Emplacement and Segment Geometry of Large, High-Viscosity Magmatic Sheets

: Understanding magma transport in sheet intrusions is crucial to interpreting volcanic unrest. Studies of dyke emplacement and geometry focus predominantly on low-viscosity, maﬁc dykes. Here, we present an in-depth study of two high-viscosity dykes (10 6 Pa · s) in the Chachahu é n volcano, Argentina, the Great Dyke and the Sosa Dyke. To quantify dyke geometries, magma ﬂow indicators, and magma viscosity, we combine photogrammetry, microstructural analysis, igneous petrology, Fourier-Transform-Infrared-Spectroscopy, and Anisotropy of Magnetic Susceptibility (AMS). Our results show that the dykes consist of 3 to 8 mappable segments up to 2 km long. Segments often end in a bifurcation, and segment tips are predominantly oval, but elliptical tips occur in the outermost segments of the Great Dyke. Furthermore, variations in host rocks have no observable impact on dyke geometry. AMS fabrics and other ﬂow indicators in the Sosa Dyke show lateral magma ﬂow in contrast to the vertical ﬂow suggested by the segment geometries. A comparison with segment geometries of low-viscosity dykes shows that our high-viscosity dykes follow the same geometrical trend. In fact, the data compilation supports that dyke segment and tip geometries reﬂect different stages in dyke emplacement, questioning the current usage for ﬁnal sheet geometries as proxies for emplacement mechanism.


Field Work and Photogrammetry
The field work for this study took place in November 2018 at the Chachahuén volcano in Argentina. To aid the field observations of the Sosa Dyke and the Great Dyke (Figure 1) we used a DJI Phantom 4 Pro UAS (Unmanned Aerial System) to take overlapping photos (Resolution: 5472 × 3648 pixels) for a subsequent photogrammetric analysis. The acquired drone photos were the input for Structure from Motion (SfM) photogrammetry analysis using Metashape (Agisoft Version 1.7.0). We apply the default workflow in Agisoft Metashape to generate two virtual outcrop models of the Sosa Dyke and Great Dyke, respectively (available via: https://v3geo.com/, accessed on 4 October 2021). In addition, we generated the orthophotos of the Sosa Dyke (285 photos, orthophoto resolution: 5 cm) and Great Dyke (498 photos, orthophoto resolution: 5 cm) from the 3D data. Subsequently, we used the software MOVE 2019 (by Petroleum Experts) to map the dyke geometries and segments. For each dyke we created virtual cross-sections orthogonal to the strike of the dyke (cross section spacing: 5 m) over its full length. Each exported cross-section contained the coordinates of the intersection points of the mapped dyke outline with the cross-section-plane. This data was imported into MATLAB 2019b (MathWorks ® ) and used to calculate the thickness of the dykes for each virtual cross-section. The MATLAB code including the two example datasets for the Sosa and Great Dyke are available via github [51] (doi:10.5281/zenodo.4784164). This allowed us to quantify the distribution of the thicknesses for each segment of the Sosa Dyke and the Great Dyke. In the case of a bifurcation in the dyke, we computed multiple thicknesses and considered the thickest measurement as part of the data. Note that parts of segments that overlap other segments of the dyke were not considered in the calculation of average thickness and standard deviation. Further, to quantify the geometry of each segment we measured the radius r tip of the segment tips, the width w (long axis) and the thickness t, (short axis) of a segment, following a method developed by Pollard et al. [25] and recently used by Stephens et al. [28].

Rock Samples and AMS Analysis
We collected oriented rock samples along two traverses across the Sosa Dyke and rock samples from each margin and centre of the Great Dyke. The two traverses across the Sosa Dyke are located ( Figure 2): in the middle of a segment traverse T1 and close to the tip of a segment traverse T2, both with a sample spacing of 5 m. The collected rock samples were drilled and 9 to 14 drill cores (24 mm diameter, 21 mm length) per sample, suitable for anisotropy of magnetic susceptibility (AMS) analysis, were extracted. Only sample T1-09 contained too many fractures such that no cores could be recovered. Low-field AMS was employed to investigate the flow direction in two of the dyke segments of the Sosa Dyke. The measurements of the in-phase anisotropy of the magnetic susceptibility (ipAMS) were performed at the Laboratory for Experimental Paleo-magnetism at the Department of Earth Sciences, Uppsala University in Uppsala, Sweden. We performed the measurements with an Agico Kappabridge MK1-FA (Agico, Brno, Czech Republic) in semi-automatic spin-mode (magnetic field: 200 A/m, frequency 976 Hz). Symmetric second rank magnetic susceptibility tensors were determined from the susceptibility measurements using the least square inversion method of Jelínek [52]. The eigenvalues and eigenvectors of the magnetic susceptibility tensor define the value and orientation of the three orthogonal, principal axes of susceptibility, k 1 ≥ k 2 ≥ k 3 , represented by a triaxial ellipsoid [53]. The following parameters are used to describe the magnetic fabric responsible for the AMS signal [54] in the Sosa Dyke: (1) K mean (K mean = (k 1 + k 2 + k 3 )/3), (2) P j the corrected degree of anisotropy: P j = exp 2 ((η 1 − η m ) 2 + η 2 − η m ) 2 + (η 3 − η m ) 2 (1) with η x = ln(k x ), x = 1, 2, 3 and η m = η 1 +η 2 +η 3 3 , where a value of 1 represents an isotropic fabric and values >1 increasing degrees of fabric anisotropy, (3) T the shape factor (T = 1, rotational oblate; T = −1, rotational prolate, T = 0, spherical) is defined as: (4) L the magnetic lineation is defined as: L = (k 1 /k 2 ), and (5) F the magnetic foliation (F = (k 2 /k 3 ) were obtained with the software Anisoft5 (Agico, Version 5.1.08) [55].
The out-of-phase AMS (opAMS) was measured on half of the cores of traverse T1 and all of the cores of traverse T2 using a KLY-5a Kappabridge at the University of St. Andrews M 3 Ore lab in 3D automatic spin mode with a magnetic field of 400 A/m and at a frequency of 1440 Hz. Out-of-phase AMS measured the viscous magnetic response as the applied magnetic field is switched off [56,57]. Out of phase AMS may therefore help single out the signal of magnetically viscous grains in ipAMS and identify the fabric of small pseudo single-domain magnetite. In some instances, opAMS can work as a substitute to anisotropy of anhysteretic magnetic remanence [57].
To characterise magnetic carriers, we performed temperature versus magnetic susceptibility measurements, demagnetisation of the natural remnant magnetization and anhysteretic remanence magnetization on selected samples. The temperature vs. susceptibility measurements were performed with the CS4 attachment to the Agico Kappabridge MK1-FA by heating powdered samples to 700 • C in an argon atmosphere and subsequently cooled back to room temperature. The samples were demagnetized and magnetized with an Agico LDA 5 AF demagnetizer FA and Agico PAM1 FA, respectively at the M 3 Ore lab at the University of St. Andrews. The sample magnetization was measured with a Agico JR6-a spinner magnetometer with an automatic sample holder. The variation in susceptibility at different temperatures and demagnetization/magnetization behaviour of the samples can help us to distinguish the ferromagnetic mineral phases responsible for the AMS signal in the dyke samples. In addition to the measurement of the AMS, we performed anisotropy of anhysteretic remanence magnetization (AARM) and anisotropy of isothermal remanence magnetization (AIRM) on selected cores to further pinpoint the carriers of the AMS fabrics. The samples were demagnetized using 150 mT alternating field (AF) with a 2-axis tumbling specimen holder, prior to imparting an ARM on six sample orientations according to orientation scheme of Hext [58] and Jelínek [59]. The ARM was imparted using a bias direct current field of 500 µT in a peak AF of 130 mT. For AIRM, the samples were imparted with an IRM in six orientations using a single, one second pulse direct current field of 20 mT for samples from traverse T1 and 15 mT on samples from traverse T2. The different IRM fields for the two traverses was based on the sample' NRM demagnetization and ARM magnetization ( Figure A4). The samples were demagnetized in a field of 150 mT using a tumbling specimen holder before imparting the IRM on each position. The results of the magnetic characterisation, AARM and AIRM measurements are presented in detail in Appendix B.

