Feeding and growth of a dyke–laccolith system (Elba Island, Italy) from AMS and mineral fabric data

Dykes feed laccoliths and sills; however, the link between feeder and intrusion is rarely observed. The felsic San Martino laccolith displays a clear feeder–intrusion link, allowing reconstruction of the influence of the size and location of feeder dykes on magma flow during formation of subhorizontal intrusions. This work uses anisotropy of magnetic susceptibility (AMS) combined with mineral shape-preferred orientations of sanidine megacrysts to examine magma flow pathways through feeders into a laccolith. Strong correlation between AMS and K-feldspar datasets indicates that alteration affecting the paramagnetic mineralogy did not influence AMS results. The well-established field relationships between feeder and laccolith provided a robust ‘geo-logical’ model for flow pathways that we have used as a framework to aid interpretation of AMS data. The position and size of the main feeder dyke helped to predict the flow paths in the overlying laccolith. Our results show that magma spread laterally from the feeding system and built the laccolith layers with propagating and inflating divergent flow where tabular particles became aligned perpendicular to the magma displacement direction. The lack of internal discontinuities indicates that the magma was injected as a single pulse or a series of quickly coalescing pulses. Supplementary material: AMS methods, AMS data and detailed fabric maps are available at www.geolsoc.org.uk/SUP18717.

Shallow igneous intrusions record the link between plutonic and volcanic processes. In particular, models of magma supply, accommodation and storage in intrusive bodies contribute to explain the evolution of felsic magma chambers (Miller & Miller 2002;Bachmann et al. 2007;Bachmann & Bergantz 2008). Recent multidisciplinary studies have led to the widely held view that igneous bodies often grow by incremental thickening or inflation of initially thin, sheet-like or tabular bodies by the addition of successive magma pulses (McCaffrey & Petford 1997;Cruden & McCaffrey 2001;de Saint-Blanquat et al. 2001;Rocchi et al. 2002;Menand 2008), overlapping of subhorizontal sheets (Horsman et al. 2005;de Saint-Blanquat et al. 2006;Morgan et al. 2008) or amalgamation of magma fingers and tongue-like lobes (Stevenson et al. 2007). Incremental magma intrusion is also a common interpretation of geophysical observations of deformation episodes at active volcanoes (Biggs et al. 2011).
This study contributes to understanding magma flow during feeding and growth of shallow-level intrusions (<3-4 km deep) by investigating their internal structures. However, rock fabrics can be difficult to disentangle for several reasons. First, the final rock fabric may result from 'pure' magmatic processes (e.g. emplacement flow, convection), late magmatic processes (thermal contraction, gravitational compaction) or tectonic processes (syn-emplacement deformation, post-emplacement deformation), or a multiple overprinting of them all. Second, magmatic fabric reflects finite strain produced by progressive magmatic flow rather than directly recording a simple flow direction (Paterson et al. 1998;de Saint-Blanquat et al. 2006).
We therefore performed a multidisciplinary analysis: relevant data are represented by field observations, structural measurements of mineral foliations and lineations, and a large collection of anisotropy of magnetic susceptibility (AMS) data. Additionally, this study focuses on fabrics in both a laccolith system and its feeder dykes, hence these data constrain the movement of magma in feeders as well as in the main sites of accumulation.
Here we use the late Miocene San Martino felsic laccolith, Elba Island, Italy, as a case study. Its geometry and emplacement-tectonic history are well defined (Dini et al. , 2006Rocchi et al. 2002Rocchi et al. , 2010Westerman et al. 2004), thanks in part to serendipitous tectonic tilting that exposed several transects of the laccolith layers from top to bottom. This igneous body offers the chance to study internal structures that are undoubtedly magmatic as it crystallized very quickly, experienced no detectable subsequent ductile deformation, and local brittle tectonics (sliding and tilting of the laccolithic complex as a single rigid body) did not affect internal structures. Geometric data for the San Martino laccolith led researchers to infer a two-stage growth model, with initial expansion of a thin sill followed by a vertical inflation stage . Our new fabric data allow the reconstruction of internal structures and present a more refined picture of magma emplacement during laccolith growth.
Elba Island, in a stack of thrust complexes ( Fig. 1; Bortolotti et al. 2001). Collision was followed by extensional processes coupled with eastward-propagating emplacement of igneous complexes (Serri et al. 1993(Serri et al. , 2001. Between c. 8.5 and 5.8 Ma, the two uppermost thrust complexes (composed of a Jurassic ophiolite sequence with its cover, and a Cretaceous-Eocene turbidite unit) were intruded by laccoliths, plutons and dykes. In western-central Elba the igneous sequence started with the emplacement of the twolayer Capo Bianco aplite (c. 8.5 Ma) (Maineri et al. 2003;Dini et al. 2007). These layers were intruded and fragmented by the four intrusive layers of the felsic, sanidine-phyric Portoferraio porphyry (7.95 Ma) . Then, intrusion of the felsic, sanidinemegacrystic San Martino porphyry followed (7.4 Ma) . Altogether, these intrusions are defined as multilayer laccoliths  based on the overall parallelism of intrusive contacts and host-rock anisotropies (Fig. 2c), tapering of visible terminations, and upward-convex roofs and flat to upwardconvex floors (Dini et al. 2006). The layers of each intrusive unit are connected by dykes (Fig. 2d), generating an overall geometry typical of a nested Christmas-tree laccolith complex (Corry 1988). Magma was emplaced at a depth between 1.9 and 3.7 km, mostly along planar anisotropies such as thrust surfaces between tectonic complexes, secondary thrusts inside complexes, and bedding within the turbidite sequence.
Emplacement of the laccoliths was followed by intrusion of the c. 2.5 km thick monzogranitic Monte Capanne pluton (7 Ma) Farina et al. 2010) and the Orano mafic dyke swarm (6.95 Ma) (Dini et al. 2008). The laccolith complex, originally intruded in the present western Elba area, was translated eastwards along with country rock by gravitationally driven tectonic collapse along the Central Elba Fault (Fig. 1). The lower section is now exposed in western Elba whereas the upper resides in central Elba (Trevisan 1950;Pertusati et al. 1993;Daniel & Jolivet 1995;Westerman et al. 2004). Following this eastward translation, westside-up normal movement occurred along the Eastern Border Fault with a throw of 2-3 km near the margin of the pluton.

