Synthesis of Cu–Zn–Zr–Al–O catalysts via a citrate complex route modified by different solvents and their dehydrogenation/hydrogenation performance

Abstract

The quaternary Cu–Zn–Zr–Al–O catalysts have been prepared by a citrate-complex method using deionized water, ethanol, and ethyl acetate as solvents. The solvents with different polarity and solubility have a prominent influence on the reaction pathway(s) of the synthesis. When using ethanol as solvent to prepare the catalysts, the selectivity and yield to ethyl acetate of 89.5 wt% and 70.6 wt%, respectively, corresponding to 78.9 wt% conversion of ethanol, are achieved. This catalyst also shows good catalytic activity and stability (>120 h at 220 °C) in the conversion of ethanol containing 7 wt% of water. Its remarkable performance is due to the existence of smaller CuO particles and their more uniform dispersion, which is beneficial to form Cu–M_xO_y (M = Zn, Zr, Al) interfaces. The catalyst showing more strong basic sites associated with the high density of O²⁻ that migrated from the interaction of Cu–M_xO_y, which is favorable for the dehydrogenative dimerization of ethanol to ethyl acetate; furthermore, the moderate chemisorption and desorption of H₂ makes this catalyst more favorable for ethanol dehydrogenation. This catalyst also proves to be effective for a series of ethyl ester hydrogenations (e.g., ethyl acetate and diethyl oxalate), indicating a general promotion of these reactions using ethanol as the solvent to prepare the catalysts.

---

1. Introduction

The dehydrogenation/hydrogenation reactions are important facets of synthetic chemistry.¹ Taking the future large availability of bioethanol into account and for reducing greenhouse gas emissions that contribute to the global warming, the use of ethanol (EtOH) based on green chemistry as feedstock can be foreseen. Direct conversion of EtOH to ethyl acetate (AcOEt) through dehydrogenative dimerization with the liberation of H₂ is an attractive alternative and represents a simple, non-corrosive, non-toxic, and economical process.^2–5 The reaction is a combination of dehydrogenation, which is preferred at low pressures, and dimerization, which is preferred at high pressures. In the case of ethanol (EtOH) dehydrogenation, the copper-based catalysts have been successfully employed because of their ability to maintain the C–C bond intact while dehydrogenating the CO–H bond. Nevertheless, there are two limitations of the Cu catalysts, namely, their inherent activity is lower than those of noble metals and irreversible deactivation mainly due to metal sintering.^6–8 To overcome these problems, methods such as the Pd–Ag membrane reactor⁹ and acceptorless dehydrogenation reactions^1,10 have recently been developed, but these systems require the extensive use of noble metals and hazardous substances such as phosphorus. From environmental and economic viewpoints, it is important to synthesize and develop new catalysts with a higher activity for the conversion of ethanol and better selectivity to ethyl acetate under mild conditions. Herein, we highlight a catalyst design for improving the activity and stability of Cu-based catalysts and a significant improvement in ethyl acetate selectivity based on the understanding of the reaction mechanism.

In a metal–oxide interface, one can obtain adsorption/reaction sites with complementary chemical properties, whereas truly bi-functional sites would be very difficult to create on the surface of a pure metal or alloy system.¹¹ Moreover, the metal–oxide interface determines the formation and stability of the intermediates present in the ethanol transformation process. Based on the study by Bueno,^12,13 a combination of the Cu⁺/Cu⁰ pair or Cu⁰ interfaced to ZrO₂, which may provide synergism, is needed to efficiently transform ethanol to ethyl acetate. In the case of the Cu–Zn–Zr–Al–O (CZZA) catalyst,^7,14 the active site for the coupling of ethanol and aldehyde is at the mixed metal–oxide surface, not at the Cu metal surface. According to many reports,^15–17 by applying a citrate complex method as an advanced wet chemical route, multicomponent catalysts can be prepared which have higher uniformity in particle size distribution and combination, and which can exhibit higher synergistic interaction and cooperative catalysis than those prepared by coprecipitation and impregnation. This technique is based mainly on the formation of metal chelate complexes in solution, followed by elimination of the solvent via drying, resulting in a gel that contains the starting cations. The organic fraction of this gel is then removed by calcination, resulting in a very fine and reactive oxide powder. More encouragingly, chelating of solvated metal ions, as an exception, is governed by a relatively high contribution of entropy, due to the replacement of many solvent molecules around a metal ion by one chelating ligand.¹⁸ Thanks to this behavior, most metal chelate complexes exhibit relatively high stability. However, a few systematic studies have been made, taking the parameters of the first step, namely, citrate synthesis, into account. For example, the contribution of different solvents used in the citrate complex method is seldom investigated. In the complexation process specifically, not only does the solvent play an important role in the hydrolysis and condensation of metal ions, but it also facilitates the nanoparticle growth of the synthesized materials.¹⁹ Along this line, adsorption characteristics of the citrate–metal complex with the aid of solvents including the interaction between hydrogen-bonded citrate chains, surface coverage, and potential intermolecular interactions between citrate anions during nanoparticle growth are significantly important.²⁰ Furthermore, the stability of the citrate complex in solution is influenced by the presence of short-range repulsive forces, i.e., steric or electrostatic repulsion, which further affect nanoparticle growth.^20,21 Therefore, the solvent used may affect not only the surface features but also the morphological characteristics of the synthesized materials.

To gain an insight into the effect of solvents on the structural properties of Cu–Zn–Zr–Al–O catalysts in relation to their performance, we prepared CZZA catalysts by citrate-complex method employing deionized water, ethanol, and ethyl acetate as solvents. The products formed during ethanol conversion on CZZA were ethyl acetate (AcOEt), acetaldehyde (AcH), diethyl ether (DEE), n-butanol (BOL), methyl ethyl ketone (MEK), crotonaldehyde (CROT), propanone (PrO), CO, CO₂, ethylene (ETE), and 1-butene (BTE). On the basis of our experiments, and with respect to maximum catalytic activities of the CZZA catalysts, ethanol still appears to be the solvent of choice for the precursor synthesis. This strategy raises the prospect of using the citrate-complex method employing ethanol as solvent to prepare catalysts for the efficient dehydrogenation of alcohols to functionalized carbonyls, and consequently contributes to the sustainable development of green chemistry. Based on the understanding of selective activation of C–O and O–H bonds, the catalysts can also be applied to the hydrogenolysis reactions of ethyl esters, which are the simple esters and good model compounds for studying the selective hydrogenation of C=O and C–O bonds to produce various chemicals.

