The Impact of Displacement Patterns on CO2 Storage in Saline Aquifers Based on Two Real Core Micromodels Authors: Changzhong Zhao, Qingping Li, Shouwei Zhou, Yongchen Song, Yi Zhang Published: 2026 ABSTRACT Due to its widespread geological distribution and substantial storage capacity, CO2 storage in saline aquifers is internationally recognized as one of the most effective methods for mitigating the greenhouse effect. The characteristics of the two-phase flow in porous media significantly affect the capacity and safety of CO2 storage. In this study, two structures at different locations of the Berea core were obtained and subsequently etched into two different micromodels to investigate the two-phase flow characteristics between CO2 and water. Using a microscopic visualization method, many experimental results have been obtained under different displacement patterns, including qualitative results of the displacement process, interface changes, and phase distributions and quantitative results of differential pressure, CO2 relative permeability, fractal dimension, and CO2 saturation. The results indicated that the maximum CO2 saturation, relative permeability, and fractal dimension were achieved when the displacement pattern was viscous fingering with CO2 predominantly existing in the micromodel as the main displacing channel. On the contrary, when the displacement pattern was capillary fingering, CO2 clusters showed the characteristics of large numbers and small areas, which led to the minimum CO2 saturation, relative permeability, and fractal dimension. Due to the simultaneous dominance of viscous and capillary forces, the flow characteristics under crossover fell between viscous fingering and capillary fingering, rendering them more complex. The phase states of CO2 and water exerted a profound influence on the differential pressure and displacement processes, primarily driven by the viscosity ratio at varying temperatures and pressures. The influence of the micromodel structure on the displacement process was mainly reflected in the local two-phase flow dynamics, resulting in numerical variations without significantly altering the general trend of change. Introduction Mitigating the greenhouse effect plays a crucial role in controlling global climate change, and CO2 geological storage is one of the carbon capture, utilization, and storage (CCUS) technologies with high feasibility and promising development prospects for mitigating the greenhouse effect. CO2 storage in deep saline aquifers is considered one of the most desirable technologies for geological CO2 storage due to its extensive geological distribution, high storage safety, and large storage capacity. , When CO2 is injected into subsurface formations to displace brine, the viscous, capillary, and inertial forces of CO2 and brine determine the displacement patterns, leading to complex fingering phenomena. The local mixing dynamics during displacement can significantly affect macroscopic flow phenomena, which in turn influence the storage capacity and safety of CO2 in aquifers. When CO2 is injected, structural trapping takes effect first. Then, the injected CO2 will be retained in pores in various forms through capillary trapping, and over time, the captured CO2 will eventually solidify in the formation through dissolution and mineralization trapping. During the entire trapping process, many factors such as the formation conditions, structural characteristics, and properties of injected fluids all play important roles on different phases’ transport paths, morphologies, and distributions. Furthermore, these flow characteristics provide an important basis for evaluating the effect of carbon sequestration in deep saline aquifers. To gain a comprehensive understanding of the two-phase flow and sequestration mechanisms of CO2 and water in porous media, an increasing number of researchers have used micromodels to carry out CO2–water displacement experiments in recent years. In 2009, Chalbaud et al. used a camera to study the variation of interfacial tension between CO2 and brine as a function of brine concentration, marking one of the earliest known experiments to visualize CO2–water flow using optical imaging technology. Jafari and Jung , obtained dynamic and static contact angles between CO2 and brine over a wide range of temperatures and pressures during displacement and imbibition processes. After that, the focus of optical imaging visualization experiments gradually shifted to the characteristics of two-phase flow and interface between CO2 and water. − Chang and Kneafsey’s team − have conducted several displacement experiments between CO2 and brine under real geological conditions using microscopy combined with camera technology since 2012. Their studies utilized both homogeneous and heterogeneous micromodels to investigate the effects of key factors, such as wettability, salinity, injection rate, structure, and initial fluid distribution, on the CO2–water flow characteristics. To overcome the limitations of