A large explosive silicic eruption in the British Palaeogene Igneous Province

Large-volume pyroclastic eruptions are not known from the basalt-dominated British Palaeogene Igneous Province (BPIP), although silicic magmatism is documented from intra-caldera successions in central volcanoes and from small-volume ash-layers in the associated lava fields. Exceptions are the Sgùrr of Eigg (58.7 Ma) and Òigh-sgeir pitchstones in the Inner Hebrides (>30 km apart), which have been conjectured to represent remnants of a single large silicic event. Currently available major element data from these outcrops differ, however, creating a need to test if the two pitchstones are really related. We employ a systematic array of methods ranging from mineralogy to isotope geochemistry and find that samples from the two outcrops display identical mineral textures and compositions, major- and trace elements, and Sr-Nd-Pb-O isotope ratios, supporting that the two outcrops represent a single, formerly extensive, pyroclastic deposit. Available isotope constraints suggest a vent in the Hebridean Terrane and available radiometric ages point to Skye, ~40 km to the North. A reconstructed eruption volume of ≥5km3 DRE is derived, suggesting a VEI 5 event or larger. We therefore argue, contrary to long-held perception, that large-volume silicic volcanism and its associated climatic effects were likely integral to the BPIP during the opening of the North Atlantic.


Geological Background And Earlier Investigations
The Sgùrr of Eigg dominates the south of the Isle of Eigg [17][18][19] (Fig. 2). The Sgùrr of Eigg was initially interpreted as a lava flow that occupies a steep-sided valley floored by a fluvial conglomerate 17 . In contrast, Harker 13,20 considered the pitchstone as intrusive, but Bailey 21 supported Geikie's lava flow interpretation. Subsequently, Allwright 14 and Emeleus 15 showed that the basal metre of pitchstone (e.g. at Bidean Boideach [NM 4412 8667]) is a welded vitric tuff, implying a pyroclastic origin, whereas the pale-coloured felsite sheets on the southern face of the Sgùrr (Fig. 2) were regarded as intrusive 14,15 . However, the most recent account of the Sgùrr of Eigg and its underlying conglomerate 16 , interprets the columnar jointing, weathering patterns and sharp undulating boundaries within the pitchstone to represent several rapidly emplaced ignimbrites from a low but sustained pyroclastic column. The pervasive, base-parallel flow-banding and associated recumbent isoclinal folds are held to indicate post-depositional 'lava-like' rheomorphism. The authors also argue, on the basis of similar mineralogy and chemical compositions between the dark pitchstone and the pale sheets on the southern face, that the latter represent devitrified zones at the tops of successive flow pulses, rather than separate intrusions. Further, Brown and Bell 16 tentatively proposed that the Sgùrr of Eigg pitchstone was sourced from Skye and envisage, a formerly extensive ignimbrite sheet. This is, however, uncommon and no such enormous outflow sheet has hitherto been recognized in the BPIP. A possible extension to the Sgùrr of Eigg pitchstone was recently suggested by Smith 22 , who identified a 5 km long, sinuous submarine ridge south of the Isle of Muck, but to date no material is available from this locality. An early age of 52.1 ± 1.0 Ma (Rb-Sr) for the pitchstone 23 , was subsequently revised to 58.72 ± 0.07 Ma (Ar-Ar) 24 . Òigh-sgeir is a group of rocky islets located ~8 km SW of Canna and >30 km WNW of Eigg and it is exclusively made of columnar porphyritic trachyte pitchstone (e.g. refs 12,13 ) (Figs 1 and 2). The petrography of the Òigh-sgeir pitchstone was investigated by Judd 18 , Geikie 17 , Harker 13 , Allwright 14 , and Emeleus 15 who all agreed that there were strong structural, petrographic and geochemical similarities between the Òigh-sgeir and the Sgùrr of Eigg. Brown and Bell 16 , however, recently provided data that show the Sgùrr of Eigg major-element composition to differ from the available Òigh-sgeir pitchstone data (Fig. S1). Notwithstanding these differences, Brown and Bell 16 support the Sgùrr of Eigg and Òigh-sgeir pitchstone to be part of a single eruption (see ref. 15 ) and on the basis of available isotope data from the Sgùrr of Eigg, they propose a source within the Hebridean Terrane. Specifically, they argue for the 40 km-distant Skye Central Complex, invoking a very large silicic eruption.   Given the possible importance of large volcanic eruptions for preparing the conditions that ultimately led to the Palaeocene-Eocene Thermal Maximum (PETM) at ~56 Ma (refs 8,25,26 ), it is important to establish the magnitude of silicic eruptions in this part of the North Atlantic Igneous Province in the run up to the PETM. To ultimately test if the Sgúrr of Eigg and Òigh-sgeir pitchstones represent the remnants of a single large silicic pyroclastic eruption, we compare the Òigh-sgeir and Sgùrr of Eigg pitchstones by employing newly acquired crystal, whole rock, and radiogenic and stable isotope data. In addition to testing for a common ancestry of the Sgùrr of Eigg and Òigh-sgeir pitchstones, we also use the new data to evaluate if a specific source within the Western Red Hills of Skye is plausible. If the pitchstones of Eigg and Òigh-sgeir can be shown to represent remnants of a single large eruption, it would be one of the largest as yet recorded from the North Atlantic Igneous Province (NAIP) and certainly the largest known in the British Palaeogene Igneous Province.

Assessing a Common Ancestry
While using mineralogy and major and trace elements to assess consanguinity of different bodies of rock may potentially be compromised in porphyritic rocks due to modal mineral variations or variable degrees of fractionation, isotopic tracers may be more reliable. This is because radiogenic isotopes are insensitive to physical changes such as temperature, pressure or crystallisation conditions, but reflect the various source compositions involved in petrogenesis (e.g. ref. 27 ). Stable isotopes, in contrast, do change with physical conditions, but fractionation factors are not very large at magmatic temperatures 28 . To establish a common ancestry between the Eigg and Òigh-sgeir pitchstones a central theme of our approach is that a positive match for radiogenic and stable isotopes is ideally coupled with a positive match for mineral types, -textures, and -composition, and also with a match for whole rock and groundmass major element trends.
As regards radiogenic isotopes, four tectono-stratigraphic terranes are traversed through the British Palaeogene Igneous Province, from Skye and Rum in the North, to Carlingford and the Mourne Mountains in the South 3 ( Fig. 1; Supplementary Information S2). These terranes are isotopically so diverse, that ascending mantle-derived magmas have been variably affected by the specific terrane through which they have erupted (cf. refs 23,27,[29][30][31] ). The Sgùrr of Eigg and Òigh-sgeir pitchstones lie within the Hebridean Terrane which is characterised by Archaean Lewisian basement (e.g. refs 23,27 ), and is separated by the Moine thrust from the Northern Highlands Terrane where additional isotopically-distinct Moine-type metasedimentary rocks occur (e.g. refs 29,30 ). This led Dickin and Jones 23 to postulate a source within the Hebridean Terrane for Sgùrr of Eigg. Isotope data from Òigh-sgeir have so far not been available.