The San Martino laccolith system
The laccolith consists of porphyritic rock, with prominent euhedral sanidine megacrysts set in a very fine-grained groundmass (<100 µm; Fig. 2a and b). Megacrysts are dominantly tabular on (010) and elongated on the c-axis; minor prismatic crystals elongated on the a-axis are also present. Megacryst abundance is 50-200 crystals m −2 with an average size of 5 × 2 × 1 cm (maximum 14 × 6 × 3 cm), corresponding to 3-12 vol%. Phenocrysts also include quartz (1-20 mm), plagioclase (1-5 mm) and biotite (1-5 mm). Groundmass consists of an equigranular, isotropic aggregate of quartz and feldspars, along with accessory apatite, zircon and monazite. Weathering and hydrothermal alteration are widespread, with replacement of plagioclase by calcite + sericite, sanidine by sericite, and biotite by chlorite, plus additional formation of titanite, anatase and/or rutile, and scattered late-magmatic tourmaline spots. The San Martino laccolith is composed of three main, westwarddipping subparallel layers cropping out in central Elba, as well as several dykes in central Elba below the laccolithic layers and in western Elba as the roots of the original laccolith complex left behind after its eastward translation (Fig. 1). The emplacement level for this unit is as shallow as c. 2 km . Contact metamorphic effects in the host rock are practically absent. The filling time of the 21 km 3 intrusion has been estimated at around 100 years, based on the size of the dykes in western Elba and assuming a conservative ascent rate of 3 × 10 −3 m s −1 . Internal magmatic layering and/or contacts are not observed.
Layer 1 is topmost and most voluminous, reaching a thickness of c. 700 m (tapering toward both the northern and southern ends) with a north-south length of 8.3 km (Fig. 3). It is characterized by branching patterns toward its margins. A prominent southern branch on the west shore of Marina di Campo Bay exposes a c. 250 m thick bottom-top section that trends north-south and dips 30° west. Its base is marked at the south end by a gently west-dipping, c. 300 m thick cross-section at Punta Mele; at the north end the base is well exposed at La Biodola Bay as a c. 500 m thick bottom-top section. Layer 2 represents less than 5% of the total laccolith volume, striking nW-SE for about 1 km in the northern half of the complex. Its thickness decreases from c. 150 to c. 100 m from south to north. The lowermost Layer 3 parallels Layer 2 with an exposure length of 2 km and a thickness of c. 250 m.
Six steeply dipping dykes of San Martino porphyry are mapped in western Elba (Fig. 1). The largest, the WnW-ESE Marciana dyke, which is 1500 m long and 10-20 m thick, is interpreted as the main feeder . In central Elba, the subvertical nE-SW-oriented Sansone dyke is the most significant, and is exposed over 400-500 m with widths of 3-20 m. Its structural location below the laccolith (Figs 1 and 2d) suggests that it locally fed Layer 3.
With the aim of describing and interpreting the fabric data in their original geometric context (i.e. before eastward translation and rotation ( Fig. 1; Westerman et al. 2004)), all central Elba data have been rotated west-side-up by 30° around a north-south horizontal axis. Data discussed in the text, therefore, refer to a body having subhorizontal basal contacts and slightly upward-convex tops, with steeply dipping dykes below.

