Capillarity and wetting of carbon dioxide and brine during drainage in Berea sandstone at reservoir conditions

The wettability of CO2-brine-rock systems will have a major impact on the management of carbon sequestration in subsurface geological formations. Recent contact angle measurement studies have reported sensitivity in wetting behavior of this system to pressure, temperature, and brine salinity. We report observations of the impact of reservoir conditions on the capillary pressure characteristic curve and relative permeability of a single Berea sandstone during drainage—CO2 displacing brine—through effects on the wetting state. Eight reservoir condition drainage capillary pressure characteristic curves were measured using CO2 and brine in a single fired Berea sandstone at pressures (5–20 MPa), temperatures (25–508C), and ionic strengths (0–5 mol kg NaCl). A ninth measurement using a N2-water system provided a benchmark for capillarity with a strongly water wet system. The capillary pressure curves from each of the tests were found to be similar to the N2-water curve when scaled by the interfacial tension. Reservoir conditions were not found to have a significant impact on the capillary strength of the CO2-brine system during drainage through a variation in the wetting state. Two steady-state relative permeability measurements with CO2 and brine and one with N2 and brine similarly show little variation between conditions, consistent with the observation that the CO2-brine-sandstone system is water wetting and multiphase flow properties invariant across a wide range of reservoir conditions.


