laboratory supercritical carbon dioxide (scCO2) injection experiments (pore water drainage experiments) on small sandstone cores investigating the effects of a single discrete fracture is presented. The orientation and aperture of the fracture are varied across the series of tests. The dynamic Young's modulus and shear modulus along the axis of cylindrical core samples and their related attenuation are determined within a sonic frequency band of ~1 kHz to ~2 kHz, using a resonant bar technique. Concurrently, the distribution of scCO2 injected into the cores is examined using X‐ray CT. The orientation of the fracture with respect to the scCO2 migration and the wave propagation direction is shown to have a large impact on how the seismic wave velocities (or moduli) and attenuations change as a function of scCO2 saturation of porous, fractured rock.
5.1. INTRODUCTION
During geological sequestration of CO2, velocity and attenuation of seismic waves can be monitored to detect the invasion of supercritical CO2 (scCO2) and to determine its saturation in the reservoir rock. The scCO2 introduced in fluid‐saturated (typically by fresh or saline water) porous rock reduces the bulk modulus of the rock, causing reductions in the compressional (or P‐wave) wave velocity. These reductions are usually related to the amount of CO2 in the pore space through simple quasi‐static rock physics relationships such as the Gassmann's fluid substitution model (Gassmann, 1951). However, these models do not always provide satisfactory results, particularly when the seismic waves used by the measurements have relatively high frequency (e.g., Cadoret et al., 1995; Azuma et al, 2013).
Many laboratory experiments for the dynamic properties of CO2‐injected rock have been conducted at ultrasonic frequencies, for correlating a variety of reservoir conditions to seismic signatures as a function of scCO2 saturation. Wang and Nur (1989) conducted laboratory ultrasonic measurements during CO2 injection into sandstone and sand cores filled with oil (n‐hexadecane). Xue et al. (2005) conducted similar tests on a sandstone core initially filled with water. Shi et al (2007) also conducted tomographic measurement on an initially water‐filled core so that the distribution of the CO2 within was imaged. Siggins (2006) conducted measurements using both gaseous and liquid CO2 injected in synthetic and natural sandstones. The experimental results indicated a tendency to match Gassmann model predictions at high effective stresses, but the agreement varied depending on rock types.
Reservoir rock often contains mesoscale (i.e., larger than grain scale but smaller than seismic wavelength scale) heterogeneities. These include patchy distribution of different fluid phases in the pore space (e.g., White et al., 1975; Dutta & Odé, 1979a, b), sedimentary layers with different poroelastic properties (Norris, 1993; Gurevich et al., 1997), and open and partially open fractures and faults (Brajanovski et al., 2005; Nakagawa & Schoenberg, 2007). During seismic wave propagation, these heterogeneities can cause local fluid pressure gradients with scales comparable to the pressure diffusion length (or, the wavelength of Biot's slow compressional waves) for given seismic wave frequencies and fluid and rock properties, resulting in large seismic velocity dispersion and attenuation. Therefore, when studying the effect of scCO2 injection on seismic wave propagation in the laboratory, employing waves with appropriate frequencies can be critically important. Additionally, compliant and open fractures are expected to play an important role in controlling migration of injected fluids in reservoir rock. When both fractures and matrix porosity connected to the fractures are present, wave‐induced dynamic poroelastic interactions between these two different types of rock porosity (high‐permeability, high‐compliance fractures and low‐permeability, low‐compliance matrix porosity) can result in complex velocity and attenuation changes in seismic waves as scCO2 invades the fractured rock. Further, these interactions are affected by the orientation of the fractures with respect to the wave propagation, resulting in different signatures of scCO2 invasion on seismic waves. Understanding these relationships can lead to better predictions of scCO2 behavior during geological carbon sequestration and enhanced oil recovery (EOR) based upon seismic monitoring.
In this laboratory study, we examined changes in dynamic elastic moduli and attenuations related to compressional and shear (torsion) waves during scCO2 injection into sandstone core samples. In the following, we will first describe an experimental setup (a modified resonant bar test system), which allows us to conduct laboratory seismic measurements at frequencies of 1–2 kHz, close to the frequencies used for monitoring of scCO2 injection in the field via crosshole tomography (e.g., Ajo‐Franklin et al, 2013). Subsequently, we will present observed changes in the seismic properties of several fractured rock samples during scCO2 injection tests, with concurrently determined distribution and saturation of the scCO2 in the samples via X‐ray CT imaging. Finally, we will discuss correlations between the changes in the seismic properties and the orientation of the fracture with respect to the scCO2 migration (which is coincidental to the wave propagation direction) as well as scCO2 saturation and distribution within the porous, fractured rock.
5.2. EXPERIMENTAL SETUP
5.2.1. Split‐Hopkinson Resonant Bar (SHRB)
Our seismic measurements employed a variant of conventional resonant bar tests, which allows us to determine the dynamic shear and Young's moduli of a core sample in the sonic frequency range near 1 kHz (Nakagawa, 2011). The basic idea behind this apparatus is that the resonance frequencies of a sample are reduced when sample size and mass are artificially increased by an attached foreign object (e.g., Tittmann, 1977). Once the resonances of the resulting “extended” sample are measured, the seismic properties of the original sample are determined via calibration, modeling, and inversion. In our experiments, a cylindrical rock core (~ 38 mm diameter) is jacketed and placed between a pair of stainless steel rods (Fig. 5.1). The resulting long composite bar is suspended by springs in a tubular aluminum cage, which is inserted into a tubular confining cell (pressure vessel). Longitudinal and torsional vibrations are induced in the bar using piezoelectric sources at one end and measured using accelerometers at the other end. Because the geometry of this apparatus is the same as the conventional Split‐Hopkinson Pressure Bar test apparatus (e.g., Kolsky, 1949), it is called the Split‐Hopkinson Resonant Bar (SHRB).
The longitudinal piezoceramic source is a single disk, and the torsion source is a group of four pie‐shaped laterally polarized shear piezoelectric slabs. Both are made of Type 5600 Navy V piezoceramics (Channel Industries). These sources are driven selectively to introduce a desired mode of vibration in the sample. At the opposite end of the other steel rod, miniature accelerometers (PCB Piezotronics, 352A24) are attached to measure the resulting vibrations. The longitudinal motion is measured by an axial accelerometer, and the torsion motion by a pair of accelerometers oriented in the tangential directions, in diametrically opposing locations. The torsion vibrations are measured by subtracting the output from one of the torsion sensors from the output of the other, resulting in cancellation of electrical noise and unwanted flexural motions contaminating the measurements. During the experiments presented in this paper, the source amplitude was adjusted so that the strains induced in the samples at resonance were in the 10–6 to 10–7 range to reduce possible nonlinear effects.
To ensure good mechanical coupling, the surfaces of the steel bars and the sample are polished flat, then a thin (30 μm) lead foil disk is placed on the interfaces. These disks are cut with a cross‐shaped pattern at the center to allow distribution of the pore fluid along the interface. With these preparations, typically 3–4 MPa of effective confining stress is sufficient to reduce the additional compliance introduced by the interface to a negligible level. The jacket is made of heat‐shrink PVC, with a thickness ranging from 150 μm to 500 μm. With appropriately machined smooth sample surfaces and with application of sufficient effective confining stress (>~1 MPa), this results in good seal at the jacket‐sample interface.
Currently, our experiments with
