Effect of Hydrophobic and Hydrophilic Metal Oxide Nanoparticles on the Performance of Xanthan Gum Solutions for Heavy Oil Recovery

Recent studies revealed higher polymer flooding performance upon adding metal oxide nanoparticles (NPs) to acrylamide-based polymers during heavy oil recovery. The current study considers the effect of TiO2, Al2O3, in-situ prepared Fe(OH)3 and surface-modified SiO2 NPs on the performance of xanthan gum (XG) solutions to enhance heavy oil recovery. Surface modification of the SiO2 NPs was achieved by chemical grafting with 3-(methacryloyloxy)propyl]trimethoxysilane (MPS) and octyltriethoxysilane (OTES). The nanopolymer sols were characterized by their rheological properties and ζ-potential measurements. The efficiency of the nanopolymer sols in displacing oil was assessed using a linear sand-pack at 25 °C and two salinities (0.3 wt % and 1.0 wt % NaCl). The ζ-potential measurements showed that the NP dispersions in deionized (DI) water are unstable, but their colloidal stability improved in presence of XG. The addition of unmodified and modified SiO2 NPs increased the viscosity of the XG solution at all salinities. However, the high XG adsorption onto the surface of Fe(OH)3, Al2O3, and TiO2 NPs reduced the viscosity of the XG solution. Also, the NPs increased the cumulative oil recovery between 3% and 9%, and between 1% and 5% at 0 wt % and 0.3 wt % NaCl, respectively. At 1.0 wt % NaCl, the NPs reduced oil recovery by XG solution between 5% and 12%, except for Fe(OH)3 and TiO2 NPs. These NPs increased the oil recovery between 2% and 3% by virtue of reduced polymer adsorption caused by the alkalinity of the Fe(OH)3 and TiO2 nanopolymer sols.


Introduction
Water-soluble polymers have been used in enhanced oil recovery (EOR) due to their ability to improve sweep efficiency by controlling water mobility, reducing water permeability in swept zones and contacting unswept zones of the reservoir [1]. Hydrolyzed polyacrylamide (HPAM) and xanthan gum are the most commonly used polymers for EOR [2]. Xanthan gum (XG) is a high molecular anionic polysaccharide produced by bacterium Xanthomonas campestris during the fermentation of a cellulosic backbone [3]. XG solution properties have been studied over the past 50 years. The main topics of interest include rheological behavior of XG solutions and their mixtures with other polymers [4][5][6][7][8][9], effect of temperature on polymer conformation [10][11][12], and effect of salinity on solution performance and polymer conformation [13,14].
In solution, XG exhibits two conformational states: an ordered helix conformation and a disordered coil conformation [13]. The conformation of the XG molecules depends on the ionic strength of the solution and the temperature. In the presence of salts, XG molecules adopt a rigid rod like structure (ordered conformation) because the negatively charged pyruvate molecules wrap

SiO 2 NP Surface Modification
A mass of 4 g of SiO 2 NPs dispersed into 80 mL of cyclohexane were mixed with 1.6 mL of MPS or 2.09 mL of OTES. The dispersion was stirred at 200 rpm for 12 h at room temperature for the reaction to take place. After treatment, the product was recovered by centrifugation at 2500 rpm for 30 min and washed three times with ethanol to remove the excess modifier. The precipitate was dried in an oven at 70 • C for 24 h. Masses of 3.44 g of SiO 2 -MPS and 3.29 g of SiO 2 -OTES NPs were recovered, since some NPs were lost during preparation.

Modified NP Characterization
Fourier-transform infrared (FTIR) spectra were recorded over the range of 4000-400 cm −1 in a FTIR spectrometer (model IRaffinity−1s, Shimadzu, Japan) using KBr for running the background spectrum. Thermogravimetric analysis (TGA), from 20 to 800 • C, were performed on Q600 TGA (TA instruments, Inc., New Castle, DE, USA) at a heating rate of 10 • C/min in air atmosphere.

