CO2 Injection-Induced Shearing and Fracturing in Naturally Fractured Conventional and Superhot Geothermal Environments

This study elucidates the possibility of CO2 injection-induced formation of a complex cloud-fracture network (CFN) in granite, along with shearing (shear slip) of natural pre-existing fractures under conventional (~ 150–300 °C) and superhot (> ~ 400 °C) geothermal conditions, potentially providing additional connecting flow paths between the stimulated natural fractures. Here, we conduct a set of experiments under triaxial stress states at 150 °C and 450 °C on cylindrical granite samples containing a sawcut (representing a natural fracture) inclined 45° from the sample’s axis. CO2 injection induced dynamic shear slip on the sawcut, with higher slip velocities at 150 °C owing to the higher elasticity of the rock than that at 450 °C. The lower viscosity of CO2 also allowed it to more uniformly pressurize the sawcut, resulting in higher slip velocities and slip displacements compared with those based on water injection in the 150 °C experiments. This implies that under conventional geothermal conditions, CO2 injection can stimulate the same volume of rock as that of water injection at lower injection-well pressures. The CFNs then formed via CO2 injection at pressures similar to those predicted by the Griffith failure criterion as the sawcut shear slip progressed at both experimental temperatures. The experiments also revealed potential challenges associated with CO2 injection in naturally fractured geothermal environments, such as narrow aperture CFNs, owing to decreasing differential stress during their formation; all these factors should be addressed in future research. CO2 injections induced larger-shear displacement on the sawcut at conventional geothermal conditions (c.a., 150–300 °C) compared to superhot geothermal conditions (> c.a., 400 °C). At conventional geothermal conditions, CO2 injection was revealed to produce larger-shear displacement on sawcut compared to water injection. CO2 injection induced a cloud-fracture network (CFN) with relatively narrow fracture apertures in granite samples with sawcut at conventional and superhot geothermal conditions. CO2 injections induced larger-shear displacement on the sawcut at conventional geothermal conditions (c.a., 150–300 °C) compared to superhot geothermal conditions (> c.a., 400 °C). At conventional geothermal conditions, CO2 injection was revealed to produce larger-shear displacement on sawcut compared to water injection. CO2 injection induced a cloud-fracture network (CFN) with relatively narrow fracture apertures in granite samples with sawcut at conventional and superhot geothermal conditions.

Keywords Carbon dioxide fracturing · Complex cloud-fracture network · Natural fracture network · Enhanced geothermal system

Introduction
Enhanced geothermal systems (EGS) are potentially feasible energy resources for providing baseload electricity (Tester et al. 2006). Over the last 2 decades, research on EGS development has expanded into the exploitation of superhot and supercritical geothermal environments (> ~ 400 °C), which could increase the productivity and sustainability of power generation (Asanuma et al. 2012;Elders et al. 2018). In addition, superhot geothermal reservoirs offer the potential benefit of geologic carbon storage because carbon dioxide (CO 2 ) is denser than the resident brine; hence, it descends (Parisio and Vilarrasa 2020).
Studies have recently demonstrated that fracturing could contribute to the creation of an EGS reservoir in superhot geothermal environments (Watanabe et al. 2017b(Watanabe et al. , 2019Goto et al. 2021), where a complex cloud fracture network (CFN), i.e., a network of permeable microfractures densely distributed over the rock body ( Fig. 1), forms in granite as a result of injecting water at its near-to supercritical temperatures. The CFN is induced at lower injection pressures than during conventional hydraulic fracturing with colder water, because of the stimulation of pre-existing microfractures by low-viscosity (< 100 μPa·s) water near or above its critical temperature. Parisio et al. (2021) suggested that fracturing in superhot geothermal environments is possible at up to ~ 600 °C, which is the theoretical limit of brittle crust in igneous provinces. Goto et al. (2021) clarified the CFN formation process of flow-induced microfracturing, where fluid flow and the resultant stimulated zone begin in areas of relatively high permeability created by pre-existing microfractures. Microfracturing then propagates as the pressure gradient between the stimulated and unstimulated zones increases, enhancing the flow into the unstimulated zones. As the penetration of low-viscosity water into pre-existing microfractures induces the CFN, Goto et al. (2021) showed that the pore pressure required to CFN formation (i.e., fracture pressure, P p,frac , MPa) can be predicted using the Griffith failure criterion which assumes fracturing initiated from pre-existing microfractures.
