Petrogenesis of the Loch Bà ring-dyke and Centre 3 granites, Isle of Mull, Scotland

The Loch Bà ring-dyke and the associated Centre 3 granites represent the main events of the final phase of activity at the Palaeogene Mull igneous complex. The Loch Bà ring-dyke is one of the best exposed ring-intrusions in the world and records intense interaction between rhyolitic and basaltic magma. To reconstruct the evolutionary history of the Centre 3 magmas, we present new major- and trace-element, and new Sr isotope data as well as the first Nd and Pb isotope data for the felsic and mafic components of the Loch Bà intrusion and associated Centre 3 granites. We also report new Sr, Nd and Pb isotope data for the various crustal compositions from the region, including Moine and Dalradian metasedimentary rocks, Lewisian gneiss, and Iona Group metasediments. Isotope data for the Loch Bà rhyolite (87Sr/86Sri = 0.716) imply a considerable contribution of local Moine-type metasedimentary crust (87Sr/86Sr = 0.717–0.736), whereas Loch Bà mafic inclusions (87Sr/86Sri = 0.704–0.707) are closer to established mantle values, implying that felsic melts of dominantly crustal origin mixed with newly arriving basalt. The Centre 3 microgranites (87Sr/86Sri = 0.709–0.716), are less intensely affected by crustal assimilation relative to the Loch Bá rhyolite. Pb-isotope data confirm incorporation of Moine metasediments within the Centre 3 granites. Remarkably, the combined Sr–Nd–Pb data indicate that Centre 3 magmas record no detectable interaction with underlying deep Lewisian gneiss basement, in contrast to Centre 1 and 2 lithologies. This implies that Centre 3 magmas ascended through previously depleted or insulated feeding channels into upper-crustal reservoirs where they resided within and interacted with fertile Moine-type upper crust prior to eruption or final emplacement.


Introduction
Within the British-Irish Palaeocene Igneous Province, a bimodal distribution of mafic and felsic compositions exists in many intrusive and eruptive suites. This phenomenon frequently raised the question as to the relationship between the respective endmember compositions and a variety of genetic links between coexisting felsic (acidic) and mafic (basic) components have been proposed (e.g. Blake et al. 1965;Walker and Skelhorn 1966;Gamble 1979;Vogel et al. 1984;Sparks 1988;Troll et al. 2004Troll et al. , 2005Troll et al. , 2008aEmeleus and Bell 2005;Meade et al. 2009Meade et al. , 2014. Whereas traditionally a closed-system relationship has often been favoured, geochemical and particularly isotope studies now suggest that endmember magma compositions in mixed-magma igneous rocks may be more likely to represent replenishments of mafic magmas into felsic magma reservoirs. Centre 3 of the Mull igneous complex (Bailey et al. 1924;Walsh et al. 1979;Figs. 1, 2) Communicated by Gordon Moore. hosts a number of micro-granites (granophyres) and the mixed-magma intrusion of the iconic rhyolite-basalt Loch Bà ring-dyke (Figs. 3,4,5). Previous work on the Loch Bà ring-dyke proposed that the mafic and felsic domains represent a closed-system lineage that reflects extreme crystal fractionation within a zoned magma chamber (e.g. Walker and Skelhorn 1966;Sparks and Marshall 1986;Sparks 1988). However, many felsic rocks in the province, have now been recognised to record open system crustal influences in addition to fractional crystallisation processes (e.g. Walsh et al. 1979; Thompson et al. 1982Thompson et al. , 1986Geldmacher et al. 1998Geldmacher et al. , 2002Troll et al. 2004Troll et al. , 2005Troll et al. , 2019Font et al. 2008;Meade et al. 2009Meade et al. , 2014Meyer et al. 2009;Nicoll et al. 2009). Here, we use Sr, Nd, and Pb isotopes from selected igneous and crustal rock samples in conjunction with major-and trace-element data to unravel the petrogenetic evolution of the Mull Centre 3 igneous rocks and to test the model of closed-system crystal fractionation for the Loch Bà rhyolites. To the best of our knowledge, we report the first Nd and Pb isotope data for the Loch Bà ring-dyke and the associated Centre 3 granites (c.f. Walsh et al. 1979;Kerr et al. 1999). We combine these new data from Centre 3 with existing and new radiogenic isotope data from surrounding crustal lithologies (Fig. 1), which include Moine and Dalradian metasedimentary rocks as well as Iona (Lewisian) gneiss and Iona Group metasedimentary lithologies.