2. Experimental

2.1 Catalyst preparation

All the reagent grade chemicals were supplied by Sinopharm Chemical Reagent Co. Ltd, China, and used without further purification. Each CZZA quaternary catalyst sample was prepared using Cu(NO₃)₂·3H₂O, Zn(NO₃)₂·6H₂O, Zr(NO₃)₄·5H₂O, Al(NO₃)₃·9H₂O, and citric acid monohydrate (CA). The method was essentially a similar procedure as for an amorphous citrate process introduced by Sato et al.,¹⁶ where deionized water,¹⁵ ethanol, and ethyl acetate²² were used as solvents. The catalysts prepared were referred to as CZZA–x, where x = am, wa, et, and eta represent the amorphous citrate process, and those using water, ethanol, and ethyl acetate as solvents, respectively. Taking CZZA–et as an illustration, the mixed nitrates of Cu, Zn, Zr and Al with a desired molar ratio (12 : 1 : 4 : 2) were dissolved in ethanol. An alcoholic solution of CA (calculated as 1/3 mol CA per g eq. of each metal plus an extra 5%) was dropped rapidly into the above mentioned mixed nitrates solution with vigorous stirring. The mixture was then evacuated under a pressure of 1.0 kPa at 60 °C. After the resulting solid had been heated in air at 170 °C for 2 h, it was calcined in air at 550 °C for 2 h to obtain a CZZA sample. Similarly, the CZZA–wa catalyst was prepared through mixing an aqueous solution of CA with nitrates of Cu, Zn, Zr and Al, which were dissolved in deionized water; the CZZA–eta catalyst was prepared by adding an ethyl acetate solution of CA to nitrates of Cu, Zn, Zr and Al, which were dissolved in ethyl acetate. The preparation of CZZA–am catalyst was essentially the same procedure as for an amorphous citrate process, which involved physical mixing of CA with nitrates of Cu, Zn, Zr and Al without the aid of solvent. Equally, the hybrid catalysts (CZZA–m) were prepared by mechanically mixing ZnO, ZrO₂, and Al₂O₃ with CuO as a base for comparison. To investigate the effects of the addition of ZnO and Al₂O₃ on the structural evolution of CZZA–et, a sample of Cu/ZrO₂ (denoted as CZr–et) with the same Cu loading (ca. 54.2 wt%) as the CZZA catalysts was prepared by the citrate complex method employing ethanol as solvent.

Before the following utilization, the raw CZZA catalyst was tabulated, crushed and sieved to obtain particles with a size range of 0.18–0.25 mm.

2.2 Characterization of catalysts

The powder X-ray diffraction (XRD) patterns in the 2θ range of 5–85° at a scanning speed of 2° min⁻¹ were collected on a DX-2700 with monochromatic Cu-Kα radiation operating at 40 kV and 30 mA. Average particle sizes were calculated by the Scherrer equation using the (111) peak position of CuO from the XRD data [eqn (1)].where K is a constant generally taken as unity (0.89), λ is the wavelength of the incident radiation (0.15405 nm), FW is the full width at half maximum and θ is the peak position.

Nitrogen physisorption isotherms were recorded using a Micromeritics TriStar II 3020 instrument. The samples were outgassed at room temperature for 24 h before the measurements. Specific surface area (S_BET) calculations were carried out using conventional BET calculations. Pore volume (V_pore) and average pore diameter (d_p) were deduced using BJH pore analysis based on the desorption branch of the nitrogen adsorption/desorption isotherm. The surface morphology and the particle size were observed by transmission electron microscopy (TEM). The TEM measurements were performed with a JEM-2100F microscope operating at 200 kV. Samples for TEM were prepared by depositing a drop of an ultrasonically dispersed solution onto a standard amorphous carbon-coated copper grid.

The reducibility of the catalysts was studied by H₂ temperature-programmed reduction (H₂-TPR) on a homemade apparatus.²³ A flow of 5% H₂/N₂ (50 mL min⁻¹) was passed through 20 mg of sample with a ramping rate of 10 °C min⁻¹ to a final temperature of 600 °C. The effluent gas formed during the TPR experiment was passed through a molecular sieve trap to remove the water. In the meantime, the hydrogen consumed was monitored continuously by a thermal conductivity detector (TCD). Calibration of the instrument by the hydrogen consumed was carried out with 20 mg standard samples of CuO (Aldrich). The dispersion and specific surface area of metallic copper (D_Cu and S_Cu) were measured by one-pulse N₂O oxidation at 50 °C using the procedure described by van der Grift et al.^24,25 The calculation was based on the total amount of N₂O consumption with 1.46 × 10¹⁹ copper atoms per m². For more details, see the ESI.† The activation and desorption properties of hydrogen were tested by the temperature-programmed desorption of adsorbed H₂ (H₂-TPD) using the same apparatus employed by the TPR experiment. Prior to H₂ adsorption, the sample (120 mg) was reduced in a H₂ (50 mL min⁻¹) gas flow at 250 °C for 4 h. The sample was then pretreated under an Ar flow (50 mL min⁻¹) at 350 °C for 1 h to sweep the surface of the reduced catalyst. After cooling to room temperature by Ar flushing, the sample was exposed to H₂ (50 mL min⁻¹) for 1 h and heated to 600 °C at a rate of 15 °C min⁻¹. The effluent formed during the experiment was purified by silica gel desiccant and a 5 Å molecular sieve to eliminate water and other substances produced in the test, and subsequently monitored by TCD. H₂O-TPD measurement was carried out to investigate the adsorption/desorption of water on the catalysts. Before any measurements, all the samples were in situ saturated with water after being reduced in H₂ (50 mL min⁻¹) for 4 h. The temperature was subsequently ramped from 50 °C to 700 °C with a linear rate of 15 °C min under He flow (30 mL min⁻¹), and the tail gas was continuously monitored on stream by a TCD. Temperature-programmed desorption of adsorbed CO₂ (CO₂-TPD) on the reduced catalysts was analyzed using a quadrupole mass spectrometer (HIDEN, Hpr-20; Pfeiffer Vacuum Technology AG) equipped with the same apparatus employing the TPR experiment. Prior to CO₂ adsorption, the sample (200 mg) was reduced in a H₂ (30 mL min⁻¹) gas flow at 250 °C for 4 h. After reduction, the sample was swept at 350 °C for 1 h and subsequently cooled to room temperature by a He gas flow (50 mL min⁻¹). Several CO₂ pulses (1.2 mL CO₂ of one pulse) were then introduced to the sample to obtain a saturated adsorption of CO₂. The sample was purged by He for 1 h at 100 °C to eliminate the physically adsorbed CO₂; then, the sample was heated to 650 °C at 15 °C min⁻¹ in He and the effluent was continuously monitored by mass spectrometer. The mass number used was 44 for the CO₂ species.

The purity of the N₂, Ar, He, N₂O, CO₂, and H₂ gases were greater than 99.99%, and all of them were pretreated by silica gel desiccant, 5 Å molecular sieve, and some of them were deoxygenated by a silver molecular sieve before use.