Griffith failure criterion (Griffith 1924;Jaeger and Cook 1979) incorporates the tensile strength of rock, σ t (MPa), and forms a non-linear failure envelope in Mohr space (where the horizontal axis is for effective normal stress, i.e., principal stress-pore pressure, and the vertical axis is for shear stress). Fracturing is initiated when Mohr circle (representing stress state) of the rock intercept the Griffith failure envelope. The diameter of the Mohr circle is determined by the difference between the maximum principal stress, σ 1 (MPa), and the minimum principal stress, σ 3 (MPa); it determines the region of the Griffith failure envelope at which the Mohr circle would intercept. Thus, P p,frac , i.e., pore pressure required to reduce the effective normal stresses so that the Mohr circle move to the left in the Mohr space, and intercept the Griffith failure criterion, is as follows (Griffith 1924;Secor 1965;Jaeger and Cook 1979;Cox 2010): where (σ 1 -σ 3 ) are the differential stresses (MPa).
Nevertheless, studies have demonstrated that water utilization in superhot EGS is characterized by specific challenges, such as difficulties in maintaining the permeability in a water-based superhot EGS owing to the competing process between free-face dissolution and the stress solution of fracture asperities (Watanabe et al. 2020), as well as concerns over fracture clogging by amorphous silica that formed during phase changes in granite-reacted water from subcritical to supercritical or superheated steam in granite (Watanabe et al. 2021a). Moreover, water utilization for fracturing, as analyzed in shale gas extraction operations, causes highwater footprints (Wilkins et al. 2016).
Hence, studies have proposed the replacement of water with CO 2 for fracturing in EGS development because it is less reactive to rock-forming minerals (e.g., Brown 2000), can significantly reduce the water footprint, as indicated by Wilkins et al. (2016), and functions as a heat-mining fluid in EGS (Brown 2000;Pruess 2006;Randolph and Saar 2011). Moreover, CO 2 injection can achieve CFN in granite under conventional (150-300 °C) and superhot geothermal conditions ) via the same mechanism and criterion described by Watanabe et al. (2017bWatanabe et al. ( , 2019 and Goto et al. (2021), as attributed to the low CO 2 viscosity under these geothermal conditions. In addition, the lower specific heat capacity of CO 2 compared with water (Brown 2000;Pruess 2006) inhibits temperature drops in superhot geothermal environments during fracturing, where high temperatures are conducive to wellbore stability (Goto et al. 2022).
In geothermal environments with natural fractures, where induced fractures do not typically occur in conventional hydraulic stimulations using colder water (Pine and Batchelor 1984;Tester et al. 2006;Ghassemi 2012), the CFN should also occur in the unfractured rock (matrix) component based on the stimulation of pre-existing microfractures by low-viscosity CO 2 . Although fluid flow in these environments is largely controlled by the stimulated natural (1) (2) fractures (Cornet and Jianmin 1995;Evans et al. 2005), the CFN could potentially provide additional connections between the stimulated natural fractures. Thus, the CFN may contribute to solving the problems arising from the conventional stimulation of natural fractures, such as poor fracture connectivity (Batchelor 1986;Wallroth et al. 1999;Tester et al. 2006;Lu 2018). In addition, the low viscosity of CO 2 allows it to stimulate a larger volume of rock compared with water at the same injection pressure.
Natural fractures are also stimulated during CO 2 injection. The stimulation of natural fractures is commonly described using the Mohr-Coulomb (MC) failure criterion (Cox 2010). Accordingly, the total normal stress (σ n , MPa) and shear stress (τ, MPa) on a fracture are calculated as follows: (3) n = 1 + 3 2 − 1 − 3 2 cos2 r , and Fracture failure occurs as follows: where C is the cohesive strength (MPa), μ s is the coefficient of static friction of the fracture, P p is the pore pressure (MPa), and θ r is the angle between the fracture plane and σ 1 direction. At present, most laboratory-scale injection-induced shearslip experiments on igneous-rock fractures were conducted using water at up to conventional geothermal temperatures (e.g., Nemoto et al. 2008;Ye and Ghassemi 2018;Huang et al. 2019). Although injection of lower viscosity-fluid, such as CO 2 , has been assumed to produce dynamic slip on a larger area compared to water injection (e.g., Rutqvist and Rinaldi 2019), which could translate to large shear displacement, this remains to be confirmed by laboratory experiments at geothermal temperatures. Moreover, injection-induced shearing experiments at superhot geothermal temperature have not yet been conducted and are necessary to identify the potential advantages and challenges associated with the high temperature. The formation of water injection-induced complex fractures that provide additional connections between pre-existing fractures in granite has been demonstrated at ambient temperatures (e.g., Ye and Ghassemi 2019). Nonetheless, the potential formation of complex fractures by injection of lower viscosity CO 2 in granite with pre-existing fractures at geothermal temperatures is yet to be elucidated. Therefore, to clarify the possibility of CFN formation via CO 2 injection in naturally fractured granite under geothermal conditions, along with the shear-slip characteristics of the natural fractures, experiments with acoustic emissions (AE) monitoring were conducted on cylindrical granite samples, each containing a sawcut inclined at 45° from its axis. First, two experiments were conducted at 150 and 450 °C, where an increase in the differential stress was used to induce a slip in the sawcut so as to clarify the μ s value of the sawcut and slip characteristics under dry conditions at each respective temperature. These experiments also allowed for the identification of the effects of pressurized fluid and fluid type in the subsequent experiments. Two experiments involving CO 2 injection were conducted at 150 and 450 °C to demonstrate and identify the characteristics of shear slip in natural fractures and CFN formation in geothermal environments with natural fractures. Finally, an experiment involving water injection was conducted at 150 °C to compare the shear-slip characteristics of the natural fractures as induced by CO 2 and water injection.

