Computational Study of CO2 Storage in Metal−Organic Frameworks

Jan 16, 2008 - Under practical application conditions, MOFs show higher CO2 storage capacity than both zeolites and carbon materials, and the suitable...
16 downloads 12 Views 321KB Size
1562

J. Phys. Chem. C 2008, 112, 1562-1569

Computational Study of CO2 Storage in Metal-Organic Frameworks Qingyuan Yang, Chongli Zhong,* and Jian-Feng Chen Department of Chemical Engineering, Key Lab for Nanomaterials, Ministry of Education, Beijing UniVersity of Chemical Technology, Beijing 100029, China ReceiVed: September 13, 2007; In Final Form: NoVember 10, 2007

In this work a systematic computational study was performed to investigate the effects of organic linker, pore size and topology, and the electrostatic fields on the adsorption and diffusion behaviors of CO2 in nine typical metal-organic frameworks (MOFs), showing that the high CO2 storage capacity achieved in MOFs is a complex interplay of these structural properties. Under practical application conditions, MOFs show higher CO2 storage capacity than both zeolites and carbon materials, and the suitable pore size is between 1.0 and 2.0 nm. For MOFs with pore size located in the above range, the larger the accessible surface area and free volume, the higher the CO2 storage capacity can be achieved in practical applications. In addition, this work shows that the self-diffusivity of CO2 in the MOFs is comparative in magnitude with that of zeolites.

1. Introduction The excessive discharge of CO2 into atmosphere, mainly due to the combustion of vast amounts of carbon-based fossil fuels, has led to one of the most serious global environment problems facing the community.1 However, in the transition toward a more sustainable energy economy, fossil fuels likely remain the dominating sources of worldwide energy supply for the foreseeable future. Thus, CO2 capture and storage from flue exhaust plays a significant role in mitigating its effect on climate change. Furthermore, the removal of CO2 for natural gas upgrade or hydrogen purification is also of economic and separations technology importance. A widely recognized strategy for the feasible CO2 capture technology is that minimal environmental impact and low costs should be achieved. The currently commercial amine-based systems used for CO2 removal bears several drawbacks such as corrosion control and considerable energy need for solvent regeneration.2 Pressure swing adsorption (PSA) technology based on porous adsorbent is known to be one of the most efficient and affordable processes for CO2 removal from flue streams.3 Many research efforts have been carried out to identify and develop a suitable porous material with high CO2 affinity and capacity.4-6 Although the conventionally involved zeolite materials like zeolites 13X and NaY have been claimed to be most adequate for CO2 capture,7,8 it is difficult to regenerate them without significant heating which leads to low productivity and great expense.9 Owing to their flexibility to design through control of the architecture and chemical functionality of the pores, metalorganic frameworks (MOFs) have emerged as a new family of nanoporous materials that offer promising applications for gas storage and separation.10 While most of the recent studies have been concentrated on the adsorption of methane11,12 and hydrogen,13-15 some MOF materials have also shown extremely high CO2 storage capacity and very desirable isotherm shapes,16 which stimulates an increasing interest toward understanding the adsorption behavior of CO2 in MOFs.17-19 * Corresponding author. Tel.: +86-10-64419862. E-mail: zhongcl@ mail.buct.edu.cn.

Computational chemistry is a powerful method to give deep insight into the phenomena studied at the microscopic level.20 Although extensive theoretical investigations have been carried out on adsorption and diffusion of hydrogen21-25 and hydrocarbons12,26-28 in MOFs using either quantum chemical calculations or molecular simulations, only a few computational studies have been performed on CO2. For example, Skoulidas and Sholl assessed the adsorption and diffusion of CO2 in MOF-5 using molecular dynamics (MD) simulations.29 Yang and Zhong performed grand canonical Monte Carlo (GCMC) simulations to study the CO2 removal in MOFs for purification of light alkanes30 and synthetic gas obtained from steam re-forming of natural gas.31 Atomic simulations were conducted to compare the storage capacity for pure CO2 and CH4 as well as the separation of their binary mixture in three different adsorbents, MOF-5, silicalite, and C168 schwarzite.32 More recently, the enthalpies of adoption at low coverage for CO2 in two versions of MIL-53(Al) material were estimated by GCMC simulations to confirm the experimental observations by Ramsahye et al.33 In this work, using a hierarchical approach combining density functional theory (DFT) calculation and molecular simulation, a systematic computational study was performed to investigate the properties of CO2 adsorption and diffusion in several typical MOFs with different pore sizes and topologies, including six isoreticular metal-organic frameworks (IRMOFs-1, -8, -10, -11, -14, and -16),11 Mn-MOF,17 MOF-177,34 and Cu-BTC.35 The purpose of this work is to explore the effects of organic linker, pore size, pore topology, and the electrostatic field existed on the adsorption and diffusion of CO2 in MOFs to provide useful information for developing new MOF materials with long-term viability in CO2 capture and storage. 2. Models and Simulation Method 2.1. MOF Structures. In this computational study, the guestfree framework structures of the selected MOFs were constructed from their corresponding experimental single-crystal X-ray diffraction (XRD) data11,17,34,35 using Materials Studio Visualizer (see Figure S1 in the Supporting Information).36 Excepting for IRMOF-11, all the porous IRMOF materials feature the same primitive cubic topology with the octahedral