two-dimensional flow in micromodels, they innovatively developed a 2.5D micromodel by varying the depth of etching. Overall, Chang and Kneafsey’s team − explored two-phase distribution, nonequilibrium dissolution, two-phase dissolution–diffusion, displacement mechanisms, and storage capacity during the displacement process through optical imaging technology. Buchgraber et al. investigated the effects of injection rate, temperature, pressure, structure, and roughness on CO2–water flow characteristics to elucidate the mechanisms of CO2 capture and dissolution during imbibition. In 2013, Sell et al. studied convection and diffusion characteristics when CO2 displaced brine in a tortuous single channel under varying salinities and pressures. Kazemifar et al. − employed dual-color microscopy combined with camera technology to study the velocity field when CO2 displaced the pure water, elucidating the intrinsic mechanisms of velocity mutations. Wang et al. examined the effects of CO2 injection rate and injection mode on saturation and displacement patterns in a homogeneous micromodel with a circular cross section at high temperature and pressure, observing the unstable displacement phenomena during the transition from capillary fingering to viscous fingering. During the period of 2011–2024, Jiang’s team − studied CO2 escape, heat transfer, and salinity precipitation in water under high temperature and pressure through simulations and experiments, while Hu et al. and Wang et al. continued to investigate salinity precipitation during CO2–brine displacement of in micromodels. Lv et al. , explored the effect of the CO2 injection direction on the displacement efficiency by changing the placement direction of micromodels. Despite numerous experimental studies on two-phase flow between CO2 and water in micromodels, the flow mechanisms in different displacement patterns remain inadequately understood, particularly the flow characteristics when the displacement pattern is crossover. Many details and phenomena of two-phase flow during the displacement process, such as the sudden change of velocity, snap-off, and water reflux, are still not fully understood. − At the same time, many existing two-phase flow experiments were conducted in homogeneous micromodels, primarily focusing on the CO2 injection rate. And the effects of other factors, such as the CO2 phase state (gaseous, liquid, and supercritical state), brine (types and concentrations of ions), and structural characteristics (porosity, permeability, wettability, heterogeneity, etc.), on two-phase flow have only been preliminarily explored without comprehensive and systematic experimental studies. Furthermore, most of the existing experiments take the displacement efficiency and CO2 saturation as important assessment indicators for two-phase flow, neglecting the safety of captured CO2 after the displacement. The safety of carbon sequestration is evaluated by more complex results, such as flow pathways, phase distributions, phase existence forms, and trapping mechanisms. To better elucidate two-phase flow mechanisms under real conditions, a microscopic visualization system suitable for high temperature and pressure was constructed in this study to measure two-phase flow characteristics. CO2 was injected into two different micromodels with different structures of the Berea core to displace water, and the effects of injection rate, CO2 phase state, and micromodel structure on the displacement process were analyzed from both temporal and spatial perspectives. Combining three displacement patterns, a comprehensive comparison and analysis were conducted on various characteristic results, including differential pressure, saturation, relative permeability, and fractal dimension. Finally, distributions and existing forms of CO2/water and the instability of the interface under three displacement patterns were clarified, laying a solid theoretical foundation for actual CO2 storage in saline aquifers. Experiment and Methodology To better simulate real geological environments, a previous microscopic visualization experimental system , has been improved for pore-scale flow under high temperature and pressure. The improved system incorporated a back pressure system, heating system, and enclosing pressure system to ensure the micromodel’s safety under high temperature and pressure, as illustrated in Figure . A pressure sensor (TESCOM 100-5000-2127) and a back pressure valve (Swagelok KPB1P0A412P20000) were connected to the outlet of the micromodel to achieve control of back pressure. To prevent CO2 backflow in the pipeline, a one-way valve (Swagelok SS-CHS2-1) was installed before the micromodel’s inlet. The enclosure system played a crucial role in achieving a high temperature and pressure. The enclosure chamber, featuring a bottom embedded with a 20 mm diameter and a 6 mm thick sapphire window, was connected to the enclosure pump via a piston container. The outer wall of the chamber was constructed as a cavity connected to the circulating