Field descriptions. The Òigh-sgeir skerries are made exclusively of pitchstone 13 and are situated on the basaltic Canna Ridge, a south-westwards submarine extension of the Skye Lava Group 15 . Columnar jointing can be seen in tidally exposed cliffs (Figs 1 and 2) and a post-emplacement fault offsets the skerries. The Eigg pitchstone, in turn, is very well exposed along a 3 km long, steep-sided ridge extending north-westwards from the Sgùrr of Eigg to Bidean Boideach (Fig. 2). There, a west-facing cliff exposes an underlying conglomerate that was filling a former steep-sided valley in the Eigg Lava Formation 15 . At the Recess [NM 4606 8462], an alcove formed by erosion below the massive pitchstone in the southeastern part of the Sgùrr of Eigg, a steep face exposes spectacularly-jointed pitchstone overlying the conglomerate with large boulders of red Torridonian sandstone and plant remains, including the 'Eigg pine' 13 . This conglomerate was widely viewed as fluvial 12,14,15,21 and an entire system of fluvial palaeo-valleys connected to the former Sgùrr of Eigg valley has been postulated 14 . The basal pitchstone is strongly brecciated and thoroughly mixed with the underlying sedimentary material. While Emeleus 15 suggested the occurrence of phreatic explosions, Brown and Bell 16 consider it a peperite. In addition, the lowest ~10 cm of the Sgùrr of Eigg pitchstone lack the otherwise abundant large feldspar crystals, but contain numerous small crystal fragments and millimetre-sized wispy basaltic fiamme in a vitreous, closely-packed matrix (Figs 3 and S2). Most recently, the Sgùrr of Eigg pitchstone was interpreted as a welded ignimbrite deposit that formed from a pyroclastic density current, and the underlying conglomerate as a debris flow 16 . The presence of mafic fiamme and re-agglutinated former glass shards documents a previous fragmented stage for these rocks, and thus corroborates an overall pyroclastic mode of formation (Figs 3 and 4).

Petrography of the pitchstones.
Here we provide a summary of previous petrological descriptions [14][15][16] and our own observations on the investigated rock samples.
The Sgùrr of Eigg pitchstone has a black, vitreous to dark matt-grey appearance on fresh surfaces and contains between 20 and 35% of 'free-floating' crystals of plagioclase (≤15%), anorthoclase (~10%), Fe-rich clinopyroxene (~2%), orthopyroxene (≤3%), Fe-Ti oxides, accessory apatite (~2%), and traces of sulphides. Notably, the rock is devoid of quartz, but fine-scale flow banding is common (Fig. 3). The Òigh-sgeir pitchstone is also black to dark brown and semi-vitreous with abundant mm-sized feldspar crystals (plagioclase and anorthoclase), plus minor augite and orthopyroxene set in pale-brown glass (no quartz present). The matrix contains microcrystals of plagioclase, oxides, sulphides, and accessory apatite (Figs 3 and 4). 'Flow banding' has locally been observed. The larger 'free-floating' feldspars are characterised by pronounced resorption textures in both the Òigh-sgeir and Sgùrr of Eigg pitchstone samples (Fig. 3). These are pronounced in the interiors of crystals, but notably not on outer crystal surfaces. The crystal assemblage in the Òigh-sgeir pitchstone is therefore extremely similar to that in the Sgùrr of Eigg and, moreover, the approximate proportions of minerals appear very similar too (Figs 3 and 4). In the basal, fragmental portion of the Sgùrr of Eigg, the occurrence of basaltic fiamme is noted (Figs 3f and S2), implying hot, mafic magma was involved in the eruption 32,33 . Petrography of plutonic inclusions. Both pitchstone groups contain abundant plutonic, broadly monzonitic, inclusions that host an almost identical mineral assemblage to that seen 'free-floating' in the glassy  Information S3). The identical mineralogical and textural crystal assemblage implies that the plutonic inclusions are the source for the crystals in the Òigh-sgeir and the Sgùrr of Eigg pitchstones and, moreover, the abundance of plutonic inclusions with disintegration textures suggests a remobilised monzonitic pluton or crystal mush as the main source of the crystals and inclusions (cf. ref. 34 ).
