Article pubs.acs.org/JPCC
Effects of Polarizability on the Adsorption of Noble Gases at Low Pressures in Monohalogenated Isoreticular Metal−Organic Frameworks Scott T. Meek,†,§ Stephanie L. Teich-McGoldrick,‡,§ John J. Perry IV† Jeffery A. Greathouse,‡ and Mark D. Allendorf*,† †
Sandia National Laboratories, Livermore, California 94551, United States Sandia National Laboratories, Albuquerque, New Mexico 87123, United States
‡
S Supporting Information *
ABSTRACT: A systematic investigation of the effects of linker polarizability on the adsorption properties of weakly interacting gases (N2, Ar, Kr, and Xe) is reported. Experimental and simulated adsorption properties were measured for a complete isoreticular series of monohalogenated metal−organic frameworks (MOFs). Variations on IRMOF-2, in which one linker hydrogen is replaced with −F, −Cl, −Br, or −I, comprise the series. Both experimental and simulated results indicate that increasing linker polarizability correlates with increased gas uptake. Evidence of increased adsorbate interaction with increased linker polarizability is also observed in the Kr/N2, Xe/N2, and Xe/ Kr selectivity data and in isosteric heats of adsorption. Unexpectedly, comparison between simulated and experimental isotherms reveals that the agreement between the two improves with the size of the adsorbate, with essentially identical results for Xe. This is apparently due to the creation of regions inaccessible to any of the noble gases as a result of halogen functionalization. Simulated adsorption isotherms are also reported for radon, which is difficult to measure experimentally due to its radioactivity.
■
microporous materials by their synthetic flexibility. Through modification of the organic linker or metal center, it is possible to tune not only the dimensions of the pore, but also its chemical properties, enabling an unprecedented ability to tailor them to fit the desired application. As a result, MOFs are promising for numerous applications10 including gas storage11 and separation,12 drug delivery,13 catalysis,14 electronics,15 and sensing.16 Thus far, there are only a few reports concerning the interactions of MOFs with noble gases. Xe adsorption isotherms measured at 292 K for IRMOF-1, -2, -3, and -6 were reported as part of an investigation using 129Xe NMR spectroscopy to probe the pore environment.17 This study shows that Xe binds to IRMOFs primarily near the carboxylate groups and tetrahedral Zn4O clusters in the corners of the cages. Additionally, the reported adsorption isotherms at 1 bar and 292 K demonstrate an uptake of 1.5−2.0 Xe molecules per cage for these materials. Farrusseng and co-workers measured the Kr and Xe heats of adsorption for IRMOF-1, IRMOF-3, and HKUST-1 using the TAP-2 method18 and also calculated heats of adsorption for these MOFs using Monte Carlo simulations, finding excellent agreement between measured and
INTRODUCTION Noble gases, which exhibit low chemical reactivity due to their full outer shell of valence electrons, pose a unique challenge in the field of gas separation. These gases serve many useful roles in industry, including cryogenics,1 carrier gases,2 anesthetics,3 insulation,4 lighting,5 and lasers.6 Rn can accumulate in buildings and is thought to be the leading cause of lung cancer in the United States among nonsmokers.7 Additionally, radioactive Xe and Kr isotopes are emitted as byproducts of nuclear fission and treatment of spent nuclear fuel.8 Thus, methods for separating noble gases from air and from one another are highly important for industrial purification, detection of radioisotopes, and waste gas treatment. The most common industrial method is cryogenic distillation, but this process is energy intensive. Separation via selective adsorption at noncryogenic temperatures onto microporous materials such as zeolites and activated carbons has also been used to accomplish this goal.9 However, these have limited usage due to the difficulty in tuning their adsorptive properties. New, microporous materials, designed for specific adsorbates, are therefore highly desirable. Metal−organic frameworks (MOFs), hybrid lattices of organic electron donors and metal cations, have enjoyed rapidly growing attention over the past decade. These materials can exhibit extremely high apparent surface areas, low density, and high thermal stability but are set apart from other © 2012 American Chemical Society
Received: April 5, 2012 Revised: August 15, 2012 Published: August 16, 2012 19765
dx.doi.org/10.1021/jp303274m | J. Phys. Chem. C 2012, 116, 19765−19772
The Journal of Physical Chemistry C
Article
predicted values. Recently, Thallapally et al. obtained Xe and Kr isotherms for MOF-74(Ni), IRMOF-1, and an activated carbon typically used in industrial Xe capture at room temperature. MOF-74-Ni (also known as CPO-27-Ni and Ni/DOBDC)19 was shown to have greater Xe uptake and higher Xe/Kr selectivity than either IRMOF-1 or the activated carbon.20 Computationally, the interactions between IRMOF-1 and Ar, Xe, and Kr were explored with Grand Canonical Monte Carlo (GCMC) simulations.21 These simulated isotherms indicate that IRMOF-1 should be effective for separating Xe from Ar or Kr mixtures. Additionally, GCMC was recently used as a screening tool to identify promising MOFs for Xe/Kr separation, focusing particularly on pore size as the distinguishing factor.22 This study highlighted a Pd-based MOF23 having pore sizes of 0.22, 0.49, and 0.58 nm as a promising candidate for Xe/Kr separation; Xe/Kr selectivity between 18 and 19.4 in the 0−1 MPa was predicted. Finally, Snurr and co-workers investigated Ar and N2 adsorption on IRMOF-1 using GCMC simulations.24 Their results significantly overpredict the experimental isotherms, an effect they attribute to defects in the experimental material. To compensate for this difference, they applied a scaling factor and were able achieve reasonable agreement. The only reported experimental separation of noble gases using a MOF was performed with HKUST-1, which was used to separate Xe from a 94/6 Kr/Xe mixture using continuous adsorption at 55 °C and 40 bar.25 Clearly, work in this area is still in its early stages and the factors