Geochemistry and Petrographic Analysis
In addition to the dyke's internal structure and propagation direction, the rheology of the transported magma is a crucial parameter for the dyke emplacement and geometry. The main parameters to describe the rheology of a magma is the chemical composition of the melt, and the size and content of the vesicles and particles, dispersed and transported in it.
From the rock samples we prepared 17 powder samples (15 g/sample) for whole rock major and trace element analysis (Appendix A.1). The major element analysis was performed via lithium borate fused bead ICP-AES (Inductively Coupled Plasma-Atomic Emission Spectroscopy) with major element oxide detection limits of 0.01 wt% (Cr 2 O 3 : 0.001 wt%; ALS Geochemistry: Me-ICP06, Vancouver, Canada) and LOI (Loss on Ignition) by furnace or thermogravimetric analysis (ALS Geochemistry: OA-GRA05/ME-GRA05).
The trace element geochemistry was analysed via fused bead lithium borate ICP-MS (Inductively Coupled Plasma-Mass Spectrometry, ALS Geochemistry: MeMS81™) with trace element oxide detection limits varying from 0.01 ppm (caesium) to 5 ppm (vanadium). In both major and trace element analysis the sample-lithium borate fused beads underwent acid digestion before being analysed in the ICP-AES/ICP-MS. The full geochemical dataset can be found in Appendix A.
We followed the recommendation of Jackson et al. [60] to not use the LOI values derived from whole rock geochemistry as proxy for the H 2 O content of the rocks, due to the potential larger error. Instead, we used the water content measured via FTIR spectroscopy of the magmatic glass [61,62]. The ERZLABOR in Freiberg/Germany prepared four doublesided polished thick sections of glass rich samples for FTIR (Fourier Transform Infrared) spectroscopy. Three of the thick sections from samples of the Sosa Dyke margin, to test the reproducibility of the values and one sample from the Great Dyke margin. The aim of the FTIR analysis was to precisely determine the H 2 O content in the glass as a proxy for the H 2 O content in the melt during emplacement. The FTIR analysis was performed at the Swedish Museum of Natural History in Stockholm/Sweden using a Bruker Vertex spectrometer equipped with a Globar source and a Potassium bromide (KBr) infrared beamsplitter. A Mercury Cadmium Telluride detector (MCT) and a Hyperion 2000 microscope were used to acquire spectra in the wavenumber range 4000-600 cm −1 with a resolution of 4 cm −1 . On each thick section sample ten glass-matrix locations were measured to ensure reproducibility of the measurements within the sample ( Figure A1). For the H 2 O-content calculation we compared two different models: (1) single extinction coefficient ε model, using the averaged ε H2O at 3550 m −1 of andesite and rhyolite [61], and (2) the endmember extinction coefficient model with two distinct ε H2O and ε OH at 3550 m −1 for molecular water and hydroxyl groups, respectively [62,63].
To investigate the size and composition of the vesicles and particles, the bubbles and phenocrysts, respectively, we used polished thin sections of the rock samples prepared at the ERZLABOR in Freiberg/Germany. The petrographic analysis comprises the main mineral determination as well as point counting (c. 300 counts per thin sections; Table A4). The point counting values were used as a proxy for the crystal content and differentiated between matrix and phenocrysts.
Finally, the results from: (1) the FTIR spectroscopy derived values of H 2 O content of the glass, together with (2) the major element whole geochemistry, and (3) the crystal content calculated from the thin sections, were used as input parameters for MAGMA (KWare Geological Software: https://www.lanl.gov/orgs/ees/geodynamics/Wohletz/ KWare/Index.htm, accessed on 4 October 2021) developed by Ken Wohletz (Los Alamos National Laboratory). The magma viscosity calculation is based on the viscosity model after Bottinga and Weill [64]. The record of input and output of the MAGMA tool is available in Supplementary Material "MAGMArecord.zip".

Sosa Dyke
The Sosa Dyke is located in the Northern part of the Chachahuén volcano, with a total length of c. 1300 m (measurement along strike; Figure 1). Its exposure suggests that the dyke is a vertical sheet (~88 • ) with an overall strike of 277 • . Due to outcrop quality, this study will focus on the eastern 950 m of the Sosa Dyke only, where we mapped three different dyke segments (I-III, Figure 2). The host rock into which the Sosa Dyke intrudes is a succession of (sub-)horizontally bedded block-and-ash-flows and dacitic to rhyolitic ignimbrites. The Eastern exposed tip of the Sosa Dyke is only c. 600 m in strike direction away from a massive igneous intrusion, the Cerro Bayo cryptodome (Figures 1 and 3A) [49].
The eastern most segment, Segment I, is at 560 m the longest mapped segment of the Sosa Dyke. It has an average thickness of around 42 m with a c. 50 m long area in the middle of the segment, which is significantly thicker (up to 65 m) compared to the rest of the segment (30-50 m; Figure 2). At this thickened mid-point, the strike of the segment changes from 299 • to 274 • . The East end of segment I exhibits a narrowed, rounded tip, whereas the Western end of segment I and the East end of segment II, are hardly exposed (illustrated by the straight border between segments I/II in Figure 2). Segment II is the shortest of the three mapped Sosa Dyke segments with a length of c. 210 m and a strike of 296 • . This segment shows a gradual increase in thickness from its tips (35 m/25 m) to the centre (60 m) with an average thickness of 48 m. The Western end of segment II displays an overall blunt tip geometry (="squared ellipse shape/superellipse"; see [65,66] for details), with two short (<10 m) pointy secondary tips at each corner ( Figure 2). Segment III strikes 277 • and has an average thickness of about 63 m. Our observations indicate that segment III is not exposed in its entirety in the field. We confidently mapped a length of 203 m for segment III, although the orthophotos indicate the total length could be 10-32% longer. Due to insufficient outcrop conditions the termination and tip geometry of segment III in the West could not be observed. The East end of segment III has an offset of 57 m to the North of segment II with a segment overlap of 53 m. Similar to segment II, the East end of segment III shows a blunt geometry. However, in contrast to segment II, we observe that only one prominent secondary tip (c. 40 m length, c. 23 m thick) has formed at the Northern corner ahead of segment III. In the field, host rock is exposed between segment II and segment III.
A common observation for the mapped segments of the Sosa Dyke seems to be the development of secondary tips towards the northeast on the East ends of the segments and towards the southwest on the West end of the segments ( Figure 2). All segments, with the exception of segment II, display overall parallel margins that are smooth at the scale of the dyke. At outcrop scale, however, we observed a number of cuspate bulges that are part of the primary dyke geometry, i.e., not caused by erosion, as they exhibit chilled margins similar to the rest of the dyke. Furthermore, we observed large scale fractures in the Sosa Dyke, of which the majority showed a convex curvature towards the East (Figures 2 and 3A,B).
The chilled margin of the Sosa Dyke is dark and up to 15 cm thick, exhibiting a glassy matrix followed by a gradual transition (10-20 cm thick) into a microlite-dominated matrix towards the centre of the dyke ( Figure 3B). The part of this transition zone between glass-and microlite-dominated matrix closest to the dyke centre is often marked by a set of parallel fractures orthogonal to the dyke margin. These band-like sets of fractures were only a few cm in length, however neither field nor optical microscope revealed any apparent reason for why their opening stopped. The outermost margin of the Sosa Dyke contains a set of joints orthogonal to the dyke contact with a centimetre to decimetre spacing and up to 25 cm in length. This chilled margin texture is accompanied by a c.25 cm wide zone of aligned hornblende and plagioclase crystals/crystal agglomerates parallel to the contact of the dyke and the host rock. Even so the alignment of crystals gets lost towards the centre of the dyke, the overall crystal content derived through point counting, stays similar with c. 35% at the margin and 37% in the centre of the Sosa Dyke ( Figure 3D).
Along the contact of the Sosa Dyke with host rock, we observed centimetre to 0.8 m thick, vein-like structures that extended from the dyke margin into the host rock for several metres, striking at an angle of 15-30 • to the dyke. The veins are filled with light grey lithics with grain sizes ranging from clay to coarse sand and exhibiting fluidized sediment textures. The lithics seem to comprise both host rock material and occasionally dyke rock (Figures 2 and 3E,F).