2.3 Catalytic activity tests

Evaluations were conducted on a stainless steel fixed-bed reactor (internal diameter = 9.8 mm). The catalyst (2 g) was placed between two layers of quartz sand and reduced in a stream of diluted hydrogen (10% H₂ in N₂) at 250 °C under atmospheric pressure for 16 h with a flow rate of 30 mL min⁻¹. The catalyst was then heated to the desired reaction temperature at steps of 2 °C min⁻¹. The liquid reactant was injected by double-plunger pump, enabling a tunable liquid hourly space velocity (LHSV). After vaporizing by a pre-heater, the vapor was mixed with N₂ or H₂. Details of the reaction conditions are presented below each result. The condensable obtained products were collected in a trap and analyzed by gas chromatography (Shimadzu GC-14C) using a capillary column (19091n-213, HP-INNOWAX, 0.25 mm × 30 m) with FID and a packed column (TDX-101) with TCD as the detectors. The gas products were periodically analyzed by online gas chromatographs (Shanghai Haixin GC-950) with a packed column filled with carbon molecular sieve. The products were determined quantitatively by calibrated area normalization. Moreover, conversion and selectivity were based on the mass basis products (wt%) and calculated by the equations given in the ESI.†

3. Results and discussion

3.1 Textural and structural properties

The X-ray patterns of all the prepared catalysts are shown in Fig. 1. All the samples exhibited main sharp peaks at 2θ = 32.6°, 35.6°, 38.6°, 48.8°, 53.6°, 58.3°, 61.6°, 66.4°, 68.1°, 72.3° and 75.1° corresponding to a crystalline CuO phase with tenorite structure (JCPDS 481548). There were no peaks obtained corresponding to ZnO and Al₂O₃, indicating that the dispersion of these was more uniform and they were present in highly disordered or amorphous states.^26,27 A decrease of the peak intensity of copper species (Fig. 1) was apparent. The average CuO crystallite sizes, estimated by Scherrer's equation, were 19.8, 14.1, 10.9, and 8.6 nm for CZZA–eta, CZZA–am, CZZA–wa, and CZZA–et, respectively. Table 1 shows the dispersion and surface area of Cu⁰, surface area, pore volume and average pore diameter values for the CZZA catalysts. Although the samples did not show much difference in V_p and d_p, S_BET and S_Cu determined by N₂O titration method (Fig. S1†) were different, and the dispersion of copper species followed the sequence CZZA–et > CZZA–wa > CZZA–eta > CZZA–am. Among the samples, CZZA–et displayed the highest surface area (84.7 m² g_cat⁻¹) and copper surface area (49.2 m² g_cat⁻¹). Probably the increase in Cu surface areas was caused by the decrease in the particle size of CuO species, as shown in the XRD patterns of CZZA–et (Fig. 1). Based on the data mentioned above, we concluded that the CuO particle sizes in the CZZA–et became the smallest and the dispersion of Cu species was more uniform, indicating that the solvents used in the preparation process had a great influence on the dispersion of the copper species.

Fig. 1 XRD patterns of calcined CZZA–x samples prepared by citrate-complex method.

Table 1 Physicochemical properties and catalytic performance of the CZZA–x catalysts

Catalyst	DCu^a (%)	SCu^a (m^2 gcat^−1)	dp^a (nm)	SBET^b (m^2 gcat^−1)	Vpore^b (m^3 gcat^−1)	(SI/SII)^c	Conversion^d (wt%)	Sel. of AcOEt^d^,^e (wt%)	Yield of AcOEt^d (wt%)	Yield of H2^d (wt%)
a Cu dispersion and surface area of Cu^0 were determined by N2O titration.b Measured by BET method for the calcined sample.c H2O-TPD peak area ratios of the lower temperature desorption peak I to the higher temperature desorption peak II of CZZA–x.d Reaction conditions: P = 0.1 MPa, LHSV = 1.0 mL gcat^−1 h^−1, and T = 220 °C.e Other condensable by-products (e.g., AcH, MEK, CROT, PrO, and H2O) and gaseous products (e.g., CH4, C2H6, CO, and CO2) were also formed.
CZZA–am	6.5	24.6	15.4	71.2	0.095	0.28	46.5	58.2	27.1	1.9
CZZA–wa	11.9	41.5	8.4	68.6	0.079	0.45	58.6	70.4	41.2	2.6
CZZA–et	14.1	49.2	7.1	84.7	0.080	0.65	63.8	84.9	54.2	3.2
CZZA–eta	8.4	29.3	11.9	65.1	0.084	0.31	44.9	52.9	23.8	1.9

---

Fig. 2a–c compares the TEM images of the calcined CZZA–x catalysts. These show that the catalysts consisted of clusters of interconnected particles with sizes of 5–50 nm (Fig. S2e†). For CZZA–am and CZZA–wa (see Fig. 2a and b), CuO particles were agglomerated into clusters larger than 15 nm in size, which was also verified by XRD. The morphology of the two calcined catalysts clearly demonstrated that the CZZA–am and CZZA–wa catalysts often appeared in larger agglomerates than the CZZA–et, which resulted in a lower BET surface area and dispersion of Cu⁰ (Table 1). The corresponding dispersion of copper species, which increased by employing ethanol as solvent, can be directly observed from the TEM image of the CZZA–et sample (Fig. 2c). Moreover, the image for the CZZA–et sample showed little obvious evidence of bulk copper species condensed on the substrate M_xO_y (M = Zn, Zr, and Al), which further confirms that the copper species were uniformly dispersed in the CZZA–et after the synthesis process and the following high-temperature thermal treatment. As can be seen in Fig. S2,† contrary to the agglomerated particles in CZZA–wa, the magnified TEM images of CZZA–et show small particles of narrow dispersity embedded in an amorphous matrix, which are, most probably, formed from the substrate M_xO_y. Furthermore, the higher agglomeration and thus higher density of particles enables additional confirmation of the crystallinity of the CZZA–wa using selected area electron diffraction (SAED) (Fig. S2e†). Moreover, in the amorphous structure of M_xO_y, the small and even structure of CuO can be further confirmed by a successive diffraction halo in an attached SAED image for CZZA–et (Fig. S2c†). Generally speaking, this was because small CuO particles adhered to the surface of M_xO_y, and hence the sintering of copper particles was obstructed, as seen in the schematic illustration (Fig. 2d). This is in agreement with the results from XRD and metallic copper surface area measurement (Table 1 and Fig. S1†). However, the metallic Cu particle size of the CZZA–wa catalyst observed by TEM was larger than the crystallite size derived from XRD and N₂O titration, indicating the polycrystalline nature of the metallic Cu particles in CZZA–wa as compared to the particles of other samples. This could probably be connected to the contribution of the water, which facilitated the intimate contacting of Cu species in the precursors by building strong hydrogen-bonded networks, and then affected the sintering behavior in the following high-temperature thermal treatment, for example, forming the agglomerated or polycrystalline of CuO. Thus, the nanostructure was not distinctive in the images of TEM. In a word, the big difference of the active copper surface area among CZZA–x may originate from the differences in copper particle size and dispersion. Moreover, the increase in Cu⁰ surface area and Cu–M_xO_y interfaces is assumed to be partially responsible for the improvement of dehydrogenation and dimerization activity.

Fig. 2 TEM images of the calcined quaternary catalysts: (a) CZZA–am; (b) CZZA–wa; (c) CZZA–et and (d) the calcined binary catalyst Cu/ZrO₂ and (e) schematic illustration of the structural evolution of the nanostructured CZZA–x catalysts as a function of solvent.

The TEM images of the Cu/ZrO₂ catalysts are shown in Fig. 2e. The morphology of the CZr–et and CZZA–et catalysts clearly demonstrate that the binary catalyst consists of small particles of similar shape, while they often appear in larger agglomerates when compared to the quaternary catalyst. The aggregation of CuO nanoparticles in CZZA–et was not significant as in CZr–et. The combination of Al and Zn oxides with Zr oxide had been described as preventive elements for sintering of Cu crystallites and therefore was considered as a structural promoter and as an alternative support.²⁷ Moreover, the introduction of ZnO and Al₂O₃ effectively stabilized copper and prevented crystallite growth. As shown in Fig. 2c, it is clear that CZZA–et catalysts exhibited better resistance to Cu/CuO particle growth upon calcination than the binary CZ–et catalyst. The results show that the addition of ZnO and Al₂O₃ had a significant influence on the particle size distribution and the structure of the catalyst with the aid of ethanol as solvent. In addition, based on “X-ray amorphous” features for CZZA–et, zinc, zirconium, and aluminum phases were present in an amorphous-like state in the quaternary catalysts.