controlling noble gas uptake in MOFs are yet to be thoroughly and systematically mapped. Most research to date has focused on surface area and pore size/shape as the controlling factors. However, other factors, such as the presence of open metal sites and the chemical functionalities on the organic linker, can strongly affect gas adsorption, as predicted by GCMC modeling of uptake of CH4 (which is similar to Kr in its adsorption properties) we recently reported.26 These aspects have received less attention in the literature, however. As a way of expanding the current understanding of the factors governing noble gas adsorption in MOFs, we recently embarked on an experimental and computational investigation of the effects of varying the polarizability of the functional groups on a MOF linker. The advantage of this approach is that functional groups can be systematically appended to an existing MOF class with relative ease, without requiring the synthesis of new MOF architectures that may be necessary to modulate surface area and pore size and shape. Although increasing the polarizability of the pore to improve uptake of weakly interacting gases makes intuitive sense, surprisingly, to the best of our knowledge, no systematic investigation of this effect has been reported (even for H2). Halogenated linkers provide an excellent series for exploring the effects of polarizability on gas adsorption. Since polarizability increases with atomic size, we expect the interactions to follow the pattern −I > −Br > −Cl > −F. Halogenated MOFs of various structures with −F,27−29 −Cl,27,30,31 −Br,32−34 and −I35,36 functionality have been reported. Previously, our group reported the IRMOF-2-X series, the first synthesis of an isoreticular series of MOFs with complete monohalogen replacement (Figure 1).37 This investigation demonstrated that for N2 at 292 K, adsorption increases with increasing polarizability of the linker. Here, we build on these results and report the effects of linker polarizability on the uptake of noble gases, reporting isotherms at several industrially relevant temperatures and the associated Henry’s constants (kH) and
Figure 1. (top) Model for the crystal structure of IRMOF-2. (bottom) Structure of the organic linkers for IRMOF-1 and IRMOF-2-X series.
isosteric heats of adsorption (Qst) for Ar, Kr, and Xe, as well as N2. Additionally, we paired our experimental effort with computational modeling, employing GCMC simulations to predict the adsorption isotherms, kH, and Qst for noble gases and N2 (serving as a proxy for air) adsorbed by the monohalogenated IRMOFs and IRMOF-1. Simulation results for Rn are also provided; in this case, computational modeling is an essential tool for screening candidate MOFs, since experimental data are difficult to obtain due to the radioactive nature of this gas. A two-pronged approach wherein modeling is used to screen and refine candidate MOFs prior to actual synthesis can greatly increase the efficiency of materials development; computational modeling of MOFs has been recently reviewed.10,38 Our primary purpose in performing these simulations is to evaluate its accuracy relative to experiment in preparation for future studies in which larger numbers of MOFs will be considered. Overall, in addition to the insights gained concerning noble gas uptake and separation, these results have implications for understanding and optimizing the adsorption of other, more thoroughly studied gases such as CH4, CO2, and H2.
■
MODELING AND EXPERIMENTAL METHODS Synthesis and Characterization of IRMOFs. IRMOFs were synthesized solvothermally and activated as reported previously.37 These MOFs were characterized by powder X-ray diffraction patterns, obtained on a PANalytical Empyrean diffractometer, and single crystal XRD measurements were made with an Agilent Supernova diffractometer. Details of these experiments are available in the Supporting Information. Isotherms were measured on 50−100 mg evacuated samples using a Quantachrome Quadrasorb-SI (Kr/MP) porosimeter (Quantachrome Instruments, USA). BET analysis was conducted from N2 isotherm measurements collected at 77 K using a liquid nitrogen bath. N2, Kr, Xe, and Ar adsorption 19766
dx.doi.org/10.1021/jp303274m | J. Phys. Chem. C 2012, 116, 19765−19772
The Journal of Physical Chemistry C
Article
Table 1. Material Properties for IRMOF-1 and the Halogenated IRMOF-2 Series37 MOF IRMOF-1 IRMOF-2-F IRMOF-2-Cl IRMOF-2-Br IRMOF-2-I
BET surface area: (m2/g) 3368 3133 2672 2461 1925
(3187)a (3172)a (3077)a (2693)a (2404)a
DA radius (Å)b
HK half pore width (Å)b
SF half pore width (Å)b
Total pore volume (cm3/g)c
percent free volumea
8.8 (7.8)67 9.0 8.8 8.8 8.7
4.3 (4.4)68 3.9 3.9 3.9 3.9
7.8 7.3 7.0 6.8 6.6
1.39 (1.19)d 1.28 1.09 0.94 0.77
78.8 78.6 77.2 76.4 75.5
a
Values calculated from crystallographic data using the Connolly method, implemented in Materials Studio with a 1.8 Å probe radius. bDubinin− Astakhov (DA),69 Horvath−Kawazoe (HK),70 and Saito−Foley (SF)71 pore width values for these IRMOFs were calculated from 77 K N2 isotherms using Quantachrome QuadraWin software. cTotal pore volume calculated at maximum adsorption using Quantachrome QuadraWin software. d Value calculated from crystallographic data.72
We find no appreciable difference between the isotherms for the two cases. We also evaluated the effect of the dihedral angle between the carboxylates and phenyl ring of the terephthalate groups. Unfortunately, there is ambiguity associated with this aspect of the IRMOF-2-halogen structures, because of disorder in the reported crystal structure and the possibility that there is some thermal motion of these planes with respect to each other at the temperatures of our measurements. Therefore, rather than make a somewhat arbitrary decision concerning this orientation, we computed Ar and Xe isotherms for IRMOF-2-I at the two extremes of the dihedral angle (structures with inplane phenyl rings and with rings rotated 90° relative to the plane of the carboxylates), for which we expect the largest effect to occur. As expected, there is almost no effect on the Ar isotherm, while for Xe there is reduced uptake by IRMOF-2-I at a dihedral angle of 90°. Nevertheless, we