Great Dyke
The Great Dyke is located in the north-western part of the Chachahuén volcano, with a total length of c. 2000 m (Figure 1, Figure 4). Similar to the Sosa Dyke, the Great Dyke is a vertical sheet (~90° dip) and weathers out as a high wall from the surrounding host rocks ( Figure 5). The Great Dyke strikes

Great Dyke
The Great Dyke is located in the north-western part of the Chachahuén volcano, with a total length of c. 2000 m (Figures 1 and 4). Similar to the Sosa Dyke, the Great Dyke is a vertical sheet (~90 • dip) and weathers out as a high wall from the surrounding host rocks ( Figure 5). The Great Dyke strikes 218 • , i.e., at an angle (c. 25 • ) to the caldera wall scarp (see Figure 1), cutting through the caldera infill consisting of block and ash flows and rhyolitic ignimbrite. Towards its southern end, the Great Dyke is exposed cutting through the entire caldera wall and pinching out in the lavas exposed in the highest elevations south of the caldera wall. The overall appearance of the dyke is a straight alignment of eight segments, with a decrease in thickness towards the outer segments ( Figure 4).
Segment I in the Northeast has a length of 78 m and a rather irregular shape with a strike of 197 • . The segment reaches up to 27 m in its central part, but decreases in thickness towards the edges. The north-eastern tip of segment I (the end of the dyke) is rounded, whereas the opposite tip is poorly defined as it is in contact with segment II (Figure 4).
Segment II is the shortest (45 m) of the eight mapped segments and the segment with the largest deviation in strike (strike: 240 • ) compared to the overall strike of the Great Dyke (218 • ). It has an average thickness of around 19 m with only minor changes in thickness towards the edges ( Figure 4).
Segment III has a strike of 211 • . Its north-eastern tip is bifurcated; one end connects to segment II the other end pinches out into a thin apophysis (length: 17 m, width: c. 7 m) with an offset towards the North. The length of segment III (excluding the apophysis) is 95 m, its thickness averages at 22 m with several irregular indentations along the margins ( Figure 4). Segment IV has a length of 145 m and a strike of 218 • . The north-eastern end of segment IV is blunt and offset towards the North in relation to segment III to which it is connected. Segment IV has an average thickness of 23 m and straight parallel margins, resulting in a box-shaped/superellipse geometry. Only at the south-western end we observe a decrease in thickness down to 10 m at the connection to segment V ( Figure 4).
Segment V strikes at 227 • with an average thickness of 25 m and a length of around 200 m (tip to tip; Figure 4). It is the only segment with a prominent bifurcation on both ends. Similar to the two previous segments, the north-eastern end of the segment is offset to the North in relation to is northern neighbour segment (i.e., segment IV). The north-eastern end of segment V consists of: (1) a North-striking apophyse of 33 m in length, c. 7 m thick and with a pointy tip, and (2) the connection with segment IV. The bifurcation at the southwestern end of segment V forms: (1) a connection with segment VI in the northern arm of the bifurcation, and (2) a South-striking apophysis of 22 m length and c. 7 m thickness and a pointy tip as the southern arm.
Segment VI is the thickest segment (average thickness 34 m) and located in the centre of the Great Dyke. It has a strike of 215 • and a length of 93 m. In contrast to the previous three segments (III-V), there is no bifurcation and offset to the North on the north-eastern end of the segment. In contrast, the south-western end of segment VI exhibits a potential secondary tip of a bifurcation towards the south, similar to segment III. The main (northern) part of this bifurcation is connected with the next segment VII.
Segment VII is by far the longest segment in the Great Dyke, with a length of 1110 m and a strike of 218 • . The thickness of segment VII is fairly homogenous (average thickness 27 m), due to the overall parallel dyke margins. However, our data shows a minor decrease in thickness to below 20 m over the last 80 m at each end. Segment VII displays the largest vertical exposure, both as a free-standing wall in the caldera and cutting through the cliff of the caldera wall. Within the cliff, no changes in thickness occur in different host rocks ( Figure 5B). In contrast to most of the other segments of the Great Dyke, segment VII does not show any bifurcation at its ends. Both ends display a blunt geometry and an incision at the contact to the neighbouring segments. Minerals 2021, 11 12 of 37

Anisotropy of Magnetic Susceptibility (AMS)
The samples in traverse T1 in the Sosa Dyke have a mean susceptibility ranging from 17 to 26 × 10 −3 SI. The degree of anisotropy Pj ranges from 1.02 to 1.04 and the shape of the AMS ellipsoid is moderately prolate to moderately oblate (T = −0.62 to 0.7). The ipAMS foliation is moderately to steeply dipping and sub parallel to the strike of the dyke in six samples of the traverse T1 (T1-01, T1-02, T1-03, T1-05, T1-06 and T1-07) ( Figure 6). The foliation in the remaining two samples are perpendicular to the strike of the dyke (T1-04 and T1-08). The k1 orientation plots on or close to the dyke plane in samples T1-05, T1-07, T1-06, whereas it is obliquely oriented to the dyke plane in T1-01, T1-02 and T1-03, and perpendicular to the dyke plane in T1-08 and T1-04 (Figure 6). The magnetic foliation dip towards the NNE in T1-01, T1-02, T1-03 and is parallel to the dyke plane T1-05, T1-06 and T1-07. The opAMS is co-axial to the ipAMS in sample T1-05 ( Figure A5). In T1-06, T1-07 and T1-08, the opAMS are non-coaxial to the ipAMS, which could indicate the presence of a magnetic subfabric carried by smaller magnetite crystals ( Figure A5). Segment VIII is exposed on a plateau, 200 m above the caldera depression. The segment is 220 m long with a strike of 217 • . The thickness of segment VIII (14 m on average) reaches a maximum of 22 m in its centre. Towards the south-western end (end of the Great Dyke) the segment pinches out entirely. Hence, segment VIII is the segment with the largest asymmetry in geometry when comparing the tips of all dyke segments.
In the outcrop, the contact between the Great Dyke and its host rocks is sharp and characterised by a chilled margin in the dyke. In some outcrops, this chilled margin is weathered away, while the dyke interior is weathering out as a wall, with the north-western side often being up to 50 m higher in altitude than the south-eastern side of the dyke. Along the dyke walls, the chilled margin is up to 20 cm thick, dark and glass-rich with centimetre to decimetre spaced fractures. The chilled margin is dominated by elongated plagioclase phenocryst aggregates, parallel to the contact of the dyke. It is followed by a transition zone (c. 25-30 cm thick) with a gradually decreasing amount of glass in the matrix and an increasing number of microcrystals towards the centre of the Great Dyke. Notably, at segment ends, the thickness of the glassy chilled margin is much higher, in some places up to <50 cm.
Throughout the Great Dyke we observe centimetre to decimetre in size, mafic enclaves in the dyke centre. Most of the enclaves are orientated with their longest axis parallel to the strike of the dyke ( Figure 5C). The crystal content is 30%. In contrast to the crystals at the margin of the Great Dyke, plagioclase agglomerates and amphibole group minerals (i.e., hornblende) in the dyke centre show no preferred orientation ( Figure 5D). In a narrow stretch of about 30-50 m along the southern margin of segment VII we observed hornblende megacrysts up to 15 cm in length (Figure 4).

