1 Introduction

CO2 storage in deep saline aquifer formation is considered an important option to mitigate CO2 emission to the atmosphere (Metz et al. 2005). After its capture from a concentrated source (e.g., coal power plant), CO2 is injected into deep permeable geological formations causing formation fluids (e.g., brine) to be displaced. Among various factors that may limit the storage efficiency, hydraulic properties of the porous media and the interfacial properties of CO2 and formation fluids are expected to play important roles in affecting fluid flow, plume development and saturation levels (Zhang et al. 2011a; Wang et al. 2013a).

The displacement process of brine by scCO2 is commonly treated as an immiscible system considering the low solubility between different phases, thus asserting a major effect on each other’s displacement. However, the mutual solubility between scCO2 and brine (Spycher et al. 2003; Lu et al. 2009) can be high, particularly at higher temperatures in deeper aquifers, which may affect the displacement process and flow properties at the pore scale: (i) the mutual dissolution affects the physical properties, such as interfacial tension, viscosity, and density. The variation of interfacial tension between the two phases alters the capillary forces, and thus the fingering patterns; (ii) scCO2 extracts water in the pore spaces and increases the salinity of brine, and even causes salt precipitation and porosity/permeability reduction, which is commonly called dry-out effect (Pruess 2009; Pruess and Müller 2009).

Recently, a growing body of modeling and experimental research across multiple spatial scales has been conducted to investigate the effects of dissolved CO2 in aqueous formations, including mineral precipitation/dissolution as well as rock porosity and permeability modification (Nordbotten et al. 2005; Xu et al. 2005; Lu et al. 2009; Pruess 2009; Pruess and Müller 2009; Zhou et al. 2010; Lopez et al. 2011; Sohrabi et al. 2011; McCaughan et al. 2013). For instance, Pruess et al. investigated the effects of CO2 dry-out resulting in salt precipitation, potentially interfering with injection operations by simulation code TOUGH2 (Xu et al. 2005; Pruess and Müller 2009). According to their study, even the water solubility in scCO2 is only a fraction of a percent, and continuous injection of DscCO2 into saline aquifers may eventually lead to complete formation dry out which will cause dissolved solids to participate. Bachu and Bennion found that capillary pressure, IFT (interfacial tension), and relative permeability depend on pressure, temperature and water salinity, as well as the pore-size distribution of the sedimentary rock in scCO2–brine systems in deep saline aquifer formations (Bennion and Bachu 2005, 2006a, b, c, 2007a, b, 2008; Stefan Bachu and Brant; Bennion 2008; Bachu and Brant Bennion 2009). However, research related to the following issues is limited: (i) potential advantages of scCO2 dry-out effect which may enhance the scCO2 sweep efficiency (i.e., the ratio of areas of displacement and reservoir). (ii) The effect of the dissolved water in scCO2 which may alter the scCO2 sweep efficiency. So far, the dissolved water in scCO2 is ignored in most works because of the relatively low solubility of water in scCO2, and the fact that below 100 °C the viscosity of water-containing scCO2 is close to that of pure scCO2. However, now it is known that the dissolved water in scCO2 drastically enhances the activities of scCO2 with rock minerals (e.g., CaCO3) (Loring et al. 2010), indicating that the dissolved water plays a significant role in scCO2–brine displacement process (Wang et al. 2013b).

In the two-phase immiscible displacement process of water displaced by CO2, the CO2 is a non-wetting phase and the water is a wetting phase. Over the last three decades, fundamental understanding of the mechanisms that control immiscible two-phase flow under ambient pressure and temperature conditions has been gained through abundant pore-scale experiments using microfabricated physical models of porous media (i.e., micromodels) and a variety of pore network modeling approaches (Lenormand et al. 1988; Ferer et al. 2004; Nordbotten et al. 2005; Cottin et al. 2010; Zhou et al. 2010; Hatiboglu and Babadagli 2010; Lengler et al. 2010; Zhang et al. 2011a, b; Zhao et al. 2011; Wang et al. 2013a; Mohammadi et al. 2013; Kazemifar et al. 2016; Roman et al. 2016; Takeshi et al. 2016). Displacement of a wetting phase by a non-wetting fluid in a 2D pore network, commonly referred to as main drainage, is often described by two dimensionless numbers, the capillary number Ca and the viscosity ratio M. Here,

