A study on the efficiency of supercritical CO2 fracturing in hot dry rocks using coupled finite-discrete element method Authors: Zeinab Aliabadian, Mansour Sharafisafa, Mojtaba Bahaaddini Published: 2026 ABSTRACT AbstractDespite widespread use, reservoir stimulation by water hydraulic fracturing has drawbacks, including difficulties in areas with water shortages. To address challenges associated with hydraulic fracturing, supercritical carbon dioxide (ScCO2) fracturing has been proposed as a suitable alternative to address water-use issues. The main characteristic of ScCO2 is its low viscosity at supercritical conditions, which makes it particularly useful as a working fluid for stimulation and fracture generation. This study utilizes a 2D fully coupled thermo-hydro-mechanical (THM) model, developed using the combined finite-discrete element method (FDEM), as the tool for fluid injection in hot dry rock (HDR). Two different models are considered: the continuum model and fracture models with discrete fracture networks (DFN). First, the employed method is verified against some available theoretical and experimental data to validate its capabilities in this regard. Once verified, models are analyzed and three influencing and controlling factors are examined: in-situ stresses, fluid viscosity, and fluid flow rate. The results show that the performance of ScCO2 injection is highly dependent on the magnitude and difference of ground stresses. It is revealed that ScCO2 performs better in anisotropic stress fields, whereas its efficiency in isotropic stress fields is lower. It is shown that changes in ScCO2 viscosity have negligible effects on fracture development patterns and breakdown pressure. In addition, fluid flow rate plays a crucial role in both breakdown pressure and the ultimate fracture pattern. The higher the flow rate, the more fluid-driven fractures are generated, the higher the breakdown pressure, and the shorter the time to reach it. Results indicate that a higher fluid flow rate is not recommended for ScCO2 injection, as this change is less efficient at improving fluid flow. In fractured reservoirs, 12 interaction types between ScCO2 and pre-existing fractures are identified; some are newly identified, such as branching at fracture tips and rebranching, making this fluid more efficient than other working fluids. Also, comparisons are made between ScCO2 and water injection. The results indicate that ScCO2 generates many more fractures, longer fracture lengths, lower breakdown pressures, and longer post-breakdown pressure times. The longer post-breakdown pressure means more stable fluid flow in the fractures, an extended fluid-immersed zone, and thus a larger stimulated area.Keywords: Supercritical carbon dioxide; hot dry rock, Finite-discrete element method, Discrete fracture networkSubject terms: Energy science and technology, Engineering, Solid Earth sciences Introduction As one of the promising resources for renewable energy, geothermal energy is attracting more attention to overcome the limitations associated with the fuel energy market. Geothermal energy resources are capable of making a considerable contribution to the sustainable energy market and assist in reducing carbon emissions. Among the geothermal energy resources, hot, dry rocks (HDR) are a great source of energy that can be found in some specific areas and depths. HDR are dense rocks and are buried at a depth of about 3 to 10 km and have an extraordinary potential to supply heat and energy because of their abundant capacity and high temperature1–3. Their temperature can surpass 150 °C and are characterized by some features such as being safe, low carbon, clean, and efficient. Moreover, they have low permeability content and a very low amount of water or steam4. However, the power generation efficiency of an HDR geothermal resource is very low because of the limited heat exchange extent between the injecting fluid and hot rock due to the tight rock texture and considerably low permeability of the rock mass with a significantly reduced rate of heat exchange between the two mediums5,6. In order to harness the maximum capacity of these energy resources, methods to artificially stimulate these reservoirs are in great demand. This goal is achieved by high-pressure fluid injection to the HRD to generate artificial fractures to improve the conductivity of the fluid, resulting in easy extraction of heat energy from the complex fracture network via cold injecting fluid and recovery of the hot fluid, which is heated by the hot rock exposure7,8. This kind of system is called an enhanced geothermal system (EGS). Fluid injection to create artificial fractures is called hydraulic fracturing which is a common way in HDR stimulation to improve the fluid flow. Nevertheless, it remains a big challenge to enhance the heat extraction rate due to the very low connection between the injection and production walls, loss of fluid injection, and challenges such as early thermal breakthrough or short circuit of heat9,10. Furthermore, some other crucial factors play a vital role in the efficiency of fluid injection in an HDR such as ground stress conditions, distribution of natural pre-existing fractures such as discrete fracture networks (DFN), aperture of fractures, type of working fluid, etc., which impact on the hydraulic connectivity between the injecting and producing walls. Water based hydraulic fracturing has shown to impose some limitations which makes this method unsuitable for some fundamental reasons. Gradual change from conventional resources to unconventional resources has led to appear disadvantages of hydraulic fracturing. The development of unconventional reservoirs often needs high water consumption which makes it hard applicable in areas having water shortages. Furthermore, hydraulic fracturing easily causes clay minerals expansion, reducing the permeability of matrix. Moreover, considerable fracture productivity deterioration can be observed due to numerous residues in the fractures as a result of incomplete flowback. In addition, environmental pollution is inevitable since a huge volume of fracture fluid flowback with large turbidity and high volume of chemical oxygen is hard to deal with11–14. To mitigate the mentioned disadvantages, waterless or low water fracture technologies such as liquid nitrogen, hydrocarbon fracture stimulation, high energy gas stimulation, and self-popping fracture stimulation have been proposed15. However, these technologies bring some serious issues such as being flammable or explosive, ultra low temperature causing huge damage to steel injection pipes, short fracture length in high-energy gas stimulation causing preventing propped fractures, non-ideal transformation of phase state in fracturing fluid, blockage of matrix pore due to the leaking fracturing fluid, etc.16. These drawbacks have led to the emergence of novel methods, such as supercritical CO2 (ScCO2) fracturing. Carbon dioxide (CO2) as an inert gas is a nontoxic and non-flammable and is noncorrosive in its pure state, making it a suitable candidate to overcome the aforementioned problems17. From the phase state perspective of CO2 in a fracture, the fracturing of CO2 can be particularly separated into a liquid state of CO2 fracturing and ScCO2 fracturing18. ScCO2 has a dual characteristic of both gas and liquid with low viscosity and surface tension, plus large density and coefficient of diffusion19, meaning that ScCO2 fracturing can gain an improved fracture stimulation impact than liquid CO2 fracturing20. ScCO2 brings some extraordinary advantages in minimizing reservoir damage and dependency on water access21. Among these advantages, CO2 has a high energy-enhancing impact, quick and full flow back after stimulation, and flow back fluid is easily treated. Moreover, a higher coefficient of diffusion and low viscosity properties are conducive to the ScCO2 flow and are capable of reducing pressure loss, making it suitable for generating complex fracture networks. Furthermore, large filtration of ScCO2 working fluid causes lower fracture initiation pressure compared to hydraulic fracturing, resulting in less fracture difficulty. To add more, ScCO2 is capable of dissolving in crude oil to decrease viscosity, replacing adsorbed methane on the surface of rocks, or reacting with hydrogen to create methane, enhancing recovery of oil and gas in the reservoir, realizing CO2 permanent storage, and mitigating the greenhouse effects22–25. Thus, ScCO2 stimulation and fracturing are more promising in developing unconventional reservoirs compared with water-less stimulation methods. To sum up, ScCo2 has been proven to have the following extraordinary benefits as a medium for heat transfer in hot reservoirs. First, ScCO2 has a high enough diffusion coefficient, which makes its flow rate reach one to six times of water. Second, its low dynamic viscosity and lesser energy loss while flowing cause easier penetration into the rock matrix than water. This working fluid is a suitable candidate for utilization in tight reservoirs with low coefficients of permeability. Third, the thermal expansion coefficient is bigger than water. Subject to the similar injecting temperature, the change in volume is larger than water. Furthermore, it can cause a stronger effect of thermosiphon, which may decrease the loss of energy effectively while pumping stages. Many HDRs are acidic and, therefore, are less reactive while containing CO2. As a result, unprovable consequences such as blockage and scaling are less likely to happen within the flow channels. Fifth, problems such as wastewater resources or limited access to water sources do not exist, and this can indirectly gain CO2 storage in geological strata. Theoretically, ScCO2 is much more suitable as a heat-transfer fluid in EGS than water26–30. The use of ScCO₂ as a working fluid in Enhanced Geothermal Systems (EGS) has attracted significant attention due to its favorable thermophysical properties, including low viscosity, high expansivity, and enhanced heat extraction efficiency compared with conventional water-based systems. Early conceptual work by Brown31 proposed replacing water with scCO₂ in hot dry rock geothermal reservoirs, while Pruess32 demonstrated the dual benefits of geothermal energy recovery and geological CO₂ sequestration. Subsequent reviews have highlighted the potential of ScCO₂-EGS technology and identified key challenges associated with reservoir stimulation, heat transfer, and long-term system performance33,34. The injection of cold ScCO₂ into hot fractured rock induces complex thermo-hydro-mechanical (THM) interactions, including thermal contraction, pore-pressure changes, stress redistribution, and fracture propagation. Experimental and numerical investigations have shown that thermal stresses can significantly influence fracture initiation and propagation in hot dry rock reservoirs, particularly during hydraulic and scCO₂ fracturing operations35. Consequently, coupled THM modelling has become essential for evaluating reservoir behavior, wellbore