Liquid CO2 Displacement of Water in a Dual-Permeability Pore

ARTICLE pubs.acs.org/est

Liquid CO2 Displacement of Water in a Dual-Permeability Pore Network Micromodel Changyong Zhang,* Mart Oostrom, Jay W. Grate, Thomas W. Wietsma, and Marvin G. Warner Pacific Northwest National Laboratory, 902 Battelle Boulevard, P.O. Box 999, MSIN K8-96, Richland, Washington 99352, United States

bS Supporting Information ABSTRACT: Permeability contrasts exist in multilayer geological formations under consideration for carbon sequestration. To improve our understanding of heterogeneous pore-scale displacements, liquid CO2 (LCO2)
water displacement was evaluated in a pore network micromodel with two distinct permeability zones. Due to the low viscosity ratio (logM =
1.1), unstable displacement occurred at all injection rates over 2 orders of magnitude. LCO2 displaced water only in the high permeability zone at low injection rates with the mechanism shifting from capillary fingering to viscous fingering with increasing flow rate. At high injection rates, LCO2 displaced water in the low permeability zone with capillary fingering as the dominant mechanism. LCO2 saturation (SLCO2) as a function of injection rate was quantified using fluorescent microscopy. In all experiments, more than 50% of LCO2 resided in the active flowpaths, and this fraction increased as displacement transitioned from capillary to viscous fingering. A continuum-scale two-phase flow model with independently determined fluid and hydraulic parameters was used to predict SLCO2 in the dual-permeability field. Agreement with the micromodel experiments was obtained for low injection rates. However, the numerical model does not account for the unstable viscous fingering processes observed experimentally at higher rates and hence overestimated SLCO2.

’ INTRODUCTION Capture of CO2 from large stationary emission sources and sequestration in depleted oil and gas reservoirs, unminable coal seams, and deep saline aquifers is being intensively studied as a strategy to mitigate CO2 emission into the atmosphere.1 In practice, CO2 in liquid or supercritical state is injected into a porous subsurface sedimentary formation and physically displaces the formation fluids such as water (brine), oil, or natural gas. The displacement process is directly affected by the physical and chemical properties of both the injected CO2 and displaced formation fluids (e.g., viscosity, interfacial tension, density, buoyancy, solubility), hydrodynamic forces such as capillary pressure endured during the process, and physical and chemical properties of the porous matrix (e.g., pore size or permeability distribution, wettability of the solid surface). The displacement stabilities, CO2 flow pathways, and saturation levels are of critical importance to better understand the trapping mechanisms of CO2 in geological media,2 aquifer storage capacity,3 as well as caprock sealing efficiency.4 Mathematical models have been developed and applied to evaluate multiphase flow, mixing and dissolution in formation brine, porous matrix heterogeneity, and trapping mechanisms involved in field scale deployment of CO2 sequestration.2,5
12 Among various factors being studied, formation heterogeneity (variation in permeability distribution) is one of the leading variables having direct impacts on aquifer storage capacity. Zhou et al.5 modeled CO2 injection into the Mt. Simon aquifer in the Illinois basin and showed that CO2 migration and plume development are highly affected by the contrasting permeability fields r 2011 American Chemical Society

in the multilayer geological formation. Kuuskraa et al.8 simulated scenarios where the CO2 injection reservoir has multiple highly permeable sandstone layers separated by low permeability shale. They showed that CO2 preferentially migrated through the high permeability layers with little to no flow through the low permeability layers.8 Han et al.12 used numerical simulations to show that residual CO2 trapping is reduced with an increasing degree of heterogeneity in correlated heterogeneous permeability fields. These studies have shown the importance of formation architecture (e.g., heterogeneity) in affecting CO2 flow and storage at the field scale. Recent laboratory studies have also focused on elucidating the role of textural heterogeneity on CO2
water or CO2
oil displacement mechanisms by using various visualization techniques to characterize phase distributions in porous media such as core samples. Several studies applied X-ray computed tomography or magnetic resonance imaging to investigate immiscible CO2
water or CO2
oil displacement in rock core samples, and showed the strong influence of subcore scale heterogeneities on steady-state migration patterns, spatial distributions, and saturation of CO2.13
16 For example, Perrin and Benson13 and Zhao et al.15 observed CO2 channeling through high porosity regions during immiscible CO2
water and CO2
oil displacement. Shi et al.14 showed a correlation Received: May 31, 2011 Accepted: July 20, 2011 Revised: July 12, 2011 Published: July 20, 2011 7581

dx.doi.org/10.1021/es201858r | Environ. Sci. Technol. 2011, 45, 7581–7588

Environmental Science & Technology

ARTICLE

Figure 1. Schematic of the dual-permeability micromodel and displacement experimental setup.

between CO2 saturation and subcore scale porosity heterogeneity, which diminished above a certain injection rate. In addition to porous media heterogeneity, viscous and capillary forces are also expected to play significant roles in affecting CO2
water displacement and subsequent CO2 distribution in porous media. Injected CO2 has a much lower viscosity than formation water (5
20% of the water viscosity),6 and the interfacial tension between CO2 and formation water is also found to be dependent on temperature, pressure and salinity.17
19 Dimensionless numbers are frequently used to describe different forces encountered by the two immiscible fluids during displacement inside pores. The capillary number, Ca = (μn  un)/(σnw  cosθ), is commonly used to describe the ratio between viscous forces to capillary forces experienced by a nonwetting fluid during immiscible fluid displacement.20 Here, μn is the viscosity of the advancing nonwetting fluid (subscript n), un is the velocity of the advancing nonwetting fluid, σnw is the interfacial tension between the wetting (subscript w) and nonwetting fluid, and θ is the fluid
fluid contact angle with the solid surface. The viscosity ratio between the advancing nonwetting fluid and the wetting fluid being displaced from the pores (M = μn/μw) is often used to describe the favorability of the displacement process,20 with logM > 0 and logM < 0 indicating favorable and unfavorable displacement, respectively.