$${\text{Ca}}={\text{ }}({\mu _{\text{n}}} \times {u_{\text{n}}})/({\sigma _{{\text{nw}}}} \times \cos \;\theta )\;{\text{and}}\;M={\mu _{\text{n}}}/{\mu _{\text{w}}},$$
(1)

where µn and un are the viscosity and velocity of the invading non-wetting fluid, respectively, σnw is the interfacial tension, θ is the contact angle, and µw is the viscosity of the initially residing wetting fluid. Depending on Ca and M, different mechanisms, including capillary fingering, crossover, viscous fingering and stable displacement, have been shown to control immiscible displacement at pore scale (Wang et al. 2013a).

For immiscible fluid displacement with low viscosity ratios (i.e., logM < 0) which is the case of the scCO2–water systems, the three mechanisms of capillary fingering, crossover and viscous fingering control the fluid flow, respectively, along with increase in Ca number (Wang et al. 2013a). In a previous study (Ferer et al. 2004), it was found that the scCO2 saturation (SscCO2) was very high during capillary fingering dominant period, confirming observations under ambient conditions by Ferer (Lenormand et al. 1988) and (Mahadevan et al. 2007; Pruess and Müller 2009). In these studies, the mutual solublities and the corresponding disturbances to each other were not considered because of the nearly zero mutual solubility between the nonaqueous phase and aqueous phase during the displacement process. However, with the drying effects of CO2, at the interfacial area of CO2 and brine, the DscCO2 will dissolve into the aqueous phase and water will dissolve into CO2, and the displacement system between DscCO2 and water cannot merely be considered as an immiscible displacement process (Spycher et al. 2003; Wang et al. 2013b). So far, the impacts of the drying effects from DscCO2 and the influence of the dissolved water in WscCO2– on the scCO2–water displacement process are unclear.

In this study, we expand our work to further interrogate the capillary forces combined with drying effects relevant to DscCO2–water displacement at the pore scale under reservoir conditions. Specifically, we focus on DscCO2– and WscCO2–water displacements, respectively, in two near homogeneous pore networks micromodels with well-defined surface property and pore geometry and different inlet and outlet designs over a broad range of injection rates (logCa = − 7.64 ~ − 4.76) under high pressure (9 MPa) and temperature of scCO2 (> 31 °C). The displacement process was observed in real time using fluorescence microscopy and SscCO2 was quantitatively evaluated from fluorescent images. Results obtained from experiments about the saturation of DscCO2 were compared with that of WscCO2 at the same Ca, to analyze the drying effects from DscCO2 and the influence of the dissolved water in WscCO2 on SscCO2.

2 Experimental methods

2.1 Fluids

Distilled and deionized water (DDI) and supercritical fluid chromatography (SFC)-grade scCO2 were used as the wetting and non-wetting phases, respectively. The non-wetting scCO2 was classified as either DscCO2 or WscCO2, with a dissolved water content of 0 and 100%, respectively. Based on our previous work, the water saturation concentration in scCO2 at 40 °C and 35 °C are 0.32 and 0.49 mol%, respectively (Choi et al. 1998; Zhang et al. 2011a; Wang et al. 2013a). WscCO2 was made by injecting 1 ± 0.1 mL water into Pump A [Figure S1(c), in Supporting Information (SI)] that was filled with 28.9 ± 1.5 mL scCO2. The presence of excess water ensures saturation level water concentration in scCO2 at 35 or 40 °C over 12 h. Table 1 summarizes the properties of the wetting and non-wetting fluids. Since the density, viscosity and interfacial tension (IFT) of WscCO2 have not been published and experimental measurement of them in our laboratory is not feasible, in this study the viscosity and density were estimated by empirical calculation of the gas mixtures (Equation S1 in SI). The results indicated that the density and viscosity of WscCO2 were almost the same as those of DscCO2. The contact angle between DscCO2 and water and that between WscCO2 and water was also determined to be approximately the same in this study (Figure S6). Similarly, the IFT between WscCO2 and water was considered to be the same as that between DscCO2 and water. From the above, those properties required to calculate the Ca of WscCO2 are almost the same as those of DscCO2.