3.2 H₂-TPR analysis

Fig. 3a compares the H₂-TPR profiles for CuO with those of the four prepared CZZA–x catalysts. All the samples displayed reduction profiles characterized by a main peak with a maximum between 160 and 270 °C, well below that of the standard bulk CuO (ca. 285 °C); this suggested the presence of a copper–oxides interaction in the present samples, which facilitated the reduction of copper oxides. In fact, the peak was sharper and much more symmetrical for CZZA–am, CZZA–wa, and CZZA–eta. The profile of CZZA–et displayed a weak reduction peak centered at ca. 162 °C, which was lower than that of the sharp one. Fig. 3b shows typical TPR profiles before and after N₂O oxidation for CZZA–et. The peak area of the first TPR profile (TPR 1) corresponded to all CuO in the sample, and that of the second TPR (TPR 2) corresponded to Cu₂O produced by N₂O oxidation. In fact, the TPR peak was around 130 °C for surface Cu⁺ to Cu⁰ compounds and 170 °C for small CuO particles to Cu⁰ in Fig. 3b. As stated, the peak occurring at 192 °C for CZZA–et was unlikely to be caused by the reduction of Cu⁺, but rather by CuO particles with small sizes.²⁸ As is known, the more facile reduction–oxidation of Cu species in the small CuO particles was presumably due to a higher degree of surface defective dominant features, which meant a poor crystallinity with plenty of weakly bonded surface oxygen ions and higher surface area exposed to H₂, whereas large crystallites would appear in the TPR as species reducible at a higher temperature because of the diffusion hindrance on the reduction process and/or their relatively lower surface area exposed to H₂.^27,29 In this case, the reducing temperature in the profiles of TPR can be used to reflect the size of the Cu particles, e.g., CuO phase with a low reduction temperature denoted a small particle size.³⁰ Second, according to Behrens et al.,³¹ the crystalline and pure CuO species are more easily reduced than the highly dispersed CuO in amorphous materials as a result of diffusion effects and the strong interactions between Cu species and M_xO_y. The higher reduction temperature of second peak as opposed to the former peak for CZZA–et indicated the presence of highly dispersed CuO species embedded into the supports of M_xO_y. Moreover, interactions with supports can decrease the rate of CuO reduction.³² The consumption of H₂ for CZZA–et was lower than that of CZZA–am and CZZA–wa, which further confirmed that highly dispersed CuO particles strongly interacted with the M_xO_y thus decreasing the rate of CuO reduction. In a word, we proposed that there were two reducible copper species in the CZZA–et catalyst. One was represented by small crystalline CuO particles, which were finely dispersed and reduced at lower temperature (peak I); another was represented by highly dispersed CuO particles, which strongly interacted with the M_xO_y supports, and reduced at higher temperature (peak II). Deconvolution of the two TPR peaks indicated that the amounts of small CuO and Cu–M_xO_y were ca. 10% and 90%, respectively. Interestingly, the two-peak H₂-TPR profile of CZZA–et was consistent with a phenomenon discussed in detail for data obtained by H₂-TPD and elsewhere³³ regarding the generation of sites for medium absorption of H₂.³⁴ More details about this will be given below. On the other hand, the product of CuO reduction strongly interacted with the support; although it existed in the other three samples, can be partially covered up by the reduction product of large CuO crystallites. As a consequence, the reduction temperature of CZZA–x catalysts increased and followed the order CZZA–et < CZZA–wa < CZZA–am < CZZA–eta. Thus, the increasing reduction temperature in the profiles of TPR indicated the increasing size of CuO particles and/or possibly a strong interaction between copper and the support, which is in accordance with XRD and TEM results. Furthermore, it should be noted that the TPR measurement can be explained partially in terms of copper dispersion, especially for the sample containing zirconia because of the introduction of ZrO₂ in the sample facilitating the reduction of CuO.^27,29 The CZZA–et catalyst displayed the lowest reduction temperature, in which both the weak reduction peak centered at ca. 162 °C and the main peak appeared at approximately 220 °C. This indicated the presence of smaller CuO particles, and hence a higher dispersion than that of other samples. These results correlate well with the metallic copper surface area measurements and XRD results.

Fig. 3 (a): TPR profiles of the CZZA–x catalysts prepared by citrate-complex method and bulk CuO; (b) reduction profiles of CZZA–et catalyst before and after N₂O oxidation at 50 °C.

3.3 H₂-TPD analysis

The H₂-TPD profiles of the reduced CZZA–x catalysts are shown in Fig. 4. Two obvious H₂ desorption peaks (centered at 152 and 460 °C) and a slight peak (centered at 580 °C) appeared in the whole T range over all the samples, which were at similar temperatures, but differed greatly in the size. According to published studies in this area,^35–37 the resolved peak at low temperatures (120–240 °C) could be ascribed to the hydrogen moderately adsorbed on Cu (hydrogen on surface Cu sites), while a considerably broader signal in the range of 350–530 °C monitored the desorption of hydrogen from the split H–H on the surface of Zn–Al–Zr–O oxides or CuHx, which denotes the hydrogen strongly adsorbed on Cu, and the weak peak after 540 °C would be due to the oxidation of the metal by support protons. Thus, the H₂-TPD patterns of CZZA–x catalysts spanning a wide range of temperature (50–670 °C) is diagnostic of different adsorption states of hydrogen species across the catalyst structure (Fig. 4). At the first stage (corresponding to peak I), sites were probably being formed that had a high efficiency for the adsorption of H₂ and thus produced moderately adsorbed hydrogen on surface Cu sites.^36,38 Apparently, these peaks were enhanced in CZZA–et and CZZA–wa, while lowered in CZZA–eta and CZZA–am because of the decrease of small CuO particles, which were liable to be transformed to Cu⁰. The high-temperature peak (corresponding to peak II, denoted as hydrogen strongly adsorbed on catalysts) was attributed to the generation of sites for facile H₂ dissociation, which facilitated the formation of CuHx. The catalytic activities test for CZZA–am indicated that it had a strengthening of peak II and weakening of peak I in the TPD profile, showing a relatively low catalytic activity for ethanol dehydrogenation (46.5 wt% conversion of ethanol and 58.2 wt% selectivity to AcOEt). The same phenomenon was observed over CZZA–m (16.5 wt% conversion of ethanol and 26.5 wt% selectivity to AcOEt, data not shown here), which only displayed surface sites for H₂ dissociation. CZZA–et, which showed the best catalytic activities (ca. 64 wt% conversion of ethanol), had the highest TPD peak area of the lower temperature peak I. Therefore, the dehydrogenation activity of the catalyst decreased with increasing amounts of strong desorption of H₂. These results suggested that hydrogen activation driven by the adsorption–desorption of H₂ for CZZA–x markedly contributed to the H₂ activation ability, and these enhanced H₂ activation abilities (by the way, the desorption of H atom would be very difficult) may impair its activity for the dehydrogenation of ethanol. Qualitatively, the key for increasing catalytic activity derives from the balance between the number of active sites and the ease of product desorption. In other words, the moderate chemisorptions and desorption of H₂ avoided the competitive adsorption of H and ethanol or acetaldehyde on metallic active sites. Moreover, the recombination of H and ethoxide yields ethanol again.^12,13

Fig. 4 H₂-TPD profiles of the reduced CZZA–x catalysts.

