this suggests that BPIP lavas were strongly controlled by sulphide depletion prior to their eruption, whereas the majority of  (1995). Sulphide depletion vector calculated from silicate/sulphide partition coefficients (Mungall and Brenan, 2014) for R factors (silicate/sulphide ratio) of 100 to 100,000. Skaergaard chilled composition (blue diamond) based on 265 ppm Cu (calculated by Thomassen and Nielsen, 2006) and 17 ppb Pd (chilled margin analysis by Vincent and Smales, 1956).
Greenlandic and Icelandic lavas were controlled by silicate magma processes only (i.e., fractional crystallisation and olivine accumulation).
6. The S-saturation status for BPIP and NAIP lavas-regional mineralisation potential The occurrence of S-saturation in a magma body may be assessed by the concentration of chalcophile elements within it (e.g., Andersen et al., 2002). Crustal contamination of magmas can trigger S-saturation, particularly if the country rocks contain significant amounts of S in the form of sulphides or sulphates (e.g., Li et al., 2002;Ripley et al., 2003;Naldrett, 2004Naldrett, , 2011Keays and Lightfoot, 2009). During S-undersaturated fractional crystallisation of silicate magma, Cu concentration increases sharply in the residual melt, but if S-saturation is achieved, the immiscible sulphide liquid formed will strongly deplete the Cu content of the residual silicate magma (i.e., a sulphide depleted trend).
Trends relating to S-undersaturated fractional crystallisation and sulphide depletion have been identified in BPIP lava suites, including the two main lava series on Mull and Skye (see Fig. 5b). Some Mull lavas are S-undersaturated (e.g., Fig. 10b), whereas all Skye lavas have experienced S-saturation prior to eruption and follow a sulphide depletion magma trend (e.g., Fig. 10c). While Mull lavas generally have higher Mg# and are therefore less fractionated than Skye lavas, the fact that some Skye lavas with Mg# 60-70 also show anomalously low Cu concentrations is noteworthy. Both the Mull and Skye lava fields erupted through thick lower crustal sequences of Lewisian basement and Moine metasediments, with variable upper crustal sequences of Torridonian and Triassic rocks of the Hebrides Basin. However, current exposures on the Isle of Mull have only thin upper crustal S-rich Lower Jurassic sedimentary units (shales and mudrocks) on the periphery of the island, and it is uncertain how much of this material underlies the lava field. By contrast, a thick package of S-rich Jurassic sediments is prevalent on the Isle of Skye. Thus, we suggest that these Jurassic S-rich mudrocks were the main control on ascending BPIP magmas achieving S-saturation. This is a particularly relevant observation in the context of other NAIP settings.
The vast majority of offshore and onshore East Greenlandic lavas display S-undersaturated (i.e., silicate fractionation only) trends on Cu vs. Mg# plots (see Fig. 5bandAndersen et al., 2002). In the case of onshore East Greenlandic lavas, this is also supported by Cu/Pd ratio (Fig. 10). Although crustal sediments such as Cretaceous-Palaeocene marine shales, sandstones and marls (Soper et al., 1976;Larsen et al., 2001) are available for contamination on the East Greenlandic continental margin, there is a lesser volume of S-rich lithologies in comparison to the Scottish BPIP and Mesozoic Hebridean Basin. Icelandic lavas predominantly show an S-undersaturated trend, although some sulphide depletion is recorded in lavas with lower Mg#. Using Mg# as a proxy for fractional crystallisation (e.g., Fig. 10b-dwheresamples have been grouped according to Mg# categories), and in the absence of S sources from continental crust on Iceland, this limited Ssaturation must be controlled by fractional crystallisation and/or magma mixing. It has been noted by Lightfoot et al. (1997) and Andersen et al. (2002) that lavas in West Greenland display geochemical features of both sulphide depletion and S-undersaturated silicate fractionation, and like the BPIP, different parts of these lava fields are dominated by flows that had undergone S-saturation prior to eruption.
