We report on measurements of the flow pattern and in-situ saturations whenn-octane displaces a brine in which surface treated silica nanoparticles are dispersed. The nanoparticles are known to stabilize octane-in-water emulsions. We find that the displacement front is more spatially uniform, and with a later breakthrough when compared to a control displacement with no in-situ nanoparticles. Pressure measurements during the displacement are consistent with generation of a viscous phase such as an emulsion. These observations suggest that a nanoparticle stabilized emulsion is formed during the displacement which acts to suppress the viscous instability. We argue that generation of droplets of nonwetting phase occurs at the leading edge of all drainage displacements. The droplets rejoin the bulk phase in the absence of stabilizing agents, but are preserved when nanoparticles adhere to the fluid/fluid interface.
 One of the key concerns for CO2 sequestration in subsurface saline aquifers is that CO2is less dense and less viscous than the in-situ brine at the depths it is planned to be emplaced [Lackner, 2003]. Without a robust seal above the saline aquifer, sequestered CO2 will potentially rise through relatively thin paths the geologic strata and be emitted back into the atmosphere [Klusman, 2003; Saripalli and McGrail, 2002; Nordbotten et al., 2005]. Nature has shown the ability to provide robust geologic seals for low viscosity, buoyant fluids as large volumes of natural gas, oil, and naturally occurring CO2 (produced at depth by heating carbonate rocks) have been trapped in reservoirs for tens of millions of years. Risk assessment of large scale geologic sequestration will include estimates of the efficacy of the seals above potential CO2 repositories (saline aquifers), and monitoring of storage sites will include regular measurements of pressures and compositions in overlying formations [Klusman, 2003]. Sequestration projects are likely to include remediation plans in case leakage occurs. Nevertheless the principle that prevention is cheaper than cure applies here, and it would be useful to engineer the sequestration in such a way that any potential leaks would be self sealing. An analogy would be “run flat” tires that are currently used on automobiles and bicycles. When these tires are punctured, the first pulse of escaping gas contains compounds which through the rapid expansion create a reaction that acts to seal the leak.
 One potential method to induce self sealing for CO2 is inspired by the observation [Zhang et al., 2010; Espinosa et al., 2010] that suitably chemically coated nanoparticles can stabilize a foam of supercritical CO2 in brine. The nanoparticles adhere to the surface of CO2bubbles and prevent their coalescence. Such foams have been created by co-injecting CO2 and brine containing dispersed nanoparticles at high rates through a porous medium [Zhang et al., 2010; Espinosa et al., 2010]. In terms of transport and sealing, the key is that the nanoparticle stabilized foams and/or emulsions (if the CO2 is in liquid form) are much less mobile than the separate pure phases. This is especially important because the current paradigm assumes CO2will be injected and stored as a supercritical phase, which has a gas-like viscosity. Thus if the CO2 spontaneously forms a foam as it moves along a leakage path, the formation of a foam will act to slow down the leak. While not a hard geologic “seal”, this can greatly reduce the mobility of the CO2 and potentially prevent catastrophic leaks.
 To use this idea for secure sequestration, one could first inject nanoparticles into the upper portion of the prospective storage structure, or into locations where risk of leakage is expected to be greatest, or even into an overlying formation. Because the particles are much smaller than the pore throats in the rocks that will be used for storage, and because the chemical coating on the particle surface can be tailored to minimize interaction with the rock surface, nanoparticles can easily be transported through the aquifer. If the buoyant sequestered CO2 were to rise through any of the nanoparticle treated rock, the nanoparticles would be attracted to the CO2/brine interface. The nanoparticles effectively armor the interface, and if the CO2 is in tiny droplets, these droplets will be stabilized (i.e., they will not coalesce) and form a CO2 in brine emulsion.
 In this paper we investigate whether this hypothesized mechanism takes place by performing core floods where n-octane displaces a brine with and without nanoparticles. In these experiments,n-octane acts as low pressure analog to high pressure supercritical CO2. We measure the displacement patterns and in-situ saturations using CT scanning, and measure the pressure drops along the core for both displacements. We find that the patterns, saturations, and pressure drops are consistent with the formation of an oil in water emulsion, even at the relatively low flow rates used in these experiments. We discuss the physical mechanisms through which a displacement into a nanoparticle suspension inherently creates an emulsion and stabilizes the overall displacement.