3.4 H₂O-TPD analysis

The H₂O-TPD profiles of the CZZA–x catalysts are shown in Fig. 5. There were two obvious H₂O desorption peaks (centered at 152 and 382 °C), which appeared for all CZZA–x catalysts. Both were similar in terms of temperature, but differed greatly in the size. Clearly, the TPD peak area of the lower temperature peak I enhanced in CZZA–et and lowered in CZZA–eta, indicating that there were more active sites for the adsorption/desorption of H₂O at the reaction temperature. The considerably broader peak II is attributed to the H₂O strongly adsorbed on Cu. This strong adsorption of H₂O competed with ethanol or acetaldehyde on the surface of CZZA–x catalysts and led to the hindering of ethanol conversion. Table 1 compares TPD peak area ratios of the peak I to the peak II for the CZZA–x catalysts. Analyses indicated that the catalyst having a high S_I/S_II ratio showed a relatively high desorption activity for water on the catalyst; moreover, it afforded more active sites for the adsorption of ethanol. As discussed above, one can suppose that CZZA–et would show better catalytic activities in the conversion of hydrous ethanol because the substantial desorption of H₂O avoided the competitive adsorption of H₂O and ethanol or acetaldehyde on metallic active sites at the reaction temperatures.

Fig. 5 H₂O-TPD profiles of the reduced CZZA–x catalysts.

3.5 CO₂-TPD analysis

As is known, zirconia possesses surface Lewis basic sites, which were able to adsorb CO₂, while ZnO enhanced the affinity of the system for CO₂.³⁶ Moreover, the combination of Al oxide with Zn and Zr oxides probably reduced the acidity of alumina.²⁷ CO₂-TPD profiles of CZZA–x samples are shown in Fig. 6. As it can be verified, these samples showed different profiles regarding their interaction with CO₂. Several obvious CO₂ desorption peaks (at 150–180 and 320–500 °C) appeared over the entire temperature range. The weak basic sites were related to a curve, which showed a maximum at a temperature of ca. 180 °C; the ones between 350 and 500 °C were strong basic sites and, finally, the peaks above 550 °C were due to the release of CO₂ from the support.³⁹ Thus, the CO₂-TPD profile of CZZA–x discloses a clear concentration of varied basic sites on the surface (Fig. 6). As it can be observed, CZZA–et showed the largest peak for strong basic sites followed by CZZA–wa and CZZA–am. There were no distinct peaks of desorption of CO₂ for CZZA–eta, while there was a broad and lower peak with maximum between 350 and 520 °C, which was similar to that of CZZA–et (curves in Fig. 6); a broad peak for the desorption of CO₂ with maximum between 400 and 520 °C appeared, which was similar to that of CZZA–am. Therefore, the CO₂-TPD profiles indicated increasing contribution of stronger basic sites associated with the interaction between Cu and M_xO_y. In other words, the high density of basic sites associated with the oxygen sites originated from the oxygen-rich interface at Cu–M_xO_y by oxygen diffusion. This implied that the location of O preferred a high coordination environment on a neutral Cu system, where the surface electron density was high, which is in line with the strong ionic bonding character of O and Cu.³² As was verified, acetaldehyde produced on the dehydrogenation catalyst migrated towards the oxides and reacted with the ethoxide species, which were generated by the oxide basic sites. The resulting hemiacetal was dehydrogenated and the ethyl acetate obtained was desorbed. Clearly, oxides with strong basic sites generated more effective systems than weak ones for the ethyl acetate synthesis.^14,40 Of all the samples used in these experiments, CZZA–et showed the best catalytic activities (Table 1). Actually, when CZZA–eta was employed, a lower selectivity to ethyl acetate was observed than that of CZZA–et (52.9 wt% vs. 84.9 wt%). The low selectivity to ethyl acetate observed from the others might be associated with the low density of the strong basic sites. Above all, we can conclude that the basic sites on the CZZA–et interacted strongly with Cu nanoparticles and formed Cu–M_xO_y, generating a catalytically active nanoenvironment for reaction coupling between alcohol dehydrogenation and dimerization.⁴¹ This suggested the existence of a synergism of metal Cu dehydrogenation and oxide basic sites dimerization, which provided the metal/oxide interface on the functionality of Cu–Zn–Zr–Al–O catalysts and enhanced the dual-site nature.^27,42,43

Fig. 6 CO₂-TPD profiles of the reduced CZZA–x catalysts.

3.6 Effect of solvent on the structure evolution of Cu–Zn–Zr–Al–O catalyst

For amorphous citrate process, the mixture of metal nitrates reacted with CA and formed hexanuclear complexes. In this process, CA molecules exist as a polymeric skeleton structure containing free carboxyl and carboxylate groups attached to the metal cation, which act as a chelating agents and help in the dispersion of the metal component. For example, the coordination of bidentate carboxylate anion on metal ion is observed in the composites of copper, alumina, titania, and zirconia with carboxylic acids.¹⁷ The presence of Cu–M_xO_y interaction in the samples but at different levels was due to this. However, the CuO in CZZA–am was liable to accumulate partially due to the absence of solvent and thus formed the heterogeneous distribution and agglomerated structure.

Solution chemistry of citrate ligands and metal ions is complicated because citrates form complexes containing metal ions of different identity (heteronuclear or mixed-metal complexes). Generally, the polarity of the solvent made a great contribution to the citric complexes of Cu–Zn–Zr–Al–O catalysts. In this case, the reduction in hydration/solvation was related to preorganization or intrinsic basicity of the ligands for the complexation of metal ions.⁴⁴ Apart from that, we observed that solvents having low dielectric constants (ethanol, 24.3) accelerated the homogeneous complexing process, while solvents having high dielectric constants (water, 80.4) did not have an equivalent effect on the process. Comparing CZZA–et with CZZA–wa, small copper oxide particles (8–10 nm) and highly dispersed copper species in a strong interaction with the M_xO_y exist in CZZA–et. This could probably be connected with the capability of solvent for stabilizing the polymeric structure of the precursors by building weak/strong hydrogen-bonded networks, which then affected the sintering behavior in the following high-temperature thermal treatment. In addition, the nanostructure was enhanced by solvents of lower polarity.⁴⁵ Despite the low polarity of ethyl acetate, it contributed little to the catalytic activities of CZZA–eta because of the low solubility of metal nitrates and CA in ethyl acetate (≈15 wt% here). For example, ethyl acetate exhibited a negative influence on the catalytic activities of CZZA–eta. The low solubility of CA in ethyl acetate, the dissolution and repolymerization might be inhibited by the non-uniform contact in the heterogeneous suspension, thereby lowering the formation of mixed-metal complexes, as was experimentally observed for CZZA–am without solvent. When water or ethanol was used as the solvent, they significantly increased the formation of mixed-metal complexes and increased the dispersion of Cu species in the M_xO_y matrix by dissolving the components and forming a uniform solution during the process of catalyst preparation. Regarding the abovementioned facts, it was obvious that the formation of agglomerated structure of CZZA particles prepared by citrate complex method was induced by the morphology of the precursor gel by using different solvents.^46,47