Methods
Experiments were conducted on different 40-mm long and 30-mm diameter cylindrical Inada granite samples (Fig. 2a). Inada granite from Ibaraki Prefecture, Japan, was selected because previous studies (Yamaguchi 1970;Lin et al. 2003;Watanabe et al. 2017a) have thoroughly examined its mechanical properties. Accordingly, the porosity, permeability, and tensile strength of the granite at near atmospheric temperature and pressure are 0.5-0.8%, 2-8 × 10 -18 m 2 , and 4-9 MPa, respectively. A sawcut inclined at 45° (θ r = 45°) from the sample's axis was made in each sample using a diamond disk saw with 80-grit diamond particles, to simulate the presence of a natural fracture. Fracture surface roughness is known to affect the hydro-mechanical behavior of a fracture during injection-induced shear slip (Ye and Ghassemi 2018). To remove the effect of surface roughness, which was irrelevant in the present experiments, samples containing sawcuts with similar surface roughness were used for all experiments (Fig. 2a). A 1.5-mm diameter borehole was then drilled parallel to the sample axis on each end face of the sample, intersecting the sawcut at a depth of 10 mm (Fig. 2a). The two boreholes allowed fluid to be injected into the sawcut and fluid production from the sawcut, as well as pressure measurements on each end side of the sample.
X-ray computed tomography (CT) scans were conducted on each sample before and after the experiments to confirm shear slip in the sawcut and observe CFN formation at the voxel scale rather than obtaining an entire network image, which requires thin section observation. The CT scans were conducted under dry, room temperature, and atmospheric pressure conditions, at an X-Ray tube voltage of 120 kV, a tube current of 150 μA, and a voxel size of 30 × 30 × 30 μm.
The experiments were conducted using an experimental system ( Pramudyo et al. (2021). The system consists of a triaxial cell that uses molten plastic (polyethylene (PE) for experiments at 120-300 °C or polyether-ether ketone (PEEK) for experiments at > 350 °C) as a confining pressure fluid, a tube with a plastic rod (PE or PEEK) inside to provide a syringe pump-controlled confining pressure, two cylindrical axial stress pistons equipped with a centered (axial) fluid flow-path and pipe at which injected fluid is heated to the experimental temperature, a syringe pump to control the confining pressure, a linear variable differential transformer (LVDT) to measure axial deformation, syringe pumps for fluid injection, and an acoustic emissions (AE) measurement system to mainly monitor the AE energy during the experiment. A hand-pump operated hydraulic jack was used to induce the axial stress, in which the exerted axial force is not controlled during experiments, allowing for axial-stress drop once deformation occurs in the sample. This strategy was inspired by Ye and Ghassemi (2018) and reasonably simulated stress relaxation due to fracturing or hydroshearing in a geothermal reservoir. Two connected syringe pumps were used to accommodate a large initial volume of gaseous CO 2 , as the injection involved large-magnitude volumetric compression during the phase change from a gas to liquid at room temperature within the pump cylinders. For the AE measurement system, Physical Acoustics Corporation's twochannel data acquisition and digital signal processing PCI-2 with an R15α 150-kHz resonant frequency sensor were used. In this AE measurement system, the AE energy was calculated as the time integral of the amplitude and is presented in arbitrary units (arb. units) after the background noise was eliminated.