stability, and permeability evolution under CO₂ injection conditions. Recent developments in fully coupled THM damage formulations have enabled the simulation of fracture growth, elastoplastic deformation, and progressive rock degradation in fractured geothermal reservoirs36–38. In particular, the THM elastoplastic damage framework proposed by Reppas et al.39 provides a robust methodology for capturing damage evolution associated with thermal loading and fluid-pressure variations, while their wellbore-scale analyses demonstrate the critical role of CO₂-induced cooling and pressure changes on wellbore integrity. These advances underscore the importance of incorporating coupled THM processes, damage evolution, and fracture mechanics into numerical models to accurately predict reservoir stimulation, fracture development, and wellbore stability in ScCO₂-based geothermal systems. Among the factors influencing the efficiency of heat extraction from HDRs is pre-existing fractures, which play a crucial role in the development of hydraulic fractures (HF) and their connectivity. The distribution of such fractures determines the HF development pattern and can impact the efficiency of fluid injection considerably. The effect of natural pre-existing fractures on the performance of fluid injection and fracturing have been the subject of many studies. Wang et al.40 studied hydraulic fracture development in HDRs containing natural fractures through simulations and concluded that the natural fracture develops continuously along the intertwined fractures, resulting in formation of multiple secondary fluid routes and widening the scale of fluid channels. Sun et al.41 investigated the effect of in-situ stresses and geometry of fracture networks on heat transfer in damaged EGS. Their results showed that system of natural fractures critical stress condition may further develop the flow channels and promote the anomality of process of heat transfer. In another study by Wang et al.42, the evolution of damage in HDRs with natural fracture systems stimulated by fluid injection was studied. They figured out that HDRs natural fractures impact significantly the pattern of fracture networks after the hydraulic fracturing process. Across the techniques used to investigate the HF development behaviour, numerical modelling has been widely employed because of their extraordinary strengths in modelling complicated mechanisms. A number of numerical techniques can be used to evaluate the mechanism of the development of the HFs under different situations and reservoir conditions. For example, the finite element method (FEM) is the most popular and utilized simulation technique. However, the FEM faces a basic problem while dealing with a discontinuity, and complicated zoning is usually needed, decreasing the modelling efficiency43. In order to comply with a discontinuity, a cohesive zone model (CZM) is an alternate to be programmed in the FEM to prevent remeshing; albeit the crack propagation path requires to be predetermined which limits the feasibility of the technique44. The CZM was then further developed to an unidentified crack path to get control of this challenge by simulating cohesive interface meshes among all the finite elements45. The extended finite element method (XFEM) has proven its extraordinary abilities in modelling discontinuities assuming a FEM simulation with no need for remeshing, causing increase the effectiveness of this technique compared to the traditional FEM. This simulation strategy has also been utilized to simulate the HF propagation in rocks46–49. Also, a finite volume method (FVM)-based phase-field model was used in50 to study CO2 fracture development in poroelastic media, showing effectiveness when dealing with a limited number of fractures. However, analyzing complicated discontinuity networks and 3D non-planar cracks by XFEM is still difficult and of little precision51. Another way is the displacement discontinuity method (DDM) which meshes the area of interest by discontinuity surface to allow for fracture propagation52. However, the DDM encounters basic problems when inhomogeneities and plastic deformation governs the behaviour of a rock mass. The discrete element method (DEM) is a robust and strong modelling technique proposed by53, which is able to simulate the propagation of cracks by employing bonded particles. Furthermore, DEM is capable of modeling fluid diffusion in discontinuities and the complicated mutual response between HFs and fluid flow in a jointed rock mass54,55. However, the computation capability of the DEM declines with the growth in the number of meshes. Therefore, the utilization of the DEM in large-scale daimonions is limited. There exist many other numerical methods; each carries some specific advantages and weaknesses while facing complicated discontinuity network development in rock masses. Thus, a robust yet powerful numerical tool needs to be selected with care considering the rock mass properties and complex interaction with the preexisting discontinuities. Each numerical method has capabilities and limitations which make them a suitable or improper tool for a particular problem. The strength of each numerical tool can be found in literature when they are employed to solve a specifics problem. The methods can be divided into two mainstreams; continuum and discontinuum methods. The strengths of the continuum methods are briefly; boundary element method: only the boundary of the domains meshed, representation of simplified data by reduced number of dimensions in space, and capable of modelling infinite domain problem. Displacement discontinuity method: introducing a single boundary condition and reducing calculation times. Finite element method: variation of property is allowed by model discretization and capable of fracture analysis in a fine scale. Finite volume method: simple, conservative and memory efficient. However, these methods impose some limitations such as complexity of the simulations when dealing with layers, edges or boundaries, inability to deal with heterogenous rocks, hydraulic fracture propagation path needs to be predefined, and introducing remeshing, cracks are only allowed to propagate through the element’s edges, and requirements for interpolation to generate compatible meshes by node splitting. On the other hand, some of the advantages of the discontinuum tools are: commutation independency for the existing or newly formed fractures, no requirement for coordination convergence and equations for displacement discontinuity, ability to preserve the features of natural fractures, capable of simulating a large number of fractures quickly, able to introduce more additional parameters impacting on fracture extension, no requirement for fracture path pre-defining, and no need for remeshing. These methods also suffer from some limitations including very large computation time to discrete elements, simplified matrix, limitation on introducing new fractures, limitation to develop to 3D space. However, the limitations are overshadowed by the strong capabilities56–59. These limitations inspire development of a method fully capable of modelling many fracture networks under various loading conditions. The aforementioned numerical tools all impose some deficiencies and exhibit obvious disadvantages when modelling complex fracture networks in terms of fracture initiation and development in rocks with complicated geological features. This challenge becomes even mor pronounced when these ground conditions are located in deep hot places with a fluid flow regime. The combined finite-discrete element method (FDEM) has been developed to overcome the deficiencies of the previous numerical tools. The FDEM primary proposed by Munjiza60 combines the strengths of the FEM for solving challenges of solid deformation and the DEM for modelling the interfaces between deforming grains. Moreover, the FDEM can model fluid flow by forming flow channels at the contacts between adjacent finite elements and these flow routes join the virtual cavities attached to the nodes of the primary finite elements. Concurrently, the FDEM simulates the crack propagation through the fracturing of embedded crack elements. The FDEM has been extensively utilized to simulate crack development in rock masses61–64. The FDEM has also exhibited extraordinary abilities in simulating fluid flow through the rock grains by modelling hydromechanical (HM) or thermo-hydro-mechanical (THM) couplings65,66. This work aims to explore the efficiency of ScCO2 injection into HDR containing DFNs by employing the combined finite-discrete element method. Fully coupled thermos-hydro-mechanical (THM) coupling is used within the framework of FDEM for a realistic simulation. First the THM login in the current FDEM algorithm is explained and related equations are described. Then the FDEM is verified against existing experimental and analytical results to validate its robustness and capabilities in modeling heat and fracture development in brittle materials. Once verified, the FDEM models will be built to implement the DFNs with various characteristics. The influence of fluid flow rate, in-situ stresses, density of the DFNs, and properties of the DFNs will be assessed to explore the fluid drive fracture patterns, fluid pressure distribution, and temperature gradient. Coupled thermo-hydro-mechanical (THM) model in FDEM In the FDEM, a simulation steps begin with an intact domain or a mix of discrete intact bodies. By stepping into the simulation, these areas (bodies) experience elastic deformation, observe translate, exhibit interaction and rotation, and undergo failure when some specific failure criteria are met, therefore forming new distinct domains or areas. The newly generated domains can then experience larger displacement and deformation, and observe more interaction, and failure. The method employed the FEM to evaluate the rock deformation and investigates the failure criterion for failure and the DEM to identify newly generated interfaces and carries out the deformation stages of the discrete domains. The FDEM benefits the continuity of the FEM and the discontinuum algorithm of the DEM. The FDEM is not only able to simulate the damage of brittle rocks before fracturing but also is capable of modelling the complicated interaction of the newly formed discontinuities. Moreover, in a HM coupling, in every hydraulic time frame, the following major mathematical steps are sequentially evaluated; identification of the volume of the cavities and their saturation degree, and also fluid flow openings and channels; employing of Darcy’s law and parallel-plate flow algorithm at every channel; employing the continuity mathematics at every cavity; and then calculation of the fluid pressure field. In the THM algorithm implemented in the FDEM, THM simulation employs the same simulation strategy used for the mechanical and hydraulic computations while forming new thermal-related variables (such as thermal energy) to the characteristic points and cavities and generating new functions to