10.1021/jp077387d CCC: $40.75 © 2008 American Chemical Society Published on Web 01/16/2008

CO2 Storage in Metal-Organic Frameworks

J. Phys. Chem. C, Vol. 112, No. 5, 2008 1563

TABLE 1: Structural and Limiting CO2 Adsorption Properties of the MOFs Studied in This Work material

space groupa

pore shapea

dporea (nm)

Fcrysta (g/cm3)

Saccb (m2/g)

Vfreeb (cm3/g)

qst0 c (kJ/mol)

IRMOF-1 IRMOF-8 IRMOF-10 IRMOF-14 IRMOF-16 IRMOF-11 MOF-177 Cu-BTC Mn-MOF

Fm3hm Fm3hm Fm3hm Fm3hm Pm3hm R3hm P3h1c Fm3hm Pna21

cubic cubic cubic cubic cubic cubic/catenation pore/channel pocket/channel cage/channel

1.09/1.43 1.25/1.71 1.67/2.02 1.47/2.01 2.33 0.70/1.20 1.08/1.18 0.50/0.90 0.55/0.45

0.59 0.45 0.33 0.37 0.21 0.76 0.43 0.88 1.59

3748 4360 4938 4800 5882 2867 4688 2368 554

1.36 1.87 2.66 2.30 4.46 0.92 1.96 0.82 0.50

13.73 12.67 11.96 13.28 10.25 20.86 14.43 25.60 24.78

a

Obtained from the XRD crystal data.11,17,34,35

b

Calculated with the Materials Studio package.36

Zn4O(CO2) clusters linked by aromatic dicarboxylate linkers.11 The catenated structure of IRMOF-11 contains doubly interpenetrated frameworks having the same underlying cubic topology.11 MOF-177 consists of octahedral Zn4O(CO2) clusters and BTB (1,3,5-benzenetribenzoate) linkers, which reveals a remarkably open three-dimensional structure with two kind cavities connected in all directions by channels.34 Mn-MOF synthesized by Dybtsev et al.17 has a three-dimensional charge neutral network, and the adamantane-like cages in this network are connected to each other via a small window to form a 1D zigzag channel. Cu-BTC has a three-dimensional channel structure connecting a system of tetrahedral-shaped cage accessible through small triangular windows.35 It should be pointed out that the crystal structure of Cu-BTC taken from Chui et al.35 includes axial oxygen atoms weakly bonded to the Cu atoms, which correspond to water ligands. In this work, dry Cu-BTC was adopted with these oxygen atoms removed.29,30 Details of the structural properties for these materials are summarized in Table 1. The accessible surface area (Sacc) and total free volume (Vfree) of each MOF material were estimated using the “Atoms Volume & Surfaces” calculation within the Materials Studio package.36 Because of the highly dependency on adsorbate size, the former was calculated by a probe molecule with diameter equal to the kinetic diameter of CO2 (0.33 nm),37 while a probe size of 0.0 nm was applied to determine the total free volume not occupied by the framework atoms. 2.2. Force Fields. Force fields play an important role in molecular simulations. In this study, CO2 was modeled as a rigid linear molecule with three charged LJ sites located on each atom. A combination of the site-site LJ and Coulombic potentials was used to calculate the CO2-CO2 intermolecular interaction. The LJ potential parameters for atom O (σO ) 0.305 nm and O/kB ) 79.0 K) and atom C (σC ) 0.280 nm and C/kB ) 27.0 K) in CO2 molecule with C-O bond length l ) 0.116 nm were taken from the TraPPE force field developed by Potoff and Siepmann.38 In this model, partial point charges centered at each LJ site (qO ) -0.35 e and qC ) 0.70 e) approximately represent the first-order electrostatic and second-order induction interactions. Such potential model has been successfully used to simulate the adsorption of CO2 in zeolites39 and MOF mateials.30,31 For the MOFs studied, similar to CO2, an atomistic representation was used to model all of them. The same potential model, that is, a combination of the site-site LJ and Coulombic potentials, was also used to calculate the interactions between adsorbate molecules and adsorbents. Due to that the all-atom OPLS (OPLS-AA) force field40 can distinguish the types of atoms in the frameworks of the MOF materials in details, as before,20,30,31 this force field was adopted in this work, and parts of the LJ parameters were refined to give better reproduction of the experimental CO2 adsorption data (see below). Since the widely used Lorentz-Berthelot combining rules were adopted

c

Obtained in this work.