Fabric results
Two approaches were taken to establish the igneous fabric within the San Martino laccolith complex: direct determination by measuring the orientation of sanidine megacrysts (Fig. 3a), and indirect determination by measuring AMS parameters (Fig. 3b). The use of multiple independent fabric determinations is of fundamental importance in validating fabric analysis.

Megacryst fabric
Shape-preferred orientations of sanidine megacrysts have been measured using (010) faces of tabular crystals, whereas magmatic lineations were determined using the c-axes of elongated-tabular crystals or the a-axes of prism-like crystals elongated on a. Tabular crystals with weak elongation yielded only foliation data.
These crystallographic features are best recognized where crystals show 3D exposures ( Fig. 2a and b), such as along weathered shoreline cliffs. Statistically controlled foliation values were  (Dini et al. 2006). Dip values, given as black numbers, refer to detailed analyses; grey numbers indicate field estimates. (b) Restored magnetic data showing magnetic foliation in red and magnetic lineation in blue. numbers beside symbols are dip values. All measured values have been processed using Stereonet v.6.3.3 of R. W. Allmendiger (http://www.geo.cornell. edu/geology/faculty/RWA/programs). determined at 48 stations and lineation at 36 stations, with both values measured at 34 of those sites. Foliation measurements were made on 25-99 contiguous crystals at each station, whereas lineation measurements were derived from 30-97 crystals. Foliation measurements based on average crystal patterns were made at an additional 19 sites. Throughout the laccolith system, the investigation of megacryst attitudes indicates well-defined magmatic foliations along with weak magmatic lineations. In fact, foliation poles in 85% of the stations (41/48) have the main eigenvalue (E 1 ) > 0.6 and almost 60% (28/48) have E 1 > 0.7. In contrast, magmatic lineations in 50% of the stations have the main eigenvalue >0.6 and in only 13% (5/36) have E 1 > 0.7.
Taken together, magmatic foliations within Layer 1 (Fig. 3a) show distinctive patterns that change progressively, emanating from the west-central part of the layer where a distinct north-south-striking foliation has been measured. This north-south attitude continues to the east, but foliation attitudes rotate clockwise toward the south and anticlockwise toward the north. In the southern part of this layer, foliations rotate progressively to a nE-SW attitude, then to eastwest, and finally to nW-SE at the southwesternmost exposures. A detailed study at the southern edge of Layer 1 ('Casa Ischia') shows changes in orientation from nW-SE in the lower portions, to northsouth in the central part, to nE-SW in the upper portion, all with variable dips. In the northern part, the rotation shows a mirrored pattern, progressing through a widespread nW-SE orientation, to east-west attitudes along the northernmost margin of the layer. A second detailed study of sanidine megacryst fabric in the north ('Lamaia sheet') reveals homogeneous fabric from the bottom to the top. Dips of foliation throughout Layer 1 are highly variable with no clear spatial patterns.
Two lower laccolith sheets are exposed in the northern part of the system. Layer 2 has a dominant fabric with foliation and lineation trending predominantly nE-SW, whereas fabric in Layer 3 is generally nW-SE for much of the unit but changes anticlockwise through east-west toward the nW terminus of the laccolith. Lineations are distributed at various attitudes, within the plane of foliation and, more commonly than not, either close to strike or running down dip. Where observed, sanidine foliation in both the Sansone and Marciana dykes is subparallel to the steeply inclined dyke walls.