water bath (Julabo CORIO CD) to control the chamber’s temperature. The injected fluids were preheated before displacement experiments, and all pipelines in this experimental system were wrapped with tin foil tape to minimize temperature loss. Physical diagram of the microscopic visualization experimental system applicable to high temperatures and pressures. Two structures at different locations of the Berea core were obtained through computed tomography (CT, inspeXio SMX-225CT). They were precisely scaled up to meet the requirements for making two distinct micromodels, as shown in Figure . The etching depth of the two micromodels in this study was 0.02 mm, and the overall porosities of the two micromodels were 0.32 (micromodel I) and 0.36 (micromodel II), respectively. The heterogeneity of micromodel I was primarily reflected in the middle region (4–9 mm from the inlet), where porosity varied significantly, creating a more tortuous flow path. In contrast, the heterogeneity of micromodel II was mainly reflected near the entrance and exit regions (0–3 and 9–12 mm from the entrance), characterized by a higher density of narrow pore throats. This difference in heterogeneity distribution rather than the absolute porosity value was the primary rationale for selecting these two structures. Prior to displacement experiments, the absolute permeabilities of the two micromodels were determined using Darcy’s law, yielding values of 2.389 and 2.660 D, respectively. Design diagram of the two micromodels: (a, b) micromodel I and (c, d) micromodel II. Most micromodels used in optical imaging techniques are two-dimensional, thus neglecting the gravitational and inertial forces of fluids. Consequently, non-mixed-phase displacement is primarily controlled by capillary and viscous forces, represented by two dimensionless numbers: the capillary number (Ca) and the viscosity ratio (M), respectively. Ca=μ2v2γcos⁡θ 1 M=μ2μ1 2 where subscripts 1 and 2 denote the displaced phase and displacing phase, respectively; μ is the dynamic viscosity; γ is the interfacial tension between phases; θ is the contact angle between phases and the porous medium; and v is the overall flow rate: v=QA 3 where Q is the volume flow rate of the displacing phase and A is the cross-sectional area of the micromodel. The Lenormand–Zhang phase diagram, proposed by Lenormand et al. − and Zhang et al., , serves as a criterion to determine the displacement pattern. According to the phase diagram, two-phase flow can be classified into four patterns: stable displacement, viscous fingering, capillary fingering, and crossover based on log Ca and log M. Based on the Lenormand–Zhang phase diagram, three CO2 volume flow rates were designed under three displacement patterns (capillary fingering, crossover, and viscous fingering) to better investigate the impact of displacement patterns on two-phase flow. In addition, to reveal the influence of the CO2 phase state on two-phase flow, under the viscous fingering displacement pattern, water was displaced by CO2 in the gaseous, liquid, and supercritical state, respectively. The phase state of CO2 was verified and defined by its position on the CO2 phase diagram, corresponding to the meticulously controlled temperature and pressure conditions of each experiment. The calculated log Ca and log M are shown in Table and Figure . Distributions of Ca and M in the Lenormand–Zhang phase diagram. 1Experimental Conditions and Calculated Ca and M under Different Volume Flow Ratesexperimental condition Q/mL/min v/10–4m/slog Ca log M 25 °C, 0.1 MPa0.0056.94–6.7800–1.77520.00811.11–6.5759–1.77520.0113.89–6.4790–1.77520.0569.44–5.7800–1.77520.08111.11–5.5759–1.77520.1138.89–5.4790–1.77520.5694.44–4.7800–1.77520.81111.11–4.5759–1.775211388.89–4.4790–1.775225 °C, 8 MPa0.5694.44–3.7800–1.124540 °C, 8 MPa0.5694.44–4.3020–1.4518 Experimental Conditions and Calculated Ca and M under Different Volume Flow Rates To verify the repeatability of flow experiments in the micromodel, CO2 was injected at a flow rate of 0.01 mL/min to displace water in micromodel II, and the phase distribution obtained from three repeated experiments is shown in Figure . The detailed image processing method had been mentioned in a previously published article. CO2 was injected from left to right, with red representing the structural skeleton, black representing CO2, and cyan representing water throughout this paper. There was a certain degree of randomness in the local phase distribution, but the overall phase distribution, flow path, and morphology of each phase showed little difference. Through image processing, it was found that the CO2 saturations after three rounds of displacement were 0.586, 0.545, and 0.567, respectively, indicating good repeatability of the experimental system. Phase distribution in micromodel II after the breakthrough when CO2 displaced water with the injection rate of 0.01 mL/min repetitively three times. Experimental Results Characteristics of Pressure during Displacement The variation of differential pressure