Forum Article www.acsami.org
Molecular Rhodium Complexes Supported on the Metal-Oxide-Like Nodes of Metal Organic Frameworks and on Zeolite HY: Catalysts for Ethylene Hydrogenation and Dimerization Varinia Bernales,†,‡ Dong Yang,†,§ Jun Yu,§ Gamze Gümüsļ ü,§ Christopher J. Cramer,‡ Bruce C. Gates,*,§ and Laura Gagliardi*,‡ ‡
Department of Chemistry, Chemical Theory Center, and Minnesota Supercomputing Institute, University of Minnesota, Minneapolis, Minnesota 55455-0431, United States § Department of Chemical Engineering, University of California, Davis, California 95616, United States S Supporting Information *
ABSTRACT: Metal−organic frameworks (MOFs) with nodes consisting of zirconium oxide clusters (Zr6) offer new opportunities as supports for catalysts with well-defined, essentially molecular, structures. We used the precursor Rh(C2H4)2(acac) (acac is acetylacetonate) to anchor Rh(I) complexes to the nodes of the MOF UiO-67 and, for comparison, to the zeolite dealuminated HY (DAY). These were characterized experimentally by measurement of catalytic activities and selectivities for ethylene hydrogenation and dimerization in a once-through flow reactor at 298 K and 1 bar. The catalyst performance data are complemented with structural information determined by infrared and extended X-ray absorption fine structure spectroscopies and by calculations at the level of density functional theory, the latter carried out also to extend the investigation to a related MOF, NU-1000. The agreement between the experimental and calculated structural metrics is good, and the calculations have led to predictions of reaction mechanisms and associated energetics. The data demonstrate a correlation between the catalytic activity and selectivity and the electron-donor tendency of the supported rhodium (as measured by the frequencies of CO ligands bonded as probes to the Rh(I) centers), which is itself a measure of the electron-donor tendency of the support. KEYWORDS: metal−organic framework nodes, ethylene dimerization, ethylene hydrogenation, rhodium complexes, supported catalyst, density functional theory
■
INTRODUCTION In contrast to molecular catalysts in solution, most solid catalysts are structurally heterogeneous and challenging to understand in depth. But when molecular species such as mononuclear metal complexes are anchored to solid supports that have nearly uniform structures, the catalytic species may be characterized incisively, by spectroscopy and theory. Thus, supported molecular catalysts are drawing increasing attention, much of it focused on those involving supports that are crystallinee.g., zeolites and metal organic frameworks (MOFs)because these are among the porous high-area supports having the best-defined structures.1−9 Among the most thoroughly investigated of the soluble metal complex catalysts are complexes of rhodium, because they are widely applied in industry, for example, in processes for alkene hydrogenation, alkene hydroformylation, and methanol carbonylation.10−13 The catalytic activities and selectivities of these and other metal complexes are often tuned and optimized by choice of the ligands bonded to the metal. When metal complexes are supported on solids, the supports themselves are ligands. Zeolites are appealing as catalyst supports not only because they offer well-defined structures and prospects for good fundamental understanding, but also because they offer the © XXXX American Chemical Society
practical advantages of porous, high-area materials that are stable at high temperatures. Some MOFs offer these same advantages, but to date they are much less well investigated than zeolites as catalyst supports. MOFs are formed by connecting nodes with organic linkers. Because there are so many node-linker combinations, there are an enormous number of possible MOF structuresfar more than the number of known zeolites.14−24 Some MOFs offer sites similar to those on metal oxides and zeolites for anchoring metal complexes. For example, the cationic Rh complexes (dppe)Rh-(COD)BF4 and (MeCN)2Rh(COD)BF4 were anchored by ion exchange onto the nodes of ZJU-28. 10 Alternatively, they were anchored to the linkers of MIL-101SO3. The resultant site-isolated Rh(I) species were reported to be precursors of catalysts for the hydrogenation of terminal and internal olefins.10 These examples are an indication of the opportunities for synthesis of site-isolated metal complexes within the porous structures of MOFs,7,25−30 and the site Special Issue: Hupp 60th Birthday Forum Received: March 18, 2017 Accepted: May 9, 2017