Table 1 Summary of experimental conditions, fluid properties, volumetric flow rates and the corresponding Darcy velocity and capillary number
Full size table

A fluorescent dye, Coumarin 153 (99.99%; Alfa Aesar, Ward Hill, MA), was added to scCO2 to distinguish scCO2 from water using fluorescent microscopy. Coumarin 153 absorbs in the near UV and emits in the broad spectral range from 450 nm to over 600 nm. Details of Coumarin 153 characteristics can be found in the literature (Chomsurin 2003; Willingham et al. 2008; Zhang et al. 2011a, b; Wang et al. 2013a). All chemicals are ACS-reagent-grade (Sigma–Aldrich). In this study, a stock solution of Coumarin 153 was prepared in methanol at a concentration of 10 mM, and then 1 mL stock solution was added to the cylinder of Pump A (shown in Figure S1(c) in SI). The methanol can be evaporated and the Coumarin 153 can be fully mixed by purging the cylinder with scCO2 for over 10 h.

2.2 Micromodels

The micromodels were fabricated using microfabrication procedures described in previous studies (Zhang et al. 2011a; Wang et al. 2013a). Two types of micromodels, M-1 and M-2 (Figure S1(a) and S1(b), in SI), were used as the pore networks in this work. The overall pore geometry of both was the same, except for the inlet and outlet boundaries in model M-2, for which the inlet and outlet channels were filled with grains to simulate the effects of different boundary conditions on the distribution of scCO2 plume. Both micromodels M-1 and M-2 have similar homogeneous network of cylindrical grains (200 µm diameter) and pore spaces (120 µm pore bodies, 26.7 µm pore throats), and average pore depth (35 ± 1 µm).

2.3 scCO2–water displacement

The high-pressure micromodel experimental system (Figure S1(c)) has been described in detail previously (Lenormand et al. 1988; Ferer et al. 2004; Zhang et al. 2011a, b; Wang et al. 2013a). A total of four different volumetric injection rates ranging from 10 to 5000 uL/h were applied (Table 1). The imposed flow rates correspond to a range in the converted Darcy velocity from 0.57 to 283 m/day, and a range in logCa from − 7.64 to − 4.94 at 40 °C or from − 7.46 to − 4.76 at 35 °C, respectively. A total of 24 scCO2–water displacement experiments were conducted in this study (Table 1), divided into three independent sets of experiments as follows:

Set I, DscCO2– and WscCO2–water displacement experiments using the M-1 micromodel at 40 °C: DscCO2 was injected into M-1 at a constant flowrate until quasi-steady state was reached (i.e., the saturation of the scCO2 did not change over time). DscCO2–water displacement experiments with four different flow rates (Q = 10, 100, 1000, and 5000 uL/h) were conducted. WscCO2 displacement experiments were performed under the same scenario as DscCO2.

Set II, DscCO2– and WscCO2–water displacement experiments using the M-2 micromodel at 40 °C: DscCO2 and WscCO2 were injected into the M-2 micromodel and performed the displacement experiments were performed under the same scenario as Set I, respectively.

Set III, DscCO2– and WscCO2–water displacement experiments using the M-2 micromodel at 35 °C: DscCO2 and WscCO2 were injected into the M-2 micromodel and the displacement experiments were performed under the same scenario as Set II, respectively.

Here, two different supercritical temperatures in Set II and III were conducted to delineate the thermal effects on the behavior of CO2 migration. A summary of experimental conditions is listed in Table 1. All displacement experiments were conducted horizontally, so density and buoyancy effects were not considered, consistent with approaches in other recent 2D micromodel experiments (Zhang et al. 2011a, b; Wang et al. 2013a; Zuo et al. 2013).

2.4 Image acquisition and analysis

Direct visualization of the scCO2 distribution in the micromodels was obtained using a fluorescent microscope and quantification of saturation in the pore network of micromodel was achieved by image segmentation and pixel counting. Details of the data treatment methods have been described in our previous work (Wang et al. 2013a).