Mechanistic understanding of immiscible fluid displacement in porous media can be gained through pore-scale micromodel experiments and numerical simulations.20
25 Lenormand et al.20 developed a numerical model and predicted the existence of three domains when a wetting fluid is displaced by a nonwetting fluid inside a two-dimensional (2D) pore network: unstable displacement including either capillary fingering or viscous fingering, and stable displacement. Capillary fingering occurs under low injection rates for a large range of viscosity ratios.20,25 Viscous fingering typically occurs during displacement at low viscosity ratios, and has been widely recognized.20,26
31 Several experimental studies also noted the difference between viscous fingering and capillary fingering.20,21,23,25 Stability diagrams identifying two fingering zones and a stable displacement zone have been constructed based on experiments performed under ambient conditions over a large range of Ca and M in oil-wet micromodels20 and water-wet micromodels,25 although the exact locations of the various boundaries and crossover zones were shown to be dependent on the pore network.25 So far, only a limited number of micromodel studies32
36 have been reported focusing on fundamental processes at the microscopic pore scale during CO2 injection into a porous formation. For instance, Er et al.34 studied porous matrix and fracture 7582

dx.doi.org/10.1021/es201858r |Environ. Sci. Technol. 2011, 45, 7581–7588

Environmental Science & Technology

ARTICLE

Table 1. Summary of Micromodel Properties, Fluid Properties and Experimental Conditionsa experimental conditions and fluid properties

micromodel properties

a

length [cm]

1.22

pressure [MPa/psi]

9.0/1305

width [cm]

1.21

temperature [°C]

22 ( 1

depth [μm]

35

viscosity (LCO2/water) [mPa 3 s]

0.076b/0.982c

porosity (H/L) [
]

0.3987/0.3987

logM [
]


1.1

grain diameter (H/L) [μm]

200/100

interfacial tension [mN 3 m
1]

28.7d

pore body length (H/L) [μm]

120/60

water
LCO2 contact angle [deg]

15.3 ( 1.9

pore throat length (H/L) [μm]

26.7/13.3

density (LCO2/water) [g/cm3]

0.825b/0.998c

hydraulic conductivity (D/H/L) [cm/s] LCO2 entry pressure (H/L) [Pa]

0.0089/0.0118/0.0066 3825/5956

injection rate [μL/h]

100, 250, 500, 1000, 2500, 5000, 7500, 10 000

H: high permeability; L: low permeability; D: dual permeability. b Nordbotten et al., 2005.7 c Clark, 1996.40 d Bachu and Bennion, 2009.17

interaction during CO2 injection into a glass micromodel initially saturated with oil, demonstrating the importance of CO2
oil interaction near the matrix
fracture interface. Chalbaud et al.33 visualized liquid and supercritical CO2 displacement of water in glass micromodels but saturations could not be quantified because of the difficulties in differentiating fluids. Riazi et al.36 studied CO2
water and CO2
oil displacement in an etched glass micromodel and showed faster CO2 breakthrough in the micromodel saturated with oil than with water. These studies are all limited to qualitative visualizations of the displacement mechanisms and quantitative information on fluid saturations and interfacial areas was not obtained. These studies also did not systematically investigate displacement mechanisms and CO2 saturations as a function of Ca and M. In this study we demonstrate a novel methodology using precisely dry-etched micromodels, fluorescently dyed liquid CO2 (LCO2), and fluorescence microscopy to quantitatively characterize water displacement by LCO2 at the pore scale. Specifically, we designed a high pressure micromodel system to investigate LCO2 displacement of water at 9 MPa in a dual-permeability micromodel with well-defined pore-scale features as a function of injection rate (i.e., Ca). LCO2 spatial distributions were imaged using fluorescent microscopy, and quantitative information on LCO2 saturations and LCO2
water interfacial areas were derived from these distributions for injection rates ranging 2 orders of magnitude. The experimental results were compared with numerical simulation results obtained with a continuum-based two-phase flow model using independently determined parameter values for porosity, permeability, and retention parameters. Comparison with continuum-based model predictions provides insights and an improved understanding of processes involved in geological carbon sequestration. To our knowledge, this study is the first to quantitatively evaluate the impacts of porous media heterogeneity and capillary forces on CO2
water displacement at the microscopic pore scale at a pressure greater than the CO2 phase-transition pressure of 7.3 MPa. By conducting the displacement studies at a high pressure, this study also represents an important first step toward elucidating the complex interfacial phenomena at the pore scale under dynamic flow conditions that characterize CO2 injection into deep saline aquifers. The interfacial phenomena occurring under reservoir conditions, including high pressures and elevated temperatures, cannot be reproduced using surrogate immiscible fluid pairs under ambient conditions.

