The metal–organic framework (MOF), NU-1000, and its metalated counterparts have found proof-of-concept application in heterogeneous catalysis and hydrogen storage among others. A vapor-phase technique, akin to atomic layer deposition (ALD), is used to selectively deposit divalent Cu ions on oxo, hydroxo-bridged hexa-zirconium(IV) nodes capped with terminal –OH and –OH2 ligands. The subsequent reaction with steam yields node-anchored, CuII-oxo, hydroxo clusters. We find that cluster installation via AIM (ALD in MOFs) is accompanied by an expansion of the MOF mesopore (channel) diameter. We investigated the behavior of the cluster-modified material, termed Cu-AIM-NU-1000, to heat treatment up to 325 °C at atmospheric pressure with a low flow of H2 into the reaction cell. The response under these conditions revealed two important results: (1) Above 200 °C, the initially installed few-metal-ion clusters reduce to neutral Cu atoms. The neutral atoms migrate from the nodes and aggregate into Cu nanoparticles. While the size of particles formed in the MOF interior is constrained by the width of mesopores (∼3 nm), the size of those formed on the exterior surface of the MOF can grow as large as ∼8 nm. (2) Reduction and release of Cu atoms from the MOFs nodes is accompanied by the dynamic structural transformation of NU-1000 as it reverts back to its original dimension following the release. These results show that while the MOF framework itself remains intact at 325 °C in an H2 atmosphere, the small, AIM-installed CuII-oxo, hydroxo clusters are stable with respect to reduction and conversion to metallic nanoparticles only up to ∼200 °C.
I. INTRODUCTION
Nanoparticles with well-defined shapes and structures hold important roles in many catalytic processes,1,2 for example, CO2 reduction3 and H2O2 production.4 Due to their high surface energy, nanoparticles usually agglomerate into larger particles (sintering) under reaction conditions, leading to the deterioration of catalytic performance over time.5 Many different strategies have therefore been developed to circumvent adverse sintering, such as using a catalyst support6 or coating the catalytic nanoparticles with a porous layer.7,8 With enhanced stability, the catalytic activity and the product selectivity of the resulting materials are usually maintained or even improved from adventitious support effects.9,10 However, controlling the size and distribution of supported and/or coated nanoparticles is non-trivial, which may change during post-deposition treatment conditions. Therefore, using a structurally well-defined porous support whereby the mobility of a nanoparticle is limited by the structural components of the support represents a promising route for the synthesis of size-uniform and sintering-resistant nanoparticles for applications in catalysis.
Metal–organic frameworks (MOFs) are a large class of crystalline microporous materials composed of inorganic nodes and organic linkers,11 which afford excellent platforms for encapsulating nanoparticles. MOFs have drawn attention in a myriad of different applications.12–17 By virtue of their reticular chemistry,18 the pore size of the MOFs can be easily tuned via ligand design. If the framework is sufficiently robust, the MOF pores can be used to define and limit the size of subsequently formed particles. MOF-encapsulated nanoparticle composites have been synthesized via two general approaches: (1) construction of a MOF around the pre-formed nanoparticles and (2) reduction of deposited metal ions in a MOF.19 In the latter approach, the presence of metal-anchoring sites and high-stability of the MOF are usually preferred for the formation of stable hybrid nanoparticles at MOF composites. High-stability Zr-MOFs are well-suited for this purpose,20 since their Zr-based nodes are excellent grafting sites for many nano-sized metal-oxide/sulfide clusters21–27 that have been used in various catalytic transformations.21,28–30 For the present study, we investigated on NU-1000, a high-porosity Zr-MOF,31–33 that has been successfully used in multiple applications such as catalysis,23,34 gas storage and separation,33 selective adsorption in solution,35 and drug delivery.36
