Sie sind auf Seite 1von 12

View Article Online / Journal Homepage / Table of Contents for this issue

This article was published as part of the

Downloaded by Indian Institute of Technology Kharagpur on 04/05/2013 11:10:32. Published on 11 February 2009 on http://pubs.rsc.org | doi:10.1039/B803498M

2009 Metalorganic frameworks issue


Reviewing the latest developments across the interdisciplinary area of metalorganic frameworks from an academic and industrial perspective
Guest Editors Jeffrey Long and Omar Yaghi

Please take a look at the issue 5 table of contents to access the other reviews.

TUTORIAL REVIEW

www.rsc.org/csr | Chemical Society Reviews View Article Online

Using molecular simulation to characterise metalorganic frameworks for adsorption applicationsw


ren,*a Youn-Sang Baeb and Randall Q. Snurrb Tina Du
Received 29th September 2008 First published as an Advance Article on the web 11th February 2009 DOI: 10.1039/b803498m Molecular simulation is a powerful tool to predict adsorption and to gain insight into the corresponding molecular level phenomena. In this tutorial review, we provide an overview of how molecular simulation can be used to characterise metalorganic frameworks for adsorption applications. Particular attention is drawn to how these insights can be combined to develop design principles for specic applications. possible linker and corner units and the possibility to postsynthetically functionalise MOFs. Molecular simulation, in which the adsorption performance of a real or hypothetical material is evaluated based on an atomistic model of its structure, can quantitatively predict the uptake of simple gases and, in the case of mixtures, predict the selectivity. Simulation also provides a detailed picture on the molecular scale that is not easily accessible from experimental methods. Therefore, molecular simulation can be a powerful tool to screen existing and hypothetical MOF structures for specic applications, thus narrowing down the search for the most promising structures. In addition, simulation allows us to rapidly study the inuence of molecular level properties such as surface area, pore size, or pore shape on the adsorption behaviour, aiding in the development of design principles. Some of the MOF structures discussed in this review are shown in Fig. 1. In this tutorial review, we provide an overview of how molecular simulation can be used to characterise MOFs for adsorption applications. Furthermore, we describe how these insights can be combined to develop design principles for specic applications. We focus on grand canonical Monte Carlo simulations as this is the molecular simulation method most widely used to predict adsorption equilibria. More

Downloaded by Indian Institute of Technology Kharagpur on 04/05/2013 11:10:32. Published on 11 February 2009 on http://pubs.rsc.org | doi:10.1039/B803498M

1. Introduction
Metalorganic frameworks (MOFs) constitute one of the most exciting recent developments in nanoporous materials, with many potential adsorption applications such as gas storage or separation as described in other articles in this issue. The major advantage of MOFs over more traditional porous materials, such as zeolites or activated carbons, is the greater scope for tailoring these materials for specic applications, which is feasible because of their modular synthesis.1 By choosing appropriate building blocks, solids with cavities of pre-dened shapes and functionalities can be created to provide optimal hostguest interaction. To date, there are tens of thousands of MOFs catalogued in the Cambridge Structural Database (CSD)2 (only a fraction of them is porous and stable upon solvent removal). However, this is only a tiny fraction of imaginable materials because of the large variety of
a

Institute for Materials and Processes, School of Engineering and Electronics, The University of Edinburgh, Kings Buildings, Edinburgh, UK EH9 3JL b Department of Chemical and Biological Engineering, Northwestern University, 2145 Sheridan Road, Evanston, IL 60208, USA w Part of the metalorganic frameworks themed issue.

Tina Duren received her PhD degree from Hamburg University of Technology, Germany, in 2002. She was a postdoctoral researcher with Prof Randall Snurr at Northwestern University, USA, from 20022004 holding a Feodor-Lynen Fellowship of the Alexander von Humboldt Foundation. In 2004, she joined the University of Edinburgh, UK, as a lecturer in Chemical Engineering. Her work focuses on employing ren Tina Du molecular simulation techniques to model the synthesis of porous materials, evaluate their performance and to tailor them for specic applications.
This journal is
 c

Youn-Sang Bae

Youn-Sang Bae was born in 1975 in Seoul, Korea. He obtained his PhD at Yonsei University, Korea, in 2006 under the guidance of Prof. C.-H. Lee, on a study concerning enantiomer separations using a simulated moving bed. Currently, he is working as a postdoctoral fellow in the laboratory of Prof. Randall Snurr at Northwestern University, focusing on gas storage and separations using metalorganic frameworks.

The Royal Society of Chemistry 2009

Chem. Soc. Rev., 2009, 38 , 12371247 | 1237

View Article Online

Downloaded by Indian Institute of Technology Kharagpur on 04/05/2013 11:10:32. Published on 11 February 2009 on http://pubs.rsc.org | doi:10.1039/B803498M

Fig. 1 Structures and linker molecules of some of the MOFs discussed in this paper. (a) IRMOF-10 and its catenated counterpart IRMOF-9. For clarity, the atoms in the networks in IRMOF-9 are shown in dierent shades of grey. (b) Organic linker molecules of the IRMOF family. (c) CuBTC. (d) The closed and the open form of MIL-53(Al).

extensive reviews of molecular modelling of MOFs including dierent simulation techniques such as molecular dynamics or DFT are given by Keskin et al.3 and Snurr et al.4

2. Simulation methods
The input to a simulation consists of models for the adsorbent and the adsorbate as well as force elds that describe the energetic interaction between them. Because MOFs are