Major elements. The Òigh-sgeir and Sgùrr of Eigg samples reported by Emeleus 15 and those used in this study form tight clusters for the whole rock and for the groundmass data on a total-alkali vs. silica diagram ( Fig. 5a; Supplementary Table S1), classifying the whole rocks as trachyte due to the absence of quartz (see also Supplementary Information S1; Supplementary Table S2). The overlap is striking, and a common origin is possible. The combined Òigh-sgeir and the Sgùrr of Eigg whole rock suite (which we term 'OSSEP') does not overlap with known felsic rocks from Rum, the closest major volcanic centre to the two outcrops ( Fig. 1)  Notably, anorthoclase is present in both plutonic feldspar groups, but is sparse in the free-floating feldspar populations of both pitchstones. In the plutonic inclusions, euhedral plagioclase cores are frequently mantled by highly resorbed anorthoclase. This implies that preferential dissolution of low Ca-anorthoclase occurred, indicating re-heating of the pitchstone magma, because low Ca-feldspar has a lower melting temperature than Ca-rich plagioclase (cf. refs 34,37 39,40 ). The OSSEP whole rocks record very high values and they straddle the boundary between S-and I-type granites at ~+10‰ (e.g. ref. 41 ). However, although late contamination of the melt that now constitutes the glassy groundmass is possible, post-eruptive alteration of the glassy groundmass is quite likely 42,43 . Irrespective of the exact cause of the high δ 18 O whole rock data, the oxygen isotopes record an identical magmatic (crystal separates) and post-eruptive (whole rock) history of the Sgùrr of Eigg and the Òigh-sgeir pitchstones.
Radiogenic isotopes. Sr, Nd and Pb isotope data for the Sgùrr of Eigg and the Òigh-sgeir pitchstones (corrected to 59 Ma, ref. 24 ) as well as for representative Rum microgranites are presented in Fig. 7a-c and Supplementary Table S4. This, to the best of our knowledge, includes the first radiogenic isotope data for the Òigh-sgeir pitchstone. Figure 7a- 35.831, respectively), and the radiogenic isotope data thus overlap within the uncertainties of the analyses. Given that the Palaeogene Igneous Centres are known to show changing degrees of assimilation with time (e.g. refs 29,31,37,39 ), a large time gap between Sgùrr of Eigg and Òigh-sgeir pitchstones is not supported by the radiogenic isotope data. Instead, a single and common source for the Òigh-sgeir and Sgùrr of Eigg pitchstones is implied.
inclusion that comprises resorbed feldspar, pyroxene and opaques. Most of the feldspars show evidence of initial to advanced resorption. Importantly, the Sgùrr of Eigg and Òigh-sgeir pitchstones are strikingly similar, displaying the same mineral-types, mineral assemblage and identical mineral textures.

Establishing the Eruption Source
The petrographic, chemical, and isotopic characteristics of the Sgùrr of Eigg and Òigh-sgeir pitchstones broadly overlap in all lines of investigation and we conclude that the two outcrops, although >30 km apart, derive from the same source. They represent the remnants of a regionally extensive large-magnitude explosive silicic eruption. Our results now fully vindicate previous suggestions for a large, regionally extensive silicic eruption in the BPIP 12,14-16,21 . To establish the exact eruption source of the OSSEP tephra, a more detailed discussion is required. The prerequisites for the OSSEP source include (i) evidence of silicic magmatism together with a major heat source capable of generating the silicic magmas through fractionation and crustal assimilation, (ii) chemical and isotopic  Information S1 and  Supplementary Table S6 for data sources and selection). Notably the combined OSSEP data overlap with the mixed-magma Marsco gabbro-granite suite of Skye and plot close to known trachyte lavas from the Skye Lava Group, while differing markedly from the rhyodacites and microgranites exposed on Rum. (b) Feldspar compositional triangles (An-Ab-Or) for Òigh-sgeir "free-floating" feldspar in the pitchstone (bottom left), feldspar in Òigh-sgeir plutonic inclusions (second from left), free-floating feldspar in Sgùrr of Eigg pitchstone (third from left), and from plutonic inclusions in the Sgùrr of Eigg pitchstone (top right). Note the compositional ranges are virtually identical. The plutonic inclusions in both pitchstone groups show some sparse K-rich compositions that are not present in the free-floating feldspar populations. This indicates that K-rich feldspar that was liberated from the plutonic inclusions was dissolved in the pitchstone melt (cf. ref. 37 ).
