Enhanced CO2 Solubility in Hybrid MCM-41: Molecular Simulations

ARTICLE pubs.acs.org/Langmuir

Enhanced CO2 Solubility in Hybrid MCM-41: Molecular Simulations and Experiments Linh Ngoc Ho,† Javier Perez Pellitero,† Fabien Porcheron,*,† and Roland J.-M. Pellenq‡,§ changeur de Solaize, BP 3, 69360 Solaize, France IFP Energies nouvelles, Rond-Point de l'E Centre Interdisciplinaire des Nanosciences de Marseille, CNRS, Campus de Luminy, 13288 Marseille, Cedex 09, France § Department of Civil and Environmental Engineering, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, Massachusetts 02139, United States † ‡

ABSTRACT: Grand canonical Monte Carlo simulations are performed in a hybrid adsorbent model in order to interpret the CO2 solubility behavior. The hybrid adsorbent is prepared by confining a physical solvent (OMCTS) into the pores of a mimetic MCM-41 solid support. As a result, simulated adsorption isotherms of CO2 nicely match the experimental data for three distinctive systems: bulk solvent, raw MCM-41, and hybrid MCM-41. The microscopic mechanisms underlying the apparition of enhanced solubility are then clearly identified. In fact, the presence of solvent molecules favors the layering of CO2 molecules within the pores; therefore, the CO2 solubility in the hybrid adsorbent markedly increases in comparison to that found in the raw adsorbent as well as in the bulk solvent. In addition, a good understanding of confined solvents’ properties and solid surface structures is essential to fully evaluate the efficiency of hybrid adsorbents in capturing CO2. The sorbent
solid interactions along with the solvent molecular size’s impact on CO2 solubility are therefore investigated in this study. We found that an ideal hybrid system should possess a weak solvent
solid interaction but a strong solvent
CO2 interaction. Besides, an optimal solvent size is obtained for the enhanced CO2 solubility in the hybrid system. According to the simulation results, the solvent layer builds pseudomicropores inside the mesoporous MCM-41, enabling more CO2 molecules to be absorbed under the greater influence of spatial confinement and surface interaction. In addition, the molecular sieving effect is clearly observed in the case of larger solvent molecular sizes.

’ INTRODUCTION Controlling and reducing the concentration of CO2 in the atmosphere has been gaining significant public and scientific attention as an effective mitigation measure against global warming. The most dominant source of CO2 has been inarguably pinpointed to be due to ever-growing industry activities in the last century, particularly from the use of fossil fuels. For instance, one important source of carbon dioxide emissions comes from coal-fired power stations where the flue gas at atmospheric pressure is predominantly composed of N2 (∼90%) with a small fraction of CO2 (∼10%). Until now, a large number of technologies have been adopted to removing CO2 from flue gas streams, for example, amine scrubbing, adsorption, membrane diffusion, and cryogenic distillation.1
5 From these conventional techniques, the combined advantages of both absorption (a high selectivity and capacitive solvent) and adsorption (absence of corrosion, reduced solvent loss) have recently been materialized in a new generation of hybrid adsorbents.6 However, the greatest challenge for this approach is to synthesize suitable adsorbents displaying high CO2 selectivity and an affordable regeneration cost. A majority of work in this area has focused on grafting amine molecules into a solid support.7
10 In most of those studies, the r 2011 American Chemical Society

authors expect a strong interaction between amines and CO2, similar to the mechanism of a conventional chemical absorption process. An additional advantage of this approach is that solids are simple to handle and do not give rise to corrosion problems. As a result, the use of physical solvents in hybrid adsorbents has yet to receive adequate attention from researchers because of their characteristic low CO2 solubility at low pressures. In recent work, Song et al.11,12 observed the apparition of enhanced CO2 solubility in a hybrid adsorbent composed of an aminated polymer confined within MCM-41. In other words, the CO2 solubility in the hybrid adsorbent was found to be greater than that in the bulk fluid and in the raw solid. The microscopic mechanisms governing this behavior, however, were not fully established. Moreover, Miachon et al.13,14 noted that a remarkable enhancement of hydrogen and light hydrocarbon solubility can be achieved when solvents are confined in mesoporous solids. For instance, by confining hexane in low-density mesoporous materials (i.e., silica aerosol), the authors observed an “oversolubility” Received: April 7, 2011 Revised: May 19, 2011 Published: June 06, 2011 8187

dx.doi.org/10.1021/la2012765 | Langmuir 2011, 27, 8187–8197

Langmuir

ARTICLE

Figure 1. (a) Final structure of MCM-41 after optimization, and (b) the structure obtained by replicating the simulation box. Color code: yellow, silicon; red, oxygen; white, hydrogen.