in the TraPPE force field to increase its transferability, the same combining rules were also used to calculate the LJ cross parameters describing the interactions of the CO2 molecules with the framework atoms in the MOFs for consistency. In addition, the atomic partial charges of the MOFs were required in the simulations, which were calculated using density functional theory (DFT) methods, as given below. 2.3. Density Functional Theory Calculations. In all simulations, atomic partial charges for the MOFs are required as input parameters. The model clusters of all the MOFs used for atom partial charge calculations are given in Figure S2 in the Supporting Information. For the cleaved cluster of Mn-MOF, the terminations are connected with metal atoms Mn. Thus, to consider the environmental effect of these metal atoms in the real system, this cluster was terminated by light metal atom Li as Sagara et al.,21 which was considered to be the simple model for the cluster bound to the metal centers.25 For the cleaved clusters of other MOFs, the terminations are all connected with the organic linkers, and the clusters saturated with -CH3 group were found to be good enough for the calculations. The electrostatic charges were used as the atomic partial charges, and the ChelpG method was adopted, which has been recognized as the most popular and reliable electrostatic charge calculation method.41 DFT calculations using the UB3LYP functional were carried out to compute the charge distributions, and the basis set LANL2DZ was used for metal atoms Zn, Cu, and Mn, while 6-31+G* was used for the remaining atoms. For heavy atoms, effective core potential (ECP) is often chosen in ab inito calculations to reduce the amount of necessary computation, and LANL2DZ is a collection of double-ζ basis sets, which is one of the most common ECP basis sets for complexes involving transition metal elements.42 This is why we selected this basis set for metal atoms Zn, Mn, and Cu. All the calculations were performed using the GAUSSIAN 03 suite of programs,43 and the calculated results are also given in Figure S2 in the Supporting Information. 2.4. Grand Canonical Monte Carlo Simulation Details. The conventional GCMC simulation technique was employed for studying the adsorption of CO2 in the MOFs. Details of this method could be found elsewhere.44 The chemical potentials needed in the GCMC simulations were calculated from NPT ensemble Monte Carlo simulations of bulk gas using the testparticle insertion method.45 On the basis of the simulated chemical potentials at various pressures, relationships between pressure and chemical potential were established to convert pressures to chemical potentials and vice versa. The simulation box representing Mn-MOF contained 18 (3 × 3 × 2) unit cells, while 8 (2 × 2 × 2) unit cells were adopted for other MOFs. The simulations with larger boxes showed that no finite-size effects existed using the above boxes. All the MOFs were treated as rigid with atoms frozen at their crystallographic positions during simulations, since the effects of the dynamics of MOFs

1564 J. Phys. Chem. C, Vol. 112, No. 5, 2008

Yang et al.

become significant only when the guests are large and/or strong guest-host interactions exist in the system at room temperature. Periodic boundary conditions were applied in all three dimensions. A cutoff radius of 1.3 nm was applied to calculate the LJ interaction energies, and the long-range electrostatic interactions were handled using the Ewald summation technique with tin-foil boundary condition.44 To increase the computational efficiency, the potential energies between the adsorbate and adsorbent were initially tabulated on a series of threedimensional grid points with grid spacing 0.015 nm. During the simulations, the potential energy at any position in the adsorbent was determined by interpolation.46 For each state point, GCMC simulation consisted of 1.5 × 107 steps to guarantee equilibration followed by 1.5 × 107 steps to sample the desired thermodynamic properties. The output of a GCMC simulation is the absolute amount adsorbed Nabs, while the excess amount adsorbed Nex is measured experimentally. To make a comparison, the conversion from absolute to excess amount was calculated by the method of Myers and Monson.47 Another important thermodynamic quantity of interest that can be obtained from a GCMC simulation is the isosteric heat of adsorption qst. According to the fluctuation theory, it could be calculated from48

qst ) RT -

〈UffN〉 - 〈Uff〉〈N〉 〈N2〉 - 〈N〉〈N〉

-

〈UsfN〉 - 〈Usf〉〈N〉 〈N2〉 - 〈N〉〈N〉

TABLE 2: Potential Parameters for the Framework Atoms of the MOFs /k (K) atom

σ (nm)

IRMOF-1

IRMOF-11

Cu-BTC

Mn-MOF

O Ccarboxyl Cbenzene Calkane Hbenzene Halkane Hcarboxyl Zn Cu Mn

0.296a

63.41b

63.41c

73.98b

0.375a 0.355a 0.350a 0.242a 0.250a 0.242a 0.246d 0.311d 0.264d

52.84a 28.18b

52.84a 31.00c 26.57c 15.10a 15.10a

44.91b 35.23a

55.02c 41.22c

15.10a

15.10a 7.55a

62.40d

62.40d 2.52d 6.54d

a Taken from the original OPLS-AA force field of Jorgensen et al.40 Taken from our previous works.30,31 c Obtained in this work. d Taken from the all-atom UFF force field50 (they are missed in the OPLS-AA force field).

b

diffusion coefficient Ds by the mean-square displacements (MSD) method.22,44 For each state point, the average diffusivity Ds and its uncertainty were calculated from the results of 10 independent simulation runs with different initial configurations. 3. Results and Discussion

(1)

where 〈 〉 indicates the ensemble average, R is the gas constant, and N is the number of molecules adsorbed. The first and second terms are the contributions from the molecular thermal energy and adsorbate-adsorbate interaction energy Uff, respectively, while the remaining term is the contribution from the adsorbentadsorbate interaction energy Usf. In the limit of zero coverage, the affinity between adsorbate and adsorbent can be evaluated by isosteric heat of adsorption qst0 at infinite dilution. In this wok, qst0 was derived from the canonical ensemble (NVT) MC simulations with a single gas molecule in one unit cell of each MOF to represent zero coverage.32 Actually, we found that the results calculated in this method are in close agreement with that extrapolated from a series of GCMC simulations at low-pressure region. 2.5. Molecular Dynamics Simulation Details. In this work, a constant temperature MD with momentum scaling method49 was used to investigate the diffusion behaviors of CO2 in the MOFs at T ) 298 K. The six-value second-order and five-value first-order Gear predictor-corrector algorithms were applied to solve the center of mass translational motion and the angular rotational motion in quaternion, respectively. The details can be found elsewhere.44 The simulation boxes used in the MD simulations were the same as those in the GCMC simulations, and periodic boundary conditions were also applied in all three dimensions. The long-range Coulombic interactions were evaluated through the Ewald summation method and LJ interactions with a 1.3 nm cutoff radius. The adsorbed CO2 densities obtained in previous GCMC simulations were used in MD simulations. Simulations were performed as follows: Molecules were randomly inserted into MOF lattices, and then relaxed using approximately 50 000 Monte Carlo moves. Following relaxation, the simulation systems were allowed to equilibrate with 100 000 MD steps before the final simulations were run for 500 000 MD steps to sample the diffusion properties of interest. The time step used for all the simulations was taken as 2.0 fs. During each simulation, the trajectory of the system was saved every 100 steps to subsequently calculate the self-