SciEnTific REPORTS | (2019) 9:494 | DOI:10.1038/s41598-018-35855-w compatibility between the OSSEP and the source, (iii) evidence of silicic-mafic magma mixing, and (iv) comparable ages between the OSSEP pitchstones and the potential source. The crustal units of the Hebridean and the NW-Highland terranes of NW-Scotland have isotopic signatures that are distinct from each other and from Palaeogene mantle compositions (Figs 1 and 7) and isotopic characteristics have successfully been used to fingerprint contamination and volcanic sources within the BPIP mafic and felsic units (e.g. refs 30,33,[44][45][46] ). In the Hebridean Terrane, ascending Palaeogene magmas encountered granuliteand amphibolite-facies rocks of the Lewisian complex (2.5 Ga) exclusively, whereas in the Northern Highland Terrane, in addition to Archaean Lewisian rocks, Proterozoic metasediments of the Moine series (1.0 Ga) are present 3 . The rhyodacites from the Isle of Rum 31,33 , and the microgranites of the Western Red Hills of Skye 27,35 , are characterised by the typical Hebridean Terrane isotopic signature 23,27,47 . In contrast, the rhyolites in the composite cone sheets of nearby Ardnamurchan show Lewisian contamination closely followed by uptake of Moine lithologies 23,30 . The Òigh-sgeir and the Sgùrr of Eigg pitchstones indicate a common and solely Lewisian (Hebridean) contamination history and the OSSEP rocks are isotopically distinct from the felsic Palaeogene rocks of the NW-Highland Terrane (as e.g. recorded in the magmatic trends from Ardnamurchan and Mull; cf. refs 23,29,30 ) (Fig. 7). However, the silicic rocks from Rum and Skye are similar to the OSSEP suite in that they are characterised by solely Lewisian contamination histories 23,27,30,31,33 .
The Pb isotope data also exclude Moine contamination, because the Sgùrr of Eigg and Òigh-sgeir pitchstones are much less radiogenic than the Moine meta-sedimentary rocks or the assumed mantle in the region (e.g. ref. 48 ). The radiogenic isotope data presented thus support the initial observation of Dickin and Jones 23 that the Sgùrr of Eigg pitchstone magma experienced Lewisian crustal input only and hence must be sourced from the Hebridean Terrane, a conclusion that now applies to the Òigh-sgeir pitchstone also. However, although Dickin and Jones 23 and Emeleus 15 favoured an eruption centre for the Sgùrr of Eigg magma on, or very close to the island of Eigg, Brown and Bell 16 advocated a single regionally extensive pyroclastic deposit and suggested Skye to the North as the source area. Voluminous silicic magma that formed dominantly from crustal anatexis is most effectively produced near a large heat source 31 , yet there is no large gravity anomaly below Eigg 49,50 . Assuming one of the nearby large igneous centres with (i) a strong positive gravity anomaly, and (ii) a characteristic 'Hebridean Terrane' isotope signal to be the source for the OSSEP eruption, then indeed Rum and Skye are the only two realistic options (cf. refs 23,33 ).
Both the Rum and, especially, the Skye central complexes have major silicic components and both have strong positive Bouguer (gravity) anomalies indicative of underlying mafic intrusions 49,51 . Notably, however, the Rum rhyodacite and microgranite samples do not match with the OSSEP data for major elements or Sr, Nd and Pb-isotope compositions and different degrees of fractionation and crustal assimilation are recorded for the Rum rhyodacites and microgranites compared to the OSSEP suite (Figs 5, 6, 7; Supplementary Tables S1 and S4). For a specific contamination history of the OSSEP suite, see Supplementary Information S4. Rum therefore does not appear to provide a fit. On the other hand, we note that the OSSEP radiogenic isotope data consistently overlap with the isotope composition of the Red Hills granitoids on Skye 23,47 . This is in accord with major element similarities, including overlap with the mixed-magma Marsco granite suite, the erupted Skye trachyte lavas, and the mixed magma Glamaig Granite on Skye 35,36 (Figs 5 and 7).