’ MATERIALS AND METHODS Micromodels. A previously reported microfabrication process involving standard photolithography, inductively

Figure 2. Water
LCO2 contact angle with thermally oxidized silicon surface as measured in the micromodel outlet channel.

coupled plasma-deep reactive ion etching (ICP-DRIE), thermal oxidation and anodic bonding techniques 37,38 was used to fabricate micromodels in silicon wafers. The micromodel used in this study consists of one inlet channel, connected to a pore network (1.22  1.21 cm), and one outlet channel (Figure 1a). The main pore network contains two distinct permeability zones each occupying one-half of the micromodel width. A thin oxide layer (∼0.1 μm) on the surface of the fabricated micromodel was obtained using a thermal oxidation technique (1100 °C under oxygen environment for approximately 1 h). Finally, a cover sheet of Pyrex glass was anodically bonded to the etched silicon micromodel in order to seal the flow channels. A digital output pressure transducer (Heise DXD 0
5 psi range, 0.02% accuracy, Heise Inc., Stratford, CT) was used to measure the pressure drop across the micromodel under different flow rates and the hydraulic conductivity of the micromodel was determined according to Darcy’s law. The hydraulic conductivities of two separately fabricated micromodels containing homogeneous arrays of large (200 μm) and small (100 μm) cylindrical posts were also determined. All pertinent micromodel properties are listed in Table 1. The entry pressures for the high and low permeability zones are estimated using the Young
Laplace equation for capillary pressure (Pc):   1 1 Pc ¼ þ σnw cos θ r1 r2 7583

dx.doi.org/10.1021/es201858r |Environ. Sci. Technol. 2011, 45, 7581–7588

Environmental Science & Technology where r1 and r2 are the measured radii of the pore throats (half the values listed in Table 1). The water
LCO2 contact angle (θ) was determined in situ using images of LCO2 blobs trapped in the outlet channel of the micromodel (Figure 2). The blob is in contact with the vertical wall of the channel with water filling the void space, and the contact angle with the micromodel surface can be determined as shown in Figure 2. A total of 18 measurements were made from images acquired in different experiments, and the average value is 15.3 degrees when measured from the water phase, with a standard deviation of (1.9 degrees. The average contact angle value was also used in the calculation of Ca. LCO2
Water Displacement Experimental Setup. A schematic of the micromodel experiment setup is shown in Figure 1. A stainless steel overburden pressure cell was constructed with a sapphire sight window (viewable diameter = 3.6 cm). The micromodel was placed horizontally on the sapphire window in the overburden pressure cell. A sufficient amount of degassed deionized (DI) water was flushed through the micromodel until it was fully saturated with water. LCO2 was injected through the micromodel inlet using Pump A (Teledyne ISCO 100 DM, Teledyne ISCO Inc., Lincoln, NE) at a constant volumetric flow rate (Q) ranging from 100 μL/h to 10,000 μL/h (Table 1), corresponding to a Darcy velocity (q) between 5.7and 570 m/day. The micromodel outlet is connected to a back pressure Pump B (Teledyne ISCO 100DM) filled with DI water. The pressure cell was filled with glycerol (Sigma Aldrich, St. Louis, MI) and connected to an overburden pressure Pump C (Teledyne ISCO 500D). For each injection rate, the LCO2 was allowed to displace water in the micromodel pore network and the displacement process was monitored in real-time using an epifluorescent microscope and camera (described in the next section). An image of the entire pore network was acquired after LCO2 reached the outlet of the micromodel. All experiments were performed in the same micromodel with three replicates at each injection rate. After each experiment, the micromodel was thoroughly cleaned using the following sequence: isopropanol, DI water, and a basic solution (DI water:NH4OH:H2O2 at 5:1:1) to effectively restore the micromodel surface properties. To facilitate visualization and quantification of LCO2 distribution in the micromodel using epifluorescent microscopy, a fluorescent dye, Coumarin 153 (99.99%; Alfa Aesar, Ward Hill, MA) was used to dye the LCO2 phase. Details of fluorescent absorption and emission properties of Coumarin 153 has been reported in a previous study.39 A stock solution of Coumarin 153 was prepared in methanol at a concentration of 10 mM. Prior to the experiment, 1 mL stock solution was added to the cylinder of Pump A, and methanol was evaporated by purging the cylinder with CO2 gas for an extended period of time (>10 h). Next, the cylinder was filled with supercritical grade CO2 to approximately 850 psi while Valve 1 connecting Pump A to the micromodel was closed. Pump A was then pressurized to 1305 psi and the temperature was allowed to equilibrate with room conditions for at least 2 h. The pressure inside the micromodel and the overburden pressure cell were sequentially increased to 1305 and 1390 psi, respectively, by simultaneously pressurizing Pumps B and C at 25 psi increments over approximately a 1.5 h interval. Due to the horizontal two-dimensional (2D) nature of the displacement process in the micromodel, density and buoyancy effects were not considered in this study. Table 1 provides the fluid properties of LCO27,17 and water,40 as well as experimental conditions. For comparison, the viscosity ratio (logM) between supercritical CO2 and water at 40 °C is approximately
1.2, which is not