Anisotropy of Magnetic Susceptibility (AMS)
The samples in traverse T1 in the Sosa Dyke have a mean susceptibility ranging from 17 to 26 × 10 −3 SI. The degree of anisotropy P j ranges from 1.02 to 1.04 and the shape of the AMS ellipsoid is moderately prolate to moderately oblate (T = −0.62 to 0.7). The ipAMS foliation is moderately to steeply dipping and sub parallel to the strike of the dyke in six samples of the traverse T1 (T1-01, T1-02, T1-03, T1-05, T1-06 and T1-07) ( Figure 6). The foliation in the remaining two samples are perpendicular to the strike of the dyke (T1-04 and T1-08). The k1 orientation plots on or close to the dyke plane in samples T1-05, T1-07, T1-06, whereas it is obliquely oriented to the dyke plane in T1-01, T1-02 and T1-03, and perpendicular to the dyke plane in T1-08 and T1-04 ( Figure 6). The magnetic foliation dip towards the NNE in T1-01, T1-02, T1-03 and is parallel to the dyke plane T1-05, T1-06 and T1-07. The opAMS is co-axial to the ipAMS in sample T1-05 ( Figure A5). In T1-06, T1-07 and T1-08, the opAMS are non-coaxial to the ipAMS, which could indicate the presence of a magnetic subfabric carried by smaller magnetite crystals ( Figure A5).
In the traverse T2 of the Sosa dyke the degree of anisotropy ranges from 1.01 to 1.04 and the shape of the AMS ellipsoid is moderately prolate to oblate (T = −0.67 to 0.83). The degree of anisotropy is lower at the margin and in the central part of the dyke compared to samples from between the margin and centre (Figure 7). The mean k 1 orientation is trending parallel to the dyke plane in samples T2-01, T2-04, and T2-05, whereas the k 1 orientation is slightly oblique to the dyke plane in T2-03, and T2-06. The magnetic foliation is dipping steeply towards the NW in T2-02, T2-03, and T2-04 and steeply to shallowly towards the SW in sample T2-05 and T2-06 ( Figure 7). The opAMS shows higher degrees of anisotropy than the ipAMS in T2 samples. The opAMS is co-axial to the ipAMS in sample T2-02, T2-04, and T2-06 however, yield a higher spread of the data both in orientation and in fabric shape ( Figure A6). In T2-01, and T2-05, the opAMS are non-coaxial to the ipAMS, which could indicate the presence of a magnetic subfabric carried by smaller magnetite crystals. ( Figure A6). All magnetic data are given in Supplementary Material "Magnetic data.xlsx".
To identify carriers of the AMS fabric, we performed temperature vs. susceptibility experiments. Those temperature vs. susceptibility experiments together with K m values of 10 −2 in our samples suggest that the AMS is dominated by Ti-magnetite (see Figure A3). The low degrees of anisotropy in our samples (Pj 1.01 to 1.04) may indicate that there are several competing ferrimagnetic components of the petrofabric contributing to the AMS fabric, such as the distribution and shape fabric of the Ti-magnetite or smaller SD magnetite [67][68][69][70][71]. The principal axes of the in-phase and out of phase AMS generally agree, although the distribution of the axes of individual sub-specimens in opAMS are more widely scattered (Figures A5 and A6). Interestingly, in a sample with oblate ipAMS, the opAMS may be prolate but the principal axes of the sub-specimens plot along the k 1 and k 2 girdle of ipAMS ( Figures A5 and A6). The opAMS therefore reflect the fabric of smaller magnetite in the groundmass of the sample, which is essentially groundmass fabric [56,57]. Thus, the agreement of the ipAMS and opAMS suggest that the AMS fabric reflect the primary magmatic flow fabric of the dyke. See appendix B for a detailed discussion on the carriers of the AMS fabric in the Sosa dyke samples.  In the traverse T2 of the Sosa dyke the degree of anisotropy ranges from 1.01 to 1.04 and the shape of the AMS ellipsoid is moderately prolate to oblate (T = −0.67 to 0.83). The degree of anisotropy is lower at the margin and in the central part of the dyke compared to samples from between the margin and centre (Figure 7). The mean k1 orientation is trending parallel to the dyke plane in samples T2-01, T2-04, and T2-05, whereas the k1 orientation is slightly oblique to the dyke plane in T2-03, and T2-06. The magnetic foliation is dipping steeply towards the NW in T2-02, T2-03, and T2-04 and steeply to shallowly towards the SW in sample T2-05 and T2-06 (Figure 7). The opAMS shows higher degrees of anisotropy than the ipAMS in T2 samples. The opAMS is co-axial to the ipAMS in sample T2-02, T2-04, and T2-06 however, yield a higher spread of the data both in orientation and in fabric shape (Figure A6). In T2-01, and T2-05, the opAMS are non-coaxial to the ipAMS, which could indicate the presence of a magnetic subfabric carried by smaller magnetite crystals. ( Figure A6). All magnetic data are given in Supplementary material "Magnetic data.xlsx".

Geochemistry
The results of the whole rock geochemistry from the sampled traverses T1 and T2 on the Sosa Dyke and the centre and margin samples of the Great Dyke show a trachyte/trachydacite composition. The Sosa Dyke samples display a more mafic composition for the centre parts of the dyke compared to the dyke margins (Figure 8; Tables A1 and A2). Rare earth element (REE) data and trace element data do not show any apparent differences between margin and centre of the dyke (Table A1). The sample GD-C6 of the Sosa dyke was taken in addition to a traverse close to one of the clast-filled veins and shows a lower alkali content than the rest of the samples (Table A2).
The whole rock composition of the Great Dyke is that of a trachyte/trachydacite. Notably, the centre sample shows the same composition as the Cerro Bayo cryptodome [49] (Figure 8), whereas the margin sample shows a composition similar to the Sosa Dyke.
The results of the whole rock geochemistry from the sampled traverses T1 and T2 on the Sosa Dyke and the centre and margin samples of the Great Dyke show a trachyte/trachydacite composition. The Sosa Dyke samples display a more mafic composition for the centre parts of the dyke compared to the dyke margins ( Figure 8; Table A1 and Table A2). Rare earth element (REE) data and trace element data do not show any apparent differences between margin and centre of the dyke (Table A1). The sample GD-C6 of the Sosa dyke was taken in addition to a traverse close to one of the clast-filled veins and shows a lower alkali content than the rest of the samples (Table A2).
The whole rock composition of the Great Dyke is that of a trachyte/trachydacite. Notably, the centre sample shows the same composition as the Cerro Bayo cryptodome [49] (Figure 8), whereas the margin sample shows a composition similar to the Sosa Dyke.