Nanopolymer Sol Preparation
The NPs were first dispersed in DI water at 0.2 wt % and ultrasonicated for 1 h. SDS was added at 0.1 wt %, and the dispersions were stirred for 30 min. Then, XG at 0.4 wt % was added and stirred for 1 h. Finally, NaCl was added to the sample to achieve a concentration of 0.3 wt % or 1.0 wt %.
For the preparation of the nanopolymer sols with in situ Fe(OH) 3 NPs, 0.2 g of the polymer and 0.05 g of SDS were dissolved into 35 mL of DI water. Then, 1.02 g of an aqueous solution of FeCl 3 .6H 2 O (19.98 g of FeCl 3 .6H 2 O in 25 mL of DI water) and 1.86 g of an aqueous solution of NaOH (5N) were added to the solution. The sample was left to mix at 150 rpm at 25 • C for 10 min. Then, 12.69 g of DI water were added. The reaction produces 0.3 wt % of NaCl, as a byproduct, according to the following reaction:

Colloidal Stability and Particle Size
ζ-potential values of each nanopolymer sol were measured at 20 • C in absence of NaCl using a Zetasizer Nano ZS unit (Malvern Instruments Ltd., Malvern, UK), with uncertainty in the order of ±1% to ±6% of the reported value. The measurements were conducted after ultrasonicating each sample for 5 min. The ζ-potential values were measured at 0.2 wt % NaCl. These values could not be measured at higher ionic strength (0.3 wt % and 1.0 wt % NaCl) due to the high conductivity values (>5 mS/cm). Particle size measurements were carried out by dynamic light scattering (DLS) using a particle size analyzer (Zetasizer nano ZS) at 20 • C. The results are presented in Table S1 of the supplementary material. A digital pH meter (Fisher Scientific, model AB 15 plus, Santa Barbara, CA, USA) with an uncertainty of less than ±0.05 of the reported value was used to measure the pH values at 20 • C.

Viscosity of the Nanopolymer Sols
The viscosities of the nanopolymer sols were measured at 25 • C using a viscometer (Thermo Scientific™ HAAKE RotoVisco 1, Santa Barbara, CA, USA). The uncertainty of the reported valued remained between ±1% to ±7%. The shear rate was varied from 5.0 to 100 s −1 .

Displacement Test in Linear Sand-Pack
A stainless-steel tube with 30.4 cm length and 2.54 cm i.d. was used as a holder for silica sand. The packing process was carried out by filling the holder with sand and frequently tapping it to ensure the sand was tightly packed. The volume of the sand grains was determined as the weight of the sand into the holder divided by the density of the grains (2.6 g/cm 3 ). The pore volume (PV) of the porous media was calculated by subtracting the volume of sand grains in the holder from the total volume of the holder (154.4 cm 3 ). Porosity was calculated as pore volume divided by total volume of the holder.
For the displacement tests, the sand-pack is initially fully saturated with brine by evacuating the pore space with a vacuum pump and allowing the DI water to imbibe under high vacuum. In order to determine the permeability of the porous media (K abs ) from Equation (1), DI water was injected at different flow rates (10,20,30, and 40 mL/min), and the corresponding pressure drop was recorded.
where L is the length of the pack, cm; A is the cross-sectional area of the pack, cm 2 ; µ is the viscosity of the fluid, mPa.s; Q is the flow rate, cm 3 /s; and ∆P is the differential pressure across the sand-pack, atm. The drainage process was carried out by injecting 2 pore volumes (PV) of silicon oil at 0.1 mL/min until the water fraction at the production end was less than 1%, and the pressure stabilized. The initial oil saturation was calculated by dividing the volume of produced water during oil flooding by the pore volume. After that, 1 PV of polymer solution or nanopolymer sol was injected followed by 2 PV of water. Cumulative oil recovery was calculated by dividing the sum of oil recovered from the chemical flood by the initial volume of oil in the sand-pack. The aqueous and oil phases were separated by heating the collected samples at 70 • C for 30 min in an oven.

