Contrasting TiO2 Compositions in Early Cenozoic Mafic Sills of the Faroe Islands: An Example of Basalt Formation from Distinct Melting Regimes

The Paleocene lava succession of the Faroe Islands Basalt Group (FIBG), which is a part of the North Atlantic Igneous Province (NAIP), is intruded by numerous basaltic sills. These can be grouped into three main categories according to their geochemical characteristics: A low-TiO2 sill category (TiO2 = 0.7-0.9), a relatively high-TiO2 sill category (TiO2 = 1.95-2.6) and an intermediate-TiO2 sill that displays major element compositions lying between the other two categories. Mantle normalised plots for the high-TiO2 and low-TiO2 sills display relatively uniform flat LREE trends and slightly steeper HREE slopes for high-TiO2 relative to low-TiO2 sills. The intermediate-TiO2 Morskranes Sill is LREE depleted. Mantle normalised trace elements of low-TiO2 sill samples define positive Eu and Sr anomalies, whereas high-TiO2 sill samples display negative anomalies for these same lements. Different Nb and Ta anomalies (positive versus negative) in many high-TiO2 versus low-TiO2 sill samples suggest various metasomatism of their sources prior to partial melting. The intermediate-TiO2 sill displays noticeably lower 87 Sr/ 86 Sr, 206 Pb/ 204 Pb and 208 Pb/ 204 Pb ratios relative to both the high-TiO2 and the low-TiO2 sill samples. Pb isotope compositions displayed by local co*ntaminated basaltic lavas imply that some of these assimilated distinct crustal material from E Greenland or basement from NW Britain, while others probably assimilated only distinct E Greenland type of crustal material. A third crustal source of E Greenland or Rockall-type basement could be required in order to explain some of the range in lead isotopes displayed by the intermediate-TiO2 Morskranes Sill. Geochemical modelling suggest that Faroese high-TiO2 sills, could have formed by ~4 to 7.5% batch melting of moderately fertile lherzolites, while 16 to 21% batch melting fertile mantle sources could explain geochemical compositions of Faroese low-TiO2 sills. The intermediate-TiO2 sill samples could have formed by a range of 6 to 7% batch melting of a depleted mantle source, probably with a composition comparable to sources that gave rise to local low-TiO2 and intermediate-TiO2 host-rocks. Most Faroese sill samples probably developed outside the garnet stabilitry field and probably formed by batch melting of mantle materials comparable in composition to those reported for the sub-continental lithospheric mantle (SCLM) previously at depths of  85 km. Relative enrichments in LREE (and LILE in general), and their varying Nb and Ta anomalies point to sources affected by metasomatism.


Introduction
Numerous hypotheses on the petrogenetic evolution of large basaltic igneous provinces (LIPs) such as the North Atlantic Igneous Province (NAIP) have been proposed earlier [1][2]. In order to distinguish between various units/suites within LIPs in general and flood basalts in particular, TiO 2 contents are commonly described [3][4][5][6][7][8][9]. Formation of low-TiO 2 basalts have at times been attributed to relatively largedegree melting of enriched material from the continental lithospheric mantle or else developed in response to melting of other mantle sources contaminated with subducted crust/sediments and/or associated expulsion of fluids [10][11][12]. Alternatively, they could also result from relatively largedegree melting (~20%) of depleted mantle sources [7][8]. By contrast, high-TiO 2 basaltic magmas probably form by relatively low-degree mantle melting from sources that may reside at relatively deeper mantle levels [9,13,14], or else the sources may originate from mantle plumes, perhaps owing some of their geochemical characteristics to recycled oceanic crust [8][9][10][11].
The lava successions of the Faroe Islands have been intruded by a number of sills. These provide insight to some of the later stages of magma generation. Although these sills have previously been characterised in terms of their geometries and stratigraphic relationships within the lava sequences [27], limited information exists regarding their petrogenesis, and in particular the likely sources of melting and how they may or may not relate to the lava sequences that they intersect. Understanding the later stages of magma generation during the evolution of the NAIP is desirable to better understand the wider temporal and spatial evolution of the province as a whole [2,28,29].
Accordingly, we focus on the compositions of primary magmas that ultimately evolved to the sills of the Faroe Islands. The sill samples are categorised according to major element compositions (especially TiO 2 ), trace elements (including REE) and isotopic compositions (Pb, Sr and Nd). The sills are then grouped into three main units defined primarily by their TiO 2 compositions. Their evolution is examined by partial melting modelling in order to help characterise their formation and mantle sources in terms of the degrees of melting and source rock compositions. The petrogenetic interpretations presented for the Faroese sills in this study are considered in the context of late-stage magmatic processes during the formation of the NAIP, and in particular we look at the effects of partial melting at relatively shallow mantle levels during the waning stages of basaltic magmatism in this LIP.

Geological Framework
The Faroe Islands Basalt Group (FIBG) was emplaced at the NW European margin as a central part of the contemporaneous NAIP magmatism [3,30,31,32]. Geophysical studies suggest that the basaltic rocks of the Faroese block lie on ~30 km thick stretched ancient continental crust [33,34,35,36], with total stratigraphic thickness of exposed and drilled lavas of this region of ~6.6 km [3,32,37,38,39] (Figure 1). The total extent onshore and offshore of these basaltic rocks have been estimated at ~120000 km 2 [32]. The Faroese lava succession was previously grouped into Upper, Middle and Lower Series basalts , but a revised nomenclature with seven formations was proposed recently [32]. From the bottom to top of the lava succession these are: the ~1075 m thick Lopra Formation; the ~3250 m thick Beinisvørð Formation; the ~9 m thick Prestfjall Formation; the 40-50 m thick Hvannhagi Formation; the 1250-1350 m thick Malinstindur Formation; the ~30 m thick Sneis Formation and finally the ~900 m thick Enni Formation [32][33][34][35][36][37][38][39] (Stratigraphic column Figure 1). Individual lava flows within the Beinisvørð, Malinstindur and Enni formations are often separated by thin volcaniclastic lithologies or weathering surfaces measuring a few centimetres, a few tens of centimetres and occasionally a few metres in thickness [32,38,39]. Basalts of the Beinisvørð Formation are generally aphyric whereas those of the Malinstindur and Enni formations include both olivine and plagioclase phyric rocks in addition to aphyric basalts [3,32,37]. Dating by the 40 Ar/ 39 Ar method has yielded ages of ~61 Ma for a lava sequence within the Beinisvørð Formation [42] and ~55 Ma for another lava sequence belonging to the Enni Formation [43]. An additional study on age(s) of Faroese lavas based on local palynoflora stated that the entire lava succession was deposited in the time span from 57.5 to 60.56 Ma [44].
Dykes are ubiquitous throughout the exposed parts of the lava succession while 'saucer-shaped' sills are confined to the uppermost parts of the Malinstindur Formation, the Sneis Formation and the lowermost parts of the Enni Formation [27,32,38,39,41] (Figure 1). High-TiO 2 lavas (TiO 2 >1.5 wt%) make up most of the volume in the lowermost ~5.5 km of the Faroese lava succession, while low-TiO 2 lavas (TiO 2 <1.5 wt%) become increasingly common in the remaining upper parts [3,8,9,32,37,45,46]. High-TiO 2 dykes and lavas of the Malinstindur and Enni formations are exposed throughout the archipelago, while low-TiO 2 dykes and lavas of these formations are most common in the northern parts of the islands . Low-TiO 2 sills are exposed in the central parts of the Faroe Islands whereas high-TiO 2 sills are exposed in the N and NE parts of the archipelago . A few lava flows of the Malinstindur Formation display anomalously high SiO 2 contents of ~54.0 wt% [26]. In the Beinisvørð Formation, clear stratigraphic trends of MgO and TiO2 contents are recorded and associated with large scale magma pulses [48]. However, in the Malinstindur and Enni Formations, the chemical stratigraphy is more erratic between individual lava flows irrespective of stratigraphic levels [8,45,46], and this has recently been demonstrated to relate to multiple inter-digitating lava flow packages fed from distinct magma batches [9].  [3,38,40] is indicated to the left of the stratigraphic column, while the revised nomenclature from [39] is shown to the right. (The figure is modified from [27]).