find that the coplanar structure provides the best agreement with the experimental results presented below, suggesting that the average orientation under measurement conditions is closer to coplanar. While the fully coplanar configuration is apparently sterically hindered based on the van der Waals radii, this may be partially overcome by hydrogen bonding between phenyl ring hydrogen atoms and carboxylate oxygens (three pairs of such interactions remain after halogen substitution), as is the case with IRMOF1, which has phenyl rings in a planar orientation.43,44 A single-site Lennard-Jones model was used for the noble gases while a three-site model, including charge, was used for N2.45−47 Parameters from the Universal Force Field48 (UFF) were used to model short-range pairwise interactions (MOF− guest and guest−guest). This combination of rigid framework and UFF parameters has been used to simulate adsorption isotherms before resulting in good qualitative agreement between simulation and experiment.49 Additionally, we calculated the interaction energy between argon and an isolated chlorobenzene ring using UFF. Our results are in good agreement with the interaction energy calculated by Oh and coworkers using ab initio methods.50 For MOF−N2 and N2−N2 interactions, an electrostatic term was also included using appropriate charges from DFT calculations.51 Fragmented clusters, containing the organic linker capped by metal clusters saturated with methyl groups, were used as models for the metal−organic frameworks. Charge distributions were determined using the ChelpG method52 and were obtained from the unrestricted B3LYP functional53,54 with the 6-31+G(d) basis set55,56 for light atoms H−Ar and the LANL2DZ ECP basis set57 for heavier atoms. Gaussian default atomic radii were used for all atoms except Zn and Br, which were modeled with radii 1.39 and 1.85 Å, respectively. Geometric mixing rules were used to calculate the interatomic interactions with a short-range
isotherms were collected at 292 K using a temperature equilibrated water bath and 273 K using an ice/water bath. For ambient temperature runs, the Po value was fixed at 760 mmHg. Isosteric heats of adsorption were calculated from linear fits of the 292 and 273 K adsorption isotherms using the Clausius−Clapeyron equation.39,40 ⎡ TT ⎤ ⎛ P ⎞ ΔH = R ⎢ 1 2 ⎥ln⎜ 2 ⎟ ⎣ T2 − T1 ⎦ ⎝ P1 ⎠
(1)
Conversion to number of molecules per cage assumes one formula unit, Zn4O(linker)3, per cage. Simulation. Grand Canonical Monte Carlo simulations were performed in the μVT ensemble (chemical potential μ, volume V, temperature T) using the MCCCS Towhee Code.41 In each Monte Carlo step, one of the following moves was chosen at random: particle translation, particle rotation (for N2 only), particle growth (for N2 only), particle insertion and deletion, and intrabox particle transfer. All moves were performed with equal probability, and the magnitudes of both the translational and rotational moves were updated throughout the simulation to achieve an acceptance ratio of one-half. Simulations were performed at 292 K for pressures between ∼0.001 and ∼2.5 atm. The Towhee code requires the adsorbate chemical potential be specified at the beginning of each simulation while isotherm results are reported as a function of pressure. The chemical potential of interest is first estimated by assuming the gas to be ideal and using a relation to equate chemical potential to pressure.42 Second, an “empty box” GCMC simulation is performed to determine the nonideal potential associated with the target pressures.21 Simulations were equilibrated for ten millions steps and then run for an additional ten millions steps. During the production phase, data were collected by block averaging with bin sizes set at 500,000 Monte Carlo steps to generate 20 bins. MOFs were modeled as rigid frameworks with atoms held fixed at their crystallographic positions, which were obtained from crystallographic information files (CIF) downloaded from the Cambridge Crystallographic Data Centre. Structures for IRMOF-2-(Cl, F, I) were created by replacing -Br in IRMOF-2Br with the halogen of interest. In the structure modeled by our simulations the halogens are located at random positions throughout the structure. Each MOF was represented by a single unit cell containing 424 atoms with a cell dimension of a = b = c = 26.0303 Å for IRMOF-2-(Br, Cl, F, I). The IRMOF-1 structure was also composed of 424 atoms but has a unit cell dimension of a = b = c = 25.669 Å. To assess the effect of halogen location on the linker aromatic ring, adsorption isotherms were computed for an IRMOF-2-I structure in which all of the halogens pointed into the corner of the pores. 19767
dx.doi.org/10.1021/jp303274m | J. Phys. Chem. C 2012, 116, 19765−19772
The Journal of Physical Chemistry C
Article
Figure 2. Experimental and simulated N2, Ar, Kr, and Xe isotherms of IRMOF-1 and the IRMOF-2-X series at 292 K.
Table 2. Henry’s Constants (1000 cm3·mol−1·atm−1) and Number of Gas Molecules Per Cage at 1 atm Calculated for the IRMOF-2-X Series and IRMOF-1 N2 MOF IRMOF-1 IRMOF-2-F IRMOF-2-Cl IRMOF-2-Br IRMOF-2-I
exp sim exp sim exp sim exp sim exp sim
Ar
Kr
Rn
kH
no.
kH
no.
kH
no.
kH
no.
3.20 4.18 2.97 4.11 3.14 4.64 3.63 4.72 3.78 5.05
0.142 0.185 0.132 0.182 0.139 0.206 0.161 0.209 0.168 0.224
3.86 4.82 2.94 4.57 3.22 5.10 3.31 5.20 3.50 5.56
0.171 0.214 0.130 0.203 0.143 0.226 0.147 0.230 0.155 0.246
9.68 9.85 7.18 9.21 8.23 10.9 8.74 11.3 9.67 12.5
0.429 0.437 0.318 0.408 0.365 0.483 0.387 0.501 0.429 0.554
27.4 25.4 24.4 24.3 30.1 31.4 33.8 33.0 40.3 38.1
1.21 1.13 1.08 1.08 1.33 1.39 1.50 1.46 1.79 1.69
cutoff of 12.5 Å. The long-range electrostatic interactions were calculated by Ewald summation with a precision of 0.0001 and a real-space cutoff of 12.5 Å. Henry’s law constants, isosteric heats of adsorption, and single component adsorption isotherms were calculated for each system. Values of kH were determined from the linear slope of the adsorption isotherm at very low pressures. Values of Qst were calculated using the ideal gas approximation where RT is the thermal energy and ⟨V⟩ is the energy of the gas/ framework interaction. Q st = RT − ⟨V ⟩
Xe kH
no.