Magnetic parameters
The second approach to determining the internal fabric of the San Martino laccolith was to determine AMS parameters, which are controlled by the orientation of crystals of the mineral(s) dominating the magnetic signal. AMS is a technique that gives a quantitative description of the crystalline fabric of a rock by determining the variation of magnetic susceptibility with direction such that the eigenvector K 1 represents the magnetic lineation whereas K 3 is the pole of the magnetic foliation (Tarling & Hrouda 1993;O'Driscoll et al. 2008).
We sampled 150 sites in the laccolith horizons and their feeder dykes. Clusters of samples were collected at selected locations to investigate the distribution of magnetic parameters at a very local scale (e.g. thin branches of the layers, outer and inner parts of dykes). The relationship between the mineral preferred orientation and magnetic fabric depends on the nature of the magnetic mineralogy, here represented by biotite, with rare tourmaline and very minor Ti-rich oxides. Biotite is commonly chloritized, but no significant formation of Fe-oxides occurred. This conclusion is supported by: (1) low measured susceptibilities (1.9 × 10 −4 to 2.2 × 10 −5 -SI), typical of rocks characterized by paramagnetic mineralogy (Tarling & Hrouda 1993); (2) K m for altered-weathered samples similar to K m of the freshest rocks; (3) heating-cooling experiments on fresh, chloritized, and chloritized-weathered samples all showing an overall paramagnetic behaviour (Fig. 4).

Magnetic fabric in dykes
Data for all the dykes indicate that both the shape parameter T and the anisotropy degree P j are variable (Fig. 5). Site mean values of T vary from -0.554 to 0.918 whereas P j values are general low (1.009-1.129). The map of the Marciana dyke in western Elba (Fig.  6) suggests an overall parallelism between magnetic fabric and dyke walls. The perpendicularity of the best-fit great circles of lineations and poles to foliation confirms that K 1 (lineation) lies within the plane of foliation. Detailed fabric analysis shows local 'normal' fabric (Rochette et al. 1992) with K 3 suborthogonal and K 1 subparallel to the outer walls; fabric close to the walls shows upward-SE imbrication in both horizontal and vertical sections. Lineations near dyke walls are generally oblique with imbrication plunging toward the walls; interior lineations are subhorizontal. Elsewhere, strong fabric asymmetry occurs with no reversal of imbrication across the dyke. Magnetic foliation and lineation in the Sansone dyke in central Elba (Fig. 6) commonly parallel the overall n55E strike of the dyke, with dips less steep than dyke walls.

Magnetic fabric in the main laccolith body
In the main laccolith body, the shape parameter T ranges from -0.836 to 0.891 (Fig. 5). Data from the middle and lower parts of the laccolith reveal an oblate range of ellipsoid shapes, whereas shapes for the upper part are predominantly oblate, illustrating that fabrics are dominated by foliation (flattening) rather than lineation (constriction). The T parameter is highly variable at both the laccolith and local scale. Values of the degree of anisotropy (P j ) in the laccolith are fairly low, ranging from 1.006 to 1.081 (Fig. 5) as is typical in granitic rocks (Horsman et al. 2005). The use of AMS allowed recognition of a well-defined magnetic fabric that is almost everywhere strong: 86% of the samples (129/150) have e 3 < 25° and 75% of the samples (113/150) have e 1 < 25°, where e 1 and e 3 are the semi-angles (measured in degrees) of the confidence ellipses around the mean value of K 1 (magnetic lineation) and K 3 (pole of magnetic foliation). Only 6% of the samples (9/150) have both e 1 and e 3 >25°.
Magnetic fabrics (Fig. 3b) are very similar to those revealed by sanidine megacryst analyses (Fig. 3a), with clockwise rotation of AMS fabric in Layer 1-south, mirrored by an anticlockwise rotation in Layer 1-north. Detailed study at Casa Ischia on the southern coast (cross-section at the southern termination of the main body) shows the same progressive bottom to top changes as described above for magmatic fabric. The lowermost 100 m has magnetic foliations striking nnW-SSE with dips 35-70° nE, whereas foliations in the uppermost 150 m strike nE-SW and dip variably. Results of a similar detailed study in the Lamaia sheet on the north shore show east-west strikes of magnetic foliation like their sanidine counterparts, but dips increase progressively from <30° at the base, to 30-60° in the core, to subvertical near the upper contact. Data for Layer 2 show consistently nE-SW-striking foliation, with gentle SE dips at the southern termination and steep dips further north. In the lowermost Layer 3 the foliation has nW-SE mean strike with variable dip.

