The slant core, together with a CT scan of the core, was examined for fractures by Gale et al. (2018), who developed criteria to identify hydraulic, natural, drilling-induced, and core-handling fractures. Here we specifically focus our analysis on hydraulic fracture morphology and its implications for the hydraulic fracture network.

In the initial slant core fracture description by Gale et al. (2018), 375 induced hydraulic fractures and 309 natural fractures were identified; here we show their distribution along the length of the cores (Figure 2C). In several locations, hydraulic fractures occur in clusters, with locally up to eight fractures present within a 1-m section. Many hydraulic fractures also exhibit fractographic features that may be linked to fracture growth and fracture segmentation such as diversion steps, twist hackles, and paired doublets (Gale et al., 2018).

Diversions where a fracture diverts along a plane of weakness for a distance (commonly less than a centimeter in core observations) before reorienting and continuing to propagate, are expressed in core as steps that extend part-way or all the way across the fracture face. In these cases, bedding planes, concretions, or natural fractures are the planes of weakness. The observed steps have variable geometries including 1) regular and sharp-angled (Figure 3A), 2) curved, smooth ramps (Figure 3B), or 3) highly irregular (Figure 3C).

Observed twist hackles in the core include both gradual twist hackles representing continuous breakdown at the fracture tip, and abrupt twist hackles representing discontinuous breakdown of the fracture tip morphologies (e.g., Figure 4; Younes and Engelder, 1999). Breakdown is observed at the parent-fracture tip, resulting in the segmentation of the fracture into an en echelon arrays typically at a transition in bedding and/or a contrasting lithology. The local direction of propagation is indicated by the twist hackles in conjunction with accompanying plumose structure (Figures 4A,C). In total, 67 hydraulic fractures were observed to have twist hackle features in the HFTS1 slant core. Although propagation directions are variable, dominant propagation direction—represented by 72% of fractures with twist hackles—was vertical (Figures 4B,D). Of twist hackles observed in the slant core, 52% of the transitions are located near lithology changes, and 48% of the transitions are contained within the parent fracture bed.

Bifurcation is the branching of a fracture into two fractures, giving rise to a closely spaced fracture pair or ‘doublet’. The double pairs appear to diverge in the direction of propagation, with propagation direction indicated by fracture surface marks (arrest lines, plumes). However, in the case that surface marks are not present to indicate growth direction, it cannot be ruled out that these features could also be caused by the coalescence of two fractures. Evidence of bifurcation in the HFTS1 core is indicated by the presence of 22 doublets with the core piece between the pair having a characteristic wedge shape, with the branchline of divergence for the doublets being a short distance (typically less than 2 cm) outside the core margin in most cases (Figure 5A). Because the divergence (i.e., branchline) was only observed in a few cases, we could not correlate divergence with specific structures and thus we did not establish a cause for the fracture branching. Determination of the direction of bifurcation for the observed doublets showed that the dominant bifurcation direction in the Wolfcamp A was to the west, and the dominant bifurcation direction in the Wolfcamp B was to the east (Figure 5B), which is consistent with propagation and bifurcation away from the adjacent horizontal well stimulation locations associated with the HFTS1 well pad. The presence of these bifurcation features in core samples collected 18–30 m from a stimulated wells shows that this process operates for a significant distance away from the wellbore. Of the 22 bifurcation doublets, 18 occur in the Wolfcamp A, and 4 occur in the Wolfcamp B.

Overall, 44% of all hydraulic fractures seen in the core show some sort of segmentation or diversion feature. This represents a substantial subset of the overall population. Observation of diversions and twist hackles indicate that mechanical heterogeneities including natural fractures, concretions, or mechanical layer boundaries seem to play important roles in the initiation of these features. However, not all examples of diversions or twist hackles could be attributed to a particular forcing mechanism.

Diversions of hydraulic fractures along natural fractures take the form of natural fracture reactivation where the natural fracture is parted (Figure 6A), or a “jog”, across the natural fracture without initiating reactivation (Figure 6B). Of the 309 natural fractures observed in the core, 26 were characterized to have been hydraulically reactivated—indicated by the visible parting of calcite cement, and/or the presence of proppant within the parted fracture—and 12 showed morphological evidence for a hydraulic fracture jog. The likelihood of a diversion occurring at an intersection with a natural fracture depends on several variables such as bonding strength of the rock-cement interface, cement thickness, angle of approach, length of the discontinuity, and burial depth (Jeffrey et al., 2009; Lee et al., 2015; Wang et al., 2018). We note that Male et al. (2021) calculated that fewer than 1 in 10 natural fractures were reactivated in the HFTS1 slant core, suggesting that they could play a less-active role in inducing fracture complexity than commonly assumed.