Geological setting
The Mull Igneous Complex is one of the chains of Palaeocene igneous centres exposed along the West coast of Scotland and in the eastern portion of the northern part of Ireland, and which developed during the early stages of the opening of the North Atlantic (Emeleus and Bell 2005;Troll et al. 2005;Meade et al. 2014). The Mull complex lies within the Northern Highlands Terrane, where Proterozoic Moine metasedimentary rocks make up a large portion of the upper crust beneath the igneous complex (Fig. 1). The rocks of the Northern Highland Terrane are separated from the Hebridean Terrane to the north and west by the shallow southeast-dipping Moine thrust and the Sound of Iona fault (e.g. Smythe 1987;Potts et al. 1995), and it has been suggested that the tectonic transition between Moine schists and Lewisian gneisses occurs at shallow levels beneath Mull and Ardnamurchan (e.g. Bott and Tuson 1973;Bott and Tantrigoda 1987;Kerr et al. 1995Kerr et al. , 1999. In the south of Mull, the Moine metasedimentary rocks are separated from Dalradian supergroup metasedimentary lithologies by a continuation of the Great Glen Fault (Bailey et al. 1924), marking the boundary of the Central Highlands (or Grampian) Terrane (Fig. 1). The Mull Igneous Complex has been the subject of intense geological investigation. Following early pioneering works (see summary in Emeleus and Bell 2005), modern geological investigation commenced with the Mull Memoir  Bailey et al. (1924). Subsequent works has focussed on the geochemical, textural, and geochronological aspects of the Mull Complex (e.g. Morrison et al. 1985;Thompson et al. 1986;Sparks 1988;Kerr 1993Kerr , 1995aKerr et al. 1995;Preston et al. 1998;Chambers and Fitton 2000;Chambers and Pringle 2001) and a detailed geochemical summary for the evolution of the entire Mull Complex was provided by Kerr et al. (1999). The age and duration of the igneous activity at Mull are constrained between c. 61 and 58 Ma (Chambers and Pringle 2001;Chambers et al. 2005). Following on after the Staffa lavas, the earliest Palaeocene activity at Mull appears to be the Mull Plateau lavas that have been dated at 60.65 ± 0.29 Ma (Chambers 2000). The plateau lavas have an estimated total thickness of ≥ 1,800 m (Bailey et al. 1924) and their geochemistry and stratigraphy have been investigated previously by Beckinsale et al. (1978), Kerr (1995a, b) and Kerr et al. (1995Kerr et al. ( , 1999. There are three geochemically distinct magma types represented within the Mull plateau lavas (Kerr et al. 1999;Chambers and Fitton 2000), reflecting how over time, the depth and style of mantle melting might have changed. The bulk of the Mull Plateau Group lavas is transitional to mildly alkaline basalts, the more magnesium-rich of which are often contaminated with small amounts (< 5%) of Lewisian crust (Kerr 1995b). These relatively uncontaminated basalts are the result of between 6 and 10% partial melting of a depleted garnet-bearing mantle source (Kerr 1995a). The Coire Gorm type-lavas (which are of an intermediate age between the Mull Plateau Group and the Central Mull Tholeiites), are somewhat more tholeiitic in nature, and appear to be the products of slightly higher degrees of partial mantle melting (8-12%) of a spinel lherzolite mantle-source (Kerr 1995a). The youngest lavas on Mull, the Central Mull Tholeiites, are considered the result of the most extensive degree of mantle melting (12-17%) of a depleted spinel lherzolite mantle (Kerr 1995a, b;Chambers and Fitton 2000).
Contemporaneous with the lavas and in part post-dating them, are the three centres exposed in central Mull. Their loci appear to have migrated in a NW-direction as the Mull Igneous Complex evolved ( Fig. 1), with migration from the oldest Glen More Centre (or Centre 1; 59.05 ± 0.27 Ma; Chambers 2000), via the Beinn Chaisgidle Centre (or Centre 2), to the Loch Bà Centre (or Centre 3; 58.48 ± 0.18 Ma) (Chambers and Pringle 2001;Emeleus and Bell 2005). Significant proportions of the exposed rocks within these three centres are of felsic (i.e., granitic, granophyric, and rhyolitic) composition (see Walsh et al. 1979;Kerr et al. 1999), which raises the question as to the relationship of the felsic magmas with the mafic lavas of the complex.

Centre 3 of the Mull Igneous Complex
We focus here on Centre 3, which marks the closing stages of the main magmatic activity of the Mull Igneous Complex (Fig. 2;Walsh et al. 1979;Kerr et al. 1999). The Glen Cannel Granite (Fig. 5) appears to be the oldest intrusion of Centre 3 and intrudes the basaltic lavas of the Mull Plateau group. It forms an oval, dome-shaped mass with a NW-trending long axis (Bailey et al. 1924;Walsh et al. 1979;Emeleus and Bell 2005). The Glen Cannel Granite is roofed by volcaniclastic rocks and extrusive felsites (rhyolites) preserved within a centrally subsided block enclosed by the Loch Bà ring-dyke, suggesting a caldera type edifice for Centre 3 volcanism (Fig. 2). The Beinn à Ghraig Granite (Fig. 6) is slightly younger than the ring dyke and is located mostly outside the perimeter of the Loch Bà ring-dyke intrusion, but is similar in petrographic character to the Glen Cannel Granite (Emeleus and Bell 2005). The Loch Bà ring-dyke intrusion itself is the youngest of the Centre 3 intrusions and is locally up to 400 m wide (Fig. 2), but varies in width around Fig. 3 a Steep outwarddipping margin of the Loch Bà ring-dyke against the Beinn a' Ghraig granite (left of red arrow), marking the ring-fault of the Centre 3 caldera. b Eutaxitic texture with highly twisted and strung out mafic inclusions (fiamme) that are frequently aligned parallel to the outer margin of the ring-dyke. Lump hammer for scale the perimeter of the intrusion. Where exposed, the ring fault is seen to dip steeply outward (Fig. 3) and has been documented to record several hundred metres of vertical displacement (Richey 1932;Lewis 1968). The Loch Bà ring-dyke is characterised by crystal-poor rhyolite that contains abundant streaky basaltic to intermediate mafic inclusions (Walker and Skelhorn 1966;Sparks 1988;Figs. 3, 4; Table 1). The mafic inclusions make up ~ 20% of the entire rock mass and occur throughout the ring-dyke, but are seen to concentrate locally to form semi-parallel bands, where they exceed 20 vol.%. The streaky inclusions display fiamme-like shapes, similar to those in welded ignimbrites (Fig. 4). The emplacement of the ring-dyke is thus believed to have involved violent magma fragmentation and mixing of magmas during chamber withdrawal and during conduit transport (Bell and Emeleus 1988;Sparks 1988;Emeleus and Bell 2005). LB-6 and LB-SE-1). Note, chilling of the mafic enclaves against the rhyolite matrix is not observed, implying the component magmas were approaching thermal equilibrium (cf. Troll et al. 2004)