Regarding oxygen isotopes, feldspar crystals from the Rum rhyodacites have δ 18 O values of ~+6.9‰ (ref. 52 ) and the granites on Skye have whole rock δ 18 O values of −4 to +7‰ (refs 53,54 ). In both, the Òigh-sgeir and the Sgùrr of Eigg pitchstones we find disequilibrium between phenocrysts and whole rock and we attribute the Figure 6. Oxygen isotope whole rock and feldspar crystal data of Sgùrr of Eigg and Òigh-sgeir pitchstones in comparison to known reference compositions. Fields for seawater, I-and S-type granites, Lewisian gneiss, Rum rhyodacite feldspar, Skye granites, Skye granite quartz (data from crystal interior) and global metasediments are given for reference. The Òigh-sgeir and Sgùrr of Eigg results overlap within the analytical uncertainty of the method. The data are, moreover, elevated relative to mantle values 28 (Fig. 6).  A final aspect is the geochronology of the region. The Rum igneous complex and its rhyodacites and microgranites are older than 60 Ma (between 60.3 and 60.5 Ma, refs 52,56 ) and can therefore not be the source of the OSSEP suite. The radiometric ages from Skye, however, compare favourably with the available radiometric age of the Sgùrr of Eigg pitchstone. The Loch Ainort Granite on Skye, for instance, has an Ar-Ar age of 58.58 ± 0.13 Ma (ref. 57 ), while the Sgùrr of Eigg pitchstone has an overlapping Ar-Ar age of 58.72 ± 0.07 (ref. 24 ). In addition, there is petrographic evidence of silicic-basic magma-mixing in the Skye granites of Glamaig and of Marsco in the Western Red Hills (e.g. refs 35,36,47 ). The available age constraints, combined with petrography, major elements and isotope data, thus argue strongly for Skye and especially the Western Red Hills as the most probable source for the OSSEP suite.

Assessing the Magnitude of the OSSEP Event
The ground between Skye and Òigh-sgeir was traversed by rivers in the Palaeogene that were in part filled with debris flows from Rum, as indicated by erosional clasts from Rum that are found in conglomerates within the lava flows of southern Skye and Canna 15,58 . Coupled with Moine-type quartzite pebbles detected in the SW-Skye conglomerates, a topographically low-lying palaeo-landscape between Skye and Rum is indicated (Fig. 8). At the time of the OSSEP eruption this river-dissected low-lying landscape (e.g. refs 14,15 ) contrasted with the older Rum edifice that was a likely significant topographic high. Accepting Skye as the vent location and Rum as an older and upstanding edifice, the exposures on Eigg and Òigh-sgeir ought to be viewed as two separate but co-eruptive flow lobes (Fig. 8). Linear run-out distances for eruptions from Skye (e.g. from the Marsco ring-dyke or the Glamaig granite) to Òigh-sgeir and Eigg would be ~45 and 41 km, respectively, which are not unreasonable for high-temperature pyroclastic density currents (e.g. ref. 59 ). Since pyroclastic flows are usually "ground-hugging", we would expect that the OSSEP deposit preferentially filled valleys and depressions. Because the OSSEP deposit is currently only preserved as former valley-fills distal to its assumed source, the original deposit must have been filling low ground elsewhere and was thus likely of considerable volume with presumably major effects on the landscape and biota of the region. The maximum established eruptive radius of 50 km (distance from Marsco to the