The growth of faults and fracture networks in a mechanically evolving, mechanically stratified rock mass: a case study from Spireslack Surface Coal Mine, Scotland

Fault architecture and fracture network evolution (and resulting bulk hydraulic properties) are highly dependent on the mechanical properties of the rocks at the time the structures developed. This paper investigates the role of mechanical layering and pre-existing structures on the evolution of strike–slip faults and fracture networks. Detailed mapping of exceptionally well exposed fluvial–deltaic lithologies at Spireslack Surface Coal Mine, Scotland, reveals two phases of faulting with an initial sinistral and later dextral sense of shear with ongoing pre-faulting, syn-faulting, and postfaulting joint sets. We find fault zone internal structure depends on whether the fault is self-juxtaposing or cuts multiple lithologies, the presence of shale layers that promote bedrotation and fault-core lens formation, and the orientation of joints and coal cleats at the time of faulting. During ongoing deformation, cementation of fractures is concentrated where the fracture network is most connected. This leads to the counter-intuitive result that the highest-fracture-density part of the network often has the lowest open fracture connectivity. To evaluate the final bulk hydraulic properties of a deformed rock mass, it is crucial to appreciate the relative timing of deformation events, concurrent or subsequent cementation, and the interlinked effects on overall network connectivity.


Introduction 23
Differences in the mechanical properties (mechanical stratigraphy) of rock layers have long been 24 recognised as influencing the style and evolution of faults (Anderson, 1951;Donath, 1961; Ranalli and 25 Yin, 1990;Ferrill et al., 2017)). However, work has tended to focus particularly on normal faults, with 26 the effect of mechanical layers in sand-shale sequences (e.g. van  Pre-existing weaknesses (e.g., joints and faults) also play an important role in the nucleation, 39 orientation, and length of later faults (Crider and Peacock, 2004;Peacock, 2001;Walsh et al., 2002). 40 The mechanical response of a pre-existing joint to faulting will depend on the orientation of the feature 41 relative to far field stress , the ratio of principal stresses (Lunn et  Photograph of the 6' Seat Earth exposed in the dip-slope to the west of the site. 102 c. 2D fracture intensity: We compared the intensity of the networks and sets within the 179 network using 2D fracture intensity (Equation 2) (P21; Dershowitz & Einstein (1988); 180 Rohrbaugh et al. (2002)). during burial diagenesis (Laubach et al., 1998). Cleat spacing (typically <2 cm) is dependent on 217 bed-thickness, coal quality and the presence of clastic material (e.g. shale partings). 218 -Mineralised shear fractures: Typically 2 to 15 cm long, but increase to greater than 1 m long as 219 stratigraphic separation increases. Fractures less than 15 cm long abut against E-W trending 220 cleats, with trace length restricted by cleat spacing. Longer fractures cut through the cleats. The 221 thickness of planar ankerite veins increases with the length of the vein. 222 -En-echelon arrays: En-echelon ankerite veins display both sinistral and dextral motion ( Figure  223 3d). Dextral arrays can occur both simultaneously with, or later than, sinistral arrays. 224 -Barren shear fractures: In addition to the cleat network, fractures that abut against all other 225 fractures are often curved and have trace lengths typically between 5 to 15 cm. These may 226 propagate from the tip of pre-existing mineralised shear fractures (Figure 3d). 227 Other lithologies observed in Spireslack SCM display a strongly developed fracture stratigraphy (c.f. 228 Laubach et al. (2009)). For example, the McDonald Seat Earth exposed in the dip-slope towards the 229 west of the site (Figure 4a), lacks a well-developed joint pattern. Instead, shear-fractures are observed in 230 relation to small stratigraphic offset, strike-slip faults (Figure 5a,b). Fractures are only found in close 231 proximity to fault strands and are either sub-parallel to fault strands in the hanging wall block, or 232 oblique to the fault strands in relay zones and fault tips. Fractures commonly display small sinistral and 233 dextral stratigraphic offsets (mm to cm) and are typically barren, although occasionally pyrite is found 234 along the fracture plane. Sandstones display bed-bound joint-sets in a similar manner to the McDonald 235 Limestone. However, there was limited bed-parallel exposure to explore the age and orientation of 236 fracture sets in sandstone lithologies. In contrast to the dip-slope, seat-earth in the high wall displays a 237 well-developed bed-bound fracture network. This suggests that mine-related stresses may have caused 238 deformation of these lithologies and that the natural network has been altered by both subsurface and 239 surface mining activities.   Segment linkage, folding, and increased fracturing between strands led to the development of a highly asymmetric damage zone ( Figure  5a, b, e). Faults typically barren, only displaying yellow alteration and occasionally pyrite. McDonald Limestone Self-juxtaposed faults, associated relay zones, and nearby N-S trending joint sets, are mineralised (calcite), display high displacement to length ratios (2.4 to 2.8), and show extensive folding of the surrounding lithologies ( Figure 5f). Strands often abut against favourably orientated pre-existing joints.