Previous Petrogenetic Models on Faroese Magmatism
Primary magmas that gave rise to high-TiO 2 basaltic dykes and lavas of the Faroe Islands are thought to have formed in response to 2.5 to 3.5% mantle melting compared to ~20% melting during production of primary magmas that yielded local low-TiO 2 rocks [7]. Relatively primitive samples of Faroese dykes and lavas fall on olivine control lines in plots of e.g. MgO versus TiO 2 [7]. Internal geochemical variations amongst relatively evolved basaltic rocks represented by local low-TiO 2 dykes have previously been interpreted in terms of low-pressure fractional crystallisation of olivine, plagioclase and clinopyroxene previously [47]. A scenario that also involved melting of a North Atlantic end-member mantle component (NAEM) was envisaged for Faroese basaltic lavas in a more recent study [8]. These authors further argued in favour of distinct sources for the low and high-Ti Faroese lavas. Most recently, a ne study identified four mantle sources to high-TiO 2 lavas of the Enni and Malinstindur formations and an additional source to low-TiO 2 lavas of the same formations [9].
Compositional variations in high-TiO 2 versus low-TiO 2 mantle sources have also been inferred for E Greenland [18][19][20] and Iceland [49]. Other authors have argued in favour of a depleted asthenospheric source to local LREE-depleted (low-TiO 2 ) basalts and a sub-continental lithospheric mantle or a deep mantle plume source to local LREE-enriched (high-TiO 2 ) basalts [23]. Coherent up-section variations in MgO, TiO 2 , Y and Zr compositions within rocks of the Beinisvørð Formation led some authors to infer magma supplies from at least two independent volcanic systems [45]. Combined geochemical and isotopic characteristics of local contaminated basaltic lavas (~54 wt% SiO 2 ) have previously been explained in terms of contamination with Lewisian-type amphibolite facies gneisses [7][8][9][10][11][12][13][14][15][16][17][18][19][20][21][22][23]. While attributed nonlinear isotopic trends of "common" tholeiitic Faroese lava/dyke samples could result from mild crustal contamination [23]; other authors have ascribed such insignificant variations to slight isotopic heterogeneities originating within their mantle sources [7]. Some authors have argued in favour of a trans-Atlantic chemostratigraphic correlation between units of the Nansen Fjord area in the central parts of E Greenland and sections of the Faroese lava pile [46][47][48][49][50]. However, a more recent study suggests that albeit the dominant progression of high-TiO 2 basaltic rocks in these regions appears broadly equivalent, these characteristics are not necessarily a direct indication of such a correlation, since Faroese high-TiO 2 lava flows can be chronostratigraphically constrained only on local (~10's km) scale [9]. They [9] argue that low-TiO 2 lava flows in these two regions could have been fed across a broader area of still relatively thick lithosphere associated with a strike-slip zone between the FIBG and Blosseville Kyst region of East Greenland.
The new data sets introduced in this paper present additional constraints on the nature of mantle sources and petrogenetic processes involved during the final known phases of magmatism in the Faroese region that ultimately resulted in the emplacement of sills within the volcanic successions. Our study provides both propositions regarding igneous processes acting during waning stages of NAIP formation and more general implications for LIP formation.

Petrography
Samples were collected from 7 Faroes sills, as well as from a number of selected local dykes and irregular intrusions in order to further constrain the petrogenesis of the magmas: Previous work [27][28][29][30][31][32][33][34][35][36][37][38][39][40][41] outline locations of sills and collected samples. Petrographically, the Faroese sills comprise ca. 40 to 55% plagioclase, 40 to 45% clinopyroxene, 5 to 10% olivine and 3 to 10% Fe/Ti oxides. More specifically, the Streymoy and Kvívík sills are feldsparphyric; the Langaregn Sill is feldspar-and olivine-phyric; the Eysturoy and Sundini sills display intergranular texture, but clinopyroxene oikocrysts are also common within these; the Svínoy-Fugloy and Morskranes sills both display ophitic to sub-ophitic textures. Late clinopyroxene crystallisation is suggested by the ophitic/subophitic texture in the Svínoy-Fugloy and Morskranes sills in particular, but also to some degree in the Eysturoy Sill. Early olivine crystallisation is suggested by the occurrences of partly altered large olivine phenocrysts in the Streymoy and Kvívík sills and by the presence of olivine microphenocrysts occurring as inclusions within larger plagioclase grains of the Langaregn Sill. The occurrences of plagioclase grains in feeder dykes, which display increasing degrees of resorption with increasing distances from contacts with their host-rocks, seem to suggest that Ca-rich plagioclases were dominating early crystallisation phases, which subsequently reacted with surrounding melts to produce fine grained Na-rich plagioclases and clinopyroxenes. Earlier studies on basalts of the Faroe Islands have indicated olivine compositions of Fo [86][87][88] in low-TiO 2 basalts and Fo 72-73 in high-TiO 2 basalts whereas An 64-70 has been measured for plagioclases representing both these rock types [7]. See Supplement 1 [51] for further details on petrography.

Analytical Methods
Following careful selection, 56 rock samples representing 7 Faroese sills and a number of local dykes and irregular intrusions selected for analysis, then crushed and milled according to standard methods in preparation for analysis. Major elements and selected trace elements representing 44 sill samples (Table 1) and 12 dyke and irregular intrusion samples (Table C in Supplement 3 [51] were determined on rock powder fused to glass discs and pressed powder tablets respectively using an ARL 8420+ dual goniometer wavelength dispersive XRF spectrometer at the Department of Earth Sciences, Open University, Milton Keynes, UK. Major elements representing 3 additional samples from the Sundini and Morskranes sills were analysed at the Geological Survey of Norway (NGU), using a PANalytical Axios 4 kW. For further details see Supplement 2 [51].
Trace elements and REEs representing 14 sill samples ( Table 2) and 8 dyke and irregular intrusion samples (Supplement 3 [51]) were determined on powdered rock samples were carried out on an Elan 6000 Perkin Elmer-Sciex inductively coupled plasma mass spectrometer (ICP-MS) at the Arthur Holmes Trace Element Laboratory, Department of Earth Sciences, Durham University, UK (for preparation and analytical methods see supplement 3 [51]). In addition, trace elements for a further three sill samples ( Table 2) were determined using a modified version of the dissolution procedure as described in [52][53]. These analyses were carried out at GEUS, Denmark, by inductively coupled plasma mass spectrometry (ICP-MS) using a Perkin-Elmer Elan 6100 DRC quadrupole ICP-MS instrument. For further details see Supplement 2 [51].
Sr and Nd isotope ratios for 8 selected sill samples and Pb isotope ratios of 7 selected sill samples (Table 3) were determined according to the preparation methods described in [54][55]. Sr and Nd fractions were taken up in 1ml of 3% HNO 3 separately and introduced into a mass spectrometer using an ESI PFA50 nebuliser and a dual cyclonic-Scott Double Pass spraychamber. Pb samples were taken up in 1ml of 3% HNO 3 and spiked with Tl prior to introduction into a mass spectrometer using an ESI PFA50 nebuliser and a dual cyclonic-Scott Double Pass spraychamber. All isotopes were analysed on a ThermoElectron Neptune Multi-collector Plasma Mass Spectrometer (MC-ICP-MS) at the Arthur Holmes Isotope Geology Laboratory, Durham University, UK.
Whole-rock MC-ICP-MS lead isotope analyses were carried out on 6 additional sill samples on a VG Sector 54-IT TIMS at the Geological Institute, University of Copenhagen, Denmark ( Table 3). Dissolution of the powdered samples was achieved in two successive, but identical steps, which consisted of a strong 8 N HBr attack that has been shown to effectively dissolve accessory phosphates [56][57], followed by a concentrated HF-14 N HNO 3 mixture, and finally by strong 9 N HCl. Chemical separation of Pb from whole rocks was performed over conventional glass stem and subsequently miniature glass-stem anion exchange columns containing, respectively 1 ml and 200 µl 100-200 mesh Bio-Rad AG 1×8 resin. Lead was analyzed in a static multicollection-mode where fractionation was controlled by repeated analysis of the NBS 981 standard using values in [58]. For further details see Supplement 2 [51].