Crustal lithologies
In addition to Centre 3 igneous rocks, we report new isotope data from Moine metasediments from Mull and nearby Ardnamurchan as well as Dalradian metasedimentary rock samples from the southern tip of the Isle of Mull. In addition, Lewisian gneissic basement and Paleoproterozoic metasedimentary rocks from nearby Iona were also sampled (Tables 2, 3).

Analytical methods
To complement existing major and trace-element data on Centre 3 igneous rock samples and crustal compositions in the region, we analysed additional eight igneous and seven crustal rock samples for major and trace elements and for radiogenic isotopes. Samples were crushed in a jaw crusher and then handpicked from fine rock chips. For samples from Loch Bà, the felsic matrix portion was separated from the mafic components by hand using a stereo microscope to ensure end-member compositions are analysed instead of analysing mixed bulk rocks. The mafic Loch Bà samples were extracted from larger mafic inclusions (≥ 5 cm) that were cut from individual rock samples and then crushed, and also hand-picked under a stereo microscope to avoid felsic materials in the picked sample. Major and traceelement values were determined using X-ray fluorescence spectrometry (XRF) on fused beads, employing an automated Philips PW1480 spectrometer at GEOMAR Research Center in Kiel, Germany following the procedure outlined in Troll and Schmincke (2002) and with full analytical details given in Abratis et al. (2002). The rare-earth element concentrations were determined by inductively coupled plasma mass spectrometry (ICP-MS), using an Agilent 7500 CE, at the Scottish Universities Environmental Research Centre (SUERC), East Kilbride, Scotland and Sr-, Nd-, and Pbisotope analyses were also conducted at SUERC. Radiogenic isotopes for Sr and Nd were analyzed on a VG Sector 54-30 thermal ionization mass spectrometer. 87 Sr/ 86 Sr was corrected for mass fractionation using 86 Sr/ 88 Sr = 0.1194. Repeat analysis of the NIST SRM-987 Sr standard gave 0.710257 ± 18 (2 sd, n = 14) for the duration of this study (see Meyer et al. 2009;Troll et al. 2019 for full analytical details). 143 Nd/ 144 Nd was corrected for mass fractionation using 146 Nd/ 144 Nd = 0.7219. During the course of this study, the SUERC internal Nd laboratory standard (JM), which is calibrated against the La Jolla Nd solution and the JNdi-1 standard, gave 143 Nd/ 144 Nd = 0.511511 ± 9 (2 sd, n = 21). Pb was separated using standard HBr-HCl anion exchange techniques, and measured on a Micromass IsoProbe MC-ICP-MS. The data were corrected for mass fractionation of 0.1% amu-1 based on replicate analysis of the NBS-981 standard. External reproducibility of the 206 Pb/ 204 Pb, 207 Pb/ 204 Pb and 208 Pb/ 204 Pb isotopic ratios is 0.2% (2 sd), and analytical blanks were < 1 ng (see Fitton et al. 1998;Ellam 2006;Meyer et al. 2009 for full analytical details). All isotope ratios were age-corrected to 58.5 Ma according to the accepted time of igneous emplacement (Chambers and Pringle 2001).