of H2 by a factor of 4 compared with that in the bulk hexane system. In recent work, we studied the apparition of enhanced solubility in hybrid adsorbents in the framework of CO2 removal.15 Hybrid adsorbents were prepared by confining physical solvents (propylene carbonate, N-methyl-2-pyrrolidone) within the porosity of a solid support (alumina) using two different impregnation methods. Experimental thermodynamic measurements (i. e., CO2 adsorption isotherms) were carried out on these hybrid adsorbents, and enhanced CO2 solubility was clearly observed as compared with that of the bulk solvent and the raw solid. In addition, GCMC simulations were performed using a simple model where the solid is modeled as a structureless slit pore and the adsorbate molecules are considered to be a Lennard-Jones fluid. The results showed that the presence of solvent molecules in the hybrid system favored the layering of CO2 within the pore; and the resulting local density profile was then markedly increased over the one found in the raw adsorbent because more carbon dioxide molecules could be structurally accommodated in the pore volume. In this study aimed at providing deeper insight into the enhanced phenomenon occurring in the hybrid adsorbent, we report molecular simulations of CO2 adsorption in hybrid MCM41. We focus on mesoporous MCM-41 because of its extended use as an efficient adsorbent for gas adsorption and separation. First, we generate realistic models as the discrete and electrostatic nature of CO2 is taken into account. Then, the structure of the porous material is described using the recently introduced MCM-41 atomic-scale pore model.16
19 Subsequently, grand canonical Monte Carlo simulations are performed to interpret the CO2 solubility behavior occurring in the hybrid systems. The experimental solubility of CO2 is measured to evaluate the performance of adsorbent models. Finally, the effects of the confined solvent’s type and solid’s surface structures are also addressed to fully evaluate the efficiency of hybrid adsorbents in capturing CO2.

’ EXPERIMENTAL SECTION Hybrid MCM-41 Adsorbent Preparation. Mesoporous MCM41 is synthesized by using ionic surfactant hexadecyltrimethylammonium bromide (CTAMBr), as a templating agent. Initially, 5.1 g of CTAMBr is dissolved in 133 mL of water and 144.35 g of ammoniac (NH4OH, 25%) and stirred at ambient temperature for 10 min. Subsequently, 22.32 g of TEOs is quickly added. The temperature is controlled to heat the mixture gradually to 35
C and held at that value for 48 h under stirring. The adsorbents are then recovered by filtration and washed with 600 mL of deionized water and then with 200 mL of

ethanol. Finally, these sorbents are dried at 100
C overnight and calcinated at 550
C for 4 h. The hybrid adsorbents are prepared by confining octamethylcyclotetrasiloxane (OMCTS, Merck) with full pore loading using the wet impregnation method.20 Solubility Measurements. The adsorption experiment system consisting of a reactor connected upstream to a gas reservoir is integrated within an oven to regulate the operating temperature and is designed to operate at pressures ranging from vacuum up to 10 bar and at temperatures up to 120
C. First, 5 g of adsorbent is loaded into the reactor cell. Before starting the experiments, the cell and the gas reservoir are vacuumed and the system is programmed to achieve a steady temperature of T = 40
C. CO2 is then injected into the gas reservoir. The valve between the reactor and the reservoir is opened and remained that way during the adsorption. To proceed to the next step of the isotherm, this valve is then closed and a small amount of CO2 is admitted into the reservoir. This procedure is repeated in a pressure domain ranging from 0 to 3 bar to determine the complete CO2 adsorption isotherm. Pressures and temperatures within the cell and the reservoir are recorded every 1 s. More details on the preparation step and the solubility measurements can be referred to in our previous work.15

’ SIMULATION DESIGN The system studied in this work contains two different adsorbates (CO2 and OMCTS solvent) confined in a porous MCM-41 material. MCM-41 Adsorbent Model. The MCM-41 pore model is generated following a procedure similar to that proposed by Pellenq et al.16
19 to prepare numerical Vycor samples. The mesoporous adsorbent is generated by carving out cylindrical pores in a 42.8  64.2  42.8 Å3 matrix of amorphous silica. The structure obtained consists of four half-pores located at four facets of a simulation box (Figure 1). After carving out the mesopores, hydrogen atoms are added in order to saturate the cleaved bonds. All of the silicon atoms linked with three hydroxyls are then removed. To ensure that the simulation box is neutrally charged and that none of the silicon and oxygen atoms have dangling bonds, the system is once again saturated with hydrogen atoms. Once the mesoporous sample has been generated, a geometric optimization based on the universal force field (UFF)21 is performed by means of the Materials Studio Forcite package.22 The pore size of the MCM-41 model is adapted in order to match the experimental CO2 adsorption isotherm at T = 40
C. Additional details about the optimization of the pore size are given in the Results and Discussion section. The final structure 8188

dx.doi.org/10.1021/la2012765 |Langmuir 2011, 27, 8187–8197

Langmuir

ARTICLE

Table 1. MCM-41 Model Properties properties

model MCM-41

no. of O atoms

1892

no. of Si atoms

774

no. of H atoms

Table 2. Lennard-Jones Potential Parameters for Wall Atoms and Fluid Molecules Used in the GCMC Molecular Simulations system CO2

684 3

σ (Å)

ε (K)

