- Copyright ©2007. The American Association of Petroleum Geologists. All rights reserved.
Gas production from the Barnett Shale relies on hydraulic fracture stimulation. Natural opening-mode fractures reactivate during stimulation and enhance efficiency by widening the treatment zone. Knowledge of both the present-day maximum horizontal stress, which controls the direction of hydraulic fracture propagation, and the geometry of the natural fracture system, which we discuss here, is therefore necessary for effective hydraulic fracture treatment design.
We characterized natural fractures in four Barnett Shale cores in terms of orientation, size, and sealing properties. We measured a mechanical rock property, the subcritical crack index, which governs fracture pattern development. Natural fractures are common, narrow (<0.05 mm; <0.002 in.), sealed with calcite, and present in en echelon arrays. Individual fractures have high length/width aspect ratios (>1000:1). They are steep (>75°), and the dominant trend is west-northwest. Other sets trend north-south. The narrow fractures are sealed and cannot contribute to reservoir storage or enhance permeability, but the population may follow a power-law size distribution where the largest fractures are open. The subcritical crack index for the Barnett Shale is high, indicating fracture clustering, and we suggest that large open fractures exist in clusters spaced several hundred feet apart. These fracture clusters may enhance permeability locally, but they may be problematic for hydraulic fracture treatments. The smaller sealed fractures act as planes of weakness and reactivate during hydraulic fracture treatments. Because the maximum horizontal stress trends northeast-southwest and is nearly normal to the dominant natural fractures, reactivation widens the treatment zone along multiple strands.
Julia Gale obtained a Ph.D. in structural geology from Exeter University in 1987. She taught structural geology and tectonics for 12 years at the University of Derby. She moved to the University of Texas at Austin in 1998, working as a research associate first in the Department of Geological Sciences and then the Bureau of Economic Geology. Her interests include fracture characterization in carbonate and shale hydrocarbon reservoirs.
Rob Reed is a research scientist associate at the Bureau of Economic Geology. He received his B.S. degree and his Ph.D. in geological sciences from the University of Texas at Austin and his M.S. degree in geology from the University of Massachusetts. His current research focuses on various aspects of the microstructure of rocks.
Jon Holder received a Ph.D. in physics from the University of Illinois at Urbana Champaign (UIUC) in 1968. He was a member of the Geology Department faculty at UIUC from 1969 until 1981, teaching and conducting research in areas of rock physics. He worked in geotechnical research in the private sector from 1981 to 1989 and then joined the research staff in the Petroleum and Geosystems Engineering at the University of Texas at Austin, where he continues to do research in mechanical behavior in porous media, with emphasis on fracture mechanics.
The Mississippian Barnett Shale gas play in the Fort Worth Basin is the largest gas field in Texas, with reserves exceeding 2.7 tcf (Montgomery et al., 2005). Success in the Fort Worth Basin has increased interest in the Barnett Shale elsewhere, for example, in the Permian Basin (Figure 1). Other Mississippian and Devonian shales are also being considered as possible Barnett-like plays. The geology of the Barnett Shale and structure of the Fort Worth Basin have been described by many workers, including Cheney (1940), Cheney and Goss (1952), Henry (1982), and Martin (1982, and articles therein). More recently, there have been contributions on geochemistry by Pollastro et al. (2003) and Hill et al. (2007), on lithologic characterization (Papazis, 2005), and on depositional setting and lithofacies (Loucks and Ruppel, 2007). Production performance of the Barnett Shale was evaluated by Frantz et al. (2005).
The terms “Barnett Formation” and “Barnett Shale” have both been used formally. “Barnett Shale” as a name is misleading because most of the Barnett Formation is a mudstone rather than shale (Loucks and Ruppel, 2007) but we adopt it here because of its wider usage. In lithologic descriptions, we use the term “mudstone” to describe a relatively nonfissile clastic rock containing dominantly noncalcareous, clay-size particles, in contrast to shale, which is characteristically fissile because of higher clay content. In the general discussion of the play type, however, we use the term “shale” to include both shale and mudstone.
An overall evaluation of the Barnett Shale play was summarized most recently by Montgomery et al. (2005), who identified several factors that make it unique compared with other gas-shale plays. These include the great depth and high pressure of the reservoir and the complex thermal history, which has influenced the geochemistry of hydrocarbon generation and storage. Montgomery et al. (2005), however, also stated that natural fractures do not appear to be essential for production and, in some cases, might reduce well performance, giving the impression that natural fractures are unimportant or undesirable. We do not concur with this general finding, although we agree that there are circumstances where they could be detrimental to well performance, for example, if natural fractures are connected to the water in the underlying Ellenburger Group. The purpose of this article is to demonstrate the nature of the natural fracture system in the Barnett Shale and explain why natural fractures can be useful for improving hydraulic fracture treatment efficiency and, thereby, gas production. Moreover, general principles learned in the Barnett Shale in the Fort Worth Basin might be applicable elsewhere.