Introduction
The aim of carbon dioxide capture and storage is to mitigate the impacts of fossil energy consumption on climate change by safely sequestering carbon dioxide (CO 2 ) in deep subsurface geologic strata such as saline aquifers and depleted oil reservoirs [Bachu et al., 1994;Orr, 2004;Bernstein et al., 2007]. The importance of carbon storage in combating climate change has been highlighted in reports by the International Energy Agency [International Energy Agency, 2012] and the IPCC indicating that the cost of CO 2 emissions avoidance through the use of alternative energy technologies will be 150% higher in the absence of CCS [Bernstein et al., 2007;Edenhofer et al., 2014]. Previous and ongoing experience with industrial-scale carbon dioxide injection has come mostly from operations of enhanced oil recovery in the United States although there are a few pilot CO 2 storage projects around the world [Global CCS Institute, 2014]. Despite its utility for oil production and the continued development of CO 2 storage, there are longstanding uncertainties about the multiphase flow properties underlying descriptions and predictions of CO 2 migration and trapping in the subsurface.
Permeable rocks have pore sizes with length scales 10 26 210 23 m. During fluid movement through constrictions at these scales, differences in interfacial energy between fluid and solid phases, r i;s (N m 21 ), give rise to the contact angle, h, where cosh5ðr nw;s 2r w;s Þ=r nw;w . The subscripts w, nw, and s refer to the wetting and nonwetting fluid phases and the solid phase, respectively. The symbol r nw;w is the interfacial tension between the two fluid phases. The interface of two fluids in a pore constriction of radius r (m) is thus curved and the nonwetting phase pressure, P nw , is elevated relative to the wetting phase, P w , with the difference being the capillary pressure P c 5P nw 2P w (Pa), described by the Young-Laplace equation, P c 5 2r nw;w cosh r : In the complex pore structures of geologic media, capillarity results in an organization of fluids into relatively large or small constrictions depending on their wetting state. This in turns results in the unique Both drainage and imbibition processes are important for a full description of multiphase flow and for CO 2 storage. We have first focused on drainage in this work, however, as it is the primary displacement process taking place while CO 2 is injected into a reservoir. For industrial storage projects, this period will last for decades and will be the key displacement process to model during the highest risk stages of the injection projects [Hesse et al., 2008;Szulczewski et al., 2012]. For latter stages of storage projects, after injection has ceased and when the CO 2 plume is migrating, both drainage and imbibition processes are taking place simultaneously, albeit at different locations in the plume [Doughty and Pruess, 2004], and for these observations of capillarity during imbibition will be important.
From equation (1), it is apparent that the wetting properties of a system will serve to strengthen or weaken capillarity and this indeed is the primary interest of research on wettability in application to CO 2 storage. In a porous rock, this weakening generally applies, although the picture is more complex. An extensive review of the impacts of the wetting state on capillarity in reservoir rocks was given in Anderson [1987]. While equation (1) was initially proposed as a way to scale measurements of capillary characteristics of rocks among various fluid pairs [Purcell, 1949], it has been shown that it does provide an accurate scaling for consolidated rocks so far as the contact angle is concerned [Dumore and Schols, 1974;Kruyer, 1958;De Witt and Sch€ olten, 1975;Swanson, 1985;Good and Mikhail, 1981;Morrow, 1976;McCaffery and Bennion, 1974;Anderson, 1987].
Decreasing the strength of the wetting state within a given rock manifests as a weakening of capillarity during drainage, but only after the wetting state has been significantly weakened from strongly wetting [Morrow, 1976;Anderson, 1987]. In other words, small differences in the affinity of fluids for the solid can give rise to strong capillarity in real rocks, and it is only when fluids have relatively similar solid interfacial energies that effects on multiphase flow are observed. This is primarily due to the options available for energy minimization and the configuration of contact lines provided by complex geometries at rock pore walls [Decker et al., 1999]. Further complicating observations for real geologic media, wettability can be distributed throughout a rock heterogeneously because of the distribution of specific minerals or the presence of hydrocarbon films [Salathiel, 1973;Jadhunandan and Morrow, 1995]. In the absence of a definitive understanding of the wetting state of a system, there is significant uncertainty around the capillary-dominated processes of multiphase flow.
The primary concern of this work is that several investigations into the wetting properties of CO 2 -brine rock systems have resulted in ambiguous determination of the wetting state [Chiquet et al., 2007;Espinoza and Santamarina, 2010;Jung and Wan, 2012;Wang et al., 2012;Kim et al., 2012;Broseta et al., 2012]. This outstanding issue has been summarized in a recent review paper by Iglauer et al. [2015]. Most observations have suggested that CO 2 is nonwetting on minerals unaltered by the presence of a hydrocarbon but differ in conclusions about the strength of the wetting state and sensitivity to pressure, temperature and salinity, and among minerals prevalent in reservoir rocks.
Observations of the contact angle between a sessile or pendant drop of fluid and a polished mineral surface have been the most common technique employed and have been reported for a range of pressure (1-10 MPa), temperature (25-508C), and brine salinity (0-5 mol kg -1 and Farokhpoor et al. [2013] saw little influence of CO 2 pressure. The observations reported in Wang et al. [2012] were sensitive to pressure and temperature, but without a discernible trend. The reported effects were generally stronger during imbibition processes, but sensitivity was also seen in brine receding contact angles representative of drainage. The work reported in Iglauer et al. [2014] also showed that the observations were sensitive to sample preparation procedures and it was suggested that this was the main source of the discrepancies. Contact angles measured in sessile and pendant drop experiments with the range of uncertainty reported cannot be used directly to assess the strength of capillarity and the impact on multiphase flow. For this  [Amott, 1959;Donaldson et al., 1969;Jerauld and Rathmell, 1997;, which are based directly on evaluations of the capillary pressure characteristic curve (USBM) or the relative fluid production after spontaneous and forced imbibition (Amott), primarily controlled by the capillary pressure characteristic curve. They are considered superior to the contact angle method when the wetting state is unknown, e.g., in the case of a native state core [Cuiec, 1975] or in the present case of CO 2 -brine systems.
This approach has been adopted for this study and has also been adopted by two groups investigating similar issues with CO 2 -brine systems in unconsolidated pure quartz sand [Plug and Bruining, 2007;Tokunaga et al., 2013]. Observations reported by Plug and Bruining [2007] of drainage and imbibition capillary pressure curves for gaseous and supercritical CO 2 -water showed little impact of the wetting state on the drainage process in a sand pack with permeability k 10 210 m 2 . Capillarity measured with CO 2 was indistinguishable from measurements made with N 2 . The imbibition process, however, suggested that the supercritical CO 2 system was less strongly water wet than the gaseous CO 2 and N 2 systems. Observations reported in Tokunaga et al.
[2013] using a sand pack with an order of magnitude lower permeability (k 10 211 m 2 ) found both drainage and imbibition capillarity to be strongly affected by pressure. The supercritical system approached an intermediate wetting state with weak capillarity as pressure varied from 8.5 to 12.0 MPa.
Observations of capillarity during drainage in consolidated rocks including Berea sandstone have been reported in Pentland et al. [2011]; Pini et al. [2012]; and Pini and Benson [2013]. Those studies were not designed to assess the impact of varying reservoir conditions but the observations suggested that drainage capillary pressure curves were not affected by variations in the wetting state in the comparison of a supercritical CO 2 -brine system to a decane-brine system [Pentland et al., 2011], between a liquid and supercritical CO 2 system  and gaseous CO 2 in comparison with supercritical CO 2 and a N 2 -brine system [Pini and Benson, 2013].
Beyond capillarity, Egermann et al. [2006] used observations of unsteady state drainage of CO 2 displacing brine at a range of pressures, temperatures, and brine salinities to evaluate effects of sensitivity to the wetting state. It was found that characterizing capillary pressure and relative permeability with mercuryair and N 2 -brine fluid pairs, respectively, were sufficient to predictively model the displacements with CO 2 (production and pressure time series) so long as the variation in thermophysical properties were  Table 2. The text BP refers to the back pressure. taken into account. It was further identified that CO 2 entry pressures scaled with interfacial tension, indicating that drainage was water wet. Using observations of relative permeability alone, Pini and Benson [2013] found the system to be consistent with the water-wet N 2 -water system whereas Levine et al. [2014] found CO 2 -brine systems to be less strongly water wet than N 2 systems, but with the CO 2 still acting as the nonwetting phase.
In this work, a core flood capillary pressure test [Ramakrishnan and Cappiello, 1991;Lenormand et al., 1993;Fordham et   . Vertical-horizontal saturation maps for Experiment 7, with gaseous CO 2 , along the central cross-sectional slice. These saturation maps confirmed that CO 2 was distributed uniformly with no dissolution, gravity segregation, or impacts of rock heterogeneity. A crosssectional image of the spacer located between sample and outlet endcap showed the fluid phases inside the spacer and confirmed a continuous water phase even at high-flow rates. Figure 4. The slice-averaged CO 2 saturation profiles along the length of the core at different injection flow rates during Experiment 1, CO 2 -water at 10 MPa and 508C.