2. Materials and Methods
 In each of the experiments the working porous medium was a cylindrical core (length 30 cm, diameter 7.5 cm) of Boise sandstone (porosity 27.5%, permeability 1D). The core was preloaded with a 2 wt% NaCl (the control case) or with 2% brine with a 10% by weight suspension of silica nanoparticles (the nanoparticle case). The nanoparticles had a silica core of diameter 5 nm with a 5 nm polyethylene glycol coating, which allows the nanoparticles to stay dispersed without aggregation [Rodriguez et al., 2009]. The treated nanoparticles have been shown to stabilize CO2 in water foams and n-octane in water emulsions through surface armoring [Zhang et al., 2010; Espinosa et al., 2010].
n-Octane was injected to displace the brine phase at a flux of 0.02 cm/min. n-Octane was used as a low pressure analog to high pressure supercritical CO2, as it mimics the key features of CO2 in the sense that both fluids a) form nanoparticle stabilized droplets [Zhang et al., 2010; Espinosa et al., 2010], b) act as the non-wetting phase in the sandstone, c) are condensed phases (have roughly the same density), d) give displacements that are capillary dominated at the pore scale (ratio of capillary forces to viscous forces), and e) are less viscous than brine. The relevant fluid parameters are listed inTable 1.
Viscosity μ, density ρ, interfacial tension with respect to brine σ, and capillary number Ca at the relevant flow rates. 5% nano and 10% nano refer to the nanoparticle suspensions in brine that were used. The supercritical CO2 properties are at T = 40 °C and P = 10 MPa; all other properties are at room temperature and pressure [Vargaftik, 2005; DiCarlo et al., 2000; da Rocha et al., 1999].
4 × 10−8
1 × 10−7
 The internal dynamics of the displacement were observed in real time by placing the core horizontally in a Universal Systems HD-350 modified medical scanner. The scanning procedure consisted of taking scans every 1 cm along the core. The resolution of the scanner was 0.30 mm in the scan plane with a scan thickness of 1 cm; adequate for seeing overall saturation patterns, but not pore-scale positioning of fluids. Before the corefloods, calibration scans of the core fully saturated withn-octane and with brine were taken. During the coreflood, the core was scanned every 5 minutes for early time behavior and every 30 minutes after the octane had broken through. Each set of scan (30 images) takes about 3 minutes. The experiments stopped after 6 hours when approximately 1 pore volume (PV) ofn-octane had been injected.
 For the experiments where the pressure was measured across the core, similar tests were performed with a slightly different protocol. A slightly smaller core of Boise sandstone was used (length 30 cm, diameter 2.5 cm), the brine was 1% NaCl, the nanoparticles suspension consisted of 5 wt%, 20 nm coated nanoparticles (Nyacol, DP 9711), and the n-octane was injected at a flux of 0.10cm/min. The pressure drop was measured (Rosemount transducers) from the inlet to the outlet to the outlet, and 3–4 PV of n-octane was injected.
3. Experimental Results
Figure 1a shows lateral CT scans longitudinally along the core after 0.1 PV of n-octane are injected with brine as the initial fluid (the control experiment). From left to right the scans start at a distance of 2 cm from the inlet and are at equal intervals of 1 cm; full brine saturation is red, n-octane is blue. These scans show that the displacement front is not uniform, there are regions of highn-octane saturation next to regions of high brine saturation. Taken as a whole, the scans show that the displacingn-octane front formed preferential flow paths through the core (fingers). In particular, going from the inlet towards the center of core, at a distance of 2–3 cm the left hand side of the column is still filled with brine, at a distance of 5–7 cm n-octane fingers are seen branching out, and at a distance of 8–9 cm a single finger of n-octane is seen in the 4 o'clock position. In the regions where the brine has been displaced the average remaining brine saturation is roughlySw ≈ 0.23, while adjacent regions have saturations of Sw ≈ 0.94 (e.g., see 3 cm position).