The timing of S-saturation is critical in terms of PGE and metal enrichment in magmatic conduits underlying lava fields, and hence the potential for Norilsk-Talnakh style conduit-hosted orthomagmatic Ni-Cu-PGE mineralisation. Lavas near the base of the Vaigat Formation in West Greenland have the most marked crustal contamination signatures (based on lithophile element ratios and radiogenic isotope compositionssee Lightfoot et al., 1997 andSaunders et al., 1997 andreferences therein). However, because they were amongst the first lavas to be erupted in the province, Lightfoot et al. (1997) suggest that they had limited potential for magmatic mixing and within-conduit 'upgrading' of metal concentrations prior to eruption. This is supported by the observation of high-Mg# lavas with low chalcophile abundances (e.g., Ni depletion- Fig. 5a). With sustained magma flow through a conduit, sulphide liquid that ponds within that conduit can become enriched in chalcophile elements, possibly to economic levels, due to its continued interaction with fresh magma flushing through the conduit (e.g., Kerr and Leitch, 2005). But according to Lightfoot et al. (1997) and based on lava geochemistry, magmatic conduits or basal intrusions relating to lavas of the Vaigat Formation provide limited potential for significant high tenor orthomagmatic mineralisation. We suggest that this negates the 'recycling' of magmatic conduits (i.e., re-activation of conduits by later silicate magma batches represented by subsequent lavas) nor does it account for sustained magma flow through conduits, or the complexity of localised conditions such as variable crustal contamination. For example, the close-association, both spatially and temporally, of lavas fields in the BPIP shows that comparatively localised districts may have low vs. high S-contamination potential (for Mull and Skye respectively) and hence magmatic conduits underlying and feeding these lava fields may be similarly multifaceted. Therefore regional 'prospectivity', as determined by whether lavas experienced S-saturation or remained undersaturated, must take into account this variability.
In subsurface magmatic conduits, where sulphur has not degassed, S-isotopic studies (δ 34 S and Δ 33 S) can identify the internal or external cause(s) for S-saturation. Hughes et al. (in press) assessed the wholerock δ 34 S signature of a suite of upper crustal narrow basaltic dykes and thick picritic sills intruded into Jurassic mudrocks on the Isle of Skye. These dykes have characteristically light δ 34 S and are clearly contaminated with crustal S. Orthomagmatic sulphides with extremely light δ 34 S occur deeper in the magmatic plumbing system in ultramafic volcanic plugs on the Isle of Rum (Power et al., 2003;Hughes et al., in press). Structural evidence discussed by Hughes et al. (under review) suggest that the vertical/subvertical volcanic plugs on Rum erupted through Jurassic sedimentary units, becoming strongly contaminated with isotopically light crustal S, inducing S-saturation and  (1995). Sulphide depletion vector calculated from silicate/sulphide partition coefficients (Mungall and Brenan, 2014) for R factors (silicate/sulphide ratio) of 100 to 100,000. Skaergaard chilled composition (blue diamond) based on 0.1 ppb Ir (estimated) and 17 ppb Pd (chilled margin analysis by Vincent and Smales, 1956).
forming an immiscible sulphide liquid which drained down through the partially crystallised conduit when activity ended (Hughes et al., under review).
Overall, the prevalence of fusible S-rich crustal rocks within the BPIP should be considered as a major exploration factor, both for the formation of immiscible sulphide liquids, but also for the collection or concentration of chalcophile elements including PGE in systems where magma input was sustained for long periods. The widespread identification of S contamination and saturation in upper crustal magmatic conduits (by δ 34 S) and lavas (by chalcophile element geochemistry) very clearly points to the presence of sulphides and enhanced mineralisation potential deeper within the Scottish BPIP magmatic plumbing system (as corroborated by Arndt, 2013) and therefore represent thus far unexplored exploration targets.