3.1. Refinement of Part Force Field Parameters. Since the parameters of the OPLS-AA force field were developed for liquid simulations, the existing parameters may not properly represent the interactions of the atoms of the solid MOF materials with the adsorbate molecules. Therefore, on the basis of the charge distributions obtained from the DFT calculations, part of the LJ parameters of the OPLS-AA force field were refined. Table 2 shows the potential parameters for the interactions of CO2 with the IRMOF-1 and Cu-BTC frameworks obtained in our previous works.30,31 To better represent the adsorption isotherms of CO2 in IRMOF-11 at 298 K and in Mn-MOF at 195 K, the energy parameters for oxygen and carbon in IRMOF-11 and those for oxygen and carbon in MnMOF were refined in this work, as shown in Table 2. Figure 1 demonstrates that the obtained parameters enable good reproduction of the corresponding experimental results.16,17 In addition, the potential parameters refined for IRMOF-1 in our previous work were used to predict the adsorption of CO2 in MOF-177. Figure 1a shows that the predicted results are in excellent agreement with the experimental data.16 This indicates that the potential parameters for IRMOF-1 are applicable to MOF-177 and other IRMOF materials (except IRMOF-11 with catenated structure) consisting of octahedral Zn4O(CO2) clusters and aromatic linkers with similar chemical functionality. 3.2. Isosteric Heats of CO2 Adsorption at Infinite Dilution in MOFs. The affinity of a MOF material represents its energetic interaction strength with the adsorbate molecules, and it could be described by the limiting properties such as the isosteric heat of adsorption qst0 at infinite dilution. Table 1 gives the calculated values of qst0 for CO2 adsorption in each MOF material. As can be seen from this table, the values of qst0 in IRMOFs-1, -8, -10, and -16 have the order IRMOF-16 < IRMOF-10 < IRMOF-8 < IRMOF-1, which is consistent with gradually decreasing sequence of pore size. As to IRMOF-14, although it has larger pore size than IRMOF-8, more aromatic carbon atoms are contained in the organic linkers of IRMOF14, resulting in higher qst0. On the other hand, qst0 for CO2 adsorption in IRMOF-11 is significantly enhanced because of its pore space considerably restricted by interpenetration of two frameworks, and MOF-177 shows larger qst0 than all the IRMOF

CO2 Storage in Metal-Organic Frameworks

J. Phys. Chem. C, Vol. 112, No. 5, 2008 1565

Figure 1. Comparison of simulated and experimental adsorption isotherms of CO2: (a) in IRMOF-11 and MOF-177 at 298 K;16 (b) in Mn-MOF at 195 K.17

Figure 2. Simulated excess adsorption isotherms of CO2 in MOFs as a function of pressure: (a) gravimetric capacity; (b) volumetric capacity.

materials considered except IRMOF-11 due to the different topology with smaller pore size. Having similar pore sizes, CuBTC and Mn-MOF show comparative qst0, which are much larger than all the other MOFs considered due to their much smaller pore sizes. Generally speaking, for MOFs the smaller the pore size, the larger the qst0. Finally, it should be noted that, for CO2 in Cu-BTC, our simulated qst0 is in close agreement with the experimental value.51 3.3. Prediction of CO2 Adsorption Isotherms in MOFs Up to 6.0 MPa. With the refined potential parameters, the excess adsorption isotherms of CO2 in these MOFs at 298 K were predicted as a function of the bulk pressure up to 6.0 MPa, as shown in Figure 2 in both gravimetric and volumetric units. The presence in all the isotherms of a maximum value is a typical feature of excess adsorption isotherms, which occurs when the density of the bulk gas increase faster than that of the adsorbate phase as a function of pressure. Figure 2 also illustrates that the sequence for gravimetric capacity is not exactly consistent with that for volumetric capacity. For example, at 4.5 MPa, IRMOF-10 shows larger gravimetric capacity than MOF-177, while the latter exhibits greater volumetric capacity than the former. In addition, the high uptake achieved at low pressure does not mean also the large uptake at high pressure. Generally speaking, besides a strong affinity with adsorbate molecules, a high-quality adsorbent should also possess large accessible surface area, high free volume, and low adsorbent framework density as well as ideal pore topology. Evidently, the complex results shown in Figure 2 are the interplay outcome of these main influencing factors. Therefore, to provide a guideline for screening new MOF materials for