between the inlet and outlet with injected pore volume of CO2 in the micromodel was first obtained through experimental measurement, as shown in Figure . Variation of differential pressure between the inlet and outlet with injected pore volume under different injection flow rates: (a) micromodel I and (b) micromodel II. The variation of the differential pressure differed across displacement patterns. When displacement patterns were capillary fingering and crossover, differential pressure accumulated continuously over time until the breakthrough. After the breakthrough, the differential pressure began to decrease and eventually oscillated and stabilized. The entire displacement process was completed without injecting excessive pore volumes of CO2. However, when the displacement pattern transitioned to viscous fingering, differential pressure throughout the displacement process only had two stages: accumulation and breakthrough. After the breakthrough, the differential pressure between the inlet and outlet remained at the maximum value in micromodel I, while it continued to rise in micromodel II, primarily due to the different locations of heterogeneity in the two micromodels: micromodel I’s heterogeneity was mainly reflected in the middle position, whereas micromodel II’s heterogeneity was mainly manifested at the inlet and outlet positions. There are three important characteristic numbers of differential pressure: breakthrough pressure (P b), stabilization pressure (P s), and cumulative injected pore volume of CO2 when reaching the breakthrough pressure. P b can optimize the injection parameters, P s is closely related to storage safety, and the cumulative injected CO2 pore volume determines the amount of captured CO2. The variation of these characteristic numbers with log Ca in the two micromodels is shown in Figure . Variation of three differential pressure characteristics with log Ca in the two micromodels: (a) breakthrough pressure, (b) stabilization pressure, and (c) the cumulative injected pore volume of CO2 when reaching the breakthrough pressure. Overall, the three characteristic numbers increased with log Ca, which were mainly attributed to the positive relationship between the viscous force and the injection velocity. However, the dominant roles of capillary and viscous forces differed in viscous displacement patterns, leading to more complex changes in the characteristic numbers. Besides the minimum injection rate, P b and P s were larger in micromodel II than in micromodel I. This was mainly because there were more narrow pores and throats at the inlet of micromodel II, increasing the pressure that needed to be overcome. The P b and P s in micromodel II are larger than those in micromodel I. To achieve a larger breakthrough pressure, the cumulative pore volume of injected CO2 is also increased. Consequently, to achieve a larger breakthrough pressure, the cumulative injected CO2 pore volume increased. The differences of these three characteristics between the two micromodels were relatively small during capillary fingering, while the differences remained relatively large during viscous fingering and decreased with an increasing injection rate when the pattern changed to the crossover. Results of Water Displacement by CO2 in Different Phases Variations of differential pressure with the injected pore volume in micromodels I and II are shown in Figure when gaseous, liquid, and supercritical CO2 displaced water with an injection rate of 0.5 mL/min. The process of gaseous CO2 displacing water was detailed above and will not be repeated here. In micromodel I, when liquid CO2 displaced water, differential pressure drastically varied with the injected pore volume, jumping back and forth between 10 and 62.2 kPa (62.2 kPa was the maximum limit of the differential pressure transducer) with high frequency, which was related to strong heterogeneity in the middle region of this micromodel. This tendency diminished after the breakthrough, and the differential pressure gradually decreased and oscillated around 0 kPa. When supercritical CO2 displaced water, differential pressure became very complex with the injected pore volume as well, and after the breakthrough, differential pressure rose sharply to the maximum and fluctuated continuously, which was unfavorable for the safety of sequestrated CO2. In micromodel II, when liquid or supercritical CO2 displaced water, the same complex situation of differential pressure occurred. In this case, differential pressure remained above the maximum limit of the differential pressure transducer for an extended period, accompanied by periodic fluctuations during the displacement process, which were more pronounced when supercritical CO2 displaced water in micromodel II. Variation of differential pressure ΔP with the injected pore volume PV in micromodels I (above) and II (below) when different phases of CO2 were injected at the rate of 0.5 mL/min to displace the water, (a) gaseous CO2, (b) liquid CO2, and (c) supercritical CO2. As