75.1
12.0
61.9
7.98
86.4
9.96
92.9
9.83
112
10.4
This was determined for a system at infinite dilution, following the method of Düren and Snurr via a NVT Monte Carlo simulation consisting of one gas molecule and the MOF framework.58
■
RESULTS AND DISCUSSION
Materials properties, including BET surface areas, and pore size distributions for the IRMOF-2-X series and IRMOF-1 are listed in Table 1.37 In general, the surface area, pore size, and total pore volume all decrease slightly as the functionality on the terephthalic acid linker increases in mass.
(2) 19768
dx.doi.org/10.1021/jp303274m | J. Phys. Chem. C 2012, 116, 19765−19772
The Journal of Physical Chemistry C
Article
pattern emerges: the fidelity of the simulated Henry’s constants with the experimental measurements improves with increasing gas size. For N2, Ar, and Kr, the simulation results always overpredict the experimental results, although the values are closer for Kr than for the other two (Table 4). In contrast,
The linear nature of both the simulated and experimental adsorption isotherms corresponds to Henry’s law behavior, indicating that the binding sites in the MOF are not saturated at pressures up to 1 atm, as seen in Figure 2 for Ar, Kr, Xe, and N2, for the IRMOF-2-X series and IRMOF-1 at 292 K. Thus, a linear fit to these isotherms (see the Supporting Information) yields the Henry’s constants, which are listed in Table 2. The predicted number of gas molecules adsorbed at 1 atm confirms the Henry’s law regime at these pressures. For Ar and N2, only 0.12−0.23 gas molecules are adsorbed per cage, while for Kr the range increases to 0.3−0.6. For Xe, 1.1−1.8 gas molecules are adsorbed, and the values computed for IRMOF-1 and IRMOF-2-Br correspond well with the published values of 1.5 and 1.8 molecules per cage respectively obtained experimentally.17 These results establish that noble gas adsorption is not saturated at 1 atm for these materials and, therefore, should be primarily influenced by the chemical environment and shape of the pore and not by surface area or pore volume. Thus, our experimental and simulation results should directly reflect the influence of linker polarizability. GCMC simulations predict two major trends in the adsorption isotherms and associated Henry’s constants (Table 3). First, for a given MOF, adsorption increases with increasing
Table 4. Ratios of Experimental and Simulated Henry’s Constants: kH(exp)/kH(sim)
N2 Ar Kr Xe
experiment −I > −Br > −H > −Cl −I > −H > −Br > −Cl −H > −I > −Br > −Cl −I > −Br > −Cl > −H
simulation > > > >
−F −F −F −F
−I −I −I −I
> > > >
−Br −Br −Br −Br
> > > >
−Cl −Cl −Cl −Cl
> > > >
−H −H −H −H
> > > >
N2
Ar
Kr
Xe
0.766 0.724 0.678 0.770 0.749
0.801 0.644 0.632 0.637 0.629
0.983 0.780 0.753 0.775 0.774
1.08 1.00 0.961 1.02 1.06
simulations and experiments for Xe yield very similar results. A similar trend was observed by Dubbledam et al. in their simulations of IRMOF-1 Ar and N2 isotherms.24 These authors explained the differences as being the result of structural irregularities and pore collapse or occlusion in the synthesized MOF material, a rationale that is reasonable based on previous work demonstrating that surface areas are highly dependent upon the method used to active the MOF.37,60 In their case, the experimental results could be brought into good agreement with the modeling by applying a constant scaling factor, which is consistent with a uniform extent of pore collapse for all samples used in their experiments. Aware of this problem, we took extra precautions in the activation of our samples, employing an optimized method described previously. As a result, our materials exhibit BET surface areas in good agreement with those predicted by the Connolly method (Table 1).37 While we are aware that molecule-accessible surface areas are more physically meaningful than Connolly surface areas,61 the effects of this difference are not significant enough to account for the discrepancies we observe in simulated and measured isotherms. Consequently, the aforementioned explanation, particularly for pore blocking and collapse, seems unlikely. In fact, the use of a constant scaling factor would not be effective in our case, since the ratio between the experimental and simulated kH is not constant either within the data for a particular MOF or across the MOF set for a specific gas. An alternative explanation for the observed difference is that in GCMC simulations, particles are inserted into locations within the MOF that might be experimentally inaccessible due to diffusion limitations and/or the structure of the MOF framework. In the IRMOF-2-X series, the pore-limiting diameter decreases as the size of the halogen increases but not sufficiently to exclude uptake of any of the noble gases. However, the halogen does point toward the Zn4O cluster, which may create pockets that can accommodate smaller gases in silico but that are not accessible to those gases experimentally. It is well established that the primary light gas and noble gas binding sites in IRMOFs reside close to the Zn− O clusters.17,24 Clearly, Xe is significantly larger than any of the other gases (kinetic diameter of 4.047 vs 3.655 Å for Kr and 3.542 Å for Ar, for example)12 and therefore could be excluded from such inaccessible pockets either in silico or in vitro. Unfortunately, such effects are difficult to verify computationally for these IRMOFs, which lack multiple pores of distinctly different sizes, as is the case in HKUST-1. Adsorbate density plots (data not shown) and Connolly surface areas with
Table 3. Comparison of the Henry’s Constants for the IRMOF-2-X Series and IRMOF-1 gas
MOF IRMOF-1 IRMOF-2-F IRMOF-2-Cl IRMOF-2-Br IRMOF2-I
−F −F −F −F