marine ridge south of Muck) and the distance between Òigh-sgeir and the marine ridge (36 km, Fig. 8) allow calculation of an area of up to 1010 km 2 that was directly affected by the OSSEP event. Although, the OSSEP rock volume removed by erosion was likely significant (e.g. refs 15,24 ) we have no information on ignimbrite-filled river lengths, over-bank deposits or distal ashes (cf. ref. 59 ) (Fig. 8). Estimation of the magnitude of this ancient eruption is thus difficult. Attempting a first-order estimate for the OSSEP eruptive volume, we can use the distances from Skye to Eigg (~41 km) and to Òigh-sgeir (~45 km), together with the pitchstone cross-section on Eigg on the face exposed at Bidean Boidheach (~150 m in width, ~120 m in thickness). Underlying the Sgùrr of Eigg are conglomerates, giving the valley a U-shape 15,16 and U-shaped cross-section area, with total former valley width of 300 m was therefore chosen (cf. ref. 15 ), but we have also considered possible wider ones of 500 m and 750 m (Fig. S5). The minimum volumes using 300 m valley widths is 1.4 km 3 for a lobe from Marsco to Eigg and 2.5 km 3 for Marsco to Òigh-sgeir (550 m wide valley). These two lobes combine to 3.9 km 3 (dense-rock equivalent; DRE) and translates to a magnitude 5 on the Volcanic Explosivity Index (VEI). When an overbank facies is considered and a probably more realistic meandering river system, with for example ~80 or 100 km valley length (as indicated by field geology), the projected volume of the combined OSSEP event exceeds 10-15 km 3 of erupted material (Supplementary Table S5) and may reflect a VEI magnitude 6 eruption or higher. Although, these projected volumes are a first-order approximation, the derived magnitude of the OSSEP eruptions compares with historical examples such as the prolific 1991 Pinatubo eruption (~5 km 3 DRE) or the infamous 1883 Krakatau eruption (11 to 15 km 3 DRE) 60,61 , two of the largest eruptions in historical times. In the absence of data for the CO 2 and sulfur contents of the pitchstone magma, the climatic effects of the OSSEP event are difficult to estimate. To constrain the maximum CO 2 released from the OSSEP eruption we used comparable data from flood basalt eruptions of similar size as the OSSEP event (5 to 10 km 3 eruptive volume), which range between 0.7 and 0.14 Gt CO 2 (e.g. ref. 62 ). These values are small compared to annual anthropogenic CO 2 emissions of ~6.3 Gt/ year in the late 20 th century 63 and since the mass fraction of CO 2 in the melt is likely less in a felsic magma due to solubility relations 64 , the effects from magmatic CO 2 released by the OSSEP eruption were likely limited.
Assuming that the sulfur content in the OSSEP magmas was similar to those of silicic eruptives elsewhere (cf. ref. 65 ), the OSSEP eruption could have ejected some 1*10 14 g of sulfur into the atmosphere (see Supplementary  Information S6). This amount of sulfur is about an order of magnitude higher than recorded in historical VEI 5 and 6 events. Due to the short degassing time-scales and large amounts of sulphur released into the atmosphere during an explosive felsic eruption (cf. ref. 66 ), the OSSEP eruption could have caused direct climate and environmental effects for several years after the eruption in form of acid rains and by atmospheric cooling as a result of aerosol formation (cf. refs 25,66 ).