Major-, trace-and rare-earth elements
We compare our new major and trace-element data from the Centre 3 Loch Bà ring-dyke and its mafic inclusions as well as from the Beinn à Ghraig and the Glen Cannel microgranites (see Figs. 6, 7, 8; Table 1) with available data from previous investigations (Marshall 1984;Sparks and Marshall 1986;Sparks 1988;Kerr et al. 1999). On a TAS diagram (Fig. 6a), our felsic Centre 3 igneous rocks show very similar compositions and form a tight high-silica cluster (SiO 2 72.5 to ~ 74 wt.%). Similarly, on most Harker plots ( Fig. 6b-h), previous whole-rock data are more spread out than our handpicked endmember compositions, implying mixing between mafic and felsic compositions, especially as characteristic inflections expected for mineral fractionation are not always present (e.g. for iron and titanium), although some oxides do show inflections (e.g., phosphorus). Exceptions are K 2 O and Na 2 O that show a degree of scatter and can reach up to 6 wt.% in the Loch Bà rhyolite samples. Although we cannot exclude a degree of alteration for some of the literature samples, the LOI values of our samples were consistently low (Table 1), and the values derived for our samples are thus unlikely to be the result of secondary alteration, but must reflects the highly evolved nature of the crystal-poor Loch Bà rhyolite magma (Fig. 6). In respect to trace elements, less data are available from previous studies, making it all the more obvious that the Centre 3 granites differ from the Loch Bà rhyolite, likely because of modal mineral variations (Figs. 4,5,7). Moreover, the mafic Loch Bà ring-dyke inclusions are compositionally separate from the felsic Centre 3 rocks, defining a compositional gap. The mafic samples also appear to show a spread from basaltic-andesite to andesite (SiO 2 from 56 to ~ 60 wt.% and MgO from 1.8 to 2.7 wt.%) and notably, no pure basalt is found amongst the more mafic enclaves. In fact, they have relatively low Ni, in line with the low MgO values, and they are thus typical for intermediate magmatic compositions from the region (cf. Emeleus and Bell 2005;Troll et al. 2019).
Chondrite-normalised (after Boynton 1984) REE concentrations of our Centre 3 samples show all of the samples are enriched in LREE relative to primitive mafic compositions from Mull and from the wider region (Fig. 8). Centre 3 samples display negative Eu anomalies with Eu/Eu* = 0.43-0.51 for felsic samples and 0.58-0.68 for the mafic Loch Bà inclusions. The Loch Bà rhyolites display a slightly more pronounced Eu-anomaly relative to the Centre 3 granite samples, consistent with a more evolved (crystal poor) composition. There is overall, however, limited difference in the REE concentrations between the Loch Bà mafic inclusions and the Loch Bà rhyolite, nor in fact between Loch Bà samples and the Centre 3 granites, and all of the samples broadly overlap in their REE spectra (Fig. 8). Remarkably, the REE patterns from the Centre 3 igneous samples are also very similar to those of local Moine pelitic schists (cf. Geldmacher et al. 1998), which would thus appear to have had a compositional influence on the Centre 3 igneous suite, consistent with the isotope data we present below.

Isotope composition of crustal samples
Isotope data of crustal lithologies are reported in Table 3 and all our crustal samples are markedly displaced from mantle-like compositions. The age-corrected data (at 58.5 Ma) show a large range of isotopic compositions. Crustal samples from different terranes or units are clearly discernible from each other in isotope space (Fig. 9), and thus they provide a framework for us to resolve specific crustal influences the Centre 3 magmas may have experienced.
Specifically, the Iona gneiss shows comparatively low 87 Sr/ 86 Sr and 143 Nd/ 144 Nd isotope ratios and plot between the previously defined Lewisian amphibolite-and granulite-facies fields (Dickin 1981;Dickin et al. 1984). The Iona Group metasedimentary rock, in turn, shows elevated 143 Nd/ 144 Nd ratios (Figs. 9, 10), whereas the Iona gneiss plots towards less radiogenic Pb isotope ratios relative to the Iona Group metasedimentary rocks, highlighting that the latter is not purely derived from the former in terms of sedimentary provenance (e.g. McAteer et al. 2014).
Available data for the Moine metasedimentary rocks show a large spread in isotope values (e.g. Thompson et al. 1986;Preston et al. 1998), which overlap with our new Moine isotope data (Table 3). Moine samples from Ardnamurchan (e.g., Geldmacher et al. 1998Geldmacher et al. , 2002 are more radiogenic in terms of 87 Sr/ 86 Sr than our Moine samples from Mull (Figs. 9, 10). For Sr and Nd isotopes, our Dalradian sample plots within the large Moine field, whereas for Pb-isotopes the Dalradian sample plots close to the Moine and the Iona metasedimentary suites (Figs. 9, 10, 11). Centre 3 Granites (Walsh et al. 1979)

Isotope geochemistry of centre 3 igneous rocks
The isotope data of the Centre 3 igneous rocks are plotted in Figs. 9, 10, 11 and 12. All Centre 3 granite and rhyolite samples are markedly enriched in 87 Sr/ 86 Sr i (0.709 to 0.716) relative to mantle-type compositions and show a spread in 143 Nd/ 144 Nd i (0.51258-0.51207). They are significantly displaced towards the Moine metasedimentary rocks of the local upper crust. Notably, the Loch Bà rhyolites and mafic inclusions are the most radiogenic of the analysed suite, and are considerably more radiogenic than the Centre 3 microgranites. This relationship precludes a closed-system magmatic evolution for the various Centre 3 magmas (e.g. Sparks et al. 1988) and documents open system assimilation of Moine-type country-rock (e.g. Walsh et al. 1979). Although, there is only a narrow range of values obtained from our Centre 3 igneous samples, the Centre 3 igneous rocks also show extremely radiogenic Pb-isotope ratios relative to mantle proxies (cf. Ellam and Stuart 2000;Upton et al. 2002;Ellam 2006) and a crustal influence can be clearly resolved (Figs. 9, 10, 11, 12). Pb-isotopes therefore confirm the 87 Sr/ 86 Sr and 143 Nd/ 144 Nd results, showing a substantial Moine crustal contribution to the Centre 3 magmas. Notably, the Loch Bà rhyolites are also the most radiogenic samples of the Centre 3 Suite in respect to Pb isotopes.