atomic size and, thus, the polarizability of the gas: Xe > Kr > Ar ≈ N2. Second, for the halogenated MOF series, adsorption increases with the increasing polarizability of the linker: −I > −Br > −Cl > −F. In general, our experimental results adhere to both of these patterns. Greater uptake is observed for Xe over Kr, and Kr over N2 and Ar; the N2 and Ar uptakes are roughly the same. Trends determined from the measured 1-atm isotherms as a function of linker functionality are also listed in Table 3. For the IRMOF-2-X series, the pattern seen in the simulation results holds experimentally, corresponding to our expectation that increasing the linker polarizability increases adsorbate−MOF interactions. The results for nonhalogenated IRMOF-1, however, only follow the simulation trend predicted for Xe; for the other gases, it has higher than expected adsorption relative to the halogenated MOFs. This inconsistent behavior by IRMOF-1 relative to the halogenated MOFs may be due to the presence of a minor amount of the interpenetrated phase of this MOF (IRMOF-1-INT),59 which has smaller pore spaces and could thus increase the uptake of the smaller gases, skewing the results. This phase is not evident in the PXRD of our samples but is difficult to detect due to the close similarity of the diffraction patterns.59 The fact that the results for Xe are not affected probably indicates that the pores in IRMOF-1-INT are not readily accessible to this larger gas. Therefore, in the following presentation of our results, we confine our discussion of trends in gas adsorption to the halogenated IRMOF series. Although the results described above are, for the most part, consistent with chemical intuition, upon comparison of the simulated and experimental data, an interesting and unexpected 19769
dx.doi.org/10.1021/jp303274m | J. Phys. Chem. C 2012, 116, 19765−19772
The Journal of Physical Chemistry C
Article
Table 6. Heats of Adsorption [kJ mol−1] Calculated from Both Experimenta and Simulationb
multiple probe sizes (see Supporting Information), for example, provide no evidence to confirm our hypothesis. Of course, another possibility is that these results highlight a weakness in the force field parameters used in this type of simulation. In some cases, van der Waals parameters for gas/framework interactions have been adjusted to compensate for overprediction of adsorption by GCMC simulation compared with experiment.62−65 While that empirical approach may succeed in a few specific cases, it is not based on more detailed quantum calculations. And as Table 4 shows, adjusting the van der Waals parameters of framework atoms to bring N2 and Ar into better agreement with the experiments is not feasible without significantly decreasing the fidelity of the remaining gases with the measured data. Better agreement with experiment is seen for Xe and Kr; it is possible that for larger, more polarizable gases the combination of general force field parameters for framework atoms with those of the gas yields more accurate framework/gas interaction energies. We also estimated gas selectivities by comparing Henry’s constants from both GCMC simulations and experimental data, as shown in Table 5. Both model and experiment indicate that
MOF IRMOF-1 IRMOF2-F IRMOF2-Cl IRMOF2-Br IRMOF2-I a b
MOF
IRMOF-2-F IRMOF-2-Cl IRMOF-2-Br IRMOF-2-I
exp sim exp sim exp sim exp sim exp sim
Ar/N2
Kr/N2
Xe/N2
Xe/Kr
1.21 1.15 0.99 1.11 1.03 1.10 0.912 1.10 0.926 1.10
3.03 2.36 2.42 2.24 2.62 2.35 2.41 2.39 2.56 2.48
8.56 6.08 8.22 5.91 9.59 6.77 9.31 6.99 10.7 7.54
2.83 2.58 3.40 2.84 3.66 2.88 3.87 2.92 4.17 3.05
N2
Ar
Kr
Xe
11.68 8.11 8.23 8.12 8.32 8.52 8.53 8.42 8.41 8.47
0.74 8.35 10.97 8.10 10.25 8.48 13.61 8.42 12.30 8.63
6.11 10.84 12.85 10.27 12.41 10.98 14.21 11.00 15.21 10.27
11.10 13.34 12.14 13.43 14.04 14.30 14.83 14.24 15.34 13.14
Experimental values listed are for the lowest loading measured. Computational values are measured at infinite dilution.
values calculated from GCMC at infinite dilution. Since these isotherms fall into the Henry’s Law regime and thus are linear, the Qst values are equivalent at all loadings in the region we studied. In general for the halogenated MOFs, the data exhibit a pattern of increasing Qst with increasing size (and polarizability) of the adsorbed gas. The experimental Qst data for Kr and Xe also increase with increasing polarizability of the linker functionality. The slightly out of trend Kr heat of adsorption for IRMOF-2-F may be due to the presence of a minor contaminant interpenetrated phase having smaller pores and, consequently, a higher heat of adsorption.66 Trends with linker polarizability are not obvious in the N2 and Ar Qst data. Comparing the simulations with the experimental data, the simulated Qst values are in excellent agreement with their experimental counterparts for N2, but the Ar, Kr, and Xe data are for the most part somewhat underpredicted by the simulations. The predicted trends in Qst also show a general increase with increasing adsorbate size. However, the trend with linker polarizability for Kr and Xe revealed in the experimental data is not reproduced by the simulations. This may be attributable to lack of explicit polarizability terms in the UFF forcefield. The combined experiments and modeling described above demonstrate that our simulation methods accurately predict trends in noble gas adsorption data, especially for larger gases, allowing us to apply the method to Rn adsorption by the IRMOF-2-X series and IRMOF-1. Results of these simulations are shown in Figure 3. For this gas, GCMC modeling is of high
Table 5. Gas Selectivity via Comparison of Henry’s Constants IRMOF-1
exp sim exp sim exp sim exp sim exp sim
Rn/N2 18.0 15.1 18.6 19.6 22.2
the selectivity of Ar verses N2 is approximately unity for all MOFs studied, and that there is no significant preference for Ar over N2 even as the polarizability of the MOF is increased. However, selective adsorption is observed in both experiment and simulation for Kr, Xe, and Rn over N2 and Xe/Kr; furthermore, this selectivity increases with increasing polarizability of the MOF. This trend is particularly notable for Xe/ N2, where an increase of 30% from 8.22 for IRMOF-2-F to 10.7 for IRMOF-2-I is observed for the experimental data. Although measured Xe/Kr selectivities are more modest, switching from −F to −I functionality yields a 23% increase in selectivity. Simulations predict a large increase in Rn/N2 selectivity, 47% from IRMOF-2-F to IRMOF-2-I; however these results are only valid while adsorption follows Henry’s law behavior (discussed below). In almost all cases except for Ar/N2, the experimental selectivity is greater than the simulated selectivity, but this can be attributed to the overprediction of the N2 and Ar isotherms, as discussed above. Nevertheless, the observed increase in selectivity resulting from increased halogen size indicates stronger adsorbate/MOF interactions with increasing polarizability of the pore environment, a result that has implications for designing MOFs for gas separations. Isosteric heats of adsorption calculated from isotherms at 292 and 273 K using the Clausius−Clapeyron equation are consistent with the trends in gas size described above for the Henry’s constants. These data are listed in Table 6, along with