10.1002/2015WR016947
Benson, 2013] was used to characterize the capillary pressure characteristic curve of the CO 2 -brine-sandstone rock system. A range of conditions were tested to assess whether CO 2 flowed as a nonwetting fluid during drainage, and if this was sensitive to reservoir conditions of pressure, temperature, and brine salinity. We performed the tests in a single Berea sandstone at eight conditions systematically varying pressure, temperature, and brine salinity. To assess the wetting state of the CO 2 systems, a test was also performed with the N 2 -water system, known to be water wet in Berea sandstone Firoozabadi, 2010a, 2010b].
The conditions were chosen to cover a range representative of subsurface conditions in real target storage reservoirs. They also allowed for observations of capillarity with CO 2 as a gaseous, low-density, and highdensity supercritical fluid, corresponding to system variations reported in the literature to have an impact on the wetting state. Relative permeability characterization tests were also performed at two conditions with CO 2 -brine systems and with the N 2 -water system to show the impact of the wetting state on multiphase flow. Finally, imbibition relative permeability observations were combined with observations of the residual trapping characteristic reported in a previous study  to provide a preliminary indication of how the drainage results reported extend toward imbibition processes.

The Rock Sample
A single homogeneous 20.32 cm long, 3.81 cm diameter Berea sandstone core sample was used for all of the tests performed in this study. Berea sandstone is an outcrop rock from the United States which is widely used in studies as a benchmark and for its utility in making comparisons with other studies of petrophysics [Oak et al., 1990;Wu and Firoozabadi, 2010a;Pentland et al., 2011]. The sample was heated at 7008C for 4 h as is standard to stabilize swelling clays that can otherwise lead to permeability changes during tests. The faces of the core were machined flat to ensure good contact with the endcaps. The sample was vacuum dried in air at 708C overnight before each test. The absolute permeability of the sample to water was 212 mD with an average porosity of 21% as measured by X-ray computed tomography (CT) scanning.  The capillary pressure characteristic curve of several samples of the Berea sandstone were initially characterized using the mercury injection capillary pressure test (MICP). The skeletal density of the sample was measured with helium pycnometer on a subsample (0.8 cm 3 ) that was cut from a section adjacent to the inlet face of the core sample and subsequently characterized by MICP using an AutoPore IV from Micromeritics. The pressure range measured was from vacuum to 240 MPa. The capillary pressure and pycnometery measurements were performed on several samples. Following a protocol described in Pini and Benson [2013], the mercury saturation in the MICP data was corrected using the true skeletal density obtained from the helium pycnometer. The true water porosity was measured by X-ray CT of the inlet face of the core.

Test Conditions
The test conditions are shown in Table 1. Carbon dioxide and nitrogen were used as the nonwetting phases in the coreflooding experiments both with 99.9% purity (BOC Industrial Gases, UK). The wetting phase fluids used were deionized water or NaCl brine. The brine solutions were made of deionized water and NaCl with total salt molality ranging from 0 to 5 mol kg 21 .
The values of interfacial tension of the fluid pairs were used to scale the capillary pressure measurements for the assessment of wettability (section 2.5). Much of the analysis is based on values of interfacial tension calculated from the empirical correlation for the CO 2 -brine system reported by Li et al. [2012], shown in Figure 1 and included in Table 1. We also include in Table 1 the range of values reported in the literature for conditions similar to those used in this work [Chun and Wilkinson, 1995;Bachu and Bennion, 2009;Chalbaud et al., 2009;Georgiadis et al., 2010;Tokunaga et al., 2013]. At some conditions, the values reported by Li et al. [2012] are significantly higher and outside of the range of values reported by the other groups. This is discussed in Georgiadis et al. [2010] where the discrepancy was attributed to fluid contamination in the measurements that resulted in lower values. Thus we evaluated the impact of the variation in reported interfacial tension on the interpretation of the measurement. This is discussed in detail in the results section 3. In most cases, it made little qualitative difference to the conclusions.
In one test, Experiment 9, the wetting phase solution was deionized water with 5 wt% of the X-ray dopant NaI. This was done to confirm the accuracy of the saturation observations without using the dopant . Because the interfacial properties were less well constrained for NaI, the other experiments were performed using NaCl solutions. Uncertainty bounds on the entry pressure are two times the standard error of the least products regression. The standard error on the values of S w;irr were all less than 0.01.  Table 1.

