Effect of Pressure and Stress Cycles on Fluid Flow in Hydraulically Fractured, Low-Porosity, Anisotropic Sandstone

Global energy consumption is dominated by fossil fuels (Chedid et al. 2007; Aydin 2015), whose demand continues to increase (Aydin 2014a, b; Chang et al. 2012). Conventional hydrocarbon resources have traditionally focused on reservoirs characterized by structural traps and featuring a porous, high permeability reservoir. In contrast, unconventional reservoirs (characterized by low permeability) (e.g. Lee and Hopkins 1994) are often developed and produced by hydraulic fracturing. Whilst in this context, hydraulic fracturing is used to intentionally fracture host rock, it is also an important natural phenomenon in the earth subsurface, exhibited across a range of processes including magma intrusion (Rubin 1993; Tuffen and Dingwell 2005) and mineral emplacement (Richards 2003). However, in the engineered environment, the method has become a standard technique, used in the petroleum industry since the mid-1950’s (Tuefel 1981), to enhance oil and gas production from tight reservoirs (characterized by low permeabilities in the microDarcy range of 10–100’s × 10^–18 mD). Hydraulic fracturing is now a common method to improve oil and gas recovery (Gillard et al. 2010; Kennedy et al. 2012; Wang et al. 2014). These new technologies have led some nations (for example the USA) to become significant producers of natural gas (Wang et al. 2014) as previously low permeable formations were fractured. However, the process is not without controversy, and additionally has been developed over years in a somewhat ‘ad-hoc’ or trial-and-error manner (Golden and Wiseman 2014). This has resulted in varying degrees of overall success due to the complexities of reservoirs that contain significant structural, sedimentological and mechanical heterogeneities. Together, these features alter the relationship between the tensile fracture mechanics needed to generate new fractures for fluid movement, as balanced against the fundamental rock physical properties and local stress field (Martin and Chandler 1993; Sone 2013; Gehne and Benson 2017, 2019).

The objective of hydraulic fracture is to increase the rock permeability through inducing new tensile fracture in the rock mass. This is achieved by pumping a pore fluid (with or without additional propping agents to keep new fractures mechanically open) into a wellbore at a sufficiently high pressure to fracture the surrounding rock mass in tension. This, in turn, requires a sufficiently high fluid flow rate to overcome the background permeability and radial fluid flow, which is a function of the permeability of the unfractured rock mass (Fazio et al. 2021). If the fluid injection is higher that the natural fluid dispersion rate, pressure builds up inside the borehole which leads to fracture, including reopening and further propagation of existing fractures when the in-situ tensile rock strength is exceeded. The resultant hydraulic fracture extends until the formation loss is greater than the pumping rate (Reinicke et al. 2010).

Different approaches have been applied to study the pressure (P_b ) at which the rock first yields (fractures), known as the breakdown pressure. The simple linear elastic approach considers a defect-free, impermeable and non-porous rock matrix around the borehole (Hubbert and Willis 1972; Jaeger et al. 2009) viawhere σ_T is the tensile strength (an inherent property of the rock), and S_h and S_H are the minimum and maximum horizontal stresses, respectively.

However, the above approach represents an ‘end-member’ case as no rock is truly impermeable: all rocks contain pores and fractures, and when saturated with pore fluid exerting a fluid pressure P₀, (Eq. 1) above is modified to:

The expression above (Eq. 2) may be further modified by adding poroelastic effects which account for the rock being both porous and permeable (e.g. Haimson and Fairhurst 1969; Jaeger et al. 2009):where (α) is the Biot poroelastic coefficient and ν is the Poisson’s ratio.

A final, minor, modification considers the role of rock matrix permeability in hydraulic fracturing. In Fazio et al. (2021), Eq. 3 is assumed to be only valid under conditions whereby the bulk rock permeability (k_w) at the interface between the injection fluids and the wall is below a critical permeability (k_wc). Adding these boundary conditions yields:

An accurate charaterisation of the fluid flow through the bulk rock mass is key to understanding reservoir properties (Tan et al. 2018). However, measuring permeability remains challenging due to its sensitivity to heterogeneity. This is further complicated by the strong anisotropy found in typical formations used for unconventional hydrocarbons (such as mudrock, shale and crossbedded/tight sandstone). Nonetheless, numerous studies using wellbore tools and core plugs have attempted to link the fracture process to permeability enhancement via numerical models (Ma et al. 2016). To calibrate these models and in situ data, laboratory measurements of flow through fractures under controlled conditions have used images of the post-test fracture aperture (e.g. Stanchits et al. 2014) or morphology of the post-test shear fracture planes (Kranz et al. 1979; Bernier et al. 2004; Gillard et al. 2010, Zhang 2015a, 2015b), as a function of flow rate or permeability. Collectively, these experiments have provided useful data on fracture behavior, but have tended to focus on mudrocks (shale) over other rock types.

