Background

Metal-organic frameworks (MOFs) constitute an important class of porous crystalline materials. Modular synthetic routes enable rational design of MOFs with precise control over pore size, connectivity, and functional groups[1]. MOFs have widespread applications, including adsorption and separation[2–4], catalysis[5–8], and sensing[9, 10]. In particular, MOFs are highly preferred for catalytic applications because of their large surface area and pore volume, tunable pore size and shape, and flexibility for diverse functionalization. As such, MOF-based catalysis has recently emerged as a burgeoning subfield in heterogeneous catalysis[2–4]. The active metal sites and/or reactive organic groups that constitute the frameworks of MOFs endow the catalytic functions of the MOFs. In addition, catalytic metal nanoparticles (NPs) incorporated into the cavities of MOFs can also provide catalytic function. The incorporation of metal NPs into porous supports, such as zeolites, mesoporous materials, and MOF, can be achieved by various methods, including solution impregnation[11–13], chemical vapor deposition[14–16], and solid grinding[17]. Although these methods have long been useful for generating heterogeneous catalysts, the high-temperature heating steps involved in these methods inevitably yield a wide distribution of particle sizes[18, 19]. Another prominent route to the supported catalysts is by a colloidal deposition method, where pre-synthesized, highly monodisperse colloidal NPs are deposited onto the supports[20, 21]. However, for full catalytic utilization of NP surfaces, this method requires judicious thermal or chemical treatments that can remove surface capping polymers or surfactants that stabilize colloidal NPs.

Recently, the conversion of the MOF into metal oxide nanoparticles has been used as a new strategy for preparing nanoscale functional entities. In this method, the secondary building units of MOF that are mostly composed of metal oxide clusters in angstrom scale were transformed into metal oxide nanomaterials by thermal treatments. For instance, Xu et al. reported the synthesis of Co3O4 nanoparticles converted from cobalt oxide subunits in a cobalt-based MOF, Co3(NDC)3(DMF)4 (NDC = 2,6-naphthalene-dicarboxylate; DMF = N,N’-dimethyl formamide) by pyrolysis in air[22]. The resulting cobalt oxide nanoparticles were utilized for an electrode material for lithium-ion batteries. More recently, MOF-5 was treated at high temperature (over 600°C) under various atmospheric conditions to produce ZnO nanoparticles and ZnO@C hybrid composites[23]. Thermal treatment of MOF-5 under nitrogen generated ZnO@C, whereas MOF-5 heated under air yielded a pure ZnO nanoparticle, indicating the combustion of organic ligands.

As an alternative scheme to the conversion of MOFs into new nanomaterials, one of the authors previously reported the redox couple-driven method where redox potential differences between foreign metal ions and redox-active MOFs (ra-MOFs) lead to the spontaneous formation of monodisperse metal NPs[24–26]. Significantly, the redox potential-driven method can readily yield highly monodisperse, surface-naked metal NPs at room temperature without the help of any reducing agent or surface capping molecules, which would be advantageous for catalytic applications. The application of this method for heterogeneous catalytic reactions, however, has not yet been exploited. In this study, we prepared Ni-based ra-MOF-supported Pd nanoparticles (Pd@ra-MOF) via the redox couple-driven method (Figure1). We found that the Pd@ra-MOF transformed into PdO x -NiO y /C nanocatalyst during gas phase CO oxidation reaction. The resulting metal oxide nanocomposite showed high and sustainable catalytic activity toward CO oxidation.

Figure 1
figure 1

Schematic representation for in situ formation of catalytically active species during CO oxidation reaction. The reaction involves thermal transformation of crystalline MOF to nanoparticles in the solid state.