ARTICLE

substantially different from the LCO2
water (logM =
1.1) considered in this study. Microscope Imaging and Quantitative Image Analysis. The pressure cell was placed horizontally on an automated microscope stage (Prior Scientific Inc., Rockland, MA) with the sapphire sight window and micromodel pore network side facing down. LCO2 injection was started by opening Valve 1 and switching Pump A (containing LCO2 at 1305 psi) to a constant flow mode at a given injection rate. All micromodel images were acquired at 1.6 μm spatial resolution using a Nikon Eclipse-2000TE (Nikon, Melville, NY) epifluorescent microscope with a 4 inverted objective. The microscope is connected to a monochrome CCD camera, controlled by a computer and imaging software (NIS-Elements; Nikon, Melville, NY). Fluorescent images of LCO2 were obtained through a Blue GFP filter set (λex = 379
401 nm, λem = 435
485 nm). A total of 48 (6  8) separate images were taken at any observation stage of the displacement process and were then montaged to form a single image that captured the entire pore network. To quantitatively evaluate the LCO2 saturation (SLCO2) in the pore network, montaged images of the pore network were analyzed by thresholding and pixel counting. The fluorescent signal intensity of pixels containing LCO2 is significantly higher than that for silicon posts and pore spaces filled with water (signal-to-noise ratio >10). A threshold value can be unambiguously determined for each image to distinguish pixels consisting of LCO2 from silicon posts and pore space filled with water. From the number of pixels, the LCO2 saturation (i.e., area) and surface area (i.e., perimeter) can be determined. The LCO2 perimeter is used as a measure for interfacial length between LCO2 and water, including liquid water films surrounding solid grains after drainage of the pore throats and bodies.33,41 Numerical Simulation Using a Two-Phase Continuum Model. Continuum-based models used for subsurface CO2 injection include buoyancy effects but ignore unstable displacement induced by viscosity gradients. To assess the repercussions of omitting unstable displacement on fluid saturations, the average experimentally derived saturations were compared with predicted results using the water
oil mode of the integrated finite difference STOMP simulator.42 A similar method was also used by Willingham et al.43 to simulate solute transport in a micromodel using a continuum-based model. The computational domain for the simulations represents the 2D micromodel system shown in Figure 1. To ensure that these simulations would not amount to a mere fitting exercise, independently determined values for fluid properties were used, as well as porosity, permeability, and fluid retention parameters values for the high and low permeability zones (Table 1). To simulate the distinct nonwetting fluid entry behavior associated with the type of micromodel used in this study, the Brooks-Corey44 saturation - capillary pressure equation in combination with the Burdine relative permeability relation45 was used. The entry pressure in the Brooks-Corey relation was computed from available information on interfacial tension and pore radii (Table 1). To be consistent with Schroth et al.,46 who found that pore geometry factor (λ) values >7 are appropriate for highly uniform porous media, a value of 8 was used. Sensitivity simulations indicated that computed saturations have converged at this value, with only minimal changes ( 8. Consistent with proper multiphase modeling practices,42 the computational domain was discretized into increasingly smaller uniform cells until the predicted average saturations converged. Cell 7584

dx.doi.org/10.1021/es201858r |Environ. Sci. Technol. 2011, 45, 7581–7588

Environmental Science & Technology

ARTICLE

Figure 3. Selected images of LCO2 (white) distribution in the dual-permeability pore network with water as the initial wetting fluid at different injection rate (expressed in terms of logCa). (Flow direction of LCO2 is from left to right).

dimensions of 0.025 cm in the x- and y-direction were used for the simulations as further refinements altered the computed LCO2 saturations for all considered injection rates by less than 1%. Initially, the system pressure was set to 9.0 MPa. The LCO2 was then injected into a high-porosity, high-permeability inlet channel located to the left of the domain (Figure 1) using the same injection rates applied in the experiments until a steadystate LCO2 saturation distribution was obtained for each of the rates. During the injection phase, the fluid pressure was kept at 9.0 MPa at the outflow boundary.

’ RESULTS AND DISCUSSION Displacement Mechanisms of Water by LCO2. Representative fluorescent images of the LCO2 distribution in the micromodel pore network at selected injection rates (expressed as logCa) are shown in Figure 3 (additional images are included in the Supporting Information (SI)). Comparison of images from three replicate experiments showed qualitative agreements (quantitative results are discussed in the following section). These results demonstrated preferential displacement through the high permeability zone, and qualitatively showed characteristics of different displacement mechanisms, that is, capillary and viscous fingering. At low injection rates (logCa e
3.96), displacement occurs only through the high permeability zone because the higher capillary force in the low permeability zone resists penetration by the nonwetting LCO2; this force is eventually overcome at higher injection rates (logCa g
3.66) as the pressure drop increases across the more viscous water phase in the pore network. At low injection rates (logCa =
5.36 and
4.96), lateral LCO2 flow (i.e., transverse to the main flow direction) is evident in the high permeability zone with large entrapped zones of water, indicating capillary fingering.20
23 For higher injection rates (logCa >
4.66), multiple LCO2 flowpaths primarily progressed forward across the pore network with limited or no lateral flow, which are characteristics of viscous fingering.20,24 At logCa =
3.66, LCO2 entered the low permeability zone through the inlet channel but did not completely break through, suggesting the displacement pressure is large enough to overcome the capillary pressure of water but not sufficient enough to maintain continuous flow through this zone. The distribution patterns of LCO2 in the low permeability zone also showed dominant features of capillary fingering with obvious lateral and backward flow. At the two highest injection rates