identify some heat transfer stages. This work takes advantages of the 2D FDEM in Irazu V. 5.1.0 for the fluid injection modelling67. Irazu has etraordinary abilities for a comprehensive THM coupling, which equips us to accurately simulate the HF propagation in complicated subsurface situations68. The coupled process has been successfully implemented into the FDEM to study fracture propagation under various loading schemes, for example works by Tatone and Grasselli69, Yan et al.70, and Wang et al.64. For heat transfer in rocks, thermal energy is transferred between the points of every triangular finite element. For the thermal conduction in fluids, thermal energy transformation is carried out between the virtual cavities filled with a fluid and is connected by flow cavities. As per the thermal advection, thermal energy is moved between the virtual cavities with the fluid through the bulk displacement of the fluid within the flow channels. Regarding the convective thermal transfer, thermal energy is moved between the fluid-filled cavities and the element end points of the triangular finite elements. For the contact thermal transfer, thermal energy is transferred through the fractured crack elements between the nodes of the interface pairs of finite element edges. The heat transfer stages explained here are depicted in the content of the computational algorithm of the Irazu using the illustration in Fig. 1. Illustration of the numerical framework selected for the thermal solver depicting various heat transfer steps: (a) solid thermal conduction, (b) fluids thermal conduction, (c) advection of thermal, (d) convective thermal transfer, and (e) thermal transfer between contacts58. The thermal transfer formulation in the contact of the finite element is according to the topology of (broken) cohesive elements, that are pre-implemented and will be activated once the failure criterion is achieved. Thus, it is only valid for the fracture small slip cases. In other words, only contacts of the edge-to-edge geometry among finite elements which are primarily jointed are inspected (Fig. 2a). The logic of coupled thermo-hydro-mechanical (THM) is according to a two way, explicit coupling technique achieved from changes among the three solvers whilst forward marching over time (Fig. 2b). In a modelling, the step for hydraulic is followed by the step of thermal calculation, which is then followed by enough steps by the mechanical logic to make sure quasi-static situation. Calculations of thermal and hydraulic steps are normally synchronised (meaning the smallest size of the two critical timestep is chosen for the two solvers) whilst the mechanical solver employs its size of timestep (very smaller). The coupled TM is according to a two way explicit coupling technique gained by changing between the thermal solver and mechanical solvers whereas forward marching over time in solid (Fig. 2c). In the coupling between TM, the temperature field of the solid leads to the contraction or expansion of the finite element and thermal stresses might normally appear in the calculation (depends on boundary conditions and or deformation of inhomogeneous thermal). (a) Influence of cracking on thermal transfer proces, (b) Explicit, twoway thermal hydraulic (TH) coupling58. In a coupled mechanical-thermal (M-T), the breakage of one or more cohesive elements (crack element) about a node/cavity position imposes the following influences (Fig. 2a): (1). the flow channel number participate to the cavity expansion, (2). splitting to group of original nodes (according to the newly formed topology of the cohesive elements) and the regrouping nodes to generate new groups, (3). pairing the number of finite element and flow channel within which transferring convective thermal occurs to increases, and (4). Increasing the pairs of number of finite element within which transferring contact thermal occurs. The coupling of thermohydraulic (TH) is according to a twoway explicit coupling technique, achieved by periodic change between the thermal and hydraulic solvers whilst marching forward over time (Fig. 2b). Thermal conduction in solids The thermal energy transport is calculated by considering heat conduction of Fourier’s law, a constitutive law related to the thermal flux vector, q, and relating to the temperature gradient of solid, ; for an isotropic and homogeneous solid: where is the solid thermal conductivity. Equation 1 is integrated numerically within surface of every triangular mesh (Fig. 3a) to approximate the thermal flow rate, Q̇, in every node considering a linearly varying temperature over the triangle mesh and using Gauss’ divergence theory. For example, considering a coordinate system locally (x, y), the term for the flow rate of thermal at node 0 is calculated by: Schematic of calculation framework for the calculation of processes of thermal transfer: (a) conduction of solid (via triangular finite elements), (b) conduction of fluid (via saturated flow channels), (c) advection (via saturated flow channels), (d) convective thermal transfer (among walls of the flow channel), and (e) thermal transfer at contact (among edges of the finite element)58. where TA, TB, TC denote the temperature at the three mid edge locations given by coordinate (xA, yA), (xB, yB), and (xC, yC), respectively, and A denotes the triangle surface area. Thermal conduction in fluids Thermal conduction within fractures filled with a fluid is ruled by an equation similar to the equation utilized for the conduction in solid but with the conductivity of thermal in the solid, , replaced by the fluid thermal conductivity, . The equation for transport is imposed to every flow channel (Fig. 3b) to calculate the thermal energy change within the virtual cavities: where L and a denote the length of flow channel and hydraulic aperture, respectively; and Tf,0 and Tf,1 denote the temperature of the fluid within the two cavities. Thermal advection Thermal energy movement through the carrying fluid bulk flow within a fracture is explained via the advection equation: where v denotes the velocity vector of the fluid. Employing again the theorem of divergence, integration is carried out on this equation for every flow channel (Fig. 3c) to calculate the thermal flow rate to the downstream cavity: where Tu and Td denote the temperatures of the fluid in the cavities located in upstream and downstream, respectively; and denote the density of the fluid and capacity of specific heat; Qf denotes the vector of the rate of fluid flow; and n shows the axis vector of the flow channel. Convective thermal transfer The heat transfer at the interface of solid and fluid is presumed to be dominated by Newton’s law of cooling: where hconv denotes the coefficient of convective heat transfer and represents the solid and fluid differential temperature. Convective heat transfer is considered to take place among the flow channels filled with the fluid and finite elements for sloid (Fig. 3d). Equation 6 is integrated on the flow channel walls length, L, to update the rate of thermal flow of the connected cavities and nodes. For instance, the rate of thermal flow at node 0 is calculated by: where Ts,0 and Tf,0 show the temperatures of solid and fluid at node 0 and cavity 0, respectively. Contact thermal transfer After failure of a cohesive element, the thermal transfer among contacted finite elements (Fig. 3e) is dominated by an equation analogous to Eq. 6 having the contact conductance, hcont to be replaced by coefficient hconv. This equation is integrated on the contact of the two finite element edges length, Lc, to figure out the rate of thermal flow, , at every node. Energy balance and temperature update The change of thermal energy is calculated by integration of each flow rate with an explicit temporal integration scheme: where denotes the size of the thermal timestep. The energy conservation is then imposed for updating the stored thermal energy at every node: where , and denote the net thermal energy changes of the node due to conduction of solid, convective transfer, and contact transfer, respectively. Similarly, the stored thermal energy in every cavity is updated as: where , , and represent the net thermal energy changes of cavity because of conduction of fluid, advection and convective transfer, respectively. Employing the thermal energy updated values at the end of the timestep, the temperatures of node (solid) and cavity (fluid) are updated as: where ms and mf denote the cavity masses of (solid) nodal and (fluid) and Cs and Cf show the specific heat capacities of solid and fluid. In Irazu 2D, the failure criterion for the THM-FDEM framework combines a maximum tensile stress (stress-based) criterion for mode I and a Mohr-Coulomb criterion for mode II, governed by the principles of nonlinear fracture mechanics. In Irazu 2D, fracture initiation and propagation are simulated using cohesive crack elements placed between discrete triangular finite elements. The failure criterion is defined by the stress state at the crack tip: Mode I (Tensile Fracturing). The criterion is stress-based. Fracturing occurs when the tensile stress normal to the crack element exceeds the material’s tensile strength (). Post-peak behavior: Once the strength is exceeded, the cohesive stress decreases according to a Mode I fracture energy () softening law. Mode II (Shear Fracturing): The criterion follows the Mohr-Coulomb failure criterion. The shear strength () is defined as: where \:C is cohesion, is normal stress acting across the crack element, denotes internal friction angle. Mixed-Mode (I-II): For scenarios involving combined tensile and shear loading, an elliptical coupling relationship is employed to determine the equivalent damage coefficient (\:D) for the crack element. It should be mentioned that the full formulation and description of the coupled THM implemented in Irazu (FDEM) can be found in Lisjak et al68.. Interested readers are referred to this article, and the complete formulation is not given in this manuscript to avoid an unnecessary lengthy paper. Verification To verify the FDEM capability in modeling heat transfer and crack development under temperature change conditions, two examples are verified: one with an analytical solution and another with an experimental quenching test. The analytical solution is developed for a model for heat conduction and advection in fluid-filled fractures. This example is used to verify the thermal transport via a saturated crack with a uniform aperture caused by advection and conduction of fluid. Employing a one way coupled H-T modeling with boundary conditions shown in Fig. 4, flow conditions with steady state assumption are achieved by employing a pressure gradient to the fracture. Next, the constant fluid temperatures of 50 °C and 20 °C are set to the ends of the fracture, while the adiabatic assumption is considered for the rock. The values of density of the fluid, dynamic viscosity, specific heat capacity, and thermal conductivity are assigned to 1000 kg/m3, 0.001 Pa.s, 4186 J/kg/°C, and 0.5918 W/m/°C, respectively. The