NU-1000 is comprised of Zr6(μ3-O)4(μ3-OH)4(H2O)4(OH)4 nodes and tetratopic 1,3,6,8-(p-benzoate)pyrene (TBAPy4−) linkers.31,37,38 The zirconium oxide based MOF can be visualized as an extended network with trihexagonal tiling defining two dimensions and channel-like pores defining the third. For trihexagonal tiling, each hexagon is surrounded by three triangles and each triangle by three hexagons (also known as Kagome network).31,39 The pore diameters for the hexagonal and triangular channels are ∼3 nm and ∼1 nm, respectively. The MOF is reasonably thermally stable (∼350 to ∼500 °C, depending on conditions)40 and also stable under considerable mechanical stress41 and in non-coordinating aqueous acids and various other demanding chemical environments.42 The Cu atoms are added via AIM (ALD-like chemistry in MOFs, ALD—atomic layer deposition) using volatile inorganic and organometallic complexes.38 AIM is a selective process where the volatile metal complexes react with the nonstructural aqua and hydroxo ligands of the eight-connected, hexa-zirconium(IV)-oxy nodes of NU-1000. A second step entailing MOF dosing with steam serves to release (as volatile neutral species) the remaining ligands of the installed precursor complex, replacing them with aqua and/or charge-balancing oxo or hydroxo ligands. Thus, AIM bypasses complexities such as purification and activation that often accompany MOF elaboration via wet chemical methods.19,43 The metal ions deposited via the vapor-phase prefer to grow in the aperture connecting the triangular and hexagonal channels.44
In catalysis and energy related applications, the stability of the catalysts at high temperature under realistic reaction conditions is a pressing question. We studied the structural evolution of Cu functionalized NU-1000 under reducing H2 environment on heating up to 325 °C. An in situ x-ray absorption study is performed to investigate the change in the oxidation state of Cu bonded to the MOFs nodes with temperature and also acquire information about the structural changes of the Cu nanoparticles. Small angle x-ray scattering (SAXS) data are collected for the fresh MOF after post-synthesis metalation with Cu-AIM and also after the temperature ramp. The study provides insight into the dynamic nature of the Cu-AIM network under reducing environment at high temperature.
II. EXPERIMENTAL SECTION
The synthesis of NU-1000 was done as described by Wang et al.,45 and Cu loading on the MOFs proceeded via atomic layer deposition as detailed in our previous reports.46 The resulting Cu doped NU-1000 is denoted as Cu-AIM. The synthesis and characterization of the MOFs are discussed in detail in the supplementary material (Sec. S1).
The structural evolution of Cu-AIM was studied under reducing environment at high temperature via combined in situ anomalous small angle x-ray scattering (ASAXS) and x-ray absorption spectroscopy (XAS) experiments, which were performed at the Advanced Photon Source 12 BM-B station. The Cu-MOF catalyst samples were loaded into a quartz capillary microreactor of 1.5-mm diameter47 for the combined SAXS and XAS measurements. The catalyst samples were kept at the center of the microreactor using quartz glass wool. The Cu-AIM catalyst was heated sequentially within the temperature range 25 °C–325 °C, following a temperature ramp shown in Fig. 1. A reducing environment was maintained by flowing H2 (4%) in helium through the capillary cell at a flow rate of 10 sccm at atmospheric pressure. The samples were treated at every temperature for 30 min and characterized at room temperature after cooling under a flowing hydrogen/helium gas mixture in order to minimize the Debye–Waller factor contribution. Cu EXAFS spectra were taken in the fluorescence mode at the Cu K-edge (8.986 keV). The data were analyzed using the ATHENA software package.48
Temperature ramp for in situ XAS and EXAFS measurement. “A” denotes the time range for XANES measurements, and “B” denotes the time range for EXAFS measurements. The sample is cooled down to room temperature after every temperature step to collect the EXAFS data.
Temperature ramp for in situ XAS and EXAFS measurement. “A” denotes the time range for XANES measurements, and “B” denotes the time range for EXAFS measurements. The sample is cooled down to room temperature after every temperature step to collect the EXAFS data.
The SAXS data were collected in transmission mode before and after the heat ramp. The x-ray beam was scattered off by the sample at a Cu K-edge energy of 8.986 keV. The two-dimensional SAXS images from the samples were analyzed by taking the horizontal and vertical cuts. The collected data were analyzed by Irena package using Igor software.49 The technique is sensitive to detect any aggregate formation with precision down to a nm (for details of the x-ray measurements, see Secs. S2 and S3 of the supplementary material).