Two separate questions regarding natural fractures in the Barnett Shale must be addressed. First, can they provide enhanced permeability or storage capacity for the reservoir? Second, do they enhance or hinder hydraulic fracture treatments? In addressing the first question, we show that whereas storage capacity of the natural fracture system is low because most small fractures are sealed, it is possible that there are large, open fractures in widely spaced clusters that may enhance permeability locally. With respect to the second question, evidence is mounting from microseismic monitoring of hydraulic fracture propagation (e.g., Fisher et al., 2004; Warpinski et al., 2005) that reactivation of the natural fracture network improves efficiency of stimulation (Figure 2). The method of completion in the Fort Worth Basin Barnett Shale has evolved so that horizontal wells are stimulated with low-proppant, high-flow-rate, water-based hydraulic fracture treatments (Warpinski et al., 2005). The wells are commonly drilled normal to the expected hydraulic fracture propagation (normal to the maximum horizontal stress, SHmax) to maximize the volume stimulated by induced fractures. Natural fractures, however, reactivate during treatment, widening the zone of stimulation. Characterization of the natural fracture system and local SHmax is therefore highly desirable to maximize the efficiency of hydraulic fracture treatment design.
Published data from the World Stress Map (Tingay et al., 2006) indicate a consistent northeast-southwest trend for SHmax in the northern part of the Fort Worth Basin near Wise County (Figure 1), which is in agreement with the known dominant hydrofracture propagation trend. By contrast, little published information on natural fracture attributes in the Barnett Shale exists, although natural fractures are common. In this study, we characterized natural fractures in terms of orientation, size, and sealing, and we measured a mechanical rock property, the subcritical crack index, which governs fracture pattern development. Our aim is to provide a basis for investigating how natural fractures might affect the play, with particular emphasis on the interaction with hydraulic fracture treatments.
Approach for Natural Fracture Characterization
Natural fracture data from the subsurface were obtained from vertical cores. Because natural fractures in the Barnett Shale are also mostly subvertical, we encountered a sampling problem where large fractures are typically more widely spaced than the diameter of a borehole and are rarely sampled. Smaller fractures in the same set may be clustered, and the apparent local intensity (fractures per unit volume, area, or scanline length) of fractures observed at any sampling point in a core or image log may not reflect the fracture intensity away from the wellbore. Fractures may be more or less intense than suggested by the sampling. Direct evidence of fracture spacing is typically lacking. Probabilistic methods using fracture data from vertical core have had success in predicting average fracture spacing where fractures are evenly spaced (Narr, 1996), but they do not address the degree of fracture clustering.
To avoid the sampling problem, several workers have used seismic attributes to measure anisotropy associated with fractures. Simon (2005) tried this approach in north Texas using new seismic attributes, including azimuthal interval velocity, seismic volumetric curvature, and interazimuth similarity. Open fractures that correlated with fractures mapped using microseismic data were detected. These fractures, orthogonal to SHmax, had been reactivated by hydraulic fracture treatments. Thus, although the technique can detect reactivated fractures and provides confirmation of such reactivation, it cannot, at this stage, provide sufficient detail to characterize the natural fracture system. Moreover, because it does not detect sealed, unreactivated fractures, it cannot be used in advance of stimulation to predict mechanical behavior of the reservoir.
An alternative approach is to use microfractures that are abundant in core to predict macrofracture attributes. We define a macrofracture as a fracture that can be observed by the eye, whereas a microfracture requires magnification greater than ×10 to be detected. An opening-mode fracture set comprises fractures across a range of sizes. Within a set, orientation (Laubach, 1997) and timing (Laubach, 2003) are consistent across the range of scales, and intensity shows a size-dependent power-law distribution (Marrett et al., 1999). Fracture population attributes of porosity and permeability (Marrett, 1996) and the sealing of opening-mode fractures (Laubach, 2003) are also size dependent. We are able to use these size-scaling relationships to predict the attributes of large fractures from observations of smaller ones in the cores. Additionally, an understanding of opening-mode fracture growth has been developed by Olson (2004) through geomechanical modeling using a measured rock property, the subcritical crack index, as an input parameter (Holder et al., 2001). We have synthesized these different aspects of fracture evolution in case studies in sandstones (Laubach and Gale, 2006) and carbonates (Gale, 2002; Gale et al., 2004), and in this study, we apply a similar approach to the Barnett Shale.