(logCa =
3.48 and
3.36), complete LCO2 breakthrough in the low permeability zone occurred as the imposed increase in the injection rate resulted in a sufficient increase in the displacement pressure so that continuous flow through this zone could be sustained. The LCO2 flowpaths in the low permeability again are dominated by capillary fingering indicating that the flow rate through this zone is lower than the high permeability zone. It is worth noting that LCO2 did not cross the boundary between the two permeability zones under low injection rates. At the higher injection rates, LCO2 flow in the low permeability zone originated from the inlet channel and fluid connections to the LCO2 in the high permeability zone can be found only at a few isolated locations (Figure S2 in the SI). Effects of Injection Rate on LCO2 Saturation. LCO2 saturations (SLCO2) in the micromodel pore network vs. logCa are shown in Figure 4. Results from three replicate experiments showed good quantitative agreement with differences within 7% for all cases except for logCa =
3.66, in which incomplete breakthrough in the low permeability zone resulted in variation in SLCO2 greater than 10%. SLCO2 in the entire pore network increased as injection rate increased (increasing logCa): the increase is relatively small (20%) were observed between logCa =
3.66 and
3.36 when LCO2 flow in the low permeability zone occurred. The rate of SLCO2 increase in the high permeability zone is small between logCa =
5.36 to
3.96, because unstable displacement through capillary and viscous fingering results in a slow increase in nonwetting fluid saturation. The greater SLCO2 increase between logCa =
3.66 to
3.36 is attributed to the displacement of large trapped water blobs by viscous fingers at higher injection rates. At the two highest injection rates when complete LCO2 breakthrough in both permeability zones occurred, SLCO2 in the low permeability zone are significantly lower (∼20%) than the high permeability zone, which can be attributed to the different displacement mechanisms: viscous fingering for high flow rates in the high permeability zone vs. capillary fingering for low flow rates in the low permeability zone. The observed displacement mechanisms on LCO2 saturations are consistent with the logCa
logM stability diagram presented by Zhang et al.25 for immiscible displacement studies in micromodels under ambient conditions. According to that diagram, displacement for logM =
1.1 for the range of logCa investigated in this study should range from a mixture of capillary and viscous fingering at low injection rates to viscous fingering at high injection rates. 7585

dx.doi.org/10.1021/es201858r |Environ. Sci. Technol. 2011, 45, 7581–7588

Environmental Science & Technology

Figure 4. Experimental and simulated LCO2 saturation vs. injection rate (expressed as logCa) in the high (SnH) and low permeability zones (SnL), and averaged over the entire pore network (Sn). Solid lines represent average saturation of three replicate experiments at each injection rate (represented by solid symbols).

Figure 5. Volume fractions of LCO2 residing in active flowpaths across the pore network, multipore blobs and small blobs as a function of injection rate.

Furthermore, the LCO2 saturations in the high permeability zone in this study and dodecane saturations in a system with logM =
1.34, observed in the ambient pressure studies by Zhang et al.,25 are comparable for similar values of logCa. The similarity in displacement mechanisms and fluid saturations for the same range in logCa and logM at high pressures is an important finding. LCO2 Distribution and Morphology. The use of Coumarin 153 as a fluorescent dye and high resolution fluorescent microscopy also made it possible to expand our analyses to quantification of LCO2 distributions and morphology. To better understand the LCO2 distribution in the pore network in relationship to the LCO2
water displacement mechanisms, the domains of LCO2 in the pore network were divided into three groups: (i) mobile phase: LCO2 residing in active flowpaths crossing the entire length of the pore network, (ii) large multipore blobs: LCO2 blobs trapped in the pore network occupying greater than five pore bodies, and (iii) small blobs: LCO2 blobs trapped in the pore network occupying less than five pore bodies. The volumetric fractions of each of the three groups were determined through image analysis and results are shown in Figure 5. At lower injection rates

ARTICLE

Figure 6. LCO2
water specific interfacial length as a function of LCO2 saturation in the dual-permeability micromodel.

(logCa =
5.36 ∼
4.66), when LCO2
water displacement is dominated by capillary fingering (Figure 3), the mobile phase accounts for approximately 50
60% of LCO2 forming active flowpaths, whereas 30
40% of the LCO2 was trapped as large multipore blobs, and less than 5% was trapped as small blobs. At logCa greater than
4.36, viscous fingering is the more dominant displacement mechanism and the majority (>90%) of LCO2 was in active flowpaths, with less than 10% in immobile small blobs. The fraction of multipore blobs was higher (20%) at logCa =
3.66 and this can be explained by the incomplete breakthrough of capillary fingers in the low permeability zone. Results from this analysis indicate that during the active injection stage, a large fraction (>50%) of the injected CO2 reside in active flowpaths, which may result in accelerated mass transfer between trapped water and injected CO2 and potentially cause formation brine solution to dry out.47 LCO2
Water Interfacial Area. The fluid
fluid interfacial area formed between CO2 and formation fluid such as brine is another important variable because it influences the interfacial mass transfer processes such as dissolution and subsequent geochemical reactions that have implications for long-term CO2 trapping in mineral carbonate phases.48 In the present study, due to the 2D nature of the micromodel, it is appropriate to use the interfacial length that can be determined from image analysis. Specific interfacial length was obtained by normalizing the LCO2 perimeter by the total pore area, and the results are shown in Figure 6. For reference, the maximum interfacial length of the dual-permeability pore network is 460 cm
1. For low injection rates, when LCO2 only displaced water in the high permeability zone (SLCO2 < 33%), the specific interfacial length increased linearly with SLCO2. As LCO2 breakthrough in the low permeability zone occurred (SLCO2 > 36%), the linearity between interfacial length and SLCO2 is preserved but the slope is 1.3 times higher. This can be attributed to the fact that the smaller pore size is associated with a higher interfacial area.49 The linear relationship between specific interfacial area and nonwetting fluid saturation has been shown in numerical simulation50 as well as column experiments.51 Numerical Simulation Results. Using independently determined hydraulic and fluid properties, LCO2 saturations in the dualpermeability micromodel were predicted by the continuum-scale 7586