3.7 Performance of CZZA–et catalyst in the reaction of ethanol conversion

As shown in Table 1, the selectivity of ethyl acetate over CZZA–x catalysts was in the order CZZA–et > CZZA–wa > CZZA–am > CZZA–eta, i.e., selectivity decreased from 84.9 wt% to 70.4 wt%, 58.2 wt%, and 52.9 wt%, respectively. This sharp decrease in selectivity can be attributed to the decreased amount of strong basic sites on the CZZA–x catalysts that catalyzed the bimolecular condensation or dimerization of EtOH to AcOEt.⁴⁰ Owing to the decreased dispersion of Cu species and increased particle size of CuO for CZZA–x catalysts, the observed conversion of ethanol decreased in the order of CZZA–et > CZZA–wa > CZZA–am > CZZA–eta along with the selectivity to AcOEt (Table 1).

Among all the catalysts, it was noted that CZZA–et exhibited the best catalytic activities for the synthesis of AcOEt (Table 1). Fig. 7a shows the changes in EtOH conversion, AcOEt selectivity, and yield of AcOEt and H₂ with LHSV at 220 °C and 0.1 MPa over the CZZA–et catalyst. At LHSV = 0.4 mL g_cat⁻¹ h⁻¹, the EtOH conversion was 78.9 wt%. The ethanol conversion decreased significantly with increasing LHSV. The ethyl acetate selectivity showed a maximum (89.5 wt%) at LHSV = 0.4 mL g_cat⁻¹ h⁻¹, which then decreased slightly with increasing LHSV. This indicated that in the formation of ethyl acetate from ethanol, the stepwise reaction requires a prolonged contact time. In the analysis of gaseous products, hydrogen and carbon dioxide were observed; more than 98 mol% of the gaseous product was hydrogen, with a small amount of CO₂. The amount of hydrogen produced corresponded to the sum of the derivatives from the dehydrogenation, and the data mentioned above was checked by the yield of H₂ and was close to 95 mol% for each experiment. Fig. 7b shows the changes in the selectivity to ethyl acetate and the conversion of the ethanol with reaction pressure. According to Inui⁶ and Santacesaria,⁸ it was advantageous that the rate of formation of ethyl acetate was accelerated at high pressure by the increase in intermolecular collision frequency. Under an ambient pressure of 0.2 MPa, the selectivity to ethyl acetate was 85.1 wt% at the temperature of 220 °C and the LHSV of 1.0 mL g_cat⁻¹ h⁻¹. The selectivity to AcOEt greatly increased with increasing reaction pressure, and reached a plateau at around 0.8 MPa (90.1 wt%), which then increased slightly regardless of the sharp increase in the pressure. In contrast, the observed conversion of ethanol decreased with increasing reaction pressure. Combining the profiles from Fig. 7b, it was inferred that an optimum pressure was about 0.8 MPa in a practical operation because of the relatively high yield of ethyl acetate (56.3 wt%) without any higher pressure.

Fig. 7 Changes in conversion of ethanol and selectivity to ethyl acetate over CZZA–et catalyst with (a): LHSV at 220 °C and 0.1 MPa; and with (b): pressure at 220 °C and LHSV = 1.0 mL g_cat⁻¹ h⁻¹.

The reactant ethanol containing 7 wt% of water was fed over CZZA–et catalysts under varied pressure (Fig. 8a). A selectivity to AcOEt of 80.1 wt%, which is based on 64.4 wt% conversion of ethanol, was achieved at 220 °C, 0.1 MPa, and LHSV = 1.0 mL g_cat⁻¹ h⁻¹. Compared with the reaction of ethanol conversion, the conversion of hydrous ethanol and formation of the diethyl ether and C4-species were decreased slightly because the aldol addition or hydration reaction was retarded by the presence of a little water. However, acetic acid was formed together with the products of the dehydrogenation of ethanol (data not shown). As the reaction pressure was increased from 0.2 MPa to 0.8 MPa, the selectivity of ethyl acetate increased from 80.1 wt% to 87.5 wt%, while the observed conversion of ethanol decreased from 64.4 wt% to 55.1 wt%, respectively. It was also found that there was no noticeable improvement in selectivity when the pressure was increased from 0.8 to 1.6 MPa, while the observed conversion of ethanol decreased to 51.5 wt%. Therefore, the higher pressure would promote the competitive adsorption of H₂O and ethanol or acetaldehyde on metallic active sites, which resulted in the decrease of the conversion of ethanol.

Fig. 8 (a) Changes in the conversion of ethanol and selectivity to ethyl acetate with pressure at 220 °C and LHSV = 1.0 mL g_cat⁻¹ h⁻¹ over CZZA–et. (b) Stability evaluation of CZZA–et for the dehydrogenative dimerization of ethanol to ethyl acetate at 220 °C and LHSV = 1.0 mL g_cat⁻¹ h⁻¹.

As is well known, water vapor increases the sintering rate of supported metals and leaching of the metal in the reaction system, while carbon deposition is another important reason for catalyst deactivation.^48–50 The stability of the CZZA–et catalyst was evaluated by a test run of 120 h under the following reaction conditions: 220 °C, atmospheric pressure, and LHSV = 1.0 mL g_cat⁻¹ h⁻¹. As shown in Fig. 8b, at initial reaction, the ethanol conversion was nearly 51.5 wt% and the selectivity to ethyl acetate was 79.3 wt%. However, the catalytic activity evidently increased with the reaction time, which gradually increased during the first 48 h of testing. The ethanol conversion reached 62.3 wt%, and the selectivity of ethyl acetate reached 80.1 wt% at 48 h. For the first 48 h of testing, the catalytic activity increased because of the instability of the Cu particle after the gradual reduction and formation of active sites for ethanol conversion.^48–50 After 48 h, the catalytic activity was stabilized at the ethanol conversion of ca. 63 wt% and the selectivity to ethyl acetate was stable at ca. 80 wt%. However, both the conversion of ethanol and selectivity to AcOEt remained unchanged, and no deactivation phenomenon was observed for CZZA–et during the 120 h stability test. In addition, mobile oxygen from the M_xO_y support may participate in the oxidation of carbonaceous species at higher temperatures, preventing carbon deposition.⁵¹

3.8 Applications for the hydrogenation of carbonyl compounds

The hydrogenation of esters is one of the most fundamental and widely employed reactions. As representative reactions, we have chosen the hydrogenation of diethyl oxalate (DEO) and ethyl acetate, which is known to be critically affected by the amount of metal-active sites on copper-based catalysts.⁵² The catalytic performances of the CZZA–x catalysts prepared using different solvents are listed in Table 2. Steady-state product compositions were obtained after about 36 h on stream and the catalytic activity of all the samples remained unchanged after 120 h on stream.