Pt/Pd ratio of NAIP magmas throughout continental break-up
Systematic differences in Pt and Pd abundance between lava suites across the NAIP are evident in Fig. 7d and e. BPIP lavas predominantly have lower Pd concentrations (at equivalent Mg#) in comparison to most Greenlandic and Icelandic lavas (Fig. 7e), and Pt vs. Mg# for BPIP lavas defines an entirely differently shaped trend to Greenlandic and Icelandic compositions, despite there being some overlap in the actual range of Pt (Fig. 7d). The sample preparation and analytical methodologies of our study for PGE in lavas are directly comparable to those of Philipp et al. (2001), Momme et al. (2002), Momme et al. (2003) and Keays and Lightfoot (2007). Therefore differences between Pd and Pt for the NAIP lava suites are unlikely to be due to analytical differences and appear to be true compositional features.
Plots of Pt vs. Pd (Fig. 12) for lavas from the NAIP are particularly instructive in assessing the compositional differences between the various areas. Lavas from the BPIP and West Greenland have Pt/Pd ratios near chondrite (1.9) and fall along a positive trend (statistically significant, r 2 N 0.89; Fig. 11a-b). At higher concentrations of either of these elements (i.e., N 9 ppb Pt or Pd) more variability around a chondritic ratio is observed. Continental flood basalts from East Greenland (onshore) have extremely variable and scattered Pt/Pd ratios (Fig. 12c) but Pd is consistently enriched over Pt in comparison to BPIP and West Greenlandic lavas. East Greenlandic SDRS (offshore) lavas have a clear positive correlation (r 2 N 0.88) between Pt and Pd, but with a different sub-chondritic Pt/Pd ratio (Fig. 12d). Icelandic lavas also have a positive correlation between Pt and Pd (r 2 N 0.90) but with a much lower Pt/Pd ratio (Fig. 12e) than the offshore east Greenland lavas.
All the lava suites discussed in this study compare transitional alkalic-tholeiitic (e.g., BPIP) and tholeiitic (e.g., Iceland) picrites and basalts (with minor andesites). The differences in Pt/Pd ratio do not correspond either with a particular compositional series of lavas, nor with fractionation. Likewise, during magma ascent, Ssaturation will not significantly fractionate Pt from Pd, as both elements have partition coefficients of a similar order of magnitude and will be partitioned correspondingly into an immiscible sulphide liquid. While in some crustal PGE-enriching environments (e.g., a long-lived conduit in which an immiscible sulphide liquid is progressively dissolved- Kerr and Leitch, 2005) Pt/Pd ratio may change, this effect is comparatively minor and would not be systematic across the various regions of the NAIP (cf. Fig. 12). In the NAIP Pt/Pd ratio appears to correspond to the age of the lava suites (i.e., the highest Pt/Pd ratio in the earliest lava suites of West Greenland and the BPIP vs. the lowest Pt/Pd in recent Icelandic lavas). The question arises whether this temporal change in the Pt/Pd ratio is indicative of a change in the plume-composition or geodynamic setting of the North Atlantic through time: from the initial eruption of picritic and basaltic magmas during early (proto-Iceland) mantle plume impingement, throughout continental rifting, the formation of oceanic lithosphere, and ultimately to the modern plume-oceanic rift setting of Iceland (Fig. 12f).
Possible causes for a systematic change in magma PGE geochemistry related to tectonic setting and/or geodynamic environment through time may include: (a) the degree or depth of partial melting; (b) changes in the mantle source region (i.e., enriched vs. depleted sources); (c) the melting regime and the shape of the melt column (i.e., cylindrical vs. triangular); (d) the mantle potential temperature; (e) the incorporation of deep mantle material in the plume; and/or (f) lithospheric contamination. Each of these possibilities is assessed below.

Degree of partial melting
Relative proportions of fertile mantle PGE concentrations are approximately chondritic (Carlson, 2005), although lower than chondrite in actual concentration by a factor of approximately 100-1000, as indicated by mantle-derived peridotite xenoliths (e.g., Luguet et al., 2003;Lorand et al., 2008;Maier et al., 2012 and references therein). During mantle melting, the first minerals to melt are sulphides, garnet and clinopyroxene, and so with increasing degrees of partial melting sulphides in the mantle source region become exhausted. However, the timing of sulphide exhaustion is primarily dependent the abundance of S in the mantle source, as well as whether an equilibrium batch melt, fractional melting, continuous melting, or dynamic melting model is assumed, and the shape and depth (pressure) of the melting profile (e.g., Rehkamper et al., 1999).
