Application and Limitations of Nanocasting in Metal–Organic

Article Cite This: Inorg. Chem. 2018, 57, 2782−2790

pubs.acs.org/IC

Application and Limitations of Nanocasting in Metal−Organic Frameworks Camille D. Malonzo,†,% Zhao Wang,†,% Jiaxin Duan,† Wenyang Zhao,† Thomas E. Webber,† Zhanyong Li,‡ In Soo Kim,§,∇ Anurag Kumar,∥ Aditya Bhan,∥ Ana E. Platero-Prats,⊥,○ Karena W. Chapman,⊥ Omar K. Farha,‡,# Joseph T. Hupp,‡ Alex B. F. Martinson,§ R. Lee Penn,† and Andreas Stein*,† †

Department of Chemistry, University of Minnesota, Minneapolis, Minnesota 55455, United States Department of Chemistry, Northwestern University, Evanston, Illinois 60208, United States § Materials Science Division, Argonne National Laboratory, Argonne, Illinois 60439, United States ∥ Department of Chemical Engineering and Materials Science, University of Minnesota, Minneapolis, Minnesota 55455, United States ⊥ X-ray Science Division, Advanced Photon Source, Argonne National Laboratory, Argonne, Illinois 60439, United States # Department of Chemistry, Faculty of Science, King Abdulaziz University, Jeddah, Saudi Arabia ‡

S Supporting Information *

ABSTRACT: Nanocasting can be a useful strategy to transfer the catalytic metal clusters in metal−organic frameworks (MOFs) to an all-inorganic support such as silica. The incorporation of silica in the MOF pores as a secondary support has the potential to extend the application of the highly tunable metal-based active sites in MOFs to high temperature catalysis. Here, we demonstrate the applicability of the nanocasting method to a range of MOFs that incorporate catalytically attractive hexazirconium, hexacerium, or pentanickel oxide-based clusters (UiO-66, (Ce)UiO-66, (Ce)UiO-67, (Ce)MOF-808, DUT-9, and In- and Ni-postmetalated NU-1000). We describe, in tutorial form, the challenges associated with nanocasting of MOFs that are related to their small pore size and to considerations of chemical and mechanical stability, and we provide approaches to overcome some of these challenges. Some of these nanocast materials feature the site-isolated clusters in a porous, thermally stable silica matrix, suitable for catalysis at high temperatures; in others, structural rearrangement of clusters or partial cluster aggregation occurs, but extensive aggregation can be mitigated by the silica skeleton introduced during nanocasting.

■

porous materials, where the final nanocast material is void of the original MOF template, including microporous titania11 and polyaniline12 templated from the MOF HKUST-1, and other porous oxides, carbon, and composites.13 While MOFs can be used as templates for other nanostructured materials, nanocasting of MOFs can also be done for a different purpose, particularly for stabilizing MOF-based catalysts. MOFs are of interest for catalysis because their structures incorporate high loadings of well-defined metal sites that can be catalytically active.14−16 The activity can also be tuned by the choice of metal during the synthesis and by postsynthetic metalation via atomic layer deposition, solutionphase deposition, or metal exchange.17−22 Catalysis with MOFs is, however, usually limited to low temperature reactions. At high temperatures and in oxidizing environments (typically >400 °C in air), degradation of the organic linkers occurs, along with the aggregation of the catalytic metal sites, which leads to loss of their catalytic activity. To address this, we developed an

INTRODUCTION Nanocasting is a method that allows the formation of nanostructured materials as direct or inverse replicas of a porous template. Combined with sol−gel techniques, it affords the synthesis of structurally tunable materials with tailorable compositions, including carbon, metal oxides, and other inorganic materials.1,2 Nanocasting involves the infiltration of a precursor solution into a template, followed by the conversion of the precursor to a solid within the template. Removal of the template by methods such as etching or calcination yields the final nanocast material. The nanocasting method has been developed extensively for the preparation of nanostructured materials from mesoporous templates.3−6 Metal−organic frameworks (MOFs) form an emerging class of porous materials that has rapidly gained interest for a variety of applications, including gas storage, separation, and catalysis.7−9 MOFs consist of metal-based nodes that are interconnected via organic linkers, forming a porous, crystalline network. They have tunable pore morphology, and pore sizes of up to 98 Å have been achieved.10 MOFs have been used as templates for nanocasting to make microporous and meso© 2018 American Chemical Society

Received: December 20, 2017 Published: February 20, 2018 2782

DOI: 10.1021/acs.inorgchem.7b03181 Inorg. Chem. 2018, 57, 2782−2790

Article

Inorganic Chemistry

Figure 1. Framework structure and the secondary building units of the MOFs used for nanocasting.