For the experiment, the chipped edges of the sawcut were first filled with fireproof putty, which has a low strength (thus not affecting the experimental results), so as to prevent molten-plastic intrusion into the sawcut during the experiment. The sample was then wrapped with a 10-μm-thick stainless-steel foil and then with a 50-μm-thick polyimide film that did not melt during the experiments; both acted as sleeves. A copper gasket, with two joined circular holes connecting the centered flow-path of the piston and off-centered borehole, was placed between the end faces of the sample and piston. The wrapped sample was then placed inside a plastic (PE or PEEK) cylinder within the triaxial cell and subsequently heated in an electric furnace. During the heating process, a relatively small axial stress (approximately 3 MPa) was maintained to prevent the PE or PEEK from flowing into the end faces of the sample once it melted. Above the plastic melting temperature (~ 120 °C for PE; 343 °C for PEEK), the confining pressure and axial stress were raised (carefully to prevent deformation of the sawcut) to the experimental stress state. Table 1 lists the temperature, initial stress state, and type of experiments conducted. To clarify the μ s and shear-slip Fig. 2 a Granite sample photographs and the diagram. A sample half shows an example of a sawcut surface. b Design of the injection-induced shear-slip experiment. Two connected syringe pumps were used to increase the pumping capacity during CO 2 injections characteristics of the sawcut, two experiments were first conducted at 450 °C (Experiment 1) and 150 °C (Experiment 2), both with an initial axial stress (corresponding to σ 1 ) and confining pressure (corresponding to the intermediate principal stress, σ 2 , and σ 3 ) of 55 and 30 MPa, respectively. In these experiments, the differential stress was increased stepby-step; each step was maintained for approximately 120 s. The differential stress was increased until several dynamic shear slips were observed, as indicated by the axial shortening and AE activities.
In this study, the shear displacement (d s , μm) and average shear velocity (v s , μm·s −1 ) of the sawcut slip were estimated from the axial deformation data using the following equations: where d x is the slip-related axial deformation (μm) and t is the slip duration (s). We note that the axial deformation data should include contraction or dilation of the rock matrix (several micrometers, depending on the temperature, stress state, and state of the rock matrix); hence, the d s and v s values reported in this study are approximate values. Ohnaka (1975) reported an average value of 0.59 for friction at the initiation of slip (i.e., μ s ) and 0.76 (i.e., the coefficient of dynamic friction, μ d ) at a displacement of 1.8 mm for the ground surface of the Inada granite near room temperature, indicating a slip-hardening behavior. However, an average of 0.53 for μ d at the ground surface of the Inada granite was obtained by Hirose et al. (2012), implying lower μ s values. Mitchell et al. (2013) reported an increase of ~0.04 in μ s from 150 to 450 °C in the ground surface of the Westerly granite; a similar magnitude of variation in μ s also occurred among the samples at the same temperature. Therefore, a μ s value of approximately 0.5 was expected in Experiments 1 and 2. Continuous (stable) slip of the sawcut was expected in our experiments as the differential stress increased, considering the slip-hardening behavior (Cornet 2016). Experiments with CO 2 injection were first conducted at 450 °C (Experiment 3) with initial axial stress and confining pressures of 90 and 40 MPa, respectively, to determine the possibility of CFN formation along with the shearing of the natural fractures in superhot geothermal environments with existing natural fractures. The initial stress state in this experiment was used in the experiments conducted by Watanabe et al. (2017b) and some of the experiments in Goto et al. (2021) and Pramudyo et al. (2021), thus allowing comparisons of the characteristics of the resulting CFNs. This allows for the identification of the differences (advantages and challenges) with respect to the presence of natural fractures. CO 2 injection was initiated with a 30-ml·min -1 flow rate at room temperature until 8-MPa injection pressure, continuing with a 1-ml·min -1 rate to avoid a large acceleration in the pressure increase that could lead to unnecessary AE noises during the CO 2 phase change from gas to liquid at ~ 7.3 MPa in the pump's cylinder. All fluid injections in the present study were conducted at relatively low flow rates to allow the temperature of the injected fluid to reach the experimental temperature before it arrived at the sample surface. The temperature of the injected fluid was measured by a thermocouple inside the injection pipe (TC-3, Fig. 2b). Preventing the injection of cold fluids was necessary in the present study, because previous studies have demonstrated that the injection of cold fluids could significantly cool the fracture surface and result in the reduction of effective normal stress and hence promote fracture slip (Bauer et al. 2016;Huang et al. 2019;Hu and Ghassemi 2020).