CO2 storage, it is necessary to systematically investigate the impacts of these factors on the CO2 adsorption capacity of MOFs. 3.4. Effect of Heat of Adsorption at Low Pressure. As is well-known, for a porous material the smaller the pore size or pore volume, the deeper overlap of potential, leading to stronger adsorption of adsorbates. A series of IRMOFs with same primitive cubic topology provides an excellent benchmark to embark this quantitatively. To make this behavior more intuitionally, Figure 3a shows the relationship between the isosteric heats of adsorption qst0 and the volumetric amounts adsorbed in IRMOFs-1, -8, -10, -14, and -16 at low pressure (P ) 0.02 MPa). As can be seen from Figure 2, there is an excellent correlation between them, which has also been found for H2 adsorption in the IRMOF materials by Frost et al.24 In their work, they found that the H2 amount adsorbed correlates with the heat of adsorption at low pressure (loading). They attributed this result to the fact that, at low loadings, IRMOF materials with the strongest enthalpic interactions to the adsorbed molecules show the highest levels of adsorption, due to small pores increasing the interactions between adsorbates and the frameworks. We also examined the dependency of adsorption amount at low pressure on the isosteric heat of adsorption qst0 in the MOF materials with different pore topologies, as shown in Figure 3b. Though there is no obvious correlation as that in Figure 3a, the general trend is the same. This illustrates that the strength of the interactions between CO2 and MOFs plays a predominant role at low pressure, while the effects of other factors become evident only at relative high pressure.

1566 J. Phys. Chem. C, Vol. 112, No. 5, 2008

Yang et al.

Figure 3. Volumetric capacities of CO2 adsorbed at 0.02 MPa vs isosteric heat of adsorption at infinite dilution in the following: (a) IRMOFs with the same primitive cubic topology; (b) MOFs with different topologies.

Figure 4. Gravimetric capacities for CO2 adsorbed at 3.0 MPa: (a) vs accessible surface area of MOFs; (b) vs free volumes of MOFs.

3.5. Effect of Accessible Surface Area and Free Volume at Moderate Pressure. For practical applications, the amount of CO2 adsorbed in MOFs at room temperature and moderate pressure (around 3.0 MPa) is of great interest. Therefore, the influences of the structural properties of MOFs on CO2 adsorption were studied and discussed in detail at 298 K and 3.0 MPa. Figure 4a shows the dependency of the adsorption amounts on the accessible surface area of MOFs with various topologies. It can be seen from Figure 4a that a good linear correlation exists between them for most MOFs except for MnMOF and IRMOF-16. Similar correlations have also been observed by Frost et al.24 for H2 adsorption in IRMOFs at intermediate pressures. Mn-MOF has very narrow pore size, low surface area, and small free volume, and its surface has already been fully occupied by CO2 molecules at pressure lower than 1.0 MPa, as shown in Figure 2. As to IRMOF-16, although it has the largest accessible surface (5882 m2/g) among the studied MOFs, due to its very large pore size, the adsorbateadsorbent interactions are too weak to physically adsorb many molecules on its solid surface at modetate pressure. Thus, the amount adsorbed in IRMOF-16 does not show evident correlation with its accessible surface area up to 3.0 MPa as depicted in Figure 4a. Therefore, for practical applications, it may be concluded that Mn-MOF and IRMOF-16 are not good candidates for CO2 storage. In addition, though IRMOF-14 has stronger interactions with CO2 than IRMOF-10 (see Figure 3), Figure 4a shows that the latter presents higher adsorption capacity because of its larger accessible surface area. This observation indicates that the intensity of interactions is not the predominant influencing factor for CO2 adsorption in MOFs at moderate pressure. Figure 4b gives the correlation between the adsorption capacity and total free volume. Although the correlation is not as good as that between the adsorption capacity and accessible surface area as shown in Figure 4a, it is clear that, at moderate pressure, CO2 adsorption capacity is also proportional to the

free volume of MOFs. Similar tendency has also been observed by Frost et al.24 for H2 adsorption in a series of IRMOF materials. As can be found from Figure 4a,b, IRMOF-10, IRMOF-14, and MOF-177, with comparative gravimetric adsorption capacities, are the most promising MOFs for CO2 storage for practical applications. Therefore, more discussion was paid to them at room temperature and 3.0 MPa. Although gravimetric capacity is often a major indicator in screening the CO2 storage materials, there are practical limits associated with the tank volume required to house the adsorbent for CO2 storage, which make volumetric capacity just a critical parameter to be considered. To compare these three materials more clearly, the simulated excess adsorption isotherms of CO2 in them as a function of pressure were shown in Figure 5a in volumetric units. Obviously, in this case, the sequence of adsorption capacity at 3.0 MPa is exactly opposite to that in gravimetric capacity, which is due to the canceling effects derived from the different crystal densities (MOF-177, 0.43 g/cm3; IRMOF-14, 0.37 g/cm3; IRMOF-10, 0.33 g/cm3) of the three MOF materials (see Table 1). Figure 5b,c shows the simulated absolute gravimetric and volumetric capacities in the three MOFs at 298 K, as a function of pressure up to 6.0 MPa. Evidently, the absolute adsorption capacities in the three MOFs show behaviors similar to the excess capacities as discussed above. On the basis of our simulations and the above discussions, it is clear that the pore size of a MOF material plays an important role on CO2 storage capacity in MOFs under practical application conditions. It seems that the suitable pore size is between 1.0 and 2.0 nm. For MOFs located in the above pore size range, large CO2 storage capacity may be achieved in practical applications, and the larger the accessible surface area and free volume, the higher the CO2 storage capacity. 3.6. Effect of Electrostatic Fields in MOFs. In our previous work,30 we found that the electrostatic interactions produced