In describing fracture attributes, it is essential to indicate the size range of fractures present. Fracture intensity should be reported with reference to a particular size range, for example, 10 fractures per meter greater than or equal to 1 mm (0.04 in.) wide. Average spacing of fractures is the inverse of intensity and, again, depends on the size of fractures being considered. In the Barnett Shale, many sealed fractures having apertures of less than 50 μm are part of the population. Although these do not contribute to the permeability of the reservoir, they are important planes of weakness that tend to be reactivated by hydraulically induced fractures. All natural fractures, including small sealed fractures and large potentially open fractures, must therefore be taken into account when predicting hydraulic fracture behavior.
Structural Geology of the Fort Worth Basin
The large-scale structure of the Fort Worth Basin contains several arches and faults (Figure 1). These are mostly associated with the late Paleozoic Ouachita orogeny. Possible mechanisms of fracture formation might be deduced from kinematic analysis of these large structures, together with burial-history data. It is generally not valid, however, to link opening-mode fractures with large structures on the basis of orientation alone. Equivalence of timing, tied with sound mechanical reasons for linking the structures, is required. Opening-mode fractures can form readily under many different conditions partly because rocks have low tensile strength. Moreover, evidence of crack-seal texture in north-south–trending fractures indicates fracture growth proceeded by multiple events. In the Fort Worth Basin, because of the complex structure and the added possibility of hydrocarbon cracking and migration, there could be several different mechanisms of fracture formation. For this reason, although the major structures are presented for the sake of context, we do not speculate on fracture origin from the range of possibilities. Detail on fracture orientation and relative timing is given where known, but it is generally insufficient to pin down the cause of fracturing.
In addition to the structures shown in Figure 1, saglike features in the Barnett Shale have been described and interpreted by Hardage et al. (1996) and McDonnell et al. (2006) as being related to paleokarst collapse in the underlying Ellenburger Group. The sags are bounded by circular fault systems, which connect down to the Ellenburger Group. The sag zones are considered undesirable for gas production because of the potential of the faults to connect into water in the Ellenburger Group. Locally, opening-mode fractures may have developed around these faults, but they would be a separate population from the regionally developed opening-mode west-northwest–trending fractures that are the subject of this article.
Natural fractures were analyzed in two Barnett Shale cores from Wise County in the Fort Worth Basin (Mitchell Energy 2 T. P. Sims and United Texas 1 Blakely) one from Erath County (Cities Service 1 St. Clair C), and one from McCullough County close to the Llano uplift (Houston Oil and Minerals MC-1 Johanson, Harold) (Figure 1). Fracture dimensions and mineral fill were recorded for individual fractures, and fracture pattern characteristics were also noted. Of the cores studied, only the T. P. Sims core was oriented. Fracture orientation information is therefore confined to this core, although it is possible to determine angles between multiple sets in unoriented core. Core may be broken along natural fracture planes; calcite crystals on the fracture surface allow distinction between natural and induced fractures. Previous work on natural fractures has been included where appropriate. Specifically, fracture orientation data from the T. P. Sims core were obtained from an FMS (Formation MicroScanner) log by Hill (1992).
Two samples from the T. P. Sims core, one from a typical mudrock layer and one from a dolomitic layer, were chosen for analysis. Some additional mudrock thin sections were examined briefly for comparison with observations made on the first mudrock sample. The T. P. Sims core is from the lower unit of the Barnett Shale. The cored interval starts 12.2 m (40 ft) below the Forestburg limestone, extends for 44.2 m (145 ft), and ends 35 m (115 ft) above the base of the lower shale unit.
Samples were examined using a variety of imaging and analytical techniques. The dolomitic thin section was first examined using cold-cathode cathodoluminescence (CC-CL) microscopy, in which CC-CL images were paired with transmitted light images. Cold-cathode cathodoluminescence was done using a Technosyn MK-II system operating at approximately 15 kV.
The scanning electron microscope (SEM)–based analytical methods used on the samples were secondary electron and backscattered electron imaging and SEM-based cathodoluminescence (SEM-CL) imaging. Energy-dispersive spectroscopy (EDS) was used to determine elemental composition of minerals present in the samples. All SEM images were done on an FEI XL30 SEM equipped with an Oxford Instruments ISIS EDS system and an Oxford Instruments MonoCL2 cathodoluminescence system. The SEM was operated at 15 or 20 kV and at large sample currents for CL imaging. Because of problems with persistent luminescence, in most cases, SEM-CL imaging of carbonates requires acquiring only shorter wavelength CL emissions, which is done most easily by inserting a broadband, short-wavelength (ultraviolet-blue) filter between the mirror and the photomultiplier tube (Reed and Milliken, 2003).