Partial melting of sulphide minerals in the mantle can be identified by fractionation of IPGE from PPGE, and therefore changes in Pd/Ir ratio (as previously outlined in Section 4.3). It is known from both experimental work and natural analogues of mantle materials that partial melting of base metal sulphides can produce Fe-rich MSS, Cu-Ni-Pd sulphide liquid with Fe-rich MSS and/or Ir-Pt alloys in the residue (e.g., Crocket, 2002;Luguet et al., 2003;Peregoedova et al., 2004). This will not only fractionate Pd from Ir, but could also fractionate Pd from Pt. We have already established that the range in Pd/Ir for each of the NAIP lava suites broadly overlaps despite lower absolute Ir concentrations due to PGE removal in sulphide liquids (Fig. 11). With the retention of Pt-bearing alloys in the mantle source region and/ or entrainment and dissolution of Cu-rich (Pd-bearing) sulphide liquid into silicate magmas produced during partial melting, Pt and Pd may be fractionated from one another causing Pt/Pd ratios to vary according to the degree of partial melting. However, coupled with other melting proxies such as MgO content and Mg# (e.g., Figs. 4-5), a systematic change in Pt/Pd ratio related to the degree of mantle partial melting cannot be identified. Therefore this option is discounted as being a credible control on Pt/Pd ratio across the NAIP.

Melting source region and depth
Changes in asthenospheric mantle sources undergoing melting have previously been proposed for the NAIP. For example, Kerr (1993Kerr ( , 1995 and Kerr et al. (1999) used REE systematics to identify variations in partial melting of depleted (i.e., due to a prior melting event) vs. enriched asthenospheric mantle at both the spinel-and garnet-stability depths during the magmatism on the Isle of Mull. Similarly other incompatible trace elements in lavas and intrusive rocks have been used to constrain the depth and degree of melting in the BPIP (see Kent and Fitton, 2000 and references therein). Starkey et al. (2009) measured REE abundances in whole-rock lava samples and melt inclusions in olivine crystals from picritic lavas for Baffin Island and West Greenland, and identified both enriched anddepletedmantlesources. Momme et al. (2006) used TiO 2 and REE contents to infer variations in melting depth and upper mantle sources for onshore East Greenlandic lavas. Isotopic studies by Kempton et al. (2000) and Murton et al. (2002) point to variations in upper mantle components from up to 6 end members. However, although variations in melt sources of enriched vs. depleted end members and melting depths are clearly identifiable within the lava suites of the NAIP, this does not correlate with the broader temporal change in Pt/Pd observed across the NAIP as a whole (see Fig. 8a-c), which must reflect a more fundamental control.  Keays and Lightfoot, 2007), (c) East Greenlandic lavas , (d) East Greenlandic SDRS lavas (Philipp et al., 2001), (e) Icelandic lavas (Momme et al., 2003), (f) all NAIP lavas. Chondritic ratio from McDonough and Sun (1995).Calculated trend lines and regressions labelled for plots (b), (d) and (e). Equivalent Pt/Pd ratios and trend line regressions labelled for each lava suite in (f).

Melting regime
Mantle melting modelling falls into two basic types; either simple batch melting or melt aggregates of near-fractional non-modal melting. In the case of the former, unreasonably high degrees of melting are required to explain the PGE content of Icelandic basalts, which cannot be reconciled with MgO content and other major and trace element systematics (Rehkamper et al., 1999). Therefore more recently, various non-modal melting models have been used to explain the combined lithophile and chalcophile element geochemistry of Icelandic and onshore East Greenlandic lavas (Momme et al., 2006). Non-modal melting models use either a triangular (analogous to spreading ridges; e.g., Momme et al., 2003) or a columnar (i.e., cylindrical; e.g., Keays, 1995;Rehkamper et al., 1999) melting regimes. For the East Greenlandic (onshore) lavas (Fig. 11c). Momme et al. (2006) suggested that lavas further inland with higher Ti contents were formed by a deeper triangular melting system in a continental rift setting, while lower Ti lavas nearer to the coast formed during continental rupture at greater degrees of shallower melting in a columnar melting regime. While these models reconcile the broad PGE (specifically Pd) and REE features of these East Greenlandic lavas, there is no correlation between Pt/Pd ratios of highand low-Ti lavas. Both have highly scattered Pt/Pd (Fig. 11c) and so the melting regime and shape of the melt column are unlikely to exert a strong control on Pt/Pd across the NAIP.