Major Elements
The 47 Faroese sill samples analysed for major elements (  Figure 2). Generally, the compositional range of major elements representing the investigated intrusions generally correlate well with published major element data on Faroese lava/dyke samples and with lava/dyke data obtained in this work (Supplement 3 [51]); however, the Streymoy and Kvívík sills in particular display higher SiO 2 , Al 2 O 3 and CaO contents and lower TiO 2 and Fe 2 O 3 contents when compared to host-rock lava/dyke samples with comparable MgO contents (insets in Figure 2). More specifically, the published host-rock data show that most local basaltic lavas/dykes display SiO 2 contents ranging from ~47 to ~49.5 wt%, their MgO generally range from ~7 to ~9 wt% (their entire MgO range span from ~4.5 to ~23 wt%), while their TiO 2 contents range from ~0.6 to ~4.0 wt% [3,7,8,[45][46][47].      [59]. [7,8,45,46,47] and from this study (see

Selected Trace Elements
Mantle-normalised trace element data from 17 representative sill samples (Table 2; Figure 3) define three main tendencies. The low-TiO 2 and high-TiO 2 sills generally display separate well-defined, relatively flat sub-parallel trends with the latter being comparatively more enriched in most elements; by contrast, the intermediate TiO 2 content sill generally exhibits a gradual depletion towards the most incompatible elements (Figure 3a). All sills display negative P anomalies and some samples of the intermediate TiO 2 sill are characterised by distinctive positive and moderately negative Rb and K anomalies. The low-TiO 2 sill samples display relatively uniform positive Ba, Rb, Th, K and Sr anomalies, whereas most of the high-TiO 2 sill samples exhibit negative anomalies for these same elements ( Figure  3). It is clear however, that some relative differences in these elements exist internally within the high-TiO 2 sills, where the Eysturoy and Sundini sills display slightly larger Sr and slightly smaller Ba, Rb, Th and K anomalies respectively compared to those of the Svínoy-Fugloy and Langaregn sills ( Local high-TiO 2 and intermediate-TiO 2 host-rock samples (analysed in this study and shown in Supplement 3 [51]) display mantle-normalised trace element trends that are broadly similar to those of most sills with similar TiO 2 contents, but local low-TiO 2 host-rocks (analysed in this study too) exhibit trends that are considerably more depleted in incompatible elements relative to those of comparable low-TiO 2 sills (insets in Figure 3a; Figure 3b).

REE
Mantle-normalised REE data representing 17 sill samples (Table 2; Figure 4) define three main tendencies. The low-TiO 2 and high-TiO 2 sills (excluding the Langaregn Sill) display moderately negatively sloping LREE trends with (Ce/Sm) N ratios ranging from 1.11 to 1.27. The low-TiO 2 sills also display relatively flat negative HREE slopes with (Sm/Yb) N ratios ranging from 1.09 to 1.26. The samples representing the high-TiO 2 sills are characterised by steeper negative HREE slopes that can be further categorised into three sub-trends, where (Sm/Yb) N ratios for the Eysturoy and Sundini sills range from 1.59 to 1.61 and from 1.97 to 2.02 for the Svínoy-Fugloy Sill compared to 2.38 for the Langaregn Sill. The Morskranes Sill is LREE depleted with an average (Ce/Sm) N ratio of ~0.6, but displays a flat HREE trend with an average (Sm/Yb) N ratio of ~1. Local low-TiO 2 basaltic host-rock samples (analysed in this study) display relatively steep positive LREE and MREE slopes in contrast to the gentle negative slopes for these same elements representing comparable low-TiO 2 sill samples. Local high-TiO 2 host-rocks (analysed in this study and shown in Supplement 3 [51] display positive LREE slopes, which is slightly at odds with the relatively flat negative slopes for these same elements that represent comparable high-TiO 2 sill samples, when the Langaregn Sill is excluded (insets in Figure 4b; Figure 4e). The REE trends of the low-TiO 2 local host-rock samples analysed in this study are virtually identical to those of local MgO-rich (including picritic) low TiO 2 host-rock samples reported in earlier studies as outlined below, but are much more depleted in their LREE and MREE compared to the low-TiO 2 samples (Figure 4c). Intermediate-TiO 2 sill samples are almost identical to those of comparable local host-rocks (Figure 4d).
With respect to published REE data on local host-rocks, low-TiO 2 dykes (0. 75 [7]. Lava flows of the Beinisvørð Formation (exposed above sea level) and the Malinstindur Formation are LREE enriched ((La/Yb) N = 1.4-3.3) with a relatively wide range in overall REE concentrations, whereas some parts of the Enni Formation display depleted MORB-like LREE trends with (La/Yb) N = 0.45-0.62 [23].

Sr and Nd Isotopes
Assuming that all sills were emplaced broadly contemporaneously and are slightly younger than the ~55 Ma determined for the upper parts of the host lava pile, the eight samples from seven sills analysed for Sr and Nd isotopes were age corrected to 54 Ma (Table 3; Figure 5). The 87 Sr/ 86 Sr(t) ratios of the sills range from 0.7028 to 0.7033, while their 143 Nd/ 144 Nd(t) ratios range from 0.512910 to 0.512976 (Nd(t) range from 6.7 to 7.9). When displayed graphically, these isotope data define a negative slope, plotting in two main clusters with two samples of the intermediate-TiO 2 sill displaying a 87 Sr/ 86 Sr(t) range from 0.7028 to 0.7029 and a 143 Nd/ 144 Nd(t) range from 0.512966 to 0.512976 compared to ranges of 0.7032 to 0.7033 and 0.51291 to 0.51296 respectively for the samples representing low-TiO 2 and high-TiO 2 sills (Figure 5a). These Sr and Nd isotope ranges plot within ranges for similar isotopes reported for samples representing other basaltic rocks of the Faroe Islands, and partly within those Sr and Nd isotopic ranges reported for Icelandic samples (inset in Figure 5a).

Element Mobility
The reliability of geochemical data as petrogenetic indicators must be evaluated carefully, as igneous rocks exposed at the Earth's surface commonly experience postmagmatic mineral break-down and associated mobilisation of major and trace elements, in addition to recrystallisation [65]. Major elements like Si, Mg and K and large ion lithophile elements (LILE) may be mobilised by low-grade metamorphism or weathering [66][67][68], while high field strength elements (HFSE) such as Th, Nb, Ta, Zr, Y and Ti commonly remain relatively unaffected during low or moderate grades of metamorphism [66][67][68].
Some of the investigated Faroese sill samples display evidences of element mobilisation at the microscopic (single grain) scale, including minute olivine grains being partially or entirely altered to phyllosilicates (see details on petrography in Supplement 1 [51]. Greenish coatings occurring in <0.5 mm wide joints represent low-temperature hydrous minerals, and suggest element mobilisation at the whole-rock scale within some parts of these intrusions. For instance, comparison between the slightly jointed sill sample 07-JSS-38 and another fresh/intact sill sample 09-JSS-02 (Table 1), collected less than 100 metres apart within the Streymoy Sill, illustrates these effects of element mobilisation and geochemical modification. Sample 07-JSS-38 is relatively depleted in Si, Al, K, Sr and Ba and relatively enriched in Mg, Fe, Ti and Y. A relative enrichment in Ti and Y thus may indicate the relative immobility of Ti and Y in the more weathering-resistant mineral phases such as Fe-Ti oxides and/or clinopyroxene. Accordingly, for the purposes of petrogenetic investigation, only samples without signs of element mobilisation have been utilised in this work.
Since no geochemical data are available for the continental basement underneath the Faroese lava succession, samples representing Proterozoic/Archaean continental basements from NW Britain, the Rockall Plateau and E Greenland were utilised to evaluate potential crustal contamination of precursor melts to Faroese basaltic rocks. In theory, less than 4-5% bulk assimilation of material similar to average basement composition for E Greenland, NW Britain and the Rockall Plateau would be required in order to shift (Zr/Nd) N and (Nb/Ce) N ratios from those of typical N-MORB and E-MORB to those representing intermediate-TiO 2 and high-TiO 2 Faroese sills respectively ( Figure 6). Substantially less assimilated basement material would be required to account for the internal variations in these elements displayed by the Faroese low-TiO 2 and high-TiO 2 sills. Around 5% crustal assimilation by Faroese sills appears quite unrealistic based on bulk geochemistry and would become apparent from additional isotopic evidences. The pronounced positive Rb and K anomalies displayed by one sample of the Morskranes Sill (Figure 3a; Figure 3c) seem to indicate selective enrichment in these elements. In theory, K-rich material could originate from residual mantle melts, or else from small-scale melting of suitable sources; but such melts also commonly tend to be SiO 2 and Na 2 O enriched. Given the relative depletion of these two major elements within this sample ( Table 2) such scenario is unlikely. Hence, potential enrichment candidates may include net assimilation of crustal materials or perhaps LILE-rich metasomatic fluids leached from other adjacent basalts (i.e. secondary processes like the infilling of vesicles), which would also be in accordance with negative Rb and K anomalies displayed by many Faroese sill samples of all TiO 2 compositions (Figure 4).