Randall Snurr is Professor of Chemical and Biological Engineering at Northwestern University. His research interests include molecular simulation of adsorption, diusion and catalysis in nanoporous materials, particularly for applications in the energy and environmental elds. He is the past chair of the AIChE Computational Molecular Science and Engineering Forum and is currently a Senior Editor of the Journal of Physical Randall Snurr Chemistry. He holds BS and PhD degrees from the University of Pennsylvania and the University of California, Berkeley, respectively, and performed post-doctoral research at the University of Leipzig under an Alexander von Humboldt Fellowship.
1238 | Chem. Soc. Rev., 2009, 38, 12371247

crystalline, the model for the adsorbent is the atomistic representation of the framework taken from its crystallographic coordinates. Here, it is typically assumed that the material is fully activated so any solvent molecules present in the structure are removed. In most simulations, the framework is treated as rigid (a valid assumption for many MOFs), although models for exible MOFs have been proposed recently.5 In order to avoid boundary or nite size eects and to allow simulations that are valid for the extended crystal lattice, periodic boundary conditions are used,6 resulting in simulations that take place in an innite, perfect structure. To model adsorbate/adsorbate and adsorbate/framework interactions, van der Waals interactions (normally modelled by Lennard-Jones potentials) and Coulombic interactions are taken into account. A systematic approach to parameterize adsorbate/adsorbate interaction potentials is to t them to experimental vapourliquid equilibrium data as in the TraPPE force eld.7 These potentials typically include Lennard-Jones parameters and partial charges for all atoms, as well as any necessary expressions for describing bond-stretching, bond-bending, and torsional motions. Depending on their nature, adsorbate molecules are sometimes modelled by a so-called united atom approach where, for example, methane is represented by a single sphere or propane by three spheres. For the framework, the Lennard-Jones parameters for the individual atoms are often taken from generic force elds, particularly DREIDING,8 UFF,9 or OPLS-AA10 which contain parameters for the whole periodic table or a subset thereof. Partial charges for the MOF atoms are typically obtained from quantum chemical calculations on either a
This journal is
 c

The Royal Society of Chemistry 2009

View Article Online

Downloaded by Indian Institute of Technology Kharagpur on 04/05/2013 11:10:32. Published on 11 February 2009 on http://pubs.rsc.org | doi:10.1039/B803498M

periodic model of the framework or on an extended cluster of atoms extracted from the periodic MOF structure. The most widely used simulation method to predict adsorption equilibria is grand canonical Monte Carlo (GCMC) simulation. Here, the chemical potential, the volume, and the temperature are kept xed while the number of molecules is allowed to uctuate. This mimics the experimental equilibrium conditions where the chemical potential and the temperature inside and outside the adsorbent have to be equal. The chemical potential is related to the gas phase pressure by an equation of state, and a whole adsorption isotherm can be simulated by running a series of simulations at increasing pressure. Each Monte Carlo simulation consists of millions of random insertions, deletions, translations, and rotations which are accepted or rejected according to criteria based on a Boltzmann-type weighting. The simulation results are then averaged over the run of the simulation. Detailed descriptions of the simulation methods including sample programs are, for example, given in the book by Frenkel and Smit.6

3. Results from molecular simulation of adsorption in MOFs


3.1 Adsorption isotherms There are many examples of using molecular simulations to predict pure component and mixture adsorption isotherms in the literature. An overview of the work to date is given by Keskin et al.3 and Snurr et al.4 Here, we highlight just two examples. However, before comparing simulation and experimental results it is important to keep in mind that the output of a simulation is the absolute number of molecules present in the framework, whereas experimental results are often reported as the excess amount adsorbed. The excess number of molecules, nex, is the number of molecules that is present in excess to the molecules that would be present in the same (pore) volume at bulk conditions. Therefore, it is related to the absolute number of molecules, nabs, by n
g ex

Fig. 2 Comparison between experimental and simulated adsorption isotherms in IRMOF-1. (a) CH4 at 298 K (open symbols: experiment, closed symbols: simulation). Reproduced with permission from ref. 12. Copyright 2004, American Chemical Society. (b) CO2. Reproduced with permission from ref. 13. Copyright 2008, American Chemical Society.

=n

abs

Vr

g g

(1)

where r is the density of the bulk gas phase calculated with an equation of state and Vg is the pore volume of the MOF framework. Experimentally the pore volume is determined from helium measurements and dierent possibilities exist to mimic this procedure by simulations.11 Fig. 2 shows a comparison between experimental and simulated adsorption isotherms for methane and carbon dioxide in IRMOF-1, one of the most widely studied MOFs.12,13 In both cases the simulation results were achieved using the generic DREIDING8 force eld for the MOF frameworks and the TraPPE7 force eld for the guest molecules without tting a single parameter. The good agreement between simulations and experiments demonstrates that molecular simulation can be a powerful tool to predict the adsorption performance of a given MOF. There are, however, several examples in the literature where simulations overpredict the experimental amount adsorbed considerably.14,15 In general, this discrepancy is not surprising: simulations use the crystal structure as an input thus simulating
This journal is
 c

adsorption in perfect, innite crystals, whereas the quality of the experimental sample depends on the synthesis and activation procedure,16,17 which can result in incomplete removal of solvents, unreacted reactants in the pores, contamination with catenated phases or partial framework collapse. 3.2 Surface areas The surface area is an important characteristic of porous materials, especially for applications in adsorption separations and gas storage. Some MOFs are reported to have very high surface areas, up to 4000 or 5000 m2 g1. However, the surface area cannot be directly observed from experiments, raising the question of whether these surface areas are physically meaningful. In experiments, the surface area is usually obtained by applying the BET theory to a nitrogen isotherm measured at 77 K. This is a standard procedure that allows for comparisons among dierent materials. However, the BET method relies on several assumptions that may break down for microporous materials.1820 Because MOFs are crystalline, the surface area can also be calculated geometrically from the crystal structure. As shown
Chem. Soc. Rev., 2009, 38 , 12371247 | 1239

The Royal Society of Chemistry 2009

View Article Online

Downloaded by Indian Institute of Technology Kharagpur on 04/05/2013 11:10:32. Published on 11 February 2009 on http://pubs.rsc.org | doi:10.1039/B803498M

Fig. 3 (a) Denition of the accessible surface area. (b) Comparison between surface areas obtained from BET analysis from experimental and simulated N2 adsorption isotherms at 77 K and the accessible surface area calculated with a N2-sized probe molecule (Figure adapted from data in ref. 19).