Table 2 Hydrogenation of diethyl oxalate and ethyl acetate over the CZZA–x catalysts

Catalyst	Conversion^a of DEO/%	Selectivity^a/%			Conversion^b of AcOEt/%	Selectivity^b/%
		EG	EtOH	EGT		EtOH	ETA
a Reaction conditions for hydrogenation of DEO: P = 2.4 MPa, T = 240 °C, H2/DEO = 25 mol mol^−1, LHSV = 1.2 mL gcat^−1 h^−1. Other by-products included ethoxyethanol (2-EE), AcOEt, ETA, 1,2-butanediol (1,2-BDO), CO, and CO2.b Reaction conditions for hydrogenation of AcOEt: P = 2.4 MPa, T = 240 °C, H2/AcOEt = 25 mol mol^−1, LHSV = 1 mL gcat^−1 h^−1. Other byproducts included AcH, MEK, CROT, PrO, CO, CO2, ETE, BTE, DEE, and BOL.
CZZA–am	51.5	25.6	48.6	19.6	81.2	92.6	1.6
CZZA–wa	63.4	22.1	41.0	26.7	81.6	91.3	2.3
CZZA–et	77.1	55.1	31.6	6.1	95.8	98.3	0.4
CZZA–eta	34.0	15.4	68.5	10.7	79.3	95.9	1.1

---

It is known that the hydrogenation of DEO proceeds via ethyl glycolate (EGT) to ethylene glycol (EG), while EG can dehydrate further to ethanol (Scheme 1).^53,54 For pressures above 1.0 MPa, with the ratio of H₂/ester more than 30/1, the formation of EG was favored.⁵⁵ The activity and durability of the catalyst under severe conditions, including low H₂/ester and high space velocity, was tentatively employed in this study. Under the harsh reaction conditions specified in Table 2, CZZA–x (x = am, wa, and eta, respectively) showed low activities for hydrogenation of DEO, while the efficiencies were greatly improved when ethanol was used as the solvent to prepare the CZZA–et catalyst. The CZZA–et catalyst, which had the highly dispersed Cu, also showed better catalytic performances for the hydrogenation of AcOEt (Table 2). The results here indicated that the prominent role of ethanol upon preparing Cu–Zn–Zr–Al–O catalyst could be generalized into hydrogenation of esters, facilitating future studies on the CZZA synergy for the hydrogenation of carbonyl groups.

Scheme 1 Reaction pathway for the hydrogenation of DEO to EGT, EG, and ethanol.

3.9 Structure and performance correlation

As can be seen in Table 1, all the samples showed dehydrogenation activity to some extent in the conversion of ethanol, even CZZA–eta which was prepared by employing ethyl acetate as solvent. Nevertheless, the catalyst employing ethanol as solvent was distinguished from the other samples because of the existence of quantitatively small CuO particles and abundant Cu–M_xO_y interfaces, as observed from XRD, TEM, TPR, H₂-TPD, and CO₂-TPD analysis. However, catalysis was controlled not only by the chemical composition and size of the catalyst used but also by the type of surface sites available at the catalyst surface. As a result, a nonlinear relationship between the dispersion of CuO and catalytic activity may be present in the dehydrogenative dimerization of ethanol to ethyl acetate. Moreover, limited by the equilibrium conversion of ethanol under the reaction conditions employed in this study (ca. 74 wt% conversion of ethanol at 0.1 MPa and 220 °C),⁶ the improvement of ethanol conversion over the as-prepared Cu–Zn–Zr–Al–O catalysts was not obvious, irrespective of the solvent used during the preparation. However, the equilibrium of the dehydrogenation of ethanol shifted toward the synthesis of ethyl acetate, which was achieved by a synergism of metal Cu for the dehydrogenation of ethanol and strong basic sites for the coupling of ethanol and aldehyde in the CZZA–et. Thus, a high yield of ethyl acetate with increasing conversion of ethanol was obtained. This method was also applied to other systems where high dispersion and small size of Cu species were important, e.g. hydrogenation of a series of ethyl esters (e.g., ethyl acetate and diethyl oxalate), indicating a general promotion of these reactions due to the highly dispersed Cu species in the CZZA–et catalyst.

4. Conclusions

The morphologies and catalytic activities of quaternary Cu–Zn–Zr–Al–O catalysts prepared by citrate-complex method were significantly altered by the solvent utilized. When a solvent with lower polarity and excellent solubility such as ethanol was employed, the catalyst showed excellent performance in dehydrogenative dimerization of ethanol to ethyl acetate. When a solvent having higher polarity such as water or low dissolving capacity such as ethyl acetate was used, the catalyst performed poorly in the reaction. Actually, the sample prepared by amorphous citrate process using no solvent also performed poorly in the reaction. Thus, the solute–solvent interaction between ethanol and precursors played an important role in the synthesis process and the nanostructure was enhanced by the lower polarity and prominent solubility of ethanol. Furthermore, the decreased hydration within the alcoholic solution led to a decrease in the size of the metal ions and thus an increased ability to form complex compounds with CA. In fact, the catalyst employing ethanol as solvent was distinguished from the other samples because of the existence of quantitatively small CuO particles and abundant Cu–M_xO_y interfaces, as observed from XRD, TEM, and TPR testing. The change in catalytic activities on both samples used was paralleled by a dramatic change in the hydrogen adsorption–desorption properties, as seen from the H₂-TPD profiles. Based on CO₂-TPD, it was verified that the selectivity to ethyl acetate increased with increase in strong basic sites. The promotional roles of solvents can be generalized into hydrogenation of esters (e.g., ethyl acetate and diethyl oxalate), where high dispersion and small size of Cu species were important, indicating a general promotion of these reactions due to the highly dispersed Cu species. Our study highlighted the design of highly active, selective, and stable Cu catalysts via a simple route, simultaneously solving the low reactivity and deactivation caused by the sintering of supported metal catalysts. The results will be useful in the development of supported metal catalysts for a range of dehydrogenation or hydrogenation reactions and have great implications for practical applications.

Acknowledgements

We would like to thank Mr Wenbin Zhang and Dr Yanting Liu for the facilitation in the equipment maintenance, Dr Shenke Zheng and Zhikai Li for the helpful discussions. This study was supported by the Natural Science Foundation of China (Grant No. 21373254) and PetroChina (Project No. 14-08-05-02).