Ambient temperature of melting or mantle potential temperature
Various estimates of mantle potential temperature exist for NAIP magmas. However, the methods by which temperatures are calculated include major and trace element thermodynamic modelling (e.g., Herzberg et al., 2007), estimates based on olivine compositions (e.g., Kent, 1995), and modelling based on geophysical constraints (e.g., seismic tomography, lithospheric uplift, buoyancy and/or estimates of active vs. passive upwelling; Brown and Lesher, 2014). All provide strong evidence for elevated mantle temperatures due to the presence of a mantle plume in this region. In particular Brown and Lesher (2014) calculate that the Palaeogene mantle was N125°C hotter than ambient mantle temperature at the time, but suggest that this temperature elevation is comparatively muted beneath present-day Iceland (85-180°C). Overall, the degree of uncertainty associated with mantle temperature estimates, coupled with the degree of uncertainty inherent in the use of differing methods of estimation, make a comparison of ambient mantle temperatures vs. Pt/Pd ratio difficult for the NAIP lava suites included in the current study. Nonetheless, the possibility that plume-related mantle potential temperatures have diminished during the formation of the NAIP and could systematically affect the Pt/Pd ratio is support by experimental studies that have shown that Pt-alloys have a significantly higher melting temperature than Pd-rich Cu sulphides or MSS (e.g., Peregoedova et al., 2004).

Incorporation of deep mantle or outer core material in the plume
Involvement of outer core material in mantle plumes could (in part) account for the elevated PGE, Ni and Fe contents observed in some plume-derived magmas, including those of the NAIP (e.g., Andersen et al., 2002). Andersen et al. (2002) calculated that PGE concentrations could be enriched 2.5 to 3 times above normal mantle contributions if as little as 0.5 to 1% of core material were contributed to a plume. Pt-Re-Os isotopic studies of plume-related magmas have widely documented an enriched 186 Os/ 187 Os signature, particularly noteworthy in lavas from West Greenland and Baffin Island, prompting suggestions that this represents incorporation of deep mantle or outer core-mantle boundary material in the plume magma (e.g., Pearson et al., 1999;Brandon and Walker, 2005). In addition, observations of elevated 3 He/ 4 He in the Icelandic plume and other plume settings (e.g., Kurz et al., 1983;Rison and Craig, 1983;Hilton et al., 2000 and references therein;Graham et al., 1998;Breddam et al., 2000), apparently provides further evidence of a deep mantle or core-mantle signature (e.g., Rison and Craig, 1983). However, more recent literature has highlighted the likelihood of elevated 3 He/ 4 He by decoupled entrainment, which implies that the He-isotopic signature cannot directly be used as evidence of core-mantle interaction (e.g., Starkey et al., 2009). Further doubt over the significance of elevated Re/Os, Pt/Os and 186 Os/ 187 Os signifying a core-mantle signature has come from W-isotopes in Hawaiian basalts, which only have bulk mantle compositions (Schersten et al., 2004;Carlson, 2005 and references therein).
Recent studies suggest that the initial volcanism in the BPIP and NAIP was unlikely to have contained a significant contribution from the outer core or from old recycled crustal material (Dale et al., 2009). Instead Os and He-isotopic studies suggest that early Baffin Island and West Greenlandic picrites were produced from a high percentage of melting in a depleted source region (leading to complete melting of mantle sulphides) at the head of the plume. Later phases of continental rifting and continued magmatism in modern Iceland contained a greater component of recycled oceanic crust melting (Dale et al., 2009). Kempton et al (2000) identified a higher contribution from a relatively enriched plume component in basalts from the SDRS offshore SE Greenland, than along the Reykjanes Ridge and Iceland today, and they suggest a diminishing plume component in the NAIP through time. Overall,ifDale et al. (2009) is accepted, then an outer core contribution to plume magmas cannot be the cause of Pt/Pd variation.