in Fig. 3, the so-called accessible surface area corresponds to the area traced out by the centre of a probe molecule as the probe is rolled across the surface of the framework atoms. In practice, the accessible surface area is calculated from a simple Monte Carlo integration where the probe sphere is randomly inserted around the surface of each framework atom in turn and tested for overlap.18,19 The fraction of probes that do not overlap with other framework atoms is used to calculate the accessible surface area. The probe should be chosen to correspond to the size of the adsorbate of interest. Comparing the experimentally determined BET surface area and the geometrically determined surface area is complicated by the fact that real experimental samples may not correspond to the perfect crystal structure due to incomplete solvent removal, etc. as described above. To avoid this problem, Walton and Snurr18 took the perfect crystal structure and simulated nitrogen isotherms using GCMC. For a series of IRMOFs, they found surprisingly good agreement between the accessible surface areas calculated geometrically and the BET surface areas obtained from the simulated isotherms (see Fig. 3(b)). It should be noted that the good agreement only holds true if the pressure range for the BET analysis is chosen taking two previously published consistency criteria into account: (1) the straight line tted to the BET plot must have a positive intercept, and (2) the pressure range should be chosen so that Vads(1 P/P0) is always increasing with P/P0 where Vads is the amount adsorbed and P/P0 the relative pressure.20 Walton and Snurr also used their simulations to probe the validity of the assumptions underlying the BET theory. The main assumption of the BET model is that adsorption takes place in multilayers as if on a free surface. Fig. 4 shows that in IRMOF-16, adsorption does, indeed, occur by multilayer formation. However, in the smaller pores of IRMOF-1, adsorption appears to be dominated by pore lling.18 Nevertheless, the accessible and BET surface areas agree well even for IRMOF-1 (Fig. 3(b)). The agreement between the BET surface areas and the accessible surface areas provides a compelling argument for using the accessible surface area to characterise crystalline nanoporous materials. Discrepancies between the accessible surface area and the BET surface area obtained from an
1240 | Chem. Soc. Rev., 2009, 38, 12371247

Fig. 4 Top: Snapshots of N2 adsorption at 77 K in IRMOF-16 at four dierent loadings. Bottom: Simulated nitrogen adsorption isotherm at 77 K for IRMOF-16. Reproduced with permission from ref. 18. Copyright 2007, American Chemical Society.

experimental nitrogen isotherm can be used to judge the quality of the experimental sample and may indicate the
This journal is
 c

The Royal Society of Chemistry 2009

View Article Online

Downloaded by Indian Institute of Technology Kharagpur on 04/05/2013 11:10:32. Published on 11 February 2009 on http://pubs.rsc.org | doi:10.1039/B803498M

presence of residual adsorbed solvent molecules, partial collapse of the pores, etc. For example, Fig. 3(b) includes experimental BET surface areas for several MOFs. The agreement with the accessible surface area is good with two exceptions, where the experimental values are much lower. For IRMOF-14, this may indicate that the experimental sample did not correspond with the perfect crystal structure; from the size of the linker molecule (see Fig. 1(b)), the experimental BET surface area should be larger than that of IRMOF-1, for example, in line with the calculated values. In the case of MIL-68, it is known that the pores are partially blocked by free terephthalic acid and DMF molecules.19 These examples demonstrate that the accessible surface area, which can be calculated quickly from the crystal structure,21 could be used as a routine characterisation tool. 3.3 Heats of adsorption

Information about energetic interactions can be obtained experimentally from microcalorimetry experiments which measure the dierential enthalpy of adsorption directly. This quantity has contributions from the adsorbate/MOF interactions and the adsorbate/adsorbate interactions. For energetically homogeneous adsorbents, the adsorbate/MOF interactions remain constant with loading, but most MOFs are energetically heterogeneous, resulting from cavities of dierent sizes, the dierent chemical nature of the framework or strong interaction sites in the corners of the framework. For these materials, the adsorbate/MOF interactions (which are negative) gradually decrease in magnitude with loading as the strongest adsorption sites are occupied rst.20 With increasing loading, attractive adsorbate/adsorbate interactions also contribute to the dierential enthalpy of adsorption. The dierential enthalpy of adsorption, Dadsh, can be directly calculated from a GCMC simulation:22 Dads h hres;bulk RT1 zbulk hvads Nads i hvads ihNads i hNads i hNads ihNads i
2

where xi is the mole fraction of species i in the adsorbed phase and yi is the mole fraction in the bulk phase. Therefore if the selectivity is larger than one, species 1 is adsorbed more strongly, and the higher the selectivity the more ecient the separation by adsorption. Liu et al. observed that the selectivity in the catenated IRMOFs-9, -11 and -13 is higher than in their non-catenated counterparts IRMOFs-10, -12 and -14 as illustrated in Fig. 5(a). Furthermore, Fig. 5(a) shows that whereas the selectivities for the non-catenated materials are nearly independent of the pressure, the selectivities for the catenated IRMOFs steeply decrease at rst and then remain constant. A very similar trend is observed for the dierences between CH4 and H2 in the portion of the isosteric heat of adsorption contributed by the adsorbate/framework interaction (Fig. 5(b)), providing an explanation for the trends observed for the selectivity. Catenation leads to various pores of dierent sizes which are all smaller than in their non-catenated counterparts, resulting in stronger adsorbate/ framework interactions. The smallest pores and therefore the

Here hres,bulk is the residual enthalpy of the bulk phase and zbulk is the compressibility factor of the bulk phase, both of which can be calculated from an equation of state, Nads is the number of adsorbed molecules and vads is their potential energy. If the gas phase can be treated as ideal, the residual enthalpy is equal to zero and the compressibility factor is equal to one, so that eqn (2) reduces to the last term. The related isosteric heat of adsorption can be derived from Dadsh by subtracting RT (note that the isosteric heat is positive and Dadsh is negative). An example showing how the isosteric heat of adsorption can be used to explain the adsorption performance of dierent MOFs is given by Liu et al.,23 who studied the separation of hydrogen and methane mixtures and compared the selectivity in several catenated and non-catenated IRMOF materials (catenated materials have two separated frameworks self-assembled within each other as illustrated in Fig. 1(a)). The selectivity, S12, is dened as follows: S12 x1 =x2 y1 =y2 3

Fig. 5 (a) Selectivity of CH4 from an equimolar binary mixture of H2 and CH4 at 298 K as a function of pressure. (b) Dierences in the isosteric heat of adsorption between CH4 and H2 as a function of pressure. Note that only the contribution to the isosteric heat from the adsorbate/framework interactions is plotted. Reproduced with permission from ref. 23. Copyright 2008, American Chemical Society.