Figure 3. Simulated adsorption isotherms for Rn. 19770
dx.doi.org/10.1021/jp303274m | J. Phys. Chem. C 2012, 116, 19765−19772
The Journal of Physical Chemistry C
Article
experimental and computational probe gases access all volume that is actually available to adsorb gas, whereas the GCMC calculation can access additional pore volume as a result of its insertion methodology. Finally, our simulated Rn adsorption data for these MOFs suggest new applications for these materials such as home air purification and Rn detection. Both the size and high polarizability of this gas favor strong binding to IRMOF pores, as shown by adsorption isotherms that begin to saturate at only 0.6 atm. This result highlights the value of simulations for understanding these complex materials and for designing new ones with improved properties. In future work, we plan to exploit these computational and experimental results to design and synthesize new MOFs with optimized properties for the adsorption and separation of noble gases, and to further probe the accuracy of GCMC simulations with a wider variety of MOFs to help improve MOF screening in silico.
value, as Rn isotherms are difficult and costly to measure due to its radioactivity. Consequently, our results provide a window into how Rn uptake would be affected by linker polarizability. At low pressures (0.6 atm), the observed trend for Rn is consistent with trends displayed by the other noble gases. However, at pressures greater than 0.6 atm, which correspond to approximately three gas atoms per formula unit, the trend reverses and gas uptake becomes inversely correlated with polarizability: IRMOF-1 > IRMOF-2-F > −Cl > −Br > −I. It appears that Rn saturates the pores at these pressures, leading to deviations from Henry’s law behavior and increasing loading dependency on surface area and total pore volume rather than pore polarizability. As is shown in Table 1, IRMOF-1 has the highest surface area and free volume of the MOFs considered and therefore has the largest accessible volume at pressures where MOF/adsorbate interactions are less important than the total volume available for the adsorbate to fill. These results represent the first simulations of Rn−MOF interactions and could lead to the design of Rn-selective MOFs for improved gas separation, decontamination, or Rn sensing.
■
ASSOCIATED CONTENT
S Supporting Information *
■
Experimental procedures for the characterization of these materials including PXRD and porosimetry, as well as kH linear fits. This material is available free of charge via the Internet at http://pubs.acs.org
CONCLUSIONS On the basis of Henry’s constants, the combined experimental and modeling results present an unambiguous picture with respect to the influence of linker polarizability on the uptake of noble gases, a conclusion that is likely applicable to other gases that interact weakly with MOF frameworks. On the basis of the measured Henry’s constants, these effects can be substantial: replacing one hydrogen on the BDC linker of IRMOF-1 with iodine produces a 47% increase in kH for Xe. The magnitude of the effect is more difficult to gauge for the other, smaller gases, however. In these cases, the observed adsorption behavior likely results from two factors: first, the presence of minor amounts of an interpenetrated impurity and, second, inaccessible regions created by the halogen substitution. The former is very difficult to avoid and quantify, but such phases are known for IRMOF-1. Even though our measured surface areas are generally in good agreement with those predicted by the Connolly method and considerably exceed those often reported in the literature,32,36 it is conceivable that an impurity with smaller pores and thus significantly higher kH and Qst could artificially raise the values for IRMOF-1. Consequently, the influence of linker polarizability is more accurately gauged from the IRMOF-2-X series only, for which kH increases from −F to −I in all cases. The second, and possibly more profound, factor influencing our results is the relatively large disagreement between simulation and experiment for gases other than Xe. Although our hypothesis that experimentally inaccessible pockets are formed as a result of halogenation is speculative at this point, it is supported by knowledge of the now well-characterized gas binding sites in IRMOF-1. The strongest of these sites is in the pore corners near the Zn4O clusters, which are the locations most easily blocked by a halogenated linker. The fact that the disagreement between experiment and modeling is fairly constant for a given gas across the five MOFs (Table 4) suggests, however, that only a minor increase in atomic radius is necessary to impede access to this site. For example, the nonbonded van der Waals distance for fluorine in UFF is 2.997 Å, while for hydrogen this diameter is 2.571 Å.48 Consistent with this analysis is that the measured and predicted surface areas of IRMOF-2-F are nearly identical, in spite of the fact that the measured N2 Henry’s constant for this MOF is ∼28% smaller than the simulated value. This suggests that both the
■
AUTHOR INFORMATION
Corresponding Author
*E-mail:
[email protected]. Author Contributions §
These authors contributed equally. S.T.M. designed, conducted, and analyzed the experiments. S.L.T.-M. conducted and analyzed the simulations. Notes
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS We would like to thank Marie V. Parkes for performing the DFT calculations. This work was supported by the Department of Energy and Sandia National Laboratories. Sandia is a multiprogram laboratory operated by Sandia Corporation, a Lockheed Martin Company, for the United States Department of Energy’s National Nuclear Security Administration under contract DE-AC04-94AL85000.