C

2.757

28.129

0.78144

O

3.033

80.500


0.39072

atoms

q (e)

skeletal density, g/cm

2.4

porosity, cm3/g

0.93

OMCTS

LJ center

7.700

351.36

17.5

MCM-41

Si

3.804

155.858

0.12220

O H

3.033 2.846

48.115 7.654


0.06157 0.03199

pore radius, Å 2

5 to 6

OH groups/nm

contains two cylindrical pores of 17.5 Å radius, has a skeletal density of 2.4 g/cm3 and a porosity of 0.93 cm3/g, and possesses 5 to 6 OH groups per nm2. Note that the value of the skeletal density is higher than that of amorphous silica (∼2.2 g/cm3). This property can be explained from the determination of the free pore volume after the geometrical optimization process as well as from the saturation of the system with hydrogen atoms. The full properties of the generated MCM-41 model are listed in Table 1. Factor λ, used to estimate the smoothness of the pore surface,23 is also determined according to λ¼

SC 2nπRp lz

ð1Þ

where SC is the accessible surface of the adsorbent; n is the total number of pores in the system; and Rp and lz are the radius and the length of the pore model, respectively. The value λ = 1.29 confirms the roughness of the MCM-41 surface. The dispersive
repulsive interactions are modeled using the Dreiding force field23,24 plus a set of electrostatic charges. CO2/Solvent Models. CO2 is represented by means of the EPM model,25 which consists of three Lennard-Jones (LJ) sites plus three electrostatic charges placed on the atomic nuclei. This model has been widely used because of its capability to reproduce both equilibrium and dynamic properties of CO2.26 Finally, because of its quasi-spherical shape and hence low dipole moment (μ = 0.22 D), the OMCTS solvent is modeled by means of a single LJ center.27 Interaction Potential Models. The interactions between the various sites in adsorbate and adsorbent models are calculated by the sum of a Lennard-Jones 12
6 potential and a Coulombic electrostatic contribution 2 !12 !6 3 ε ε ij ij 5 þ qi qj
ð2Þ Uij ðrÞ ¼ 4εij 4 rij rij 4πε0 where r is the distance between the centers of any pair of sites; ε is the depth of the potential well; σ is the site diameter; q is the partial electrostatic charge of each site; ε0 is the vacuum permittivity; and the i, j subscripts represent the different species of the system (S = solid, A = CO2, and B = solvent). The cross interaction parameters (σij, εij) between unlike sites are calculated using the Lorentz
Berthelot and the Kong combination rules depending on the difference in molecular size.28 The Lorentz
Berthelot combination rules are given by 8 < σ ¼ σii þ σ jj ij 2 ð3Þ : εij ¼ pffiffiffiffiffiffiffiffi εii εjj

where (σii, εii) and (σjj, εjj) are the pure compound LJ parameters. As mentioned above, because of the large size difference between CO2 and solvent molecules, the Kong combination rule29,30 is used for the energetic cross interaction parameter εij: qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi εAB σ6AB ¼ εAA σ6AA εBB σ 6BB ð4Þ Finally, in the case of solvent
substrate interactions, the cross interaction parameter is adjusted to reproduce the experimental adsorption isotherm of CO2 in hybrid MCM-41 at T = 40
C and then the transferability of the model is verified by comparing the simulation isotherm to the experimental data at T = 30
C. The Lennard-Jones potential parameters for the totality of the force centers used in the model are summarized in Table 2. Simulation Methodology. We perform grand canonical Monte Carlo (GCMC) simulations to compute the adsorption of carbon dioxide molecules in the hybrid solid model, which is prepared by confining the solvent to MCM-41. Periodic boundary conditions are applied in all three dimensions. The van der Waals interactions are evaluated with a cutoff radius of rc = 21.4 Å, and long-range electrostatic interactions are taken into account by employing the Ewald summation technique.31 To address the effect of the pore length, a longer MCM-41 pore (i.e., lz = 213.9 Å) is considered. The simulation results of two different pore length models correctly predict the experimental isotherms of CO2, indicating that a pore length of lz = 42.8 Å is large enough to neglect the effect of finite system size. Subsequently, the hybrid adsorbent is created by filling the pore volume with the solvent molecules. The first series of GCMC simulations is performed by adjusting the chemical potential of the reservoir in order to simulate the solvent loading process and reproduce the experimental concentration of the solvent in the pore (i.e., about 3 mmol/cm3). The resulting configuration is used as an initial configuration prior to simulating the CO2 adsorption process. After preparing the hybrid adsorbent model, we perform GCMC simulations of CO2 adsorption. Throughout the simulation process, the adsorbent is considered to be rigid and the number of OMCTS molecules is kept constant. The conformations of the adsorbate molecules are sampled by means of translation for both CO2 and OMCTS and also by rotation, insertion, and deletion Monte Carlo trials for CO2 molecules. Typical runs consist of 1  107 MC steps for each pressure point to guarantee equilibration, followed by 1  107 production MC steps to calculate averages of the thermodynamic properties of interest. All simulations are performed at temperatures of T = 30 and 40
C with a pressure range of PCO2 = 2
300 kPa. The theoretical adsorption isotherm expressed in terms of the volume 8189

dx.doi.org/10.1021/la2012765 |Langmuir 2011, 27, 8187–8197

Langmuir

ARTICLE

Figure 2. Experimental (filled symbols, dashed line) and simulation (open symbols) absorption isotherms of CO 2 in bulk OMCTS at T = 40
C.