Full size image

Methods

Materials and characterization of samples

All chemicals and solvents used in the syntheses were of reagent grade and they were used without further purification. Infrared spectra were recorded with a Thermo Fisher Scientific Nicolet 6700 FT-IR spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). Elemental analyses were performed at the UNIST Central Research Facilities Center in Ulsan National Institute of Science and Technology. Palladium content on ra-MOF was analyzed with an inductively coupled plasma optical emission spectrometer (Varian 720-ES, Varian Inc., Palo Alto, CA, USA). X-ray photoelectron spectroscopy was performed using a Thermo Scientific K-Alpha XPS spectrometer. X-ray diffraction (XRD) patterns were recorded with a Rigaku D/MAZX 2500 V PC diffractometer (Rigaku Corporation, Tokyo, Japan) at 40 kV and 100 mA with Cu-Kα radiations (1.54059 Å) with a scan speed of 2°/min and a step size of 0.02° in 2θ at room temperature. JEOL JEM-2100 F transmission electron microscope (JEOL Ltd., Akishima, Tokyo, Japan) and an Oxford INCA EDS unit (Oxford Instruments, Abingdon, Oxfordshire, UK) were used to examine the morphology of nanostructured catalysts before and after catalytic reactions. Thermogravimetric analysis (TGA) was performed under N2(g) atmosphere at a scan rate of 5°C/min using Q50 from TA instruments (New Castle, DE, USA). N2 sorption isotherms of ra-MOF and Pd@ra-MOF were obtained by BELSORP-max (BEL Japan Inc., Osaka, Japan) at 77 K to estimate the specific Brunauer-Emmett-Teller (BET) surface areas.

Preparation of ra-MOF

[Ni(C10H26N6)](ClO4)2 (C10H26N6 = LCH3) was prepared according to the preparation conditions in a previous report[27]. [NiLCH3(bpdc) (ra-MOF; bpdc2− = 4,4′-biphenyldicarboxylate) was synthesized by the modified method from the previous reports[24]. Synthetic detail is as follows: (NiLCH3)(ClO4)2 (0.80 g, 1.64 mmol) was dissolved in water/pyridine (48 mL, 2:1 v/v), and an aqueous solution (16 mL) of Na2bpdc (0.56 g, 2.10 mmol) was added. The solution was stirred over 20 min at room temperature, and 150 mL of methanol was added to it. The mixture was stirred for 6 h, forming pale purple microcrystalline precipitates which were isolated by filtration, washed with methanol, and dried in air. The as-prepared metal-organic framework was desolvated at 120°C under vacuum for 3 h resulting in a purple color. The yield is 57%.

Preparation of the Pd@ra-MOF

The desolvated solid (0.43 g, 0.81 mmol) was immersed in 3.12 × 10−2 M acetonitrile solution (82 mL) of Pd(NO3)2· × H2O at room temperature and hand-shaken for 10 min, which resulted in a light brown solid. The resulting light brown powder was isolated by filtration, washed with acetonitrile, and dried in air.

Catalytic activity test

The catalytic tests for CO oxidation were performed in a fixed bed reactor at atmospheric pressure, containing 0.06 g of catalyst samples. A feed mixture, prepared using mass flow controllers (MKS Instruments, Inc., Wilmington, MA, USA), contained 3.0% CO and 8.5% O2 and was balanced with He. The total flow rate of the feed mixture was 52 mL min−1, and the gas hourly space velocity was 1,7316 h−1. The effluent gas stream from the reactor was analyzed online by the thermal conductivity detector of parallel gas chromatography (Younglin Instrument Co., Ltd, Anyang, Korea) with a Carboxen 1000 column. In order to determine the conversion, the products were collected during 40 min of steady-state operation at each temperature. The empty reactor (without catalyst) showed no activity under identical conditions.

Results and discussion

[NiLCH3(bpdc), which is composed of Ni(II) hexaaza macrocycle and carboxylate and has one-dimensional channels with a 7.3-Å pore opening (Additional file 1:Figure S1), was selected as the ra-MOF[24]. It has previously been shown that the MOF constructed by Ni(II) macrocyclic complexes and multidentate carboxylate ligands exhibits redox-active properties that originate from the six-coordinated Ni(II) sites[24–26]. The ra-MOF, [NiLCH3(bpdc) (bpdc = 4,4′-biphenyldicarboxylate), was prepared by the self-assembly of [NiLCH3(ClO4)2 and Na2bpdc in a H2O/pyridine mixture, yielding [(NiLCH3)3(bpdc)3·2pyridine·6H2O, and it was subsequently activated at 120°C under vacuum for 3 h. Pd NPs were spontaneously formed in the ra-MOF solid - without the use of any reducing agent - by soaking the ra-MOF in an acetonitrile solution of Pd(NO3)2 at room temperature for 10 min. Because the oxidation potential of Ni(II) to Ni(III) in the monomacrocyclic complexes ranges from 0.90 to 0.93 V[28], the redox reaction of a ra-MOF possessing Ni(II) macrocyclic complexes with Pd ions leads to the simultaneous oxidation of Ni(II) to Ni(III) and the reduction of Pd(II) ions to the metallic Pd NPs.