Major and trace elements
The Centre 3 micro-granites and the Loch Bà rhyolites form a high SiO 2 cluster on most major and trace-element plots (Figs. 6, 7), whereas the Loch Bà mafic inclusions spread out in the intermediate compositional range. Using our handpicked end-member samples for the enclaves and the Loch Bà rhyolites, we note that a pronounced compositional gap (Bunsen-Daly gap) becomes apparent between these groups, for major elements as well as for trace elements, which was not as obvious from the previous studies where bulk samples were used (cf. Fig. 6). Moreover, the linear trend for the previously published Loch Bà data suite in e.g., TiO 2 , FeO t , against SiO 2 is likely a function of magma mixing between the Loch Bà rhyolite and mafic magma(s), as the characteristic 'kinks' of fractional crystallisation are not observed on these graphs (cf. Geldmacher et al. 1998;Kerr et al. 1999;Troll et al. 2004). Indeed, all Centre 3 samples  Kerr et al. 1999), Loch Scridain sills (Preston et al. 1998), the gneisses from Iona and Tiree (Walsh et al. 1979;Kerr et al. 1995), and picrite dyke M9 (Upton et al. 2002). Symbol sizes are larger than associated analytical uncertainties. Note Pm (between Nd and Sm) is not included. The Centre 3 samples show surprisingly similar REE patterns to the local Moine-type crust (see also Preston et al. 1998). See text for details are also significantly LREE enriched relative to mantle-like compositions and seem to share a strong trace-element affinity with the Moine pelite schists (Fig. 8), i.e., they mirror a strong enrichment of La to Sm (cf. Thompson et al. 1986). Moreover, all Centre 3 rocks show negative Eu anomalies (Eu/Eu* = 0.43-0.68), which may reflect plagioclase fractionation, or, more probably, selective crustal melting of Moine rocks that left behind a plagioclase-rich restite (cf. Walsh et al. 1979;Duffield and Ruiz 1998), which together with the La and Sm enrichment is thus unlikely to reflect crystal fractionation. Interaction between Centre 3 magmas with Iona Lewisian gneiss cannot be fully ruled out, but Moine-type rocks were clearly the dominant crustal influence on the REE patterns of the felsic Centre 3 magmas. A virtually continuous spectrum of phenocryst compositions was reported in the Loch Bà suite by Sparks (1988) and was interpreted to reflect a continuous liquid line of descent. This, he argued, supports continuous fractional crystallisation as in a Skaergaard-type scenario (e.g. McBirney 1975). The presented REE data, together with the new isotope results rule out a closed-system evolution, however, implying magma-crust interaction as an important process during the evolution of Centre 3 magmatic compositions (cf. Walsh et al. 1979).

Isotopes
Iona Lewisian gneisses and Iona Group metasedimentary rocks are isotopically distinct from each other (Figs. 9, 10, 11, 12). The Iona gneiss shares characteristics with granulite-and amphibolite-type Lewisian gneisses exposed farther north on the Scottish mainland and on Tiree (Dickin 1981;Kerr et al. 1995;Friend and Kinny 2001;Kinny et al. 2005). However, neither REE data (Walsh et al. 1979) nor the isotope data of this study show that Iona gneiss or the Iona Group metasedimentary rocks left a significant imprint on the Centre 3 igneous rocks (Figs. 9, 10, 11, 12). In contrast, earlier rocks from the Mull igneous complex, e.g. Mull Plateau lavas and cone-sheets of Centre 1, do clearly record signs for interaction with Lewisian-type materials (Thompson et al. 1986;Kerr 1995a, b;Preston et al. 1998;Kerr et al. 1999). In addition, isotope modelling (Figs. 11,12), suggests that Dalradian metasediments have also not significantly contributed to the Centre 3 magma compositions, which is not unreasonable given the centre's position north of the Great Glen fault and the > 20 km geographical distance between Centre 3 and the nearest occurrence of Dalradian rocks in outcrop. The Moine Supergroup is seemingly the only significant crustal influence on Centre 3 rocks  (Preston et al. 1998), the Mull lava and cone sheet data (Thomson 1986;Kerr et al. 1999), the Ardnamurchan cone-sheet data and Ardnamurchan Moine crust (Geldmacher et al. 1998), the Skye Red hills data (Dickin et al. 1981), and gneiss from Tiree (Kerr et al. 1995 and although small early additions of Lewisian cannot be ruled out, the mixing trajectories shown in Fig. 9 preclude large amounts of Lewisian gneiss materials to have played a role. The Moine Supergroup suite is divided into three stratigraphic subunits (Strachan et al. 2002). The oldest Morar Group is exposed on Ardnamurchan (e.g. Geldmacher et al. 1998), while the younger (but higher metamorphic grade) Glenfinnan Group is exposed in SW-and E-Mull. From the Moine isotope data obtained in this study, it seems these groups can be broadly distinguished on isotopic grounds (e.g. using Sr-Nd isotope ratios, Figs. 9, 10). Provided this distinction is systematic, this realisation then allows an attempt to quantify the respective influences on Centre 3 magmas.