CO2 Storage in Metal-Organic Frameworks

Figure 5. Simulated adsorption capacities in the three MOFs as a function of pressure: (a) excess volumetric capacity; (b) absolute gravimetric capacity; (c) absolute volumetric capacity.

by the atomic partial charges of the MOF framework can greatly enhance the separation of mixtures in which the components have largely different polarities. To investigate the effects of electrostatic interactions on the CO2 storage capacity in the above three well-behaved MOFs, additional GCMC simulations were performed by switching off the electrostatic interactions between the CO2 molecules and the framework atoms. We define the contribution of the CO2-MOF electrostatic interactions to the total adsorption capacity as (Ntotal - NVdw)/Ntotal, where Ntotal and NVdw denote the absolute amount adsorbed with and without the CO2-MOF electrostatic interactions, respectively. The simulation results shown in Figure 6a demonstrate that, for IRMOF-10 and IRMOF-14 with the same primitive cubic topology, the low-coverage regime is enhanced approximately by 20%, while, for MOF-177, due to its different topology, the contribution of CO2-MOF-177 electrostatic interactions shows larger values (about 30%) than both IRMOF-10 and IRMOF-

J. Phys. Chem. C, Vol. 112, No. 5, 2008 1567 14. In the work of Garberoglio et al.,23 they concluded that the charge-quadrupole interactions enhanced substantially (approximately 30-40%) the adsorption of H2 in MOFs at 77 K and low pressures but did not significantly change the amount adsorbed at room temperature. However, this work shows that for CO2 the case is different. On the basis of the partial point charges used in the potentials for CO2 and H2 molecules, the absolute value for the quadrupole moment of CO2 (-1.51 × 10-39 C m2) is much larger than that of H2 (2.05 × 10-40 C m2). Thus, the electrostatic interactions between CO2-MOF are much stronger than those between H2-MOF, resulting in the CO2-MOF electrostatic contribution up to 30% at room temperature and low pressures as shown in Figure 6a. Interestingly, for the three MOFs, the electrostatic contributions all decreases monotonically with increasing pressure (loading) and contribute only a few percent at high pressures (under 3%). Figure 6b shows the simulated pressure (or loading) dependency of isosteric heat of adsorption qst for CO2 in MOF-177 at 298 K, where the contributions from both the CO2-CO2 and CO2-MOF-177 interactions are also given. According to the fluctuation theory, the isosteric heat of adsorption is related to the derivative of adsorption energy with respect to loading, which can be used to deduce information on the adsorption process. Figure 6b demonstrates that the contribution from the adsorbate-adsorbent interactions decreases with increasing pressure (loading), indicating the heterogeneous nature of the MOF-177 surface, since CO2 molecules first occupied the more energetically favorable sites, followed by the less favorable sites, leading to a decreasing contribution from the adsorbateadsorbent interactions. An examination of Figure 6a,b shows that the heterogeneity of MOF-177 surface comes partly from the heterogeneity in electrostatic field. The contribution from the adsorbate-adsorbate interactions shows the opposite tendency, since the higher the pressure, the closer packed of the adsorbate molecules. The behavior of the total qst is the result of the cooperative contributions from the two parts. 3.7. Comparison of MOFs with Other Adsorbents. Extensive investigations have been carried out on the storage of CO2 in other porous adsorbents such as zeolites and carbonbased materials. Figure 7 shows the comparison of the adsorption isotherms of CO2 in the three well-behaved MOFs with those in zeoltes 13X7 and carbon-based material MAXSORB4 in terms of both gravimetric capacity and volumetric capacity. Note that both of the referencing materials have been reported to have quite high CO2 uptake, especially zeolite 13X has been claimed to show the highest gravimetric capacity for CO2 adsorption among zeolites.8,16 As can seen from Figure 7, although at the low-pressure region both zeolite 13X and MAXSORB can adsorb more CO2 than the three MOFs, for practical applications (around 3.0 MPa), the three MOFs can store much more CO2 than both zeolite 13X and MAXSORB. As shown in Figure 7a, the adsorption capacities are 35.6, 34.2, and 32.7 mmol/g at 3.0 MPa for IRMOF-10, IRMOF-14, and MOF-177, respectively, which are approximately 4.7 times that in zeolite 13X (7.3 mmol/g) and 1.5 times in MAXSORB (23.5 mmol/g). In addition, Figure 7b shows that the volumetric adsorption capacities are 262.4, 288.0, and 312.4 cm3 (STP)/ cm3 for IRMOF-10, IRMOF-14, and MOF-177, respectively, which are 1.6 times that in zeolites 13X (185.3 cm3 (STP)/ cm3) and 2.0 times that in MAXSORB (142.6 cm3 (STP)/cm3). Thus, it can be concluded that MOFs are promising materials for capturing and storage of CO2. 3.8. Diffusivity of CO2 in MOFs. The diffusivity of CO2 in MOFs is an important property in practical applications. Thus, the self-diffusivities of CO2 in the three MOFs were further investigated by MD simulations at 298 K. The simulation results,

1568 J. Phys. Chem. C, Vol. 112, No. 5, 2008

Yang et al.