Mechanical Rock Properties
Subcritical Crack Growth
Mechanical rupture of a material occurs when the mode I (normal opening) stress intensity factor, KI, is equal to the critical stress intensity factor or fracture toughness, KIC, at which time the fracture propagates with a velocity slightly below the shear-wave velocity. However, fractures can also propagate at stresses well below this level at velocities several orders of magnitude slower than the rupture velocity. During this regime of crack growth, the material is strained at levels below that necessary for breaking bonds, but the strained bonds are weaker and more prone to chemical attack (Lawn, 1975; Atkinson and Meredith, 1981). The bonds are further weakened and ultimately broken by thermally activated chemical interactions. An empirical power law, introduced by Charles (1958), provides a good correlation between subcritical fracture velocity, V, and the stress intensity factor (or, because of its proportionality to KI, applied load, P): where n is the subcritical crack index, and k0 and A are constants. This is the defining equation for the subcritical index.
In tectonically stressed crustal rocks, subcritical crack growth can be significant, and studies have shown that fracture-spacing length distributions, connectivity, and fracture aperture can be controlled by this process (Olson et al., 2001). For low values of n (<20), computed natural fracture patterns exhibit small spacing relative to bed thickness. At high values (n ≥ 80), fractures are spatially arranged in widely spaced swarms or clusters. Intermediate values (20–80) result in more regular fracture spacing that is roughly proportional to layer thickness. Discrete fracture flow modeling of geomechanically generated fracture patterns has demonstrated variations in effective permeability with subcritical index, primarily through its influence on fracture length distributions (Philip et al., 2002; Rijken, 2005).
The subcritical index can be measured in the laboratory (Atkinson and Meredith, 1981; Atkinson, 1984), and values for a range of rocks and testing environments have been reported (Atkinson and Meredith, 1981; Holder et al., 2001; Rijken et al., 2002; Rijken, 2005). The experiments are conducted in ambient laboratory conditions at zero depth. Theoretically, this should not affect the subcritical crack index, whose governing equation is only marginally affected by confining pressure. For most studies, including the present study, values for the subcritical indices are determined from measurements of load decay in a dual torsion beam configuration at constant displacement (Williams and Evans, 1973). This technique is based on empirical evidence that the effective specimen compliance, S, is a linear function of crack length, a. This is equivalent to the ratio of displacement normal to the plane of the test specimen, y, to the normal load, P: In this expression, B is a constant; S0 is compliance when a = 0; and the crack velocity, V, is given by the time derivative of a. For a constant normal displacement, y0, the crack propagation velocity is determined by the rate of change of the load, P. where C is another constant. The crack velocity can then be determined from numerical differentiation.
Subcritical Crack Index Measurement
Subcritical crack index measurements were made on two samples from the T. P. Sims core. The core samples were sliced parallel to bedding, and several test specimens with dimensions of 2.5 cm × 5 cm × 1.5 mm (1 × 2 × 0.6 in.) were cut from the slices, polished on one side, and grooved along the center of the nonpolished surface.
Measurements of subcritical crack propagation were conducted in the following test sequence:
The specimen is loaded in steps of approximately 0.23 kg (0.5 lb), holding the applied load constant during each interval by means of a programmed stepper motor.
When crack formation and propagation were indicated by increases in vertical displacement of the loading ram, further displacements were stopped.
The load was allowed to decay for approximately 10 min, attaining an approximately constant value.
Microsoft Excel's SOLVE option was used to determine the parameters from a least-squares fit of all load decay data. This process is set up in an Excel template, and the entire fitting procedure is conducted in a few seconds.
A summary of fracture attributes for the four cores examined is given in Table 1.
Fracture Characterization, T. P. Sims Core
All natural fractures in this core are sealed with calcite and are typically present in en echelon groups (Figure 3). Offsets between adjacent en echelon fractures are typically less than 5 mm (0.2 in.), and the sense of offset may change within a group. Fractures range from less than 50 μm to 0.265 mm (0.01 in.) in width in the mudstone and up to 2.15 mm (0.08 in.) in a concretion (Figure 4a). The tallest fractures extend up to 81 cm (32 in.) in height before continuing out of the core, so this represents a truncated value for height (Figure 4b). True heights of the tallest fractures are not known. For each fracture, the nature of upper and lower terminations was recorded as being gradually tapered, abrupt at bedding, abrupt at a concretion, or out of core. Of the 84 fractures measured in 110 ft (34 m) of core, 40% of terminations were gradually tapered, 4% abrupt at bedding or at concretion, and 56% out of core. Mechanical boundaries within the shale are most commonly associated with changes in the carbonate content, but carbonate layers are not always barriers to fractures. Aperture-height relationships were plotted for all fractures with a true measured height, but were distinguished on the basis of whether they terminated at a mechanical boundary or they gradually tapered (Figure 4c). No correlation exists; fractures of a given aperture have a wide range of heights.
Examples of induced fractures are present but are distinguished from natural fractures by the absence of calcite cement and, in the case of drilling-induced petal centerline fractures, on the basis of geometry. Fractures trend dominantly northeast-southwest (Hill, 1992) (Figure 5a). One dip-slip fault trending 109°/55°SSW was identified in the T. P. Sims (Hill, 1992). Other unoriented dip-slip faults with slickensides and calcite-cemented breccias have been observed in other Barnett Shale cores (Papazis, 2005).