cases for advection and conduction applying on the sane (model A) and opposite (model B) directions are assumed by varying the direction of the hydraulic pressure gradient. The modeled velocity for steady state flow via the 1 m long fracture of 0.001 mm width is 4.2e-7 m/s. The steady-state temperature distribution for the verification example described is given by: (a) Illustration of geometry and boundary conditions of the conduction and advection via fracture domain, (b) result of temperature distribution of fluid through the fracture (c) comparing the analytical solution and numerical simulation of the advection and conduction of fracture. where TA and TB are the temperatures at the end positions of the fracture, L denotes the fracture length, and Pe denotes the Peclet number expressed by: where ρf, Cf, and qf are the density, specific heat capacity, and thermal conductivity of the fluid, respectively; λf is the specific discharge of fluid through the fracture. Figure 4c shows a good match between the analytical solution and modeled fluid temperature profiles along the fracture. Results also depict that thermal advection acts by shifting the temperature distribution in the fluid flow direction. The second verification problem is to simulate fracture development in a hot ceramic plate quenched in water. The quenching experiments are given here to verify the FDEM capabilities further. In the experiments conducted by Jiang et al62., a ceramic plate having an initial temperature (300 °C, 350 °C, 400 °C, 500 °C, 600 °C) is quenched into cold water, and some parallel cracks are generated. The FDEM models are rectangle domains with dimensions equal to the experiments (50 mm×10 mm). the boundaries of the models are subjected to quenching via heat conduction. The assigned material properties for the model are taken from Jiang et al71. and Yan et al70. presented in Table 1. The nominal mesh size is 0.25 mm. Table 1Physical and mechanical properties of ceramic as input for the FDEM models.ParameterValueDensity, ρ (kg/m3)3980Young’s modulus, E (GPa)370Poisson’s ratio, ν0.3Friction, fr0.6Cohesion, c (MPa)200Tensile strength, ft (MPa)180Viscous damping (kg/(m.s))19,200Mode I fracture energy (J/m2)100Mode II fracture energy (J/m2)1000Fracture penalty (GPa)400Normal contact penalty (GPa.m)400Tangential contact penalty (GPa/m)400Specific heat capacity (J/kgoC)880Thermal conductivity (W/moC)31Linear thermal expansion coefficient (1/oC)7.5 × 10− 6 Physical and mechanical properties of ceramic as input for the FDEM models. The temperature distribution and crack propagation patterns of the quenching experiments are modeled. At the early stages of the quenching, the edge boundaries of the samples are cracked due to the severe induced thermal stress caused by thermal change. Subject to the effect of thermal shock, thermally induced cracks emerge and develop uniformly with almost equal spacing and are almost perpendicular to the edges. Primarily, the thermal cracks develop quickly to the inner of the samples. Then, the development speed decreases moderately with the thermal stress release. Some cracks stop propagation to a short distance due to the strain energy decline, which is unable to help all cracks to more propagation. Thus, the remaining cracks get further driving forces to facilitate further development. Figure 5 compares the crack patterns of the experiments and the FDEM simulations. From the compared images, it is evident that the obtained results of the FDEM simulations are in good agreement with the quenching tests. Composition of thermally induced fractures in the ceramic plate under different initial temperature in the (a) quenching experiments and (b) FDEM simulations. Modeling strategy The main objective of this study is to investigate the fluid-driven fracture (FDF) propagation in deep HDRs with and without preexisting natural fractures under the influence of ScCO2 injection as a working fluid. The first step is to model the ScCO2 injection in an HDR model without any pre-existing fracture network to evaluate the performance of such fluid injection in hot rocks. The influence of in-situ ground stresses, working fluid injection flow rate, and fluid viscosity is studied, and fracture development patterns and fluid pressure histories are extracted and discussed. Next, the influence of the discrete fracture network (DFN) on the ScCO2 fracturing is studied. Considered DFN density is 0.1 m/m2, with a standard deviation (m/m2) of 1e− 5. Besides, the influence of ground in-situ stresses, working fluid flow rate, ScCO2 viscosity, in both continuum, and fractured reservoirs with DFNs aperture are investigated in detail. The models are 2D rectangle domains with a dimension of 10 m×10 m and a borehole with a 20 cm diameter. The models have about 34,500 triangle elements. The first stage of the simulations is to run the thermo-mechanical (TM) FEM logic to establish the in-situ stresses and rock temperature to a specific value of 200 °C. Once these two inputs are fully established and reached the equilibrium, the thermos-hydro-mechanical (THM) FDEM is activated and ScCO2 is applied to the borehole nodes to start the fluid injection. The models are shown in Fig. 6, and full details of rock, fluid, and DFN are presented in Table 2. It should be mentioned that the ScCO2 properties in Table 2 are the values under 12 MPa pressure, and the viscosity changes will be adjusted according to pressure changes up to 32 MPa, as explained in72. Geometry of the continuum and fractured models with DFNs. Table 2Input values for the HDR, ScCO2,and DFN.ParameterValue Rock properties Density, ρ (kg/m3)2650Young’s modulus, E (GPa)60Poisson’s ratio, υ0.25Cohesion, c (MPa)20Tensile strength, ft (MPa)10Mode I Fracture energy, GIc (N/m)26Mode II Fracture energy, GIc (N/m)260Fracture penalty, Pf (GPa)10×ENormal penalty, Pn (GPa.m)10×ETangential penalty, Pt (GPa/m)10×E Fluid properties Fluid density, ρf (kg/m3)475.5Kinematic viscosity, µ (m2/s)7.00E-08Fluid bulk modulus, Kf (GPa)0.02Initial saturation degree0Specific heat capacity, Cp (J/kg°C)1800Thermal conductivity, K (W/m°C)0.03 Thermal properties of rock Specific heat capacity (J/kgoC)732Thermal conductivity (W/moC)0.2Linear thermal expansion coefficient (1/oC)6.20E-06 Parameters of microcracks Fracture penalties, Pf (GPa)10×ENormal contact penalties, Pn (GPa·m)10×ETangential contact penalties, Pt (GPa/m)10×EFriction coefficient, fr0.5Initial aperture, ai (mm)0.1Residual aperture, ar (mm)0.1Threshold aperture, at (mm)0.1 Input values for the HDR, ScCO2,and DFN. Results and Discussions ScCO2 injection in continuum reservoirs This section is devoted to studying the influence of in-situ stresses on the ScCO2 fracturing in HDR models without DFN. Six in-situ stress arrangements are considered; σH-σh = 30 MPa-30 MPa, 50 MPa-50 MPa, 70 MPa-70 MPa, 30 MPa-20 MPa, 40 MPa-30 MPa, 70 MPa-40 MPa. All models are under the flow rate of 5 kg/s. Figure 7 illustrates the zoomed failure pattern and fracture development under the mentioned ground stresses. Under the low isotropic stress state of 30 MPa, many fluid drive fractures develop in all directions by an analogous maximum length of 2.2 m. Moreover, fracture branching occurs in all directions and creates new flow channels. Increasing the ground in-situ stresses to isotropic 50 MPa state shows shorter fractures (maximum fracture length of 1.7 m), but the density of the fractures increases around the injection borehole. Under the isotopic state of 70 MPa ground stresses, the maximum length of the fractures becomes much shorter than the two other cases and extends to about 1.1 m. Although the higher ground stress levels lead to shorter final FDFs, they cause excessive fracturing near the borehole. The number of fractures increases significantly by increasing ground stress levels due to the greater resistance of the rock to fracturing against the ScCo2 injection. Therefore, the injected working fluid stops further flow and thus causes extensive fracturing. Under the anisotropic stress state of 30 MPa-20 MPa, long fluid drive cracks propagate in the direction of maximum principal stress, with some fractures deviating from this direction. ScCO2 penetrates the fractures easily and helps the ground stress generate long fractures in a specific direction. Increasing the ground stresses to higher amounts of 40 MPa-20 MPa leads to more directional fractures parallel to σH and longer fractures. However, the increased ground stresses do not always favor the long fractures, as observed for the ground with 70 MPa-40 MPa stresses. Under such a high anisotropic stress state, the fractures are inhibited to short lengths, but the accumulation of fractures becomes more excessive around the borehole due to the high directional stress orientation and higher accumulation of ScCO2 in the newly formed fractures. Zoomed ScCO2 fracturing pattern under various in-situ ground stresses. The histories of fluid pressure over injection time for the cases in Fig. 7 are plotted in Fig. 8. The graphs clearly show that the breakdown pressure increases significantly under higher isotropic stress states, and more fluid energy is required to break the borehole and initiate fractures. This trend is also true for the anisotropic stress states. When the difference between σH and σh increases, higher fluid energy is needed to break the rock and generate FDFs. It is also observed that the breakdown pressure for isotropic stress states is much higher than that of anisotropic ones because the large difference between the ground stresses helps higher stress concentration, lowering the need for fluid energy. Borehole fluid pressure histories subject to different ground in-situ stress magnitudes under 5 kg/s flow rate. It is known that fluid viscosity plays an important role in hydraulic fracturing or any other fluid fracturing. Some studies assessed this factor and realized its role in the length and width of hydraulic fractures and also the breakdown pressure, such as Wang et al33., Wrobel et al64., and Shimizu et al65.. The existing studies focus on the viscosity of water as the working fluid, and outcomes indicate the strong influence of change in fluid viscosity on the hydraulic fracture pattern. However, to the best of the authors’ knowledge, there is no systematic study to evaluate the influence of ScCO2 viscosity on fluid fracturing. In this section, we study the influence of ScCO2viscosity change on the FDF pattern and breakdown pressure. Seven viscosities are selected from Shen et al63., and a change in density and pressure changes the viscosity. Table 3 presents the input parameters for this section. Figure 9 illustrates the FDFs under the seven viscosities. It should be noted that in this figure, the ground stresses are considered isotropic 30 MPa, and the injection flow rate is 5 kg/s. All other input parameters are the same as those in the previous section. The fracture paths in Fig. 9 exhibit minimal difference between the models and demonstrate that in the case of ScCO2 injection as the working fluid, change in viscosity, which