Magma-crust interaction at Mull
Parental magmas of a depleted mantle-like isotope composition have previously been proposed for much of the British-Irish Palaeocene Igneous Province (e.g. Gamble et al. 1992;Kerr et al. 1999;Ellam and Stuart 2000;Upton et al. 2002) and we envisage such an 'unradiogenic' parental magma composition for the Mull complex. Parental magmas of this type would be highly susceptible to contamination from radiogenic crustal rocks, especially if high concentration partial country rock melts are envisaged (Ellam and Stuart 2000;Troll et al. 2005). To provide a first order quantitative assessment, we use a MORB-type picrite composition from the Isle of Rum (Upton et al. 2002;Meyer et al. 2009) as a potential mantle-type end-member and couple this composition with Moine metasedimentary rocks from the Isle of Mull as crustal input (samples Mul-Moine-1, Mul-Moine-CRG-3; Table 3). We then quantify the degree of interaction through assimilation and fractional crystallization-(AFC) calculations, as well as through two-component binary mixing calculations (cf. DePaolo 1981;Figs. 11, 12). Employing these end-member compositions, Sr and Nd isotopic ratios of the Loch Bà rhyolites can be reproduced by an AFC-type  Table 3). b Same plot using a different Moine composition as dominant contaminant (Mul-Moine-1). In either case, the AFC and binary mixing calculations support an apparent absence of deep basement influences for Centre 3 compositions, contrasting e.g. the Mull Plateau Group lavas (MPL), the early acidic cone-sheets (EACS), the early basaltic cone-sheets (EBCS), and the Loch Scridain sills (Thomson 1986;Preston et al. 1998)  scenario that assimilates Moine metasedimentary rocks similar to sample Mul-Moine-1 (r = 0.7). Notably, such an AFC curve also intersects the isotopic ratios of the Loch Bà mafic inclusion (Figs. 11,12). The modelled concentrations of Sr and Nd ppm (see Nicoll 2008) do, however, not accurately match the measured sample values of the Loch Bà rhyolite (measured Sr = 74 ppm, Nd = 54; modelled Sr = 280 ppm, Nd = 30 ppm), and partial country rock melts (instead of whole-rock melts) are a possible reason for this mild mismatch, e.g. when considering micas and some feldspar contributed to the partial melt but some feldspar and also pyroxene was residual (cf. Duffield and Ruiz 1998;Troll et al. 2005).
In contrast, binary mixing trajectories between the MORB-type isotopic end-member and the Dalradian metasediments, or between MORB and the Ardnamurchan Moine metasedimentary rocks show that Centre 3 igneous samples do not follow these trajectories. The Sr-Nd isotopic data of the Centre 3 samples can be reproduced, however, by mixing between a MORB-type magma and Moine metasedimentary rocks from Mull. This binary mixing model suggests between 10 and 30% input of Mull Moine incorporation for the Glen Cannel and Beinn à Ghraig granites, whereas the Loch Bà rhyolite requires between 40 and 60% of this local Moine component. The mafic Loch Bà inclusions show lesser degrees of crustal involvement, ranging from ~ 20 to ~ 30% for Sr and Nd isotopes. The uptake of local Moine metasedimentary crustal melts by Centre 3 magmas can be modelled as either binary mixing or indeed as high assimilation-rate AFC-style process (see Figs. 11,12).
Moreover, our Pb isotope data, the first published for Centre 3 igneous rocks, are also consistent with Moine crustal additions (Figs. 10, 12). The Centre 3 data plot on a straight line that connects mantle-type compositions with the Moine compositional field (e.g. Ellam and Stuart 2000). Notably, the Glen Cannel and the Beinn a' Ghraig granites plot closer to mantle-proxies (cf. Ellam and Stuart 2000;Upton et al. 2002) relative to the Loch Bà suite and quantitative mixing models suggests ~ 10% Moine-type crustal incorporation for the granitic samples (Figs. 11, 12). The Loch Bà rhyolites, in contrast, require ≥ 45% of local Moine metasedimentary incorporation, while the Loch Bà mafic inclusions record variable (20-50%) Moine metasedimentary involvement. The mixing ratios derived for Pb isotopes data imply that most Pb in Centre 3 samples is crustal in origin and the mixing ratios derived from the Sr-Nd isotope data presented above, are exceeded when using Pb isotope modelling. This observation confirms variable degrees of mixing between mantle-derived and local crustal components (Walker 1975;Kerr et al. 1999) and supports the notion of partial melting of crustal compositions (cf. Thompson et al. 1986;Kerr et al. 1995;Troll et al. 2005;Meade et al. 2009Meade et al. , 2014 to explain the Centre 3 isotope results. Assimilation of basement gneiss, as sampled from Iona, or as exposed in the Lewisian complex elsewhere in NW-Scotland, is not evident in Centre 3 igneous rocks, including potentially deeper Lewisian "granulite-facies-type" material, as e.g. reported from the Isle of Skye (Dickin 1981;Thompson et al. 1986;Font et al. 2008). In fact, the Mull Centre 3 rocks do not record an influence of an unradiogenic 206 Pb/ 204 Pb component that is characteristic of Lewisian granulite-type compositions (Dickin 1981;Kerr et al. 1999) and thus contrast the earlier Mull Centre 1 intrusives, the older Mull Plateau lavas (Walsh et al. 1979;Morrison et al. 1985;Thompson et al. 1986), and also the Ardnamurchan cone sheets (Geldmacher et al. 1998(Geldmacher et al. , 2002. Whereas the presented data on Centre 3 support a lack of interaction of Centre 3 magmas with Lewisian-type basement, or a very minimal interaction only (see Fig. 9), Moine metasedimentary rocks are clearly documented as a major crustal influence. The Centre 3 igneous rocks thus share similarity with the earlier Mull acidic cone-sheets (EACS) described by Thomson (1986), or the Loch Scridain sill described by Preston et al. (1998), which seem to also have escaped significant interaction with Lewisian basement at depth, but record upper crustal magma contamination. This can be explained by either (1) rapid ascent of the Centre 3 magmas to shallow crustal levels, (2) shielding from contact with deep country rock due to earlier igneous intrusions at depth, or (3) via progressively inhibited crustal assimilation ("PICA" concept) where easily fusible components in the gneiss were previously extracted (cf. Gamble et al. 1992;Upton et al. 1998;Kerr et al. 1999;Meade et al. 2014), e.g. during Centre 1 and 2 activity (Fig. 13).