Lithospheric contamination
When assessing lithospheric contamination it is important to distinguish between crustal contamination and contamination from the subcontinental lithospheric mantle (SCLM). Deep crustal contamination of magmas by silicic basement is widely documented in lavas from across Greenland and the BPIP, using trace elements and isotopic compositions (see reviews in Saunders et al., 1997 andKerr et al., 1999). However, the concentration of PGE within silicic crustal rocks is typically very low (see Carlson, 2005 and references therein) and they are therefore extremely unlikely to contribute to the PGE budgets of ascending mantle-derived magmas. There is no correlation between Pt/Pd and Nb/La (or other trace element proxies for contamination e.g., Ba/Nb, Nb/Th-see Fig. 8d-e) and so crustal contamination cannot be responsible for changing Pt/Pd ratio across NAIP lavas.
Pt/Pd ratios have previously been utilised to infer a metasomatic and 'on-craton' signature for various magmas, with an increased Pt/Pd ratio resulting from input of a metasomatic fluid-alteration component in the lithospheric mantle (Maier and Barnes, 2004). However, this does not reflect the physical nature of entrainment of PGE-bearing minerals and/or alloys affecting Pt/Pd ratio (alongside other PGE ratios), as originally proposed in the instance of kimberlites (McDonald et al., 1995). Nonetheless, metasomatic sulphides (with or without a demonstrable association of 'fluid'-rich alteration) are common in mantle peridotites and pyroxenites of the lithospheric mantle, as has been demonstrated in oceanic mantle, abyssal peridotites and alpine peridotites (e.g., Luguet et al., 2003;Alard et al., 2005;Luguet et al., 2008).  reports 3 distinctive populations of native base metal sulphides in the SCLM underlying northwest Scotland (Loch Roag, Isle of Lewis). One of these sulphide groups is systematically found in frozen melt pockets along grain boundaries (i.e., areas that would be most susceptible to melting). These sulphides have the highest abundance of PGE relative to the other sulphide groups in the xenolith, and include discrete micron-scale PtS (cooperite) platinum-group minerals. The bulk rock geochemistry of the Loch Roag peridotite xenoliths show chondritic Pt/Pd ratios, and chondrite-normalised PGE spectra show the xenoliths to be PPGE-rich with Pt and Pd at higher concentrations than that found in cratonic xenoliths from Western Greenland (Wittig et al., 2010)and'primitive upper mantle ' (McDonough and Sun, 1995;Palme and O'Neill, 2004). Although the age of these PtS-bearing sulphides remains unquantified, based on a combination of whole-rock Re-Os isotope analyses and sulphide-specific Re/Os ratios,  proposes that this group of sulphides are likely to be pre-Palaeogene, and could record a much older sulphide-bearing magmatic/metasomatic event.
Incorporation of fusible SCLM into plume magmas or ahead of a plume melting front has been suggested previously (e.g., Thompson and Morrison, 1988;Kerr, 1993;Saunders et al., 1997 and references therein;Foulger et al., 2005). This SCLM fusion is particularly feasible given the likelihood of lithospheric thinning or delamination inferred to take place during plume impingement on the base of continents in the North Atlantic (e.g., Kerr, 1994;Saunders et al., 1997). Hence the higher Pt/Pd ratio recorded in the older NAIP lava suites (e.g., BPIP and west Greenland) could reflect assimilation or mixing of ascending asthenospheric magmas with a (complex or multi-phase) Pt-rich lithospheric mantle end member. Given that sulphides are readily fusible and one of the first mineral phases to melt, we suggest that this SCLM interaction could have had a substantial effect on the precious metal budget of plume-derived asthenospheric magmas in the NAIP, essentially dictating the Pt/Pd ratio (NMORB) recorded spatially and temporally across the province.