Figure 6. (a) Effect of electrostatic interactions on CO2 adsorption at 298 K in the three MOFs. (b) Various contributions to the isosteric heat of adsorption for CO2 in MOF-177 at 298 K, as a function of pressure.

Figure 7. Comparison of the adsorption isotherms of CO2 in MOFs with those in zeoltes 13X7 and carbon-based material MAXSORB4 at 298 K: (a) gravimetric capacity; (b) volumetric capacity.

Figure 8. Self-diffusivities of CO2 in IRMOF-10, IRMOF-14, and MOF-177 at 298 K as a function of the following: (a) loading; (b) pressure.

as a function of both loading and pressure, are shown in Figure 8. Obviously, the dependency of the self-diffusivity of CO2 on loading or pressure in the three MOFs is similar, showing a monotonic decrease with increasing loading or pressure. This behavior is one of the most common forms of loading (pressure) dependency observed in nanoporous materials, arisen from a natural consequence of steric hindrance between diffusing molecules.29 Figure 8 also shows that the diffusivities in the three MOFs are comparative in magnitude within the studied loading and pressure ranges. Furthermore, the magnitude is comparable to that in IRMOF-129 and silica zeolites.52,53 Skoulidas54 has observed that the diffusion of Ar in Cu-BTC is

very similar to Ar diffusion in silica zeolites in magnitude, concentration, and temperature dependence. Thus, the current result agrees with his finding. 4. Conclusions The simulations show that, for a MOF material to have high CO2 adsorption capacity, a suitable pore size between 1.0 and 2.0 nm is essential and, for MOFs in this pore size range, the larger the accessible surface area and free volume, the higher the CO2 storage capacity can be achieved in practical applications. Although the CO2-MOF electrostatic interaction can

CO2 Storage in Metal-Organic Frameworks contribute as large as 30% to the total adsorption capacity at low pressures, it decreases monotonically with increasing pressure and contributes only a few percent at high pressures. This work also shows that the adsorption heterogeneity of MOF surface is partly attributed to the heterogeneity in electrostatic field in MOFs. Finally, this work demonstrates that MOFs show higher CO2 storage capacity than most zeolites and carbon materials with comparative self-diffusivity in magnitude. Acknowledgment. The financial support of the NSFC (Grant Nos. 20725622, 20706002, 20325621) and the Young Scholars Fund of BUCT (Grant No. QN0604) is greatly appreciated. Supporting Information Available: Details of the crystal structures for the studied MOFs and DFT calculations. This material is available free of charge via the Internet at http:// pubs.acs.org. References and Notes (1) Khoo, H. H.; Tan, R. B. H. EnViron. Sci. Technol. 2006, 40, 4016. (2) Aaron, D.; Tsouris, C. Sep. Sci. Technol. 2005, 40, 321. (3) Gomes, V. G.; Yee, K. W. K. Sep. Purif. Technol. 2002, 28, 161. (4) Himeno, S.; Komatsu, T.; Fujita, S. J. Chem. Eng. Data 2005, 50, 369. (5) He, Y. F.; Seaton, N. A. Langmuir 2006, 22, 1150. (6) Maurin, G.; Llewellyn, P. L.; Bell, R. G. J. Phys. Chem. B 2005, 109, 16084. (7) Cavenati, S.; Gradne, C. A.; Rodrigues, A. E. J. Chem. Eng. Data 2004, 49, 1095. (8) Harlick, P. J. E.; Tezel, F. H. Microporous Mesoporous Mater. 2004, 76, 71. (9) Bourrelly, S.; Llewellyn, P. L.; Serre, C.; Millange, F.; Loiseau, T.; Fe´rey, G. J. Am. Chem. Soc. Soc. 2005, 127, 13519. (10) Snurr, R. Q.; Hupp, J. T.; Nguyen, S. T. AIChE J. 2004, 50, 1090. (11) Eddaoudi, M.; Kim, J.; Rosi, N.; Vodak, D.; Wachter, J.; O’Keeffe, M.; Yaghi, O. M. Science 2002, 295, 469. (12) Du¨ren, T.; Sarkisov, L.; Yaghi, O. M.; Snurr, R. Q. Langmuir 2004, 20, 2683. (13) Zhao, X. B.; Xiao, B.; Fletcher. A. J.; Thomas, K. M.; Bradshaw, D.; Rosseinsky, M. J. Science 2004, 306, 1012. (14) Chen, B. L.; Ockwig, N. W.; Millward, A. R.; Contreras, D. S.; Yaghi, O. M. Angew. Chem., Int. Ed. 2005, 44, 4745. (15) Kubota, Y.; Takata, M.; Matsuda, R.; Kitaura, R.; Kitagawa, S.; Kato, K.; Sakata, M.; Kobayashi, T. C. Angew. Chem., Int. Ed. 2005, 44, 920. (16) Millward, A. R.; Yaghi, O. M. J. Am. Chem. Soc. 2005, 127, 17998. (17) Dybtsev, D. N.; Chun. H.; Yoon, S. H.; Kim, D.; Kim, K. J. Am. Chem. Soc. 2004, 126, 32. (18) Natesakhawat, S.; Culp, J. T.; Matranga, C.; Bockrath, B. J. Phys. Chem. C 2007, 111, 1055. (19) Llewellyn, P. L.; Bourrelly, S.; Serre, C.; Filinchuk, Y.; Fe´rey, G. Angew. Chem., Int. Ed. 2006, 45, 7751. (20) Yang, Q. Y.; Bu, X. P.; Zhong, C. L.; Li, Y. G. AIChE J. 2005, 51, 281.