FTIR and TGA Measurements of the Modified Silica NPs
The FTIR spectra of unmodified and modified silica NPs after normalization of the peak area are shown in Figure 1. The asymmetric and symmetric stretching vibrations of C-H groups of MPS for the SiO 2 -MPS NPs were clearly observed around 2951 and 2481 cm −1 , and the stretching vibration of C=O, methylene (=CH 2 ), and vinyl (-CH=CH 2 ) bending vibrations were observed at 1703, 1456, and 1406 cm −1 [41], respectively. For SiO 2 -OTES NPs, two peaks appear at 2926 cm −1 and 2858 cm −1 due to the intense symmetric and asymmetric stretching vibrations of the C-H bonds in the octyl groups [42]. A peak also appears at 1392 cm −1 , which is assigned to the asymmetric deformation vibration of C-H bonds due to a slight substitution of the octyl groups in place of the -OH groups of the SiO 2 NPs [43]. . Porosity was calculated as pore volume divided by total volume of the holder. For the displacement tests, the sand-pack is initially fully saturated with brine by evacuating the pore space with a vacuum pump and allowing the DI water to imbibe under high vacuum. In order to determine the permeability of the porous media (Kabs) from Equation (1), DI water was injected at different flow rates (10,20,30, and 40 mL/min), and the corresponding pressure drop was recorded.
where L is the length of the pack, cm; A is the cross-sectional area of the pack, cm 2 ; µ is the viscosity of the fluid, mPa.s; Q is the flow rate, cm 3 /s; and ΔP is the differential pressure across the sand-pack, atm.
The drainage process was carried out by injecting 2 pore volumes (PV) of silicon oil at 0.1 mL/min until the water fraction at the production end was less than 1%, and the pressure stabilized. The initial oil saturation was calculated by dividing the volume of produced water during oil flooding by the pore volume. After that, 1 PV of polymer solution or nanopolymer sol was injected followed by 2 PV of water. Cumulative oil recovery was calculated by dividing the sum of oil recovered from the chemical flood by the initial volume of oil in the sand-pack. The aqueous and oil phases were separated by heating the collected samples at 70 °C for 30 min in an oven.

FTIR and TGA Measurements of the Modified Silica NPs
The FTIR spectra of unmodified and modified silica NPs after normalization of the peak area are shown in Figure 1. The asymmetric and symmetric stretching vibrations of C-H groups of MPS for the SiO2-MPS NPs were clearly observed around 2951 and 2481 cm −1 , and the stretching vibration of C=O, methylene (=CH2), and vinyl (-CH=CH2) bending vibrations were observed at 1703, 1456, and 1406 cm −1 [41], respectively. For SiO2-OTES NPs, two peaks appear at 2926 cm −1 and 2858 cm −1 due to the intense symmetric and asymmetric stretching vibrations of the C-H bonds in the octyl groups [42]. A peak also appears at 1392 cm −1 , which is assigned to the asymmetric deformation vibration of C-H bonds due to a slight substitution of the octyl groups in place of the -OH groups of the SiO2 NPs [43]. The TGA thermograms shown in Figure 2 correspond to the unmodified and modified silica NPs. The thermogram of unmodified SiO2 NPs shows the desorption of physisorbed water up to 400 °C and the dehydroxylation of adjacent -OH groups between 400 and 800 °C [44]. The weight losses of each region were 25.71% and 6%, respectively. For modified silica NPs, the weight losses detected up to 100 °C were attributed to volatile solvents and unbound water. The weight loss for SiO2-OTES The TGA thermograms shown in Figure 2 correspond to the unmodified and modified silica NPs. The thermogram of unmodified SiO 2 NPs shows the desorption of physisorbed water up to 400 • C and the dehydroxylation of adjacent -OH groups between 400 and 800 • C [44]. The weight losses of each region were 25.71% and 6%, respectively. For modified silica NPs, the weight losses detected up to 100 • C were attributed to volatile solvents and unbound water. The weight loss for SiO 2 -OTES and SiO 2 -MPS in this region were 13.37% and 15.37%, respectively. The onset of oxidation of SiO 2 -OTES and SiO 2 -MPS was found at 200 • C. Temperatures for maximum oxidation were 245 • C for SiO 2 -OTES, and 300 • C for SiO 2-MPS, and the weight loss was 15.48% and 23.41%, respectively.