Coupled with the occurrence of widespread ash beds in E-Greenland, Denmark and the North Sea as well as with eroded ignimbrite vents elsewhere in the BPIP 4,[9][10][11]33,67,68 , our results imply that large-scale explosive silicic eruptions have likely been common during both phases of volcanic activity during the opening of the North Atlantic, including the British Palaeogene Igneous Province. This realisation paints a more violent picture of the rift to drift transition of the North-Atlantic region between 61 and 56 Ma than previously assumed. Moreover, the identification of the OSSEP event implies that ash-beds of this event, and likely similar events of the first phase of NAIP, may be present elsewhere, but poor resistance to weathering and erosion resulted in relatively few examples of this silicic volcanism being preserved outside the central complexes. Although individual silicic eruptions may have had only temporary climate effects, latest research indicates that several events and sources of volatiles are required to explain the build-up of climate active volatiles towards the Palaeocene-Eocene Thermal Maximum at ~56 Ma 8,26,69,70 . We therefore argue that the collective effects of repeated silicic volcanism from Palaeogene magmatic centres exposed in Scotland, Ireland and Greenland and those nowadays submerged in the N-Atlantic (e.g. the Blackstone bank) may have provided a so far underappreciated contribution towards the severe climate effects around the Palaeocene-Eocene boundary.

Methods
Samples. Samples SR303A and B (Òigh-sgeir pitchstone) and SR490, SR562 (Sgùrr of Eigg pitchstone) were selected from the sample catalogue in the BGS Memoir on Rum and the Small Isles 15 . The Sgùrr of Eigg pitchstone samples were collected by C.H. Emeleus and both outcrops are described in detail in Emeleus 15 . Reference specimens of these samples can be inspected with the British Geological Survey (www.bgs.ac.uk).
Major and trace elements. Major and trace element analyses were undertaken on the Sgùrr of Eigg pitchstone samples SR490 and SR562 and on Òigh-sgeir pitchstone samples SR303A and SR303B. Samples were analysed on a Spectro X-Lab EDP XRF at the University of St. Andrews. Analyses were carried out on fused beads and all analyses were performed with a Rh tube. Calibration was performed using international geological reference standards. The data are compared to the range of available compositions from these and related localities (e.g. refs [14][15][16]35,36   step size between every single analysis in the grid was 20 or 30 µm and the analysed area in a single grid was thus between 100 × 100 µm to 230 × 230 µm. The analyses were subsequently averaged for the respective sample and normalized. Spectrometers 1 and 2 used TAP crystals to analyse Na, Al and Si, Mg respectively. Radiogenic isotope analysis. Sr, Nd, and Pb isotope analyses were conducted on two whole-rock samples from the Sgùrr of Eigg pitchstone (SR490, SR562) and two from the Òigh-sgeir pitchstone (SR303A and SR303B) (Supplementary Table S3). The analyses were performed at the Scottish Universities Environmental Research Centre (SUERC), East Kilbride, UK, following procedures as outlined in Meyer et al. 31  , and analytical blanks were <1 ng. All isotope ratios have been age-corrected to 59 Ma (ref. 24 ).

Stable isotope analysis.
Oxygen isotope ratios in whole rock and mineral samples were acquired using a conventional silicate extraction line (for whole rocks) and a laser fluorination line (for crystals), combined with a Thermo DeltaXP mass spectrometer at the University of Cape Town, South Africa. Approximately 10 mg of powdered sample was dried in an oven at 50 °C and degassed under vacuum on a conventional silicate extraction line attached to externally heated Ni vessels at 200 °C (refs 41,72 ). Silicates were reacted with ClF 3 (ref. 73 ), and the liberated O 2 was converted to CO 2 using a hot platinised carbon rod. For analytical details of the laser line see Vennemann and Smith 74 . The results are reported in standard δ-notation, where δ = (R sample /R standard − 1) × 1000 and R = 18 O/ 16 O. Samples were run in tandem with duplicate samples of the internal quartz standards (MQ) which calibrates the raw data to SMOW (Standard Mean Ocean Water; e.g. ref. 72 ), and is equivalent to V-SMOW using a δ 18 O value of 10.1 for MQ (calibrated against NBS-28). During the course of this study, the analytical error for δ 18 O is estimated to be ± 0.2‰ (2σ) for whole rocks and ± 0.3‰ (2σ) for mineral analysis, based on long-term duplication of MQ.

Data Availability Statement
All data generated or analysed during this study are included in this published article (and its Supplementary Information files).