is a function of change in density and pressure, does not significantly impact the pattern and length of fractures. This phenomenon is mainly because the viscosity of ScCO2 is much lower compared to other working fluids, such as water and is in a supercritical state under critical pressure and temperature, and a change in viscosity does not change fracture propagation pattern. This negligible change is also reflected in the fluid pressure graph, and breakdown pressure shown in Fig. 10. As can be observed in Fig. 10, the overall shape of the graphs is the same but with minimal differences. The difference is in the breakdown pressures, which rise from 84.4 MPa for the lowest viscosity to 87.5 MPa for the highest viscosity. Moreover, decreased viscosity increases the time needed to reach the breakdown pressure. Nevertheless, the changes are not pronounced in this case, and it can be said that the viscosity change in a ScCO2 injection does not considerably influence the fracturing pattern. Table 3Input values for models with change in ScCo2 viscosity63.Pressure (MPa)Density (kg/m3)Viscosity (10− 6 Pa.s)8198.520.1012475.534.8316664.952.5620743.562.6924792.170.1628828.076.4332856.581.95 Input values for models with change in ScCo2 viscosity63. ScCO2 fracturing pattern under various fluid viscosities, varying with pressure and density. Borehole fluid pressure histories subject to different fluid viscosities. Another influencing factor on fracturing under fluid injection is the rate of fluid injection, which has been demonstrated to play a crucial role in a successful injection. This controlling factor has been studied experimentally and numerically by some researchers, including experiments conducted by Zhuang et al66. and Cheng et al67. and simulations by Liang et al68.. These existing studies prove the dominating role of fluid flow rate on the pattern and breakdown pressure of hydraulic fracturing. However, the current studies focus on water as a working fluid, which is fundamentally different from a supercritical fluid. Therefore, in this section, we aim to explore the influence of the ScCO2 flow rate on FDF patterns. Figure 11 depicts the fracture pattern under a flow rate of 10 kg/s. This figure should be compared with Fig. 7 (cases with 5 kg/s flow rate) for better understanding. Under the isotropic stress fields, the number of fractures has increased considerably in all directions because the high injection rate increases the number of fractures initiated from the borehole, facilitating fluid penetration into the primary fractures and expanding them. However, the length of the FDFs does not experience significant perceptible compared to Fig. 7. This negligible change is due to the rapid increase of the fluid pressure around the borehole, which causes excessive fracturing, thus making the fluid circulate hard into the fracture and avoiding them to propagate longer. In this situation, fluid leak-off occurs frequently, and the efficiency of fluid fracturing decreases, which increases the cost of fluid injection. However, in the state of anisotropic stress, the story is different. Subject to anisotropic stress fields (regardless of the magnitude difference), the length and number of the fluid-drive fractures increase meaningfully. It is also observed that the effect of the high anisotropic stress differences becomes weaker and the fractures tend to propagate in the direction of σh compared to Fig. 7 where 5 kg/s injection rate directs the fractures parallel to the direction of σH. It is also observed under the higher flow rate and larger in-situ stresses difference (70 MPa-40 MPa), the severity of the fractures around the borehole reduces, resulting in less leak-off and better fluid diffuse. This observation demonstrates that ScCO2 working fluid injection in anisotropic stress fields works better and more efficiently than in isotropic stress fields. The performance of a ScCO2 in anisotropic stress fields facilitates the generation of more and favorably directed new fluid channels which helps smoother flow circulation, better heat exchange between the hot dry rocks and cold fluid, and more effective fracture connection. Therefore, ground in-situ stress measurements should be carefully assessed before any fluid injection operation design. ScCO2 fracturing pattern under the high 10 kg/s injection flow rate and different in-situ ground stresses. Figure 12 shows the failure modes of the fracturing patterns shown in Fig. 11. Red colors indicate tensile failure while blue colors represent the mixed mode (shear-tensile fracturing). When the stress difference (the contrast between the maximum and minimum in-situ principal stresses) is too large, shear failure becomes the dominant mechanism over tensile failure in hydraulic fracturing. This shift occurs because a high differential stress creates a geological environment in which the rock’s effective stress state favors slip along planes of weakness. The increased stress difference can cause two primary effects: activation of natural fractures, and large stress contrasts promoting shear slip along existing natural fractures or bedding planes rather than opening new tensile fractures against the minimum principal stress. This is particularly important in naturally fractured reservoirs. Direct shear fracturing of the rock matrix: even in intact rock, a sufficiently high stress difference can cause the rock matrix to fail in shear directly. Experimental studies show that when the differential stress exceeds a certain threshold (relative to confining pressure), hydraulic fluid diffusion reduces the effective confining pressure until the rock fails by shear fracture instead of a pure tensile crack. While hydraulic fracturing is traditionally thought to create tensile fractures perpendicular to the minimum stress, this understanding is based on ideal, isotropic conditions. In reality, particularly in tectonically active regions or deep reservoirs, the high stress anisotropy means that shear failure is not just possible, but often more likely to occur, making it a critical consideration for predicting fracture complexity and reservoir stimulation effectiveness. Failure modes of ScCO2 fracturing pattern under the high 10 kg/s injection flow rate and different in-situ ground stresses. The fluid pressure graphs over injection time are plotted in Fig. 13 and compared with the injection rate of 5 kg/s. Two notable observations can be figured out in Fig. 13. First is the breakdown pressure difference between the injection rates. Overall, the breakdown pressure increases by about 4 MPa for each case. The second important finding is the breakdown time. Under the higher injection flow rate of 10 kg/s, the breakdown pressure and, thus fracture initiation time reduces significantly. For example, in the case of 70 MPa-70 MPa, the breakdown pressure time decreased from 5.19 s to 3.16 s, which implies that less fluid is required for fracture initiation. This advantage coexists with the disadvantages of high flow rate in isotropic stress fields, therefore, all scenarios need to be taken into account when designing fluid injection operation by a supercritical fluid. Borehole fluid pressure histories under two injection flow rates subject to different ground in-situ stress magnitudes. ScCO2 injection in fractured reservoirs Having understood the behavior of continuum reservors subject to ScCO2 working fluid injection, this section is devoted to studying the performance of injection of such working fluid in fractured reservors with DFNs. The dimensions, properties, ground stresses, and controlling factors are identical to the previous section. DFN properties are listed in Table 2 and the following functions are enabled in the FDEM simulation for realistic simulations: flow through intact DFN, thermal conduction through fluid inside fractures, thermal advection, impermeable matrix and permeable fractures, and convective heat transfer allowing for thermal exchange between fluid in fracture and solid. Moreover, fluid pressure change is enabled in response to the mechanical deformation of fractures. These assumptions would produce simulations involving fluid flow in a hot rock as precisely as possible. Before we begin analyzing the fractured models, identifying the interaction between ScCO2 working fluid and DFN is worth identifying. Figure 14 presents the identified interaction types in a model with the DFN sets. At the center of Fig. 14 is a model’s ultimate fracturing pattern, and the zoom-in images are captured from different corners of the central image, shown by arrows. Twelve interaction types were identified by closely viewing the fracture propagation within the model. The first interaction type is a direct connection, which is fluid flow through one fracture and initiating an FDF from its tip and developing to an adjacent tip. The second is rebranching which is due to high fluid pressure dispersion into a new fracture and leading to the development of new fractures. Third, is offset crossing because of rapid fluid flow through an existing long fracture and initiating new fractures from the surface of the fracture. Low viscous fluid initiates some new fractures before reaching the tips of the long fracture. The fourth is due to the interaction of the cold, low viscous fluid with the hot rock, causing rapid heat exchange and generating many short thermal shrinkage fractures. The fifth one shows the quick crossing of a fluid fracture from an existing one without activating it. The sixth type arrests the FDF and stops further propagation when this new fracture meets an existing fracture. The seventh type is a hybrid one, combining activation of a pre-existing fracture and making new offset fractures. The eighth one is also a combined one, made up of both crossing and activating. The ninth one represents dilation, fluid dispersion into a pre-existing fracture without making new fractures from its surface. The tenth type is an interesting, notable one; the high-pressure low viscous ScCO2 branches the FDF into two from the tip of a pre-existing fracture. The eleventh type illustrates a complex one, comprising dilating the pre-existing fracture, making offset fractures, and generating FDF surface branching. The twelfth interaction type shows high fluid dispersion into the pre-existing fracture, which makes combined interaction made up of dilation, offset crossing, and activation. These interaction types will help to better analyze the FDF development in fractured models in the following section. Identified interaction types between ScCO2 and DFN; (1) direct connection, (2) rebranching, (3) offset crossing, (4) thermal branching, (5) crossing, (6) arresting, (7) activation+offset crossing, (8) activation+ crossing, (9) dilation, (10) crack tip branching, (11) dilation+offset crossing+surface branching, (12) dilation+offset crossing+activation. Figure 15 shows the fracture propagation pattern in the fractured models under different ground in-situ stresses. Under the isotropic stress field, the extent of the FDFs is considerably