Isotopic Constraints on Potential Crustal
Contamination The Pb isotopic compositions of presumed uncontaminated Faroese basaltic rocks compared with those of local contaminated basaltic lavas, of the Enni Formation, indicate involvement of two contamination sources characterised by different 208 Pb/ 204 Pb ratios in particular, but also with various 207 Pb/ 204 Pb ratios ( Figure  7b, Figure 7c; Figure 7d). While the low 208 Pb/ 204 Pb contaminated samples could owe their Pb isotopic characteristics to assimilation of materials comparable to those reported for felsic and intermediate granulites or amphibolitic gneisses from E Greenland and NW Britain, the contaminated basalts with higher 208 Pb/ 204 Pb ratios most likely assimilated felsic or intermediate materials with Pb isotopic compositions corresponding to those of basement samples reported for E Greenland (Figure 7b, Figure 7c; Figure 7d). If precursor melts to the contaminated Faroese basalts (silicic basalts) originally possessed Sr and Nd isotopic compositions similar to those of the present-day MgO-rich Faroese dykes (e.g. Basalt Formation from Distinct Melting Regimes sample Sv-12 in [7]), then between 10 and 20% assimilation of materials similar to representative Proterozoic/Archean continental basement samples from E Greenland or NW Britain would be required in order to modify them to the isotopic range currently measured for these contaminated basalts (Figure 7a).

5% of average E Greenland and/or NW Britain basement sources to relatively primitive Faroese basalts could in theory explain some of the Sr and Nd isotopic variations of Faroese sill samples (inset). (b), (c) and (d) Measured Pb isotopic ratios. Two distinct contamination sources are required in order to explain the configuration of contaminated silicic basalts versus average Faroese basalts namely, one purely E Greenland-like basement source and one E Greenland-like and/or NW Britain-like basement source. Arrows point towards Pb isotope compositions of their probable contamination sources. Contamination with materials similar to those of the Rockall Plateau and those of E Greenland could explain variations of Pb isotope ratios of some samples of the Morskranes Sill, if these heterogeneities were not inherited from their mantle source (insets). Basement data for E
Greenland are from [72,83,84]; NW Britain basement data are from [70,76,77,85,86,87]; Rockall Plateau basement data are from [78]. Datasets on calculated mixing trends in a) are shown in [79].
At present, it remains un-clear whether the noticeable variations in 207 Pb/ 204 Pb ratios between some samples of the small Morskranes Sill, as well as those within the Eysturoy and Sundini sills (Figure 5b, Figure 5d), reflect characteristics inherited from their mantle source(s), or if contamination with crustal materials is the most likely source. Indeed, noticeable differences in Pb isotopic compositions between melt inclusions and crystals versus their host phenocrysts and matrix respectively (thought to reflect isotopic heterogeneities at short length scales in their mantle sources) have been measured at a number of sites worldwide [80][81]. However, it is noticeable that samples of the Morskranes Sill with decreasing 207 Pb/ 204 Pb ratios also display increasing MgO contents (e.g. samples: 08-JMS-14 → 16-JMS-18 → 08-JMS-16 display increasing MgO contents and decreasing 207 Pb/ 204 Pb ratios in direction of arrows), i.e. Table 1; Table 3.
If assimilation indeed modified the Pb isotopic composition of the Morskranes Sill in particular and (to a lesser degree) that of the Eysturoy and Sundini sills, minor involvement of crustal materials with Pb isotopic compositions comparable to those of basement samples of the Rockall Plateau and of E Greenland could be an explanation (inset in Figure 7b). In this context, it is worth noting that isotopic compositions of Rockall granites led [82] to conclude that there existed an isotopic/genetic relationship between Rockall granites and alkali-rich intrusives of E Greenland. The inset in Figure 7d does not conclusively point to any particular of the three potential contamination sources, discussed in this sub-Sect.
If variations in Sr and Nd isotopic ratios of Faroese sill samples reflect assimilation of crustal material into their precursor melts, then less than 0.5% contamination with basement materials comparable to those of E Greenland and NW Britain would be required in order to shift the Nd isotopic compositions from those of the Faroese LREE depleted MgO-rich lava/dyke samples to those of LREE depleted samples of the Morskranes Sill (inset in Figure  7a). In theory, the Sr and Nd isotopic variations within high-TiO 2 and low-TiO 2 sills of this study, as well as Sr and Nd isotopic differences between high-TiO 2 and low-TiO 2 sills versus intermediate-TiO 2 sill samples, could have been generated by ≤ 0.25% assimilation of materials comparable to those of Proterozoic/Archean continental basement material from E Greenland, Rockall Plateau or NW Britain (e.g. Figure 7a including inset).
To summarise, the isotopic compositions of the contaminated silicic basalts of the Faroe Islands together with some samples of the Morskranes Sill in particular (if differences in their isotopic signatures are not source related) suggest that at least three distinct crustal contamination sources could have affected their precursor melts during ascent through the local palaeo crust ( Figure  5; Figure 7).

Implications from Mixing Calculations
Calculations (not shown) show that basaltic rocks with initial Pb concentrations of ~0.5 ppm and initial average 208 Pb/ 204 Pb ratios of ~38 that experience ~0.5% assimilation of basement material with Pb concentrations of ~15 ppm and 208 Pb/ 204 Pb ratios being ~5 higher and lower respectively than those of the target basalts (i.e. 208 Pb/ 204 Pb ratios of ~43 and ~33 respectively) will experience an increase and decrease of ~0.65 in their 208 Pb/ 204 Pb ratios to ~38.65 and ~37.35 respectively. Accordingly, contributions with fractions of a percent of high or low 208 Pb/ 204 Pb basement materials comparable to some of those reported for E Greenland (e.g. samples KS 60 and 229642 in [72,84] respectively) to melts comparable to Faroese dykes or lavas [7,8] could account for the poor correlation between their 208 Pb/ 204 Pb ratio plots. The same would apply for basaltic rock samples from Iceland and E Greenland (e.g. insets in Figure 5c; Figure 5d). A relatively recent study based on geochemistry and isotopes suggested that low-TiO 2 tholeiitic Faroese lavas with the highest probabilities of contamination would require less than 1% contributions with crustal components if they assimilated local continental material [8]. Similar arguments on potential contamination in low-TiO 2 basalts of E Greenland have been inferred earlier [18]. Collectively, the results of this sub-Sect. suggest that parts of the Morskranes Sill could have been affected by crustal contamination, while it is somewhat more uncertain if the high-TiO 2 and low-TiO 2 sills were affected as well. Small-scale crustal contamination of these two latter sill categories would probably have resulted in slightly elevated La and Ce values, but assimilation of ≤ 1% crustal components would hardly be detectable in normalised REE diagrams [41] and cannot account for the observed differences between LREE in sills versus their host-rocks with similar TiO 2 compositions (e.g. Figure 4).

Constraints on Faroese Tholeiite Formation from Fractional Crystallisation
MgO-rich primary basaltic melts typically evolve to less magnesian varieties by fractional crystallisation of olivine [88]. Further melt modifications may occur in response to e.g. plagioclase, clinopyroxene or garnet fractionation . Basaltic magmas can also evolve in response to complex RTF processes   Hence, it is not likely that clinopyroxene was a dominating fractionating phase during early stages of magmatic evolution of melts that gave rise to most Faroese sills, if the geochemical compositions of their primary melts resembled those of local picrites. Moreover, the similarities in Sc concentrations between sills and local picrites also argue against any noticeable fractionation of garnet, provided that the picrites and primary melts that developed to the Faroese sills displayed matching compositions.

