Magma pathways in sill-complexes

This report documents the findings of a one-week fieldtrip, and associated lab analyses, to study the Loch Scridain Sill-complex (LSSC) on the Isle of Mull, Scotland. We gratefully acknowledge the support of the VMSG, who funded this work through the Henry Emeleus Award (2019). Magee designed the project and was aided in the field by Martin, who worked on the area as part of his PhD, and MacDonald. Rock magnetic experiments and anisotropy of magnetic susceptibility analyses were conducted at the M3Ore laboratory in the University of St Andrews. All raw data acquired during this work are archived at Figshare (link provided at report end) and freely available.


Introduction
With over 600 million people globally living near active volcanoes, understanding how volcano plumbing systems control eruption location, and style of precursor warning signals, is a fundamental issue facing volcanologists (e.g., Sparks et al. 2012;Cashman & Sparks 2013;Cashman et al. 2017). The deeply eroded roots of ancient volcanoes preserved in the British and Irish Palaeogene Igneous Province (BIPIP) arguably birthed our understanding of subvolcanic plumbing systems (e.g., Bailey 1924;Richey & Thomas 1930;Anderson 1936Anderson , 1937Emeleus & Bell 2005), and it is thus integral to constantly revisit this archetypal area and test new magma emplacement hypotheses. For example, classical models of ring dyke and cone sheet emplacement, developed through examination of BIPIP central complexes, have been revisited and tested (e.g., O'Driscoll et al. 2006;Stevenson et al. 2007;O'Driscoll et al. 2008;Stevenson & Bennett 2011;Emeleus et al. 2012;Magee et al. 2012). Whilst interpretations of magmatic processes within the BIPIP broadly conform to a traditional view of plumbing systems, involving (sub-)vertical transport of magma between reservoirs and eruption sites ( Fig. 1A) (e.g., Tibaldi 2015;Cashman et al. 2017), a consensus is now emerging that laterally injected intrusions (dykes or sills) can link adjacent volcanic systems ( Fig. 1A) (e.g., Biggs et al. 2016;Magee et al. 2016b;Schofield et al. 2017). In particular, recent studies champion a new paradigm whereby magma flow through a network of sub-horizontal (sills) and inclined sheets (i.e. a sill-complex) can connect volcanoes across 1-1000's km (Fig. 1A) (e.g., Svensen et al. 2012;Magee et al. 2016b;Eide et al. 2017;Schofield et al. 2017), challenging our understanding of how volcanic systems work.
Sill-complexes cover areas up to >1.5×10 6 km 2 , and comprise thin (<100 m thick) sills and inclined sheets injected over ~0.5-15 Myrs (e.g., Chevallier & Woodford 1999;Thomson & Hutton 2004;Planke et al. 2005;Cartwright & Hansen 2006;Leat 2008;Svensen et al. 2012;Coetzee & Kisters 2016;Magee et al. 2016b;. Seismic reflection data reveal that flow pathways within sill-complexes can follow tortuous routes (e.g., Cartwright & Hansen 2006;Magee et al. 2016b;Schofield et al. 2017). These convoluted pathways are expected to increase the time magma takes to reach the surface (Magee et al. 2016b). Conversely, because flow occurs through relatively thin intrusions, magma may be expected to solidify quickly (months to a few thousand years) (e.g., Holness & Humphreys 2003;Annen 2011;Annen et al. 2015), certainly before such extensive sill-complexes can form (Magee et al. 2016b). A solution to this paradox is that sill-complexes form via incremental magma input, where solidified yet hot injections elevate the local temperature, allowing younger pulses to remain molten and flow for longer (e.g., Magee et al. 2016b). It has been hypothesised that these young injections of into sill-complexes are focused either: (i) in distinct channels in a sill (Fig. 1B); or (ii) along margins of earlier sills (i.e. accretion; Fig. 1C) (e.g., Annen et al. 2015;Magee et al. 2016b). These untested hypotheses imply only certain flow pathways are active at any one time.
Therefore, although previous studies have examined the broad structure of sill-complexes and how individual intrusions inject (e.g., Hansen et al. 2004;Thomson & Hutton 2004;Cartwright & Hansen 2006;Thomson & Schofield 2008;Schofield 2009;Schofield et al. 2012;Magee et al. 2014;Magee et al. 2016b;Schofield et al. 2017;Magee et al. 2019a), we do not know how entire sill-complexes are built or whether they actually allow connected volcanoes to dynamically interact.
The Loch Scridain Sill-complex on the Isle of Mull affords an opportunity to examine how magma moves through and builds sill-complexes (Holness & Humphreys 2003). For example, field and petrological evidence indicates the Tràigh Bhàn na Sgùrra Sill is segmented and comprised several thick channels, separated by thin sill portions, that facilitated longer-lived magma flow (Holness & Humphreys 2003;Stephens et al. 2017). We integrate structural field analysis and rock magnetic techniques in an attempt to identify and characterise magma flow pathways in the LSSC, with a view to expanding the study to understand how the entire sill-complex was constructed.