Coal
Fault strands are characterised by a fault core comprising of a 5 to 20 cm thick zone of ankerite, with occasional calcite mineralisation, brecciated coal and pyrite ( Figure 5c). The fault core is discontinuous along strike, with displacement transferring to other strands after 1 to 5 meters ( Figure 5c). The gentle folding of the bed between strands is taken up by a symmetric zone of damage consisting of increased fracturing, en-echelon veining and mineralised shear fractures. The structures represent a continuation of the processes discussed in Section 4.1.1.

Shale
Fault strands are rarely observed. High angle thrusts (40° to 60°) dominate, with bed parallel folding picked out by ironstone concretions (Figure 5d), which themselves can display internal deformation (tension gashes). Near self-juxtaposed faults a cleavage is developed sub-parallel to the fault plane, which combined with slickenfibers on competent bedding planes suggests bed-parallel slip.

Fracture relationships at low fault intensity: 376
The interpretation area in Figure 8a is dominated by large trace-length, NE trending, joints and smaller 377 trace length NNW trending joints. Abutting relationships suggest these formed as four distinct phases, 378 with two phases occurring at each orientation. The fault network displays two orientation sets (N and 379 NNW) of sinistral faults with low connectivity, trace length, and intensity (Table 3) (Table 3). Abutting relationships of faults in this panel suggests that the majority of strands represent 393 reactivated Phase 2 (orange) or Phase 4 (purple) joints. The majority of faulted Phase 2 joints display 394 sinistral offset or evidence of reactivation, while Phase 4 joints display predominantly dextral offsets. 395 Abutting relationships suggest that faulting occurred as two phases, with joint development occurring 396 both between (Phase 5 and 6) and following (Phase 7 and 8) the formation of dextral faults. 397

Fracture relationships where both phases of faulting is present 398
The interpretation area in Figure 8c is located close to the major NW-trending dextral fault (Figure 4), 399 and includes two self-juxtaposed faults towards the bottom and top of the studied section (Figure 8c). 400 The panel displays a complex fracture evolution, however, many of the features observed in the 401 previous panels are visible. Phase 1 to 4 joints are still easily identified; however, their trace length has 402 further decreased due to increased fault intensity (I = 1.9 f/m). Unlike figures 8a and 8b, the fault 403 network is well connected in this panel (Pc = 0.71), with individual fault strands linking to form locally 404 complex relay zones (e.g. the bottom left of Figure 8c). Abundant sinistral, dextral, and reactivated fault 405 strands are observed, with Phase 2 and 4 joints regularly becoming reactivated and linked by new fault 406 strands. Locally, Phase 6 joints are also reactivated in a dextral sense (e.g., the relay zone in the NE of 407 Figure 8c). The number of joints that abut against faults and the pre-existing joint sets (Phase 1 to 4) is 408 greatly increased in this panel, with several Phase 7 and 8 joints identified. 409

Summary of structures 410
As fault intensity increases, the complexity of age relationships in the fault-fracture network also 411 increases (Figure 8 reasonably consistent across this section of the limestone pavement, as fault-meshes begin to form, age 425 relationships become increasingly complex and spatially variable (Figure 6a). This suggests a highly 426 heterogeneous stress field, which was rotated relative to locally active fault strands. An increase in fault 427 throw also affects the intensity, trace-length and connectivity of the network. 428  The fault and fracture network is highly variable in SA3, with complex relationships between preexisting joints, faulted joints, faults, and fracture corridors.   (Fig. 5), prior to being slightly rotated prior to the formation of Phase 3 & 4 joints. Joint Phases 1-4 represent the pre-faulting fracture state (Fig. 5a) Bedding became folded towards the SE, with the early influence of sinistral wrench tectonics (stage 2) possibly causing some NS orientated folds to develop (Fig. 3c).
Bed-parallel shear in shale (Fig 4d) was associated with regional folding and probably continued into Stage 2.