an example, the flow process in micromodel I for three phases of CO2 displacing water is detailed in Figures –. When gaseous CO2 displaced water, CO2 bubbles with different sizes formed in large pore spaces. When water backflow occurred, the number of CO2 bubbles increased, but when CO2 displaced water again, these bubbles merged into the main flow path. As CO2 continued to be injected, snap-off of CO2 tended to occur at smaller throat locations, leading to the formation of CO2 clusters of different shapes and sizes. Flow process when gaseous CO2 was injected at a rate of 0.5 mL/min to displace water in micromodel I. Flow process when liquid CO2 was injected with a rate of 0.5 mL/min to displace water in micromodel I. Flow process when supercritical CO2 was injected with a rate of 0.5 mL/min to displace water in micromodel I. When liquid CO2 displaced water, a large number of small CO2 droplets formed in the pore space initially, and the displacement process was dominated by the merging of these small droplets, which constantly moved with the main flow path. Even after the breakthrough, more small CO2 droplets still remained near the main flow path. The phenomenon was similar to the initial displacement process of gaseous CO2 displacing water when the temperature of CO2 was supercritical. However, with continuous CO2 injection, CO2 flowed in a viscous form, primarily related to the viscosity and interfacial tension due to the CO2 phase change. The viscosity of supercritical CO2 was between the gaseous and liquid CO2, while water had the smallest viscosity at te corresponding temperature and pressure among the designed experimental conditions. It was also observed that there was continuous flow of supercritical CO2 above the studied field of the micromodel after the breakthrough but no impact on the distribution and saturation of the formed CO2 and water in the flow path. Distribution and Saturation of CO2 at the End of Displacement Phase distribution and saturation in the micromodel were obtained when CO2 was injected at different rates to displace water after the breakthrough, as shown in Figures and , with numbers in brackets denoting capillary number (log Ca) and CO2 saturation (S CO2), respectively. The average S CO2 was calculated for three displacement patterns separately. The calculated average S CO2 in micromodel I was capillary fingering (0.2396) < crossover (0.4395) < viscous fingering (0.6859), and in micromodel II, it was capillary fingering (0.4754) < crossover (0.4981) < viscous fingering (0.7924). Regardless of the micromodel structure, the largest average S CO2 was obtained when the displacement pattern was viscous fingering followed by crossover, and the smallest was obtained for capillary fingering. It was important to note that the high saturation observed in the viscous fingering regime in this 2D pore-scale study referred to the saturation within the developed flow path and the surrounding pores it accessed. This should not be conflated with macroscopic sweep efficiency in a 3D reservoir, where viscous fingering typically led to lower overall sweep and storage efficiency due to the bypassing of large volumes of the pore space. Phase distribution in micromodel I when CO2 was injected at different rates to displace water after the breakthrough. Phase distribution in micromodel II when CO2 was injected at different rates to displace water after the breakthrough. In micromodel I, the displacement pattern was viscous fingering when CO2 displaced water with three injection rates of 1.0, 0.8, and 0.5 mL/min. In this case, viscosity force was dominant, and S CO2 decreased with a decreasing injection rate. There was a large CO2 cluster (1–15 mm2) at all three injection rates as the main flow path. The area of this type of CO2 cluster increased with an increasing injection rate, reaching up to 18.35 mm2 when the injection rate was 1.0 mL/min, leading to the highest CO2 saturation at this injection rate. Additionally, there were many small CO2 clusters (0–1 mm2) in the micromodel, and the number and cumulative area of these clusters decreased with an increasing injection rate. Overall, when the displacement pattern was viscous fingering in micromodel I, CO2 was captured as one large cluster, accompanied by a large number of small CO2 clusters. The displacement pattern was crossover when CO2 displaced water with three injection rates of 0.1, 0.08, and 0.05 mL/min. After the breakthrough, CO2 predominantly existed in the form of small clusters, but there was no longer a strong relationship between the number and cumulative area of small clusters and the injection rate. This was related to the simultaneous dominance of two forces, which also contributed to the complex relationship between the CO2 saturation and capillary number. When the displacement pattern was capillary fingering, most of the CO2 was captured in the form of small clusters. Some CO2 was trapped in small pores through snap-off at the upper part of the micromodel, and some CO2 was trapped in