The sawcut in this study was assumed to be cohesionless (C = 0 MPa). Therefore, dynamic shear slip was expected at a P p value of approximately 15 MPa for μ s of approximately 0.5 in Experiment 3 based on Eqs. (3)-(5). A pressure gradient was expected inside the sawcut, as CO 2 flowed from the injection to the production borehole through a narrow aperture sawcut, as observed by Ye and Ghassemi (2018). In this study, to simplify the calculation, P p was assumed to be the average between the injection and production pressure values, following Ye and Ghassemi (2018). The shear slip was also predicted to cause axial stress drops, increasing P p,frac and causing a narrow aperture in the resulting CFN . This was because the axial stress was applied by a hydraulic jack. Based on Eqs. (1) and (2), P p,frac in Experiment 3 was between 42.7 and 47 MPa for a σ t value of 7 MPa and a 10-20-MPa differential-stress drop Pramudyo et al. 2021). In this study, the P p,frac values were taken from the injection borehole pressure, similar to Watanabe et al. (2017b), Watanabe et al. (2019), Goto et al. (2021), and Pramudyo et al. (2021). They considered the flow-induced microfracturing process , where CFN formation began and propagated from an area with sufficient pressure. CO 2 injection continued until the injection pressure reached 55 MPa to ensure CFN visibility in the X-ray CT observation. A pressure higher than the predicted P p,frac would increase the fracture aperture in the resulting CFN .
To determine the shear-slip characteristics of the natural fractures induced by CO 2 and water injection, as well as to clarify the possibility of CFN formation in naturally fractured rock in conventional geothermal environments, two experiments were conducted at 150 °C: one with CO 2 injection (Experiment 4) and the other with water injection (Experiment 5). Both experiments were conducted under an initial axial stress and confining pressures of 90 and 40 MPa, respectively. Experiment 4 used the same injection method as Experiment 3. In Experiment 5, water was injected at 1 ml·min -1 because no phase change was expected within the cylinder of the pump during pressurization. Dynamic shear slip was predicted to occur at P p value of approximately 15 MPa in both experiments based on a μ s value of ~ 0.5 (Eqs. 3-5). Nonetheless, differences in the slip characteristics were expected because water has higher viscosities than CO 2 at this temperature (e.g., 184.8 μPa·s versus 22.8 μPa·s for CO 2 at 10 MPa). The CFN was only expected in Experiment 4 at P p,frac values from 42.7 to 47 MPa for a σ t of 7 MPa (Eqs. 1 and 2). Figure 3 shows the time evolution of the stresses, axial deformation, and AE energy in Experiment 1. The dynamic slip of the sawcut initiated at a differential stress of 43 MPa, as indicated by large axial shortening, an AE energy peak (approximately 5 × 10 2 arb. units), and an increase in the cumulative AE energy. The dynamic slip corresponded to μ s = 0.54 (Eqs. 3-5), which was similar to the predicted value of approximately 0.5. The dynamic slip appeared to occur continuously, as indicated by the continuous AE activity, with a v s value of ~ 3.9 × 10 -1 µm·s -1 during the first second after initiation. The slip decelerated to a non-dynamic slip (e.g., v s of ~ 5 × 10 -3 µm·s -1 between an elapsed time of 650-700 s) when the differential stress was held constant in a step of differential stress. The next step of the increase in the differential stress raised the velocity again so that the acceleration and deceleration of the slip velocity occurred as the differential stress was increased stepwise. This behavior suggests an approximately 0.08-increase in μ s for each increase in the differential stress, which is also supported by the ratio of shear stress to effective normal stress on the sawcut (Fig. 3b). Slip hardening, where μ d > μ s and μ d increases as slip progresses, has been proposed as a result of an increase in the actual contact area of the fracture by slip-induced flattening of the fracture surfaces (asperity degradation) (Byerlee 1970;Biegel et al. 1992). Visual observation of the sawcut surfaces post-experimentation in this, as well as all subsequent experiments, generally revealed the presence of wear materials