Microfracture Analysis, T. P. Sims Core
Mineralized fractures range from 50 μm to 0.2 mm (0.008 in.) width in the samples chosen for detailed study. Macroscopically nonmineralized, apparently induced fractures are also present in both. Minimal micrometer-scale fracture porosity exists in the small macrofractures. Bowker (2003), in trying to explain the abundant Barnett Shale gas, yet low-permeability suggested that natural microfractures might be present that could enhance permeability. We see no evidence of widespread open natural microfractures; fractures that are present are sealed. Conversely, extensive SEM-based examination of open, northeast-trending fractures in both samples showed no evidence of mineralization. Open microfractures in core samples are most likely induced by drilling or core removal and handling. Material present near the tips of these fractures is a mixture of clay and barite, apparently drilling mud that was not removed prior to epoxy impregnation.
The mudstone sample is composed primarily of clay-size quartz and feldspar with subsidiary dolomite, clay minerals, organic material, pyrite, and microfossils and fossil fragments. Two sets of fractures are present within the mudstone sample (Figure 5b). One set trends east-northeast and is apparently drilling induced. One set trends northwest and is mineralized and, thus, natural. No unequivocal evidence of a second set of mineralized fractures has been observed in any of the mudstone core examined for this study. Fracture fill in Barnett Shale mudstone fractures is mostly calcite. Other phases are developed locally, near compositional anomalies in the host rock, or near fracture tips. Changes in type, amount, and proportion of mineralization are common near fracture tips in other clastic rocks (Laubach, 2003).
The west-northwest–trending macrofracture set in the mudstone layer has an atypically large assemblage of fracture-lining minerals: calcite, quartz, albite, pyrite, barite, and dolomite (Figure 6). Albite and quartz dominate at the fracture tips. Away from the fracture tips, calcite is the dominant fracture-filling cement. Papazis (2005) noted all these fracture cement types (plus sphalerite), but not the occurrence of all six in a single fracture. Quartz, which forms partial or complete bridges, is less common than albite. The SEM-CL imaging reveals complex zoning and shows the connection of albite and quartz cement with grains in matrix.
The SEM-CL imaging of calcite mineralization shows faint zoning parallel to fracture walls in some places. It could be growth zoning, but it is more likely to indicate limited crack-seal texture, in which the fracture opened in three or more small steps. Calcite formed after at least some quartz. Single crystals of calcite fill that are in optical continuity extend for several millimeters along the fracture length, although some fibrous structures are visible. Pyrite is present in large patches containing micrometer-scale inclusions of albite and quartz. Pyrite shows euhedral faces against calcite. Limited development of pyrite hinders textural interpretation. Barite is confined to the center of the fracture in dispersed small patches, commonly greater than 10 μm in diameter, but a few are up to 0.5 mm (0.02 in.) in diameter. Barite is most commonly surrounded by albite and is rare in calcite-filled areas of the fracture. Dolomite is relatively rare and found in association with both calcite and albite. Dolomite rhombohedra within calcite fracture fill are late, and the crystal shape suggests that dolomite is replacing calcite instead of growing into fracture porosity.
The paragenetic sequence within the fracture is complex, with evidence of synchronous growth of some phases. Quartz and albite are partly synchronous and early, forming syntaxial crystals that nucleated on grains in the matrix. Some albite predates calcite; in calcite-dominated segments of the fracture, albite forms partial bridges, with gaps filled by calcite. At least some calcite predates pyrite, where the pyrite may be replacing calcite. Barite postdates most albite, and because it is confined to the middle of the fracture fill, we interpret it as a late phase.
Carbonate concretions, which locally can be several tens of centimeters in height, are developed throughout the Barnett Shale. Concretions are commonly more fractured than the mudstones, but fractures terminate within individual concretions. Fractures in concretions typically have complex geometries and multiple phases of fill and are unlike fractures in the mudstones. Because they are local to individual concretions, these fractures are not considered important with respect to hydrocarbon production, and we did not study them further.
The dolomitic sample comprises mostly ferroan dolomite and calcite, but with significant pyrite, phosphatic material, and clay. Fossil fragments and microfossils of several compositions are present, along with small amounts of albite. This lithology is one of the thin coarse-grained accumulations noted by Papazis (2005), specifically a ripple cross-laminated interval. Loucks and Ruppel (2007) interpret these layers to have been starved ripples formed by contour currents in deep water.