Colloidal Stability
The colloidal stability of NP dispersions can be predicted by the magnitude of the ζ-potential. Dispersions with ζ-potential values greater than +30 mV or less than −30 mV typically have a high degree of stability [45]. NP dispersions with a low ζ-potential value will eventually agglomerate under the effect of interparticle attractions. Accordingly, most of the nanofluids (NPs dispersed in DI water) should potentially be unstable, as reported in Table 1. The mechanisms causing the instability of these NPs can be hydrophobic interactions among the modified silica NPs, or the hydrogen bonding between the silanols groups of the unmodified NPs.
It was observed that the unmodified silica nanofluid was stable despite its low ζ-potential value. The same observation was reported by Gun'ko et al. [46]. They suggest that the hydration layer formed between the silanols groups on the silica surface, and the water molecules through hydrogen bonding, prevent the agglomeration of the NPs. Effective steric hindrance leads to net repulsive interparticle forces despite the lack of charges on the silica surface [47]. The hydration effect was, nevertheless, not observed in the other hydrophilic NPs (i.e., Al2O3, TiO2, and Fe(OH)3). TGA analysis showed that the weight loss associated with the desorption of physisorbed water (up to 400 °C) for unmodified SiO2 NPs was 25.71%, while the weight losses for Al2O3, TiO2, and Fe(OH)3 were 0.49%, 3.78%, and 5%, respectively. The instability of Al2O3 and TiO2 in water was previously reported by Hendraningrat and Torsae ter [38].
The ζ-potential for Al2O3 NPs in DI water is +16.4 mV at pH 6.75, which is relatively high, consistent with the fact that this pH is far from their isoelectric point (pH IEP = 9.1) [48]. When alumina is dispersed in water, it behaves like a basic oxide which consumes H + ions. Thus, it possessed positive surface charges [48]. Upon adding SDS, the potential changes sign (−26.9 mV), which could be explained by the adsorption of SDS onto the surface of the positively charged NPs as a bilayer, with the sulfate headgroups exposed to the solution [49]. Formation of a surfactant bilayer with concomitant charge reversal of the alumina particles with anionic sodium dodecyl sulfonate surfactant was previously reported by Somasundaran and Fuerstenau [50]. In general, the addition of SDS increased the negative magnitude of the ζ-potential values of the NPs, but did not improve their stability.
The nanopolymer sols prepared in this study are stable (Table 1), mainly because of the stabilizing effect of the adsorbed polymeric chains onto the NPs surface inducing steric repulsion. In addition, the stability of the suspension increased due to the increase in the viscosity of the dispersion medium [51]. In general, polymer adsorption results from electrostatic and non-electrostatic