J. Phys. Chem. C, Vol. 112, No. 5, 2008 1569 (21) Sagara, T.; Klassen, J.; Ganz, E. J. Chem. Phys. 2004, 121, 12543. (22) Yang, Q. Y.; Zhong, C. L. J. Phys. Chem. B 2005, 109, 11862. (23) Garberoglio, G.; Skoulidas, A. I.; Johnson, J. K. J. Phys. Chem. B 2005, 109, 13094. (24) Frost, H.; Du¨ren, T.; Snurr, R. Q. J. Phys. Chem. B 2006, 110, 9565. (25) Hu¨bner, O.; Glo¨ss, A.; Fichtner, M.; Klopper, W. J. Phys. Chem. A 2004, 108, 3019. (26) Jiang, J.; Sandler, S. I. Langmuir 2006, 22, 5702. (27) Pan, L.; Olson, D. H.; Ciemnolonski, L. R.; Heddy, R.; Li, J. Angew. Chem., Int. Ed. 2006, 45, 616. (28) Amirjalayer, S.; Tafipolsky, M.; Schmid, R. Angew. Chem., Int. Ed. 2006, 45, 1. (29) Skoulidas, A. I.; Sholl, D. S. J. Phys. Chem. B 2005, 109, 15760. (30) Yang, Q. Y.; Zhong, C. L. Chem. Phys. Chem. 2006, 7, 1417. (31) Yang, Q. Y.; Zhong, C. L. J. Phys. Chem. B 2006, 110, 17776. (32) Babarao, R.; Hu, Z.; Jiang, J. W.; Chempath, S.; Sandler, S. I. Langmuir 2007, 23, 659. (33) Ramsahye. N. A.; Maurin, G.; Bourrelly, S.; Llewellyn, P.; Loiseau, T.; Fe´rey, G. Phys. Chem. Chem. Phys. 2007, 9, 1059. (34) Chae, H. K.; Siberio-Pe´rez, D. Y.; Kim, J.; Go, Y.; Eddaoudi, M.; atzger, A. J.; O’, Keffe, M.; Yaghi, O. M. Nature 2004, 427, 523. (35) Chui, S. S-Y.; Lo, S. M-F.; Charmant, J. P. H.; Orpen, A. G.; Williams, I. D. Science 1999, 283, 1148. (36) Accelrys, Inc., Materials Studio, 3.0 V; Accelrys, Inc.: San Diego, CA 2003. (37) Beck, D. W. Zeolite molecular sieVes; John Wiley & Sons: New York, 1974. (38) Potoff, J. J.; Siepmann, J. I. AIChE J. 2001, 47, 1676. (39) Goj, A.; Sholl, D. S.; Akten, E. D.; Kohen, D. J. Phys. Chem. B 2002, 106, 8367. (40) Jorgensen, W. L.; Maxwell, D. S.; Tirado-Rives, J. J. Am. Chem. Soc. 1996, 118, 11225. (41) Heinz, H.; Suter, U. W. J. Phys. Chem. B 2004, 108, 18341. (42) Foguet-Albiol, D.; O’Brien, T. A.; Wernsdorfer, W.; Moulton, B.; Zaworotko, M. J.; Abbound, K. A.; Christou, G. Angew. Chem., Int. Ed. 2005, 44, 897. (43) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; et al. GAUSSIAN 03, rev B.1; Gaussian, Inc.: Pittsburgh, PA, 2003. (44) Frenkel, D.; Smit, B. Understanding molecular simulation: from algorithms to applications; Academic Press: San Diego, CA, 2002. (45) Shing, K. S.; Chung, S. T. J. Phys. Chem. 1987, 91, 1674. (46) Bates, S. P.; van Well, W. J. M.; van Santen, R. A.; Smit, B. J. Am. Chem. Soc. 1996, 118, 6753. (47) Myers, A. L.; Monson, P. A. Langmuir 2002, 18, 10261. (48) Do, D. D.; Do, H. D. J. Phys. Chem. B 2006, 110, 17531. (49) Zhou, J.; Wang, W. C. Langmuir 2000, 16, 8063. (50) Rappi, A. K.; Casewit, C. J.; Colwell, K. S.; Goddard, W. A., III; Skid, W. M. J. Am. Chem. Soc. 1992, 114, 10024. (51) Wang, Q. M.; Shen, D. M.; Bu¨low, M.; Lau, M. L.; Deng, S. G.; Fitch, F. R.; Lemcoff, N. O.; Semansin, J. Microporous Mesoporous Mater. 2002, 55, 217. (52) Makrodimitris, K.; Papadopoulos, G. K.; Theodorou, D. N. J. Phys. Chem. B 2001, 105, 777. (53) Skoulidas, A. I.; Sholl, D. S. J. Phys. Chem. B 2001, 105, 3151. (54) Skoulidas, A. I. J. Am. Chem. Soc. 2004, 126, 1356.