3 Results and discussion

3.1 Water–scCO2 displacement characteristics

Fluorescent images of scCO2 distribution in the micromodel at each injection rate (expressed as logCa) for the three experiment sets are shown in Fig. 1. Each image was obtained after one pore volume (PV) of scCO2 was injected into the pore network, considering at all displacement experiments that (i) CO2 breaks through the micromodel, (ii) the displacement is at quasi-steady state, and (iii) scCO2 saturation in the micromodel does not significantly change after 1.0 PV of scCO2 injection.

Fig. 1
figure 1

scCO2 (blue) distribution at various injection rates (logCa) for the DscCO2– and WscCO2–water experiments in: a M-1 micromodel at 40 °C. b M-2 micromodel at 40 °C. c M-2 micromodel at 35 °C. The scCO2 flow is from left to right. The numbers indicate the logCa and the coresponding injection rate Q

Full size image

In the DscCO2–water displacement experiments at low injection rates, DscCO2 entered the pore network as a relatively uniform front, followed by randomly distributed forward and lateral migration paths with clusters of entrapped water (Fig. 1, logCa ≤ − 6.64 for Set I and II, logCa ≤ − 6.46 for Set III). Approximately halfway through the micromodel, the DscCO2 flow path transitioned into one gradually narrowing finger migrating to the outlet. This displacement mechanism can be classified as capillary fingering dominant. At high injection rate, the DscCO2 entered the pore network at multiple locations in the form of narrow viscous fingers (i.e., 1 ~ 3 pore bodies) (Fig. 1, logCa = − 4.94 for Set I and II, logCa = − 4.76 for Set III). These viscous fingers were distributed across the entire width of the micromodel and mainly progressed forward from the inlet to the outlet. This displacement mechanism can be classified as viscous fingering preferential. At medium injection rate, the DscCO2 front started rather evenly, similar to the formation of capillary fingers, and gradually transitioned to narrower fingers, i.e. viscous fingers (Fig. 1, logCa = − 5.64 at 40 °C, logCa= − 5.46 at 35 °C). These DscCO2 flow paths include a mixture of wide lateral paths and narrow forward progressing paths throughout the entire pore network. This displacement mechanism could be defined as crossover. The same phenomena have been observed by others (Zhang et al. 2011a, b; Wang et al. 2013a).

In the WscCO2–water displacement experiments, at low and medium injection rates, the WscCO2 front started rather evenly and gradually transitioned to one or two narrow fingers (i.e., 1 ~ 3 pore bodies) in a pattern of crossover. At high injection rate, the WscCO2 entered the pore network and displaced water in a pattern of viscous fingering (Fig. 1).

Comparing the DscCO2– with the WscCO2–water displacement experiments, it was observed that at low injection rates, the scCO2 sweep efficiency was much higher in the DscCO2 displacement experiments than that in the WscCO2 displacement experiments. During the stage of capillary fingering, the DscCO2 appears to enhance the lateral and back migration of CO2 in the micromodel. As a result, the drying effects of DscCO2 combined with the capillary forces are expected to enhance capillary fingering substantially, so that the DscCO2 can positively and forwardly spread and occupy much more pore spaces (Wang et al. 2013a). On the contrary, the dissolved water in WscCO2 appears to limit lateral migration using narrow flow paths, and WscCO2 occupies less pore spaces than DscCO2. At medium and high injection rates, both show similar trends. The crossover and viscous fingering mechanisms were prominent, leading to similar distribution patterns of DscCO2 and WscCO2 displacements. Under condition of 1.0 PV CO2 injection, DscCO2 might have had limited time of contact with water, affecting the interfacial boundary. Thus, DscCO2 would not impact the stages of crossover and viscous fingering patterns dramatically. The dry-out effects from DscCO2 in such short period or with such small volume injection (1.0 PV) can be ignored due to the small impacts on the displacement processes under the patterns dominated by crossover and viscous fingering.

3.2 scCO2 saturation

The relationship between the SscCO2 of the above images (Fig. 1a–c) and the injection rate (logCa) for the above three sets of DscCO2– and WscCO2–water displacement experiments are shown in Fig. 2. The SscCO2 of all experiments are listed in Table S2 (SI).