Figure 3. Experimental (filled symbols, dashed lines) and simulation (open symbols) adsorption isotherms of CO2 in MCM-41 at T = 30
C (circles) and T = 40
C (squares).

concentration is constructed by computing the average CO2 density inside the pore model, FCO2, as a function of the pressure of the gas phase in equilibrium with the adsorbed CO2 phase, yielding FCO2 ¼

ÆNCO2 æ Vbox

ð5Þ

where ÆNCO2æ is the average number of CO2 molecules in the simulation box. To understand the microscopic mechanisms of CO2 adsorption occurring in the model hybrid adsorbent, the local molecular distribution profiles of CO2 and solvent are also computed.

’ RESULTS AND DISCUSSION Model Validation. Before considering the simulation of CO2 adsorption in the hybrid adsorbent, the consistency of the molecular models used in this work needs to be validated. For this purpose, the adsorption behavior of CO2 has been compared with the experimental data in three different models: OMCTS bulk solvent, raw MCM-41, and a hybrid solid (OMCTS confined in MCM-41). First, we determine that the EPM2 CO2 model correctly reproduces the experimental isotherm of carbon dioxide in the bulk OMCTS, proving that the CO2 and solvent models used in this work are adequately representative of the CO2 absorption in the bulk solvent (Figure 2). Then, the models are validated for the case of the adsorption of CO2 in raw MCM-41 where the experimental isotherms are available at two different temperatures. As mentioned in the previous section, the size of the pore is adapted to match the experimental isotherm at T = 40
C. Subsequently, the consistency of the solid model is confirmed by comparing the simulated result with the experimental one at T = 30
C. Figure 3 shows excellent agreement between simulations and experiments over the whole pressure range at the two aforementioned temperatures. This clearly demonstrates that the employed pore model is sufficiently realistic to represent the MCM-41 material. Finally, the adsorption of CO2 in hybrid MCM-41 is considered. Because there is no available experimental information for the case of solvent
solid interactions, both Lorentz
Berthelot and Kong combination rules are tested. The results

Figure 4. Experimental (filled symbols, dashed lines) and simulation (open symbols) adsorption isotherms of CO2 in hybrid MCM-41 at T = 30
C (circles) and at T = 40
C (squares).

show that neither Lorentz
Berthelot nor Kong combination rules can fit the hybrid adsorbent experimental isotherm at T = 40
C, implying that the original combination rules may not be totally applicable to our particular system without modification. Consequently, the solvent
solid cross interaction parameter εSB is slightly varied to match the experimental data. The optimized value not only reproduces the experimental solubility data at T = 40
C but also gives excellent agreement for that at T = 30
C, as seen in Figure 4. In conclusion, the models used in this work are precise enough to represent CO2 sorption isotherms in different systems: a solid support, a bulk solvent, and a hybrid adsorbent. The local organization of CO2 and OMCTS molecules in the hybrid adsorbent at P = 3 bar is reported along with the CO2 profile in the raw adsorbent (Figure 5). Clearly, the CO2 molecular profile in the hybrid adsorbent is rather different than the one found in the raw solid, thereby proving that the two adsorption mechanisms are different. We also observe that the OMCTS-MCM-41 hybrid adsorbent displays a lower adsorption capacity than the raw solid. Thus, in the remainder of this article, 8190

dx.doi.org/10.1021/la2012765 |Langmuir 2011, 27, 8187–8197

Langmuir

ARTICLE

Figure 5. Molecular distribution profiles for OMCTS (open squares), CO2 in the model of raw MCM-41 (b), and hybrid MCM-41 (2) at PCO2 = 3 bar and T = 40
C.