Core-Flooding Experimental Setup
Capillary pressure and relative permeability measurements were conducted using a two fluid phase core flooding system shown in Figure A1 in Appendix A. The experimental setup was designed to replicate in situ temperature and pressure conditions of up to 1208C and 30 MPa. The system included, six pumps (Teledyne Isco, Model 500D, and 100DX), a two phase separator, two high-accuracy pressure transducers (Oilfilled Digiquartz Intelligent transmitter, Model 410K-HT-101), and a Pheonix Instruments core holder wrapped in a silicone heating tape with temperature measured in the confining fluid. The setup was a closed loop system which allowed for a pre-equilibration of CO 2 and brine prior to initiation of drainage in the rock core, creating an immiscible displacement, and allowed fluids to be coinjected into the core sample at controlled rates (constant rate injection). The core holder was positioned horizontally through a medical X-ray CT scanning instrument refurbished for petrophysical work (Universal Systems, HD-350).

Measuring the Capillary Pressure Characteristic Curve
We used a core flooding test to measure the capillary pressure at the inlet of the rock core during drainage [Ramakrishnan and Cappiello, 1991;Fordham et al., 1993;Lenormand et al., 1993;Pini et al., 2012;Pini and Benson, 2013]. A schematic showing the principles of the method is shown in Figure 2. A detailed description of the startup and operation of the experiment is provided in Appendix A and summarized here.   Table 3. The numbers next to the data points indicate the experiment number. To make the measurements, CO 2 was injected into an initially water or brine-saturated core sample at a steady and low-flow rate. The pressure drop across the sample set the capillary pressure at the inlet face of the core where the saturation was measured using an X-ray CT scanner using standard techniques [Akin and Kovscek, 2003] with a detailed procedure provided in Appendix B. To ensure a reservoir of brine at the outlet boundary [Ramakrishnan and Cappiello, 1991], a spacer was used at the downstream end of the core (Figure 2). A capillary pressure curve was constructed by repeating these measurements at increasing CO 2 flow rates.
Flow rates were initially selected using the aid of a numerical simulation of the core flood. This is detailed in Appendix C. Example simulation results of pressure profiles of CO 2 along the length of the core at various fluid flow rates are shown in Figure 2. The injection flow rates ultimately used for the experiments varied   Table 2. The values of N c for the highest flow rates applied in this study ranged between 10 27 and 10 26 . These capillary numbers were below the critical value for ensuring a capillary dominated distribution of fluids in the pores. For Berea sandstone, this has been observed to be in the range N c > 10 25 210 24 [Taber, 1969;Foster, 1973;Gupta and Trushenski, 1979].

Assessing Wetting State Variation From Capillary Pressure Observations
The Brooks-Corey model was fit to the capillary pressure observations [Brooks and Corey, 1964], The entry pressure is given by P e (Pa) (i.e., the minimum pressure required for the entry of CO 2 into the pores of the rock), the subscript w represent the wetting phase, the subscripts irr represents the irreducible saturation, and k tunes the curvature of the model.
To assess the impact of the wetting state, a curve was first fit to the N 2 -water data to obtain a value for k. Then curves were fit to the CO 2 -brine data using P e and S w;irr as fitting parameters while holding k constant. All fits were performed using least products regression, considering both pressure and saturation observations as prone to error. Finally, these curves were scaled by the values of interfacial tension in Table 1. Plots of P c =r against S w were used as the basis for identifying the impact of reservoir conditions on wettability [Morrow, 1976;Plug and Bruining, 2007;Tokunaga et al., 2013].
The dimensionless J-function was also used to scale capillary pressure curves to compare the observations made for this work with data reported in the literature,  Results showing the effect of water salinity with capillary pressure curves scaled by interfacial tension and to quantify potential impacts of wettability variations. Nitrogen-water capillary pressure was used as a wetting reference. The values of applied in this graph are displayed in Table 4. The slice-averaged saturation profiles along the length of the core shown in Figure 4 for a CO 2 test and Figure 5 for a N 2 test showed the negative saturation gradient expected from the capillary end effect. In this case, the N 2 experiment exhibited a higher capillary end effect corresponding to the lower viscosity of N 2 compared with CO 2 at these test conditions.