The Ernst Tinaja arroyo in Big Bend National Park (West Texas) cuts NE-SW across Cuesta Carlota and drains the Ernst Basin half-graben along the western edge of the Sierra Del Carmen. The arroyo crosses a large SW-dipping panel exposing a substantial section of the lower and upper Cretaceous strata (Maxwell et al., 1967; Moustafa, 1988; Turner et al., 2011). This exposure includes a complete section of the Ernst Member of the Boquillas Formation, which is the lateral equivalent of the Eagle Ford Formation (Maxwell et al., 1967; Lehrmann et al., 2019). The Cretaceous strata in this area experienced SW-NE directed contraction during the Laramide orogeny in the early Paleogene (70–50 Ma; Lehman, 1991), followed by SW-NE directed Basin and Range extension in the Cenozoic (25–2 Ma; Turner et al., 2011). Laramide deformation features observable at the Ernst Tinaja exposure include contractional folds, thrust faults, and tectonic stylolites (Ferrill et al., 2016), and Basin and Range deformation includes normal faults and abundant extension fractures (McGinnis et al., 2017), with relative timing constrained by cross-cutting relationships (Ferrill et al., 2016, Ferrill et al., 2021).

Within this outcrop of the Boquillas Formation, we focus on two well-exposed pavements (bed-parallel exposures) within the lower portion of the Ernst Member that include both bedding-plane and profile exposures through their fracture networks (Figures 7A,C). The beds are composed of lime packstone and grainstone bounded by calcareous mudrock beds (Lehrmann et al., 2019; Frébourg et al., 2016). Pavements 1 and 2 represent tops of limestone beds at stratigraphic heights of 7.8 and 7.0 m in the measured section of Lehrmann et al. (2019) and the rebound profile of McGinnis et al. (2017) (Figure 7B). Beds throughout the exposure are cut by NW/SE-striking bed-perpendicular opening-mode fractures of Basin and Range origin, with spacings observed to be larger in the limestone beds than in the mudstone (McGinnis et al., 2017). Many of these fractures in the limestone beds exhibit twist hackles near bed boundaries—these twist hackles are most easily observed from above in map-view, and several can be seen in profile view along the top and/or bottom portions of beds (Figure 8). One-dimensional scanline surveys (azimuth 225°) were constructed perpendicular to the dominant opening-mode fracture orientation (parallel to the dip direction of bedding) to determine fracture spacing within the network (e.g., Priest and Hudson, 1981; Priest, 1993; Watkins et al., 2015; Gale et al., 2018). Scanline orientation was chosen to minimize the need for geometric corrections to spacing data. Following this work, the fracture pavement was characterized to assess the abundance, spacing, and orientations of twist hackles, and studied in profile to assess the vertical height of twist hackle extents and separation magnitudes. This combined fracture and twist hackle data was collected to provide insights and assist in interpreting similar fracture morphologies observed in the slant core.

The scanline survey for Fracture Pavement 1 extended for 4.7 m and encountered 28 fractures. Of the sampled fractures, 19 (64%) contain twist hackles. Average bed thickness and fracture height for the Pavement 1 bed is 12 cm, fracture lengths (visible) range from 17 to 230 cm, and average fracture spacing is 16.9 cm (Table 1). The scanline for Fracture Pavement 2 extended for 4.8 m and encountered 29 fractures, of which 21 fractures (69%) contain twist hackles. Average bed thickness and fracture height for the Pavement 2 bed is 10 cm, fracture lengths (visible) range from 10 to 88 cm, and average fracture spacing is 16.7 cm (Table 1).

In total across both pavements, 46 fractures were observed to have twist hackles. Hackles per-fracture range from 3 to 88, and varying locations of initiation along the parent fracture trace were observed. Using the height and length data collected for the twist hackles, we calculate that these en echelon arrays add between 6 and 20% of additional surface area to the fracture network compared with single, planar fractures with the same overall height and length dimensions (Table 1). Twist hackles on 49 fractures exhibit counterclockwise rotation with respect to the parent fracture and associated right-stepping arrangements, whereas only 6 fractures—all within pavement 2—show clockwise rotation and left-stepping arrangement (Figure 9A). The dominance of counterclockwise rotation and right-stepping arrangement of twist hackles is most consistent with an overall 26.5° counterclockwise rotation of the extension direction from an azimuth of ∼ 252° at the time of parent fracture formation, to ∼ 225.5° at the time of formation of the bulk of the twist hackles. The Pavement 2 fractures with clockwise-rotated twist hackles suggest another local stress reorientation to ∼ 099°.