Figure2 (a) and (b) display the XRD patterns of the as-synthesized ra-MOF and Pd@ra-MOF. The XRD peak intensities of Pd@ra-MOF decreased compared to those of the as-synthesized ra-MOF, which are presumably due to the partial occupation of Pd NPs in the pores of the ra-MOF as well as the partial destruction of the crystalline framework of the ra-MOF. This change was also observed with N2 adsorption measurements. The specific BET surface areas calculated from N2 adsorption decreased from 462 to 14 m2g−1 upon the formation of Pd NPs in the ra-MOF (Additional file 1:Figure S2). No characteristic peaks of Pd metal appeared in the XRD pattern of Pd@ra-MOF, which indicated that small-sized Pd NPs were well dispersed throughout the ra- MOF support. The ra-MOF and Pd@ra-MOF were further characterized by transmission electron microscopy (TEM) (Figure2 (c) and (d)) and energy-dispersive X-ray spectroscopy (EDS) (Additional file 1:Figure S3). The TEM image clearly showed that uniform small Pd NPs that were 1.8 ± 0.3 nm in diameter were well dispersed on the ra-MOF. A high-resolution TEM image (Figure2 (d), inset) showed the lattice fringes of Pd, indicating the single crystalline nature of Pd NPs. As determined by inductively coupled plasma analysis, the loading of Pd metal in the ra-MOF was calculated to be 3.6 wt.%.

Figure 2
figure 2

XRD patterns and TEM images of (a, c) ra -MOF, and (b, d) Pd@ ra -MOF, respectively.

Full size image

The catalytic properties of Pd@ra-MOF were explored using CO oxidation as a probe reaction, and the as-prepared ra-MOF without Pd NPs was also tested for comparison. The catalytic oxidation of CO to CO2 has long been a benchmark reaction in heterogeneous catalysis; it is a continuous subject of fundamental mechanistic studies as well as of practical importance in many industrial processes, including the reduction of CO in automobile exhaust gases and the selective oxidation of fuel streams for polymer electrolyte fuel cells[29–31]. CO oxidation was performed in a fixed bed flow reactor at elevated temperatures under atmospheric pressure in a mixed gas composed of 3.0% CO, 8.5%O2, and 88.5% He. As shown in Figure3, the as-synthesized ra-MOF exhibited poor catalytic activity, which is evidenced by a very high CO conversion temperature, with T50 and T100 (temperatures for 50% and 100% conversion, respectively) being 364°C and 500°C, respectively. Successive four runs showed similar catalytic behaviors. In contrast, the loading of Pd NPs into ra-MOF resulted in markedly lower conversion temperatures of T50 and T100 at 222°C and 250°C, respectively, indicating the Pd@ra- MOF had enhanced catalytic activity over the ra-MOF for CO oxidation. The catalytic abilities of Pd@ra-MOF in the first run, however, are not superior to other Pd catalysts, and this probably comes from the sharp decrease of surface area after formation of larger-sized Pd NPs than the channel size of ra-MOF, which lessens the accessibility of CO and O2 to Pd NPs embedded in ra-MOF. The CO oxidation activities over Pd@ra-MOF were repeatedly tested, and the Pd@ra-MOF exhibited further lowered conversion temperatures in the second run, with T50 and T100 dropping down to 125°C and 150°C, respectively, thus revealing significantly enhanced CO oxidation activity as compared to the first run (Figure3). This implied the formation of new species from the ra-MOF and Pd NPs during the first catalytic reaction. As shown in Figure3, three more reaction runs over the Pd@ra-MOF produced the same results as the second run, demonstrating its long-term stability.

Figure 3
figure 3

Catalytic activities and conversion temperatures. (a) Catalytic performance for CO oxidation of the as-prepared ra-MOF and Pd@ra-MOF. (b) Comparison of T100 and T50 for ra-MOF and Pd@ra-MOF according to the number of cycles.