■
REFERENCES
(1) Al-Garni, A. Z. J. Thermophys. Heat Transfer 1996, 10, 119−125. (2) Kuo, D.-H.; Cheung, B.-Y.; Wu, R.-J. Thin Solid Films 2001, 398− 399, 35−40. (3) Liu, L. T.; Xu, Y.; Tang, P. J. Phys. Chem. B 2010, 114, 9010− 9016. (4) Kebabian, P. L.; Romano, R. R.; Freedman, A. Meas. Sci. Technol. 2003, 14, 983−988. (5) Yeralan, S.; Doughty, D.; Blondia, R.; Hamburger, R. Proc. SPIEInt. Soc. Opt. Eng. 2005, 5740, 27−35. (6) Stein, J. D.; Challa, P. Curr. Opin. Ophthalmol. 2007, 18, 140− 145. (7) http://www.epa.gov/radon/healthrisks.html (accessed April 4, 2012). (8) Pence, D. T.; Paplawsky, W. J. Proc. DOE Nucl. Air Clean. Conf. 1981, 16, 161−176. (9) Izumi, J. Waste gas treatment using zeolites in nuclear-related industries. In Handbook of Zeolite Science and Technology; Auerbach, S. M., Carrado, K. A., Dutta, P. K., Eds.; Marecel Dekker: New York: 2003. 19771
dx.doi.org/10.1021/jp303274m | J. Phys. Chem. C 2012, 116, 19765−19772
The Journal of Physical Chemistry C
Article
(10) Meek, S. T.; Greathouse, J. A.; Allendorf, M. D. Adv. Mater. 2011, 23, 249−267. (11) Murray, L. J.; Dinca, M.; Long, J. R. Chem. Soc. Rev. 2009, 38, 1294−1314. (12) Li, J. R.; Kuppler, R. J.; Zhou, H. C. Chem. Soc. Rev. 2009, 38, 1477−1504. (13) Horcajada, P.; Chalati, T.; Serre, C.; Gillet, B.; Sebrie, C.; Baati, T.; Eubank, J. F.; Heurtaux, D.; Clayette, P.; Kreuz, C.; et al. Nat. Mater. 2010, 9, 172−178. (14) Lee, J.; Farha, O. K.; Roberts, J.; Scheidt, K. A.; Nguyen, S. T.; Hupp, J. T. Chem. Soc. Rev. 2009, 38, 1450−1459. (15) Allendorf, M. D.; Schwartzberg, A.; Stavila, V.; Talin, A. A. Chem.Eur. J. 2011, 17, 11372−11388. (16) Kreno, L. E.; Leong, K.; Farha, O. K.; Allendorf, M.; Van Duyne, R. P.; Hupp, J. T. Chem. Rev. 2012, 112, 1105−1125. (17) Pawsey, S.; Moudrakovski, I.; Ripmeester, J.; Wang, L. Q.; Exarhos, G. J.; Rowsell, J. L. C.; Yaghi, O. M. J. Phys. Chem. C 2007, 111, 6060−6067. (18) Farrusseng, D.; Daniel, C.; Gaudillere, C.; Ravon, U.; Schuurman, Y.; Mirodatos, C.; Dubbeldam, D.; Frost, H.; Snurr, R. Q. Langmuir 2009, 25, 7383−7388. (19) Dietzel, P. D. C.; Panella, B.; Hirscher, M.; Blom, R.; Fjellvag, H. Chem. Commun. 2006, 959−961. (20) Thallapally, P. K.; Grate, J. W.; Motkuri, R. K. Chem. Commun. 2012, 48, 347−349. (21) Greathouse, J. A.; Kinnibrugh, T. L.; Allendorf, M. D. Ind. Eng. Chem. Res. 2009, 48, 3425−3431. (22) Ryan, P.; Farha, O. K.; Broadbelt, L. J.; Snurr, R. Q. AIChE J. 2011, 57, 1759−1766. (23) Xamena, F.; Abad, A.; Corma, A.; Garcia, H. J. Catal. 2007, 250, 294−298. (24) Dubbeldam, D.; Frost, H.; Walton, K. S.; Snurr, R. Q. Fluid Phase Equilib. 2007, 261, 152−161. (25) Mueller, U.; Schubert, M.; Teich, F.; Puetter, H.; Schierle-Arndt, K.; Pastre, J. J. Mater. Chem. 2006, 16, 626−636. (26) Zeitler, T. R.; Allendorf, M. D.; Greathouse, J. A. J. Phys. Chem. C 2012, 116, 3492−3502. (27) Devic, T.; Horcajada, P.; Serre, C.; Salles, F.; Maurin, G.; Moulin, B.; Heurtaux, D.; Clet, G.; Vimont, A.; Greneche, J. M.; et al. J. Am. Chem. Soc. 2010, 132, 1127−1136. (28) Achmann, S.; Hagen, G.; Kita, J.; Malkowsky, I. M.; Kiener, C.; Moos, R. Sensors-Basel 2009, 9, 1574−1589. (29) Chen, B. L.; Yang, Y.; Zapata, F.; Qian, G. D.; Luo, Y. S.; Zhang, J. H.; Lobkovsky, E. B. Inorg. Chem. 2006, 45, 8882−8886. (30) Deng, H. X.; Doonan, C. J.; Furukawa, H.; Ferreira, R. B.; Towne, J.; Knobler, C. B.; Wang, B.; Yaghi, O. M. Science 2010, 327, 846−850. (31) He, M. Y.; Chen, S. C.; Zhang, Z. H.; Huang, K. L.; Yin, F. H.; Chen, Q. Inorg. Chim. Acta 2009, 362, 2569−2576. (32) Rowsell, J. L. C.; Yaghi, O. M. J. Am. Chem. Soc. 2006, 128, 1304−1315. (33) Jones, S. C.; Bauer, C. A. J. Am. Chem. Soc. 2009, 131, 12516− 12517. (34) Farha, O. K.; Malliakas, C. D.; Kanatzidis, M. G.; Hupp, J. T. J. Am. Chem. Soc. 2010, 132, 950−952. (35) Furukawa, H.; Kim, J.; Ockwig, N. W.; O’Keeffe, M.; Yaghi, O. M. J. Am. Chem. Soc. 2008, 130, 11650−11661. (36) Burrows, A. D.; Fisher, L. C.; Richardson, C.; Rigby, S. P. Chem. Commun. 2011, 47, 3380−3382. (37) Meek, S. T.; Perry, J. J.; Teich-McGoldrick, S. L.; Greathouse, J. A.; Allendorf, M. D. Cryst. Growth Des. 2011, 11, 4309−4312. (38) Han, S. S.; Mendoza-Cortes, J. L.; Goddard, W. A. Chem. Soc. Rev. 2009, 38, 1460−1476. (39) Kaye, S. S.; Long, J. R. Chem. Commun. 2007, 4486−4488. (40) Ma, S. Q.; Zhou, H. C. J. Am. Chem. Soc. 2006, 128, 11734− 11735. (41) Martin, M. G.; Siepmann, J. I. J. Phys. Chem. B 1999, 103, 4508− 4517.