Figure 1b shows the lateral CT scans with nanoparticle suspension as the initial fluid at 0.1 PV. The same core is used as the control experiment. Comparing to the control experiment in Figure 1a, these images show much less lateral variations of n-octane and thus little or no fingering. In particular, then-octane phase has moved a 20% less distance even though the same volume has been injected, there is no evidence of fingers (the images are uniform), and behind the front there is a mixture ofn-octane and brine (uniform green color rather than red or blue). Also, in the regions behind the front that are most filled withn-octane the water saturation is still rather high atSw ≈ 0.42. Since n-octane and brine spontaneously separate, this high residual saturation is likely the sign of an emulsion forming in the core, although the CT images do not have the resolution to distinguish what fraction of the water is in an emulsion, and what fraction of the water is in a separate phase.
Figure 2 (top) shows the overall n-octane saturation as a function of distance along the core at different times; here the saturation is averaged over each image. Here we observe a continuous drop in water saturation behind the front, and the images show that the fingers continue to grow laterally with time, eventually almost merging together. Behind the front, the overall slope of the saturation isdSw/dz ≈ 0.06 cm−1, and with no distinct saturation front when averaged over the whole core.
Figure 2 (bottom) shows the overall n-octane saturation as a function of distance along the core at different times. Comparing to the control experiment, here we observe a distinct saturation front; this front has a saturation gradient ofdSw/dz ≈ 0.12 cm−1at the front, twice as steep as the control experiment. This type of piston-like front is what is expected in a stable displacement. Behind this front, the saturation is uniform laterally and longitudinally with a water saturation ofSw ≈ 0.4.
Figure 3shows the overall pressure drop as a function of time for both the control and nanoparticle cases. For ease of comparison, the pressure drop was normalized by the pressure drop of single-phase brine through the column at the same flow rate. For the control case, the pressure drop increased by a factor of two during multi-phase flow in the core, followed by a decrease as then-octane saturation increased, before finally settling down at residual water saturation to a pressure drop slightly lower than for brine. For the case with in-situ nanoparticles, the pressure drop showed a similar structure to the control case (an increase during the multi-phase flow portion followed by a decrease and a plateau). But for the nanoparticle case, the pressure drop was roughly 1.5–2.5 times greater than that seen in the control case, with the largest differences during the multi-phase flow portion of the displacement.
 The results above show that in the presence of nanoparticles, the displacement front becomes stable or self-regulating. This is associated with a higher water saturation behind the displacement front and a higher pressure drop when compared to the control case. In the following, we offer an explanation of these results through the formation of a nanoparticle stabilized emulsion at the main displacement front. This emulsion lowers the mobility of the invading phase (mobilityλ = Kkr/μ, where K is the permeability, kr is the relative permeability, and μ is the viscosity), which in turn suppresses the fingering and stabilizes the front.
 We begin the discussion by working backwards from the observed displacement behavior. Viscous fingering [Homsy, 1987; Saffman and Taylor, 1958] is a well known phenomenon and occurs in the control case because n-octane has a lower viscosity than brine (seeTable 1). Since the mobility of the invading phase is greater than the defending phase, this makes the front viscously unstable, i.e., any perturbation to the front will grow with time, and any variations in the local permeability of the core will be magnified [Homsy, 1987; Saffman and Taylor, 1958]. If the invading phase (CO2 or n-octane) can be made to have a mobility less than the brine, any perturbations will diminish with time, and effect of variations in permeability will be lessened.
 The observed stabilization for the nanoparticle experiment can only occur if the invading n-octane creates a phase that has a lower mobility in the nanoparticle displacement than in the control. As the major effect of nanoparticles is to stabilize emulsions, we hypothesize that the decrease in mobility due to the formation of an emulsion at the invading front. The in-situ emulsion can decrease the mobility by either increasing the effective viscosity of the invading phase, or by decreasing the relative permeability (e.g., the discontinuous droplets can act to block pores). We next look into how an emulsion can be created at the relatively low superficial velocity (flux) used in these experiments.
 In these displacement experiments, the creation of the emulsion can be due to a phenomena dubbed “Roof snap-off” [Roof, 1970; Rossen, 2003]. It is known that at the pore-scale a local increase in nonwetting phase saturation happens through Haines jumps, that is, when the pressure become enough for the nonwetting phase to pass through the throats, it jumps into the adjacent pore.Roof showed that this Haines jump will create a temporary decrease in the local capillary pressure at the throat. This decrease is great enough so that the wetting phase can re-coalesce, or snap-off, in the throat. This temporarily disconnects the non-wetting phase in the pore. Without the nanoparticles, the non-wetting phase reconnects once the nonwetting phase re-enters the throat (due to the continual forcing of non-wetting phase). But in the presence of nanoparticles, once the nonwetting phase becomes disconnected, it becomes armored with nanoparticles [Binks and Horozov, 2006; Zhang et al., 2010], and remains as a stable emulsion.