Full size image

To examine the transformation of the Pd@ra-MOF during a CO oxidation reaction, the XRD patterns and TEM images of the as-synthesized ra-MOF and Pd@ra-MOF after catalytic reactions were obtained and compared (Figure4). The XRD pattern of the ra-MOF after the fifth run of the catalytic reactions (Figure4 (a)) was completely different from that of the as-prepared ra-MOF (Figure2 (a)). The diffraction peaks were assigned as cubic NiO of the Fm3^-m space group (JCPDS 47–1049). It appears that during CO oxidation, building blocks of ra-MOF (hexaaza macrocyclic ligands, Ni ions, and bpdc2− ligands) were transformed into different species under oxidative reaction conditions. Thermally unstable hexaaza macrocyclic ligands, LCH3, coordinating Ni ions in the ra-MOF could be easily decomposed and removed during a high-temperature catalytic reaction. This was substantiated by the X-ray photoelectron spectroscopy (XPS) of the ra-MOF, whose nitrogen content approached to nearly zero after the catalytic reaction (Additional file 1:Figure S4). Ni ions were oxidized to form NiO and Ni2O3 species under an oxidative reaction environment, which was proven by Ni 2p3/2 XPS spectra that indicated the formation of Ni2O3 (855.7 eV) as well as NiO (853.9 eV) (Figure5 (b)) from N-coordinating Ni(II) (854.9 eV) (Figure5 (a)). Lastly, the organic ligand bpdc2− was decomposed and converted to a carbogenic support by heating up to 500°C, as evidenced by a TGA trace (Additional file 1:Figure S5) and infrared (IR) spectra before and after catalytic reactions (Additional file 1:Figure S6). Overall, the resultant material after the catalytic reaction was a NiO x /C nanocomposite. The TEM image also revealed the formation of spherical crystalline NiO x NPs on a carbogenic support (Figure4 (c)). The size of the NiO NPs was estimated by applying the Debye-Scherrer equation to the (200) reflection at 2θ = 43.2° of the XRD pattern (Figure4 (a)), and the derived diameter of crystalline NiO NPs was 9.8 nm, which was consistent with the size estimated by TEM.

Figure 4
figure 4

XRD patterns, TEM images, and EDS mapping. XRD patterns and TEM images of (a, c) ra-MOF and (b, d) Pd@ra-MOF, respectively, after CO oxidation reactions. (e) EDS mapping of Pd@ra-MOF after CO oxidation reaction.

Full size image
Figure 5
figure 5

XPS spectra. (a) Ni 2p3/2 of ra-MOF, (c) Ni 2p3/2 of Pd@ra-MOF, and (e) Pd 3d5/2 of Pd@ra- MOF before and (b, d, f) after CO oxidation reaction (fifth run).

Full size image

A similar phenomenon occurred in the Pd@ra-MOF catalyst during CO oxidation. The XRD pattern of the catalyst after five consecutive CO oxidation reactions up to 300°C (Figure4 (b)) indicated the formation of NiO and PdO as well as the presence of pre-existing Pd(0) species. The NiO species were formed by the thermal transformation of ra-MOF (as described above), and the PdO NPs were generated by the oxidation of metallic palladium during the reaction. The XRD peaks for NiO and PdO species were broad, and the crystalline sizes determined by the Scherrer equation were 8.9, 9.9, and 10.3 nm for NiO, PdO, and Pd, respectively. The TEM observation of Pd@ra-MOF before and after the CO oxidation (Figure4 (d)) also confirmed the formation of spherical NiO and PdO nanoparticles. Contrary to the complete decomposition of ra-MOF at 500°C, the organic ligands, bpdc2− in Pd@ra-MOF, were partially decomposed to carbogenic supports and partially intact at a relatively low temperature, 300°C as an IR spectrum showed peaks for νO-C=O at 1,593(s) cm−1 and νC=C(aromatic) at 1,528(s) cm−1, respectively (Additional file 1:Figure S7). As shown in Figure4 (e), EDS mapping of the PdO x -NiO y /C catalyst after CO oxidation indicated that the PdO x and NiO y NPs were well dispersed on the carbogenic support. The morphology of the PdO x NPs was maintained after repeated CO oxidation runs, as evidenced by TEM images (Figure4 (d), right).

This might be attributed to the immobilization of the PdO x and NiO y NPs by carboxylate ligands coexisted on the carbogenic support and the consequent suppression of migration and aggregation of NPs, even under the harsh CO oxidation reaction condition.