Three or more sets of fractures occur in the dolomitic layer (Figures 5c, 7). Three compositional variants (dominantly calcitic, dominantly dolomitic, and those with highly variable cements) do not correspond strictly to the three fracture sets. All northwest-trending fractures are dominantly calcite, as is the corresponding set in mudrock. Also present are two sets of roughly north-south–trending fractures that are difficult to differentiate. Similar fracture fills are present in both sets: dolomite + calcite ± pyrite and quartz. In a couple of fractures in the sample, pyrite is locally the dominant cement. Crack-seal texture was observed in one of the predominantly dolomitic north-south–trending fractures (Figure 8). Where present, quartz forms partial bridges with chaotic CL zoning.
Fracture-filling pyrite in the dolomitic sample contains numerous micrometer-scale inclusions of calcite, suggesting partial replacement of calcite by pyrite. Pyrite growths fill the fracture and then continue into the host rock, exceeding the width of the fracture where dolomite + calcite ± quartz is the fracture fill. This relationship also suggests that pyrite is replacing some fracture-lining mineral (probably calcite) instead of growing into the fracture porosity.
The more westerly trending set seems to be younger, although timing relations are not definitive. In at least one case, this set appears to reactivate parts of the earlier set, contributing to the complexity of fracture cement in the resulting composite fracture. Both sets of north-south–trending fractures are cut by northwest-trending fractures, establishing the north-south fractures to be older than the northwest-trending fractures.
Extensive examination of the mudstone layer in the T. P. Sims core using numerous imaging methods failed to produce any evidence of the two early sets of generally north-south–trending fractures seen in the dolomitic layer. One possible explanation is that because the dolomitic layers consolidated prior to the mudstone, they retain evidence of earlier fracturing events. Another possibility is that the distribution of these fractures in the mudstone layers is such that they do not occur in the size of sample we examined. This could be caused by either low overall intensity or a high degree of clustering with the sample in the study being outside of a cluster.
Fracture Characterization, Blakely Core
The Blakely core contains upper and lower Barnett Shale, separated by the Forestburg limestone. Most natural fractures in this core are found in the Forestburg interval, for example, at 7132 and 7134.25 ft (2173 and 2174.51 m) (Figure 9a, b, respectively), or are associated with concretions (Figure 9c). Only one natural fracture was observed in the shale. Fractures in the Forestburg interval are arranged in subvertical clusters, terminating within the limestone. Two natural fracture sets in the Forestburg limestone exist with trends differing between 21 and 31° (Figure 9b). These may be related to the two north-south fracture sets in the dolomitic layer in the T. P. Sims core (Figure 5c). Aperture and length data for fractures in the Forestburg limestone are plotted together with data from the T. P. Sims for comparison (Figure 4). Several fractures in both cores have similar dimensions, but there are also wider and longer fractures in the Forestburg interval than in the Barnett Shale in the T. P. Sims core. It is possible that more fractures are present in the mudstone than are suggested by the limited sample provided by this vertical well. Alternatively, fracturing may have occurred preferentially in the Forestburg limestone because it was more brittle than the surrounding mudstone. Although the Barnett Shale is considered to be relatively brittle in comparison with other shales because of its low clay content, it is, nevertheless, likely to be less brittle than the Forestburg limestone.
Fracture Characterization, MC-1 Johanson Core
The Barnett Shale section in this core from close to the Llano uplift is just 13 ft (4 m) thick. The core, which is 2 in. (5 cm) in diameter, has parted along bedding planes but is relatively well preserved. No natural fractures were observed in the mudstone, which is locally rich in skeletal debris. Two calcite-sealed fractures approximately 1.5 cm (0.6 in.) tall and less than 1 mm (0.04 in.) wide that are present in concretions do not extend into the mudstone. Fracture intensity in this location is apparently low, although if fractures are clustered, the local intensity could be misleading.
Fracture Characterization, St. Clair Core
A Barnett Shale section 23 ft (7 m) thick at a depth of 4973–4996 ft (1515–1522 m) is present in this core from the Erath County. Three natural fractures were observed in the mudstone, each being less than 0.05 mm (0.001 in.) wide and sealed with calcite. The fractures terminate outside the core, so the heights of 9, 4, and 3.5 cm (3.5, 1.5, and 1.3 in.) are minimum values (Figure 4c). The fractures are steeply dipping and resemble those in the T. P Sims core, but the core is not oriented, and the fracture orientations are therefore unknown.
Subcritical Crack Index Measurements
Representative load decay data from two of the Barnett Shale tests in this study, for which subcritical indices are included in parentheses, are shown (Figure 10). The total time for each test is approximately 10 min. Note that the ranges in load differ significantly between the two indices, but variations are smooth and fit well to the behavior predicted by equation 3. To determine the magnitude of the subcritical index from equation 1, the load decay curves are numerically differentiated to obtain the velocities. From a log-log plot of velocity against load, the index, n, is given by the slopes of the curves, independent of all the constants in the equations given in the methods section of this article (Figure 11).