This journal is

 c

The Royal Society of Chemistry 2009

Chem. Soc. Rev., 2009, 38 , 12371247 | 1241

View Article Online

Downloaded by Indian Institute of Technology Kharagpur on 04/05/2013 11:10:32. Published on 11 February 2009 on http://pubs.rsc.org | doi:10.1039/B803498M

strongest interaction sites are occupied rst, causing a strong decrease in the isosteric heat of adsorption with loading for both methane and hydrogen. As the larger decrease happens for methane (the bigger molecule), a strong decrease is also observed in the dierence plotted in Fig. 5. In contrast to this, the structures of the non-catenated IRMOFs, which consist of large cubic pores, are much simpler and the pressure is not yet high enough for packing eects to play a role, which might also lead to a variation in the selectivity. The dierential enthalpy of adsorption and therefore the strength of the interaction between adsorbate molecules and the MOF framework depend on the pore size and shape. Jumps in dierential enthalpy curves can therefore be an indicator for changes of the framework structure triggered by the uptake of adsorbate molecules. By performing GCMC simulations of CO2 in the open and closed forms of MIL-53(Al) (Fig. 1(d)) and combining the results, Ramsahye et al.24 showed that the experimentally measured steps in the adsorption isotherm and in the dierential enthalpy data can be explained by a breathing eect of the framework. This was later conrmed experimentally by in situ diraction experiments performed at synchrotron radiation facilities.25 3.4 Molecular level insights into adsorption mechanismsenergetic interactions and sitings In addition to predicting macroscopic phenomena such as adsorption isotherms and heats of adsorption, GCMC simulations can also provide a detailed picture on the molecular scale, as the positions and potential energies of all adsorbate molecules are known over the duration of the simulation. This information can be further analysed to get information about energetic interactions, preferential siting, and ultimately the adsorption mechanism. Potential energy maps can be easily determined by placing an adsorbate molecule at positions throughout the MOF framework and calculating the adsorbate/framework interaction, i.e. without conducting a GCMC simulation. These potential maps illustrate the regions in the framework with the lowest interaction energy and therefore the preferential adsorption sites (for a threedimensional example, see the work by Vishnyakov et al.26). Whereas the potential energy maps show the distribution of potential energies in the framework, they do not contain information on how these sites are lled up with increasing loading. This information is obtained by determining energy histograms during a GCMC simulation. Especially if determined for several pressures, these distributions contain valuable information on how quickly specic sites saturate and the relative importance of individual interaction sites. Fig. 6 shows the fraction of methane molecules that experienced certain methane/framework interaction energies for three IRMOF frameworks at 0.1 and 50 bar. Large negative values indicate strong interaction and more favourable adsorption sites. At 0.1 bar, the two non-catenated structures IRMOF-1 and IRMOF-10 show a distinct plateau between 11 and 10 kJ mol1, which corresponds to the corner regions, as well as a peak around 7 kJ mol1 corresponding to the less favourable linker sites. In the case of IRMOF-10, which has larger cavities than IRMOF-1 (compare Fig. 1(b)), molecules
1242 | Chem. Soc. Rev., 2009, 38, 12371247

Fig. 6 Distribution of methane/framework potential energies for individual methane molecules in dierent IRMOFs at 298 K ( IRMOF-1, n IRMOF-9, IRMOF-10) (a) at 0.1 bar, (b) at 50 bar. The lines were added to guide the eye. Reproduced with permission from ref. 27. Copyright 2007, Elsevier.

can also be found in the centre of the cavities where the energy is close to zero. The energies that a methane molecule experiences in the catenated IRMOF-9 framework are much more favourable resulting from the smaller cavities due to the catenation (Fig. 6(a)). An increase of the pressure and therefore the loading leads to a shift in the relative importance of adsorption sites as the size of the preferential adsorption sites is limited. Fig. 6(b) shows that whereas the positions of the peaks remain the same, energetically less favourable sites such as the centre of the cavity for IRMOF-10 become more important, as the most favourable adsorption site (around 20 kJ mol1) can only accommodate a small fraction of the methane molecules present in the framework.27 Molecular simulation is an ideal tool to investigate what-if scenarios that cannot be investigated experimentally. For example, switching o the electrostatic interaction in a simulation allows investigating in detail what inuence it has on the uptake of a gas or on the separation of
This journal is
 c

The Royal Society of Chemistry 2009

View Article Online

Downloaded by Indian Institute of Technology Kharagpur on 04/05/2013 11:10:32. Published on 11 February 2009 on http://pubs.rsc.org | doi:10.1039/B803498M

Fig. 7 Eect of the electrostatic interactions on the CO2/CH4 selectivity from equimolar mixtures of CO2 and CH4 in CuBTC at 298 K. Case 1: all electrostatic interactions switched o, Case 2: only CO2/CO2 electrostatic interactions considered, Case 3: CO2/CO2 and CO2/framework electrostatic interactions considered. Reproduced with permission from ref. 28. Copyright 2006, Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim.