as evidence of slip-induced flattening of sawcut surfaces. This implies that slip behavior may be altered after several millimeters of displacement because wear materials built up within the fracture, as suggested by Byerlee (1967). Figure 4 shows the time evolution of the stresses, axial deformation, and AE energy from Experiment 2. Here, small axial shortening (~ 8 µm) with sparse and low-energy AE activity (< ~ 60 arb. units) were initially observed as the differential stress was increased to 30 and 35 MPa, possibly due to non-dynamic slip in the sawcut. The sawcut dynamic slip initiated when the differential stress reached 39 MPa, as indicated by a large axial shortening (correlating to a d s of approximately 33.3 µm) within 0.5 s (v s of ~ 66.7 µm·s -1 ) and an AE energy peak with a magnitude of two orders higher (~ 4.7 × 10 4 arb. unit) than that in Experiment 1 at the initiation, which corresponds to a µ s of 0.48. The dynamic slip decelerated to a non-dynamic slip when the differential stress was held constant at a step. The observed μ s value, although slightly lower, was in agreement with the predicted value of 0.5. The μ s value in this experiment was also slightly lower than the μ s value in Experiment 1, which was likely caused by a higher Young's modulus (i.e., lower deformability of sawcut asperities, where the stresses were concentrated), which led to a smaller contact area within the sawcut at this temperature (Mitchell et al. 2013), as compared to that at 450 °C. The same factor also appeared to cause the higher velocities and AE energies during the dynamic slip at this temperature compared with those at 450 °C. Other dynamic slips with similar characteristics to the first slip (v s of ~ 86-138.8 µm·s -1 ) were then observed when the differential stress was increased to 42 and 50 MPa, corresponding to μ s values of 0.54 and 0.68, respectively.

Shear Slip at Conventional and Superhot Geothermal Conditions
Although slip hardening was observed in Experiments 1 and 2, significant differences in the dynamic-slip characteristics were observed. In Experiment 1, the dynamic slip of the sawcut occurred continuously with velocities of ~ 10 -1 µm·s -1 . Meanwhile, in Experiment 2, dynamic slips occurred within ~ 1 s with velocities of 10 1 -10 2 µm·s -1 . Therefore, the dynamic-slip characteristics in Experiment 1 were called continuous slow slip, whereas the dynamic slips observed in Experiment 2 were called rapid slips. These results may imply that the shearing of natural fractures in superhot geothermal conditions could yield smaller displacements compared with shearing under conventional geothermal conditions, assuming the same stress state. Figure 5 shows the stresses, pressures, axial deformation, and AE energy against time during Experiment 3. In the beginning, low energy (~ 10 arb. units) AE activity and negligible axial deformation occurred, possibly due to nondynamic slips within the sawcut (Ye and Ghassemi 2020). Small increase in aperture (which were not measurable by the axial LVDT) likely occurred in the sawcut due to the pressurized CO 2 , as indicated by acceleration in the increase of production pressure and deceleration in the increase of injection pressure at an elapsed time of ~ 1700 s. The small increase in aperture resulted in a smaller differential pressure between the injection and production sides, and subsequently a more homogeneous reduction in effective normal stress on the sawcut, allowing for a rapid slip at P p = 29 MPa (time T1, Fig. 5), with a v s of ~ 24 µm·s -1 and AE activity (energy of ~ 4 × 10 2 arb. unit). The slip velocities and energies of the AE activities decreased as it progressed. Stress state on the sawcut at rapid-slip initiation corresponded to a µ s of 0.69, which was higher than expected. Nonetheless, this was due to large heterogeneity in the sawcut-effective normal stress before the rapid slip. The dynamic slip of the sawcut resulted in cumulative differential-stress drop of ~ 7 MPa at P p = 55 MPa. The changes in the permeability of the sawcut could not be determined experimentally in this study. Nonetheless, permeability could increase in natural fractures as a result of dynamic shear slip (e.g., Tester et al. 2006;Ghassemi 2012;Ye and Ghassemi 2018).