Correlation of AMS and megacryst fabric data
AMS fabric data and shape-preferred orientations of sanidine megacrysts, along with structural reconstructions, allow development of an internally consistent model of magma flow and laccolith growth. Before presenting the model, some concerns will be addressed. First, some recent work suggests caution in interpreting flow structures in intrusive rocks, owing to possible subsolidus development of phenocryst-bearing texture in cases of thermal cycling (Mills et al. 2011). However, this does not apply to the San Martino laccolith, which underwent unidirectional rapid cooling as supported by the sanidine structure of its K-feldspar megacrysts.
Second, many previous studies have illustrated how AMS can be used successfully to determine magmatic fabric patterns by direct correlation between fabrics from mineral shape-preferred measurements and AMS fabrics (Guillet et al. 1983;Bouillin et al. 1993;Darrozes et al. 1994;de Saint-Blanquat et al. 2001Horsman et al. 2005). nevertheless, we tested the correlation of igneous foliation preserved by sanidine megacryst attitudes and that of biotite as revealed by AMS for the San Martino laccolith. Results from 25 sites where both AMS and detailed megacryst fabric data were collected reveal general concordance (Table 1): the angle between magnetic foliation and the megacryst foliation is <30° in 16/25 FEEDInG AnD GROWTH OF A DyKE-LACCOLITH SySTEM 419 (65%) stations and the angle between magnetic and megacryst lineation is <30° in 6/7 (85%) stations. Given that highly oblate biotite crystals generate the AMS fabric, the observed parallelism of tabular sanidine and AMS fabrics indicates that both crystal sets recorded similar strains (magmatic flow).
Correlations between lineations derived from the two methods are not so easily explained, as the AMS lineation comes from biotite crystals, which are not elongated. We conclude that the K 1 lineation is most probably due to the platy biotite crystals being preferentially oriented along a 'zone axis' within the plane of mineral foliation (Bouchez 1997). This requires that the highly oblate biotite crystals wobble within the plane of foliation, and also that this line be coincident with the line along which elongated sanidine crystals trend, most probably corresponding to the axis of maximum stretching during magma flow.

Relation of magmatic fabric to magma flow
Having established that AMS fabric in the San Martino system mimics megacryst attitudes, the next assessment concerns how such petrofabric data can preserve evidence of magma flow paths. This is problematic because (1) fabric can result from multiple events (flow, tectonic deformation, hydrothermal activity, etc.), and (2) fabric reflects finite strain generated by differential stress owing to progressive magmatic flow (Paterson et al. 1998).
Based on the absence of appreciable signs of solid-state deformation, an overprint of igneous AMS fabrics by regional deformation can be ruled out, even though the post-emplacement history included tectonic translation from western to central Elba. Additionally, regularly varying patterns of fabric within the intrusions show that stresses were local and, therefore, not a record of regional palaeostress. Other processes able to impart a fabric, such as filter pressing or porous flow, are also unrealistic in this case owing to the low percentage and homogeneous distribution of phenocrysts in the magma during emplacement flow. Additionally, this laccolith was emplaced rapidly  and was quickly solidified (very fine-grained matrix). We can thus infer that the observed fabrics reflect the final increments of strain as the magma was moving and solidifying.
The lack of magmatic layering, internal magmatic contacts and/ or internal shear zones in the laccolith suggests that the magma was injected to form the different layers as either a single pulse or as batches coalescing shortly after or during injection. In the latter For foliation, the angle between the two datasets is <30° at 71% of stations; for lineation, the angle between the two datasets is <30° at 83% of stations. hypothesis, the time gap between pulses had to be shorter than the solidification time of the preceding pulse. Although this thermal requirement is more feasible for slowly cooling, deep-seated igneous bodies , there are examples of such processes in shallower, but more mafic, igneous bodies (depth c. Saint-Blanquat et al. 2006). These constraints suggest that our fabric markers formed during the waning stages of a single episode of flow: the fabric represents only the strain occurring in the final stages of emplacement, making it difficult to test the two-stage (horizontal spreading then vertical inflation) model for this system . Given these conditions, it is fundamental to define which fabrics can be generated by the different types of magmatic flow (Paterson et al. 1998). If magma flows in any way other than with a uniform velocity field, such that crystals are not forced to rotate to a preferred orientation, then a stress field will be produced that will orient tabular and linear crystals. Tabular crystals become oriented perpendicular to the direction of maximum shortening and linear crystals become aligned parallel to the direction of maximum stretching. Three end-members of non-uniform flow may be considered here: (1) convergent flow, occurring when magma moves to a progressively narrowing region with associated velocity increase; crystals align their longest axes and largest crystal faces with the particle path; (2) divergent flow, occurring when magma spreads in progressively widening regions with divergence of flow lines and velocity decrease; planar fabrics develop in the plane of flattening perpendicular to particle paths, and lineation develops in the flattening plane, parallel to the stretching direction; (3) noncoaxial flow, generated by drag along a boundary surface affected by simple shear, with velocity increasing away from a boundary surface; fabric forms at a variable angle to that surface.