Results and Discussion
Capillary pressure-saturation pairs for Experiment 1 are shown in Figure 6 alongside saturation maps from the inlet slice of the core. Also shown are the anticipated capillary pressure observations from the numerical simulation which used the MICP curve, scaled by interfacial tension, as an input to select the flow rates.
The pore-size distribution index obtained from the Brooks-Corey model fit to the N 2 -water test, equation (2), was k51:396:03. For the CO 2 tests, best fit Brooks-Corey curves were obtained using the entry pressure, P e , and the irreducible water saturation, S w;irr as the fitting parameters, but maintaining k51:4. Best fit values for all of the capillary pressure tests are reported in Table 3. All of the data are plotted scaling using the dimensionless J-function, equation (3), in Figure 7 with the values of interfacial tension calculated from the correlation of Li et al. [2012]. In this scaling, no adjustments were made using a term, e.g., cosh, reflecting the wetting state.
A plot of the best fit entry pressures from Table 3 against the interfacial tension is shown in Figure 8. The data show a positive correlation with a linear proportionality constant in the range of what would be expected for, e.g., 10 lm pore throats and a relationship with interfacial tension only. Experiment 4, with the highest salinity brine was a significant outlier, while entry pressures obtained for Experiments 1 and 6 were lower than expected for the interfacial tension and could indicate a small effect of the wettability state. These variations are discussed in more detail in the following sections. a The only parameters used to fit the curves were n w and n nw . The endpoint parameters were assumed. See text for details. Figure 13. Steady state (a) drainage and (b) imbibition relative permeability. Solid lines are Brooks-Corey curves fit to the data and an arbitrary endpoint value for the nonwetting phase.
Residual saturation corresponding to the drainage turnaround in Figure 13b are shown on the horizontal axis. See the text and Table 1 Pini and Benson [2013] in consolidated Berea sandstone, in which the comparison of drainage capillary pressure characteristic curves between CO 2 -brine and a separate fluid pair were found to scale with interfacial tension alone. This is also consistent with the conclusions in Egermann et al.
[2006] that drainage core flood displacements of brine by supercritical CO 2 could be predictively modeled based on a characterization of flow properties with other fluid systems, e.g., N 2 -water and mercury-air.
This suggests that the sensitivity of the wetting state observed in sessile and pendant drop contact angle measurements [Chiquet et al., 2007;Espinoza and Santamarina, 2010;Jung and Wan, 2012;Wang et al., 2012;Kim et al., 2012;Broseta et al., 2012] and in unconsolidated sand [Plug and Bruining, 2007;Tokunaga et al., 2013;Wang and Tokunaga, 2015] has little impact during drainage in consolidated sandstone. This observation was also supported by the results of the drainage and imbibition relative permeability tests, discussed below, which showed little variation between test conditions.
3.1. The Impact of Pressure, Temperature, and Brine Salinity on Capillarity Experiments, 1, 7, and 8, performed at 5, 10, and 20 MPa, respectively, and 508C and 0 M, were used to evaluate the effect of pressure on the wetting state. The characteristic curves are shown in Figure 9. The pressure range in the tests was such that the CO 2 phase was gaseous, low-density supercritical CO 2 , and high-density supercritical CO 2 at the respective conditions ( Figure 1). With increasing pressure, the interfacial tension between the fluids is reported to decrease from a range of 41-48 mN/m to a range of 24-30 mN/m, with the correlation of Li et al. [2012] showing a decrease from 47 to 30 mN/m, Table 1.
The data reflect the trend of increasing capillary strength with interfacial tension. Figure 9 shows the data scaled by the interfacial tension values reported in Li et al. [2012] alongside the curve fit scaled by the full range of interfacial tension values found in the literature at similar conditions (Table 1). The observations scaled with interfacial tension, including the observations made with the N 2 -water system. This was indicative of no observable difference in multiphase flow due to impacts on the wetting state, both among conditions and in comparison to the water wet N 2 system.
The impact of temperature can be observed comparing Experiments 1, 3, 5, and 6 performed at 25 and 508C, with the CO 2 phase in a liquid or supercritical state. Two pairs of experiments were performed at these conditions-one with deionized water (Experiments 1 and 5) and one with a 3 mol kg 21 NaCl solution (Experiments 3 and 6). In the deionized water systems, the interfacial tension increased from a reported range of 28-34 mN/m to a range of 30-36 mN/m with the increase in temperature. In the systems with NaCl in solution, the interfacial tension increased from a reported range of 29-38 mN/m to 35-41 mN/m.
The characteristic curves are shown in Figure 10, including the curve for the N 2 -water system. A plot is again shown scaling the data by the interfacial tension values of Li et al. [2012] and the Brooks-Corey curves with the full range of reported values. In this case the observations do not strictly show an increase in capillary strength with interfacial tension. At each salinity, increasing the temperature, and the corresponding transition from liquid to supercritical CO 2 , resulted in a small weakening of the capillary strength despite the corresponding, albeit small, increase in interfacial tension. This effect was small, however, and within the range of uncertainty with respect to the values of the interfacial tension. Comparison with the N 2 -water curve again suggested that the systems were water wet.
The impact of salinity can be observed by comparing the results of Experiments 1, 2, 3, and 4 in in which the salinity varied from deionized water, to 5 mol kg 21 NaCl, at constant temperature and pressure. In this case, the interfacial tension varied from a range of reported values of 30-36 mN/m with deionized water to a range of 44-46 mN/m in the 5 molal solution.
The characteristic curves are shown in Figure 11 and exhibit an increasing capillary strength with increasing interfacial tension. There is more variation, however, in the CO 2 -brine data than expected from effects of Water Resources Research

10.1002/2015WR016947
interfacial tension alone. The Brooks-Corey curves fit to the data are shown scaled by the full range of interfacial tension values reported at a given salinity. Variation in the reported values of interfacial tension does not substantially resolve the difference in the scaled curves. It is thus possible that there was some sensitivity in the wetting character of the system with salinity. In this case, increasing salinity corresponded to an increasingly water wet system. All of the data were clustered around the data from the N 2 system and thus the displacement was still characteristic of a water wet drainage process.