■

RESULTS AND DISCUSSION General Considerations. Although some large pore MOFs have been prepared, such as IRMOF-74-XI (pore size = 98 Å),10 most MOFs have pores within a few nanometers in size,7 built from simpler linkers such as benzene-1,4-dicarboxylate (e.g., UiO-66) 36 and benzene-1,3,5-tricarboxylate (e.g., HKUST-1).37 The small pore size in MOFs warrants the use of small precursors for casting to enable infiltration and to incorporate as much precursor as possible in the pores. For silica, tetramethyl orthosilicate (TMOS) is a good precursor because of its small size (∼8 Å). TMOS is also a liquid, which discounts the need for solvents that would reduce the maximum loading of the pores with silica. On the basis of the relative densities and molar masses of TMOS and silica and assuming that the MOF pores are completely filled with TMOS after infiltration, up to ∼18% of the total MOF pore volume could be occupied by silica when using TMOS as a precursor (see the Supporting Information). TMOS converts to silica by hydrolysis and condensation in the presence of water and an acid (or base) catalyst. Casting fluids typically consist of a solution of these three components. Methanol is also often added as a cosolvent to slow down TMOS hydrolysis during template infiltration. For the MOFs investigated here, the small pore size necessitates that the silica precursor stays monomeric during infiltration, since even short condensed precursor units could be too big to infiltrate the MOF pores. Thus, the presence of a catalyst in the casting fluid may not be ideal. This was observed previously when NU-1000, a MOF with 3 nm pores, was cast using a solution of TMOS, HCl(aq), and methanol, which did not introduce sufficient silica into the MOF pores to provide a stabilizing scaffold, even after multiple infiltration steps.23 As a result, after the linkers were removed at high temperatures, the hexazirconium oxide clusters of the MOF had aggregated. To keep TMOS monomeric during infiltration, the acid catalyst was removed from the precursor solution used for infiltration, and was introduced afterward via vapor-phase HCl treatment of the infiltrated MOF particles. This allowed a silica skeleton to form in the mesopores of NU-1000 and helped to prevent cluster aggregation. Successful incorporation of silica in the MOF pores can be evaluated by techniques such as infrared spectroscopy, gas physisorption analysis, and difference envelope density (DED) analysis38,39 Apart from the presence of silica, however, other

approach to stabilize the metal sites in MOFs by nanocasting with silica.23 Nanocasting introduces a thermally stable silica layer in the pores of the MOF that can serve as a secondary support for the metal sites, maintaining their site isolation and catalytic activity even after the linkers are lost at high temperatures. Thus, through nanocasting, the highly tunable and well-defined metal sites in the MOFs can be used for high temperature catalysis. We have successfully applied this nanocasting method to the mesoporous MOF, NU-1000,23 as well as to bimetalated NU1000 samples.24 In both cases, the presence of silica in the MOF pores enabled maintenance of the site isolation, accessibility, and catalytic activity of the metal sites after high temperature treatment that removed the organic linkers from the structure. The current work describes our nanocasting experiments with other MOF systems with different pore sizes and metal node compositions (Figure 1). Here, the MOFs UiO-66, (Ce)UiO-66, (Ce)UiO-67, (Ce)MOF-808, DUT-9, and postmetalated NU-1000 MOFs were nanocast with silica. NU-1000, UiO-66, (Ce)UiO-66, (Ce)UiO-67, and (Ce)MOF808 feature isostructural hexazirconium (Zr6) or hexacerium (Ce6) oxide nodes. The Zr6 node in UiO-66 is nominally coordinated with 12 carboxylate groups (i.e., 12-connected) from the organic linkers, whereas, in NU-1000, this cluster is nominally only 8-connected, and the rest of the coordination sites are occupied by −OH and −H2O groups (Figure 1). The Ce6 nodes in (Ce)UiO-66 and (Ce)UiO-67 are also nominally 12-connected, similar to the Zr6 nodes in UiO-66 and UiO-67, but only 6-connected in (Ce)MOF-808.25,26 In DUT-9, the Ni5 nodes are 6-connected, with the four remaining coordination sites occupied by solvent molecules, e.g., DMF. This work demonstrates that the MOF nanocasting method can provide access to other catalytically interesting metal site compositions (e.g., Ni,27−31 In,32−34 and Ce,35 of potential interest for the high temperature catalytic manipulation of C−H and C−C bonds). Additionally, it is shown here that the method is applicable to both mesoporous (e.g., DUT-9) and microporous MOFs (e.g., UiO-66), but among the latter, larger micropores (e.g., (Ce)MOF-808) provide a more stabilizing nanocast skeleton. Considerations regarding the compatibility between the nanocasting conditions and MOF stability and the associated method optimizations are also discussed. 2783