The rebound profile for the Ernst Member of the Boquillas Formation at Ernst Tinaja (Figure 5 in McGinnis et al., 2017) documents mechanical stratigraphy using mechanical rebound (measured with an N-type Schmidt hammer) as a proxy for Young’s modulus and unconfined compressive strength. The profile shows that the Pavement 1 and 2 limestones are significantly more competent than the surrounding laminated calcareous mudrock (Figure 7B). In addition to mechanical layering, pre-existing deformation of Laramide age is also present in the outcrop (Ferrill et al., 2016; Ferrill et al., 2021), including thrust faults, tectonic stylolites, and folds. Both pavements studied crop out immediately west of a small fold. Such local structures as well as larger scale structural position with respect to the Basin and Range normal fault system in the Sierra del Carmen are likely to have combined to influence near-field stress orientations during fracture nucleation and reactivation, producing the observed twist hackle pattern (Figure 7A). We also see the combined effect of mechanical stratigraphy and preexisting deformation on a smaller scale in Pavement 2, where a bed-parallel stylolite serves as a transition for fracture breakdown and twist hackle formation (Figure 9B). We interpret that the parent fracture initially cut the lower portion of the bed, terminating against the bed-parallel stylolite. Subsequent renewed propagation upward through the bed in a stress field represented by a clockwise-rotated extension direction produced twist hackles in the upper part of the bed that are rotated clockwise from the parent fracture.

The majority of blocks examined in this study are in the form of 30 cm × 30 cm cubes poured in three layers, each with 10 cm average thickness. The bottom and top layers were made of hydrostone, which is a mixture of 75.5% cement and 24.3% water by weight (Bahorich et al., 2012). These hydrostone layers sandwich a central layer of gypsum plaster, which were 63% gypsum cement and 37% water by weight at the time the layers were poured (Bahorich et al., 2012). Due to their mechanical properties, the hydrostone layers are meant to serve as fracture barriers to the initiated hydraulic fracture, containing them within the middle layer (Bahorich et al., 2012). In total, 33 blocks were studied—of these, the middle layer in 18 of the blocks are homogeneous, and the middle layers in the other 15 blocks contain wafer-like inclusions of sandstone, glass, or plaster to simulate the presence of pre-existing natural fractures. Each block was hydraulically fractured using flat jacks on each face of the cube to simulate anisotropic stress conditions, with the vertical stress being dominant (Bahorich et al., 2012). Blocks contained anywhere from 1 to 4 vertically oriented “wellbores” made of 1 cm diameter aluminum tubing. Each wellbore contained one perforated interval which consisted of pre-drilled holes in the tubing oriented perpendicular to the wellbore. These perforations were filled with pipe cleaners that extended 2 cm out of the perforations into the block medium during the initial pouring and were removed through the wellbore after the block had hardened and prior to fracturing (see Bahorich et al., 2012). This provided an open cavity in the block for fluid to leave the wellbore and pressurize the surrounding medium. Following the hydraulic fracture of the blocks in the lab, each block was broken open to observe the hydraulic fractures inside, which are dyed red from dye included in the pressurized fluid.

Overall, we found that 57% of the experimental block population (19 blocks) show evidence of fracture segmentation (Table 2). Simulated natural fractures appear to play the largest role in introducing complexity to the fracture network. Commonly, the interaction with natural fractures resulted in both fracture diversion, kinking, and the breakdown into twist hackles (Figure 10). These blocks show that 82% of observed twist hackles and 83% of observed diversions are the result of interaction between hydraulic fracture and natural-fracture simulants (wafer-like inclusions of sandstone, glass, or plaster), which also caused reactivation of the natural-fracture simulants.

In blocks with little heterogeneity (that is, few or no simulated natural-fractures), induced fractures also show non-uniform, branched shapes. Here the perforation, or in several cases the wellbore, appear to have served as the locus for generating non-uniformity in the fracture trace. Of the 18 observed homogeneous blocks (those without simulated natural fractures), 83% contained complex fracture morphology. A substantial source of the fracture-face irregularity observed in this lab study came from the perforations. Irregularity seems to arise from this contact point with the host medium, and then either grows into step and ramp features, or segments into multiple fractures (Figure 11). A lack of sealing around the wellbore was also observed to create step and ramp features in several blocks (Figures 11A,B). In some cases, fractures propagated in oblique orientations from the perforation (Figures 11C-F), and overlapped to create the appearance of multiple closely spaced fractures (Figure 12). This geometry looks very similar to that observed in the bifurcation doublets in the HFTS1 slant core. Most of these observations concerning fracture curving and overlap occurred in homogeneous blocks with 3 or more wellbores. It is likely that this behavior results from the presence of these multiple closely spaced wellbores, which creates a modified stress state, resulting in separate fracture wings growing from a single perforation in order to avoid each other. This is commonly referred to as the stress shadow effect (Pollard and Aydin, 1988).