we will investigate the influence of several energy and geometric parameters in order to optimize the capacity of the hybrid adsorbent. Effect of Energy Interactions on the Mechanisms of CO2 Adsorption. In this section, we aim to find the most efficient hybrid system for capturing CO2. In fact, the mechanism of CO2 adsorption in hybrid adsorbents is a very complex phenomenon that depends on many parameters such as thermodynamic conditions, the solid surface structure, and the interaction between the adsorbed molecules in the pores (solvent
solvent and solvent
CO2) as well as between the surface and these adsorbates (surface
solvent and surface
CO2). Therefore, it is imperative to understand the correlation among interactions of various components in the mechanism of adsorption in hybrid systems. The effect of confined solvent type and solid surface structures is subsequently addressed in order to evaluate fully the efficiency of hybrid adsorbents in capturing CO2. Effect of Solvent
Solid and CO2
Solvent Interactions on the CO2 Adsorption Behavior. First, to investigate the effect of solvent
solid interaction on the CO2 adsorption behavior, the Lorentz
Berthelot combination rules are applied to like fluid and CO2
substrates interactions whereas the modified Kong rules are used in the case of the CO2 solvent (as discussed in the model validation section). The dependence of CO2 solubility on the solvent
solid interaction will then be examined using our model with different values of εSB (εSB = kAε0SB, where ε0SB is obtained from the Lorentz
Berthelot combination rule). The results exhibit decreasing solubility with a stronger interaction between solvent and solid (Figure 6, top). We then focus on the effect of solvent
CO2 interactions. The value of the solvent
solid interaction (kA = 1) is then fixed, and the value of the interaction between solvent and CO2 (εAB) is allowed to vary (εAB = kBε0AB, ε0AB is applied from the Kong combination rule). As a result, we clearly observe enhanced CO2 solubility with increasing interactions between solvent and CO2 molecules (Figure 6, bottom). In conclusion, by screening the interaction parameters, properties of the ideal solvent used to prepare an efficient hybrid system to capture CO2 can be highlighted. The solvent should exhibit a weak interaction with the solid support but a substantial affinity for CO2. Effect of Solvent
Solid Interaction on the CO2 Local Organization. To obtain an in-depth view of the microscopic structure of CO2 confined in the hybrid solid model, the local

Figure 6. CO2 solubility in hybrid MCM-41 with different solvent
solid interactions (top) and different solvent
CO2 interactions (bottom) at PCO2 = 3 bar and T = 40
C. The line is intended to guide the eye.

molecular distribution profiles for the adsorbates are generated at a pressure of PCO2 = 3 bar for various interactions of solvent
solid systems and two intensities of the CO2
solvent interaction kB = 1.0 and 1.7 (Figure 7). The profiles show an oscillatory damped behavior reflecting the organization of adsorbed molecules in discrete layers on the solid surface. Besides, the positions of adsorbates molecules in each type of system are almost the same (e.g., the first peak is located about 12.25 and 15.75 Å from the pore center for solvent and CO2 species, respectively). As the solvent
solid interaction increases, the intensity of the solvent’s first peak develops while that of CO2 shrinks. This may be due to the fact that as the solvent
solid affinity becomes stronger, more solvent molecules adsorb on the surface of MCM-41, leading to decreasing CO2 packing at the contact layer. This clearly illustrates the competitive adsorption mechanism between the two adsorbates on the surface of the modeled system. We first focus on the systems with the lowest CO2
solvent interaction (kB = 1.0). For the hybrid adsorbent displaying weak solvent
solid interactions (Figure 7a), there is only one significant layer of adsorbed CO2 molecules whereas another layer of solvent is formed further away from the solid wall. When the solid
solvent interaction increases, we observe the formation of a second layer of CO2 about 7 Å from the pore wall. Another noticeable result is 8191

dx.doi.org/10.1021/la2012765 |Langmuir 2011, 27, 8187–8197

Langmuir

ARTICLE

Figure 7. Molecular distribution profiles for solvent (0) and CO2 (2) with varying interactions of solvent
solid systems (a and a0 , kA = 0.25; b and b0 , kA = 0.33; c and c0 , kA = 0.45; d and d0 , kA = 1; e and e0 , kA = 3) for a weak solvent
CO2 interaction (left, kB = 1) and a strong solvent
CO2 interaction (right, kB = 1.7) for PCO2 = 3 bar and T = 40
C.

8192

dx.doi.org/10.1021/la2012765 |Langmuir 2011, 27, 8187–8197

Langmuir that, contrary to the first peak, the second peak of the solvent diminishes as the second peak of CO2 increases (Figure 7e). This additional CO2 layer crops up between two solvent layers, suggesting the apparition of a CO2 adsorption mechanism in a pseudopore created by the layers of solvent molecules. This observation leads to a very important feature of the mechanism governing adsorption on a hybrid adsorbent. In fact, the pore of the hybrid system can be divided into two regions: one near the surface ruled by the adsorption mechanism of CO2 on the solid surface and one away from the MCM-41 wall, where the adsorption between two layers of solvent prevails. In other words, it could be reasonably concluded that the sorption process in hybrid adsorbents simultaneously follows two different sorption mechanisms. As illustrated by the results, the intensity of the first peak of CO2 transfers to the second one, corresponding to the transformation of CO2 adsorption on the solid to one on the solvent layers in the hybrid system. Because we now focus on the system with a higher solvent
CO2 interaction intensity (kB = 1.7), we can observe a competition between the two sorption mechanisms. For a weak solvent
solid interaction, the result confirms the dominant effect of adsorption on the surface solid over adsorption on the layers of solvent (Figure 7a0 ). However, when the solvent
solid interaction increases, the simulation results show a significant growth of the second peak of CO2, as carbon dioxide molecules now adsorb in a pseudopore of solvent layers (Figure 7e0 ). Effect of Solvent
CO2 Interaction on the CO2 Local Organization. Figure 8 displays the evolution of adsorbates’ local molecular distribution profiles at a pressure of PCO2 = 3 bar as a function of the solvent
CO2 interaction modification while the solvent
solid cross interaction parameter is kept constant at kA= 0.25. As a result, the increased solvent
CO2 interaction consistently yields not only a significant improvement in the CO2 adsorption capacity between the solvent layers but also a greater number of CO2 molecules adsorbed at the contact layer near the surface. Furthermore, we notice the constant positions of solvent molecules whereas these of CO2 are shifted toward the solvent layers. As previously discussed in the work of Tripathi et al.,32 the equilibrium distribution of fluid molecules is determined by the competition between the energetics of the system and the entropy penalty. Indeed, the adsorbed layer forming as a result of fluid
fluid association tends to stand away from the wall to avoid an entropy penalty, and the surface
fluid association tends to attract fluid molecules toward the surface to reduce the system’s energy. However, the introduction of solvent molecules into the hybrid system creates an additional potential field, thereby disturbing the predictable distribution of CO2 molecules and enabling more CO2 molecules to pile up at the interface. In other words, the solvent acts as a “structuring promoter” for CO2 molecules. In a region further away from the surface, the CO2
surface interaction is weaker and hence CO2 molecules are mainly under the influence of the potential field created by solvent molecules. A higher solvent
CO2 interaction would then result in a greater number of CO2 molecules adsorbed in this area. Hybrid System Optimization for Enhanced CO2 Solubility Based on the Energetic Interactions. As discussed above, the intricate relationship among interactions of solvent
solid as well as solvent
CO2 with respect to the CO2 adsorption capacity of the hybrid adsorbent confirms that an efficient hybrid system, where the greatest solubility of CO2 occurs, should possess a mild