mixtures. Yang and Zhong performed a computational study of binary and ternary mixture adsorption of CO2, CH4 and C2H6 in CuBTC.28 To study the underlying mechanisms of CO2/CH4 separation in CuBTC, the authors performed three GCMC simulations: (1) all electrostatic interactions are switched o, (2) only the electrostatic interactions between CO2 molecules and the CuBTC framework are switched o thus taking into account the CO2/CO2 electrostatic interactions, and (3) all electrostatic interactions are considered (Fig. 7). Their ndings show that electrostatic interactions resulting from the atomic partial charges of the MOF framework can greatly enhance the separation of mixtures of components with largely diering polarities. Karra and Walton have recently performed atomistic GCMC simulations to assess the role of open metal sites in CuBTC on the separation of carbon monoxide from binary mixtures containing CH4, N2 or H2.29 The results showed that electrostatic interactions between the CO dipole and the partial charges on the MOF atoms dominate the adsorption mechanism. The simulations predicted that CuBTC is quite selective for carbon monoxide over hydrogen and nitrogen for three dierent mixture compositions at 298 K. On the basis of the pure-component and low-concentration behaviour of CO, the results indicate that MOFs with open metal sites have the potential for enhancing adsorption separations of molecules of diering polarities, but the pore size relative to the adsorbate size also plays a signicant role. A recent experimental study by Bae et al. showed that the open metal sites in a carboranebased MOF can help the separation of (quadru)polar/ nonpolar pairs such as CO2/CH4.30 Further information about the adsorption mechanism comes from analysing the positions of adsorbate molecules in MOF frameworks during a GCMC simulation with snapshots or density distributions. Snapshots represent a single conguration during the simulation and visualise, for example, how full the pores are and where molecules are located as illustrated in Fig. 4. Density distributions are
This journal is
 c

Fig. 8 Contour plot of the centre of mass probability density of benzene in IRMOF-1 in a plane close to the corners of the framework. Reproduced with permission from ref. 31. Copyright 2007, Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim.

calculated by averaging the positions of the adsorbate molecules over the whole simulation run and provide complementary information about the sitings of adsorbate molecules. Fig. 8 shows the density distribution of benzene in IRMOF-1 at 303 K.31 In this gure, A corresponds to the larger cavities where the linker molecules are pointing out and B to the smaller cavities where the linker molecules are pointing in. As the green/blue areas indicate, benzene molecules are located preferentially in the corners of the A cavities, and hardly any are located in the centres of the B cavities. The solid arrow in the contour plot indicates that diusion occurs as molecules hop from one A cavity to the next A cavity without thermalising in the smaller B cavities. The dashed arrow shows intracavity hopping, with a lower energy of activation. Density distributions also can be directly compared to experimental results from X-ray diraction studies as demonstrated for example by Dubbeldam et al. for argon in IRMOF-1.15 To investigate which experimentally determined adsorption sites are present at innite dilution and which are created by loading eects, they performed energy minimisations of a single argon molecule in the framework and identied three adsorption sites. This implied that ve additional sites observed in GCMC and by XRD are caused by loading eects. By studying the sitings at dierent loadings, they managed to elucidate in which order the dierent adsorption sites are lled. With one exception, the adsorption sites determined by simulation agreed very well with the experimentally determined sites.

4. Using simulation to develop design principles


Qualitative trends and quantitative correlations between the properties of MOFs and their adsorption performance can be very helpful for screening and tailoring existing and
Chem. Soc. Rev., 2009, 38 , 12371247 | 1243

The Royal Society of Chemistry 2009

View Article Online

hypothetical MOFs. Combining information about the properties of MOFs such as types of functional groups, surface area, free volume and heat of adsorption, with detailed analysis of the adsorption mechanism provides the opportunity to develop design principles for MOFs to tailor them for specic applications. These eorts are still in their infancy. In the following sections, we present three examples: the storage of methane and hydrogen, and adsorption of CO2. 4.1 Methane storage

Downloaded by Indian Institute of Technology Kharagpur on 04/05/2013 11:10:32. Published on 11 February 2009 on http://pubs.rsc.org | doi:10.1039/B803498M

Natural gas, which consists mainly of methane, is a promising alternative fuel to supplement or replace gasoline and diesel fuel in vehicular applications. To encourage the vehicular application of methane, the U.S. Department of Energy (DOE) has dened a storage target of 180 v/v (v/v: the volume of gas adsorbed at standard temperature and pressure per volume of the storage vessel) at 35 bar. Du ren et al. investigated the adsorption characteristics of methane in

several IRMOFs, as well as molecular squares, zeolites, MCM-41, and carbon nanotubes.12 As shown in Fig. 9, they found a correlation between the adsorption of methane at 35 bar and 298 K with the surface area, showing a band of points. Du ren et al. suggested that the ideal adsorbent for methane storage should have a large surface area, high free volume, low framework density, and strong adsorbent/ methane interactions. Taking these ndings into account, they proposed new, not yet synthesized, IRMOF structures having 1,4-tetrabromobenzenedicarboxylate and 9,10-anthracenedicarboxylate as linker molecules. These structures were predicted to show substantially higher uptake of methane than IRMOF-6, which showed the best performance among all tested materials at that time. Building on this work, Wang performed a simulation study to investigate the adsorption of methane in MOFs with dierent topologies, including ve IRMOFs (IRMOF-1, 6, 8, 10 and 14), CuBTC, two CPL MOFs (CPL-28 and CPL-522), and Cu(AF6)(bpy)2 (A = Si and Ge).32 Similar to the results of Du ren et al., this study showed that the surface area is more important than other factors for methane adsorption at room temperature and moderate pressure. Wang obtained a linear correlation of the heat of adsorption with the pore sizes for IRMOFs and argued that the heat of adsorption plays a main role in methane adsorption at low loading. He also found that the free volume and surface area correlate well with the amount adsorbed at high pressures conrming earlier ndings for hydrogen adsorption.33 4.2 Hydrogen storage Hydrogen storage is also important for potential automotive applications, and MOFs have been suggested as promising hydrogen adsorbents. Many researchers have suggested that increasing the surface area of MOFs will lead to improved hydrogen storage. There have, however, been disagreements in the literature about whether hydrogen uptake is correlated with the surface area or not. To answer this question, Frost et al. performed molecular simulations of hydrogen adsorption in MOFs at 77 K over a wide range of pressures to see the eects of surface area, free volume, and heat of adsorption on hydrogen uptake.33 They selected a series of ten IRMOFs, which all have the same framework topology and surface chemistry but varying pore sizes (compare Fig. 1(b)). Investigating the absolute amount adsorbed at 77 K, the authors found three adsorption regimes where the isosteric heat, surface area, and free volume each dominates the amount of hydrogen adsorbed. At low loading and therefore low pressure, the hydrogen uptake correlates with the heat of adsorption; at intermediate loading, the uptake correlates with the surface area; and at the highest loadings, the uptake correlates with the free volume of the MOF as illustrated in Fig. 10. Later, Frost and Snurr tested these correlations for H2 adsorption at room temperature and analyzed both absolute and excess adsorption.34 The results showed that the correlations that had been developed for absolute H2 adsorption at 77 K do not hold well for absolute H2 adsorption at 298 K due to very weak adsorption. Even at low loadings, the absolute adsorbed amount of H2 at 298 K mainly correlates with the
This journal is
 c