Evaluating the Wetting State
The most variation among data scaled by the interfacial tension was observed with variation in salinity. To quantify shifts in this data, possibly due to wetting, we used a constant multiplier, , the Arabic letter ''Ayn,'' such that scaling P c =ðr ) collapsed the data to a single curve. A value of was given to the N 2 -water P c curve to serve as the reference water wet system.
The scaled data are shown in Figure 12. A list of best fit values of for the scaling is shown in Table 4. Values of > 1 suggested that a system was more strongly water wet than the N 2 -water system and < 1 indicated that a system was less strongly water wet. As is clear from the preceding discussion of the results, the possible effects of wetting are small. The multiplier ranges from a value of 0.77 < < 1.43. This indicates that potential wetting effects were limited to result in a maximum decrease in capillary pressure at a given saturation by a factor of about 2 between the systems with the strongest and weakest capillarity, respectively.

Drainage Relative Permeability
The results of the drainage relative permeability tests are shown in Figure 13a. Axial saturation maps and averaged saturation profiles are provided in Appendix D and showed that the fluids were uniformly distributed for observations at every fractional flow, and not strongly affected by capillary end effects for all except the injection of 100% CO 2 at the end of the drainage cycle. As is common for studies with CO 2 and other low viscosity gases, the irreducible water saturation at the capillary endpoint to drainage was not achieved. This is because the achievable capillary pressure is limited in core floods by the maximum viscous pressure gradient that can be obtained in the core . Thus endpoint parameters for the Brooks-Corey models were assumed at values that have been obtained in past studies using Berea sandstone, e.g., Oak et al. [1990]. The Corey parameters were the only fitting parameters and suggest that there was little impact of changing conditions and fluid pairs.
The two drainage relative permeability curves for CO 2 -brine were indistinguishable from each other, consistent with the observation that there was little impact of wetting state changes on capillarity with changes in brine salinity. The data for the N 2 -water curve fall to the right of the CO 2 -brine observations. This could be indicative of a difference in wetting between the fluids [Berg et al., 2013], but could also be caused by subtle effects that the interplay between rock heterogeneity and interfacial tension can have on observations of relative permeability [Krause, 2012;Kuo and Benson, 2013]. Subsequent observations at higher fluid flow rates are currently being undertaken to determine this. The crossover point of the curves are above the 50% water saturation mark, characteristic of drainage in water wet systems [Craig, 1971]. As with the capillary pressure data, the findings were consistent with those of Egermann et al. [2006] which reported that relative permeability for N 2 -brine systems could be used to model drainage in CO 2 -brine core floods, so long as the thermophysical properties were taken into account.

Imbibition Relative Permeability
Observations of imbibition relative permeability are shown in Figure 13b for the CO 2 -brine and N 2 -water tests. The imbibition tracking curve is dependent on the saturation turnaround point at the end of drainage, which was different for the three experiments. Thus the hysteresis tracking curves themselves were not directly comparable. However, initial-residual data for these conditions were reported in a separate study on the same rock by Niu et al. [2015]. Using the maximum saturation obtained at the end of drainage as the turnaround point, the estimated residual saturations are shown on the horizontal axis in Figure 13b. The hysteresis tracking curves approach the predicted value of the residual. That study found the IR curve to be equivalent for CO 2 -brine and N 2 -brine systems, and invariant across a range of reservoir conditions. Thus the imbibition tracking curves obtained for this work were consistent with imbibition governed by water wet displacement physics. Further observations of imbibition processes, particularly of the capillary

Conclusions
This study investigated the impact of conditions of pressure, temperature, and brine salinity on the drainage capillary pressure characteristic curve of CO 2 -brine systems in sandstone. Eight capillary pressure tests were performed in a Berea sandstone at a range of conditions that were generally representative of the range of conditions in target CO 2 storage reservoirs worldwide.
Observed variation between the capillary pressure curves were largely explained by the corresponding variation in interfacial tension at those conditions, suggesting little impact of a change in wetting state. A comparison with the water wet N 2 -water system also showed the CO 2 -brine systems to be water wet.
The only variation in the scaled capillary pressure curves that exceeded the range of uncertainty from reported values of interfacial tension was seen with salinity variation from 0 to 5 molal NaCl brine solutions. In this case, the observed variation resulted in a maximum strengthening of capillarity (shifting upward of the capillary pressure curve) at the highest salinity of a factor of two over the system with dieonized water.
The observations were complemented with steady state drainage and imbibition relative permeability observations in CO 2 -brine systems at 0 and 3 mol kg 21 NaCl. The drainage curves for CO 2 and brine at the two conditions were indistinguishable but slightly shifted with respect to a set of N 2 -water curves. Imbibition curves exhibited hysteresis and were consistent with the characterization of the initial-residual characteristic of the rock previously published in Niu et al. [2015], consistent with water-wet displacement physics during imbibition in Berea sandstone.