ARTICLE

Figure 8. Molecular distribution profiles for solvent (0) and CO2 (2) with varying solvent
CO2 interactions (a, kB = 0.67; b, kB = 1; c, kB = 1.7; d, kB = 2; e, kB = 2.5) at a fixed solvent
CO2 interaction (kA = 0.25) at PCO2 = 3 bar and T = 40
C.

solvent
solid interaction and an intense CO2
solvent interaction. To obtain a clearer picture of the enhanced CO2 solubility 8193

dx.doi.org/10.1021/la2012765 |Langmuir 2011, 27, 8187–8197

Langmuir

ARTICLE

Figure 10. CO2 adsorption isotherms in the models of bulk solvent (9), raw MCM-41 (2), and hybrid adsorbent (b) at T = 40
C for kA = 0.45 and kB = 1.7.

Figure 9. CO2 solubility difference between hybrid MCM-41 and a raw solid (O) and bulk solvent (9) with different solvent
solid interactions (top, kB = 1.7) and different solvent
CO2 interactions (bottom, kB = 0.25) at PCO2 = 3 bar and T = 40
C.

phenomenon, simulations of CO2 adsorption in the hybrid system are compared with that in the raw solid and bulk solvent. To make the presentation more comprehensible, we plot only the results of CO2 adsorption at PCO2 = 3 bar, but the overall trend in adsorption is consistent along the whole isotherm. For a fixed solvent
CO2 interaction (e.g., kB = 1.7), the hybrid system loses its superior adsorption performance over that of the raw solid and the bulk solvent with increasing solvent
solid interactions (Figure 9, top). However, for a constant solvent
solid interaction (e.g., kA = 0.25), following the increasing trend in solvent
CO2 interactions, we observe not only an enhanced adsorption performance over that of the raw support but also an optimum solvent
CO2 interaction where a maximized enhanced CO2 solubility as compared with that of the bulk is observed (Figure 9, bottom). The adsorption trend steadily increases but markedly drops after the solvent
CO2 interaction reaches a value of εAB = 1.7ε0AB. This observation had not been identified in our previous work with a simple hybrid adsorbent model (i.e., the CO2 solubility of confined solvent was always greater than that of bulk solvent).15 The strength of the potential force formed by fluid
fluid association in a bulk environment is

Figure 11. CO2 solubility in hybrid MCM-41 (kA = 0.45, kB = 1.7) with different solvent sizes at PCO2 = 3 bar and T = 40
C. The line is intended to guide the eye.

stronger than that created by a combination of the fluid
wall association and an additional “structuring” field in a hybrid sorbent. The results hence suggest a limited enhancement in the adsorption mechanism in hybrid systems in which not all strong solvent
CO2 interactions would result in an efficient hybrid adsorbent. According to the above analysis, we obtain a system (kA = 0.45 and kB = 1.7) displaying enhanced solubility in the hybrid system. Figure 10 reports the simulation results of CO2 adsorption at T = 40
C in the different systems for this set of parameters. The average density of CO2 in decreasing order of hybrid adsorbent > solid support > solvent is in good qualitative agreement with the results of our previous study.15 Effect of Solvent Size on the Mechanisms of CO2 Adsorption. In this section, we investigate the effect of the solvent size on the CO2 adsorption behavior. For this purpose, GCMC simulations of CO2 adsorption at PCO2 = 3 bar and T = 40
C are carried out with the optimized hybrid MCM-41 model (i.e., kA = 0.45, kB = 1.7) and different solvent sizes (σB). The resulting CO2 density displays a maximum for σB = 25 Å. Above this value, the 8194

dx.doi.org/10.1021/la2012765 |Langmuir 2011, 27, 8187–8197

Langmuir

ARTICLE

Figure 12. Molecular distribution profiles for solvent (0) and CO 2 (2) with varying solvent size (a, σB = 10 Å; b, σB = 13 Å; c, σB = 18 Å; d, σB = 25 Å; e, σB = 32 Å) P CO 2 = 3 bar and T = 40
C.