DOI: 10.1021/acs.inorgchem.7b03181 Inorg. Chem. 2018, 57, 2782−2790

Article

Inorganic Chemistry Table 1. Compositional and Structural Details of MOFs Used for Nanocasting MOF UiO-66 (Ce)UiO-66 (Ce)UiO-67 (Ce)MOF-808 DUT-9

linkera

node 12+

[Zr6(μ3-O3)4(μ3-OH)4] [Ce6(μ3-O)4(μ3-OH)4]12+ [Ce6(μ3-O)4(μ3-OH)4]12+ [Ce6(μ3-O)4(μ3-OH)4]12+ [Ni5(μ3-O)2]6+

2−

bdc bdc2− bpdc2− btc3− btb3−

surface areab,c (m2 g−1)

pore size (Å) 11, 8 11, 8d 23, 11d 18 (ref 26); 22 (this work) 25, 13

1580 1282 2064 1725

(ref 36); 1357 (this work) (ref 25); 1569 (this work) (this work) (ref 26); 1333 (this work)

pore volumeb (cm3 g−1) 0.70 0.50 0.95 0.62 2.18

(this work) (ref 25); 0.76 (this work) (this work) (ref 26); 1.03 (this work) (ref 46); 1.21 (this work)

a

bdc = benzene-1,4-dicarboxylate, bpdc = biphenyl-4,4′-dicarboxylate, btc = benzene-1,3,5-tricarboxylic acid, btb = benzene-1,3,5-tribenzoate. bN2 sorption isotherms of the MOFs used for the evaluation of BET surface area and pore volumes are shown in Figure S3. c(Ce)UiO-67 surface area not measured in ref 25 due to decomposition at reduced pressure. DUT-9 does not have a linear range for evaluation of BET surface area.46 BET surface areas in this work were evaluated according to the guidelines in ref 47. BET range: UiO-66 up to p/p0 = 0.0746; (Ce)UiO-66 up to p/p0 = 0.0739; (Ce)UiO-67 up to p/p0 = 0.125. dIndicated values are the pore sizes of the corresponding Zr MOFs: UiO-66 and UiO-67.36