Fig. 2
figure 2

scCO2 saturation vs. injection rate (logCa) for the three sets of DscCO2– and WscCO2–water displacement experiments

Full size image

From Fig. 2 and Table S2, in DscCO2–water displacement experiments, SscCO2 first increased with increasing logCa during the stage of capillary fingering when the flow pattern was controlled by capillary forces coupled with drying effects from DscCO2 (Fig. 2, upper three curves). Then SscCO2 decreased sharply as logCa increased further to the stage of crossover. The decrease is caused by the transition from broad capillary fingers to narrow viscous fingers shortly after scCO2 entered the pore network (Fig. 1). Finally, when viscous fingering dominates the unstable flow at high injection rates, SscCO2 increased again as logCa increased. The observation is consistent with previous studies (Zhang et al. 2011a, b; DeHoff et al. 2012; Wang et al. 2013a).

In WscCO2–water displacement experiments, SscCO2 increased generally with injection rates (shown in Fig. 2, lower three curves). First, it increased a little when the flow pattern of crossover was observed (i.e., at low and medium logCa) and then was followed by a large increase when viscous fingering was observed (i.e., at high logCa). When the displacement was at viscous fingering dominating pattern, SscCO2 can increase by 0.35 ~ 0.47 per logCa. But when the displacement is at a crossover pattern, SscCO2 increases 0.04 ~ 0.09 per logCa. It was considered that the dissolved water in scCO2 hindered scCO2 from extracting water any more (i.e., no more drying effects) during the process of WscCO2–water displacement.

We also investigated the influence of the dissolved water in WscCO2 and the drying effects from DscCO2 on SscCO2 by comparing the saturation fractions, which were determined by the differential saturation between each pair of DscCO2– and WscCO2–water displacement saturations. The decreased SscCO2 fraction due to block of the dissolved water for the three sets of DscCO2– and WscCO2–water displacement experiments are shown in Figure S2(a) (SI). The enhanced SscCO2 fraction due to drying effects in all three sets of experiments is shown in Figure S2(b) (SI). The influences of drying-out effect of DscCO2 and dissolved water in WscCO2 are mainly reflected in capillary fingering preferential periods. At the low injection rates, DscCO2 saturation increased up to 4.4 times as compared to that of WscCO2 (Figure S2(a)) and WscCO2 saturation was decreased 0 ~ 0.77 in proportion to that of DscCO2 (Figure S2(b)). However, at medium and high injection rates, SscCO2 is approximately the same in DscCO2– and WscCO2–water displacement experiments when crossover and viscous fingering dominate the two-phase flow. Because the dry-out effects from DscCO2 are limited in such short contact duration between CO2 and water, they make little impact on the displacement processes when the flow patterns are dominated by crossover and viscous fingering.

3.3 Specific interfacial area

The specific interfacial area between scCO2 and water can be approximated by normalizing the surface area of the scCO2 phase obtained from the micromodel images by the total area of the pore network, on the basis that the wetting phase (water) surrounds the solid grains (Zhang et al. 2011a, b). The relations of interfacial area vs. injection rate are shown in Fig. 3. The results of each interfacial area for the three sets of experiments are listed in Table S2 (SI).

Fig. 3
figure 3

scCO2 specific interfacial area vs. flow rate (logCa) for the three sets of DscCO2– and WscCO2–water displacement experiments

Full size image

From Fig. 3 and Table S2, in DscCO2–water displacement experiments, the specific interfacial area of the DscCO2 displacements first increased slowly when the flow pattern was controlled by capillary forces coupled with drying effects from DscCO2. Then a large decrease in the DscCO2 specific interfacial area was observed in the crossover scenario. When viscous fingering dominated the unstable flow at high injection rates, the DscCO2 specific interfacial area increased again (Fig. 3, upper three curves). In WscCO2–water displacement experiments, the specific interfacial areas were relatively low with a slight fluctuation in the experiments of low and medium injection rates and increased subsequently at high injection rates (Fig. 3, lower three curves).