CO 2 Injection in Naturally Fractured Geothermal Environments
A continuous AE activity with an increasing energy (maximum of ~ 2 × 10 2 arb. unit) was then observed, starting at an injection pressure of 43 MPa (time T2 , Fig. 5), indicative of CFN formation. In addition, there was a slight increase in the rate of axial shortening at 43-MPa injection pressure, likely due to axial contraction in the rock matrix with the gradual formation of the CFN. Hence, 43 MPa was considered the fracture pressure, which is similar to that predicted by Eq. (1) for the corresponding differential stress, with σ t = 7 MPa. Post-experimentation X-ray CT images (Fig. 6a) confirmed shearing of the sawcut, with the amount of deformation on the same (10 2 -μm) order as that derived from the LVDT data. The CT images also revealed a CFN with narrower fracture apertures compared to those formed at 450 °C and the same initial stress state reported by Pramudyo et al. (2021). This may have been the result of decreasing differential stress during CFN formation, as demonstrated by Goto et al. (2021) and Pramudyo et al. (2021). Figure 7 shows the time evolution of the pressures, stresses, axial deformation, and AE energy in Experiment 4. As the injection pressure increased, sparse and low energy (< 50 arb. units) AE activity occurred without significant axial shortening, likely as a result of non-dynamic slips 1 3 of the sawcut. Subsequent deformation of the sawcut was characterized by multiple rapid slips (v s of ~ 36-88 µm·s -1 ), which gradually turned into a mostly continuous slow slip. The rapid slips first started at P p = 14.9 MPa (time T1), which corresponds to µ s = 0.5. Each rapid slip was accompanied by high energy (up to ~ 6.6 × 10 4 arb. units) AE activity. The continuous slow slip started at P p of ~ 40 MPa and generated continuous AE activity with energies of approximately 50 arbitrary units. Relatively small (< 10 MPa) differential pressures were observed throughout the injection, which likely caused the multiple rapid slips. Cumulative differential-stress drop of ~ 12 MPa was observed at a P p of 55 MPa due to the dynamic shear slip.
The higher Young's modulus of the rock in this Experiment 4 likely suppressed the normal deformation of sawcut asperities, causing slightly wider apertures within the sawcut throughout the injection experiment compared to those at 450 °C and leading to relatively small differential pressures most of the time. Furthermore, the relatively constant differential pressure also suggested that the CO 2 pressureinduced increase in the aperture of the sawcut was negligible. Ultimately, the amount of sawcut shear displacement in this experiment was larger than that observed in Experiment 3 within the same interval of increasing injection pressure. This implies that the injection-induced shearing of natural fractures could produce a larger shear displacement under conventional geothermal conditions than that under superhot geothermal conditions.
The CFN formed at an injection pressure of 43 MPa (time T2, Fig. 7), following the onset of an increase in the AE energy (0.5 to 3 × 10 2 arb. unit) during the continuous slip. In addition, there was an increase in the rate of axial shortening, likely due to both contraction in the rock matrix during CFN formation and the temporal acceleration of sawcut shear slip as a result of injection-production pressure homogenization. Post-experimentation X-ray CT images (Fig. 6b) confirmed CFN formation, as well as shear displacement of the sawcut with the amount of deformation on the same order as that revealed by the LVDT data. The resulting CFN had narrow apertures as a result of formation at both lower temperatures and lower differential stresses ); also, perhaps as a result of ineffective shear deformation of microfractures in the presence of the continuously sliding sawcut. Figure 8 shows the time evolution of the stresses, pressures, axial deformation, and AE energy in Experiment 5. The higher viscosity of water at this temperature (e.g., 184.8 μPa·s versus 22.8 μPa·s of CO 2 , at 10 MPa) caused a large pressure gradient within the sawcut; parts of the sawcut near the injection borehole more rapidly approached the slip condition. This situation apparently led to a continuous slow slip that started at P p of ~ 17 MPa, yielding a μ s of 0.51 [Eqs.
(3)-(5)]. However, as suggested by the results of Experiment 3, the calculated μ s value could be inaccurate in an experiment with large differential pressure. The continuous slip only generated AE activity with an energy of 20-80 arbitrary units, and caused a cumulative differential-stress drop of ~ 5 MPa as injection pressure reached 55 MPa. These results clarify that under conventional geothermal conditions, water injection was less effective at inducing shear slip, especially with a large displacement, on natural fractures compared with CO 2 injection. X-ray CT images of the sample from Experiment 5 (Fig. 6c) confirmed the shear slip of the sawcut, with an amount of deformation on the same order as that of the LVDT data; as expected, no CFN was observed.

Implications for Enhanced Geothermal System Development
The permeability enhancement and thermal extraction of an EGS in naturally fractured geothermal environments have been mainly dependent on stimulated natural fractures (Pine and Batchelor 1984;Batchelor 1986;Rinaldi et al. 2015); therefore, environments with closely spaced natural fractures are preferable (Tester et al. 2006). Nevertheless, previous studies have reported on the poor connectivity between the injection and production wells in networks of natural fractures stimulated by water injections (e.g., Wallroth et al. 1999;Tester et al. 2006;Lu 2018).