km) such as the Black Mesa intrusion (de
All of these may be present over short distances to define units of flow where non-coaxial flow is combined with either convergent or divergent flow, as in flow lobes (Stevenson et al. 2007). nevertheless, laccolith emplacement is characterized by lateral spreading and filling, with the cross-sectional area of the feeding dyke in western Elba with strikes of K 1 -K 2 planes (magnetic foliation) in red and trends of K 1 (magnetic lineation) in blue, measured at 19 sites including a complete transverse section. Owing to the variable strike of the dyke and to simplify reading of data, stereographic projections of foliations and lineations are plotted in relation to strike of the dyke rotated to east-west orientation for every site. The Marciana dyke remained in western Elba below the décollement surface of the Central Elba Fault, therefore these data have not been rotated for any tectonic restoration. (b) Map of Sansone dyke with symbols as above, measured at 10 sites, with three complete transverse sections. Stereographic projections of magmatic foliations and lineations have been restored by 30° clockwise rotation around a horizontal north-south axis. system remaining constant (i.e. a feeder dyke) while the crosssectional area of the laccolith grows. Such conditions would make divergent flow the norm within filling laccoliths. Transitions in fabric-flow relationships have been documented in experiments (Kratinová et al. 2006) where magnetic fabric inside the feeder (i.e. constricted) was parallel to the transport direction of the magma, but further away (i.e. diverging), the fabric rotated by 90° to become perpendicular to the transport direction. On the other hand, in thin dykes where all the magma is relatively close to the walls, non-coaxial flow generates an imbricated foliation along dyke walls. Similarly, in subhorizontal sheets, where the centre of each igneous sheet flowed more rapidly than the edges (Correa-Gomes et al. 2001;Gil-Imaz et al. 2006), symmetrical imbrication of foliation planes develops at the upper and lower contacts (Komar 1972(Komar , 1976, as shown also by analogue modelling (Kratinová et al. 2006).
This discussion on fabric-flow relationships relates to the issue that published papers commonly present a seamless transition between maps showing shape-preferred orientation and/or AMS fabrics, and maps or diagrams presenting the magma flow history as deduced from the fabrics. However, a variety of fabric-flow relationships are used in interpreting these fabric data, according to the different types of inferred magma flow.