Fig. 9 Adsorbed amounts of methane at 35 bar and 298 K as a function of the accessible surface area (a) per volume and (b) per mass (, IRMOFs; J, molecular squares; n, carbon nanotubes; *, zeolites and MCM-41). Reproduced with permission from ref. 12. Copyright 2004, American Chemical Society.

1244 | Chem. Soc. Rev., 2009, 38, 12371247

The Royal Society of Chemistry 2009

View Article Online

Downloaded by Indian Institute of Technology Kharagpur on 04/05/2013 11:10:32. Published on 11 February 2009 on http://pubs.rsc.org | doi:10.1039/B803498M

Fig. 10 Simulated hydrogen adsorption in a series of 10 IRMOFs at 77 K. (a) Adsorbed amounts at 0.1 bar vs. isosteric heat of adsorption. (b) Adsorbed amounts at 30 bar vs. accessible surface area. (c) Adsorbed amounts at 120 bar vs. free volume. E, IRMOF-1; , IRMOF-4; m, IRMOF-6; , IRMOF-7; *, IRMOF-8; K, IRMOF-10; +, IRMOF-12; n, IRMOF-14; B, IRMOF-16; &, IRMOF-18. Reproduced with permission from ref. 33. Copyright 2006 the American Chemical Society.

free volume. The authors also tested the correlations for excess adsorption of H2 at 298 K and found that at low loading the excess amount adsorbed correlates well with the heat of adsorption again, but at high loadings, the adsorption correlates better with the surface area than with the free volume. Explanations for these nding are given in ref. 34. The interaction between hydrogen and MOF frameworks to date is very weak. In order to investigate how much the heat of adsorption must be increased to meet current targets set by the US Department of Energy (DOE) for hydrogen storage at room temperature, Frost and Snurr articially increased the hydrogen/MOF Lennard-Jones interaction parameters.34 From their results, they found a correlation between the heat of adsorption and the density of hydrogen in the pore void volume. Using this correlation, they developed a graph showing the required heats of adsorption as a function of the free volume to meet specied gravimetric and volumetric storage targets at room temperature and 120 bar. Fig. 11 provides design targets for obtaining desired gravimetric or volumetric H2 loadings within MOFs (and perhaps other similar microporous materials). For example, a reasonable goal of 9 wt% and 30 g L1 could be obtained with MOFs if a new material can provide an isosteric heat of 15 kJ mol1 or higher while maintaining a free volume of 2.5 cm3 g1 and void fraction of 85%. Fig. 11 clearly illustrates the tradeo between free volume (or void fraction) and heat of adsorption.

By comparing the catenated IRMOF-9 and its noncatenated counterpart IRMOF-10, Frost and Snurr suggested that catenation is not a promising option to increase the H2 uptake. For this pair of MOFs, the increase in heat of adsorption gained through catenation does not oset the loss in free volume, so hydrogen uptake is larger in the non-catenated MOF. The authors recommended alternative methods of increasing the heat of adsorption, such as open-metal sites and charged frameworks. It is clear that the interaction energy between H2 and the adsorbent must be increased, while maintaining sucient free volume in order for hydrogen storage by physisorption to become viable.34 4.3 CO2 adsorption The capture of CO2 from the ue exhaust of power plants may play a signicant role in mitigating the eect of CO2 on climate change in the future. Adsorption in porous materials is known to be energy ecient and is a promising technology for post-combustion capture of CO2 from ue gases. Recently, systematic computational studies have been performed by several research groups to investigate the eects of various factors on CO2 adsorption in MOFs.35,36 These studies have provided useful information for developing new MOF materials for CO2 capture. Yang et al. investigated the eects of organic linker, pore size, pore topology and electrostatic elds on the adsorption properties of CO2 in MOFs.35 They selected

Fig. 11 (a) Requirements for target gravimetric H2 loadings at 120 bar and 298 K. (b) Requirements for target volumetric H2 loadings at 120 bar and 298 K. Results for existing materials are marked with symbols: n, IRMOF-1; B, IRMOF-10; J, IRMOF-14; +, IRMOF-16; &, CuBTC. Figure adapted from ref. 34. Note that the volumetric storage units were reported incorrectly in the version of this graph in ref. 34.