Colloidal Stability
The colloidal stability of NP dispersions can be predicted by the magnitude of the ζ-potential. Dispersions with ζ-potential values greater than +30 mV or less than −30 mV typically have a high degree of stability [45]. NP dispersions with a low ζ-potential value will eventually agglomerate under the effect of interparticle attractions. Accordingly, most of the nanofluids (NPs dispersed in DI water) should potentially be unstable, as reported in Table 1. The mechanisms causing the instability of these NPs can be hydrophobic interactions among the modified silica NPs, or the hydrogen bonding between the silanols groups of the unmodified NPs.
It was observed that the unmodified silica nanofluid was stable despite its low ζ-potential value. The same observation was reported by Gun'ko et al. [46]. They suggest that the hydration layer formed between the silanols groups on the silica surface, and the water molecules through hydrogen bonding, prevent the agglomeration of the NPs. Effective steric hindrance leads to net repulsive interparticle forces despite the lack of charges on the silica surface [47]. The hydration effect was, nevertheless, not observed in the other hydrophilic NPs (i.e., Al 2 O 3 , TiO 2 , and Fe(OH) 3 ). TGA analysis showed that the weight loss associated with the desorption of physisorbed water (up to 400 • C) for unmodified SiO 2 NPs was 25.71%, while the weight losses for Al 2 O 3 , TiO 2 , and Fe(OH) 3 were 0.49%, 3.78%, and 5%, respectively. The instability of Al 2 O 3 and TiO 2 in water was previously reported by Hendraningrat and Torsaeter [38].
The ζ-potential for Al 2 O 3 NPs in DI water is +16.4 mV at pH 6.75, which is relatively high, consistent with the fact that this pH is far from their isoelectric point (pH IEP = 9.1) [48]. When alumina is dispersed in water, it behaves like a basic oxide which consumes H + ions. Thus, it possessed positive surface charges [48]. Upon adding SDS, the potential changes sign (−26.9 mV), which could be explained by the adsorption of SDS onto the surface of the positively charged NPs as a bilayer, with the sulfate headgroups exposed to the solution [49]. Formation of a surfactant bilayer with concomitant charge reversal of the alumina particles with anionic sodium dodecyl sulfonate surfactant was previously reported by Somasundaran and Fuerstenau [50]. In general, the addition of SDS increased the negative magnitude of the ζ-potential values of the NPs, but did not improve their stability.
The nanopolymer sols prepared in this study are stable (Table 1), mainly because of the stabilizing effect of the adsorbed polymeric chains onto the NPs surface inducing steric repulsion. In addition, the stability of the suspension increased due to the increase in the viscosity of the dispersion medium [51].
In general, polymer adsorption results from electrostatic and non-electrostatic interactions and the balance between these forces. The adsorption of XG chains onto the surface of the unmodified SiO 2 , Fe(OH) 3 , and TiO 2 NPs likely occurs through hydrogen bonding between -OH groups on the surface of the NPs and the carboxylic groups on the side chains of the XG, and on the positively charged Al 2 O 3 NPs through electrostatic attraction forces. The adsorption onto the modified SiO 2 -MPS and SiO 2 -OTES NPs likely occurs through hydrophobic interactions between the polymer backbone and the R-CH 3 chain of the modifiers on the silica surface. It was observed that the addition of 0.2 wt % NaCl to the nanopolymer sols reduced their ζ-potential values, but did not destabilize them.