The influence of the dissolved water in WscCO2 and the drying effects from DscCO2 on the scCO2 specific interfacial area were also investigated by comparing scCO2 specific interfacial area fractions. Those specific interfacial area fractions were determined by the differential specific interfacial area between each pair of DscCO2– and WscCO2–water displacement experiments. The decreased scCO2 specific interfacial area fractions due to the dissolved water for the three sets of DscCO2– and WscCO2–water displacement experiments are shown in Figure S3(a) (SI), and the enlarged scCO2 specific interfacial area fractions due to the drying effects are shown in Figure S3(b) (SI). The maximum WscCO2 specific interfacial area fraction due to the dissolved water was reduced to 0.9. The maximum DscCO2 specific interfacial area fraction due to the drying effects increased up to 16 times compared to that of WscCO2.

Thus, the drying effects of DscCO2 combined with capillary forces play a dominate role in specific interfacial area as well as scCO2 saturation enhancement when capillary fingering dominates the displacement process. Conversely, the dissolved water in WscCO2 adversely influences WscCO2–water displacement, reducing the specific interfacial area and saturation of WscCO2. Thus, the WscCO2 specific interfacial area was lower than that of DscCO2 considerably at lower injection rates. The drying effects from DscCO2 enhanced the specific interfacial area, especially for Set II where the temperature was slightly higher, resulting in a larger dissolved water solubility. In the pattern of crossover from capillary to viscous fingering, a sharp drop in the value of DscCO2 specific interfacial area was observed, indicating the drying effects cannot play an effective role any longer. In addition, From Figs. 2 and 3, the trends of each curve for CO2 saturation and the corresponding curve for CO2 specific interfacial area are almost the same, indicating that the larger the CO2 specific interfacial area is, the more the CO2 saturation is, vice versa.

3.4 scCO2 mobility analysis

Two fractions of the scCO2 phase in the pore network can be identified from the fluorescent images: a mobile fraction forming active flowpaths through the entire length of the network, and a trapped fraction consisting of clusters of scCO2 disconnected from the active flowpaths (Wang et al. 2013a). For example, Figure S4 (SI) shows the active flowpaths caused by scCO2 displacing water in micromodels. Figure 4 shows the volume fractions of mobile scCO2 at different injection rates. Figure S5(a) and (b) (SI) shows the decreased (due to presence of the dissolved water) and increased (due to the drying effects) scCO2 volume fraction in active flowpaths across the pore network with the injection rate (logCa) for the three sets of DscCO2– and WscCO2–water displacement experiments. The results of volume fraction of mobile scCO2 for each experiment are listed in Table S2.

Fig. 4
figure 4

scCO2 volume fraction in active flow paths across the pore network vs. flow rate for the three sets of DscCO2– and WscCO2–water displacement experiments

Full size image

From Fig. 4 and Table S2, at low injection rates in DscCO2–water displacement experiments, where the displacement process was dominated by capillary fingering combined with drying effects from DscCO2, the mobile fraction accounted for more than 95% of scCO2 in the pore network, with the remainder trapped as blobs. At high injection rates, where viscous fingering was the dominant displacement mechanism, approximately 89 ~ 98% of scCO2 was in active flow paths, with the rest as immobile blobs, irrespective to DscCO2– or WscCO2–water displacement processes. In the crossover from capillary to viscous fingering, differences between the DscCO2– or WscCO2–water displacements were also small. For DscCO2–water displacements, the mobile fraction was 66.5–80.9%, while values of 19.1–33.5% obtained from the main flow paths as inactive fingers. In the WscCO2–water displacement experiments, the mobile fraction accounted for 55–89% of scCO2 in the pore network, with 11–45% as immobile blobs when the displacement process was dominated by crossover. At logCa = − 4.64 (40 °C), almost 89% of scCO2 was in active flowpaths, which means only 11% was as immobile blobs in the pattern of viscous fingering. From Figure S4 (a) and (b), slight differences in scCO2 mobile fraction were observed between WscCO2– and DscCO2–water displacement processes. The variations of the mobile fractions for experiments Sets I and II under 40 °C were similar, however the trend for experiments Sets II and III at 35 °C differred greatly.

For experiment Set III, scCO2 mobile fraction accounted for the highest (> 90%) in DscCO2—water and the lowest (< 25%) in WscCO2–water displacement at low flow rates. The scCO2 mobile fraction decreased by up to 70% due to the dissolved water in WscCO2 and increased by as high as 3.7 times due to the dry-out effects from DscCO2. Large fractions of scCO2 in active flow paths could potentially influence mutual dissolution between the injected CO2 phase and trapped water, causing a large initial increase in water acidity and resulting in dry-out of the formation brine. Meanwhile, the drying effects from DscCO2 further enhanced the scCO2 volume fraction in active flow paths.