There is a large body of laboratory research examining the controlling elements that affect the propagation of hydraulic fractures, such as stress controls, injection parameters, and interactions with preexisting structures (e.g. bedding planes and/or fractures). Hubbert and Willis (1957) were the first to explore stress controls on fracture propagation, determining the anticipated orientation of fractures with regard to tectonic stresses, assuming tensile (Mode I) failure. Chitrala et al. (2013) found that both shear and tensile failure modes are prevalent in hydraulic fracturing, as revealed by focal mechanism data from Acoustic Emissions (AEs), while Solberg et al. (1977) found that whether shear or tensile failure is the primary mechanism is related to stress ratio. The fluid viscosity, pressurisation (injection) rate, and, more recently, cyclic injection schemes have all been noted as key injection parameters.

Data from Ishida et al. (2004), Stanchits et al. (2015), and Zoback et al. (1977) all indicate that high viscosity fluid is more likely to lead to stable fracture propagation, likely due to high viscosity fluids being less able to easily penetrate tight fractures. Breakdown pressures have also been reported to be influenced by the rate of pressurisation or injection, with higher injection rates leading to higher breakdown pressures (e.g., Cheng et al. 2020; Haimson and Zhao 1991; Lockner and Byerlee 1977; Zhuang et al. 2019). Finally, fluid injection is not limited to constant pressure or flow rates. Hofmann et al. (2018), Patel et al. (2017), and Zhuang et al. (2019, 2020) presented experimental and field work on cyclic injection systems, noting that fracture breakdown pressure is generally lower than comparable constant pressurisation methods, and likewise resulting in a lower maximum amplitude of associated AE events produced by fracture formation. Such methods may be useful for lowering seismic energy releases in a production environment.

The analysis of fracture propagation with respect to anisotropic mechanical qualities and preexisting interfaces is a key challenge, given that the rocks most targeted for unconventional oils (shale and tight sandstone) have pervasive layered structure. This bedding, from m to mm in scale, is a key factor that leads to anisotropy in rocks (Vernik and Nur 1992; Hornby 1998) in terms of both rock physics and permeability (e.g. Benson et al. 2003, 2005). Anisotropy of the rock also affects the strength (Amann et al. 2012; Ulusay 2014; Zhou et al. 2008) which invariably control the fracture orientation during tensile fracture due to hydraulic fracture. Experimental and numerical studies have revealed that a larger differential horizontal stress induces dominant cross‐cutting hydraulic fractures (Hou et al. 2019; Tan et al. 2017; Xu et al. 2015). Fluid flow through this fleshly generated tensile fracture is then controlled by the fracture properties such as aperture, length, asperity and tortuosity (Kamali and Ghassemi 2017; Ye et al. 2017). Permeability enhancement in rocks through hydraulic fracture processes is a key application and has been widely reported (Nara et al. 2011; Zhang et al. 2015a, b; Patel et al. 2017) to measure the increased effectiveness of tight oil and gas reservoirs (Tan et al. 2019).

However, the focus of the bulk of past research in this area has tended to be on shale. Here, we report a new laboratory study designed to measure the fluid flow rate through tensile fractures in a tight anisotropic sandstone (Crab Orchard), with respect to its anisotropy, generated mainly by mm-scale crossbedding. Whilst less extensive than shale, such tight sandstone is frequently encountered in a range of hydrocarbon exploration scenarios. Fractures are freshly generated in the tensile mode using water, via the method of Gehne and Benson (2019) before fluid flow data are taken, up to simulated reservoir conditions to 0.5 km. Fracture aperture data are then imaged post-test using X-ray Computed Tomography (CT) to analyze the final fracture aperture to measured flow rate. Our laboratory setup is designed to eliminate the possibility of altering the fracture properties when extracting the fractured sample as flow rate data is taken immediately after the main macro-scale fracture, and so allows better comparison between the fluid-driven tensile fracture processes (and the associated flow enhancement), to reservoir conditions. Finally, we link these fracture mechanics and fluid flow through the fracture to the accompanying Acoustic Emission (AE, the laboratory proxy to tectonic seismicity) as an additional guide to the timing and development of fracture properties with respect to the mm-scale crossbedding.