influenced by the magnitude of the stresses. Under the low 30 MPa isotropic stress field, FDFs propagate in all directions, hit the DFNs, and interact with the existing fractures to make new fluid channels by generating the abovementioned interaction types. Since the ground stress state is relatively low, ScCO2 can flow easily inside the DFNs and dilate them, causing large DFN openings and better fluid dispersion. This creates a better environment for generating new FDFs and thus longer fluid immersion and longer FDFs. When the magnitude of the isotopic stresses increases, not only does the extent of high fluid pressure and FDFs decrease significantly, but the interaction types also change. By stress increase, offset and activation types decrease, arresting increases notably, and thermal shrinkage cracks show remarkable growth. This behavior occurs because the large ground stresses keep the DFNs close and do not allow for easy fluid flow inside the DFNs, resulting in difficult fluid dispersion inside the pre-existing fractures. Therefore, continuous fluid injection and interaction with the HDR in the close vicinity of the borehole cause excessive thermal cracking and lower the efficiency of fluid injection. The FDF behavior shows notable differences when the ground stresses are anisotropic. Under such ground stresses, the extent of the FDFs is much larger than that of isotropic ones. Under any anisotropic stresses, the fractures’ extent propagates toward the models’ boundaries, regardless of the magnitude of σH and σh. When the horizontal stress increases, the FDFs tend to propagate in the direction of σH and are reluctant to develop in other directions. Furthermore, the interaction types vary by change in the difference between σH and σh. Under the low σH−σh difference, the DFNs can open and dilate easily, facilitating the fluid dispersion into them and, thus, frequent dilation, offset crossing, and activation to occur. However, the increased difference between σH−σh results in more arresting and shrinkage cracking since the high horizontal stress prevents proper DFN opening and dilation, and fluid flow barely occurs.Figure 16 compares the fluid pressure histories of the model in Fig. 15. Similar to the fluid pressure histories of intact domains, the ground stresses highly influence the breakdown pressure, the higher the magnitude of ground stresses is, the larger the breakdown pressure becomes since more pressure is needed to overcome high in-situ stresses and rock breakage. This observation is somehow true for anisotropic stress fields; however, the higher the σH-σh difference becomes, the lower the breakdown pressure. This might be attributed to the effect of the DFNs, which facilitate fluid flow in certain directions by increasing the difference between the ground stresses. ScCO2 fracturing pattern under various in-situ ground stresses in a reservoir with DFNs. Borehole fluid pressure histories subject to different ground in-situ stress magnitudes under 5 kg/s flow rate in fractured reservoirs. In the previous section, we studied the influence of ScCO2 viscosity change on the FDF propagation in continuum reservoirs and observed insignificant changes in the ultimate fracturing pattern. Here, we examine the effect of this parameter change on the fracture pattern of fractured reservoirs, and the results are shown in Fig. 17. As can be clearly observed in this figure, the change in ScCO2 viscosity does not make a tangible impact on the ultimate fracturing patter. In all models with change in the viscosity, the extent of FDFs is almost the same, interaction types are similar, and fluid-immersed and pressurized domains are analogous. This, along with the observation for the continuum models in the previous section, confirms the inconsiderable impact of ScCO2 viscosity change on the performance of fluid injection in such reservoirs. ScCO2 fracturing pattern under various fluid viscosities, varying with pressure and density in fractured reservoirs. Figure 18 is the comparison of the fluid pressures and breakdown pressure for the domains under ScCO2 viscosity change. As expected from Fig. 16 above, the breakdown pressures do not exhibit a notable change by viscosity change. There is a minimal change in breakdown pressures, decreasing from 88.4 MPa for the highest viscosity to 86.4 MPa for the lowest viscosity. The only obvious difference is the breakdown pressure time, increasing from 3.6 min for the highest viscosity to 5.1 min for the lowest. This implies that the higher ScCO2 viscosity can reduce the costs by quicker borehole breakdown and fracture initiation. Borehole fluid pressure histories subject to different fluid viscosities in reservoirs with DFN. The fractured models are also assessed under the higher flow rate of 10 kg/s to see how they react to the higher fluid pumping. Figure 19 depicts the FDF development pattern under the higher flow rate in the reservoirs with DFNs. To make a reasonable investigation, Fig. 19 should be compared with Fig. 15. Once the fluid flow doubles under the isotropic stress fields, two noticeable differences can be observed; first is the extent of fluid pressure which becomes larger, the FDFs grow longer, and fluid immersed zone is wider. Furthermore, the number of FDFs increases meaningfully, particularly when the magnitude of the ground stresses increases. This is because the high ground stresses avoid opening and dilation of the DFNs, while the high fluid flow injection infiltrates the narrow DFNs at the close vicinity of the borehole, fluid pressure makes up excessively and generates numerous FDFs in a narrow area. This behaviour causes the unwanted leak and excessive fracturing around the borehole, preventing longer fracture propagation. Under the anisotropic stress states, the fracture propagation pattern is quite similar to the 5 kg/s injection rate in Fig. 14, in which the FDFs extend to longer distances parallel to the direction of σH, with some minor differences, such as more fracture deviation from the direction of σH, more branching, activation, and shrinkage cracking. Therefore, based on these observations, a high injection flow rate is not recommended for ScCO2 injection. ScCO2 fracturing pattern under the high 10 kg/s injection flow rate and different in-situ ground stresses in fractured reservoirs. Figure 20 compares the fluid pressure histories of the DFN models under both 5 kg/s and 10 kg/s flow rates. The difference between the breakdown pressures is not significant, and only about a 3 MPa increase occurs when the flow rate is doubled. However, the main difference is the breakdown pressure time in which the breakdown time for the higher flow rate is much less than that of 5 kg/s, meaning that fracture initiation occurs rapidly compared to the lower flow rate. This is an advantage, but it should be considered by other factors, such as the magnitude and direction of in-situ stresses, rock temperature, fluid viscosity, etc., for an optimal and efficient fluid injection operation. Borehole fluid pressure histories under two injection flow rates subject to different ground in-situ stress magnitudes in fractured reservoirs. Discussion The previous results showed the performance of ScCO2in HDR under different scenarios. It is worth comparing the results of a supercritical fluid injection with water as a working fluid. To begin with, a model is built to model the lab tests carried out by Yang et al73. that ScCO2 injection was done within a borehole inside a rock specimen. In their test, a granite specimen with dimensions of 50 mm by 50 mm by 85 mm was fabricated. Further details including sample fabrication, setup of the tests, injecting fluid, in-situ stress setting, adjusting the temperature, etc. may be read in their study. The properties assigned to the material in this study are taken as a density of 2595 kg/m³, uniaxial compressive strength of 240 MPa, and uniaxial tensile strength of 11.71 MPa. However, these parameters are insufficient for FDEM modeling; some other input parameters must be estimated based on the available empirical and experimental equations. The estimated values are E = 60 GPa, ν = 0.26, cohesion = 25 MPa, Mode I fracture energy 1300 µN/mm, and Mode II fracture energy 13,000 µN/mm. Other input parameters are the same as those listed in Table 2. Many endeavors have been made for an accurate estimation of the parameter according to many trial and error runs, and the provided simulation inputs illustrate the true granite sample with a large matching range. The following are the input data from the experiments. The reported experimental temperature was the ScCO₂ fracturing temperature, 40 °C, and the hydraulic fracturing (water) temperature, room temperature (≈ 20–25 °C). Since the critical temperature of CO₂ is 31.1 °C, the selected temperature ensured that the injected CO₂ remained supercritical throughout the experiment. The shale specimens were loaded in a true-triaxial loading system. The study investigated different horizontal stress regimes to evaluate fracture propagation. The applied principal stresses were: case, vertical stress (σv), maximum horizontal stress (σH), and minimum horizontal stress (σh) were equal at 20 MPa, 20 MPa, and 20 MPa. Moderate stress differences of 20 MPa, 25 MPa, and 15 MPa. Large stress differences of 20 MPa, 30 MPa, and 10 MPa. These boundary conditions were used to examine the effect of horizontal stress contrast on fracture geometry and fracture-network complexity. The injection was performed using an ISCO syringe pump at a constant flow rate (displacement-controlled injection). The reported injection rate was 0.5 mL/s, equivalent to 30 mL/min or 1.8 L/h. The pump continuously injected fluid until fracture initiation (breakdown) occurred. These conditions are frequently used as benchmark laboratory conditions for comparing water fracturing and ScCO₂ fracturing in shale, particularly for studying fracture initiation pressure, acoustic emission activity, and fracture-network morphology. Mesh and time-step sensitivity analyses demonstrated numerical convergence of the FDEM model. An average element size of approximately one-tenth of the borehole diameter and a time step corresponding to 20–50% of the critical CFL limit were sufficient to accurately capture breakdown pressure, fracture propagation, and fracture-network complexity during supercritical CO₂ fracturing of hot dry rock. Time-step sensitivity analyses were performed using time steps ranging from s to s. The breakdown pressure and fracture length differed by less than 2% and 4%, respectively, between the two smallest time steps. Therefore, a time step of s was adopted for all simulations. The thermophysical properties of ScCO₂ are highly dependent on both pressure and temperature, particularly near its critical point (31.1 °C and 7.38 MPa). As pressure increases at constant temperature, the density, viscosity, and bulk modulus of ScCO₂ generally increase, while its compressibility decreases as the fluid becomes more liquid-like. Conversely, increasing temperature at constant