Viscosity
The whole rock geochemistry, total water content (FTIR), crystal content and crystal size (thin sections) from this study (Table 1) are combined with

Viscosity
The whole rock geochemistry, total water content (FTIR), crystal content and crystal size (thin sections) from this study (Table 1) are combined with the temperature and pressure estimates from Burchardt et al. [49]. They obtained temperatures of 790 • C and 850 • C based on the lowest temperature estimates for plagioclase and amphibole crystallization, respectively. The pressure of 10 MPa was estimated using the lithostatic pressure with an approximate emplacement depth of several hundred metres, derived from amphibole breakdown rims for the Cerro Bayo cryptodome [72]. Given the proximity of the Sosa Dyke and the Cerro Bayo (Figure 1), we used the same pressure estimate. These values of temperatures and pressure were used as input parameters for the MAGMA Version 2.50.0160 (Kware) software tool to calculate the viscosity of the two dykes. The results of this calculation provide an average effective viscosity of η eff of 2.42 × 10 6 Pa·s (790 • C) and 1.17 × 10 6 Pa·s (850 • C) for the Great Dyke and 5.20 × 10 6 Pa·s (790 • C) and 2.54 × 10 6 Pa·s (850 • C) for the Sosa Dyke.

Segment Offset-Initial Dyke Geometry or Tectonic Feature?
The interpreted orthophoto ( Figure 2) displays a significant offset between segment II and III of the Sosa Dyke. Perez and Condat [48] interpreted the segment offset to be associated with a post-emplacement tectonic fault due to its alignment with large-scale regional lineaments in the area. However, the occurrence of a continuous chilled margin around both ends of segments II and III, which are related to the fast cooling (quenching) of hot magma in contact with a cold host rock during the initial phase of dyke emplacement [73], argues against faulting between the dyke segments. The absence of a chilled margin in contact with the host rock, i.e., along a fault plane, would indicate that the magma was already cooled down, and most likely solid, to accommodate the tectonic stresses in a brittle manner (fault). This shows, that the observation of offsets between segments alone is not conclusive evidence for faulting. Many field and seismic studies showcase step-features and segment offsets related to initial flow geometries [21,25,28,35,[74][75][76] rather than postemplacement tectonic events. The recent work by Galland et al. [27] explains that intrusion segment offsets can be easily misinterpreted as tectonic fault-related offsets rather than an intrusion emplacement feature. Thus, considering our detailed field observations (cf. Figure 1), we conclude that the segmented shapes of the dykes, including the long offset between segment II and III of the Sosa Dyke, are primary and related to the emplacement of the dykes, rather than to post-emplacement tectonic faulting.

Interpretation of Magma Flow Direction in the Sosa Dyke
Despite its large size, the Sosa Dyke does not show any internal borders, such as chilled margins or sudden changes in magma composition, i.e., hybridisation [77] or phenocryst content which would be indicative for its emplacement as a composite dyke [77][78][79]. Thus, we conclude that the Sosa Dyke was emplaced as one continuous flow rather than multiple successive flow injections.
The outcrop of the Sosa Dyke revealed zones of aligned crystals and mafic enclaves parallel to the strike of the dyke ( Figure 3E), and oriented curved fractures across the dyke (Figures 2 and 3), both suggesting a lateral magma flow direction towards the East. An indepth analysis for magma flow related magmatic fabrics of our traverse T1 samples showed that the magnetic foliation strikes E-W in the southern half of the dyke. In the northern half of the dyke, the magnetic foliation strikes NE-SW. In samples with dominantly prolate fabric (in the centre of the dyke), the lineation is sub-parallel to the strike of the dyke and dips moderately to the West ( Figure 6). Overall the intersection between the steeplydipping magnetic foliations indicates a dominantly lateral flow direction towards NNE. In the traverse T2 samples, the magnetic fabric on the northern half of the dyke is steeply to moderately dipping and strikes WNW-ESE. On the southern half of the dyke, the magnetic foliation is steeply dipping and strikes NE-SW (Figure 7). The abundance of other studies showcase that AMS is a commonly used method to interpret the flow direction in dykes [33][34][35]53,78,[80][81][82]. It is expected that the AMS in sheet intrusions are imbricated due to the interaction between the magma and the intrusion wall [33]. The intersection between the AMS fabric at the opposite walls of a sheet intrusion therefore reveals the magma flow vector. Thus, we conclude that the magma flow indicators in the Sosa Dyke record a dominantly lateral magma flow towards the East.
The lateral magma flow direction inferred from the AMS data is potentially in contradiction with the segment shape of the dykes. Segmentation of igneous sheet intrusions has been extensively documented through seismic data ( [74,75,83] and references therein) and field observations ( [25,27,29,83] and references therein), [84][85][86]. Segments are usually interpreted as magma flow indicators, as supported by (1) the three-dimensional shape of radial finger distribution on seismic data(e.g., [74,87], (2) structural field measurements indicating segment bridges are magma flow markers [27], and (3) laboratory experiments [88]. According to these studies, the segment geometry of the Sosa Dyke indicates a dominantly upward magma flow, which is contradictory with the AMS measurements that indicate overall lateral magma flow. One explanation of such a mismatch is that the segment geometry might indicate the overall magma propagation direction while AMS fabrics might indicate the stage of lateral inflation of the upward propagating finger [85], such as evidenced by laboratory experiments [89]. Alternatively, the works by Poland et al. [35], Townsend et al. [90], and Healy et al. [18] on mafic to felsic dykes show that dominant lateral magma flow can also result in dyke segment geometries similar to what was observed in this study. Further studies on changes in the magma flow direction during sheet propagation and subsequent magma transport through established sheets (e.g., [91]) is necessary to potentially solve the discrepancy between the observed dyke geometries and flow indicators.

High-Viscosity Sheet Intrusions and the Effect of Different Host Rocks on the Intrusion Geometry
We first verify that the Sosa and Great Dyke are good examples for the geometries of thick, high-viscous magmatic sheets. Therefore, we compare the overall dyke geometry, represented by the maximum thickness and length values, with other examples of highviscous sheet intrusions in the world. The 67 m thickness/1300 m length and 39 m thickness/2000 m length, for the Sosa and Great Dyke, respectively, are a good match with the well-studied high-viscosity silicic dykes from Summer Coon volcano in Colorado/USA (50 m thickness/2000-7000 m length; e.g., [35,92] and references therein). In addition, the dimensions and field observations of the Sosa and Great Dyke are in agreement with the data collection by Cruden et al. [2], plotting in between long mesoscale sheets and thin laccolith. Thus, we conclude that the Sosa and Great Dyke are valid representatives of commonly known, thick, high-viscosity dykes/(sub-)vertical sheet intrusions.
To further strengthen the representativity of the Sosa and Great Dyke, we calculated the effective viscosities for each sample, which are on the order of~10 6 Pa·s for both the Sosa Dyke and the Great Dyke. This places the Sosa and Great Dyke well in the range of viscosities for intermediate-to high-viscosity magmas (10 4 to 10 8 Pa·s) with a phenocryst content of up to 50% (e.g., [93,94]). It is important to note that the calculated viscosities provide only a bulk value for the magma viscosities in the Sosa and Great Dyke, whether viscosity changed spatially (position in the dyke) and with time (initial intrusion to final solidification) cannot be resolved.
Magma viscosity is only one important parameter that affects the geometry of the Sosa and Great Dyke, another one is the host rock they are emplaced in. Our field observations show that both the Sosa and the Great Dyke emplaced into various host rocks, i.e., basaltic lavas, block and ash flows, of different mechanical strengths (e.g., [95]) without a systematic change in dyke geometry (Figures 2, 5 and 9). This suggests that host rock lithology and strength play a minor role during the emplacement of the studied dykes. This is in contrast to the findings of recent field studies [21,27,96] and laboratory/numerical modelling studies [37,[97][98][99][100][101][102], which associate mechanically different host rock types with different intrusion geometries. Recent studies e.g., [95,103,104] show that the variation in mechanical strength between different rock types is commonly smaller than one order of magnitude. Hence, the range of host rock strengths is significantly smaller than the range of magma viscosities observed in nature, which can easily vary by six orders of magnitude (10 2 to 10 8 Pa·s) [93,105] or more [106]. The models of e.g., Galland et al. [99] and Schmiedel et al. [98] suggest that the emplacement of magma is governed not only by the magma viscosity or the host rock strength, but by the balance between the viscous stresses within the flowing magma and the deforming host rock. Given the high viscosity of the magma emplaced in the Sosa and Great Dykes, these models propose that the viscous stresses within the magma dominate the emplacement of the dykes over the host rock properties.