Sample assemblies were mounted within a conventional servo-controlled triaxial machine capable of confining pressures up to 100 MPa (Fig. 2). Four 100 MPa servo-controlled pumps provide: (i), axial stress through a piston-mounted pressure intensifier to provide a maximum of 680 MPa, (ii), confining pressure up to 100 MPa. Both these pumps use heat transfer oil (Julabo Thermal HS) as pressurizing medium. Two pore pumps independently provide fluid pressure to (iii), the bottom of the sample (via the lower waterguide) and (iv), receive water through the generated tensile fracture and exiting via the fluid outlet. After fracture, pumps (iii) and (iv) are set to maintain a set pressure gradient and thus establish steady fluid flow through the freshly generated tensile fracture. The final flow rate value is only taken when the flow between the two pumps have achieved a steady, but equal and opposite rate to signify no leaks in the system and to allow transients to settle (approximately 2 min).

Mechanical data (stress, strain, fluid pressures) are recorded at both a ‘low’ recording rate of 1 sample/second and high sampling rates (10 k samples/s), for axial strain and fluid injection pressure only, to record fast changing transients (Gehne et al. 2019). In addition, a suite of 11 acoustic emission sensors, fitted to ports in the engineered rubber jacket (Fig. 1C), received Acoustic Emission (AE) data to monitor fracture speed and progress. The AE signals are first amplified by 60 dB and then received on an ASC “Richter” AE recorder at 10 MHz. For accurate seismo-mechanical data synchronisation during the dynamic tensile fracture, the fluid injection pressure output is split across both mechanical and a single channel of the AE data acquisition systems through an amplified circuit as described by Gehne (2018). This allows data synchronization with an accuracy of ± 0.01 ms.

The experimental procedure spans three stages (Fig. 3). First, hydrostatic pressure is established by increasing the confining pressure and the axial pressure concomitantly to attain the target pressure, and a pre-fracture measurement of fluid flow is taken by setting a differential pressure of 2 MPa between central conduit and the fluid outlet port. Second, pore fluid injection was activated at a constant flow rate of 5 mL/min resulting in an increasing conduit pressure, until failure (hydraulic fracture) occurred (Fig. 3). Evidence of fracture development includes a sharp decrease in injection (pore) pressure, accompanied by a swarm of AE. Third, after tensile failure, a fluid pressure gradient (differential fluid pressure of 2 MPa) was re-established between the conduit pressure and the fluid outlet port to initiate a steady-state flow through the freshly generated tensile fracture(s). The volume of the two pressure pumps were monitored independently; steady-state flow is reached when the volume change with time is equal and opposite for the two pumps, averaged across a 4-min time period and after an initial 2 min elapsed to allow transient effects to decay away. This procedure was repeated as a function of confining pressure increase (and decrease) to investigate the effect of confining pressure and pressure hysteresis on flow rate.

The experimental procedure is summarised in the flowchart (Fig. 4). Initially, the sample assembly is loaded in the the triaxial apparatus, the AE sensors are installed and system integrity is tested for leaks by pressurising the chamber with Nitrogen gas. If there is no leak, indicated by pressure communication between the chamber and the injection pressure pump, the chamber is filled with oil and pressurised until the initial pressure conditions (axial stress and confining pressure) are established. During this process of initial setup, a servo feedback loop is used to maintain a differential stress (Axial stress – Confining pressure) of 0.5 MPa to hold the assembly securely. The target confining pressure is then set for the experiment, with axial stress tracking confining pressure and set higher by approximately 5 MPa. An initial (pre hydraulic fracture) flow rate is measured for about 10 min; during this time the AE activity is monitored and decays to background level. The hydraulic fracture (HF) experiment is then performed by injecting water into the sample chamber using a constant injection rate of 5 mL/Min until breakdown is recorded. Finaly, the post-experiment flow rate through the fracture is measured by setting a differential pressure of 2 MPa between central conduit and the fluid outlet. The chamber is de-pressurised, sample retrieved, and XCT scan is conducted for fracture visualisation and analysis.

Six experiments were conducted on COS at initial confining pressures (before injection) of 5 MPa, 11 MPa, and 16 MPa. At each pressure, a pair of samples were cored with long axis either parallel or perpendicular to bedding. As detailed above, for each sample an initial fluid flow is measured by setting a differential pore pressure (difference between conduit and outlet pressure) and measuring at the upstream and downstream reservoir (Fig. 3). These initial flow rate data are tabulated in Table 1.