large pores at the lower right in the form of strips or fingers with an area of 0–1 mm2. Overall, most trapped CO2 was concentrated in intermediate regions with a strong heterogeneity. Compared to the other two displacement patterns, more displacing channels were formed at the inlet of the micromodel when CO2 was injected at low rates, which reduced displacement efficiency, combined with the occurrence of water backflow. When water was displaced by CO2 with injection rates of 1.0 and 0.8 mL/min in micromodel II, the obtained CO2 saturation and distribution were essentially equal. At this time, the displacement pattern was viscous fingering, and CO2 was no longer predominantly present in small clusters. Conversely, the area of the main displacing channel represented more than 95% of the captured CO2, and most residual water was present at the upper left of micromodel II due to narrow pores and throats around the inlet position. When the displacement pattern was crossover, the overall distribution of the CO2 clusters exhibited characteristics of large numbers and small areas after the breakthrough. The accumulated area of CO2 clusters in the range of 0–1 mm2 was the largest, accounting for about 50% of the total cumulative area, which increased with decreasing injection rate. The captured CO2 essentially existed in stripe or finger-like forms, even appearing in locations where water could not be displaced for viscous fingering, correlating with flow properties for crossover. When injection rates were 0.005 and 0.008 mL/min, CO2 displace water with the pattern of capillary fingering, achieving close CO2 saturation but different distributions. When the injection rate was 0.005 mL/min, the cumulative area of CO2 clusters in the range of 0.1–1 mm2 was the largest, accounting for more than 50% of the total cumulative area. The number of CO2 clusters in the range of 0.01–0.1 mm2 was the largest, primarily existing as thin stripes, and water near the inlet and outlet of micromodel II was essentially not displaced by CO2, which was conducive to the stable CO2 storage. When the injection rate was 0.008 mL/min, the accumulated CO2 clusters in the range of 1–10 mm2 had the largest area, occupying more than 40% of the total area of all CO2 clusters. The number of CO2 clusters in the range of 0.01–0.1 mm2 was the largest, and the total number of CO2 clusters at this injection rate was more than that at the injection rate of 0.005 mL/min, so the CO2 clusters obtained at 0.008 mL/min at the same CO2 saturation were more dispersed. Analysis and Discussion The CO2 relative permeability K rg at different injection rates in both micromodels was calculated using stabilization pressure after the breakthrough, and the variation with log Ca is shown in Figure . The result showed that the trend was not affected by the micromodel structure: K rg increased with log Ca. Near the maximum and minimum log Ca values in this study, K rg varied very little with log Ca, whereas K rg changed strongly with log Ca when the displacement pattern was crossover and conversion. Therefore, the smaller the injection rate is, the stronger is the stability of captured CO2, and the more favorable it is for safe CO2 storage. The influence of the micromodel structure on K rg was mainly shown in the value. K rg ranged from 0 to 0.42 in micromodel I, which was essentially twice that in micromodel II. The reason was that the heterogeneity of micromodel II was mainly reflected near the entrance and exit positions, limiting the diffusion and migration capacity of captured CO2. Variation of K rg with log Ca at stabilization in two micromodels. The fractal dimension D f of the two micromodels at stabilization was obtained by using the software ImageJ, as shown in Figure . The D f quantified the complexity and space-filling characteristics of the CO2 cluster pattern. A higher D f, as observed in viscous fingering, indicated a more compact, thicker, and less tortuous main flow path, which correlated with a higher relative permeability and saturation. Conversely, a lower D f in capillary fingering reflected a more ramified, dendritic, and tortuous pattern composed of many thin fingers. This highly tortuous path increased flow resistance (lower relative permeability) and promoted capillary trapping through snap-off events, leading to a larger number of isolated, smaller clusters. Thus, D f served as a quantitative bridge between the observed displacement pattern and the resulting trapping morphology and flow capability. In this study, the calculated D f of micromodel I ranged from 1.6620 to 1.7387, with an average relative error of 0.026, and the calculated D f of micromodel II ranged from 1.7035 to 1.7728, with an average relative error of 0.026. Therefore, phase distributions in micromodel II were more complex than those in micromodel I. Moreover, the micromodel structure caused different trends of D f with log Ca. As shown in Figure , it can be observed that the trend of D f in micromodels I and II was opposite with the