This journal is

 c

The Royal Society of Chemistry 2009

Chem. Soc. Rev., 2009, 38 , 12371247 | 1245

View Article Online

nine MOFs with dierent pore sizes and topologies, including six IRMOFs (IRMOFs-1, -8, -10, -11, -14 and -16), a Mn-MOF, MOF-177 and CuBTC. From their simulation results, they argued that the pore size of a MOF material plays an important role in CO2 capacity. MOFs having pore sizes between 1.0 and 2.0 nm (IRMOF-1, IRMOF-8, IRMOF-10, IRMOF-14 and MOF-177) showed large CO2 capacity at room temperature and around 30 bar (note that the pressure of ue gas is much lower). For MOFs in this pore size range, the CO2 capacity increased with accessible surface area and free volume. Moreover, they investigated the eects of electrostatic interactions on the CO2 capacity in three MOFs by performing GCMC simulations with and without electrostatic interactions between the CO2 molecules and the framework atoms. The results showed that, although the electrostatic interactions between CO2 and the MOF framework can contribute as much as 30% to the total adsorption capacity at low pressures (at room temperature), it decreases monotonically with increasing pressure and contributes only a few percent at high pressures. From these results, they argued that the strength of the interactions between CO2 and MOFs plays a predominant role at low pressure, while the eects of the surface area and free volume become evident only at relatively high pressures. Finally, this work suggests that MOFs show larger CO2 capacity than most zeolites and carbon materials at high pressures. In a related study, Babarao and Jiang used simulations to investigate the eects of the metal oxide, organic linker, functional group and framework topology on CO2 in MOFs.36 The authors chose a series of IRMOFs (IRMOF-1, Mg-IRMOF-1, Be-IRMOF-1, IRMOF-1-(NH2)4, IRMOF-10, IRMOF-13 and IRMOF-14) as well as UMCM-1, a uorous MOF (F-MOF-1), and a covalent-organic framework (COF-102) as illustrated in Fig. 12. Their results showed that the adsorption capacity in IRMOFs can be changed by varying the functional group, metal oxide, organic linker and framework topology. By adding a functional group into IRMOF-1

to form IRMOF-1-(NH2)4, a higher heat of adsorption was obtained at low pressures due to the increased anity of the framework for CO2. Also, compared to the non-catenated IRMOF-14, the catenated IRMOF-13 showed enhanced heat of adsorption at low pressures because of the presence of smaller pores. However, their results also showed that at high pressures catenation and extra functional groups can decrease adsorption due to the decreased free volume and surface area. In addition, Babarao and Jiang created the hypothetical structures Mg-IRMOF-1 and Be-IRMOF-1 by replacing the metal oxide Zn4O in IRMOF-1 with Mg4O or Be4O.36 They found that the isosteric heats at innite dilution of Mg-IRMOF-1 and Be-IRMOF-1 are almost identical to those of IRMOF-1. This indicates that the framework/CO2 interaction at low pressures is not strongly aected by varying the metal oxide corner in IRMOF-1. Their results also showed that Mg-IRMOF-1 and Be-IRMOF-1 exhibit larger gravimetric CO2 adsorption especially at moderate and high pressures compared to IRMOF-1 due to the presence of the lighter metals Be and Mg, demonstrating the importance of the low framework density for gravimetric uptake. They also commented that the reduction in framework density leads to the increase of the free volume and surface area per unit mass, despite identical porosity. Another important insight from this work is that the organic linker plays a critical role in tuning the free volume and accessible surface area and largely determines CO2 adsorption at high pressures. They found that the longer and bigger organic linkers in IRMOF-10 and IRMOF-14 compared to IRMOF-1 (see Fig. 1(b)) signicantly increase the free volume and accessible surface area. A similar increase in the surface area and free volume is observed in UMCM-1 using two dierent organic linkers. Considering both gravimetric and volumetric capacities, IRMOF-10, IRMOF-14 and UMCM-1 turned out to be the best for CO2 adsorption at room temperature and 50 bar among the MOFs studied in this work.36

Downloaded by Indian Institute of Technology Kharagpur on 04/05/2013 11:10:32. Published on 11 February 2009 on http://pubs.rsc.org | doi:10.1039/B803498M

Fig. 12 Schematic tailoring of the metal oxide and organic linker in IRMOF-1. Color code: Zn, green; Mg, cyan; Be, purple; O, red; N, blue; C, gray; H, white. Reproduced with permission from ref. 36. Copyright 2008, American Chemical Society.

1246 | Chem. Soc. Rev., 2009, 38, 12371247

This journal is

 c

The Royal Society of Chemistry 2009

View Article Online

Downloaded by Indian Institute of Technology Kharagpur on 04/05/2013 11:10:32. Published on 11 February 2009 on http://pubs.rsc.org | doi:10.1039/B803498M

Fig. 13 Gravimetric and volumetric CO2 capacities of all MOFs studied by Babarao and Jiang at 30 bar as a function of (a) the framework density, (b) the free volume, (c) the porosity and (d) the accessible surface area. Solid circles and curves: gravimetric capacity. Open circles and dashed curves: volumetric capacity. Reproduced with permission from ref. 36. Copyright 2008, American Chemical Society.

Both gravimetric and volumetric capacities at 30 bar correlate well with the framework density, free volume, porosity and accessible area as shown in Fig. 13. These results indicate that both gravimetric and volumetric capacities can be enhanced by decreasing the framework density or increasing the free volume, porosity or accessible surface area. However, the authors also mentioned that not all of these options are practically feasible and there is a complex interplay between these properties.

5. Conclusions
Molecular simulation is playing an increasingly important role in increasing our understanding of the fundamentals of adsorption in MOFs and is contributing to the development of MOFs toward practical applications. For rigid well-characterized MOFs and relatively simple adsorbate molecules, GCMC simulations predict adsorption isotherms, heats of adsorption, and siting of adsorbate molecules that are generally in good agreement with experiment. Current developments in the eld suggest that these capabilities will be extended to exible MOFs and more complex adsorbate molecules in the near future. The work done to date underlines the usefulness of molecular simulations as a screening tool to nd new candidates for gas storage and separations and to guide the design of new MOFs. From structurefunction correlations, the gas adsorption can be estimated without time-consuming simulations or expensive experiments and, more importantly, unknown materials can be rapidly screened or designed for gas storage and separations.