Magma flow in feeder dykes
Magma flow in dykes has traditionally been inferred with the assumption that K 1 is oriented parallel to the direction of magma flow (Rochette et al. 1991), with the sense of flow determined using the symmetrical imbrication of K 1 (Knight & Walker 1988). However, magnetic lineation can be perpendicular to magma flow (Rochette et al. 1991(Rochette et al. , 1999Dragoni et al. 1997) and the intersection of magnetic foliations can result in an apparent magnetic lineation (Callot & Guichet 2003). For these reasons some researchers (Geoffroy et al. 2002) established that imbrication of magnetic foliation better constrains magma flow direction than does simple magnetic lineation.
Flow histories for the Marciana and Sansone dykes have been interpreted from both AMS and sanidine petrofabric data, using the theoretical and empirical bases noted above. Investigation of internal structures has mainly focused on interpretation of the attitudes of foliation owing to the planar and oblate nature of the magnetic carrier (biotite), along with the tabular shape of most sanidine megacrysts (Fig. 6). The Marciana dyke in western Elba lies beneath the former location of the San Martino laccolith. The subvertical fabric, general parallelism, and imbricated orientation of AMS foliation and lineation near the dyke walls, combined with moderately inclined foliations (and lineations) in the dyke core with respect to the walls, suggest that the dominant magma flow was subvertical. Fabrics in the Sansone dyke at the base of the laccolith in central Elba are also suggestive of vertical magma flow, in that foliation is typically subparallel and imbricated with respect to dyke walls, and lineation is generally steeper near the dyke walls than in the core.

Magma flow in laccolith horizons
When an intrusion grows in two stages, such as a sill inflating to a laccolith as at Papoose Flat, the relationships between fabric and magma flow also evolve through time (de Saint-Blanquat et al. 2001). A foliation parallel to the sill shape develops during the sill formation stage. During inflation and transition to laccolith shape, when flow is mainly vertical, foliation develops perpendicular to flow, retaining the pattern with foliation parallel to the upper contact in the core of the body away from the solid-state fabric. Lineations close to the contact (<1 m) are parallel to flow owing to wall-rock interaction and shear, whereas below that, lineations are parallel to the stretching direction perpendicular to flow. The case of magma flowing in lobes is illustrated by the Trawenagh Bay granite, where AMS fabric data define frozen lobes of granitic magma (Stevenson et al. 2007). Foliations are aligned parallel to the lateral margins of the lobes and wrap concentrically around lobe noses, whereas lineations trend parallel to the elongation of lobes. In thin sills with fingers, as in the Henry Mountains, magnetic foliation trends subparallel to contacts with lineations presenting a radiating pattern in the fingers off the main body of the sill (Horsman et al. 2005). In a nearby small, flat pluton, strong parallelism of concentrically arranged foliations from both AMS and field fabric data has been reported (de Saint-Blanquat et al. 2006).
These interpretations of flow history seem entirely plausible, yet the rules for getting from fabric to flow are not generally presented. The rules used appear to vary considerably, largely because the assumptions used to interpret flow patterns are not always clearly stated. Our approach to interpreting the San Martino flow history has been to start with the structural data (shape, geometry, location of feeders, etc.) and postulate a reasonable emplacement model to be tested using multiple fabric datasets and basic fabric-flow principles. We have thus far established that tabular and elongated sanidine megacrysts and biotite carrying the paramagnetic AMS signal have similar shape-preferred orientation that varies widely but in organized patterns. After considering available explanations for this coherence of fabric, we have concluded that crystal alignment recorded the strain produced by the stress field acting during the waning stages of magma flow.
The magnetic fabric is mainly oblate throughout the laccolith, probably linked to the dominance of flattening processes during the intrusion growth. Flattening in thick sheets is usually associated with divergent flow where particles align their largest faces orthogonal to flow directions. For this reason it has been assumed that the magma displacement direction (magma flow) in the laccolith layers was orthogonal to the foliation. On the basis of these considerations, AMS and megacryst fabric data can be used together to depict models for magma flow through feeder dykes and into the laccolith layers.
The Marciana dyke in western Elba is assumed to have been the primary feeder for the San Martino Christmas-tree laccolith system above, with smaller dykes serving as connectors between laccolith horizons. A schematic inset in Figure 7 (upper left) illustrates the diverging particle paths in magma spreading horizontally from a dyke with length less than the laccolith diameter. We assumed that flow within the sheets was away from the centrally located east-west feeding system. Given that the cross-sectional area of the feeder dyke was of the order of 0.2 km 2 , whereas the horizontal area of the laccolith sheets reached 55 km 2 with multiple layers up to 700 m thick , divergent flow is assumed to have been the norm during laccolith growth. Figure 7 presents a map of interpreted flow directions assuming that flow was in the direction perpendicular to magmatic foliation (i.e. the plane of flattening), and parallel to the pole to such foliations. It should be noted that foliation values on this map are corrected for subsequent tectonic rotation, whereas the map itself presents the current distribution of the laccolith sheets that dip 30° westward on average. nevertheless, logical patterns develop when one assumes that poles to foliation preserve particle paths, and therefore that foliation dips in the direction of upwardly inclined flow but dips away during downwardly inclined flow.
To interpret this map and the resulting emplacement model, we start along the western edge near the roof of the uppermost sheet of FEEDInG AnD GROWTH OF A DyKE-LACCOLITH SySTEM 423 the system. Flow arrows plunge shallowly toward the east as a result of upward flow being directed eastward near the roof of the laccolith. Further east (and lower in the section), arrows diverge to show both northward and southward movement of the magma near the eastern termini of the sheet. Southern central Elba is dominated by southeastward flow, locally inclined upward, but predominantly plunging in the direction of flow. Flow paths have been confirmed in this area where strained quartz phenocrysts in the outer 1 cm skin of the sheet are aligned nW-SE, parallel to the magmatic fabric measured several metres below.
Further to the south, flow rotates to predominantly due south, with local divergence. Detailed study near Casa Ischia shows southward flow in the lowermost part of the sheet, with flow of the upper (western) part to the ESE. This sense is confirmed by strained quartz phenocrysts in the upper contact exhibiting differential flow. Magma below the skin flowed ESE to produce bookshelf structures of quartz, strained with aspect ratios up to 40. Further rotation of foliation at the southwesternmost exposures of San Martino porphyry indicates flow toward the SW. north of the central feeder system, flow patterns show particle paths reflecting northward movement of magma, with local divergence above and below a large septum of host flysch, and predominant flow toward the nE along much of the base of the uppermost San Martino sheet. Two smaller underlying sheets show general filling by nE-directed flow with divergence. northernmost exposures, much like their mirror counterpart to the south, show the maximum rotation of flow to the nW. Figure 7 schematically presents the relationships between fabric, flow and position in the reconstructed laccolith system. Magma flowed subvertically within a central feeder dyke, as indicated by symmetrical imbrications of the subvertical AMS markers in sections orthogonal to the dyke plane. The dyke fed a laccolithic main body by lateral spreading of the magma, during which the oblate sanidine and biotite crystals became parallel to the plane of flattening that developed perpendicular to the magma displacement direction before the melt solidified to form the porphyry matrix. Reconstruction of reasonable patterns of filling for the laccolith horizons and the 3D patterns of flow within them was based on (1) correspondence of sanidine megacryst fabric and the biotite AMS fabric, and (2) a model generated from detailed maps and reconstructions of the geology.