increase of log Ca. However, regardless of the structure, the largest D f was obtained when viscous fingering occurred. Variation of D f with log Ca at stabilization in the two micromodels. The variation of the CO2 saturation S CO2 with log Ca was similar to the fractal dimension in two micromodels, as shown in Figure . Overall, S CO2 increased with increasing injection rate; i.e., the maximum average S CO2 was obtained when the displacement pattern was viscous fingering followed by crossover, and the minimum was capillary fingering. For the same displacement pattern, the variation of S CO2 with log Ca was complex, and the micromodel structure had an impact on this trend. In terms of CO2 saturation, greater S CO2 was obtained in micromodel II than in micromodel I, with an average S CO2 of 0.589 for micromodel II and 0.455 for micromodel I. Combined with other experimental results, such as differential pressure characteristics, CO2 clusters, phase distributions, CO2 relative permeability, and fractal dimension, it was determined that the displacement effect was better in micromodel II under the measurement conditions of this study, and the maximum S CO2 could be achieved at an injection rate of 0.8 mL/min in micromodel II. Variation of S CO2 with log Ca at stabilization in two micromodels. The pore-scale phenomena observed here have direct implications for field-scale CO2 storage performance. The identification of displacement regimes via the log Ca–log M phase diagram provides a predictive tool for estimating the initial plume morphology. For instance, operations designed to be in the capillary fingering regime (low injection rates) will promote a more dispersed plume and higher residual trapping via snap-off, enhancing short-term security but potentially yielding lower initial mass saturation. Conversely, viscous fingering (high rates) achieves high local saturation along dominant flow paths, which is beneficial for injectivity and maximizing structural trapping in the target zone, but may reduce the surface area for dissolution and increase leakage risk if the caprock is heterogeneous. The complex and unstable pressure signals observed in the crossover regime suggest that operational conditions near this transition should be carefully managed to avoid uncontrolled flow instabilities. Therefore, optimizing storage projects requires a balanced consideration of these pore-scale controls on capacity, injectivity, and trapping safety. Conclusions Base on real core structures, displacement experiments of CO2/water were conducted in two micromodels to obtain the flow processes under three displacement patterns. Then, differential pressure characteristics, CO2 relative permeability, fractal dimension, and CO2 saturation were comprehensively compared. Combining displacement patterns, the effects of micromodel structure and CO2 phase state on two-phase flow were analyzed to comprehensively explore the impact of multiple factors on capacity and safety of CO2 storage.1.Through analysis of differential pressure between the inlet and outlet, it was found that the change of differential pressure under viscous fingering was different from that under capillary fingering and crossover; i.e., there were only two stages during the entire displacement process (accumulation and breakthrough) when gaseous CO2 was injected to displace water for viscous fingering. The breakthrough pressure and the cumulative injected pore volume of CO2 increased with the injection rate. When CO2 phase state changed to liquid and supercritical states, the change of differential pressure was more complex, varying dramatically between 0 and 62.2 kPa.2.The effect of CO2 phase state on the displacement process was investigated to find that the snap-off of CO2 tended to occur at locations with a smaller throat. Unlike gaseous CO2, when liquid and supercritical CO2 displaced water, a large amount of spherical CO2 occurred, which continuously merged to form the main flow path. At the late stage of displacement, supercritical CO2 displaced water in a viscous form, and the two were well mixed and dispersed in the micromodel in a discrete broken form, which was not conducive to achieving large CO2 saturation.3.The relationship between CO2 saturation at stabilization and displacement patterns was analyzed. Overall, the largest CO2 saturation was obtained in viscous fingering and the smallest in capillary fingering, influenced by the dominant forces in different displacement patterns. Combined with the displacement process and phase distribution, when the displacement pattern was viscous fingering, the main CO2 flow path was dominant, accompanied by a large number of small-area CO2 clusters. When the displacement pattern changed to capillary fingering, the main CO2 flow path was restricted to be narrowed, and the existence of CO2 was dominated by the larger independent clusters caused by snap-off.4.CO2 relative permeability, saturation, and fractal dimension were compared in the two micromodels. The results showed that the CO2 saturation and fractal dimension obtained in micromodel II were larger and the relative CO2 permeability was smaller, indicating that more CO2 could be captured in micromodel II, and the sequestrated CO2 had a more complex distribution and higher stability, which was more suitable for CO2 injection and storage. 