dx.doi.org/10.1021/es201858r |Environ. Sci. Technol. 2011, 45, 7581–7588

Environmental Science & Technology two-phase flow simulator. The results of the predictions are shown in Figure 4 and additional simulation results are included in SI Figure S3. In general, the numerical model adequately captured the increase in the LCO2 saturation at the lower injection rates in the high permeability pore network where capillary fingering was experimentally observed. It correctly predicts the breakthrough into the low permeability zone at the three highest injection rates, where the LCO2 invades by capillary fingering. However, the simulated LCO2 saturations are generally higher than those measured in the high permeability zone for the higher injection rates (SLCO2 is overestimated by as much as up to 22%), where viscous fingering was observed. This discrepancy is attributed to the fact that continuum-scale models do not explicitly account for the pore geometry and the unstable displacement by viscous fingering mechanism, and hence do not accurately predict the increases in saturation as a function of injection rate. Implications. CO2 sequestration in saline aquifers involves displacement of formation water by CO2 under unfavorable viscosity ratios with logM between approximately
1.6 and
0.7.6 Under these conditions, heterogeneity in the formation permeability and injections rates are expected to have major impacts on the resulting distributions of CO2, and are likely to affect both the shortterm hydrodynamic trapping mechanisms as well as the long-term mineral trapping. In this study, we used a micromodel with a dualpermeability pore network as a simple model of a confined aquifer system and investigated the influence of injection rate on LCO2 water displacement mechanisms, LCO2 saturations, interfacial areas, and LCO2 distributions. These pore-scale studies at 9 MPa, using precise dry-etched structures and florescence imaging, confirmed the importance of displacement mechanisms governed by viscosity ratios and capillary numbers, and showed a predictable transition from capillary to viscous fingering as flow rates (and hence capillary numbers) increased. In addition, the results demonstrated that heterogeneous permeability at the microscopic pore scale may directly influence aquifer storage capacity. In the dual-permeability pore network, a permeability contrast of approximately 2 resulted in preferential displacement of water by LCO2 in the high permeability zone. Although LCO2 was found to break through the low permeability zone under sufficiently high experimental injection rates, these rates (corresponding to a Darcy velocity of >280 m/day) may not be practically feasible in field-scale deployment of CO2 sequestration. Results from this study demonstrated that unstable displacement can be expected due to the unfavorable viscosity ratio between CO2 and water. It is possible for unstable capillary fingering to occur in places where injection rates and viscous forces are relatively small, for example, at some distance from an injection well. Near an injection well, displacement is likely to occur in the high capillary number range and unstable viscous fingering may become a concern. Comparison of the micromodel experiments with continuum model simulation results demonstrated that by using independently determined input parameters to characterize the model, the continuum-scale model accurately predicted LCO2 saturations in the dual-permeability field for the lower injection rates. The continuum model was also effective predicting when the LCO2 entered the low permeability zone by overcoming the nonwetting fluid entry pressure. Since the effects of unstable displacement on fluid saturations are not explicitly accounted for, LCO2 saturations are overestimated at high injection rates in the high permeability zone where viscous fingering was observed in the micromodel. The differences between modeling

ARTICLE

and experimental results at high flow rates in the high permeability zone illustrates the limitations of continuum modeling to accurately represent displacement and saturations under unstable viscous fingering conditions. Since injection of liquid of supercritical CO2 is expected to occur under unstable flow conditions, these results emphasize the importance of pore scale studies to better understand and predict large scale effects in carbon sequestration.

’ ASSOCIATED CONTENT

bS

Supporting Information. Images of LCO2 displacement of water under all injection rates, close-up images of the interface between low and high permeability zones, and continuum-model simulated LCO2 saturations are included in the Supporting Information. This material is available free of charge via the Internet at http://pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

*Phone: 1-509-371-6659; fax: 1-509-371-6354; e-mail: Changyong. [email protected].

’ ACKNOWLEDGMENT This research is supported by the Pacific Northwest National Laboratory Directed Research and Development Program under PNNL’s Carbon Sequestration Initiative. The experiments were conducted in the William R. Wiley Environmental Molecular Sciences Laboratory, a United States Department of Energy (DOE) scientific user facility operated for the DOE by PNNL. ’ REFERENCES (1) IPCC. IPCC Special Report on Carbon Dioxide Capture and Storage, Prepared by Working Group III of the Intergovernmental Panel on Climate Change. Metz, B., Davidson, O., De Coninck, H. C., Loos, M., Meyer, L. A., Eds.; Cambridge University Press: Cambridge and New York, Ny, 2005. (). 2005. (2) Class, H.; Ebigbo, A.; Helmig, R.; Dahle, H. K.; Nordbotten, J. M.; Celia, M. A.; Audigane, P.; Darcis, M.; Ennis-King, J.; Fan, Y. Q.; Flemisch, B.; Gasda, S. E.; Jin, M.; Krug, S.; Labregere, D.; Beni, A. N.; Pawar, R. J.; Sbai, A.; Thomas, S. G.; Trenty, L.; Wei, L. L. A benchmark study on problems related to CO2 storage in geologic formations. Comput. Geosci. 2009, 13 (4), 409–434. (3) Keating, G. N.; Middleton, R. S.; Stauffer, P. H.; Viswanathan, H. S.; Letellier, B. C.; Pasqualini, D.; Pawar, R. J.; Wolfsberg, A. V. Mesoscale carbon sequestration site screening and CCS infrastructure analysis. Environ. Sci. Technol. 2010, 45 (1), 215–222. (4) Wollenweber, J.; Alles, S.; Busch, A.; Krooss, B. M.; Stanjek, H.; Littke, R. Experimental investigation of the CO2 sealing efficiency of caprocks. Int. J. Greenhouse Gas Control 2010, 4 (2), 231–241. (5) Zhou, Q. L.; Birkholzer, J. T.; Mehnert, E.; Lin, Y. F.; Zhang, K. Modeling basin- and plume-scale processes of CO2 storage for full-scale deployment. Ground Water 2010, 48 (4), 494–514. (6) Nordbotten, J. M.; Celia, M. A.; Bachu, S. Injection and storage of CO2 in deep saline aquifers: analytical solution for CO2 plume evolution during injection. Transp. Porous Media 2005, 58 (3), 339–360. (7) Nordbotten, J. M.; Celia, M. A.; Bachu, S.; Dahle, H. K. Semianalytical solution for CO2 leakage through an abandoned well. Environ. Sci. Technol. 2005, 39 (2), 602–611. (8) Kuuskraa, V. A.; Koperna, G. J.; Riestenberg, D.; Esposito, R. Using reservoir architecture to maximize CO2 storage capacity. Energy Procedia 2009, 1 (1), 3063–3070. (9) Hovorka, S. D.; Choi, J.-W.; Meckel, T. A.; Trevino, R. H.; Zeng, H.; Kordi, M.; Wang, F. P.; Nicot, J.-P. Comparing carbon sequestration 7587