(42) Frenkel, D.; Smit, B. Understanding Molecular Simulation From Algorithms to Applications; Academic Press: Waltham, MA, 2002. (43) Zhou, W.; Yildirim, T. Phys. Rev. B 2006, 74, 180301. (44) Gonzalez, J.; Devi, R. N.; Tunstall, D. P.; Cox, P. A.; Wright, P. A. Microporous Mesoporous Mater. 2005, 84, 97−104. (45) Potoff, J. J.; Siepmann, J. I. AIChE J. 2001, 47, 1676−1682. (46) Pellenq, R. J. M.; Levitz, P. E. Mol. Phys. 2002, 100, 2059−2077. (47) van Loef, J. J.. Phys. B+C 1981, 103, 362−364. (48) Rappe, A. K.; Casewit, C. J.; Colwell, K. S.; Goddard, W. A.; Skiff, W. M. J. Am. Chem. Soc. 1992, 114, 10024−10035. (49) Vishnyakov, A.; Ravikovitch, P. I.; Neimark, A. V.; Bulow, M.; Wang, Q. M. Nano Lett. 2003, 3, 713−718. (50) Oh, J. J.; Park, I.; Wilson, R. J.; Peebles, S. A.; Kuczkowski, R. L.; Kraka, E.; Cremer, D. J. Chem. Phys. 2000, 113, 9051−9059. (51) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.; Robb, M. A.; Cheeseman, J. R.; Scalmani, G.; Barone, V.; Mennucci, B.; Petersson, G. A.; et al. Gaussian, Inc.: Wallingford CT, 2009. (52) Breneman, C. M.; Wiberg, K. B. J. Comput. Chem. 1990, 11, 361−373. (53) Becke, A. D. J. Chem. Phys. 1993, 98, 5648−5652. (54) Lee, C.; Yang, W.; Parr, R. G. Phys. Rev. B 1988, 37, 785−789. (55) Francl, M. M.; Pietro, W. J.; Hehre, W. J.; Binkley, J. S.; Gordon, M. S.; DeFrees, D. J.; Pople, J. A. J. Chem. Phys. 1982, 77, 3654−3665. (56) Hehre, W. J.; Ditchfield, R.; Pople, J. A. J. Chem. Phys. 1972, 56, 2257−2261. (57) Hay, P. J.; Wadt, W. R. J. Chem. Phys. 1985, 82, 299−310. (58) Düren, T.; Snurr, R. Q. J. Phys. Chem. B 2004, 108, 15703− 15708. (59) Chen, B.; Wang, X.; Zhang, Q.; Xi, X.; Cai, J.; Qi, H.; Shi, S.; Wang, J.; Yuan, D.; Fang, M. J. Mater. Chem. 2010, 20, 3758−3767. (60) Kaye, S. S.; Dailly, A.; Yaghi, O. M.; Long, J. R. J. Am. Chem. Soc. 2007, 129, 14176−14177. (61) Düren, T.; Millange, F.; Ferey, G.; Walton, K. S.; Snurr, R. Q. J. Phys. Chem. C 2007, 111, 15350−15356. (62) Liu, B.; Smit, B. J. Phys. Chem. C 2010, 114, 8515−8522. (63) Perez-Pellitero, J.; Amrouche, H.; Siperstein, F. R.; Pirngruber, G.; Nieto-Draghi, C.; Chaplais, G.; Simon-Masseron, A.; Bazer-Bachi, D.; Peralta, D.; Bats, N. Chem.Eur. J. 2010, 16, 1560−1571. (64) Jorge, M.; Lamia, N.; Rodrigues, A. E. Colloid Surface A 2010, 357, 27−34. (65) Yang, Q. Y.; Zhong, C. L. J. Phys. Chem. B 2006, 110, 17776− 17783. (66) Kim, H.; Das, S.; Kim, M. G.; Dybtsev, D. N.; Kim, Y.; Kim, K. Inorg. Chem. 2011, 50, 3691−3696. (67) Liu, Y. Y.; Ng, Z. F.; Khan, E. A.; Jeong, H. K.; Ching, C. B.; Lai, Z. P. Microporous Mesoporous Mater. 2009, 118, 296−301. (68) Saha, D. P.; Wei, Z. J.; Deng, S. G. Sep. Purif. Technol. 2009, 64, 280−287. (69) Rouquerol, F.; Rouquerol, J.; Sing, K. In Adsorption by Powders and Porous Solids; Academic Press: London, 1999, p 219−236. (70) Horvath, G.; Kawazoe, K. J. Chem. Eng. Jpn. 1983, 16, 470−475. (71) Saito, A.; Foley, H. C. AIChE J. 1991, 37, 429−436. (72) Poirier, E.; Dailly, A. J. Phys. Chem. C 2008, 112, 13047−13052.
19772
dx.doi.org/10.1021/jp303274m | J. Phys. Chem. C 2012, 116, 19765−19772