A
DOI: 10.1021/acsami.7b03858 ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX
Forum Article
ACS Applied Materials & Interfaces
of H4TBAPy was added to this solution, and the mixture was sonicated for 20 min. The yellow suspension was held in an oven at 353 K for 24 h. After cooling to room temperature, yellow polycrystalline material was isolated by filtration (35 mg of activated material, 54% yield) and washed with DMF and subsequently activated with HCl. Then, the solid was washed twice with DMF and six times with acetone. The powder was dried at room temperature The resulting Zr-NU-1000 and Hf-NU-1000 were found to have BET surface areas of 2100 and 1700 m2/g, respectively. Synthesis of MOF-Supported Rh(C2H4)2 and Ir(C2H4)2. Sample synthesis and handling were performed with air and moisture exclusion by use of a double-manifold Schlenk line and an argonatmosphere glovebox. The precursor Rh(C2H4)2(acac) (Strem, 99%) (acac is acetylacetonate) was used as received. Supported rhodium complex catalysts were prepared by bringing Rh(C2H4)2(acac) into contact with activated UiO-67, Zr-NU-1000, or Hf-NU-1000 powder (activated under vacuum (1 × 10−7 Torr) at 393 K for 12 h before use) in a Schlenk flask. The mixture was slurried in 20 mL of dried n-pentane (Fisher, 99%) at room temperature. After 12 h (if not specified, each of the reactions was conducted for 12 h), the solvent was removed by evacuation for a day, leaving all the rhodium bonded to the MOF. The resultant solids (e.g., that containing 2.5 wt % rhodium, with approximately 0.53 Rh atoms per node for UiO-67, 0.54 Rh atoms per node for Zr-NU-1000, and 0.67 Rh atoms per node for Hf-NU-1000), was stored in an argon-filled glovebox. In other experiments, rhodium loadings were varied and inferred from the conditions of the syntheses, whereby all the added rhodium remained in the MOF after removal of the solvent. The synthesis of the UiO-67-supported iridium complex from Ir(C2H4)2(acac) was carried out under the same conditions, as before;1 the iridium loading was 5.0 wt %, about 0.58 Ir atoms per node. Synthesis of Zeolite-Supported Rh(C2H4)2. DAY zeolitesupported Rh(C2H4)2 complexes were synthesized similarly, by reaction of Rh(C2H4)2(acac) with dealuminated Y zeolite (DAY, SiO2/Al2O3 ratio = 30).45−47 The rhodium loading was 1.0 wt %. Synthesis of MgO-Supported Rh(C2H4)2. Similarly, Rh(C2H4)2 complexes were synthesized on MgO powder (BET surface area 70 m2/g). The rhodium loading was 0.4 wt %. Determination of Numbers of Node Defect Sites. The numbers of defects per node in the MOFs were determined by differential thermal analysis, as described elsewhere.41 Infrared Spectroscopy. A Bruker IFS 66v/S spectrometer with a spectral resolution of 2 cm−1 was used to collect transmission IR spectra of pressed powder samples. Approximately 20 mg of solid sample in the glovebox was pressed into a thin wafer and loaded into a cell that also served as a flow reactor (In-Situ Research Institute, South Bend, IN). The cell was sealed and connected to a flow system, and spectra were recorded as reactant gases flowed through the cell at various temperatures. Each spectrum is the average of 64 scans. X-ray Absorption Spectroscopy. X-ray absorption spectra were collected at X-ray beamline 2−2 of the Stanford Synchrotron Radiation Lightsource. The storage ring electron energy was 3 GeV and the ring current ∼300 mA. A double-crystal Si(220) monochromator was detuned by 15%−20% at the Rh K edge to minimize the effects of higher harmonics in the X-ray beam. Samples were transferred to the X-ray absorption spectroscopy cells in an argon-atmosphere glovebox and sealed before mounting in the X-ray beam. Data were analyzed as before,13 and details are given in the Supporting Information. Catalytic Reaction Experiments. Ethylene conversion catalysis was carried out in a conventional laboratory once-through tubular plug-flow reactor at 298 K and 1 bar. The catalyst (10−30 mg) was mixed with 10 g of inert, nonporous α-Al2O3 powder and loaded into the reactor in the argon-filled glovebox. The feed partial pressures were 50 mbar of C2H4, 50 mbar of H2, and 900 mbar of helium, with a total feed flow rate of 100 mL(NTP)/min. Products were analyzed with an online Agilent Model 6890 gas chromatograph. The ethylene conversions were kept low, in the differential conversion range.