attempted to use the silica nanocast material derived from NU-1000 for high temperature alkane dehydrogenation (610 °C), for which Lewis acidic oxides like Ga2O3 have been shown as active catalysts.45 However, the apparent rates of cyclohexane consumption and benzene/cyclohexene formation were small with this Zr6O8 cluster-based catalyst (Figure S1). Such low catalytic activity can be addressed by adding another catalytic metal to the clusters or by changing the clusters themselves. To test the applicability of the nanocasting method to MOFs with smaller pore size and to access other catalytically active metal site compositions, we applied nanocasting to five other MOFs: UiO-66, (Ce)UiO-66, (Ce)UiO-67, (Ce)MOF-808, and DUT-9. Table 1 summarizes some important compositional and structural information on each MOF. The MOFs used for nanocasting in this work were prepared according to reported procedures.25,26,36,46 The XRD patterns of the synthesized samples are consistent with simulated patterns from the published cif files of each MOF, as shown in Figure 2. For (Ce)UiO-66 and (Ce)UiO-67, the patterns are shifted to lower angles compared to their Zr counterparts, indicating a larger d-spacing in these Ce MOFs. The SEM images and N2 sorption isotherms (Figures S2, S3, and S4) are also consistent with literature data, confirming the successful synthesis and activation of each MOF. Similar nanocasting conditions as we previously reported for NU-100023 were initially applied to the MOFs in this study. The casting fluid consisted of a TMOS mixture with water and methanol (97% by volume TMOS). This was followed by vapor phase HCl treatment, condensation at 60 °C, and, last, linker removal by calcination. The XRD patterns of the nanocast samples are shown in Figure 3. Among the samples, only the nanocast UiO-66 has the broad features associated with amorphous silica in its XRD pattern, while the others exhibit sharp peaks indicative of aggregated clusters. TEM images of the UiO-66 nanocast sample (Figure 4) also do not show large ZrO2 aggregates, indicating that the clusters in this sample likely exist as site-isolated, individual clusters, similar to our previous study with NU-1000. We found that the clusters are also accessible and that the nanocast material was mesoporous by pyridine sorption and N2 sorption analyses, respectively (Figure S5). This result of nanocasting UiO-66 shows that the method reported for mesoporous NU1000 can be applied to MOFs with micropores. Even though diffusion of TMOS through the small windows in UiO-66 may be slow, the infiltration time of 24 h used here was sufficient to fill the pores and provide a stabilizing skeleton throughout the UiO-66 particles. Without TMOS infiltration, extensive aggregates of ZrO2 would be generated during calcination of UiO-66 (Figure S6).

criteria must be met for nanocasting to be considered successful, especially for the purpose of stabilizing the metal clusters in MOFs for catalytic applications. One is that the site isolation of the clusters must be maintained in the nanocast material after removal of the organic linkers, in the same way they are isolated in the original MOF. This ensures that the high loadings of catalytically active sites are maintained. Techniques such as X-ray diffraction (XRD), selected area electron diffraction (SAED), and pair distribution function (PDF) analysis40 can be used to differentiate between aggregated and site-isolated metal sites in the nanocast samples. The second criterion is accessibility. The metal sites need to remain accessible to reactant molecules and not be fully coated with silica in the nanocast material. This is facilitated by removal of the organic linkers in the structure where the original points of attachment to the metal clusters are presumably open after linker removal. Four equivalents of methanol are also released from the hydrolysis of each TMOS unit, and can introduce additional porosity in the silica matrix formed in the nanocast samples. Gas physisorption can be used to determine the porosity of the nanocast material. However, it is also important to test the accessibility of the actual clusters using probe molecules that can be chemisorbed onto these metal sites, for example, pyridine, ammonia, or hydrogen. The last criterion is catalytic activity, such as selectivity or rate of product formation for the target reaction. The catalytic activity of the metal sites must be maintained as they are transferred from the MOF to the nanocast silica matrix. Comparison of activity with the original MOF can be done using low temperature model reactions. Notably, this activity is also reflective of the site isolation and accessibility of the metal sites in the nanocast sample. Among the above methods, XRD is a convenient initial check to rule out the presence of large aggregated clusters in the material. When the metal sites in MOFs aggregate into larger oxide crystallitesthe XRD patterns of the samples consist of sharp peaks. In the case that the aggregated metal oxides are amorphous, transmission electron microscopy (TEM) can be used to see the aggregates, particularly for metal oxide compositions with high atomic number (Z) contrast with silicon.41 Nanocasting Different MOFs. Our previous work on nanocasting focused on NU-1000, which consists of Zr6O8 clusters as nodes. The clusters are found to be active for low temperature Lewis acid-catalyzed reactions.42−44 We showed that the catalytic activity of these clusters (approximate rate of fructose formation, fructose yield at a given glucose conversion) is practically retained after nanocasting using glucose isomerization as a model Lewis acid-catalyzed reaction.23 We 2784

DOI: 10.1021/acs.inorgchem.7b03181 Inorg. Chem. 2018, 57, 2782−2790

Article

Inorganic Chemistry

Figure 3. XRD patterns of nanocast MOFs prepared by nanocasting utilizing an acid catalyst. The XRD pattern was taken after the final calcination step at 500 °C in air. Average grain sizes estimated from line broadening using the Scherrer equation are ∼17 nm of CeO2 for nanocast (Ce)UiO-66, ∼7 nm of CeO2 for nanocast (Ce)UiO-67, ∼5 nm of CeO2 for nanocast (Ce)MOF-808, and ∼73 nm of NiO for nanocast DUT-9.