 There is a debate if the Roof snap-off events are the sole mechanism for creating high quality emulsions, as the first Roof snap-off discourages subsequent events [Rossen, 2008; Kovscek et al., 2007]. Regardless, other proposed mechanisms of emulsion generation all involve pore-scale displacements of the wetting phase by the invading nonwetting phase [Rossen, 2008]. These pore-scale displacements are forced to occur in the displacement experiments presented here, but not in co-injection experiments [Espinosa et al., 2010].
 There are various equations to estimate the viscosity of an emulsion in a porous medium [Kokal et al., 1992], but all depend on the quality (the fraction of nonwetting phase dispersed in the wetting phase) of the in-situ emulsion. In our case, the quality is unknown; we were unable to obtain measurable quantities of the emulsion, most likely due to the capillary end effect (the buildup of wetting phase near the outlet of a porous medium). Instead we assume that our effective emulsion phase in the porous medium consists of a mixture of continuous phases ofn-octane and brine along with the emulsion, and the effective mobility of this phase is found from the pressure drops shown inFigure 3. Since the average pressure drop is approximately 2.5 times greater for the experiment with nanoparticles, we infer that the effective emulsion phase has a mobility 2.5 times smaller than the n-octane in the control experiment. Thus, the mobility of the invading phase is then less than that of the brine phase and the flow is stable. This smaller mobility is also likely the result of relative permeability effects. The CT scans show that behind the front the water saturation is much higher in the nanoparticle compared to the control case. At these intermediate saturations, the relative permeability of both phases is low, leading to an overall lower mobility.
 The observed decrease in multi-phase mobility is enough to stabilize the front of the invading analog fluid (n-octane). As shown inTable 1, the viscosity ratio between supercritical brine and CO2 is much greater than for the n-octane. As the proposed mobility reduction mechanism by nanoparticles is physically based, we also expect a decrease in mobility for CO2 displacements, but with unknown magnitude. We are currently setting up to perform identical experiments using supercritical CO2, in order to determine the magnitude of the mobility decrease, and if the decrease is enough to stabilize displacements for the larger viscosity ratio.
 There are two important distinctions between these experiments and typical emulsion or foam generation experiments in porous media. First, most of the emulsion generation experiments consist of the phases being co-injected; here, the experiments consist of a displacement experiment. For co-injection experiments, the phases can flow in separate channels requiring very high velocities to obtain the required shear rates between the phases. For displacement experiments, the phases displace each other, and the mixing and shear can be created by the snap-off mechanism. Second, in most displacement experiments in which a surface active compound are used (be it a surfactant or nanoparticle suspension), the surface active compound is in the invading phase (typically the wetting phase) not the displaced phase, as is the case here.
 In terms of engineering safer CO2storage, ideally one would like to pre-position the nanoparticles along potential leakage paths, and the CO2 encounters such a path (e.g., a fracture in the caprock), a CO2/brine foam will form, reducing the effective permeability of the leakage path. Care must be taken to flush the nanoparticles from the wellbore before injecting the CO2 as creating an emulsion near the well bore could seriously impact the injectivity of the well.
 In addition to performing the same experiments with CO2, we are also setting up experiments with a larger permeability contrast (e.g., with a fracture), and determining the dependence of the front dynamics on the concentration of nanoparticles. These experiments will determine the amount of nanoparticles needed to stabilize and seal the CO2 displacement as a function of caprock geometry and CO2 flow rate. The results can eventually be used in engineering CO2 injection protocols.
 We wish to thank S. Gabel, H. Dehghanpour, and M. Mirzaei for experimental assistance, and C. Doughty and an anonymous reviewer for useful comments. This work was supported by the Center for Frontiers of Subsurface Energy Security (CFSES), an Energy Frontier Research Center funded by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences under award DE-SC0001114.
 The Editor thanks the anonymous reviewer for their assistance in evaluating this paper.