Because of multiple preexisting joint sets and a welldeveloped mechanical stratigraphy, trace length of individual fault strands is low and strain is taken up by several small faults.
Joint sets 1 and 3 restricted the growth of Phase 1 faults and favorably orientated Phase 2 joints were reactivated. Calcite mineralization commonly observed (Figs 4, 6) with evidence of multiple crack seal events (Fig. 2c). Joint sets 5 and 6 formed between Phase 1 and Phase 2 faults (Fig. 5b, c).
Fault-zones became well connected through the linkage of through-going faults (Fig 5c). Where Phase 1 and 2 faults interact, complex fault meshes developed (Fig 3a, 6a).
Reactivation of cleats and stage 2 features was accompanied by a further phase of ankerite mineralization, and local kink-band development (Fig 2 df).

Stage 4: Paleogene intrusions.
Basaltic dykes were intruded, with the orientation suggesting it is associated with the British Tertiary Igneous Provence 17 .
No fragments of dyke are observed within the fault core in Fig 7a. This suggests the dyke broke through the fault core out of the plane of observation, as the dyke post-dates faulting.
Although not visible in the Muirkirk 6' coal, in the high wall the Muirkirk 9' coal becomes altered to white trap, a common trend in the Western Ayrshire Coalfield 18 . No evidence of extensional reactivation is observed other than along the edge of the dyke.
Brecciation of the edge of the major dyke and surrounding limestones was coupled with dip-slip reactivation, suggesting post-intrusion extensional reactivation occurred.
No evidence of late stage reactivation observed.  increase in Phase 1 joints in comparison to smaller trace length Phase 3 or 5 joints (Reed et 478 al., 2008). 479 The fact that some joints show evidence of preferential reactivation and subsequent cementation, while 480 others remain barren, suggests that certain joint sets indicate the past-connectivity of mineral rich fluids 481 through the network, which at Spireslack SCM was dominated by faults ( Figure 8). Barren joints 482 typically post-date mineralisation (Peacock, 2001;, however, Phase 1 483 and 3 joints at Spireslack are often offset by faults or reactivated as faulted joints (Figure 8). This 484 suggests joints were present at the time of faulting, however, not all joints sets were hydraulically 485 connected to the mineralising fluids. This could be either due to: fluid flow being dominated by vertical 486 flow associated with fault-valve behaviour during slip events (Sibson, 1990(Sibson, , 1992 5, 6) may act as a baffle or barrier to flow (e.g. 496 (Skurtveit et al., 2015). It is therefore more appropriate to consider the 'joint network' when assessing 497 the modern-day network connectivity at the site. While joint intensity increases as fault-intensity 498 increases (Table 3) this is not the case for connectivity. Where faulting intensity was low, joints are well 499 connected (SA1, Pc = 0.96). However, as fault intensity increased and the number of faulted joints 500 increases, connectivity drops to Pc = 0.90. For example, in SA3, where fault intensity is 1.9 f/m, the 501 connectivity of the joint network drops to Pc = 0.77. Additionally, connectivity depends on the 502 orientation of the fractures, with NW trending features being the most connected (Table 3, Figure 9). 503 The modern day stress orientation in Scotland (roughly northerly trending maximum compressive stress 504 (Baptie, 2010;Heidbach et al., 2008)) would act to reduce the aperture of these large trace length joint 505 sets and further reduce the permeability of the network. This leads to the counter-intuitive observation 506 that although joint intensity increases in areas associated with faulting (Table 3) The internal structure of faults at Spireslack SCM is greatly affected by the level of lithological 517 juxtaposition, with different properties observed for self-juxtaposed faults (Section 4.2.1) and those that 518 cut multiple lithologies (Section 4.2.2). Self-juxtaposed fault-strands cutting lithologies without pre-519 existing joints are typically relatively planar, develop relay zones, and only display local iron staining 520 along fault planes (e.g. the 6' seat earth; Figure 5a This suggests that faults in these lithologies initiated as a segmented fault-fracture mesh (Sibson, 1996) Sibson, 1990Sibson, , 1992. Mineralised Phase 1 faults that cut multiple lithologies also 529 display multiple slip events ( Figure 6c) and matrix (calcite) supported fault breccias located within relay 530 zones (Figure 6d), intersections between Phase 1 and 2 faults (Figure 6c), and at asperities along the 531 principle slip zone (Figure 6a). This suggests fault valve behaviour was also present along non-self-532 juxtaposed faults (Peacock et al., 2019;Sibson, 1990). 533 Where faults cut multiple lithologies, shale accommodates the rotation of bedding, leading to rotated 534 blocks and multiple generations of curved slickensides ( Figure 6). As shale is buried and compressive 535 stresses increase, the ratio of pre-consolidation stress and compaction-related stresses control the 536 behaviour or shales and mud rocks (Yuan et al., 2017;Nygård et al., 2006). As a general rule, shales are 537 ductile during burial, and brittle during exhumation where they experience stresses below the maximum 538 stress they have encountered. Ductile behaviour of the shales at the time of faulting suggests that both 539 phases of faulting occurred prior to maximum burial, which is estimated at <3,000 m at around 60 Ma 540 for the Limestone Coal Formation (Monaghan, 2014). 541 Fault cores at Spireslack SCM also differ between self-juxtaposed faults ( Figure 5) and those that cut 542 multiple lithologies (Table 2). Wilkins et al. (2001), studying growth of normal faults through jointed 543 lithologies, found similar observations to those at Spireslack SCM with little fault rock development 544 (Figure 3a, 5f), and considerably smaller displacement/length ratios than that expected for faults which 545 do not cut jointed lithologies (Figure 5f, 6a, d). While fault core at Spireslack SCM is typically thin 546 (Table 2) Our data demonstrates that the evolution of faults and fault zone structure, and therefore the bulk 554 hydraulic properties of the rock mass at Spireslack SCM, varied both through time and as faults cut 555 multiple lithologies. The abundance of faults within competent lithologies that cannot be traced into 556 shale interbeds suggests faults at Spireslack SCM initiated as segmented fault strands within competent 557 lithologies (e.g. limestones) and the coals (Figures 5, 8), with shale interbeds. They restricted fault 558 growth and instead accommodating ductile deformation. Despite faulting being dominantly strike slip, 559 the oblique orientation of faults to bedding across the site meant that many fault-growth models derived 560 from observations and modelling of normal faults in mechanically layered sequences appear to be valid 561 (e.g. (Childs et al., 1996;Ferrill et al., 2017;Schöpfer et al., 2006Schöpfer et al., , 2007Schöpfer et al., , 2016 . However, it is also 562 clear that the initial segmented fault network within the competent layers was strongly controlled by the 563 presence and evolution of the joints and mineralised fault zones (Figure 8). It is therefore helpful to 564 consider the concept of lithological juxtaposition, the presence and behaviours of shale interbeds, and 565 the relative timing of deformation, when considering the growth and internal structure of fault growth in 566 mechanically layered sequences. 567