isolation might be expected to mitigate against the aggregation of metal complexes that can occur in solution,25−27,31 although it is not guaranteed to prevent it. We now report rhodium complex catalysts anchored to MOFs with nodes that are essentially zirconium oxide clusters. These MOFs, UiO-67 and NU-1000, offer the advantages of high thermal and chemical stability. UiO-67 incorporates nodes represented as [Zr6(μ3-O)4(μ3-OH)4]12+, each of which may be coordinated to 12 biphenyl-4,4′-dicarboxylate linkers, although defects are common, and the node coordination number is typically less than 12. The MOF NU-1000 incorporates only 8 linkers and comprises [Zr6(μ3-O)4(μ3-OH)4(OH)4(OH2)4]8+ nodes modulated with tetratopic 1,3,6,8-tetrakis(p-benzoate)pyrene linkers that provide three distinguishable large pores (Figure S7). Two of the largest pores of NU-1000 are associated with channels having diameters of about 31 and 10 Å, respectively.32,33 The nodes on these MOFs that are not coordinated by carboxylates (either by design or because they are defects) present reactive −OH/−OH2 groups that can be functionalized with a transition metal complex by atomic layer deposition from the gas phase or by chemisorption from the liquid phase.34−38 NU-1000 further offers opportunities for tuning the secondary coordination environment around the transition metal by ligand modification to tune the reactivity and catalytic properties.29,30,32−36,39−41 Control of the number of defect sites in UiO-67 can be achieved in part by control of the number of organic linkers.41 Taking advantage of the well-defined structures of NU-1000 and UiO-67, we undertook an experimental and computational investigation of rhodium complex catalysts supported on their nodes, using Rh(C2H4)2(acac) (acac is acetylacetonate) as a precursor to react with the node surfaces. The goals of our research were to gain fundamental understanding of the supported rhodium complexes as catalysts for ethylene hydrogenation and dimerization and to compare the MOF nodes as ligands that serve as catalyst supports with the wellinvestigated DAY zeolite. The results lend themselves to comparison with both computational and experimental results characterizing analogous metal complex catalysts on other supports42−44 and provide new evidence of a correlation between the catalytic properties of metal complexes on various supports and the stretching frequencies of CO ligands on the metalsa correlation that links the catalytic properties to the electron-donor tendency of the support as a ligand.
■
EXPERIMENTAL SECTION
Synthesis of UiO-67 with HCl Modulator. ZrCl4 (67 mg, 0.27 mmol) and 0.5 mL of concentrated HCl modulator (37 wt %) were dissolved in 15 mL of DMF in an 8 dram vial as ultrasound was applied for 5 min. Biphenyl-4,4′-dicarboxylic acid (0.09 g, 0.38 mmol) was added to the resultant solution and dispersed by ultrasound applied for about 15 min. The vial was held in an oven at 353 K for 24 h. A white precipitate formed that was isolated by centrifugation after cooling to room temperature. The solid was washed three times with DMF to remove unreacted precursors and six times with acetone to remove DMF. The resultant powder was dried in air at room temperature. The UiO-67 was found to have a Brunauer−Emmett−Teller (BET) surface area 2230 m2/g. Synthesis of Zr-NU-1000 and Hf-NU-1000. A sample consisting of 97 mg of ZrOCl2·8H2O (0.30 mmol) for Zr-NU-1000 or 123 mg of HfOCl2·8H2O (0.30 mmol) for Hf-NU-1000 and 2700 mg (22 mmol) of benzoic acid was mixed in 8 mL of DMF (in a 6 dram vial) and ultrasonically dissolved. The clear solution was incubated in an oven at 353 K for 1 h. After cooling to room temperature, 40 mg (0.06 mmol) B
DOI: 10.1021/acsami.7b03858 ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX
Forum Article
ACS Applied Materials & Interfaces Similar experiments have been reported for the zeolite- and MgOsupported rhodium complex,45,47,48 but not at exactly the same conditions used for the experiments reported here.
■
thermochemical quantities at 298.15 K. All frequencies below 50 cm−1 were replaced by 50 cm−1 when vibrational partition functions were computed. The geometry optimizations and vibrational frequency analysis were carried out for the samples in the gas phase as well as in n-pentane (the vibrational frequencies obtained from npentane-optimized structures are reported in the Supporting Information).54 The SMD solvation model was used to account for solvation effects for the calculations carried out with the samples in npentane.54
MODELS AND COMPUTATIONAL METHODS
Cluster Models. The UiO-67, NU-1000, and DAY zeolite were modeled as finite clusters obtained from previous work (Figure 1).37,38,41 A linker-deficient UiO-67 model was constructed by
■
RESULTS
Hydroxyl Groups on UiO-67, Zr-NU-1000, and Hf-NU1000. The reactive surface sites on UiO-67 nodes are defect sites (missing linker sites), and the number of such sites determined by TGA is 1.2/12 linkers. The surface sites on NU1000 nodes are at structural vacancies, with 4.0/12 per node. The hydroxyl groups on the missing linker sites of UiO-67 and Zr-NU-1000 have been inferred to be hydrogen-bonded OH/OH2 groups, with characteristic IR bands at 2745 cm−1 (Figure 2 and Figure S1).29,37,41,55 The sharp band at 3674 cm−1 observed for UiO-67 and Zr-NU-1000, with a weak shoulder for Zr-NU-1000 at 3653 cm−1, is inferred to result
Figure 1. Cluster model used for (a) UiO-67, (b) NU-1000, and (c) DAY zeolite. The zeolite atoms shown in the ball and stick representation were allowed to relax fully during the optimization process. Color code: oxygen atoms, red; carbon, gray; hydrogen, white; zirconium, cyan; aluminum, pink (in DAY-zeolite); silicon, teal. truncation to a single inorganic building unit, followed by replacement of 11 organic linkers with acetate (CH3CO2−) groups. The defect site was passivated with a hydration model previously proposed by Planas et al.37 The resultant UiO-67 cluster model contains 100 atoms. For NU-1000, the cluster model was constructed in a similar manner, but owing to the native structure of NU-1000 no linker defects were created. The resultant NU-1000 cluster model contains 94 atoms, with the organic linkers again truncated to acetate (CH3CO2−) groups. Moreover, for NU-1000, the mix-S proton topology of the Zr6-based node reported by Planas et al. was employed.37 The same protocol to generate the acetate model employed in previous investigations30,49 was followed. For DAY zeolite (SiO2/Al2O3 atomic ratio = 30), we used a 33T cluster model (T = Si, Al) that consists of 13 interconnected four-ring systems that face the supercage of the zeolite. The zeolite rings that coordinate the supported Rh atom were allowed to relax, with the remaining atoms held fixed, as shown in Figure S8. For more details, see the Supporting Information. In the representation of the rhodium-decorated clusters, a proton from the support was replaced with L2Rh(I) (L = CO or C2H4), essentially as shown in Scheme 1.