Incorporation of PGE from the SCLM has previously been suggested for Pt-rich deposits such as the Bushveld Complex and Stillwater Complex (see Maier and Groves, 2011 and references therein). However, in these cases the involvement of SCLM in the PGE budget has been suggested by proxy, either by non-PGE geochemical evidence (e.g., elevated Fig. 13. Cratonic block reconstruction for the North Atlantic (based on Bleeker, 2003) vs. NAIP lava Pt/Pd ratio. Thick orange lines highlight cratonic margins (nominally to a width of 100-150 km, as a distance of b100 km was highlighted as being 'near craton margin' by Begg et al., 2010, and the Bushveld Complex sits approximately 130 km from the Kaapvaal cratonic margin). lithophile element concentrations), or by radiogenic isotopes, and always in the absence of sulphide compositional analyses and petrography of SCLM material (i.e., mantle xenoliths) corresponding to that underlying such deposits. By contrast, between this study and  we can clearly demonstrate that PGE-rich sulphides containing PtS (cooperite) exist in fusible lithologies (and along silicate grain boundaries) in parts of the Scottish SCLM and would have been available for contamination of asthenospheric plume magmas during the formation of the BPIP. While an ancient metasomatic process may ultimately have been responsible for the pre-enrichment of this cratonic SCLM by sulphide liquids, we suggest that such metasomatic enrichment was itself not coeval with the LIP magmatism responsible for PGE (especially Pt-rich) deposit formation (c.f., Barnes et al., 2010;Maier and Groves, 2011). Hence the dichotomy of whether SCLM-or plume-based melting is the dominant control on PGE prospectivity and relative Pt/Pd abundances, may obscure a deeper truth. Instead the preconditioning of cratonic SCLM may be an essential precursor to LIP magmatism in defining certain PGE deposit geochemical characteristics. This means that while the majority of magma is derived from asthenospheric mantle (possibly plume) source regions, during ascent precious metal ratios (e.g., Pt/Pd) may be modified by melting and contamination by multiple 'populations' of SCLM sulphides in specific petrographic settings (e.g., pre-existing and preserved melt pockets).
This raises the question-which regions of the SCLM are prone to this preconditioning, and why? For the NAIP lava suites underlain by ancient cratonic gneiss basement and SCLM, the higher Pt/Pd ratio in BPIP and West Greenlandic lavas at first appears to conflict with the more scattered Pt/Pd ratios of East Greenlandic (onshore) lavas. Craton reconstructions in the North Atlantic show that the NAC is bordered to the north by the Nagssugtoqidian orogenic belt (1900( -1680van Gool et al., 2002;Kolb, 2015) and that this Palaeoproterozoic mobile belt forms the southern border of the Rae Province-see Fig. 13. The Lewisian gneiss complex of Scotland, and the underlying SCLM (as represented by Loch Roag mantle xenoliths) occur on the very margin of the NAC, and are known to have undergone Palaeoproterozoic orogenesis (e.g., Loch Maree arc terranes c. 1900 Ma; Park, 2002). Lavas from West Greenland (Disko Island) also sit close to this mobile belt, although they actually overlie the southern margin of the Rae Province. By contrast NAIP lavas in East Greenland span the full width of the Rae Province and in particular lavas of the Sortebre Profile (analysed for PGE by Momme et al., 2002 and included in this study) are notably further from the Nagssugtoqidian orogenic belt than the West Greenlandic lavas.
Hence we suggest that depending on where the plume head impinges on the lithosphere, parts of the LIP may involve melting of SCLM regions that have experienced older metasomatism which introduced Pt-rich sulphides. This predisposition of Pt-enrichment is expected to be restricted along lithospheric mantle lineaments such as cratonic block boundaries and/or neighbouring orogenic belts (e.g., Nagssugtoqidian orogenic belt vs. NAC and Rae) and is not inherent in the depleted keel that makes up the majority of Archaean cratons themselves (e.g., Pearson et al., 2003). While the bulk rock Pt/Pd ratio of depleted cratonic keel mantle xenoliths may be broadly chondritic (e.g., Wittig et al., 2010) this does not inform us about PGE 'fusibility' which will be maximised when PGE-rich sulphides are developed along grain boundaries. In contrast, if the bulk rock PGE budget is strongly controlled by sulphides included within olivine or Cr-spinel, these are unlikely to contaminate passing asthenospheric magmas. Similarly, if sulphides along silicate grain boundaries only carry low PGE concentrations, their effect on asthenospheric magma PGE composition will also be negligible, despite their more feasible entrainment.