Conclusions 568
The exceptional exposures of the Limestone Coal Formation at Spireslack SCM provides an excellent 569 opportunity to examine the role of lithology and pre-existing structures on fault evolution, internal 570 structure and connectivity. Careful mapping to unpick cross-cutting relationships has revealed a 5 stage, 571 complex geological evolution for the Spireslack SCM succession consisting of two phases of faulting 572 and eight phases of joint development: 573 Stage 1: Cleats and multiple sets of joints formed during burial of the fluvial-deltaic host rocks and 574 formation of the regional Muirkirk syncline. 575 While the overall fracture density increases around the larger faults, counter-intuitively the modern day 586 network connectivity decreases in these areas due to the cementation of faults and joints. 587 We find that the fault zone internal structure at Spireslack SCM depends on: a) whether the fault is self-588 juxtaposed or cuts multiple lithologies; b) the presence and ductility of shale layers, which in turns leads 589 to bed-rotation and fault-core lens formation; and c) the orientation of open and mineralised joints/coal 590 cleats at the time of faulting. Self-juxtaposed faults are strongly affected by the orientation, and 591 mineralisation of pre-existing joint-sets and coal cleats, causing them to grow as multiple segmented 592 fault strands within competent lithologies. Self-juxtaposed faults only become well connected where 593 fault intensity is high. Faults that cut multiple lithologies are strongly affected by the presence of shale 594 interbeds and display a complex and heterogenous fault structure with fault length limited by the 595 presence of pre-existing faults. Therefore, it is crucial to appreciate the relative timing of deformation 596 events, concurrent or subsequent cementation and the degree of lithological juxtaposition when 597 considering the mechanical and hydraulic properties of a mechanically stratified succession. 598 Acknowledgements 599 This work was funded through BJA's PhD studentship, supported by the Environmental and Physical 600 Sciences Research Council (EPSRC, award number EP/L016680/1). LMcK is supported by a 601

University of Strathclyde Environmental and Physical Science Research Council (EPSRC) Doctoral 602
Training Partnership (DTP) award (award reference 1904102). We would like to thank Dave Healy for 603 the use of the high-resolution photomontage of the McDonald Limestone dip slope and the British 604 Geological Society for the use of the photomontage of the high wall. 605