4 Implications

Due to the unfavorable viscosity ratio (logM < 0), unstable displacement of formation brine by scCO2 has been identified as one of the leading causes for the low storage efficiency factor of carbon storage in saline aquifers. Results from the DscCO2–water displacement experiments indicated that, consistent with the results of our previous study (Metz et al. 2005), there are three mechanisms that control DscCO2–water two-phase flow at the pore scale: (i) capillary fingering at low injection rates, (ii) crossover from capillary to viscous fingering at medium injection rates, and (iii) viscous fingering at high injection rates. However, for WscCO2–water displacements, only two mechanisms were observed: (i) crossover at low and medium injection rates, and (ii) viscous fingering at high injection rates. Capillary fingering might occur only at far lower injection rates [e.g., logCa < − 7.64 (40 °C)]. With the capillary force combined with the drying effect from DscCO2 affecting the scCO2–water displacement process, larger saturation, specific interfacial areas and mobile of DscCO2 fractions were observed when the capillary fingering dominates the displacement process. The capillary forces combined with drying effects also were identified as one of the leading causes for obtaining a higher scCO2 sweep efficiency.

Results from this study demonstrate the importance of understanding pore-scale displacement mechanisms and the effects of water dissolution on interface of two-phase flow and eventually on sweep efficiency of CO2. The drying effects from DscCO2 would prompt to extract water from interfacial surface between CO2 and water, break the interfacial surface and allow CO2 to penetrate through current pore throat and to the next pore throat as well as to form a new interface whose specific interfacial area, for example, changes from 28.239 to 96.933 cm− 1 in Set I. Thus, such mutual solubility obviously modifies the radius of wetted perimeter of CO2 in the micromodel. The specific interfacial length between CO2 and water doubled and even tripled during the capillary fingering process. In this way, the saturation and sweep efficiency factor of CO2 have increased dramatically. Furthermore, the dry-out effects of CO2 can alter capillary boundary or “capillary pressure lock” at the pore throat, enhancing the CO2 displacement efficiency and CO2 saturation. One of the reasons why dry-out effect is more efficient in capillary fingering scenario might be that at the scale of micromodel, the slow CO2 flow is capable of carrying a lot of water moisture out of the micromodel and enhancing the processes of mutual solution and capillary fingering in the micromodel. As a result of variation of water–CO2 interface at pore scale and difference of CO2 distribution at microscale caused by dry CO2, the rock properties measured by laboratory experiments, such as capillary pressure curve and relative permeability, should be used carefully for different state of CO2. Previous literatures have shown that these properties are dependent on lots of factors, such as flow velocity, geochemistry and flow rate history (Wang et al. 2013b; Loring et al. 2011; Loring et al. 2012; Kazemifar et al. 2016; Roman et al. 2016; Takeshi et al. 2016). Besides, this study implies that the state of CO2, whatever dry or wet, will certainly affect these properties. Therefore, conventional calibration curves of capillary pressure and relative permeability obtained without consideration of CO2 state might not adequately describe the complete CO2–brine displacement process, especially near the wellbore. Thus, proper modifications to the current models are needed to accurately account for the effect of CO2 saturation enhancement caused by the dry-out effect. For instance, to design an appropriate injection scheme, it will be important to (i) measure the rock properties in both dry or wet scCO2; (ii) utilize dry-out effects in the flow patter of capillary fingering so as to enhance the CO2 saturation and storage efficiency near the borehole or a certain distance from the borehole; and (iii) build more detailed models considering the dry-out effects and the migration of CO2.

It should be noted that this study was performed in a nearly homogeneous 2D pore network with well-controlled surface properties. The heterogeneity from the fabrication process likely contributed a little to the complex fingering shape of various migration patterns of CO2 during the displacement processes. Large variations in pore geometry with different surface wettability, stemming from various lithology and mineralogy, can be expected in geological formations, which will have more pronounced impacts on pore level CO2 saturation and overall storage capacity.