Trace Element Constraints on Fractional
Crystallisation and Assimilation While the pronounced positive Sr anomalies displayed by samples of the low-TiO 2 Streymoy -Kvívík sills and the likewise conspicuous negative Sr anomalies of the high-TiO 2 Eysturoy -Sundini sills (Figure 3) clearly indicate the involvement of plagioclase at some point during melt evolution, there remains a slight uncertainty regarding the reliability of the weak Eu anomalies of samples from these same sills (Figure 4) given the small sizes of these compared to analytical error. Olivine fractionation and accumulation from/to modelled primitive basaltic melts (partition coefficients and the equation used are shown in Supplement 5 [51]) do not result in noticeable changes of their Sr/Sr* and Eu/Eu* ratios (definitions outlined in caption to Table 2), but around 15 wt% fractionated olivine in combination with minor amounts of fractionated magnetite and ilmenite would increase their Nb, Ta, Er and Y concentrations slightly. Similar calculations involving clinopyroxene fractionation from similar melts would not affect their Eu/Eu* ratios and would only result in very minor changes in their Sr/Sr* ratios. Addition of ~25 wt% plagioclase from external sources to magmas, generated by ~20% mantle melting (modelled), could explain much of the span in Sr/Sr* and Eu/Eu* ratios displayed by the low-TiO 2 Streymoy and Kvívík sills, while around 20 wt% plagioclase fractionation from magmas, generated by 10 to 12% mantle melting (modelled), could explain much of the range in the Sr/Sr* and Eu/Eu* ratios displayed by the Eysturoy and Sundini sills (Table 2; Figure 8). Plagioclase fractionation in the range 10 to ~20 wt% from magmas formed by ~10% mantle melting are required in order to recreate the Sr/Sr* and Eu/Eu* ratios of the Svínoy-Fugloy and Langaregn sills (Figure 8). The differences in calculated plagioclase accumulation and fractionation required to recreate Sr/Sr* versus Eu/Eu* ratios representing the Faroese sills, as shown in Figure 8, could in theory stem from either analytical error during analyses of Faroese sill samples, or the actual Sr/Sr* and Eu/Eu*ratios of these elements in their respective mantle sources differed from those used in the modelling. Around 20 wt% plagioclase accumulation and fractionation to/from liquids with geochemical compositions broadly similar to those of Faroese low-TiO 2 and high-TiO 2 sill samples would decrease/increase their overall REE concentrations respectively by amounts corresponding to around 2% larger/lesser degrees of partial melting of their respective mantle sources, while ~10 to ~15 wt% olivine fractionation would increase their overall REE concentrations by amounts corresponding to around 1% lesser degree of partial mantle melting (calculations not shown).  [93]). Residual minerals were the same as in a). See Table 2 for definition of Sr/Sr* and Eu/Eu*. Mineral abbreviations are from [94]. Numbers along dotted lines indicate melting percentages and numbers along dashed lines indicate plagioclase fractionation/accumulation percentages. Datasets on calculated partial mantle melting, calculated fractionation and calculated accumulation of plagioclase are shown in [79].

Constraints on Fractionation/Accumulation from Mass-balance Calculations
Mass-balance calculations, involving all major elements apart from MnO and P 2 O 5 , suggest that fractionation and accumulation of plagioclase from/to basaltic melts with compositions comparable to Faroese low-TiO 2 and high-TiO 2 sills chiefly affect their Al 2 O 3 , Fe 2 O 3 and MgO compositions and to some degree their CaO and TiO 2 compositions, while the other major elements remain relatively unaffected (Table  4). If low-TiO 2 sills of this study evolved by ~25 wt% plagioclase accumulation, their precursor melts would possess significantly lower Al 2 O 3 and higher Fe 2 O 3 contents relative to values measured for these sills, while evolvement of their high-TiO 2 counterparts by ~20 wt% plagioclase fractionation would produce parental melts being significantly enriched with respect to their Al 2 O 3 and depleted with respect to their Fe 2 O 3 contents relative to values measured for these particular sills (Table 4). Consequently, in contrast with measured relative abundances of Al 2 O 3 and Fe 2 O 3 in Faroese high-TiO 2 sills versus low-TiO 2 sills, which define positive and negative slopes respectively when connected and plotted against MgO, the relative abundances of these major elements in their calculated parental melts would define negative and positive partial melting slopes respectively when connected and plotted against MgO (Figure 9a; Figure 9b). The scenario is different for the TiO 2 and CaO contents of the same samples, when exposed to the same accumulation/fractionation calculations as discussed above. Both these major elements in the calculated parental melts maintain their original graphical trends (i.e. negative and positive slopes respectively) when connected and plotted against MgO, but the gradients for their calculated parents are gentler than those of the measured samples (Figure 9c; Figure 9d; Table  4). Configurations of Al 2 O 3 , Fe 2 O 3 , TiO 2 and CaO versus MgO in calculated parents to low-TiO 2 and high-TiO 2 sills resemble results from previous experimental studies on partial mantle melting to produce low-TiO 2 and high-TiO 2 basaltic rocks [95][96][97] and would also support inferences in the discussion on trace elements above. Insets in Figure 9 illustrate similarities between calculated high-TiO 2 versus low-TiO 2 melts of this study and experimental basaltic melts, produced at roughly 5 to 20% melting and extrapolated to lower MgO contents in response to 12-15 wt% olivine fractionation. The differences in slopes and concentrations of experimental versus calculated melts are due to low pressures (1 GPa) and the hydrated nature of the experimental mantle materials. The discussion above strongly suggests that noticeable accumulation and fractionation of plagioclase in mid crustal staging chambers affected the melts, which eventually evolved to low-TiO 2 and high-TiO 2 Faroese sills. A scenario where plagioclase, originating from different sites of storage and differentiation during magma ascent, accumulated in magmas, which subsequently evolved to the Faroese low-TiO 2 sills, would be in accordance with earlier studies on basaltic magmas of the NAIP [25]. More specifically, [98] suggested that three criteria must be fulfilled in order for melts containing accumulated plagioclase to reach the upper crust: 1) the initial plagioclase-free melt must pass through conduits in which plagioclase cumulates are present; 2) the ascent velocity of the magma within the conduit must be greater than the settling velocity of the entrained phenocrysts and 3) the magma must not travel through a conduit system containing an axial magma chamber, which would halt the upward ascent of the magma and allow denser plagioclase crystals to segregate.  Table 4 [96], with melt percentages spanning from ~5 to ~20%. It is assumed that olivine fractionation shifted/extrapolated MgO contents in both calculated and experimental primary melts roughly horizontally to lower MgO values, as displayed in this figure. See main text and Table 4 for more details.

Constraints on Depths of Formation by Partial
Melting Since the Faroese archipelago rests on top of an Archaean continental crust, the melts that gave rise to their olivine tholeiites must have formed beneath an ancient microcontinent [33][34]. Such olivine tholeiitic basaltic magmas may form by partial melting of a range of mantle compositions under various T and P [95,97,99,100], but the lack of HREE depletion in Faroese low-TiO 2 and most high-TiO 2 sill samples points to formation of their precursor melts by partial melting outside the garnet stability field, i.e. at P ≤ ~2.8 GPa corresponding to depths of ≤ ~85 km. Moreover, as melting of plagioclase-bearing mantle material would produce quartz tholeiitic basalts [99][100], the olivine tholeiitic sills of the Faroe Islands probably formed at depths outside the plagioclase stability field too, i.e. at P ≥ ~0.9 GPa corresponding to depths of ≥ ~30 km [102][103], thus leaving mantle sources within the spinel stability field as their most likely origin.