III. RESULTS AND DISCUSSION
A. Cu-AIM: Post-synthesis metalation of NU-1000
The SAXS data collected for fresh NU-1000 MOFs contain diffraction peaks corresponding to a pore diameter of 34.15 Å, as shown in Fig. 2(a) and Fig. S1. The diffraction peaks shift on metalation with Cu and expand to 35.70 Å. The increase in the hexagonal pore diameter of the MOFs by 1.55 Å for Cu-AIM reflects its dynamic nature (Table I). The observed results for the ring expansion show a reverse trend as compared with what has been predicted for Zn-AIM and Al-AIM by density functional theory (DFT) calculations.38 The ALD process is self-limiting, and the MOFs donot over-saturate if the vapor-phase metalation is continued over different time scales. As has been reported from a recent study, an average of 4.5 Cu atoms can be accommodated on each Zr6 node.50
(a) Ex situ SAXS measurement for NU-1000 (green), Cu-AIM (red), and Cu-AIM after the temperature ramp (blue). (b) Experimental (black) and simulated (red) data curve for Cu-AIM after the temperature ramp. (c) The number and volume distribution for the aggregated Cu nanoparticles derived from the simulated curve for Cu-AIM after the temperature ramp.
(a) Ex situ SAXS measurement for NU-1000 (green), Cu-AIM (red), and Cu-AIM after the temperature ramp (blue). (b) Experimental (black) and simulated (red) data curve for Cu-AIM after the temperature ramp. (c) The number and volume distribution for the aggregated Cu nanoparticles derived from the simulated curve for Cu-AIM after the temperature ramp.
Data analysis from SAXS. Possibly (100), (111), and (200); assuming (100) at q = 0.184, d = 34.15 Å and q = 0.176, d = 35.70 Å, with a displacement of ∼1.55 Å.
MOF . | . | MOF-Cu . | . |
---|---|---|---|
0.184 | 1 | 0.176 | 1 |
0.319 | 0.3045 | ||
0.368 | 2 | 0.352 | 2 |
0.38 |
MOF . | . | MOF-Cu . | . |
---|---|---|---|
0.184 | 1 | 0.176 | 1 |
0.319 | 0.3045 | ||
0.368 | 2 | 0.352 | 2 |
0.38 |
The oxidation state and composition of the Cu clusters at the nodes of the MOFs are derived from an in situ XAS measurement. The x-ray absorption near-edge structure (XANES) spectra collected at the Cu K-edge of 8.979 keV resemble characteristics of both bulk CuO and Cu(OH)2 species, as shown in Fig. 3(a). On performing a linear combination fitting (LCF) of the XANES spectra using bulk Cu standards shown in Fig. S3, we find that the Cu clusters are present in the Cu2+state with about 65% contribution from Cu(OH)2 and 25% from CuO, as shown in Fig. 3(b) (see Fig. S4a for the XANES data fitting by LCF). LCF is an efficient technique for finding the oxidation state of the small Cu clusters.51
(a) XANES spectra collected during temperature ramp shown in Fig. 1 at all temperatures, (b) linear combination fit (LCF) done on the XANES spectra using bulk standards shown in Fig. S1, the contribution from different standards shown, and (c) evolution in the average oxidation state of the clusters with temperature estimated from the LCF results. In (a), “blue dots” marks the edge location for XANES spectra with shift designating the change in oxidation state of the clusters. The bottom spectrum was obtained at 25 °C and the top spectrum at 325 °C.
(a) XANES spectra collected during temperature ramp shown in Fig. 1 at all temperatures, (b) linear combination fit (LCF) done on the XANES spectra using bulk standards shown in Fig. S1, the contribution from different standards shown, and (c) evolution in the average oxidation state of the clusters with temperature estimated from the LCF results. In (a), “blue dots” marks the edge location for XANES spectra with shift designating the change in oxidation state of the clusters. The bottom spectrum was obtained at 25 °C and the top spectrum at 325 °C.