Indices for all the specimens tested range from 109 to 326, with means of 276 ± 54 for the specimens from 7692 ft (2344 m), and 122 ± 20 for the specimen from 7749 ft (2361 m) (Table 2). The results for specimens at a depth of 7692 ft (2344 m) are from tests on different specimens. Data from the 7749-ft (2361-m) depth are from multiple tests on the same specimen because other specimens failed in preparation. The range of indices is near the high end of measurable values; the specimens behave as almost perfectly brittle. High indices indicate a rapid transition from zero propagation to almost rupture-crack velocity as load increases by a small amount (equation 1). The indices determined for these shale samples are comparable to indices for dolostones and chalk (Table 3), but are high relative to those for sandstones, which have means of approximately 55 (Rijken, 2005).
DISCUSSION AND INTERPRETATION
Natural opening-mode fractures in the Barnett Shale are most commonly narrow, sealed with calcite, and present in en echelon arrays. The narrow fractures are all sealed and cannot contribute to reservoir storage or enhance permeability. Individual fractures have high aspect ratios. The host rock has a high subcritical crack index. These characteristics are also seen in fractures in the Austin Chalk, a well-known fractured reservoir (Gale, 2002) (Figure 12a, b). We consider the Austin Chalk to be a good analog for the Barnett Shale with respect to natural fracture patterns. It is similar in the sense that the Barnett Shale consists of mudstones interspersed with carbonate layers, and the Austin Chalk comprises chalk-marl couplets. Both are fine-grained, layered systems with low matrix permeability. Moreover, their mechanical rock properties are similar (Table 3).
In the Austin Chalk, fracture aperture sizes follow a power-law distribution, and commonly only the largest fractures, above an 11-mm (0.43-in.) emergent threshold, are open (Figure 12c). The emergent threshold is the size above which fractures in a given population are likely to retain porosity (Laubach, 2003). We have not directly observed large open fractures in the Barnett Shale, although we note that Simon (2005, p. 47, his figure 31) reported partly open natural fractures in an image log. We suggest here that fracture apertures in the Barnett Shale may follow a power-law distribution where, without stimulation, only the largest fractures are open to flow. In the Austin Chalk, the largest fractures are at least 10 cm (4 in.) wide and are mostly sealed, but they have openings up to 1 cm (0.4 in.) wide (Figure 12d, e). Horizontal wells drilled in the Austin Chalk exploit these open fracture clusters. By analogy, horizontal wells drilled normal to natural fractures in the Barnett Shale might intersect an open fracture cluster. If the fractures are contained within the shale, then this could be useful in enhancing permeability. They could be problematic with respect to hydraulic fracture treatments, however, because open natural fractures can capture treatment fluids and prevent new fractures from forming. If the large natural fractures connect to water in the underlying Ellenburger Group, they could be detrimental. The effect of drilling into an open, natural fracture cluster is therefore partly dependent on the height of the fracture system and whether it connects to the Ellenburger Group.
The maximum aperture of elastic fractures scales with the smallest dimension normal to aperture, which is generally the fracture height. In the case of the Austin Chalk, mechanical layer thickness, which controls height, can be several tens of meters. Mechanical layer thickness has not yet been determined for the Barnett Shale. The upper constraint is the entire thickness of the formation, including the Forestburg limestone (>300 ft [>92 m] in the thickest part). Alternatively, the upper and lower Barnett thicknesses may provide an upper limit. These are on the order of 100–250 ft (30–76 m) thick, respectively. Internal carbonate-rich layers provide smaller scale mechanical boundaries for propagation of some fractures (Figure 9c), but not for others (Figure 12b). Vertical persistence of fractures is affected by the relative thicknesses of the fracturing layer and the bounding layers, together with the size of the propagating fracture when it arrives at the boundary.
The Forestburg limestone is approximately 38 ft (12 m) thick in the vicinity of the Blakely well (Loucks and Ruppel, 2007). Fractures could attain this height before being constrained by the surrounding shale. At a height of 12 m (39 ft) and a height/width aspect ratio of 1000:1 (Figure 4c), fractures could grow to be 12 mm (0.47 in.) wide. These fractures may retain significant porosity and permeability. If this is the case, then although the Forestburg is regarded as being a barrier for hydraulic fractures (Hill, 1992), isolated, open, natural fracture swarms might connect the upper and lower Barnett shales if they are tapped into during hydraulic fracture treatment.