Hydraulic fracturing has been established as a key process in both a natural environment (e.g. magma intrusion, and mineralization) as well as the engineered geo-environment, most frequently to develop hydraulic fractures in unconventional reservoirs (Guo et al. 2013; Gehne and Benson 2017, 2019; Tan et al. 2018). The ultimate aim of these methods is to generate a higher permeability in the rock mass for developing the reservoir that would otherwise be uneconomic. However, whilst there have been a large number of studies investigating the fluid flow and permeability properties of highly anisotropic rocks such as shale (e.g.; Walsh 1981; Benson et al. 2005; Gehne and Benson 2017), studies investigating the fracture mechanics of shale (e.g. Hubbert and Willis 1972; Zoback et al. 1977; Teufel and Clark 1981; Rubin et al. 1993; Reinicke et al. 2010), and studies combining these two elements (e.g. Fredd et al. 2001; Guo et al. 2013; Zhang et al. 2015a, b), there are far fewer studies investigating low porosity or ‘tight’ sandstone. This is important as the hydraulic properties of low porosity rocks, like shale, is also significantly modified by both pressure and are often highly anisotropic due to small scale crossbedding, such as in COS (e.g. Gehne and Benson 2019). In addition, like unconventional shale reservoirs, tight sandstone (and limestone) reservoirs are increasingly being targeted for new hydrocarbon exploration.

Here, we have conducted a series of hydraulic fracture experiments in a tight sandstone (nominally 5% porosity and 10^–18 m² permeability) with fluid flow measurement directly after this stage in order to assess fluid flow enhancement as a function of anisotropy across cycles of confining pressure. In our experiments, we note a distinct interplay between the inherent anisotropy of the fracturing materials, with samples cored with long axis perpendicular having a higher breakdown pressure than those parallel to bedding, and the effect of the overall confining pressure. We develop our discussion along these two lines of enquiry below. In general, the cycles of effective pressure have a largely irreversible effect on fluid flow. This is consistent with past studies, including from large sample volumes (Guo et al. 2013; Tan et al. 2018).

In this study, we have investigated the influence of confining pressure and anisotropy on fluid flow through tensile fracture under simulated in situ pressures relevant to hydraulic fracture in a low porosity (tight) sandstone (Crab Orchard). We find that a general increase trend in breakdown pressure and cumulative acoustic emission when confining pressure increases, which leads to an irreversible decrease in fluid flow through the tensile fracture when confining pressure is cycled. In addition, breakdown pressure is higher in experiments with samples cored parallel to bedding at a lower confining pressure (5 MPa), this effect decreases at higher confining pressure (11 MPa and 16 MPa) at injection. We conclude that anisotropy is a significant contributing factor to both the fluid flow hysteresis effect and breakdown stress, with the tortuosity a key factor rather than fracture aperture alone in describing fluid flow rate through the fracture.

In general, the fluid flow is higher in experiments with samples cored parallel to bedding and additionally has weaker recoverability when confining pressure is ‘re-set’. We observed two stages of flow rate reduction during in the two cycles of confining pressure. The first cycle of confining pressure is identified by a rapid decrease in flow rate (e.g. 97% for COSx-1 and 95% for COSz-1), while the second cycle is characterized by a slow decrease in flow rate (e.g. 79% for COSx-1 and 86% for COSz-1). We conclude that it is likely that a combination of mechanisms operate, and must be considered in determining the overall permeability of tight sandstone to regional stresses during burial and uplift (expressed as confining pressure cycles and ‘re-set’). This is not limited to tight sandstone but also a low permeability anisotropic rock material such as shale and mudstone. Finally, we suggest that the open fracture compliance is also important, particularly with regards to cyclical pressure and stress, which is further complicated for rocks such as Crab Orchard that have significant clay content.

This study also highlights the effect of scale and heterogeneity. The smaller grain and finer layering of shale has likely led to more consistent and reliable experiments as previously reported by Gehne et al. (2020) compared to the tight sandstone used in this study. With coarser, mm-scale anisotropy, as seen here in the Crab Orchard Sandstone, it is likely that cm-scale samples are below the minimum size for reliable measurements of breakdown pressure and hydraulic fracture. As such we recommend that larger samples of the dm-scale are used for future studies of hydraulic fracture in coarser grained (250micron and above) samples.