Appendix A: Experimental Procedure
The first stage in each experiment was to remove air from the flow loop by flowing CO 2 through all parts of the experimental setup at an elevated pressure and released to the atmosphere. This was done at different stages by isolating different sections of the flow loop to insure that air was removed entirely. The system was then pressurized with CO 2 and heated to the experimental pressure and temperature. This step was required to obtain the CO 2 saturated core background scan. Next, water or brine, depending on the experimental fluid, was injected into the loop to displace and dissolve all of the CO 2 out of the core at the experimental pressure and temperature. A scan of the brine-saturated core was obtained for porosity measurements.
The absolute permeability to water was measured before each experiment. Permeability was measured by injecting water or brine, based on the experimental fluid, through the core sample at different constant flow rates and recording the pressure drop across the core. Permeabilities varied less than 5% throughout the series of tests suggesting minimal or no changes (i.e., fines migration, chemical reaction) in the core sample throughout the entire course of experiments performed in this study.
Water and CO 2 were circulated at experimental conditions in a closed loop bypassing the core. Carbon dioxide dissolved into the water until equilibration was reached. This resulted immiscible displacement. The dissolution of CO 2 into water was monitored by observing the mass balance of the fluid phases. The equilibration process started with a decreasing trend of the free-phase CO 2 in the system with subsequent stabilization of the phase confirming fluid equilibration. Equilibrium was generally established in this setup after 6 h of circulating the fluids at constant flow rates of 20 mL min 21 . Circulation was continued for at least 12 h. At this point, the CO 2 flow was stopped and only CO 2 equilibrated brine was circulated to ensure all free CO 2 was pushed through the lines to the two phase separator. Then the CO 2 equilibrated brine was injected through the line into the core. Several pore volumes of brine were injected into the core sample to displace the initial brine with brine equilibrated with CO 2 . This was confirmed by taking X-ray scans during injection until a slight increase in CT number corresponding to the higher density brine was observed. Once Water Resources Research 10.1002/2015WR016947 the core was fully saturated with CO 2 equilibrated brine, another background X-ray CT scan of the core was taken. At this stage, capillary pressure or relative permeability tests were performed.
The capillary pressure observations were initiated by injecting CO 2 into the core. For each capillary pressure measurement, CO 2 was injected at a constant flow rate until a constant pressure drop across the core was reached. Then, an X-ray scan was performed at the inlet face of the core sample. As described earlier, 10 scans over 10-30 s were taken of the inlet slice and then averaged for each saturation measurement to reduce the uncertainty associated with the computed saturations.
All steady state relative permeability experiments were performed at a constant total volumetric rate of 15 mL min 21 . The total volumetric flow rates were kept constant while changing the fractional flow. End effects were evident in the extreme fractional flows and these data points tend to have large error bars.

Appendix B: Porosity and Fluid Saturation From X-ray Imaging
Porosity and experimental fluid saturations were measured under each experimental condition. Both porosity and saturation measurements were obtained by a combination of background and experimental X-ray scans [Akin and Kovscek, 2003;Withjack, 1988]. The X-ray imaging parameters applied in this study were as follows: a voxel dimension of about (0.23 3 0.23 3 1) mm 3 , a display field of view of 12 cm, an energy level of the radiation of 120 keV, and a tube current of 200 mA. The porosity was measured using four sets of Xray scans. The first two were one scan of the air and one scan of the experimental brine at ambient conditions. The brine scans were made for every experiment so that the CT number of each particular brinesaturated core was obtained. The third scan was a scan of the dry core at ambient conditions where the pore space was filled with air. The fourth was a scan of the core saturated with brine at experimental conditions. The same confining pressure was applied in last two scans. The porosity was then calculated using where CT and I were the X-ray CT scanner attenuation coefficient converted to numerical values in Hounsfield units. These values were assigned for each voxel in the X-ray image. The parameters CT brine and CT dry were CT numbers from the brine-saturated core and the dry core, respectively. The parameters I brine and I air were CT numbers obtained from scanning brine and air alone, respectively. Fluid saturation measurements in the core sample required two background scans in addition to the experimental scan. The background scans were one scan of the core saturated with CO 2 and one scan of the core saturated with CO 2 -equilibrated brine. Both scans were taken at experimental conditions and with confining pressure. Carbon dioxide saturation during CO 2 flow was calculated using where the subscripts exp, CO 2 , and brine1CO 2 refer to scan values obtained during CO 2 injection through the core during core-flooding, when the core was saturated with CO 2 and when the core was saturated with CO 2 equilibrated brine, respectively. The slice-averaged porosity and saturation properties were calculated using slice-averaged CT numbers. A total of 10 scans were taken for each set of CT numbers used in equations (B1) and (B2). The repeated scans were then averaged to reduce the uncertainty associated with the computed porosities and saturations.