adsorbed amount markedly decreases (Figure 11). The reason for achieving an optimum solvent size could be better understood

by studying the adsorbates’ molecular distribution profiles, as shown in Figure 12. 8195

dx.doi.org/10.1021/la2012765 |Langmuir 2011, 27, 8187–8197

Langmuir

ARTICLE

model drops to zero in cases of larger solvent sizes (>33 Å) simply because of the molecular sieving effect.

Figure 13. CO2 solubility difference between hybrid MCM-41 and the bulk solvent with different solvent sizes at PCO2 = 3 bar and T = 40
C. The line is intended to guide the eye.

For instance, when σB = 10 Å, Figure 12a shows the presence of both adsorbates inside the pore where two monolayers are formed in the hybrid system for each species: one near the surface and the other close to the center of the pore. However, the first layer of CO2 molecules is strongly adsorbed on the MCM-41 surface because of their smaller size and stronger interaction with the surface atoms, as compared with solvent molecules. The second layer of CO2 is then created between two layers of solvent, governed by the adsorption mechanism on the pseudopore. When the size of the solvent increases to σB = 13 Å (Figure 12b), the adsorbates’ distribution profiles look similar to a certain extent to those observed for σB = 10 Å, but the position of the solvent layer shifts toward the center of the pore as a result of a larger solvent size. Moreover, the number of solvent molecules in the pore decreases as the solvent size increases because the solvent density is fixed during the GCMC simulations. This arrangement hence leaves more vacancies for carbon dioxide molecules to be fitted in at the contact layer, resulting in a greater amount of CO2 adsorbed in the pore. When σB = 18 Å, only one layer of each adsorbates is observed in the pore. We observe a significant increase in CO2 density at the contact layer (Figure 12c). The formation of pseudopores by the solvent molecules now induces the adsorption of CO2 molecules in a pseudomicropore. Hence, the adsorbed amount of CO2 in these pores significantly increases as a result of additional support for spatial confinement and surface interaction. Furthermore, by increasing the solvent’s size up to σB = 25 Å, the CO2 molecules reach the highest packing density. As a result, complete monolayer formation can be achieved. Besides, the pore size is large enough to facilitate more CO2 molecules to be adsorbed continuously to form a multilayer, resulting in a broader peak in Figure 12d. This optimum point would yield the highest differential CO2 adsorption between the hybrid adsorbent and the bulk solvent (Figure 13). However, the absorbed number of CO2 molecules in the hybrid pore is significantly reduced when the solvent’s size increases further (Figure 12e). The pseudomicropores then probably become too small for CO2 molecules to pass through freely. In other words, there is a limitation on the capacity of pseudomicropores to accommodate CO2 in the hybrid system. In fact, the adsorption capacity of CO2 in the hybrid MCM-41

’ CONCLUSIONS To interpret the CO2 solubility behavior in a modeled system of hybrid adsorbent comprehensively, an atomistic model representing mesoporous MCM-41 silica material is constructed and used in the grand canonical Monte Carlo simulations. Besides, CO2 is simulated with a three-atom molecular model, and the solvent is presented as a spherical nonpolar molecule. To make the model as realistic as possible, electrostatic interactions are also taken into account by adopting the Ewald summation technique. Good qualitative agreement is found between the simulated and experimental data, proving that our methodology adequately reproduces the adsorption isotherms of CO2 for various systems: the MCM-41 support, a bulk solvent, and the hybrid MCM-41 adsorbent. In this work, the effects of confined solvent type and solid surface structures are further addressed to evaluate fully the efficiency of hybrid adsorbents in capturing CO2. By examining the interplay among interactions of solvent
solid and solvent
CO2 on the adsorbed amount of CO2 in the hybrid system, we have found that an efficient hybrid system, where the greatest solubility of CO2 is achieved, should possess a weak solvent
solid interaction but a strong solvent
CO2 interaction. In good agreement with the results from our previous study,15 the microscopic mechanisms clearly demonstrate the apparition of enhanced solubility in an optimized hybrid system because the presence of solvent molecules favors the layering of CO2 within the pore. The resulting local CO2 density profile in the hybrid adsorbent markedly increases compared to that in the raw solid as well as in the bulk solvent. Moreover, we found that the size of the confined solvent plays a very important role in the adsorption performance. The number of CO2 molecules adsorbed in the hybrid sorbent displays an optimum as the size of the solvent increases. The solvent layers ultimately create additional micropore channels inside the mesoporous MCM-41 model, which facilitates more CO2 molecules to be adsorbed. At the optimum solvent size, the CO2 molecules are possibly adsorbed continuously, forming a multilayer, displaying the highest packing density. Work is in progress to study the potential of these sorbents in CO2 capture by performing experiments and molecular simulations on hybrid adsorbents using different solvents. ’ AUTHOR INFORMATION Corresponding Author

*E-mail: [email protected].