Segment Geometry and Dyke Emplacement
Dyke segment tips are often used to infer the model of emplacement e.g., the linear elastic fracture mechanics model (LEFM), viscous indenter model, etc., correlated to magma viscosity and host rock strength (e.g., [22,36,107]). The tips of the segments of the Sosa and Great Dyke tend to be oval (Figure 9). The comparison of segment geometry data from this study with data of other sheet intrusion segments with viscosities ranging from lowviscosity dykes (mafic; [21,25,28]) to intermediate/high-viscosity sills [27] does not allow for a distinction in segment (tip) shape towards our high-viscosity dykes. Furthermore, even though the sheet intrusion data sets record magma emplacement into various host rocks, no correlation between host rock type and sheet segment geometry is apparent. These results question the common practice of using dyke segment geometries as proxy for a single intrusion emplacement mechanism (cf. [28,40]). In order to reveal the emplacement and propagation mechanism, it is therefore necessary to also describe the host rock structures associated with the creation of space for the magma [26,27,36].
Minerals 2021, 11 21 of 37 Figure 9. Dyke segment geometries. The plot is redrawn and modified after Pollard et al. [25] and Stephens et al. [28]; squares represent data from this study. Roman numbers correspond to the segments of the Sosa Dyke (SI-SIII) and the Great Dyke (I-VIII) in Figure 2 and Figure 4, respectively (Data: Appendix C). Colours represent different host rocks. The vertical line indicates two data points belong to the same segment (e.g., left and right tip of a segment); no line, means only one data point (tip) was observed for that segment. tsegment thickness, w-segment width, rtip-radius of segment tip curvature. (inset) Schematic simplification of the different types of observed intrusion segment shapes and their parameters; squared ellipses can be expected in the plot above the "oval" line, whereas sharp ellipse shapes are displayed below the "ellipse" line.

Conclusions
Our field observations on the geometry of the two exceptionally well exposed magmatic sheets: the Sosa Dyke and the Great Dyke, in the Miocene Chachahuén volcano in Argentina combined with magma flow indicators (fractures, crystals, AMS) within the dykes, and their calculated magma viscosity lead us to conclude that: Figure 9. Dyke segment geometries. The plot is redrawn and modified after Pollard et al. [25] and Stephens et al. [28]; squares represent data from this study. Roman numbers correspond to the segments of the Sosa Dyke (SI-SIII) and the Great Dyke (I-VIII) in Figures 2 and 4, respectively (Data: Appendix C). Colours represent different host rocks. The vertical line indicates two data points belong to the same segment (e.g., left and right tip of a segment); no line, means only one data point (tip) was observed for that segment. t-segment thickness, w-segment width, r tip -radius of segment tip curvature. (inset) Schematic simplification of the different types of observed intrusion segment shapes and their parameters; squared ellipses can be expected in the plot above the "oval" line, whereas sharp ellipse shapes are displayed below the "ellipse" line.
A challenge of interpreting the shapes of segment and their tips in terms of magma emplacement is that outcrops offer the end-product of a complex, potentially multi-stage phenomenon. Stephens et al. [28] propose that distinct intrusion segment geometries can represent successive stages during the emplacement of a sheet intrusion rather than showcasing distinct emplacement mechanisms: (1) initiation: segments thin tapered tips (Mode I, tensile opening), which later on can be transformed into (2) oval/squared ellipse tip shapes (inelastic deformation or viscous indenter model), governed by the magma viscosity and host rock strength. These authors only tested their model on outcrops of low-viscosity sheet intrusions. The newly acquired data sets from this study seem to corroborate their "multistage-model" also in the range of high-viscosity sheet intrusions (Figure 9). Even if the multistage model seems to fit most of the field observations of the data sets investigated (see Figure 9 for references) it remains unclear if the proposed order of stages is always the same. The elasto-plastic, analytical model of Scheibert et al. [108] for example suggests the opposite, i.e., a more prominent inelastic deformation during the earlier stage of emplacement.
A key observation on the Sosa and Great Dyke is that the walls of each segment are (sub-)parallel, i.e., each segment exhibits a relatively constant thickness along-strike, except close to the tips. This shape is really similar to those of igneous fingers described by Spacapan et al. [26], who show that their entire emplacement was accommodated by pushing of the host rock entirely without Mode I opening. Moreover, the parallel-wall shape is also very similar to the laboratory viscoelastic fingers of Bertelsen et al. [32], which are emplaced in relatively weak host matrix. Conversely, the laboratory intrusions of Bertelsen et al. [32] into stronger host rock exhibit tapered shapes only at the end of the experiments. This discussion highlights that the final intrusion shape is not sufficient to reveal the entire emplacement history of magmatic sheet intrusions. One potential solution to overcome this limitation could be to better integrate rock physical methods into models for magma emplacement aided by detailed field observations of rock failure mechanisms and structures in magma and host rock, which would enable the performance of a time-resolved, quantitative structural back restoration.

Conclusions
Our field observations on the geometry of the two exceptionally well exposed magmatic sheets: the Sosa Dyke and the Great Dyke, in the Miocene Chachahuén volcano in Argentina combined with magma flow indicators (fractures, crystals, AMS) within the dykes, and their calculated magma viscosity lead us to conclude that:

•
The segment geometries observed in this study are related to the emplacement of the dykes, rather than the result of post-emplacement tectonic deformation; • AMS and curved fractures record a dominant lateral magma flow direction in the Sosa Dyke towards the East, thus contradicting expected vertical magma flow derived from dyke segment geometry. It remains unsolved if the flow indicators thus record a late-stage of a dominantly vertical dyke emplacement or an overall lateral dyke emplacement; • The comparison between low-to high-viscosity magma sheet segment geometries reveal a significant similarity. Thus, we question the common practice of using sheet segment geometries as proxies for different intrusion emplacement mechanisms of low-and high-viscosity sheet intrusions; • The similarities between low-and high-viscosity sheets emplaced within host rocks of different strengths suggest that magma sheet emplacement is governed not only by the magma viscosity or the host rock strength, but by the balance between the viscous stresses within the flowing magma and the deforming host rock. The final intrusion shape is not sufficient to reveal the entire emplacement history of magmatic sheet intrusions.
To overcome the limitation of field observations as the final snapshot of an intrusion emplacement history, future work should concentrate on the time-resolution during magma emplacement of potentially various different emplacement mechanisms. One potential solution could be the better integration of rock physical methods into models for magma emplacement aided by detailed field observations of rock failure mechanisms and structures in both magma and host rock. This would help with the undertaking of a quantitative back restoration of structures and related processes and thus help to improve current geophysical and geodetic models for eruption forecasts. In addition, for the growing field of planetary volcanology with surface imagery as the main data source, we need a solid understanding of related magma transport in the subsurface to be able to interpret extra-terrestrial processes e.g., [109].