For structural analysis of Cu-AIM, we use the EXAFS data, which at room temperature demonstrate peaks at 1.1 Å and 1.5 Å, as shown in Fig. 4. The O–H bond length in Cu(OH)2 is ∼1 Å,52 and the Cu–O bond lengths for Cu(OH)2 and CuO are ∼1.7–1.8 Å.53–55 Thus, the Cu clusters at the Zr6 nodes are present as a mixture of CuO and Cu(OH)2, as has also been deduced from XANES data. The Cu⋯O and Cu⋯OH bond could form at two possible ends: (1) between a Cu ion and the terminal = O and –OH ligands at the Zr6 nodes and (2) between neighboring installed copper ALD precursors that have subsequently intentionally been treated with steam in the ALD reactor. The results are reminiscent of what has been seen previously for supported Cu clusters of different sizes after exposure to humid air.56
EXAFS data collected in situ at each temperature step (i.e., step “B” from the temperature ramp in Fig. 1) during the measurement.
EXAFS data collected in situ at each temperature step (i.e., step “B” from the temperature ramp in Fig. 1) during the measurement.
B. Evolution of Cu-AIM under reducing reaction environment
During the temperature ramp (see Fig. 1) performed under H2, a gradual change was observed in the XANES spectra as Cu(OH)2 started decreasing at 175 °C and Cu2O appeared, as shown in Figs. 3(a) and 3(b). Above 200 °C, Cu was reduced primarily to metallic state with an average oxidation state of 0.5, as shown in Fig. 3(c) (see Figs. S4b–S4d for the XANES data fitting by LCF). The change in composition occurred as a sequential step with the formation of Cu2O as an intermediate product. The H2 flow into the capillary was kept at 10 sccm and the Cu atoms at the nodes of the MOFs had even more restricted interaction with H2, as the gas needs to migrate through the pores. As has been previously reported with a high flow rate of H2 in the reactor, a direct conversion from Cu2+ to Cu0 precedes, whereas with a lower flow rate, suboxides form as intermediate products.57
The EXAFS spectra show an increase in bond length from 2.2 Å to 2.4 Å, which is a strong indication of the formation of Cu–Cu bonds.55 At 225 °C and 275 °C, and as Cu starts to get reduced and aggregation onsets, a bimodal peak centered at the metallic Cu–Cu bond length is observed. The concomitant reduction of Cu to metallic state and the formation of metallic bond can suggest that the Cu atoms are losing contact with the nodes of the MOFs, becoming mobile on the interior or exterior surface of the MOFs, and are aggregating to form Cu nanoparticles. The individual peaks in the bimodal distribution lie at 2.2 Å and 2.4 Å and could arise from the Cu–Cu bonds between the Cu atoms that are attached to the Zr6 nodes and the ones that have detached from the MOFs. Powder x-ray diffraction (PXRD) measurements (see Figs. S5 and S6) showed that (a) the crystallinity of the metal–organic framework is retained after Cu(0) nanoparticle formation and (b) the expected wide-angle diffraction peaks for crystalline copper metal peaks are present.
Chen et al. have shown that sub-nanometer sized Pt clusters loaded within the copper(II)-based MOF, HKUST-1, display strong activity in reaction with dihydrogen, which dissociates on the cluster surface and spills over on the network to efficiently break the Cu–O bonds that connect the nodes of HKUST-1 and linkers.58 A similar mechanism can be proposed here where the dissociated H on the Cu cluster surface attacks the Cu⋯O and Cu⋯OH bonds and breaks them, which helps in the release of Cu atoms bonded to the MOFs as well. The Cu cluster’s surface being catalytically active dissociates H2 to H.59 H spills over the Cu surface and reacts with the oxy and hydroxy ligands to be released as H2O and, in the process, also weakens the bonds between the Cu clusters and MOFs. With the increase in temperature, the metallic Cu component rises, and at 325 °C, a single peak corresponding to the Cu–Cu bond length of 2.25 Å for metallic Cu persists.