Batch Melting Calculations
Batch melting calculations (partition coefficients and the equation used are shown in Supplement 5 [51]) were carried out in this work, in order to estimate melting percentages of suitable mantle sources, which gave rise to the different types of melts that ultimately evolved to the Faroese sills. In these calculations, only realistic existing figures for mantle compositions are used, which have been reported for various sites worldwide previously. In addition, shallow mantle sources of late melts are considered, as hypothesized in earlier studies [104].
Calculations in this study suggest that partial melting of slightly LREE enriched mantle sources have the potential to reproduce the REE trends characterising Faroese low-TiO 2 and high-TiO 2 sills, while LREE and MREE depleted mantle sources are required in order to recreate the REE trend(s) of the intermediate-TiO 2 Morskranes Sill. Residual mineral assemblages utilised in these calculations are dominated by olivine (~80 to ~84%) and orthopyroxene (~8 to ~16%) that may be associated with minor amounts of clinopyroxene (0 to ~7%) and spinel (0 to ~1%), i.e in accordance with the mineralogy of many naturally occurring peridotites worldwide [105][106][107][108].   [60]. Mineral abbreviations are from [94]. See Supplement 5 [51] for partition coefficients and equation used in calculations and Supplement 6 [51] for details on mineral residues from calculations shown in (d), as well as similar calculations on the same sill samples using a Zr versus Y/TiO2 diagram. Datasets on calculated melting trends are shown in [79].
Trial partial melting calculations, performed in order to recreate measured REE trends of low-TiO 2 and most high TiO 2 sills, were carried out on a large number of mantle compositions from localities worldwide and from estimated average mantle values and include published material reported by e.g. [60,93,106,107,109,110]. Melting percentages ranging from 15 to 20% of fertile mantle material from [110] and ranging from 8 to 10.5% of likewise fertile mantle material from [93] best reproduce the ranges in REE compositions of low-TiO 2 and high TiO 2 sills respectively (Figure 10a; Figure 10c). It is clear that the measured REE trend representing the Langaregn Sill doesn't quite fit in with those of the rest of the high-TiO 2 sills ( Figure  4e; Figure 10c). Its actual REE trend can be recreated by ~7% partial melting of a fertile source (not shown), but it is likely that some garnet was a residual phase and that its source displayed slight depletion with respect to LREE when compared to sources to other local high-TiO 2 sills. Regarding measured (LREE depleted) REE in intermediate-TiO 2 sill samples, trial calculations included published material reported by e.g. [111][112][113], where 7 to 8% of average depleted mantle from [112] best recreate the REE Basalt Formation from Distinct Melting Regimes compositions of these samples (Figure 10b).
When the effects on REE compositions representing Faroese sills from supposed fractionation of olivine and plagioclase and accumulation of plagioclase (previous Sect.) are taken into account, partial melting percentages to produce melts that gave rise to the Faroese sills should be corrected slightly. With respect to the low-TiO 2 sills, olivine fractionation and plagioclase accumulation would shift the range of partial melting percentages in their source(s) by +1% from 15-20% to 16-21. With respect to the high-TiO 2 sills, olivine and plagioclase fractionation would shift the range of partial melting percentages in their source(s) by -3% from 8-10.5% to 5-7.5%, while the calculated melting for the Langaregn Sill should be shifted from ~7 to ~4%. With respect to the intermediate-TiO 2 sill, olivine fractionation would shift the range of partial melting percentages in its source by -1% from 7-8% to 6-7%.
Batch melting calculations based on selected REE, which are plotted in a binary (Yb) N versus (Ce/Sm) N diagram, are perhaps better suited to illustrate the effects of various residual mineralogies and general differences in source compositions on modelled melts. With respect to LREEenriched mantle sources [93][94][95][96][97][98][99][100][101][102][103][104][105][106][107][108][109][110], the results shown in the binary REE plots support the inferences above regarding formation of the low-TiO 2 and high-TiO 2 sills by different degrees of partial melting of enriched sources that left residual mineralogies composed mainly of olivine and orthopyroxene (curves i-iv in Figure 10d). Similarly, partial melting of moderately LREE-depleted sources [112][113] with residual mineral assemblages dominated by olivine and orthopyroxene most realistically reproduces the REE composition of the intermediate-TiO 2 Morskranes Sill (curves v and vi in Figure 10d). In addition, strongly LREE depleted mantle sources [111] do not have the potential to reproduce REE compositions matching those of the Morskranes Sill, when exposed to partial melting, irrespective of which residual mineral assemblages are chosen (curves vii-ix in Figure 10d). More details on Figure  10d as well as additional modelling are shown in Supplement 6 [51].

General Considerations
Given that the mantle sources used in the partial melting calculations in this study were primarily selected because they best reproduce the REE trends representing the actual sills, the actual mantle sources to these sills could, in theory, have been less or more fertile relative to the chosen compositions; if so, then the outcome would require slightly lesser or greater degrees of partial melting relative to the calculated values. In cases where precursor melts to the actual sills experienced small-scale assimilation of felsic crustal materials upon ascent, it could have increased their LREE concentrations very slightly [41]. However, contamination with ≤ 1% crustal material would not have had noticeable effects on the modelling carried out in this work. Altogether, the partial modelling and trace element characteristics of actual sill samples combined (Figure 3; Figure 10) point to distinct mantle sources for intermediate-TiO 2 versus high-TiO 2 and low-TiO 2 sill samples. In turn, some heterogeneities probably exist between sources to high-TiO 2 and low-TiO 2 sill samples and probably also to some degree within sources to the high-TiO 2 sills themselves. Here the Eysturoy and Sundini sills, the Langaregn Sill and the Svínoy-Fugloy Sill respectively probably originated from three slightly different mantle sub-sources. These findings are in accordance with the isotopic characteristics recorded for the actual sills ( Figure 5). Interestingly, four slightly different mantle sub-sources were inferred for high-TiO 2 lavas of the Faroese Enni and Malinstindur formations in a recent study [9].
Processes involving partial melting of sources, comparable to those reported for SCLM, to produce Faroese sills, as is suggested by our modelling above, are at odds with previous theories on formation of their host-rocks [7, 8, 46,].

Primary Melts
Previous estimates on the MgO percentages present in primary magmas that evolved to basaltic rocks of the N Atlantic have come to various conclusions: 10.0-13.0 wt% and 17.0-18.5 wt% respectively for W Greenland ; 12.0-13.6 wt% and 16.6 wt% for E Greenland ; 13.0-15.0 wt% for NW Britain [122]; 13.5-17.7 wt% for Iceland and adjacent mid-ocean ridges [121] and 16.0-19 wt% MgO for the Faroe Islands [7][8]. MgO contents of 10-15 wt% have been calculated for primary magmas that gave rise to average global Ocean Ridge Basalts [119].
Based on the results from the partial melting modelling and presumed subsequent fractional crystallisation of olivine and plagioclase, as well as inferred plagioclase accumulation, melting percentages of ~16 to 21 and ~4 to 7.5% to produce primary melts that developed to Faroese low-TiO 2 and high-TiO 2 sills respectively appear to be reasonable estimates, if the following criteria are fulfilled: 1) fractionation of clinopyroxene did not play a major role during the early stages of their magmatic evolutions; 2) around 10 to 15 wt% olivine and minor amounts of magnetite and ilmenite fractionated from their primary melts, which contained ~15 wt% MgO and 3) ~15 to 20 wt% plagioclase fractionated from melts that gave rise to high-TiO 2 sills while about 15 to 25 wt% plagioclase accumulated in melts that evolved to the low-TiO 2 sills. The 16 to 21% partial melting range to produce precursor melts to the Faroese low-TiO 2 sills correspond roughly to earlier estimates on melting percentages for primary melts that developed to low-TiO 2 basaltic rocks of W Greenland [15] (Larsen and Pedersen, 2009), E Greenland [17], the Faroe Islands [7], southern Brazil [120] and global ocean ridge basalts [119]. The 4 to 7.5% partial melting range proposed for production of Faroese high-TiO 2 sills correspond roughly to melting percentages estimated for primary melts, which evolved to high-TiO 2 basaltic rocks of southern Brazil [120] and E Greenland [17], but reach slightly higher values than those estimated for primary melts that gave rise to other high-TiO 2 basalts of the Faroe Islands [7] and are slightly lower than those inferred for global ocean ridge basalts [119]. However, as most sills of this study, being termed high-TiO 2 basaltic rocks, display TiO 2 compositions of only 2.0 to 2.5 wt% compared to 2.5 to 3.7 wt% for other high-TiO 2 basalts of e.g. the Faroe Islands [7][8], the proposed range in melting percentages for high-TiO 2 primary magmas in this study appear to be rather similar to other high-TiO 2 rocks within the N Atlantic area after all.
The 6-7% melting range calculated/estimated for the LREE depleted primary melts giving rise to the Morskranes Sill contrast somewhat with the higher degrees of melting thought to have generated comparable LREE depleted Central Mull Tholeiites from the British Tertiary Igneous Province (BTIP) and with the melting percentages measured for experimentally produced MORB-like rocks [100][101][102][103][104][105][106][107][108][109][110][111][112][113][114]. It should be noted however, that the basalts reported by these authors presumably formed at relatively low pressures, e.g. 1.5 GPa for the experimental rocks [100]. If the melts that evolved to the Morskranes Sill also experienced 10-15 wt% plagioclase fractionation, as could perhaps be suggested by its Eu/Eu* ratio (Figure 8b), then slightly higher range in melting percentages of 7-9% could be a more correct approximation. All the above estimated melting percentages rely on the assumption that crustal contamination did not significantly affect the overall trace element compositions used in the modelling.