Acknowledgments:
The authors are grateful for the field assistance, hospitality and help from the Sosa family and Rubén Castillo. We thank Sam Poppe and Tara Stephens for easy access to background information on their published data. Moreover, we are grateful for the support from Emma Rhodes, Hélène Breton, and Bjarne Almqvist with sample preparation, software and equipment at Uppsala University. Comments from participants of the LASI IV conference, as well as discussions with Taylor Witcher and Christoph Hieronymus have helped to develop the contents presented in this manuscript. The Swedish Geological Survey is acknowledged for facilitating core drilling of samples. The authors acknowledge the use of the Move Software Suite granted by Petroleum Experts academic software initiative. We thank Simon Buckley and Conor Lewis from the Virtual Outcrop Geology Group (VOG Group) for the fast and uncomplicated publication of our 3D outcrop models on their platform V3Geo (https://v3geo.com/, accessed on 4 October 2021).

Conflicts of Interest:
The authors declare no conflict of interest.

Appendix A. Whole Rock Geochemistry and Fourier Transform Infrared Spectroscopy (FTIR) Analysis
Appendix A.1. Whole Rock Geochemistry Table A1. Whole-rock major (wt%) and trace element (ppm) ICP analyses of samples from the Sosa dyke traverse T1 for sample locations see Figure 6.   (GD)   SAMPLE  T1-01  T1-02  T1-03  T1-04  T1-05  T1-06  T1-07  T1-08  T1-  The results for both of the tested FTIR models to determine the total H2Ocontent from the glass phase in the thick section samples are in general higher than the LOI values determined through whole rock geochemistry. In detail, for the Great Dyke the total H2O-content calculated by the FTIR single extinction coefficient model shows an on average a 112% higher value for the total water content compared to the LOI value determined by the whole rock geochemistry (0.6 wt%; Figure A2). For the FTIR endmember extinction coefficient model, using two different extinction coefficients for molecular water and for the hydroxyl groups, we see an on average 150% higher value of total H2O-content compared to the LOI value. In case of the Sosa Dyke, the determined values for water are minimally higher (3%) for the FTIR single extinction coefficient model and 24% higher for the FTIR endmember extinction model compared to the values derived from whole rock geochemistry ( Figure  A2). The results for both of the tested FTIR models to determine the total H 2 O-content from the glass phase in the thick section samples are in general higher than the LOI values determined through whole rock geochemistry. In detail, for the Great Dyke the total H 2 O-content calculated by the FTIR single extinction coefficient model shows an on average a 112% higher value for the total water content compared to the LOI value determined by the whole rock geochemistry (0.6 wt%; Figure A2). For the FTIR endmember extinction coefficient model, using two different extinction coefficients for molecular water and for the hydroxyl groups, we see an on average 150% higher value of total H 2 O-content compared to the LOI value. In case of the Sosa Dyke, the determined values for water are minimally higher (3%) for the FTIR single extinction coefficient model and 24% higher for the FTIR endmember extinction model compared to the values derived from whole rock geochemistry ( Figure A2).

Appendix B. Characterisation of AMS Carriers in the Sosa Dyke Rocks
The temperature vs. susceptibility (T-χ) experiments of the Sosa dyke samples reveal an initial increase in magnetic susceptibility to between 600 to 700 × 10 −6 SI at 290 °C. Subsequently the susceptibility decreases to about 400 to 500 × 10 −6 SI at 450 to 500 °C where the decrease is briefly levelling off before the susceptibility again drops more rapidly ( Figure A3). Several of the samples from the interior of the dykes display an additional bump in the T-χ susceptibility curve at about 200 × 10 −6 SI between 600 to 620 °C. During cooling, the samples display a slow increase in susceptibility to 600 °C when the susceptibility starts to increase more rapidly until 500 °C, after which the increase levels of. The bulk susceptibility of the samples of the Sosa dyke at room temperature are up to 90% lower after cooling compared to before heating ( Figure  A3). The presence of low Ti-Magnetite in the samples is indicated by the observed drop in susceptibility of around ~550 °C [110], whereas the presence of Maghemite is indicated by the bump in the heating curve at around 300 °C [111][112][113]. The non-reversable behaviour of the samples during heating and