Geological Setting
Igneous rocks are widespread on the Isle of Mull, Scotland and primarily represent the plumbing system or eruptive products of magmatic activity related to the formation of the ~62-55 Myr old British and Irish Palaeogene Igneous Province (BIPIP), which forms part of the North Atlantic Igneous Province ( Fig. 2A) (e.g., Bailey 1924;Emeleus & Bell 2005). Sheet intrusions on Mull include regional dykes, cone sheets, and the LSSC. The ~58 Ma LSSC is a wellexposed, mafic in composition, and relatively small (~100 km 2 ) sill-complex likely linked to the formation of the Mull Central Complex, which is considered the deeply eroded roots of an ancient volcanic edifice (Figs 2A and B) (Bailey 1924;Preston et al. 1998;Holness & Humphreys 2003;Stephens et al. 2017). Palaeostress analysis particularly reveal the LSSC was emplaced during horizontal radial compression induced by central complex formation (Stephens et al. 2017). Here, we focus on the Tràigh Bhàn na Sgùrra Sill, which is located along the SW coast of Mull where it is intruded into steeply to vertically dipping, metasedimentary psammites and pelites of the Precambrian Moine basement (Figs 2B and C) (Holness & Humphreys 2003;Stephens et al. 2017). The petrology and chemistry of the Tràigh Bhàn na Sgùrra Sill has been well characterised and comprises olivine-poor and olivine-free dolerite, which commonly hosts Fe-Ti oxides (see Holness & Humphreys 2003 and references therein). Critically, sections where sill thickness exceeds ~3.5 m have no chilled margin and a wide contact aureole, in contrast to thinner sections, suggesting magma was channelized into these thick zones (Fig. 2C); modelling of these observations is consistent with sustained magma flow within thicker sections of the sill for 5-6 months, with the thinner sections crystallising in days (Kille et al. 1986;Wartho et al. 2001;Holness & Humphreys 2003). Lineations are also visible throughout the sill at various structural levels and have been interpreted to reflect magma flow ( Fig. 2C) (Holness & Humphreys 2003).
Other possible field flow indicators include steps (Stephens et al. 2017), which are connectors between two subhorizontal sill elements.

Methods
As magma intrudes, it cools and begins to crystallise; injected magma may also carry a crystal cargo extracted from a previous storage site. The arrangement, size, shape, and orientation of these crystals, i.e. the rock texture, record kinetic (e.g., magma flow fabrics) and equilibration processes (e.g., Holness & Humphreys 2003;Higgins 2006;Magee et al. 2013). It may therefore be expected that accretion and channelization, where parts of the sill-complex remain hotter for longer, produce a distinct textural record that can be evaluated via integrated rock fabric analyses.
Of the 17 total block samples collected (B1-B18; there is no B11), most were acquired from vertically oriented transects across the sill, targeting areas where the sill is thin and thick (Holness & Humphreys 2003), whereby a sample was taken from the top, middle, and base of the intrusion (one transect only has samples from the top and base). The rationale for this sampling strategy was to analyse whether magnetic fabrics differed across the intrusion and along its strike, particularly between areas where magma flow within the sill was longer-lived (thick sections) and those where it was limited (thin sections) (Holness & Humphreys 2003). We also collected one sample from a step connector between two segments (B1), and one from an area where lineations at multiple levels within the sill are visible in the field (B5); the aim of these was to test whether the connector or lineation long axes aligned with the magnetic fabric. Here, we use anisotropy of magnetic susceptibility (AMS) to measure magnetic fabrics from 12 oriented block samples collected from the Tràigh Bhàn na Sgùrra Sill of the LSSC; a further five samples were collected but their analysis was delayed by the Covid-19 pandemicdata for these will be added once acquired. Right cylinder sub-specimen drill core samples measuring 22 mm by 25 mm were extracted from each block sample using a rock coring drill press and a diamond-tipped, non-magnetic saw blade. Each sub-specimen was analysed using an automated 3D rotator attached to a KLY-5a Kappabridge operating at a low alternating field of 400A/m at 1220 HZ at the University of St. Andrews M 2 Ore Laboratory, and the results averaged for each block using Agico Anisoft 42 software.