Petrogenesis of Mull Centre 3
On the basis of major element correlations in available mineral phases, Sparks (1988) proposed a closed-system fractional crystallization model for the evolution of the Loch Bà ring-dyke suite. This "extreme closed-system crystal fractionation" from a mafic parent was inspired by the closed-system fractionation model suggested for the Skaergaard intrusion in SE-Greenland (McBirney 1975;Hunter and Sparks 1987). However, the trace element and isotope data presented here do not support a closed-system scenario for the Mull Centre 3 compositions. Instead, magma-crust interaction is highlighted as a significant additional process in the petrogenetic history of the Centre 3 magmas and we propose a model whereby differentiation of ascending Centre 3 magmas was accomplished by assimilation of fusible parts of the local upper crust (cf. Walsh et al. 1979;Patchett 1980;Thompson et al. 1986;Preston et al. 1998;Kerr et al. 1999). Variable degrees of AFC-type processes involving Moine whole rocks and variable portions of partial melts, coupled with magma mixing, took place, consistent with the trace element and isotope evidence presented and with the relatively low-melting temperatures expected for Moine schist-type compositions that would allow considerable portions of Moine-derived crustal melts at relatively modest temperatures (cf. Patchett 1980; Thompson et al. 1986;Huppert and Sparks 1988;Meade et al. 2014). In this respect, we note that a simple AFC scenario does likely not apply as the Loch Bà rhyolites and the micro-granites show similar Sr contents and silica composition, yet show different Sr-isotope ratios. Moine pelitic schist is the most distinct source of crustal input for Centre 3 magmas according to our isotope data, with the Loch Bà rhyolite showing the highest portion of crustal input. This contrasts most earlier felsic rocks from the Mull complex that usually also record initial Lewisian crustal inputs followed by upper crustal influences Walsh et al. 1979;Kerr et al. 1999). Prolonged deep storage seems thus to be recorded in the contamination pattern of many earlier Mull igneous rocks, whereas the main storage level recorded in the contamination signal for the Centre 3 magmas was much shallower, giving the impression of an overall upward migration of magma storage with time (cf. Walker 1975;Kerr et al. 1999;Troll et al. 2008b). This realisation is consistent with the observation that both Centre 3 microgranite bodies display drusy cavities and granophyric intergrowth (see Fig. 5), which is usually a sign of a shallow crustal emplacement of perhaps no more than 1.5-2 km below the former land surface (Emeleus 1961). These considerations would suggest an intrusion level right below the Mull plateau lavas and their underlying thin Mesozoic sediment cover (~ 200 m) and into the substantial thickness of Moine metasedimentary rocks (~ 4-5 km) below the thin Mesozoic sediments. Progressive supply of basic and intermediate magmas to this shallow crustal level could produce considerable quantities of crustal melt from the fertile Moine-type metasedimentary rocks (cf. Patchett 1980;Huppert and Sparks 1988;Meade et al. 2014), although such evolved crustal melts may not readily mix with mafic liquids (Blake et al. 1965;Walker and Skelhorn 1966;Kerr et al. 1999;Troll et al. 2004). We argue that this is the reason for the incompletely blended (stirred and mingled) Loch Bà rhyolite-andesite association. Finally, repeated injection of mafic magma into a shallow crustal rhyolitic reservoir represents a likely eruption trigger and may have caused violent pyroclastic eruptions and associated caldera collapse at Centre 3 (cf. Sparks et al. 1977;Sparks and Marshall 1986;Troll et al. 2004). While the two-component magmas may have been residing simultaneously in the shallow magma reservoir for some time (perhaps weeks to months), they were not side by side for too long, as otherwise isotopic equilibration would have been expected (cf. Troll and Schmincke 2002;Troll et al. 2004).