’ REFERENCES (1) Figueroa, J. F.; Fout, T.; Plasynski, S.; McIlvried, H.; Srivastava, R. D. Int J. Greenhouse Gas Control 2008, 2, 9–20. (2) Abanades, J. C. Chem. Eng. 2002, 90, 303–306. (3) Rochelle, G. T. Science 2009, 325, 1652–1654. (4) Versteeg, G. F.; van Duck, L. A. J.; Van Swaaij, W. P. M. Chem. Eng. Commun. 1996, 144, 113–158. (5) Porcheron, F.; Gibert, A.; Mougin, P.; Wender, A. Environ. Sci. Technol. 2011, 45, 2486–2492. (6) Chaffee, A. L.; Knowles, G. P.; Liang, Z.; Zhang, J.; Xiao, P.; Webley, P. A. Int J. Greenhouse Gas Control 2007, 1, 11–18. (7) Choi, S.; Drese, J. H.; Jones, Ch. W. Chem. Sus .Chem. 2009, 2, 769–854. 8196

dx.doi.org/10.1021/la2012765 |Langmuir 2011, 27, 8187–8197

Langmuir

ARTICLE

(8) Knowles, G. P.; Delaney, S. W.; Chaffee, A. L. Ind. Eng. Chem. 2006, 45, 2626–2633. (9) Huang, H. Y.; Yang, R. T.; Chinn, D.; Munson, C. L. Gas. Ind. Eng. Chem. 2003, 42, 2427. (10) Sayari, A.; Harlick, P.J. E. Ind. Eng. Chem. 2006, 45, 3248–3255. (11) Song, C. S.; Xu, X. C.; Andresen, J. M.; Miller, B. G.; Scaroni, A. W. Surf. Sci. Catal. 2004, 153, 411–416. (12) Song, C. S. Catal Today 2006, 115, 2–32. (13) Miachon, S.; Syakaev, V. V.; Rakhmatullin, A.; Pera-Titus, M.; Caldarelli, S.; Dalmon, J.-A. ChemPhysChem 2008, 9, 78–82. (14) Pera-Titus, M.; El-Chahal, R.; Rakotovao, V.; Daniel, C.; Miachon, S.; Dalmon, J.-A. ChemPhysChem 2009, 10, 2082–2089. (15) Ho, N. L.; Porcheron, F.; Pellenq, R.J.-M. Langmuir 2010, 26, 13287–13296. (16) Pellenq, R. J. M.; Rousseau, B.; Levitz, P. E. Phys. Chem. Chem. Phys. 2001, 3, 1207. (17) Pellenq, R. J.-M.; Levitz, P. E. Mol. Phys. 2002, 100, 2059–2077. (18) Coasne, B.; Pellenq, R. J.-M. J. Chem. Phys. 2004, 120, 2913– 2922. (19) Coasne, B.; Pellenq, R. J.-M. J. Chem. Phys. 2004, 121, 3767– 3774. (20) Hogendoorn, J. A.; Van Swaaij, W. P. M.; Versteeg, G. F. Chem. Eng. Sci. 1994, 49, 3421–3438. (21) Rappe, A. K.; Casewit, C. J.; Colwell, K. S.; Goddard, W. A.; Skiff, W. M. J. Am. Chem. Soc. 1992, 114, 10024. (22) Materials Studio; Accelrys Inc.: San Diego, CA, 2003. (23) Zhuo, S.; Huang, Y.; Hu, J.; Liu, H.; Hu, Y.; Jiang, J. J. Phys. Chem. C 2008, 112, 11295–11300. (24) Mayo, S. L.; Olafson, B. D.; Goddard, W. A. J. Phys. Chem. 1990, 94, 8897. (25) Harris, J.; Yung, K. H. J. Phys. Chem. 1995, 99, 12021–12024. (26) Nieto-Cragui, C.; De Bruin, T.; Perez-Pellitero, J.; Bonet, J. B.; Mackie, A. D. J. Chem. Phys. 2007, 126, 064509. (27) Ayappa, K. G.; Mishra; Ratan, K. J. Phys. Chem. B 2007, 111, 14299–14310. (28) Delhommelle, J.; Millie, P. Mol. Phys. 2001, 99, 619–625. (29) Kong, C. L. J. Chem. Phys. 1973, 59, 1953. (30) Kong, C. L. J. Chem. Phys. 1973, 59, 2464. (31) Jorge, M.; Seaton, N. A. Mol. Phys. 2003, 100, 2017–2023. (32) Tripathi, S.; Chapman, W. G. Condens. Matter Phys. 2003, 6, 523–540.

8197

dx.doi.org/10.1021/la2012765 |Langmuir 2011, 27, 8187–8197