The experimental results in this study suggest that the injection of lower viscosity CO 2 into naturally fractured rock, assuming the absence of faults, could be more advantageous than water injection. Under conventional geothermal conditions, CO 2 injection was experimentally proven to produce a larger shear displacement on a larger fracture surface area, which potentially leads to a higher fracture permeability based on the association between the magnitude of the induced micro-earthquake and permeability enhancement (Ishibashi et al. 2016). Moreover, the injection of low-viscosity CO 2 may require a lower injection pressure to stimulate the same volume of reservoir compared to that when using water. Under superhot geothermal conditions, CO 2 is advantageous over water for being capable to maintain low viscosities and capable to induce large shear displacement (i.e., in the event of an injection-temperature drop or when the injection borehole must be maintained as cool as possible for engineering reasons). The formation of a CO 2 injection-induced CFN within naturally fractured geothermal environments, of which the possibility was clarified in this study, would also be beneficial at EGS sites where natural fractures are sparse; the numerous permeable microfractures within the CFN could provide additional connecting flow paths between the natural fractures and thus potentially contribute to a permeability enhancement (Ye and Ghassemi 2019). These hypotheses require clarification in future research.
Nevertheless, CO 2 injection in a naturally fractured geothermal environment presents challenges, as revealed by the experimental results of this study. First, excessive dynamic shear slip on natural fractures could also impair fracture permeability due to wear-material built up, as implied by Byerlee (1967). Second, the resulting fractures within the CFN have narrow apertures in both conventional and superhot geothermal environments, reducing their potential to provide additional connections for the natural fractures. Therefore, further stimulation, such as a longer pressurization duration  or the use of chemical stimulation to dissolve rock-forming minerals at conventional geothermal temperatures (Watanabe et al. 2021b), may improve the permeability of the CFN. Third, the low viscosity of CO 2 may also impose challenges during field-scale operation: CO 2 could flow far within the natural fracture network, especially under conventional geothermal conditions, such that increasing the pressure is difficult. These points require clarification in future studies.

Conclusions
This study clarified that CO 2 injection into granite with natural fractures under conventional and superhot geothermal conditions potentially results in the formation of a CFN, along with the shearing of the natural fractures, through a set of experiments with AE measurements on cylindrical Fig. 7 a Fluid pressures, b principal stresses, c stresses on sawcut, d axial deformation, and e AE energy (red line) and cumulative AE energy (blue line) during Experiment 4. Vertical scale for AE energy is limited to 550 to emphasize the AE activity produced by CFN formation. T1: dynamic slip, T2: initiation of CFN formation (color figure online) granite samples with sawcuts inclined 45° from their axis at 150 and 450 °C under a conventional triaxial stress state.
We first clarified that the higher deformability of the rock (owing to a lower Young's modulus) at 450 °C increased the actual contact area between the microscopically rough sawcut surfaces, leading to a slight increase in the coefficient of static friction (μ s ) and lower dynamic-slip velocities compared to those at 150 °C. CO 2 injection produced small increase in aperture of the sawcut in the experiment at 450 °C, allowing for faster slip velocities once dynamic slip progressed, as compared to that without a pressurized fluid. Nonetheless, CO 2 injection did not cause an increase in the aperture of the sawcut at 150 °C, which could be attributed to the higher Young's modulus at this temperature; instead, it caused multiple rapid slips and an even larger shear displacement compared to the dynamic shear slip without fluid injection and those in the experiments at 450 °C. This may imply that the amount of injection-induced dynamic shear slip in natural fractures under superhot geothermal conditions is more restricted than that under conventional geothermal conditions.
Based on the existing studies, experiments at 150 °C implied that CO 2 injection potentially produce a larger magnitude permeability enhancement within the stimulated natural fractures under conventional geothermal conditions; it induced a larger shear displacement over a larger fracture area, owing to lower viscosity of CO 2 under these conditions, as compared to water. Moreover, the lower viscosity of CO 2 potentially allows for the stimulation of the same volume of rock as water at lower injection pressures.
The dynamic shear slip of the sawcut caused a differential-stress drop during injection in this study, increasing the formation pressure of the CFNs, following the Griffith failure criterion. The resulting CFNs also had narrow apertures compared with those formed in intact rocks under the same or similar initial conditions owing to the decreasing differential stress during formation and perhaps the less effective shear deformation within the microfractures in the presence of pre-existing fractures. Therefore, further stimulation methods, such as a higher and longer pressurization duration or chemical stimulation in conventional geothermal environments, could be considered. In addition, low-viscosity CO 2 may flow far from the borehole in field-scale operations, such that pressurizing the reservoir becomes difficult; hence, further considerations are required.