Fig. 13
Conceptual sketch of the evolution of the Palaeogene Mull volcanic Complex. The crustal structure is shown schematically and involves the Lewisian-like basement beneath the Moine metasedimentary suite of Mull. Also schematically represented are the gneisses and metasediments of Iona, located to the west of Mull which are uplifted relative to the gneissic basement beneath Mull, therefore representing a likely window into the deeper crustal structure. The Great Glen Fault marks the boundary to the Dalradian metasedimentary suite to the south, which seemed to have played no part in the igneous evolution of Centre 3. Magma ascent for Centre 3 was either via the shielded and or depleted plumbing region of Centre 1 followed by lateral transport to the NW, or via rapid ascent (through dykes) to high levels directly underneath Centre 3. All samples of the Centre 3 magmatic suite record a dominantly Moine-type contamination signal derived from the local country rocks at shallow crustal levels. The main magma storage for Centre 3 was thus confined to the upper portions of the crust, probably ponding under the thermal and density barrier of the earlier Mull lavas The role of Centre 3 within the evolution of the Mull complex Morrison et al. (1985) and Kerr et al. (1999) proposed that the style of crustal contamination at the Mull Complex changed over time (Fig. 13). Prior to activity at Centre 3, the Mull igneous complex had already been active for perhaps ~ 2 Ma (cf. Chambers and Pringle 2001) and some of the magmas of the earlier Mull lavas seem to have stalled and fractionated at high pressure, e.g. at the base of the crust, where they assimilated Lewisian gneiss materials (Thompson et al. 1982;Morrison et al. 1985;Kerr et al. 1995Kerr et al. , 1999. As the magmatic system beneath Mull matured, the focus of magma storage appears to have migrated upwards, and so the style of contamination changed because of the fusible Moine-type compositions that became available at shallow crustal levels. The presented REE and isotope data of Centre 3 rocks suggest a close affinity with the Moine psammites exposed in Eastern and Western Mull, which are present beneath Centre 3, and below what must have been a thick cover of Palaeocene basalt lava (> 1 km; Bailey et al. 1924).
Indeed, Kerr et al. (1999) advocated that the Mull Igneous Complex has a very similar deep architecture to that seen at Skye (Bott and Tuson 1973) and Rum (Emeleus 1997;Emeleus and Troll 2014). Geophysical studies by Bott and Tantrigoda (1987) indicate that a large + 50 mGal gravity anomaly exists beneath the Isle of Mull, which they interpreted as a thick (6.5-13 km) mafic or ultramafic pluton with a volume of between 2,000 and 3,600 km 3 . The granitic rocks on Mull appear to form a volumetrically very minor cap of maybe 1-2 km thickness only, beneath which the large mass of mafic to ultramafic rock commences. Following this line of thought, the granites of Centres 1, 2 and 3 represent perhaps 10-20% of the total intrusion mass of the Mull complex (cf. Bott and Tantrigoda 1987;Kerr et al. 1999), confirming inferences from other well-studied centres in the region, such as Rum or Skye (Walker 1975;Emeleus 1997;Emeleus and Bell 2005). This realisation implies that acidic (felsic) rocks are a notable but ultimately volumetrically limited by-product of large-volume basaltic magmatism, even if they exceed pure fractionation derived volumes of felsic magma due to the mobilisation of felsic continental crust.

Conclusion
Mull Centre 3 is composed of several shallow-level microgranite and rhyolite intrusions that mark the closing stages of a major and prolonged period of Palaeocene igneous activity at Mull. Distinct compositional gaps in the major elemental data preserve a record of magma mixing, while REE patterns show the Centre 3 magmas also reflect melting and incorporation of Moine-type metasedimentary rocks from the local upper crust. Isotope data confirm magma-crust interaction for Centre 3 and the rhyolites of Loch Bà show the largest degree of crustal involvement with ~ 50% Moine-derived Sr, Nd and Pb. Likely, the Loch Bà rhyolite magma represents the culmination of heat input into the upper crust during Centre 3 activity. The Loch Bà mafic inclusions generally show lower crustal additions (≥ 20%), as do the surrounding Centre 3 granites (up to 15%).
The successive Mull Igneous Centres (1-3), together with the extensive and successive lava units, appear to record an overall upward migration of their main magma storage zones that is recorded in their crustal assimilation patterns, with Centre 3 representing one of the final and shallowest episodes of activity. The interaction of Centre 3 magmas with shallow Moine-type metasedimentary crust only contrasts the earlier episodes of the Mull complex, where lower crustal Lewisian gneiss input is frequently detected. By the time Centre 3 was emplaced, magma pathways in the lower crust were likely well established and lined conduits may have shielded ascending magmas from interaction with gneissic materials. Alternatively, readily fusible material was perhaps already extracted from the lower crust during the earlier activity at the Mull complex and little fertile material may have been left at this point in the lower crust beneath the larger Mull igneous system.
Funding Open Access funding provided by Uppsala University.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat iveco mmons .org/licen ses/by/4.0/.