Scheme 1. Simplified Representation of the Reaction of Rh(acac)(C2H4)2 with UiO-67 Node Surfaces.56
Methodology. All local minima and transition-state structures were optimized using the M06-L50 density functional as implemented in Gaussian 09.51 For the UiO-67 and NU-1000 clusters, the def2-SVP basis set for C, H, and O atoms, and the def2-TZVPP52,53 basis set with associated effective core potentials for Rh and Zr atoms, were employed. In all cases, the positions of all atoms were optimized, except for the methyl groups of the acetate linkers. The geometry optimization for DAY zeolite was performed at the same level of theory as described above for the MOFs, with the def2-TZVPP52,53 basis sets for Si and Al. All geometry optimizations were carried out without symmetry constraints. Energy and geometry were computed with a “tight” convergence criterion; ultrafine grids were used. The natures of all stationary points were verified by analytical computation of vibrational frequencies, which were also used to compute
Figure 2. IR spectra characterizing (A) OH region and (B) hydrogenbonded H2O and OH region characterizing bare UiO-67 (black) and the samples formed by chemisorption of Rh(C2H4)2(acac) (red). C
DOI: 10.1021/acsami.7b03858 ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX
Forum Article
ACS Applied Materials & Interfaces
Table 1. Selected EXAFS Parameters and DFT Bond Lengths Characterizing Rh(C2H4)2+ Supported on UiO-67 Nodesa experimental N
shell Rh−Rh Rh−Zr Zr−OH Rh−OH Rh−μ3-OH/μ3-O Rh−CC2H4
EXAFS R (Å)
103 × Δσ2 (Å2)
calculated, DFT optimized ΔE0 (eV)
b
b
b
b
2.0 N/A 2.2 2.2 4.0
3.65 N/A 2.04 3.32 2.22
3.78 N/A −0.70 −0.90 1.24
3.70 N/A −5.05 −2.40 4.04
gas-phase R (Å)
n-pentane R (Å)
3.87 2.15 2.13 3.07/4.13 2.09
3.89 2.15 2.13 3.08/4.13 2.09
Δ[theory − exp] 0.22 (0.24)c 0.09 −0.28 (−0.29)c,d −0.13
Notation: OH atoms of terminal hydroxo groups on Zr6 node; CC2H4, ethylene carbon; μ3-O(H) are the bridging μ3-O or μ3-OH group on Zr6 node; N, coordination number; R, distance between absorber and backscatterer atoms; Δσ2, disorder term; ΔE0, inner potential correction. Estimated EXAFS error bounds: N, ± 20%; R, ± 0.02 Å; Δσ2, ± 20%; ΔE0, ± 20% (errors characterizing the Rh−Zr contribution are greater than these); fit range: 3.85 < k (wave vector) < 10.0 Å −1; 0.5 < R < 4 Å; goodness of fit value = 7.5. bContribution not detectable. cValues in parentheses correspond to DFT-optimized structures in n-pentane. dComparison to only the shorter theory distance is made. a
information, because the bands characterizing the organic linkers in the range of 3000−3150 cm−1 are so intense that they would swamp the weak bands characterizing the ethylene ligands (see the Supporting Information for details). In Table 1, we present a comparison of EXAFS and DFT structure parameters characterizing Rh(I) supported on the nodes of UiO-67. The agreement is generally good, although the distances between the Rh atoms and terminal/bridging O or Zr atoms of the node are overestimated by 0.2−0.3 Å by DFT; these results are consistent with results reported in analogous combined DFT/EXAFS investigations.29,58,59 The coordination numbers are in agreement with our DFT modeled clusters, with Rh having a total coordination of ∼6 at distances