Geomechanical modeling of growing fracture systems has shown the importance of the subcritical index for controlling the degree of fracture clustering (Olson, 2004). Large indices are linked with strongly clustered fractures. Because the subcritical index in the Barnett Shale is large, we interpret the fractures to be strongly clustered. Geomechanical models using high subcritical indices indicate that the spacing between clusters may be two to three times the mechanical layer thickness. For the largest fractures, a mechanical layer thickness of the whole Barnett Shale might be appropriate. Fracture spacing would then be between 600 and 900 ft (184 and 276 m), a value comparable to Austin Chalk cluster spacing, which is in excess of 250 m (820 ft) (Gale, 2002). If cluster spacing is similar to that in the Austin Chalk, then we predict that large open fractures will be seen on horizontal well image logs, and evidence of them should be present during drilling. They are unlikely to be sampled in vertical core or on vertical well image logs, however. If the mechanical layer thickness is much smaller, then the maximum fracture aperture will be smaller, and most fractures will probably be sealed. Simon (2005) saw a marked northwest-oriented, fastest azimuthal interval velocity around a well in the Barnett Shale where fractures were seen in the image log. If the fractures were clustered, then Simon's (2005) inference that fractures outside the image volume add to the fastest azimuthal interval velocity signal may be correct.
Large natural fractures perpendicular to present-day SHmax may be open, although the stress regime is unfavorable. This is because fracture cement props the fracture open, and cement that precipitated in the host rock during or after fracturing renders the rock less compliant and preserves the opening (Laubach et al., 2004).
Why Natural Fractures Reactivate
Hydraulic fractures follow SHmax, until they encounter natural fractures. In the case of the Barnett Shale, microseismic monitoring (Fisher et al., 2004; Warpinski et al., 2005) has shown that the hydraulic fractures stimulate the natural fractures to open, producing a complex network (Figure 2). In many reservoirs where hydraulic fractures encounter natural fractures, the hydraulic fractures are blocked from further propagation (Warpinski and Teufel, 1987), but this is not a problem in the Barnett Shale (Warpinski et al., 2005). We propose that natural fractures are not a barrier because the tensile strength of the contact between the calcite fracture fill and the shale wall rock is low. The strength is low because calcite in the fracture is not growing in crystallographic continuity with grains in the wall rock. No crystal bond exists between the wall rock and the calcite cement. This is in contrast to quartz cement in fractures in tight-gas sandstones (Laubach, 2003). We know the fracture-host boundary in the Barnett Shale fractures is weak because fracture-cement fills in the cores are commonly parted from the wall rock (Figures 3, 12). Although these cases are caused by core-handling damage, we suggest that elevated fluid pressures from a hydraulic fracture treatment cause similar failure in the subsurface. Thus, when natural fractures are encountered, they are opened up to fluids, and they provide a network connected to the wellbore. Natural fractures in this scenario are advantageous to optimal stimulation of the well.
Natural, regionally developed, opening-mode fractures in the Barnett Shale can reactivate during hydraulic fracture treatments, providing a larger rock volume in contact with the wellbore than would be the case with a single hydraulic fracture. The natural fracture system must therefore be characterized, and in-situ stress must be determined for hydraulic-fracture treatments to be optimized.
Although natural fractures observed in the Barnett Shale are mostly sealed, they probably follow power-law aperture size distributions, so that a few wide fractures may be open. Barnett Shale has a high subcritical crack index, indicating that fractures are highly clustered. In the Fort Worth Basin, at least two sets of natural fractures are present, an older north-south–trending set and a dominant, younger, west-northwest–east-southeast–trending set. Cements in the fractures are not generally templated onto grains in the wall rock, and the fractures act as planes of weakness that can reactivate. The in-situ stress in the Fort Worth Basin is well known, with SHmax trending northeast-southwest. SHmax trends are less consistent to the west in the Permian Basin because this region is at the junction of several modern-day stress province boundaries. When the Barnett Shale–type play is extended outside the Fort Worth Basin, the natural fracture system should be characterized, and in-situ stress should be measured.
J. F. W. Gale thanks Bob Loucks and Steve Ruppel for encouragement to work on the Barnett Shale. Peggy Rijken did the subcritical testing. Randy Marrett, Steve Laubach, and Jon Olson provided the foundation for ideas concerning fracture scaling and sealing and fracture pattern evolution. In-situ stress data were obtained from the World Stress Map, an open-access database project of the Heidelberg Academy of Sciences and Humanities and the Geophysical Institute at Karlsruhe University. Helpful reviews by Ron Hill, Ron Nelson, and Gary Prost, together with editorial comments from Ernie Mancini, helped us to improve the manuscript. The State of Texas Advanced Resource Recovery program supported J. F. W. Gale. J. Holder and R. M. Reed were supported by the University of Texas Fracture Research and Application Consortium. Additional support was provided in part by the John A. and Katherine G. Jackson School of Geosciences and the Geology Foundation at the University of Texas at Austin. This article is published with permission of the director of the Bureau of Economic Geology, University of Texas at Austin.
- Manuscript receivedJune 1, 2006.
- Revised manuscript receivedAugust 24, 2006.
- Revised manuscript receivedOctober 13, 2006.
- Final acceptanceNovember 1, 2006.