pressure typically reduces density and the bulk modulus, but slightly increases viscosity and thermal expansion. Near the critical point, even small changes in pressure or temperature can cause substantial variations in density and compressibility, making ScCO₂ highly responsive to reservoir conditions. In geothermal and hot dry rock (HDR) reservoirs, where temperatures commonly range from 100 to 250 °C and pressures from 20 to 40 MPa, ScCO₂ exhibits densities of approximately 400–900 kg/m³ and viscosities of only 0.04–0.12 mPa·s. These characteristics provide ScCO₂ with excellent penetration into microfractures and natural discontinuities, promote rapid pressure diffusion, and enhance fracture complexity compared with conventional water-based fracturing fluids. Consequently, accurate numerical simulations of ScCO₂ fracturing should account for pressure- and temperature-dependent fluid properties using appropriate equations of state rather than assuming constant fluid parameters. Figure 21 makes a comparison between the outcome of injecting pressure in time between the tests and FDEM model simulations subject to both ScCO2 and water-based fluid, under different temperature values, ranging from temperature of room (RT) to 300 °C with a 100 °C gap, subject to σh = 60 MPa. For a better comparison, the rate of injection is assigned to 5 ml/min for ScCO2 and water. Despite the injection rate being the same, the pressure of breakdown and time varies. For instance, when the sample’s temperature equals room temperature (RT), the needed time for breaking pressure subject to water-based fluid injection is around 65 s, whereas the required time for ScCO2 fluid is around 210 s. This outcome is observed for all elevated temperatures. This observation happens because of the fact that the density of water-based fluid is larger than of ScCO2, and water-based fluid has very high viscosity than SCCO2, in turn rapidly increases pressure of fluid at the injection, leading to quick fracture development. The difference between the parameter also causes the fracturing pressure of water-based fluid too bigger than ScCO2, resulting in quick fracturing. The general outcomes of the tests and FDEM models fall in the reasonable range, but not exact values. The differences may arise due to some reasons, such as inaccuracy in input value estimations, distribution of various-sized grains in the granite specimen in the tests, the two-dimensional assumption of models against the three-dimensional nature of tests, etc. Another highlighted difference between water-based fluids and ScCO2 is the development pattern of FDF. Graphs of injection pressure over injection time for granite specimens under water and ScCO2 injection experimentally69 and compared with FDEM simulations. Figure 22 depicts the comparison of the developed fracturing patterns in the sample under assumptions of σh = 20 MPa and temperature of 300 °C, resulting from the tests and simulated by FDEM, for both water-based fluid and ScCO2 fluid. Under the water-based fluid injection, one long FDF propagates with no branching. In this sample, the performance of fluid fracturing is questionable, and pivotal parameters should be reassessed and adjusted to guarantee a successful operation. This fluid-injection strategy may results in partial fracturing of rock, increasing operation costs. On the other hand, injecting ScCO2 shows a notable outcome. Subject to the same assumptions, ScCO2 fluid fracturing creates many fracture branches from a single initial FDF, generating new channels for fluid flow and for efficient fluid fracturing. In this scenario, the low-viscous fluid penetrates into the rock rapidly, fractures it to generate FDFs, permeates the newly generated FDFs, and causes branching. Thus, ScCO2 is a superior alternative to water-based fluids for stimulation of the HDR, yielding an efficient and reliable outcome. (a) experimental hydraulic fracturing, (b) experimental ScCO2 fracturing, (c) FDEM hydraulic fracturing, (d) FDEM ScCO2 fracturing. Figure 23 is the results of water and ScCO2 injection in a model with a single fracture positioned 1 m away from the center of the borehole. In-situ stresses are anisotropic 30 MPa-50 MPa to direct the FDF toward the pre-existing fracture. 5 kg/s is selected as the input flow rate and all inputs are the same as Table 2. In Fig. 23a, water performance is shown, and a single FDF propagates toward the natural fracture; water permeates partially into it, dilates it, and finally crosses. In contrast, ScCO2 fracturing is shown in Fig. 23b. as shown, a single FDF propagates toward the natural fracture, permits into it, fully dilates it, generates offset crossing fracture (marked 1), activates at both tips (marked 2), and rebranches at the intersection with the existing fracture (marked 3). These observations further demonstrate the extraordinary performance of ScCO2 as the working fluid in HDRs. Interaction of a single closed joint with the working fluid, (a) water injection, (b) ScCO2 injection. To further discuss the differences, models with the same DFN, as shown in Fig. 6 with 30 MPa-30 MPa, 40 MPa-20 MPa, and 70 MPa-40 MPa ground stresses, are injected with water and ScCO2 at two flow rates of 5 kg/s and 10 kg/s, and the results are illustrated in Fig. 24. Subject to in-situ isotopic 30 MPa-30 MPa stresses and for 5 kg/s BD 10 KG/S flow rateS, compared to ScCO2 fracturing, hydraulic fractures are significantly shorter, their number is considerably lower, and interaction types are limited. A few DFNs are activated underwater injection, and the fluid’s ability to flow further is restricted due to the large water viscosity. Another observation is that with water injection, the number of fractures initiated from the borehole is almost double the ScCO2 injection. This confirms the very high fluid pressure water imposes on the borehole surface. When the in-situ stresses are anisotropic, the FDF pattern difference becomes even more distinct. By ScCO2 injection, FDFs grow to a long distance and interact more with DFNs, making new fluid channels frequently. However, water injection, as the working fluid, exhibits short FDFs, much fewer fractures, and thus very low interaction with DFNs, making the success of fluid injection unlikely. Comparison of water and ScCO2 injection performance under various conditions. Figure 25 compares the fluid pressure histories of the cases studied in Fig. 24. Water histories show pronounced time reduction for breakdown pressure, meaning that fluid begins fracturing much earlier than ScCO2 and FDFs start to appear quickly. Furthermore, the breakdown pressure for water injections is much larger than that of ScCO2, leading to rapid fracturing and more fractures to initiate from the borehole. It is also observed that the post-breakdown histories of ScCO2 are gentler than that of water and the graph drop is more gradual than that of water, which shows abrupt drops. This behavior indicates that ScCO2 pressure remains stable for longer in the fractures, resulting in more FDFs and longer DFN fluid immersion. Fluid pressure histories for cases with water and ScCO2 injection under various situations. In hydraulic fracturing, thermal parameters influence fracturing performance through thermally induced tensile stresses, but their role differs fundamentally between water and supercritical CO₂ (ScCO₂) due to differences in fluid viscosity, compressibility, and heat transfer efficiency. For water fracturing, thermal parameters play a secondary, assisting role: the injected water’s lower temperature relative to the hot reservoir creates a moderate thermal gradient, inducing tensile thermal stresses that generate a narrow zone of micro-cracks primarily near the wellbore wall, which enhances near-wellbore permeability and aids fracture initiation, but because water’s high viscosity and low compressibility allow it to efficiently transmit hydraulic pressure deep into the fracture, the primary fracture propagation remains dominated by mechanical pressurization rather than thermal effects, with the thermal stress contribution being significant only for reducing breakdown pressure modestly and creating minor secondary cracks. In contrast, for ScCO₂ fracturing, thermal parameters become a primary and synergistic fracturing mechanism: ScCO₂’s extremely low viscosity (similar to a gas) and high diffusivity allow it to penetrate instantly into microscopic pores and existing micro-cracks throughout the rock matrix, while its relatively high specific heat capacity and thermal conductivity enable rapid heat extraction from the rock upon injection, creating a large temperature difference (ΔT) between the cold ScCO₂ and the hot reservoir; this large ΔT generates intense, uniformly distributed tensile thermal stresses across a wide volume of rock, significantly lowering the rock’s effective tensile strength and reducing breakdown pressure by up to 50–60% compared to water, while also promoting the initiation of multiple fractures in various directions rather than a single planar fracture, ultimately creating a complex, multi-branched, highly tortuous fracture network with greater stimulated reservoir volume and higher fracture conductivity, especially in high-temperature reservoirs where the thermal shock effect is most pronounced. Figure 26 compares failure modes of the rock domains discussed in Fig. 24. Red colors indicate tensile failure, while blue colors represent the mixed mode (shear-tensile fracturing), and black lines are DFNs. The figure presents a direct side-by-side comparison of induced fracture networks and associated failure modes in rock under identical conditions. ScCO₂ (right column) induces significantly more mixed-mode (shear-tensile, blue) fracturing and greater overall complexity compared to water across all tested conditions. Water tends to favor more planar, tensile-dominated propagation, especially under higher stress differentials. 30 MPa-30 MPa under 5 kg/s (low stress anisotropy) water (left): predominantly tensile (more red/less branching) with moderate mixed-mode branching near the injection point. ScCO₂ (right): much higher proportion of mixed-mode shear-tensile (blue) fractures. Extensive branching and interaction with pre-existing DFN (black lines). ScCO₂ activates shear slippage more effectively.30 MPa-30 MPa under 10 kg/s (higher injection rate) water (left): limited increase in complexity; still mostly tensile with some mixed-mode zones. Fracture network remains relatively simple. ScCO₂ (right): Dense network dominated by mixed-mode failures. Higher rate enhances shear-tensile interactions dramatically. clear advantage for ScCO₂ in promoting shear. Third row 40 MPa–20 MPa under 5 kg/s (moderate stress contrast) water (left): fractures align strongly with the maximum stress direction; predominantly tensile with minimal branching (less mixed-mode). ScCO₂ (right): retains substantial mixed-mode (blue) components despite stress anisotropy. More torturous paths and DFN interactions. ScCO₂ better resists stress-controlled planar growth. 