Appendix B. Characterisation of AMS Carriers in the Sosa Dyke Rocks
The temperature vs. susceptibility (T-χ) experiments of the Sosa dyke samples reveal an initial increase in magnetic susceptibility to between 600 to 700 × 10 −6 SI at 290 • C. Subsequently the susceptibility decreases to about 400 to 500 × 10 −6 SI at 450 to 500 • C where the decrease is briefly levelling off before the susceptibility again drops more rapidly ( Figure A3). Several of the samples from the interior of the dykes display an additional bump in the T-χ susceptibility curve at about 200 × 10 −6 SI between 600 to 620 • C. During cooling, the samples display a slow increase in susceptibility to 600 • C when the susceptibility starts to increase more rapidly until 500 • C, after which the increase levels of. The bulk susceptibility of the samples of the Sosa dyke at room temperature are up to 90% lower after cooling compared to before heating ( Figure A3). The presence of low Ti-Magnetite in the samples is indicated by the observed drop in susceptibility of around~550 • C [110], whereas the presence of Maghemite is indicated by the bump in the heating curve at around 300 • C [111][112][113]. The non-reversable behaviour of the samples during heating and cooling indicate that Ti-Magnetite partially altered to Ti-Maghemite is the main magnetic phase in the samples [cf. 112]. Furthermore, the decrease in susceptibility recorded in samples T1-04 to T1-06 at temperatures higher than 600 • C could indicate either the presence of Ti-Maghemite or Ti-Hematite [112]. The alteration of maghemite during heating explains why the magnetic susceptibility curves are non-reversable during cooling [111] The interior of the Great dyke displays a similar T-χ curve as the Sosa dyke. The samples from the margin of the dyke, on the other hand, lack a distinct Maghemite bump in the heating curve. However, a small bump can be observed ( Figure A3). During cooling, the susceptibility increases rapidly from <100 × 10 −6 SI at 600 • C to 980 × 10 −6 SI at 420 • C before steadily decreasing to 50 • C. The cooling curve indicates the presence of Ti-rich magnetite or Ti-Hematite formed during the heating experiment [112,114].
NRM AF demagnetisation of selected sub-specimens from the Sosa dyke show a different demagnetisation response for traverses T1 and T2. Samples from traverse T1 display a gradual decrease in magnetization. About 50% of the magnetization is lost in fields in between 30 to 50 mT. Subsequently, the magnetization linearly decreases until 200 mT. Samples from traverse T2 lose 80% of their magnetization in an applied magnetic field below 40 mT before the magnetization lies mostly constant with the increasing field. The samples of traverse T1 of the Sosa dyke indicate the presence of a magnetic mineralogy that contain both low and high coercivity phases. The high coercivity phase is likely the Tihematite that is indicated by the T-χ experiments. ARM acquisition was also performed on the sub-specimens ( Figure A4). All samples are 90% saturated by an AC field of between 30 and 50 mT. The ARM magnetization shows that magnetite is the main remanence carrying grain that is picked up with AARM and AIRM measurement in the specimens.
Ti-Maghemite forms during subaerial (or near surface) low-temperature alteration of Ti-magnetite and can be altered to Ti-Hematite during heating [114]. The presence of Ti-hematite in the samples therefore suggest a shallow emplacement depth leading to an oxidizing environment of the magma, which may have led to alteration of original Ti-magnetite to Ti-hematite.
In order to identify the carriers of the ipAMS fabric in the Sosa dyke samples, we employed AARM, AIRM and opAMS to distinguish eventual magnetic subfabrics. Representative sub-specimens from each sample were selected for comparison between different magnetic fabrics. AARM and AIRM are controlled by magnetic remanence and are used to measure the fabric of ferrimagnetic minerals in isolation from other rock forming minerals [70,[115][116][117]. Low-field AIRM can single-out the contribution of low-coercivity grains such as multi-domain magnetite to the AMS fabric [118][119][120], whereas AARM also can be used to identify the fabric of both MD ferrimagnetic particles and higher coercivity single-domain ferrimagnetic particles [70,[115][116][117]. Modelling by Hrouda et al. [57] show that ipAMS, opAMS and remanence magnetization may be carried by different types of magnetic particles. In a sample where ipAMS, opAMS and Anisotropy of Magnetic Remanence (AMR) are coaxial, the AMS fabric is likely carried by MD magnetite. However, when the measured magnetic fabrics are non-coaxial, magnetic subfabrics carried by different grain sizes and minerals may be present in the sample. Whereas ipAMS may be carried by several different particle fractions, opAMS in particular may be carried by viscous multiparticle systems with single-domain and superparamagnetic particles. AARM are not carried by viscous multiparticle systems but can be carried by stable single domain particles. Based on the information presented above one can identify the magnetic carriers in the samples of our traverses.  In traverse T1, samples T1-05 and 06, the ipAMS k1, AARM and AIRM R1 axes plot close to each other and the ipAMS k2 and k3, AARM and AIRM R2 and R3 axes plot on the same girdle ( Figure A5). The opAMS k1 axes are offset about 30 degrees from the ipAMS k1 axes. The coaxial ipAMS and AMR fabrics suggest that MD magnetite is the main magnetic carrier in these two samples. In sample T1-07, the ipAMS and AARM principal axes plot close to each other, and conversely the AIRM and opAMS plot close to each other ( Figure  A5). The fabric suggests that the AMS is dominated by MD magnetite and SD magnetite, occurring in inclusions and breakdown rims in amphibole. The opAMS and AIRM likely show the fabric of viscous multiparticle and MD grains in the groundmass of the sample. In sample T1-08 the ipAMS k1 and k2, AARM and AIRM R1 and R2 axes plot on the same girdle and the ipAMS k3, AARM and AIRM R3 axes plot close to each other ( Figure A5). The opAMS principal axes are non-coaxial relative to the other magnetic fabrics in the sample. The general coaxial AMS and AMR show that the samples ipAMS is dominated by multidomain magnetite and likely also the amphibole shape fabric enhanced by SD magnetite in breakdown rims and in inclusions (cf. [121][122][123]). The opAMS fabrics likely represent the groundmass fabric as evidenced by the wide girdle distribution of opAMS tensors. In traverse T1, samples T1-05 and 06, the ipAMS k 1 , AARM and AIRM R 1 axes plot close to each other and the ipAMS k 2 and k 3 , AARM and AIRM R 2 and R 3 axes plot on the same girdle ( Figure A5). The opAMS k 1 axes are offset about 30 degrees from the ipAMS k 1 axes. The coaxial ipAMS and AMR fabrics suggest that MD magnetite is the main magnetic carrier in these two samples. In sample T1-07, the ipAMS and AARM principal axes plot close to each other, and conversely the AIRM and opAMS plot close to each other ( Figure A5). The fabric suggests that the AMS is dominated by MD magnetite and SD magnetite, occurring in inclusions and breakdown rims in amphibole. The opAMS and AIRM likely show the fabric of viscous multiparticle and MD grains in the groundmass of the sample. In sample T1-08 the ipAMS k 1 and k 2 , AARM and AIRM R 1 and R 2 axes plot on the same girdle and the ipAMS k 3 , AARM and AIRM R 3 axes plot close to each other ( Figure A5). The opAMS principal axes are non-coaxial relative to the other magnetic fabrics in the sample. The general coaxial AMS and AMR show that the samples ipAMS is dominated by multidomain magnetite and likely also the amphibole shape fabric enhanced by SD magnetite in breakdown rims and in inclusions (cf. [121][122][123]). The opAMS fabrics likely represent the groundmass fabric as evidenced by the wide girdle distribution of opAMS tensors. Minerals 2021, 11 29 of 37 Figure A5. Comparison between different magnetic fabrics analysed on samples in traverse T1 of the Sosa dyke.
In traverse T2, the ip and opAMS, AIRM and AARM principal axes are coaxial in samples T2-02 and T2-06 and plot close to each other in T2-03 and T2-04 ( Figure A6). This indicates that MD magnetite dominate the magnetic fabrics in these samples. In sample T2-01, opAMS k1 and k2, AARM and AIRM R1 and R2 axes plot on the same girdle and the k3 and R3 occur in a cluster ( Figure A6). The ipAMS k1 plots on the opAMS k1 and k2 girdle, whereas ipAMS k2 plots close to opAMS k3. The ipAMS may therefore reflect the amphibole fabric in the rock enhanced by magnetite inclusions, which produce a strong lineation, while the opAMS and AARM and AIRM reflect the distribution of MD magnetite in the groundmass of the sample. In sample T2-05, ipAMS principal axes are coaxial with AARM principal axes. Both AIRM and opAMS principal axes are oriented oblique to ipAMS and AARM principal axes ( Figure A6). The general coaxial AMS and AARM show that the samples ipAMS is dominated by multidomain magnetite and likely also the amphibole shape fabric enhanced by SD magnetite in breakdown rims and in inclusions. The opAMS and AIRM fabric in these samples may be controlled the fabric of smaller magnetite crystals in the groundmass of the sample. To summarise, the magnetic fabrics in the Sosa dyke traverse are complex. However, the general coaxial AARM and ipAMS fabric show that it is not magnetically inverse and likely represents the MD and silicate fabric in the samples. All magnetic data are given in the Supplementary material "Magnetic data.xlsx" (separate document). In traverse T2, the ip and opAMS, AIRM and AARM principal axes are coaxial in samples T2-02 and T2-06 and plot close to each other in T2-03 and T2-04 ( Figure A6). This indicates that MD magnetite dominate the magnetic fabrics in these samples. In sample T2-01, opAMS k 1 and k 2 , AARM and AIRM R 1 and R 2 axes plot on the same girdle and the k 3 and R 3 occur in a cluster ( Figure A6). The ipAMS k 1 plots on the opAMS k 1 and k 2 girdle, whereas ipAMS k 2 plots close to opAMS k 3 . The ipAMS may therefore reflect the amphibole fabric in the rock enhanced by magnetite inclusions, which produce a strong lineation, while the opAMS and AARM and AIRM reflect the distribution of MD magnetite in the groundmass of the sample. In sample T2-05, ipAMS principal axes are coaxial with AARM principal axes. Both AIRM and opAMS principal axes are oriented oblique to ipAMS and AARM principal axes ( Figure A6). The general coaxial AMS and AARM show that the samples ipAMS is dominated by multidomain magnetite and likely also the amphibole shape fabric enhanced by SD magnetite in breakdown rims and in inclusions. The opAMS and AIRM fabric in these samples may be controlled the fabric of smaller magnetite crystals in the groundmass of the sample. To summarise, the magnetic fabrics in the Sosa dyke traverse are complex. However, the general coaxial AARM and ipAMS fabric show that it is not magnetically inverse and likely represents the MD and silicate fabric in the samples. All magnetic data are given in the Supplementary Material "Magnetic data.xlsx" (separate document). Minerals 2021, 11 30 of 37 Figure A6. Comparison between different magnetic fabrics analysed on samples in traverse T2 of the Sosa dyke.
Appendix C. Data Figure 9 Table A4. Sosa and Great Dyke data for Figure 9.