Figure 2. XRD patterns of the synthesized MOFs, showing patterns consistent with published data of each MOF. Compared to UiO-66 (d = 1.18 nm), the highest d-spacing for (Ce)UiO-66 is 1.23 nm. Similarly, for (Ce)UiO-67, the highest d-spacing is 1.58 nm, compared to 1.54 nm for UiO-67. COD refers to the Crystallography Open Database (www.crystallography.net) and CCDC to the Cambridge Crystallographic Data Center (www.ccdc.cam.ac.uk).

Unlike UiO-66, nanocasting DUT-9 resulted in aggregated NiO on silica (∼73 nm grain size NiO, calculated from XRD peak broadening using the Scherrer equation). Several conditions in the nanocasting method are incompatible with DUT-9 and could have resulted in the aggregation of the pentanickel oxide nodes. We found that DUT-9 is unstable toward capillary forces. During its synthesis, DUT-9 cannot be dried from conventional solvents used in activation of MOFs. Drying DUT-9 from acetone (surface tension, γ, of 23 dyn cm−1), for instance, leads to structural collapse of the MOF (Figure S7). Instead, DUT-9 must be supercritically dried (γ of supercritical CO2 is 0.6 dyn cm−1).46 This instability toward capillary forces is also seen during infiltration where TMOS (γ = 23 dyn cm−1) and other solvents cause structural collapse in the MOF (Figure S7). One way to avoid this collapse during the infiltration step is through solvent exchange with the casting fluid. After DUT-9 is

Figure 4. Low and high magnification TEM images of nanocast UiO66 prepared by nanocasting utilizing an acid catalyst. No large aggregates of ZrO2 can be seen from the high magnification image.

synthesized and washed with DMF, the DMF can be replaced with the casting fluid for exchange. However, modifying the infiltration step in this manner was not sufficient to prevent the 2785

DOI: 10.1021/acs.inorgchem.7b03181 Inorg. Chem. 2018, 57, 2782−2790

Article

Inorganic Chemistry aggregation of the clusters in the nanocast DUT-9 (Figure S8). Apart from capillary forces, DUT-9 is also incompatible with the HCl catalyst, as shown by the loss of the characteristic XRD pattern of the MOF after exposure to the acid (Figure S9). Therefore, to cast DUT-9, the procedure should be carried out by solvent exchange and without using an acid catalyst. A modified casting method was applied, consisting of solvent exchange with a solution of TMOS with water and methanol (70% by volume TMOS). Increasing the proportion of water in the mixture was necessary to achieve hydrolysis and condensation in a reasonable amount of time. After 24 h, the casting fluid was decanted and the solvent-exchanged particles were washed once with methanol. After gelation and linker removal, the XRD pattern of the sample shows only very broad, low intensity features that can be related to NiO, indicating that cluster aggregation was substantially reduced compared to the sample prepared by nanocasting using an acid catalyst (Figure 5). (It should be noted that multiple washing steps with

Figure 5. XRD pattern of nanocast DUT-9 prepared by nanocasting without an acid catalyst. The silica precursor mixture was introduced by solvent exchange instead of infiltration. The XRD pattern was taken after the final calcination step at 500 °C in air.

methanol after TMOS infiltration resulted in more extensive cluster aggregation, likely due to partial loss of the TMOS precursor from the pores.) The grain size estimated from simulations of the XRD pattern of NiO is less than 2 nm (Figure S10). The TEM images (Figure 6a,b) also do not show large NiO domains, indicating that the modified casting procedure was successful at maintaining small nickel oxide clusters within the silica matrix support. A PDF analysis of local structure revealed the presence of single-layer Ni(OH)2 slabs with a lateral size of approximately 2.8 nm (Figure 7). This observation suggests that, while significant, extended cluster aggregation in three dimensions was prevented by the infiltrated silica skeleton, adjacent [Ni5(μ3-O2]6+ nodes were able to react with each other in two dimensions to form the Ni(OH)2 slabs, possibly between layers of silica. During the calcination process, the Ni(OH)2 slabs could then further react to form the small (