1.Through analysis of differential pressure between the inlet and outlet, it was found that the change of differential pressure under viscous fingering was different from that under capillary fingering and crossover; i.e., there were only two stages during the entire displacement process (accumulation and breakthrough) when gaseous CO2 was injected to displace water for viscous fingering. The breakthrough pressure and the cumulative injected pore volume of CO2 increased with the injection rate. When CO2 phase state changed to liquid and supercritical states, the change of differential pressure was more complex, varying dramatically between 0 and 62.2 kPa. Through analysis of differential pressure between the inlet and outlet, it was found that the change of differential pressure under viscous fingering was different from that under capillary fingering and crossover; i.e., there were only two stages during the entire displacement process (accumulation and breakthrough) when gaseous CO2 was injected to displace water for viscous fingering. The breakthrough pressure and the cumulative injected pore volume of CO2 increased with the injection rate. When CO2 phase state changed to liquid and supercritical states, the change of differential pressure was more complex, varying dramatically between 0 and 62.2 kPa. 2.The effect of CO2 phase state on the displacement process was investigated to find that the snap-off of CO2 tended to occur at locations with a smaller throat. Unlike gaseous CO2, when liquid and supercritical CO2 displaced water, a large amount of spherical CO2 occurred, which continuously merged to form the main flow path. At the late stage of displacement, supercritical CO2 displaced water in a viscous form, and the two were well mixed and dispersed in the micromodel in a discrete broken form, which was not conducive to achieving large CO2 saturation. The effect of CO2 phase state on the displacement process was investigated to find that the snap-off of CO2 tended to occur at locations with a smaller throat. Unlike gaseous CO2, when liquid and supercritical CO2 displaced water, a large amount of spherical CO2 occurred, which continuously merged to form the main flow path. At the late stage of displacement, supercritical CO2 displaced water in a viscous form, and the two were well mixed and dispersed in the micromodel in a discrete broken form, which was not conducive to achieving large CO2 saturation. 3.The relationship between CO2 saturation at stabilization and displacement patterns was analyzed. Overall, the largest CO2 saturation was obtained in viscous fingering and the smallest in capillary fingering, influenced by the dominant forces in different displacement patterns. Combined with the displacement process and phase distribution, when the displacement pattern was viscous fingering, the main CO2 flow path was dominant, accompanied by a large number of small-area CO2 clusters. When the displacement pattern changed to capillary fingering, the main CO2 flow path was restricted to be narrowed, and the existence of CO2 was dominated by the larger independent clusters caused by snap-off. The relationship between CO2 saturation at stabilization and displacement patterns was analyzed. Overall, the largest CO2 saturation was obtained in viscous fingering and the smallest in capillary fingering, influenced by the dominant forces in different displacement patterns. Combined with the displacement process and phase distribution, when the displacement pattern was viscous fingering, the main CO2 flow path was dominant, accompanied by a large number of small-area CO2 clusters. When the displacement pattern changed to capillary fingering, the main CO2 flow path was restricted to be narrowed, and the existence of CO2 was dominated by the larger independent clusters caused by snap-off. 4.CO2 relative permeability, saturation, and fractal dimension were compared in the two micromodels. The results showed that the CO2 saturation and fractal dimension obtained in micromodel II were larger and the relative CO2 permeability was smaller, indicating that more CO2 could be captured in micromodel II, and the sequestrated CO2 had a more complex distribution and higher stability, which was more suitable for CO2 injection and storage. CO2 relative permeability, saturation, and fractal dimension were compared in the two micromodels. The results showed that the CO2 saturation and fractal dimension obtained in micromodel II were larger and the relative CO2 permeability was smaller, indicating that more CO2 could be captured in micromodel II, and the sequestrated CO2 had a more complex distribution and higher stability, which was more suitable for CO2 injection and storage.