dx.doi.org/10.1021/es201858r |Environ. Sci. Technol. 2011, 45, 7581–7588

Environmental Science & Technology in an oil reservoir to sequestration in a brine formation—Field study. Energy Procedia 2009, 1 (1), 2051–2056. (10) Ennis-King, J.; Paterson, L. Role of convective mixing in the long-term storage of carbon dioxide in deep saline formations. Soc. Pet. Eng. J. 2005, 10 (3), 349–356. (11) Lengler, U.; De Lucia, M.; K€uhn, M. The Impact of Heterogeneity on the Distribution of CO2: Numerical Simulation of CO2 Storage at Ketzin. Int. J. Greenhouse. Gas Control 2010, 4 (6), 1016–1025. (12) Han, W. S.; Lee, S.-Y.; Lu, C.; McPherson, B. J. Effects of permeability on CO2 trapping mechanisms and buoyancy-driven CO2 migration in saline formations. Water Resour. Res. 2010, 46 (7), W07510. (13) Perrin, J. C.; Benson, S. An experimental study on the influence of sub-core scale heterogeneities on CO2 distribution in reservoir rocks. Transp. Porous Media 2010, 82 (1), 93–109. (14) Shi, J.-Q.; Xue, Z.; Durucan, S. Supercritical CO2 core flooding and imbibition in tako sandstone-influence of sub-core scale heterogeneity. Int. J. Greenhouse Gas Control 2011, 5 (1), 75–87. (15) Zhao, Y. C.; Song, Y. C.; Liu, Y.; Liang, H. F.; Dou, B. L. Visualization and measurement of CO2 flooding in porous media using MRI. Ind. Eng. Chem. Res. 2011, 50 (8), 4707–4715. (16) Suekane, T.; Soukawa, S.; Iwatani, S.; Tsushima, S.; Hirai, S. Behavior of supercritical CO2 injected into porous media containing water. Energy 2005, 30 (11
12), 2370–2382. (17) Bachu, S.; Bennion, D. B. Interfacial tension between CO2, freshwater, and brine in the range of pressure from (2
27) MPA, temperature from (20 to 125) degrees C, and water salinity from (0 to 334 000) Mg 3 L
1. J. Chem. Eng. Data 2009, 54 (3), 765–775. (18) Chiquet, P.; Daridon, J. L.; Broseta, D.; Thibeau, S. CO2/water interfacial tensions under pressure and temperature conditions of CO2 geological storage. Energy Convers. Manage. 2007, 48 (3), 736–744. (19) Espinoza, D. N.; Santamarina, J. C. Water-CO2-mineral systems: Interfacial tension, contact angle, and diffusion-implications to CO2 geological storage. Water Resour. Res. 2010, 46. (20) Lenormand, R.; Touboul, E.; Zarcone, C. Numerical-models and experiments on immiscible displacements in porous-media. J. Fluid Mech. 1988, 189, 165–187. (21) Cottin, C.; Bodiguel, H.; Colin, A. Drainage in two-dimensional porous media: from capillary fingering to viscous flow. Phys. Rev. E 2010, 82 (4). (22) Chang, L. C.; Tsai, J. P.; Shan, H. Y.; Chen, H. H. Experimental study on imbibition displacement mechanisms of two-phase fluid using micro model. Environ. Earth Sci. 2009, 59 (4), 901–911. (23) Ferer, M.; Ji, C.; Bromhal, G. S.; Cook, J.; Ahmadi, G.; Smith, D. H. Crossover from capillary fingering to viscous fingering for immiscible unstable flow: experiment and modeling. Phys. Rev. E 2004, 70 (1). (24) Lovoll, G.; Jankov, M.; Maloy, K. J.; Toussaint, R.; Schmittbuhl, J.; Schafer, G.; Meheust, Y. Influence of viscous fingering on dynamic saturation-pressure curves in porous media. Transp. Porous Media 2011, 86 (1), 335–354. (25) Zhang, C.; Oostrom, M.; Wietsma, T. W.; Grate, J. W.; Warner, M. G. Influence of viscous and capillary forces on immiscible fluid displacement: pore-scale experimental study in a water-wet micromodel demonstrating viscous and capillary fingering. Energy Fuels 2011 Epub ahead of print, DOI: 10.1021/ef101732k. (26) Chen, J. D.; Wilkinson, D. Pore-scale viscous fingering in porous-media. Phys. Rev. Lett. 1985, 55 (18), 1892–1895. (27) Lenormand, R.; Zarcone, C.; Sarr, A. Mechanisms of the displacement of one fluid by another in a network of capillary ducts. J. Fluid Mech. 1983, 135 (OCT), 337–353. (28) Lenormand, R.; Zarcone, C. 2-Phase flow experiments in a twodimensional permeable medium. Physicochem. Hydrodyn. 1985, 6 (5
6), 497–506. (29) Dias, M. M.; Payatakes, A. C. Network models for 2-phase flow in porous-media 0.1. Immiscible microdisplacement of nonwetting fluids. J. Fluid Mech. 1986, 164, 305–336. (30) Dias, M. M.; Payatakes, A. C. Network models for 2-phase flow in porous-media 0.2. Motion of oil ganglia. J. Fluid Mech. 1986, 164, 337–358. (31) Homsy, G. M. Viscous fingering in porous-media. Annu. Rev. Fluid Mech. 1987, 19, 271–311.