Geochemical Implications from the Literature
Depleted mantle materials probably represent residue(s) following earlier phases partial melting of primordial mantle reservoirs to produce basaltic melts [111,112,115,116]. By contrast, fertile mantle materials could result from metasomatic processes, where primordial mantle materials were contaminated with ascending low-degree basaltic magmas or with fluids expelled from these [93,115,117,118], or they could originate from assimilation of recycled oceanic crust [2,88,97,121,123]. Total normative Fe-oxide values of ≥ 13 wt% in basaltic rocks are at times interpreted in terms of derivation from mantle sources contaminated with recycled oceanic crustal components [97]. The Faroese high-TiO 2 sills display average measured Fe 2 O 3 values of ~14.7 wt%, but their calculated parental melts contain less than 12 wt% of this oxide, compared to average measured Fe 2 O 3 values of ~10 wt% for the Faroese low-TiO 2 sills and ~12.8 wt% for their calculated parental melts (Table 4; Figure 9b). The iron contents of the actual sills also plot well within ranges displayed by common mid-ocean ridge and ocean island tholeiites [124].
Previous studies have attributed the trace element characteristics displayed by Faroese basaltic lavas and dykes to an origin from a heterogeneous mantle plume comprising distinct geochemical zones [7][8], while asthenospheric and/or SCLM sources have been inferred for Faroese lavas by other authors [23]. Based chiefly on Zr, Nb and Y concentrations, [125] indicated an enriched plume component, a depleted astenospheric component and minor contributions from an additional EM 1-like component, entrained within the asthenosphere, as the main sources to Icelandic basalts. [126] demonstrated that an additional local depleted Iceland plume (DIP) component was required in order to adequately account for the Zr, Nb and Y characteristics of the Icelandic basalts combined. In theory, mantle sources broadly similar to those suggested by [125][126] for the Icelandic basalts could have been involved during formation of basaltic rocks of the Faroe Islands, as samples representing Faroese sills, lavas and dykes share many of the geochemical characteristics defined by the Icelandic basalts (Figure 11a).
Importantly, it is evident from the inset to Figure 11a that all Faroese low-TiO 2 and most high-TiO 2 sill samples combined define a "missing link" between fields outlining the bulk of Faroese low-TiO 2 and high-TiO 2 host-rocks, whilst the intermediate-TiO 2 Morskraness Sill and the high-TiO 2 Langaregn Sill plot within the fields for low-TiO 2 and high-TiO 2 host-rocks respectively. Overall, the bulk of Faroese basaltic rocks, including the local sills, define a trend quite comparable inn shape to average values reported for Iceland previously (i.e. the bold curve in Figure 11a, adopted from [125]), but which is relatively rotated by around 10° clockwise.

Geochemical Implications on Metasomatised Mantle
Sources Negative Nb and Ta anomalies in basaltic rocks have commonly been associated with mantle sources that were metasomatised previously with hydrous fluids, C-bearing fluids or low-degree basaltic melts, while positive Nb and Ta anomalies are commonly associated with residual dryer mantle materials from which the above-mentioned metasomatic agents were extracted previously . Negative Nb and Ta anomalies in suites of low-TiO 2 flood basalts from the Siberian Traps, and from Central Nicaragua, being virtually identical to low-TiO 2 sills from this study; were previously attributed to an origin from metasomatised upper mantle sources in subduction-zone environments [10][11][12], while negative Nb and Ta anomalies in low-TiO 2 basaltic rocks from the Emeishan Igneous Province, SW China, are thought to reflect partial melting of enriched SCLM materials [11]. Basaltic rocks of the Emeishan Igneous Province with higher TiO 2 contents that display positive Nb and Ta anomalies are interpreted in terms of partial melting of mantle sources being significantly affected by plume-derived components [11], while relatively high-TiO 2 basaltic rocks from Central Nicaragua possessing positive Nb anomalies are thought to have developed from a mantle wedge comparable in composition to an enriched sub-oceanic mantle and unaffected by subduction zone processes [10].
Therefore, if the moderately negative Nb and Ta anomalies displayed by the Faroese low-TiO 2 sills resulted from melting of mantle sources affected by metasomatism, the moderately positive Nb and Ta anomalies of the high-TiO 2 sills ( Figure  3) could indicate magma tapping from dryer Nb and Ta enriched residual mantle sources, from which metasomatic agents were already extracted. The discussion above, suggesting formation of the Faroese low-TiO 2 sills by higher degrees of partial melting relative to the melting percentages required to produce their high-TiO 2 counterparts, would be in accordance with more metasomatised materials in sources to the former sills, as the presence of metasomatic agents supposedly enhance mantle melting [96,104,128]. However, as all the investigated Faroese high-TiO 2 sills, except for the Langaregn Sill, are LREE enriched (Figure 4), the formation of these sills by low-degree melting of a relatively dry Nb and Ta enriched source, which had already experienced a partial melting event, may appear to be problematic, if no other processes were involved, as the LREE within their source rocks would have been strongly partitioned into the first melts, leaving a relatively LREE depleted residue. Refertilisation of relatively dry mantle sources to the high-TiO 2 Faroese sills by other highly LREE enriched metasomatic agents could be a plausible explanation and would be in accordance with geochemical modification of upper mantle sources by multiple metasomatic events, as reported for other igneous regions previously [129]. Based on the gentle REE slopes of most high-TiO 2 and low-TiO 2 sill samples and the modelling above, in addition to their Nb and Ta characteristics, we propose that these sills originated from melting of variously metasomatised mantle materials with compositions comparable to those reported for SCLM materials previously. Formation of basaltic rocks by melting of such materials would fit tentative suggestions by [23] regarding petrogenesis of some of the Faroese lavas and would also fit inferences by [130] regarding origin of alkaline lavas from W Greenland, but would be at odds with earlier theories favouring magma supplies to Faroese lavas/dykes from a deep-rooted mantle plume [7][8].