Lithospheric lineaments as conduits for ascending asthenospheric magmas and/or fluids have previously been highlighted as a control on the location of PGE mineralisation (e.g., Begg et al., 2010). We suggest that this is coupled with a fundamental geochemical control too, and that enrichment in precious metals along craton lineaments or margins is not a simple scenario. For example, we see no clear evidence to support a straightforward 'in-out box model' situation in which SCLM partial melting and S-depletion (thereby enriching PGE in the residual mantle sulphide) is followed by second-stage melting of the same SCLM (as proposed by Hamlyn and Keays, 1986). Instead the variety of sulphides reported widely in xenolith suites worldwide and from a range of tectonic settings (e.g., Guo et al., 1999;Lorand et al., 2004;Pearson et al., 2003 and references therein;Lorand et al., 2008;Delpech et al., 2012;Lorand et al., 2013) are testament to the true complexity of S-bearing events recorded in the lithospheric mantle. In the case of the BPIP and its relationship to the NAC, the multiple 'populations' of Loch Roag xenolith sulphides  demonstrate the multi-stage enrichment that marginal cratonic lithospheric mantle, or cratonic block/boundary lineaments in the SCLM, may trace. The spatial relationship between ascending plume magmas and whether these magmas interact with pre-enriched (Pt-rich) cratonic boundaries or not, may exert a strong control over the Pt/Pd ratio of LIP magmas.

Conclusions
Analyses of the platinum-group element geochemistry of a series of lavas from the BPIP, in conjunction with other published studies of Greenlandic and Icelandic NAIP lavas, have enabled us to tackle two main themes pertaining to mineralisation potential and petrologic constraints on plume-SCLM magmatism and interaction.
1 The apparent enrichment (relative to MORB) in PGE abundances of the NAIP lavas and differing S-saturation controls across the region have been used to assess the mineralisation potential for orthomagmatic Ni-Cu-PGE sulphides in Greenland and Iceland. The widespread opportunity for crustal contamination to take place (particularly in the upper crust with widespread S-rich sediments of the Hebrides Basin) means that the BPIP represents one of the most fertile regions of the NAIP for Ni-Cu-PGE mineralisation. 2 Pt/Pd ratios change systematically across the NAIP, both spatially and temporally, such that the early NAIP lavas in West Greenland and the BPIP have the highest Pt/Pd ratios (approximately chondritic; 1.8). In comparison lavas erupted during the initiation of continental rifting and formation of oceanic crust have progressively lower Pt/Pd, such that modern Icelandic lavas are comparatively Pd-rich. This broad shift in Pt/Pd ratio coincides with changes in the geodynamic setting of the NAIP throughout its evolutionfrom the first formation of plume-derived magmas beneath thickened SCLM of the North Atlantic Craton and Rae Province, to modern oceanic rifting. Although this geochemical characteristic may be strictly plume controlled and therefore reflect a change in the dominance of the (proto-) Icelandic plume through time, the existence of preexisting Pt-enriched SCLM underlying the margins of the NAC, and underlying early high Pt/Pd NAIP lavas suggests that there is a fundamental interaction between the plume and lithospheric mantle. 3 We suggest that Pt-enriched SCLM is restricted to the margins of cratons and along cratonic block lineaments such as Palaeoproterozoic inter-block mobile belts. We therefore favour a model whereby fusible sulphides in a fertile and Pt-rich marginal lithospheric keel become mobilised (via partial melting or wholesale lithospheric delamination) and incorporated into ascending asthenospheric magmas. The spatial relationship between intercratonic Proterozoic mobile belts and/or cratonic lineaments vs. the path of ascent of asthenospheric derived (plume) magmas dictates the PGE metal ratios inherent in LIP magmas and hence mineralisation within these provinces.