Rock magnetic experiments
As exemplified by the recognition that magnetic fabrics may be normal, intermediate, or inverse, the interpretation of AMS requires knowledge of the mineralogy controlling its signature. The interpretation of AMS data is subject to several caveats. We conducted thermomagnetic experiments on one sample (i.e. B1) to provide insight into the composition, grain size, and domain state of the mineral phase(s) controlling the AMS response (e.g., Dunlop & Özdemir 2001). These experiments involve powdering a sub-specimen in a ceramic pestle and mortar to acquire a 0.3 g powder sub-sample. The powder was cooled in liquid nitrogen and magnetic susceptibility was measured as the sample heated from -194°C to 0°C using a KLY5 CS-L cryostat. The sample was placed in a KLY5 CS4 furnace and, in an argon atmosphere, magnetic susceptibility was measured from 25°C to 700°C and back to 40°C. Finally, the susceptibility was again measured from -194°C to 0°C in the KLY5 CS-L. The results were reduced in Cureval 8 software where a continuous temperature vs susceptibility plot was generated. From our thermomagnetic experiment we estimated the Curie point using the inflection point method (Tauxe et al., 1998); this transition can be used to interpret the magnetic mineralogy of each sample and determine what mineral phases control the AMS response (e.g., Dunlop & Özdemir 2001).

Rock magnetic experiments
The heating curve shows a steady increase in susceptibility from room temperature to 320°C where a sharp decrease in susceptibility occurs (Fig. 3); this profile is characteristic of titanomagnetite and the absense of an abrupt hopkinsons peak is consistent with the dominance of multidomain-to-pseudosingle domaign ferromagnetic grains.
From its Curie temperature we determine that the Ti component of the Fe²⁺(Fe³⁺, Ti)₂O₄ titanomagneite is ~0.41 (Akimoto 1962). The heating and cooling curves are not reversable and we note that the Curie point shown on the cooling curve occurs at a much lower temperture than the heating curve, yet the measured starting and final bulk suscetpability values are very closely comparable (Fig. 3). These differences in the heating and cooling curve profiles indicates that a mineralogical change occurred during the experiment and that this change did not produce or destroy substantial volumes of magnetite; we interpret these results to show the resorption of titanium into titanomagnetite.

Observations
Here we present data for vertical sampling transects across a headland exposure of the Tràigh Bhàn na Sgùrra Sill, which currently includes samples B2-B4, B8-B10, B12, B14, and B17-B18; once analysed we will add the remaining samples. The AMS sub-specimen data for each block sample are consistent, producing small and well-defined confidence ellipses, with Km overall ranging from 1.4-2.94×10 -2 ( Fig. 4; Table 1). Most samples, with the exception of B2 (T = 0.53; oblate), contain triaxial-to-prolate magnetic farbics (T = -0.002 --0.735) ( Fig. 4; Table 1). The only full transect currently available is samples B2-B4, where the sills is ~2 m thick, which all show south-easterly (143.5-155.9°) trending K1 axes that increase in plunge from 5.7° at the top of the sill to 20.2° (middle) and 29.7° at the base ( Fig.4; Table 1). Conversely, for B2-B4, the magnetic foliation dip decreases from 67.8° at the top of the sill to 45.8° (middle) and 38.1° at the base; magnetic foliation strike changes from 158.3° at the top of the sill to 164.5° (middle) and 011.6° at the base ( Fig. 4; Table 1). The contact orientation at the top of the sill near B2 (158/68° SE = magnetic foliation) is, on average, 035/17° SE, whereas the contact orientation at the top of the sill near B2 (012/38° SE = magnetic foliation) is, on average, 056/20° SE ( Fig. 4; Table 1). For both B2 and B4, K1 is located close to the plane of the intrusion (defined by contact orientations) (Fig. 4). The intersection between the magnetic foliations of B2 and B4 coincides with the B4 K1 axis (Fig. 4). For other samples from vertical transects there is a variable relationship between the orientation of K1 and the magnetic foliation relative to the adjacent sill contact (Fig. 4). For example, the magnetic foliations of B12 and B17 broadly strike parallel to the sill contacts and, although they dip SE instead of NW, they dip ate similar values ( Fig. 4; Table 1). Conversely, for samples B9 and B18, the magnetic foliation is near orthogonal to the sill strike and dip ( Fig. 4; Table 1). In no sample does K` closely parallel features identified in the field as possible magma flow indicators (Fig. 4).