Geochemical Implications on Heterogeneous Mantle
Sources While differences between some major elements of the low-TiO 2 Faroese sills in particular and local lavas/dykes (insets in Figure 2) could reflect post-melting processes where plagioclase was accumulated in precursor melts to the actual sills, some of the observed differences in trace element characteristics between data for Faroese lavas/dykes and those of the investigated sills (insets in Figure 3; Figure 4; Figure 11) most likely point to differences between their respective mantle sources.
The higher concentrations of LREE in high-TiO 2 sills and higher concentrations of LREE and MREE in low-TiO 2 sills relative to those of older local basaltic dykes and irregular intrusions with exactly similar TiO 2 contents (insets in Figure  4) could indicate a temporal enrichment of these elements in the local upper mantle in Early Cenozoic times, if all these intrusions originated at broadly similar mantle depths. Alternatively, the Faroese sills formed at relatively shallower depths from distinct mantle sources that may or may not have been geochemically affected/metasomatised by the earlier melting to produce melts parental to local dykes and irregular intrusions at deeper mantle levels. The scenario where the enrichment of LREE in high-TiO 2 sills and enrichment of LREE and MREE in low-TiO 2 sills relative to REEs in their host-rock counterparts with similar TiO 2 contents (i.e. Figure  4) could result from small-scale crustal contamination is not realistic, as this would require the contribution from at least a few percent of crustal materials [41]. The steeper HREE slopes in Faroese high-TiO 2 dykes and irregular intrusions relative to those of most high-TiO 2 sills could indicate some residual garnet or garnet fractionation during their formation, suggesting that they may indeed have developed at mantle levels deeper than those of the sill sources.
Evolution of the Faroese high-TiO 2 sills by RTF processes from local low-TiO 2 magmas, as suggested for Early Cenozoic basalts of e.g. W Greenland [15], could explain many of the differences in trace element concentrations between these two basalt groups. However, as Nb and Ta are not noticeably fractionated relative to e.g. La and Ce during common rock-forming processes, it is not likely that the moderately negative Nb and Ta anomalies in the Faroese low-TiO 2 sills developed to the moderately positive Nb and Ta anomalies, characterising many of the Faroese high-TiO 2 sill samples, by RTF processes.
All things considered, it is tentatively suggested that the compositional characteristics of Faroese low-TiO 2 and high-TiO 2 sills developed in response to tapping of fertile SCLMlike sources affected by metasomatism in various ways, which resulted in slight geochemical differences between the sources to these two categories and also to some degree internally within the high-TiO 2 category. There is abundant evidence that metasomatised lithosphere domains exist in other regions of the NAIP such as W Greenland too [130]. If some of the inferred metasomatic agents to low-TiO 2 and high-TiO 2 Faroese sill sources were not supplied from local Early Cenozoic mantle melting, geological activities associated with the complex geological history within this part of the North Atlantic area could be an alternative explanation. A potential alternative to metasomatic source enrichment would be magma mixing at deeper mantle levels involving recycled crustal materials. The origin of basaltic rocks within a fertile and metasomatised SCLM source during waning stages of regional magmatism, as we tentatively suggest for melts parental to Faroese lowTiO 2 and high-TiO 2 saucer-shaped sills, would be in accordance with inferences regarding late stage magmatism in other LIPs too [104]. [104] pointed out that a thermal anomaly from some sort of a mantle plume at the base of the lithosphere in the Paraná-Etendeka province once provided the heat necessary for basaltic melt production within the overlying fertile metasomatised lithospheric mantle. He inferred melting at successively shallower lithospheric levels with time from the arrival of the presumed mantle plume, provided that overall lithospheric thicknesses exceeded 100 km. [ [63]; HIMU and FOZO [131]; DMM [132]; IE 1 [49]; NAEM [8][9][10][11][12][13][14][15][16][17][18][19][20][21][22]; FDC [8]. (t) and (0) refer to age corrected and measured isotope ratios respectively. See text for more details.
towards higher 207 Pb/ 204 Pb ratios relative to other local sill samples in diagrams, where these lead isotope ratios are plotted against 208 Pb/ 204 Pb and 206 Pb/ 204 Pb ratios ( Figure  5b; Figure5d). The lead isotope particularities of these Faroese sills could stem either from tapping from mantle reservoirs displaying isotopic heterogeneities (at short length scales in the case of the Morskranes Sill), or they could implicate crustal contamination (i.e. discussion in sub-Sect. 5.2 above). * [13] suggested a link between increasing TiO 2 contents in OIBs worldwide and relative enrichments of their Sr and Nd isotopes, due to decreasing degrees of partial mantle melting of eclogite-bearing peridotites (~1 to ~10% eclogite) with increasing mantle depths, in accordance with the lid effect [13,14,120]. The lid effect has also been considered as an important process in parts of the NAIP [8,9,16,17,18]. The distinct TiO 2 compositions, displayed by samples representing the various Faroese sills, could in theory have evolved in response to similar mantle processes, but no clear systematic Sr and Nd isotopic enrichments exist for Faroese high-TiO 2 versus low-TiO 2 sills (although the low TiO 2 sills display slightly elevated 143 Nd/ 144 Nd ratios relative to all high-TiO 2 sills). Also, the limited Sr and Nd isotopic data available for these sills (Figure 5a) render it difficult to assess with any certainty whether they developed in accordance with the lid effect. If these sills indeed formed broadly contemporaneously according to the lid effect, it would imply a very steep or discontinuous local lithospheric base, as lateral distances between most low-TiO 2 and high-TiO 2 Faroese sills only range from 10 to 15 km (Figure 1). A discontinuous local lithospheric base in the Faroese area could have developed in response to previous regional or provincial tectonic activity. Alternatively, some of the magmas may have experienced noticeable lateral transport, like it has been suggested in recent studies for other low-TiO 2 Faroese lavas and during recent volcanic eruptions in Bali, Indonesia [9-134].

Summary and Concluding Remarks
The basaltic sills of the Faroe Islands can be grouped into three main categories according to their TiO 2 contents. The Streymoy and Kvívík sills define a low-TiO 2 category, the Eysturoy, Sundini, Langaregn and Svínoy-Fugloy sills define a high-TiO 2 category, while samples of the Morskranes Sill are intermediate-TiO 2 between these two. It is likely that the high-TiO 2 sills developed from three slightly dissimilar mantle sub-sources, i.e. the Eysturoy and Sundini sills, the Langaregn Sill and the Svínoy-Fugloy Sill respectively tapped mantle reservoirs of slightly different geochemical compositions Calculations suggest that differences in Al 2 O 3 , Fe 2 O 3 , TiO 2 , CaO, MgO, Sr, Eu compositions in particular within the actual intrusions can be explained by various degrees of partial melting, olivine fractionation, as well as plagioclase fractionation and accumulation. Moderately positive and negative Nb and Ta anomalies in some of the sills seem to suggest that their mantle sources were affected by the addition or extraction of metasomatic agents at some point. The inferred petrogenetic sequences for the actual sills can briefly be summarised as follows: 1. The Streymoy and Kvívík sills evolved by fractional crystallisation of mainly olivine (~15 wt%) and net accumulation of plagioclase (15 to 25 wt%) upon ascent. Their REE compositions can be explained by 16 to 21% melting of a moderately fertile mantle at depths of ≤ 85 km and (e.g. Figure 12a). 2. The Eysturoy, Sundini, Svínoy-Fugloy and Langaregn sills evolved by fractional crystallisation of mainly olivine (~15 wt%) and plagioclase (15 to 20 wt%), while en-route to the upper crust. The REE compositions of the first three sills can be accounted for by 5 to 7.5% melting of a moderately fertile mantle at depths of ≤ 85 km, while ~4% melting of a slightly different mantle composition(s) at deeper levels is required for the Langaregn Sill (Figure 12b; Figure 12c). 3. The primary magmas, which eventually evolved to the Morskranes Sill by fractional crystallisation of mainly olivine (~15 wt%) and perhaps some plagioclase (≤ 10 wt%) during ascent, most likely formed by 6 to 7% melting of a moderately depleted mantle at depths of ≤ 85 km (Figure 12d). Isotopic constraints suggest that parts of the Morskranes Sill have experienced slight crustal contamination, or their mantle source(s) displayed slight isotopic heterogeneities a short length scales. Such a contaminant would be of similar composition to that reported earlier for the Rockall Plateau basement and/or E Greenland basement (please see captions Figure 7 for references). 4. The Pb isotope range of local contaminated silicic lavas can be explained in terms of assimilation of materials comparable to those reported for E Greenland, while part of this isotope range could be explained by contamination with materials comparable in isotope compositions to those reported for NW Britain basement. Hence, E Greenland-type basement is apparently ubiquitous beneath the Faroe Islands (please see captions Figure 7 for references). Based on the limited data of this work, isotopic compositions of most Faroese sill samples can probably be explained in terms of the two well-known isotopic mantle reservoirs NAEM and FOZO, where low-TiO 2 and high-TiO 2 sills probably also contain slight amounts of materials originating from an EM-like mantle component. Hence, it is likely that at least three main end-member isotopic mantle reservoirs contributed to the isotopic characteristics displayed by the Faroese sills. It is proposed that melt generation to produce the Faroese sills occurred at relatively shallow mantle levels during the waning stages of basaltic magmatism in the NAIP, much similar to what has been envisaged previously for some Early Cretaceous Paraná-Etendeka flood basalts [104]. The garnet stability field is from [101,102]. The plagioclase stability field is from [103,102]. Abbreviations where x represents the letters a-d are: xm = partial melting; xa = plagioclase accumulation; xf = mineral fractionation; xe = sill emplac ement. See main text for a more detailed explanation.
Data availability. All information on international and/or in-house standard geological samples, used in association with analyses related to this research (i.e. as mentioned in Supplement 2 [51], are available at the actual laboratories: at the Open University, Milton Keynes, UK, and at NGU, Trondheim, NO, for XRF analyses; at Durham University, Durham, UK, and at GEUS, Copenhagen, DK, for the ICP- Author contributions. The actual manuscript is based upon a relatively recent Ph.D. thesis in geology, defended at Durham University in 2012 by the first author. The actual research and writing/preparation was performed by the first author Jógvan Hansen. All listed co-authors have acted as supervisors at various stages of thesis and paper preparation. In addition, Dougal Jerram initiated the actual research project at Durham University, while Mike Widdowson and Christopher Ottley ran XRF and ICP-MS laboratories at the Open University and at Durham University respectively, where the actual analyses were performed.
Competing interests. The authors declare that they have no conflict of interest.
Disclaimer. The authors declare that they have nothing to disclaim.