Implications
The possibility of interpreting all the megacryst and AMS fabric data in a unique frame of flow suggests that each laccolith layer grew in a single inflation episode. This inference implies that spreading and inflation were simultaneous as suggested on a theoretical basis (Michaut 2011), supporting laccolith emplacement as modelled by the elastic plate theory (Bunger & Cruden 2011;Michaut 2011). Alternatively, traces of the initial horizontally expanding sill are lost or yet to be documented; for example, by collecting data along contacts where quenched magma shows deformation features compatible with extreme stretching.
Additionally, a fabric compatible with a single inflation episode could imply that either laccolith filling was by a single magma pulse or, alternatively, filling was by means of multiple, yet quickly coalescing pulses. Both possibilities agree with the short time scale inferred for laccolith formation based on the size of the feeding system , as well as on a theoretical basis (Michaut 2011). On the other hand, significantly longer times are suggested by calculating minimum filling rates for magma chambers (Annen 2011) or by dividing a laccolith volume by isotopically determined, highly resolved emplacement times (Michel et al. 2008). However, these timings have to be considered as averages, probably composed

Conclusions
The strong correlation between megacryst and magnetic fabrics strengthens the use of AMS as a magma strain indicator. Furthermore, whereas megacrysts commonly give poor lineation data, AMS provides the magmatic lineation as a 'zone axis '. Fabric (strain) in the rock and magma flow are closely related, owing to fast emplacement and cooling, as well as to the lack of post-emplacement tectonic deformation. The magma feeding the laccolith layers flowed subvertically from a sizeable central dyke. Magma then spread laterally as a single pulse or a series of pulses that quickly coalesced.