Appendix C: Experimental Design Using a Numerical Simulator
Numerical simulation was used to estimate drainage flow rates that would result in even coverage of data across a range of capillary pressures and saturations achievable within the flow rate limitations of the experimental apparatus. We describe this model in detail here but emphasize that the simulation was simply used to analyze the assumptions underlying the technique and to aid with the initial choice of fluid flow rates, not to interpret the experimental observations. A model of the rock core was constructed for use in the adaptive implicit-explicit black oil simulator (IMEX) from the computer modeling group (CMG). The fluids CO 2 and brine were immiscible by default in the black oil simulator, consistent with the conditions created in the experiments in this work. The simulations were performed on a two-dimensional homogeneous model constructed to represent an axial slice from the middle of the Berea sandstone rock core. The model had 200 grid cells along the flow direction and 50 in the vertical direction. This corresponded to the minimum image thickness obtained from our CT scanner and allowed us to anticipate the potential impact of gravity segregation. The Brooks-Corey model, equations (D2) and (D3), was used to describe the relative permeability using data from experiments reported in Krevor et al. [2012]. The capillary pressure characteristic curve was estimated from the MICP data scaled to the CO 2 -brine system using the Young-Laplace scaling, equation (1), ignoring the contact angle term. The mercury-air interfacial tension applied was 485 mN/m [Ritter and Drake, 1945]. After conversion, the data were fit with the Water Resources Research

10.1002/2015WR016947
Brooks-Corey model, equation (2), for use in the simulator. Boundary slices were added to both inlet and outlet ends of the core, having the same dimension and properties as the rock slices following a similar procedure outlined in Kuo and Benson [2013]. Constant injection rate-controlled wells and constant bottom hole pressure controlled wells were completed in the grid elements of the inlet and outlet boundary slices, respectively. In addition, linear relative permeabilities were used for the inlet and outlet slices to allow full mobility for each phase. The capillary pressure between CO 2 and brine was set to zero at the inlet. For the outlet boundary condition, the capillary pressure was set to a constant equal to the entry capillary pressure. This arrangement produced simulation results most consistent with our observations of the impact of the inlet and outlet boundaries.

Appendix D: Steady State Relative Permeability Measurement Technique and Saturation Maps
Three steady state relative permeability tests were performed in this work to complement the findings from the capillary pressure characteristic curve tests. The one-dimensional form of the Figure D1. Vertical-horizontal central cross-sectional slice saturation maps of CO 2 and N 2 for some of the fractional flows applied to measure steady-state relative permeability Experiments 11 and 13, respectively.

10.1002/2015WR016947
extended Darcy's law [Dullien, 1991] was used to calculate the relative permeability at a given fractional flow.
q 1 52 Akk r;1 ðS 1 Þ l 1 where q 1 is the volumetric flow rate of a particular phase, A is cross sectional area of the core sample, k is the absolute permeability, k r;1 ðS 1 Þ is the phase relative permeability as a function of its saturation, l 1 is the phase viscosity, Dx is the length of the core, and DP is pressure drop across the core.
The saturation in the core sample was controlled by the ratio of injected phases (i.e., the fractional flow). In this study, in situ saturation was measured along the entire sample with X-ray CT for each fractional flow once a constant pressure drop (i.e., steady state) was achieved. For the drainage test, CO 2 saturation was incrementally increased from zero by increasing the fractional flow, f CO2 . The fractional flow was controlled by simultaneously changing the relative volumetric flux of CO 2 and water while holding the total volumetric flow rate (q CO2 1q water ) constant. Axial saturation maps and averaged saturation profiles are shown in Figures D1, D2 and D3.
The Brooks-Corey model was fit to the drainage relative permeability observations [Brooks and Corey, 1964] k r;w 5k r;w ðS nw;r Þ s w 2s w;irr 12s w;irr 2s nw;r nw (D2) k r;nw 5k r;nw ðS w;irr Þ 12s w 2s nw;r 12s w;irr 2s nw;r nnw : The parameters n w and n nw are unit less fitting parameters for wetting and nonwetting phase curves, the subscripts nw and w refer to the nonwetting and wetting phase, respectively, k r;w ðS nw;r Þ and k r;nw ðS w;irr Þ are the corresponding relative permeability endpoints. Figure D3. Slice-averaged N 2 saturation profiles along the length of the core at 15 cm 3 min 21 at different fractional flows in a steady state relative permeability experiment of N 2 -water at 10 MPa and 258C. Figure D2. Slice-averaged CO 2 saturation profiles along the length of the core at 15 cm 3 min 21 at different fractional flows in a steady state relative permeability experiment of CO 2 -water at 10 MPa and 508C.