70 MPa-40 MPa under 5 kg/s (high confining pressure plus large differential) water (left): Strongly tensile-dominated, forming long planar/elongated fractures aligned with principal stress with very limited shear-tensile branching. ScCO₂ (right): more mixed shear-tensile failures, with noticeable off-plane branching and complexity even under high pressure. ScCO₂ maintains shear activation capability. In summary, tensile failure is more dominant in water injection, especially under higher differential stresses. Leads to simpler, more predictable, planar fractures. Mixed-mode shear-tensile is far more prevalent in ScCO₂ across all scenarios. This results in complex, branched, and interconnected networks with greater interaction with pre-existing fractures (black DFN lines). The overall trend is that ScCO₂ promotes greater shear involvement, leading to higher fracture complexity and a larger stimulated volume. Water behaves more classically (tensile-dominated hydraulic fracturing). This behavior is consistent with literature on ScCO₂ fracturing: its lower viscosity, higher diffusivity, and different pressure transmission allow easier penetration into microcracks and activation of shear along weaknesses/DFN, whereas water tends to create dominant tensile cracks. Comparison of fracturing modes of water and ScCO2 injection performance under various conditions. ScCO2 is superior to water for hydraulic fracturing primarily because it solves water’s three major problems: formation damage, water blocking, and high environmental impact. While water has higher viscosity for proppant transport, ScCO2’s extremely low viscosity (0.03–0.10 mPa·s) enables it to infiltrate microcracks and natural fractures, creating complex fracture networks that dramatically increase stimulated reservoir volume while reducing breakdown pressure by over 50%, lowering pumping energy and seismic risk. Unlike water, ScCO2 is non-polar and does not cause clay swelling or fines migration, preserving natural permeability in clay-rich shales. It also eliminates water blocking because it vaporizes upon pressure release, allowing immediate hydrocarbon flow without the weeks of cleanup required with other methods. Furthermore, ScCO2 actively enhances production by dissolving organic matter and reducing oil viscosity (enhanced oil recovery), with no freshwater or toxic additives required. The only drawback- poor proppant transport due to low viscosity-is being solved with thickened ScCO2 formulations. Thus, water merely fractures and damages, whereas ScCO2 fractures, cleans, and enhances flow with a dramatically lower environmental footprint. Conclusion This study focuses on the performance of supercritical carbon dioxide (ScCO2) as the working fluid for stimulation of hot dry rocks (HRD) analyzed via a fully coupled thermo-hydro-mechanical (THM) scheme implemented into the 2D combined finite-discrete element method (FDEM). Two scenarios are considered: continuum reservoirs without pre-existing fractures and reservoirs with discrete fracture networks (DFN). First, a series of verification examples are presented to ensure the method’s capability in modeling fluid flow in a solid medium. Next, a comprehensive search in the literature lists rock, fluid, and DFN properties. For each type of reservoir, changes in three controlling factors- ground in-situ stresses, fluid viscosity, and fluid flow rate- are examined in relation to the performance of the ScCO2. Furthermore, the difference between water and ScCO2 injection is discussed. The following main conclusions are drawn: Based on a close observation of ScCO2 dispersion in fracture reservoirs, twelve interaction types between the fluid and DFN are identified. Some can be formed in water injection, but some specific ones are identified only with ScCO2 injection, such as crack tip branching and rebranching, and are not observed in water injection. Based on a close observation of ScCO2 dispersion in fracture reservoirs, twelve interaction types between the fluid and DFN are identified. Some can be formed in water injection, but some specific ones are identified only with ScCO2 injection, such as crack tip branching and rebranching, and are not observed in water injection. It is observed that doubling the flow rate causes excessive fluid driven fractures (FDF), resulting in fluid lose and unnecessary leak-off. Therefore, based on these observations, a high injection flow rate is not recommended for ScCO2 injection. Furthermore, increasing the flow rate increases the breakdown pressure and decreases the breakdown time. It is observed that doubling the flow rate causes excessive fluid driven fractures (FDF), resulting in fluid lose and unnecessary leak-off. Therefore, based on these observations, a high injection flow rate is not recommended for ScCO2 injection. Furthermore, increasing the flow rate increases the breakdown pressure and decreases the breakdown time. Viscosity change is shown to have a limited impact on the FDF pattern. Seven viscosities adopted from the literature for ScCO2 were selected and examined in both continuum and fractured reservoirs. It was observed that viscosity change imposes negligible influence on both fracture pattern and breakdown pressure. This observation is valid for ScCO2 and cannot be expanded for any other working fluid. Viscosity change is shown to have a limited impact on the FDF pattern. Seven viscosities adopted from the literature for ScCO2 were selected and examined in both continuum and fractured reservoirs. It was observed that viscosity change imposes negligible influence on both fracture pattern and breakdown pressure. This observation is valid for ScCO2 and cannot be expanded for any other working fluid. Compared to water injection, ScCO2 generates much more fractures under identical conditions. It was observed that this kind of fluid can produce more FDFs by branching into multiple fractures, which makes new flow channels and better fluid dispersion. Moreover, compared to water, ScCO2 is a perfect alternative to generate longer FDFs by more DFN activation and dilation, resulting in a bigger fluid immersion domain. Compared to water injection, ScCO2 generates much more fractures under identical conditions. It was observed that this kind of fluid can produce more FDFs by branching into multiple fractures, which makes new flow channels and better fluid dispersion. Moreover, compared to water, ScCO2 is a perfect alternative to generate longer FDFs by more DFN activation and dilation, resulting in a bigger fluid immersion domain. A comparison of fluid pressure and breakdown histories of the model under water or ScCO2 injection revealed that water injection causes much higher breakdown pressure, which creates more fractures initiated from the borehole. On the other hand, post-breakdown pressure history is more stable and gentler than water’s, meaning that ScCO2 can generate more extended fractures even after the breakdown pressure is passed. A comparison of fluid pressure and breakdown histories of the model under water or ScCO2 injection revealed that water injection causes much higher breakdown pressure, which creates more fractures initiated from the borehole. On the other hand, post-breakdown pressure history is more stable and gentler than water’s, meaning that ScCO2 can generate more extended fractures even after the breakdown pressure is passed. To conclude, this study demonstrates the excellent performance of ScCO2 as the working fluid in deep hot dry rocks and is promising in successful reservoir stimulations. However, this study can be regarded as a starter and further studies are needed to fully characterise this kind of fluids and address challenges. For ScCO₂ fluid fracturing in HDR geothermal reservoirs, several geological, geomechanical, thermal, and operational conditions can limit the effectiveness of fracture initiation, propagation, and reservoir stimulation. These limitations should be carefully considered during reservoir selection and project design. These limitations are explained briefly. When the difference between the maximum and minimum principal stresses is large ( fractures tend to propagate as a single planar fracture rather than a complex network, which results in reduced fracture branching, smaller stimulated reservoir volume (SRV), and limited activation of natural fractures. Very deep HDR reservoirs may experience confining stresses exceeding 70–100 MPa, with consequences including higher breakdown pressure requirements, closure of natural fractures, reduced permeability enhancement, and increased pumping requirements. Highly homogeneous rock masses are another problematic issue. ScCO₂ performs best when pre-existing discontinuities exist. In very homogeneous granites, few natural weaknesses are available; fracture propagation becomes relatively planar, and the benefits of ScCO₂-induced branching are reduced. Excessive heterogeneity can also be problematic, leading to unpredictable fracture paths, localized stress concentrations, and non-uniform stimulation. Another problem arises in the absence of natural fractures. ScCO₂’s low viscosity promotes penetration into existing microcracks. If natural fractures are scarce, limited fracture complexity develops, and fracturing behavior approaches conventional hydraulic fracturing. When large connected fractures already exist, significant fluid leak-off, poor pressure buildup, and reduced fracture extension are consequences. Reservoir temperature is another factor contributing to the problem. Extremely high temperatures (> 250–300 °C) reduce CO₂ density, alter thermophysical properties, and decrease fluid storage capacity. Consequences: are lower fracturing efficiency, more difficult pressure control, and potential thermal degradation of well components. Formation permeability also contributes. While extremely low permeability is generally desirable for stimulation, it can lead to poor pressure transmission and delayed fracture initiation. Excessively high permeability can cause rapid pressure dissipation and limit fracture growth. Therefore, all limitations and strengths should be considered simultaneously for a successful fluid injection. Author contributions ZA: Conceptualization, Methodology, Software, Formal analysis, Investigation, Funding acquisition, Writing – original draft, Visualization.MS: Conceptualization, Methodology, Software, Formal analysis, Investigation, Writing – original draft, Visualization, Supervision.MB: Conceptualization, Methodology, Validation, Writing – review & editing. Funding This work is based upon research funded by Iran National Science Foundation (INSF) under project No.4028887. Data availability The datasets used and/or analysed during the current study available from the corresponding author on reasonable request. Declarations Conflict of interest The corresponding author declares no conflict of interests on behalf of all authors. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement The datasets used and/or analysed during the current study available from the corresponding author on reasonable request.