Following the in situ measurement, the sample was measured using SAXS in the transmission mode to examine the change in structure and dimensions of Cu-AIM. Interestingly enough, we found that the Cu atoms developed into larger particles with ∼6 nm as the particle diameter, as shown in Figs. 2(b) and 2(c). Cu particles within the MOF hexagonal channel grow to ∼3 nm, which is the maximum pore diameter.50,60 Cu particles can also form on the exterior surface of the crystalline MOF. Here, sintering of particles with increasing temperature is not constrained by the channel width, and formation of particles larger than 3 nm is possible. A visible color change of the Cu-AIM sample had also been observed above 200 °C, which could be associated with the formation of larger Cu nanoparticles on the MOF exterior surface. Although not investigated here, it is conceivable that at the highest temperatures examined, a fraction of the channel-confined particles migrate to the termini of the channels and onto the external surface of the MOF crystallite, thereby contributing to the population of particles capable of sintering to form particles larger than ∼3 nm. The particle size distribution is relatively narrow, spanning between a minimum particle size of ∼3 nm and a maximum size of ∼10 nm. The peak at the particle size diameter of 6 nm could be formed due to the aggregation of four Cu nanoparticles of 3 nm diameter or a mixture of particles of 3 nm and 1.5 nm, where the latter could form within the MOF’s triangular pores.61 The reduced Cu0 particles are more mobile on the MOF interior and exterior surfaces than the small node-grafted Cu-oxo clusters. Most interestingly, at high temperature, the channel width reverts to ∼34 Å, i.e., its width before introduction and reduction of copper. This finding would be consistent with speculation about possible expulsion of metal nanoparticles from MOF channels and onto crystallite exteriors. Formation of a mixture of nanoparticles has previously been reported for Pd clusters in MIL-100(Al) MOF crystallites where comparatively small nanoparticles of ∼1–3 nm diameter were detected within the MOFs channels, whereas 6 nm and 8 nm particles formed on the MOF external surface.62
IV. CONCLUSION
The evolution of the mesoporous NU-1000 MOFs during metalation using ALD and subsequent hydrogenation at high temperature have been studied. The hexagonal pore diameter of the parent NU-1000 MOF increases by 1.55 Å after Cu ALD. A combination of in situ XAS and EXAFS reveals that at room temperature, the Cu atoms stay as a combination of CuO and Cu(OH)2. The =O and –OH binding to the Cu atoms could attach directly to the surface of the Cu clusters or form a bridge with the Zr6 nodes. The –OH terminal ligands set up the MOFs scaffold for selective adsorption of metal atoms and clusters and also set a self-limiting condition on metal loading.38 As Cu is reduced, one also sees a simultaneous metallic Cu–Cu bond formation. The results imply that H2 assists in breaking the bonds between Cu clusters and the nodes within the MOFs releasing the neutral Cu atoms, which are mobile within the MOF channels and on their surface and thus easily aggregate to form larger nanoparticles. The pore diameter of the MOFs restores exactly to the size before metalation. The results suggest that even though NU-1000 is stable under thermal treatment up to nearly 500 °C, the metal atoms doped by ALD get evolved above 200 °C under H2 environment. This sheds light on the limiting conditions for the metal AIM derived MOFs under high temperature treatment.
This technique could be utilized for the synthesis of size-controlled larger nanoparticles.63 However, further improvements are necessary if the metal AIM is to be used for catalysis in the reducing environment at high temperatures.
SUPPLEMENTARY MATERIAL
See the supplementary material for descriptions of materials synthesis, x-ray scattering experiments, and x-ray absorption experiments and figures showing plots of experimental data from SAXS, EXAFS, XANES, and PXRD measurements.
ACKNOWLEDGMENTS
The work at Northwestern University was supported by the Inorganometallic Catalyst Design Center, an EFRC funded by the DOE, Office of Basic Energy Sciences (Grant No. DE-SC0012702). The work at Argonne (A.H., B.Y., M.J.P., and S.V.) was supported by the US Department of Energy, BES Materials Sciences, under Contract No. DEAC02–06CH11357 with UChicago Argonne, LLC, operator of Argonne National Laboratory, and the work at the Advance Photon Source (S.L., beamline 12-BM) was supported by the US DOE, Scientific User Facilities, under Contract No. DEAC02-06CH11357. S.V. also acknowledges support from the European Union’s Horizon 2020 Research and Innovation Programme under Grant Agreement No. 810310, which corresponds to the J. Heyrovsky Chair project (“ERA Chair at J. Heyrovský Institute of Physical Chemistry AS CR – The institutional approach towards ERA”) during the finalization of the paper. The funders had no role in the preparation of the article.