Acknowledgements
We gratefully acknowledge the U.S. Dept. of Energys Oce of Science (Grant No. DE-FG02-01ER15244) and the European Commission (DeSANNS SES6-CT-2005-020133).

References
1 U. Mueller, M. Schubert, F. Teich, H. Puetter, K. Schierle-Arndt , J. Mater. Chem., 2006, 16, 626. and J. Pastre 2 J. L. C. Rowsell and O. M. Yaghi, Microporous Mesoporous Mater., 2004, 73, 3. 3 S. Keskin, J. Liu, R. B. Rankin, J. K. Johnson and D. S. Sholl, Ind. Eng. Chem. Res., 2008, DOI: 10.1021/ie800666s. . Yazaydin, D. Dubbeldam and H. Frost, in 4 R. Q. Snurr, A. O MetalOrganic Frameworks: Design and Application, ed. L. R. MacGillivray, Wiley-VCH, in press. 5 F. Salles, A. Ghou, G. Maurin, R. G. Bell, C. Mellot-Draznieks rey, Angew. Chem., Int. Ed., 2008, 47, 8487, and and G. Fe references therein.

6 D. Frenkel and B. Smit, Understanding of Molecular Simulation: from Algorithms to Applications, Academic Press, San Diego, 2002. 7 M. G. Martin and J. I. Siepmann, J. Phys. Chem. B, 1998, 102, 2569. 8 S. L. Mayo, B. D. Olafson and W. A. Goddard, J. Phys. Chem., 1990, 94, 8897. 9 A. K. Rappe, C. J. Casewit, K. S. Colwell, W. A. Goddard and W. M. Ski, J. Am. Chem. Soc., 1992, 114, 10024. 10 W. L. Jorgensen, D. S. Maxwell and J. Tirado-Rives, J. Am. Chem. Soc., 1996, 118, 11225. 11 A. L. Myers and P. A. Monson, Langmuir, 2002, 18, 10261. 12 T. Du ren, L. Sarkisov, O. M. Yaghi and R. Q. Snurr, Langmuir, 2004, 20, 2683. 13 K. S. Walton, A. R. Millward, D. Dubbeldam, H. Frost, J. J. Low, O. M. Yaghi and R. Q. Snurr, J. Am. Chem. Soc., 2008, 130, 406. 14 G. Garberoglio, A. I. Skoulidas and J. K. Johnson, J. Phys. Chem. B, 2005, 109, 13094. 15 D. Dubbeldam, H. Frost, K. S. Walton and R. Q. Snurr, Fluid Phase Equilib., 2007, 261, 152. 16 J. Liu, J. T. Culp, S. Natesakhawat, B. C. Bockrath, B. Zande, S. G. Sankar, G. Garberoglio and J. K. Johnson, J. Phys. Chem. C, 2007, 111, 9305. 17 J. Hazovic, M. Bjrgen, U. Olsbye, P. D. C. Dietzel, S. Bordiga, C. Prestipino, C. Lamberti and K. P. Lillerud, J. Am. Chem. Soc., 2007, 129, 3612. 18 K. S. Walton and R. Q. Snurr, J. Am. Chem. Soc., 2007, 129, 8552. rey, K. S. Walton and R. Q. Snurr, 19 T. Du ren, F. Millange, G. Fe J. Phys. Chem. C, 2007, 111, 15350. 20 F. Rouquerol, J. Rouquerol and K. Sing, Adsorption by Powders and Porous Solids, Academic Press, San Diego, 1999. 21 Web site for obtaining the source code of a program to calculate the accessible surface area: http://www.see.ed.ac.uk/Btduren/ research/surface_area/. 22 T. Vuong and P. A. Monson, Langmuir, 1996, 12, 5425. 23 B. Liu, Q. Yang, C. Xue, C. Zhong, B. Chen and B. Smit, J. Phys. Chem. C, 2008, 112, 9854. 24 N. A. Ramsahye, G. Maurin, S. Bourrelly, P. L. Llewellyn, rey, Chem. Commun., 2007, 3261. T. Loiseau, C. Serre and G. Fe 25 C. Serre, S. Bourrelly, A. Vimont, N. A. Ramsahye, G. Maurin, P. L. Llewellyn, M. Daturi, Y. Filinchuk, O. Leynaud, P. Barnes rey, Adv. Mater., 2007, 19, 2246. and G. Fe 26 A. Vishnyakov, P. I. Ravikovitch, A. V. Neimark, M. Bulow and Q. M. Wang, Nano Lett., 2003, 3, 713. 27 T. Du ren and R. Q. Snurr, Studies in Surface Science and Catalysis: 7th International Conference on the Characterisation of Porous Solids, Aix en Provence, 2005, p. 161. 28 Q. Y. Yang and C. L. Zhong, ChemPhysChem, 2006, 7, 1417. 29 J. R. Karra and K. S. Walton, Langmuir, 2008, 24, 8620. 30 Y.-S. Bae, O. K. Farha, A. M. Spokoyny, C. A. Mirkin, J. T. Hupp and R. Q. Snurr, Chem. Commun., 2008, 4135. 31 S. Amirjalayer, M. Tapolsky and R. Schmid, Angew. Chem., Int. Ed., 2007, 46, 463. 32 S. Y. Wang, Energy Fuels, 2007, 21, 953. 33 H. Frost, T. Du ren and R. Q. Snurr, J. Phys. Chem. B, 2006, 110, 9565. 34 H. Frost and R. Q. Snurr, J. Phys. Chem. C, 2007, 111, 18794. 35 Q. Y. Yang, C. L. Zhong and J. F. Chen, J. Phys. Chem. C, 2008, 112, 1562. 36 R. Babarao and J. W. Jiang, Langmuir, 2008, 24, 6270.

This journal is

 c

The Royal Society of Chemistry 2009

Chem. Soc. Rev., 2009, 38 , 12371247 | 1247

Das könnte Ihnen auch gefallen