References

1. C. Gunanathan and D. Milstein, Science, 2013, 341, 249.
2. C. Angelici, B. M. Weckhuysen and P. C. Bruijnincx, ChemSusChem, 2013, 6, 1595–1614.
3. A. L. Wang, H. Xu, J. X. Feng, L. X. Ding, Y. X. Tong and G. R. Li, J. Am. Chem. Soc., 2013, 29, 10703–10709.
4. J. Llorca, N. S. Homs, J. Sales and P. R. R. de la Piscina, J. Catal., 2002, 209, 306–317.
5. T. G. Kelly and J. G. Chen, Green Chem., 2014, 16, 777–784.
6. K. Inui, T. Kurabayashi and S. Sato, J. Catal., 2002, 212, 207–215.
7. K. Inui, T. Kurabayashi and S. Sato, Appl. Catal., A, 2002, 237, 53–61.
8. E. Santacesaria, G. Carotenuto, R. Tesser and M. Di Serio, Chem. Eng. J., 2012, 179, 209–220.
9. G. Zeng, T. Chen, L. He, I. Pinnau, Z. Lai and K. W. Huang, Chem.–Eur. J., 2012, 18, 15940–15943.
10. M. Nielsen, H. Junge, A. Kammer and M. Beller, Angew. Chem., Int. Ed., 2012, 51, 5711–5713.
11. J. Graciani, K. Mudiyanselage, F. Xu, A. E. Baber, J. Evans, S. D. Senanayake, D. J. Stacchiola, P. Liu, J. Hrbek, J. F. Sanz and J. A. Rodriguez, Science, 2014, 345, 546–550.
12. A. G. Sato, D. P. Volanti, I. C. de Freitas, E. Longo and J. M. C. Bueno, Catal. Commun., 2012, 26, 122–126.
13. I. Freitas, S. Damyanova, D. Oliveira, C. Marques and J. Bueno, J. Mol. Catal. A: Chem., 2014, 381, 26–37.
14. A. B. Gaspar, F. G. Barbosa, S. Letichevsky and L. G. Appel, Appl. Catal., A, 2010, 380, 113–117.
15. J. H. Ryu, S. M. Koo, J. W. Yoon, C. S. Lim and K. B. Shim, Mater. Lett., 2006, 60, 1702–1705.
16. S. Sato, K. Koizumi and F. Nozaki, J. Catal., 1998, 178, 264–274.
17. S. Samantaray, D. K. Pradhan, G. Hota and B. G. Mishra, Chem. Eng. J., 2012, 193–194, 1–9.
18. D. W. Kim and S. G. Oh, Mater. Lett., 2005, 59, 976–980.
19. R. A. Lucky, R. Sui, J. M. H. Lo and P. A. Charpentier, Cryst. Growth Des., 2010, 10, 1598–1604.
20. J. W. Park and J. S. Shumaker-Parry, J. Am. Chem. Soc., 2014, 136, 1907–1921.
21. Z. S. Pillai and P. V. Kamat, J. Phys. Chem. B, 2003, 108, 945–951.
22. J. Słoczyński, R. Grabowski, P. Olszewski, A. Kozłowska, J. Stoch, M. Lachowska and J. Skrzypek, Appl. Catal., A, 2006, 310, 127–137.
23. H. Boer, W. Boersma and N. Wagstaff, Rev. Sci. Instrum., 1982, 53, 349–361.
24. J. Evans, M. Wainwright, A. Bridgewater and D. Young, Appl. Catal., 1983, 7, 75–83.
25. C. van der Grift, A. Wielers, B. Jogh, J. van Beunum, M. de Boer, M. Versluijs-Helder and J. Geus, J. Catal., 1991, 131, 178–189.
26. L. R. W. Dai, in Handbook of Heterogeneous Catalysis, ed. G. Ertl, H. Knozinger, F. Schuth and J. Weitkamp, Wiley-VCH Verlag GmbH, 1st edn, 2008, vol. 3, pp. 1064–1077.
27. J. Agrell, H. Birgersson, M. Boutonnet, I. Melián-Cabrera, R. M. Navarro and J. L. G. Fierro, J. Catal., 2003, 219, 389–403.
28. R. X. Zhou, T. M. Yu, X. Y. Jiang, F. Chen and X. M. Zheng, Appl. Surf. Sci., 1999, 148, 263–270.
29. W. P. Dow, Y. P. Wang and T. J. Huang, J. Catal., 1996, 160, 155–170.
30. M. F. Luo, Y. J. Zhong, X. X. Yuan and X. M. Zheng, Appl. Catal., A, 1997, 162, 121–131.
31. M. Behrens, I. Kasatkin, S. Kühl and G. Weinberg, Chem. Mater., 2010, 22, 386–397.
32. A. Sato, D. Volanti, D. Meira, S. Damyanova, E. Longo and J. Bueno, J. Catal., 2013, 307, 1–17.
33. C. S. Polster, H. Nair and C. D. Baertsch, J. Catal., 2009, 266, 308–319.
34. F. Arena, K. Barbera, G. Italiano, G. Bonura, L. Spadaro and F. Frusteri, J. Catal., 2007, 249, 185–194.
35. S. Xia, Z. Yuan, L. Wang, P. Chen and Z. Hou, Appl. Catal., A, 2011, 403, 173–182.
36. F. Arena, G. Italiano, K. Barbera, S. Bordiga, G. Bonura, L. Spadaro and F. Frusteri, Appl. Catal., A, 2008, 350, 16–23.
37. H. Lin, X. Duan, J. Zheng, X. Zheng, P. He, Y. Yuan and Y. Yang, RSC Adv., 2013, 3, 11782–11789.
38. H. Wilmer, T. Genger and O. Hinrichsen, J. Catal., 2003, 215, 188–198.
39. Z. Y. Ma, C. Yang, W. Wei, W. H. Li and Y. H. Sun, J. Mol. Catal. A: Chem., 2005, 227, 119–124.
40. P. C. Zonetti, J. Celnik, S. Letichevsky, A. B. Gaspar and L. G. Appel, J. Mol. Catal. A: Chem., 2011, 334, 29–34.
41. F. Wang, R. Shi, Z. Q. Liu, P. J. Shang, X. Pang, S. Shen, Z. Feng, C. Li and W. Shen, ACS Catal., 2013, 3, 890–894.
42. S. H. Kang, J. W. Bae, H. S. Kim, G. M. Dhar and K. W. Jun, Energy Fuels, 2010, 24, 804–810.
43. H. W. Lim, M. J. Park, S. H. Kang, H. J. Chae, J. W. Bae and K. W. Jun, Ind. Eng. Chem. Res., 2009, 48, 10448–10455.
44. A. Martell, R. Hancock and R. Motekaitis, Coord. Chem. Rev., 1994, 133, 39–65.
45. E. Krupicka, A. Reller and A. Weidenkaff, Cryst. Eng., 2002, 5, 195–202.
46. R. Takahashi, S. Sato, T. Sodesawa, M. Kawakita and K. Ogura, J. Phys. Chem. B, 2000, 104, 12184–12191.
47. R. T. Wang, L. B. Kong, J. W. Lang, X. W. Wang, S. Q. Fan, Y. C. Luo and L. Kang, J. Power Sources, 2012, 217, 358–363.
48. C. H. Bartholomew, Appl. Catal., A, 2001, 212, 17–60.
49. J. T. Sun, I. S. Metcalfe and M. Sahibzada, Ind. Eng. Chem. Res., 1999, 38, 3868–3872.
50. R. Hughes, Chem. Eng. Sci., 1989, 44, 1747–1748.
51. H. Song and U. S. Ozkan, J. Catal., 2009, 261, 66–74.
52. Y. N. Wang, X. Duan, J. Zheng, H. Lin, Y. Yuan, H. Ariga, S. Takakusagi and K. Asakura, Catal. Sci. Technol., 2012, 2, 1637–1639.
53. D. J. Thomas, J. T. Wehrli, M. S. Wainwright, D. L. Trimm and N. W. Cant, Appl. Catal., A, 1992, 86, 101–114.
54. Y. Cui, C. Wen, X. Chen and W. L. Dai, RSC Adv., 2014, 4, 31162–31165.
55. G. H. Xu, Y. C. Li, Z. H. Li and H. J. Wang, Ind. Eng. Chem. Res., 1995, 34, 2371–2378.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c5ra13778k

---