ARTICLE

(32) Bahralolom, I.; Bretz, R. E.; Orr, F. M. J. Experimental investigation of the interaction of phase behavior with microscopic heterogeneity in a CO2 flood. SPE Reservoir Eng. 1988, 3 (2), 662–672. (33) Chalbaud, C.; Robin, M.; Lombard, J. M.; Martin, F.; Egermann, P.; Bertin, H. Interfacial tension measurements and wettability evaluation for geological CO2 Storage. Adv. Water Resour. 2009, 32 (1), 98–109. (34) Er, V.; Babadagli, T.; Xu, Z. H. Pore-scale investigation of the matrix-fracture interaction during CO2 injection in naturally fractured oil reservoirs. Energy Fuels. 2010, 24, 1421–1430. (35) Riazi, M.; Sohrabi, M.; Jamiolahmady, M. Experimental study of pore-scale mechanisms of carbonated water injection. Transp. Porous Media 2011, 86 (1), 73–86. (36) Riazi, M.; Sohrabi, M.; Bernstone, C.; Jamiolahmady, M.; Ireland, S. Visualisation of mechanisms involved in CO2 injection and storage in hydrocarbon reservoirs and water-bearing aquifers. Chem. Eng. Res. Des., 2011, 89 (9), 1827–1840. (37) Chomsurin, C.; Werth, C. J. Analysis of pore-scale nonaqueous phase liquid dissolution in etched silicon pore networks. Water Resour. Res. 2003, 39 (9). (38) Willingham, T. W.; Werth, C. J.; Valocchi, A. J. Evaluation of the effects of porous media structure on mixing-controlled reactions using pore-scale modeling and micromodel experiments. Environ. Sci. Technol. 2008, 42 (9), 3185–3193. (39) Biswas, R.; Lewis, J. E.; Maroncelli, M. Electronic spectral shifts, reorganization energies, and local density augmentation of coumarin 153 in supercritical solvents. Chem. Phys. Lett. 1999, 310 (5
6), 485– 494. (40) Clark, M. M. Transport Modeling for Environmental Engineers and Scientists; John Wiley & Sons: New York, 1996. (41) Grate, J. W.; Zhang, C. Y.; Wietsma, T. W.; Warner, M. G.; Anheier, N. C.; Bernacki, B. E.; Orr, G.; Oostrom, M. A Note on the Visualization of wetting film structures and a nonwetting immiscible fluid in a pore network micromodel using a solvatochromic dye. Water Resour. Res. 2010, 46. (42) White, M. D.; Oostrom, M. Stomp: Subsurface Transport over Multiple Phases, Version 4.0. User’s Guide. Pnnl-15782; Pacific Northwest National Laboratory: Richland, WA, 2006. (43) Willingham, T.; Zhang, C. Y.; Werth, C. J.; Valocchi, A. J.; Oostrom, M.; Wietsma, T. W. Using Dispersivity values to quantify the effects of pore-scale flow focusing on enhanced reaction along a transverse mixing zone. Adv. Water Resour. 2010, 33 (4), 525–535. (44) Brooks, R. H., Corey, A. T. Hydraulic Properties of Porous Media: Hydrology Papers; Colorado State University: Fort Collins, 1964; p 24. (45) Burdine, N. T. Relative permeability calculations from pore-size distribution data. Trans. Am. Inst. Min., Metall. Pet. Eng. 1953, 198, 71–77. (46) Schroth, M. H.; Ahearn, S. J.; Selker, J. S.; Istok, J. D. Characterization of miller-similar silica sands for laboratory hydrologic studies. Soil Science Soc. Am. J. 1996, 60 (5), 1331–1339. (47) Pruess, K.; Muller, N. Formation dry-out from CO2 injection into saline aquifers: 1. Effects of solids precipitation and their mitigation. Water Resour. Res. 2009, 45. (48) Zhang, C.; Dehoff, K.; Hess, N.; Oostrom, M.; Wietsma, T. W.; Valocchi, A. J.; Fouke, B. W.; Werth, C. J. Pore-scale study of transverse mixing induced CaCO3 precipitation and permeability reduction in a model subsurface sedimentary system. Environ. Sci. Technol. 2010, 44 (20), 7833–7838. (49) Zhang, C. Y.; Werth, C. J.; Webb, A. G. A magnetic resonance imaging study of dense nonaqueous phase liquid dissolution from angular porous media. Environ. Sci. Technol. 2002, 36 (15), 3310–3317. (50) Oostrom, M.; White, M. D.; Brusseau, M. L. Theoretical estimation of free and entrapped nonwetting-wetting fluid interfacial areas in porous media. Adv. Water Resour. 2001, 24 (8), 887–898. (51) Porter, M. L.; Wildenschild, D.; Grant, G.; Gerhard, J. I. Measurement and prediction of the relationship between capillary pressure, saturation, and interfacial area in a napl-water-glass bead system. Water Resour. Res. 2010, 46. 7588

dx.doi.org/10.1021/es201858r |Environ. Sci. Technol. 2011, 45, 7581–7588