Viscosity Measurements
The XG polymer solutions and nanopolymer sols exhibited shear-thinning behavior, which is attributed to the uncoiling and partial alignment of the XG chains at the high shear rate region (Figure 3a). The 4000 ppm XG solutions with 0.3 wt % and 1 wt % salt had slightly higher viscosity than the salt-free solution. This behavior was previously reported by Wyatt and Liberatore [13,52]. They found that, below a critical polymer concentration (C c~2 000 ppm), the viscosity decreases upon addition of salt. The addition of counterions (Na + ) contributes to neutralizing the electrostatic interactions between charges along the backbone of the XG chain, which allows the chain to fold, causing a reduction in the hydrodynamic size of the macromolecules. Since the XG chains decrease in size, the number of interactions with neighboring chains also decreases. For polymer concentrations >C c , the interaction among the XG chains increases, and its effect on the viscosity of the solution becomes more important than the negative effect caused by the decrease in the hydrodynamic size of the macromolecules.
The addition of unmodified silica, SiO 2 -MPS, and SiO 2 -OTES NPs to the XG-SDS solutions has a positive effect on the viscosity values at all salt concentrations (Figure 3). For the salt-free nanopolymer sols, the increment in viscosity can be attributed to the interaction between -OH groups on the surface of the NPs (modified and unmodified) with the carboxylic groups on the trisaccharide side chains of the XG through hydrogen bonding [25] and hydrophobic interactions between the modifiers and the backbone of the polymer. Then, the unmodified and modified silica NPs act as physical crosslinker between polymer chains. When NaCl is added to these nanopolymer sols, the hydrophobic interactions between SDS micelles, NPs, and the XG chains increases, due to the screening of the electrostatic repulsion between them [53,54]. The repulsion forces between the attached SDS Nanomaterials 2019, 9, 94 7 of 13 micelles and SDS-coated NPs will prevent the coiling of the XG chains, facilitating the formation of a tridimensional network with other NP-XG-SDS complexes, increasing the viscosity.
The reduction in viscosity of XG-SDS solution caused by the addition of Fe(OH) 3 , Al 2 O 3 , and TiO 2 NPs suggests high adsorption of polymer molecules at the NP surface. High polymer adsorption on the NP surface reduces the viscosity of the polymer solution because it reduces the polymer concentration in the liquid phase and diminishes the bridging of polymer molecules with the NPs. The reduction in the polymer viscosity was observed at all salinities. Nanomaterials 2019, 9, x FOR PEER REVIEW 7 of 14 hydrophobic interactions between SDS micelles, NPs, and the XG chains increases, due to the screening of the electrostatic repulsion between them [53,54]. The repulsion forces between the attached SDS micelles and SDS-coated NPs will prevent the coiling of the XG chains, facilitating the formation of a tridimensional network with other NP-XG-SDS complexes, increasing the viscosity. The reduction in viscosity of XG-SDS solution caused by the addition of Fe(OH)3, Al2O3, and TiO2 NPs suggests high adsorption of polymer molecules at the NP surface. High polymer adsorption on the NP surface reduces the viscosity of the polymer solution because it reduces the polymer concentration in the liquid phase and diminishes the bridging of polymer molecules with the NPs. The reduction in the polymer viscosity was observed at all salinities. The viscosity data of XG polymer solutions and nanopolymer sols exhibits a good fit to Ostwaldde Waele model (Equation 2). Nevertheless, a previous model developed by the authors [55] and based on multilayer perceptron (MLP) neural network can also be used for predicting the viscosity of nanopolymer sols. The rheological parameters (n and K) are presented in Table 2).
where n is the flow index, dimensionless, and K is the consistency factor, Pa.s n .  The viscosity data of XG polymer solutions and nanopolymer sols exhibits a good fit to Ostwald-de Waele model (Equation 2). Nevertheless, a previous model developed by the authors [55] and based on multilayer perceptron (MLP) neural network can also be used for predicting the viscosity of nanopolymer sols. The rheological parameters (n and K) are presented in Table 2).
where n is the flow index, dimensionless, and K is the consistency factor, Pa.s n .  Table 3 presents the absolute permeabilities and porosities of the sand-packs used in this study and Figure 4 shows the cumulative oil recovery curves for water, polymer, and nanopolymer sols floods. The r 2 values of the ∆P/L vs. Q/A curves used to calculate the absolute permeabilities were between 0.9970 and 0.9999. The cumulative oil recovery for waterflooding (WF) and polymer flooding was 38% and 67%, respectively. The addition of NPs to the salt-free XG solution increased its cumulative oil recovery between 3% and 9%. The increment on oil recovery is attributed to the improvement of the injected fluid and oil mobility ratio. The highest oil recovery was obtained with MPS-and OTES-modified SiO 2 NPs. At 0.3 wt% NaCl, NPs increased the cumulative oil recovery between 1% and 5%, except for SiO 2 -MPS. The oil recovery of SiO 2 -MPS was 2% lower than that of the XG solution. The reduction of the incremental oil recovery showed by all NPs can be attributed to NP and polymer adsorption on the sand grains induced by the addition of NaCl. Since the nanopolymer sols have a pH higher than 2 (between 6.04 and 10.62), the sand grain surface is likely negatively charged. Then, the formation and adsorption of NPs aggregates, and the adsorption of XG chains on the surface of the sand grains [56,57] is promoted by the counterions (Na + ) . The ∆P values during nanopolymer flooding increased as salinity increased, due to the permeability reduction caused by the adsorption of NPs aggregates and XG chains within the porous medium ( Table 3). The highest ∆P value during nanopolymer flooding was obtained by the injection of SiO 2 -MPS nanopolymer sol, which is in agreement with its low cumulative oil recovery. The highest oil recovery was obtained with SiO 2 -OTES NPs, which showed higher increment on XG viscosity and lower ∆P values.  At 1.0 wt % NaCl, the addition unmodified SiO2, SiO2-MPS, SiO2-OTES, and Al2O3 reduced the cumulative oil recovery between 5% and 12%. However, the addition of Fe(OH)3 and TiO2 NPs increased the cumulative oil recovery between 2% and 3%. The performance of the Fe(OH)3 and TiO2 NPs can be attributed to lower polymer adsorption on the surface of the sand grains. The alkaline nature of both nanopolymer sols (Fe(OH)3 pH = 10.62, and TiO2 pH = 8.15) increases the negative charge density of the sand grains [57] and the negative charges on the XG chains through the At 1.0 wt % NaCl, the addition unmodified SiO 2 , SiO 2 -MPS, SiO 2 -OTES, and Al 2 O 3 reduced the cumulative oil recovery between 5% and 12%. However, the addition of Fe(OH) 3 and TiO 2 NPs increased the cumulative oil recovery between 2% and 3%. The performance of the Fe(OH) 3 and TiO 2 NPs can be attributed to lower polymer adsorption on the surface of the sand grains. The alkaline nature of both nanopolymer sols (Fe(OH) 3 pH = 10.62, and TiO 2 pH = 8.15) increases the negative charge density of the sand grains [57] and the negative charges on the XG chains through the deprotonation of the carboxylic acids [12], thus, the repulsive forces between the XG chains and the sand grains increases, which leads to lower adsorption of the polymer. The high ∆P values during the injection of Fe(OH) 3 and TiO 2 nanopolymer sols are related mainly to the adsorption of NPs aggregates on the sand grains. In general, Fe(OH) 3 and TiO 2 nanopolymer sols were the only nanopolymer sols that show the same incremental oil recovery at 0.3 wt % and 1.0 wt % NaCl. It seems that the increment of the repulsive forces between the XG chains and the sand grains can mitigate the negative effect caused by the NP adsorption on the sand grains.