Interpretation
The Km range of 1.4-2.94E×10 -2 is consistent with the thermomagnetic experiments results that indicate multidomain titanomagnetite-to-pseudosingle is the principal contributor to the magnetic signature of the samples (Fig. 3).
However, we note that some magnetic fabrics (e.g., B9 and B18) are near-orthogonal to the plane of intrusion, which may suggest that these few fabrics are inverse and could be dominated by a single-domain magnetite population (e.g., Ferré 2002); i.e. for normal magnetic fabrics, K1 and K2 are expected to be within or near the plane of intrusion.With most samples displaying triaxial-to-prolate fabrics, interpretation of magnetic foliations has to be considered carefully; critically we note that although prolate fabrics occur because the magnitude of K2 and K3 are similar, the well-defined confidence ellipses we derive suggest that we can constrain K2 and thus use the K1-K2 plane as the magetic foliation (Fig. 4). Importantly, we show magnetic fabric orientations are variable (Fig. 4), suggesting fabrics within the sill have not been overprinted by syn-or post-emplacement deformation. Based on the variation in fabric orientation and the evidence the magnetic signature is carried by titanomagnetite, with little influence from other phases, we suggest the magnetic fabrics likely have a magmatic origin.
We currently lack the full set of samples to determine whether flow patterns were different between thicker, and perhaps longer-lived (Holness & Humphreys 2003), portions of the sill and other sections. If we consider the magnetic foliations of B2 and B4, which sample the sill margin adjacent to the host rock contact, are normal (Ferré 2002) and define a flow-related imbrication our results would imply magma, at least locally, towards the NW (see Geoffroy et al. 2002). Whether we can relate K1 to the magma flow direction (e.g., Knight & Walker 1988;Callot et al. 2001;Magee et al. 2012) is difficult to determine yet due to its different plung relative to that of visible field flow indicators (e.g., assumed flow lineations and step axes) (Fig. 4). This difference between the field flow indicator orientations and magnetic farbics could be because: (i) they each record separate stages of magma emplacement; (ii) the block samples where collected in an area where magma flow was different (i.e. magma flow may have been heterogeneous over short distances; Magee et al. 2013;Magee et al. 2016a); or (iii) one or both do not record magma flow, but instead reflect other processes.

Observations
Here we test further whether the magnetic lineations relate to possible indicators of magma flow observed in the field.

Interpretation
Sheet intrusions such as sills commonly emplace as discrete elements that through continued magma flow and element infaltion, coalesce to form a throughgoing sheet (Fig. 6) (see Magee et al. 2019b and references therein).
These element typically occur at slightly differnet stratigraphic and/or structural levels, such that when they coalesce, they connect via the formation of a step or briken bridge connector (Fig. 6) (Magee et al. 2019b;Köpping et al. 2021).
Critically, the long axes of these connectors is parallel to the sheet propagation axis (Fig. 6) (Magee et al. 2019b;Köpping et al. 2021). The parallelism between the sample B1 K1 trend and that of the intrusive step it is associated with suggest the former may be considered a proxy for the sheet emplacement axis (Fig. 4). Lineations observed along surfaces exposed within and throughout (i.e. not just at the host rock contact) the Tràigh Bhàn na Sgùrra Sill have previously been inferred to record magma flow axes (Holness & Humphreys 2003).
Our work, particularly the disparity we show between the magnetic fabric and the lineations of sample B5, questions whether this inferred link between the visible lineations and magma flow is valid.

Concluding Remarks and Future Work
The aim of this study was to test whether there is a systematic variation in the magnetic fabric across the the Tràigh Bhàn na Sgùrra Sill relative to areas where the sill was short-and long-lived. Although we have not been able to analyse all samples yet, our results currently are sufficient to indicate: (i) fabrics do vary, implying they are not related to a tectonic overprint; (ii) deciphering the magnetic mineralogy will be key to understanding why fabrics may vary, particulalrly those magnetic lineations and foliations oriented near-orthogonal to the sill plane; and (iii) there is rarely a clear relationship between magnetic lineation and features observed in outcrop that may be interpreted as mimicking the magma flow axis (i.e. flow lineations and steps). Future work within the scope of our study required to improve our understanding of magma flow patterns and channelization in the Tràigh Bhàn na Sgùrra Sill include: (i) finish analysing the AMS of collected blocks, which will allow us to identify any imbrication of magnetic fabrics from which magma flow direction can be inferred; (ii) conduct further rock magnetic experiments to better characterise the magnetic mineralogy of the sill (e.g., more thermomagnetic experiments, Lowrie-Fuller test); (iii) analysing a few samples for anisotropy of anhysteretic remanent magnetisation, which can help identify whether AMS fabrics are the product of sub-fabric averaging; and (iv) petrological studies to determine how the magnetic fabric relates to the orientation of silicate phases and whether textural variations (e.g., crystal size distribution, equilibration) vary across the sill.