Changes in the chemical states of Ni and Pd species in the Pd@ra-MOF before and after the catalytic reaction were monitored by XPS. Before the CO oxidation reaction, Pd@ra-MOF reasonably contained N-coordinating Ni(III) species (855.4 eV) as well as Ni(II) (854.7 eV), which were generated by the redox reaction with Pd(II) (Figure5 (c)). However, after the CO oxidation reaction, Ni2O3 (855.9 eV) and NiO (854.1 eV) were formed by the thermal transformation of Ni(II/III) macrocyclic complexes, similar to the case of the as-prepared ra-MOF catalyst (Figure5 (d)). Before the Pd@ra-MOF underwent a catalytic reaction, the small-sized Pd0 clusters (337.7 eV), which were strongly interacting with aromatic ligands of the ra-MOF[32], coexisted with unreduced Pd(NO3)2 species (338.3 eV) (Figure5 (e)). After the reaction, most of the Pd species were oxidized to PdO (336.9 eV) under a highly oxidative reaction condition, while the rest remained in the reduced Pd0 state (335.3 eV). It is noteworthy that the binding energy of the latter was shifted to a lower energy level compared to that of Pd@ra-MOF before the catalytic reaction (Figure5 (f)). This shift was due to the enlargement of the Pd NPs from approximately 2 nm to approximately 10 nm, which was consistent with the XRD results (Figure4 (b)). Based on the XPS, XRD, and TEM results, the Pd@ra-MOF was transformed into the PdO x -NiO y /C nanocomposite (0 ≤ x ≤ 1, 1 ≤ y ≤ 3/2) catalyst during the catalytic reaction, where each species showed the synergetic catalytic effect to convert CO to CO2 at a very low temperature. In addition, the PdO x -NiO y /C catalyst after the fifth run showed a higher BET surface area (35 m2/g) than that of Pd@ra-MOF, which can result in improved catalytic performance by providing active sites for catalyzing surface reaction (Additional file 1:Figure S2). However, the surface area is still very low, which implies that the significant portion of active sites of PdO and NiO NPs can be buried and covered by carbon support. This might result in the loss of catalytic activity from the potentially expected.

The various metal oxides supporting metal NP catalysts have been investigated as CO oxidation reaction catalysts. Machida et al. reported the CO oxidation activity of metallic Pd NPs supported on CeO2, which was significantly enhanced by thermal aging of the catalyst[33]. It was revealed that the thermal treatment caused the strong metal-support interaction via Pd-O-Ce bonding, which prevented the sintering of Pd oxide species at high temperature and promoted CO adsorption to react with oxygen. Haruta and other groups studied on supported gold catalysts on metal oxides such as TiO2, Fe2O3, Al2O3, CuO, La2O3 NiO, and Y2O3[34–36]. The activation of O2 molecules has been shown to occur at the perimeter between Au NPs and the metal oxide support, highlighting the importance of metal-metal oxide interface in promoting CO oxidation[37]. In this study, the catalytically active species for the CO oxidation reaction, PdO x -NiO y /C nanocomposite, was generated in situ in the reaction environment, and its stable, sustainable activity under cycled conditions could be ascribed to the synergistic interaction among the PdO, NiO NPs, and carbogenic supports. In addition, compared to the conventional methods, our in-situ generation method of MO x @C catalysts has an advantage due to the simple preparation procedure which can provide a strong metal-metal oxide interaction as well as carbon-metal oxide interaction during the reaction. In general, MO x @C catalysts are synthesized by loading metal precursors on carbon supports and successive pyrolysis, which is mostly conducted at high temperature for a long time[38]. In the present work, the spontaneously formed strong interactions not only provide the enhancement of catalytic properties, but also reduce mobility of metal precursors on the carbon and metal surfaces, which results in small and monodispersed metal oxide nanocrystals.

Conclusions

In conclusion, we reported the preparation of ra-MOF-supported Pd NPs, Pd@ra-MOF, via the redox couple-driven method, which yields unprotected metallic NPs at room temperature within a few minutes without the use of reducing agents. We found that during the CO oxidation reaction, the Pd@ra-MOF was transformed into a PdO x -NiO y /C nanocomposite that showed sustainable and enhanced catalytic activity through the synergistic stabilization of catalytically active PdO species. This study creates new opportunities for taking advantage of the MOF’s vulnerable point to develop a novel class of heterogeneous catalysts by utilization of MOFs or metal NPs@MOF as precursors.