Conclusions
In this study, formulation and characterization of nanopolymer sols of XG with hydrophobic and hydrophilic metal oxide NPs is reported. The properties of the nanopolymer sols were explored using ζ-potential and viscosimetry. The performance of all nanopolymer sols was evaluated by conducting heavy oil recovery tests in linear sand-packs. The role of the physical and chemical interactions between NPs and XG solutions, and the relationship between their hybrid structure, properties, and their performance were investigated.
The colloidal stability of the NPs dispersed in DI water and XG solution was evaluated by the ζ-potential measurements. The NP dispersions in deionized (DI) water exhibited low ζ-potential values, which suggested low dispersion stability. Nevertheless, the addition of XG induced steric stabilization. It was observed that the addition of untreated silica, SiO 2 -MPS, and SiO 2 -OTES NPs improved the thickening behavior of the XG solution. However, Fe(OH) 3 , Al 2 O 3 , and TiO 2 NPs decreased the viscosity of the polymer solution. The displacement tests showed that the breakthrough time occurred later with the injection of the nanopolymer sols, since the reduction of mobility ratio increased the macroscopic efficiency. At 0 wt % and 0.3 wt % NaCl, the addition of NPs increased the cumulative oil recovery between 3% and 9%, and between 1% and 5%, respectively. At 1.0 wt % NaCl, unmodified SiO 2 , SiO 2 -MPS, SiO 2 -OTES, and Al 2 O 3 reduced the cumulative oil recovery between 5% and 12%, whereas Fe(OH) 3 and TiO 2 NPs increased it between 2% and 3%.
Supplementary Materials: The following are available online at http://www.mdpi.com/2079-4991/9/1/94/s1, Table S1: Average particle size measurement using DLS. Acknowledgments: Laura M. Corredor gratefully acknowledges Ecopetrol S.A for the scholarship to pursue her graduate studies, and Bernie Then, Amitabha Majumdar and Ola Jabar for their technical assistance.

Conflicts of Interest:
The authors declare no conflict of interest.

Abbreviations
The following abbreviations are used in this manuscript: