**Gas-Phase Hydrogenation of Furfural to Furfuryl Alcohol over Cu-ZnO-Al2O3 Catalysts Prepared from Layered Double Hydroxides**

**Guillermo R. Bertolini 1 , Carmen P. Jiménez-Gómez 2, \*, Juan Antonio Cecilia 2, \* and Pedro Maireles-Torres 2**


Received: 15 April 2020; Accepted: 27 April 2020; Published: 29 April 2020

**Abstract:** Several layered double hydroxides (LDHs) with general chemical composition (Cu,Zn)1−xAlx(OH)2(CO3)x/2·mH2O have been synthesized by the co-precipitation method, maintaining a (M2+/M3+) molar ratio of 3, and varying the Cu <sup>2</sup>+/Zn <sup>2</sup><sup>+</sup> molar ratio between 0.2 and 6.0. After calcination and reduction steps, Cu/ZnO/Al2O<sup>3</sup> catalysts were synthesized. These catalysts were characterized by X-ray diffraction (XRD), transmission electron microscopy (TEM), H<sup>2</sup> thermoprogrammed reduction (H2-TPR), N<sup>2</sup> adsorption-desorption at −196 ◦C, N2O titration, X-ray photoelectron miscroscopy (XPS), NH3-thermoprogramed desorption (NH3-TPD) and CO2 thermoprogrammed desorption (CO2-TPD). The characterization data revealed that these catalysts are mainly meso-and macroporous, where Cu, ZnO and Al2O<sup>3</sup> are well dispersed. The catalytic results show that these catalysts are active in the gas-phase hydrogenation of furfural, being highly selective to furfuryl alcohol (FOL) and reaching the highest FOL yield for the catalyst with a Cu <sup>2</sup>+/Zn <sup>2</sup><sup>+</sup> molar ratio of 1. In an additional study, the influence of the aging time on the synthesis of the LDHs was also evaluated. The catalytic data revealed that the use of shorter aging time in the formation of the LDH has a beneficial effect on the catalytic behavior, since more disordered structures with a higher amount of available Cu sites is obtained, leading to a higher yield towards FOL (71% after 5 h of time-on-stream at 210 ◦C).

**Keywords:** layered double hydroxides; Cu-based catalysts; Cu/ZnO/Al2O3; furfural; furfuryl alcohol

#### **1. Introduction**

Layered double hydroxides (LDHs), also known as anionic-clays or hydrotalcites, are a group of inorganic lamellar compounds of basic nature. In the last decades, much attention is being paid to the study of these inorganic materials due to their chemical and structural properties, which make them useful in interesting applications, such as adsorbents, catalysts, anion exchangers or flame retardants, among others [1–3]. These materials exhibit high chemical stability, good biocompatibility and pH-dependent solubility [4]. The first LDH reported in the literature, discovered in 1842, was the mineral hydrotalcite [Mg6Al2(OH)16]CO3·4 H2O. This inorganic structure results from stacked brucite layers, Mg(OH)2, where some Mg <sup>2</sup><sup>+</sup> ions can be replaced by Al <sup>3</sup><sup>+</sup> ions, thus generating an excess of positive charge in layers, which must be counterbalanced by the presence of anions, mainly carbonates (CO<sup>3</sup> <sup>2</sup>−), in the interlayer spacing [1–3].

The general formula of LDH is [M(II)1−xM(III)x(OH)2]<sup>x</sup> + [A<sup>n</sup> <sup>−</sup> <sup>x</sup>/n]x−·mH2O, where M(II) are divalent cations (Mg2+, Zn2+, Cu2+, Ni2+) and M(III) are trivalent cations (Al3+, Fe3+, Cr3+). An<sup>−</sup> is an anion of charge *n*, and *m* is the stoichiometric value of co-intercalated water [1].

LDHs can be synthesized through several synthesis methods, such as co-precipitation, sol-gel, hydrothermal or urea hydrolysis, although the most usual method is co-precipitation, which can be performed at variable or constant pH [3].

Focusing only on the catalytic properties of LDHs themselves, these inorganic compounds can display both acid and basic sites [5,6]. However, the amount and strength of these acid/basic sites can be modified by thermal treatment, where dehydroxylation and decarbonation processes cause the collapse of the layered structure, leading to the formation of their respective metal oxides [6]. In addition, if any of M(II) or M(III) cations is easily reducible, it is feasible to obtain catalysts with metallic, acid and basic sites, in such a way that they can find a large spectrum of catalytic applications.

Among the great variety of catalytic processes that have emerged in the last century, the biomass valorization to produce high value-added chemicals is attracting the interest of many research groups, as a sustainable alternative to the use of fossil-based raw materials. This great interest has been prompted by the depletion of fossil fuels, which has led to the search and development of alternative sources that can satisfy both chemical and energy demands [7,8]. Biomass is the only feedstock that can replace fossil fuels, but many efforts are still required for the implementation and integration of these processes in the forthcoming biorefineries. However, the selection of a biomass source must be carried out with care and responsibility, since this biomass could interfere with the food chain, causing serious speculation problems and social imbalances. Taking into account these premises, lignocellulosic biomass, coming from agricultural waste, has emerged as an abundant, sustainable and non-edible source of energy, biofuels and chemicals [9]. Lignocellulose is composed by cellulose (40–50%), hemicellulose (20–35%) and lignin (15–25%), which can be extracted selectively by using several thermal, physical and chemical treatments [10].

Focusing only on the hemicellulose fraction, this can be isolated under mild hydrolysis treatment, leading, after depolymerization and hydrolysis, to the respective monosaccharides, mainly xylose [11,12]. Xylose, in turn, can also be dehydrated through homogeneous and/or heterogeneous catalysts to obtain furfural (FUR) as the main product [12,13]. After bioethanol, furfural is the second most produced compound in the sugar platform. The great interest in this organic compound lies in its chemical structure (an aldehyde group and a furan ring with α,β-unsaturations), which confer it a high reactivity [12]. Thus, furfural can be used as feedstock to give rise to a wide range of products with applications in fields, such as polymers, pharmacy, cosmetics, among others, through hydrogenation, oxidation, dehydration, decarbonylation or condensation reactions [12,14,15]. Thus, for instance, different environmentally friendly processes have been reported aimed at the furfural derivatization for the synthesis of valuable chemicals in aqueous media [16,17], as well as in in eco-sustainable removable media as deep eutectic solvents, by using homogeneous [18] or heterogeneous catalysts [19,20].

Among the products that can be derived from FUR, furfuryl alcohol (FOL) is the most important. It has been estimated that about 62% of FUR production is employed for the synthesis of FOL due to its importance in resin manufacture for the foundry industry and for chemicals [15]. Industrially, FOL has been synthesized through furfural hydrogenation using a copper chromite catalyst [21–24]. Another product obtained in high proportions during FUR hydrogenation is 2-methylfuran (MF) [12], which is also considered a valuable product as a biofuel additive and for the synthesis of heterocycles [15].

Despite the good performance of the commercial copper chromite catalyst, in the last decade, environmental awareness has led to the search for chromium-free catalysts, which can be more sustainable. As alternative, transition metal-based catalysts, e.g., Cu [25–35], Ni [28,35–41] or Pd [28,42,43] have been proposed to replace copper chromite. Generally, these metals have been dispersed on different supports. Both the hydrogenating character of the metal and the acid/base/redox properties of the support play a determining role in the activity and the product pattern, mainly for those processes where FUR hydrogenation takes place in the gas phase.

In this context, the design of LDHs as catalyst precursors, where some of their cations can be easily reducible, is a suitable approach for the development of active catalysts for FUR hydrogenation to obtain high value-added chemicals. Several authors have reported the synthesis of Ni-based hydrotalcites as efficient catalysts for this reaction, although the high hydrogenating nature of Ni sites [28] leads to a wide range of products, including a high proportion of carbonaceous deposits in many cases [35,36]. A suitable alternative could be the use of Cu-based hydrotalcites, since Cu active centers display a lower hydrogenating capacity, in such a way that the selectivity pattern can be easily controlled. In this sense, previous research has evaluated the role of Al source in the synthesis of Cu/ZnO/Al2O<sup>3</sup> from their respective hydroxides, after calcination and reduction [44]. These authors achieved well dispersed Cu centers, which were selective towards FOL [44]. In the present work, a series of [CuZn1−xAlx(OH)2]<sup>x</sup> + [A<sup>n</sup> <sup>−</sup> <sup>x</sup>/n]x−·mH2O hydrotalcites have been synthesized by coprecipitation method, with different Cu/Zn molar ratio, although the M(II)/M(III) molar ratio (M(II) is Cu(II) + Zn(II)) was 3 in all cases, since this value favors the formation of ordered LDH structures [1–3]. Moreover, the effect of other parameters, like aging time, was also evaluated for the synthesis of LDHs. Precursors were calcined and reduced, and then characterized and tested in the FUR hydrogenation, with special emphasis on the correlation between the amount of available metal sites and their catalytic behavior. − − − –

#### **2. Characterization of the Catalysts**

The identification of crystalline phases was carried out by powder X-ray diffraction. In all cases, the XRD patterns of LHDs synthesized (Figure 1) are similar to those reported in the literature [1,45]. The most intense peaks correspond to the typical basal (00*l*) planes, thereby, those located at 2θ ( ◦ ) ≈ 12, 24 and 34 are attributed to the (003), (006) and (009) planes. These can be indexed in a rhombohedral symmetry, which is assigned to a (Cu,Zn)1−xAlx(OH)<sup>2</sup> (CO3)x/<sup>2</sup> m H2O layered double hydroxide (Powder Diffraction File: PDF: 00-37-0629). In addition, less intense peaks ascribed to non-basal planes of hydrotalcites are also observed, together with other low intense peaks, mainly in those LDHs whose Cu <sup>2</sup><sup>+</sup> or Zn <sup>2</sup><sup>+</sup> content is higher. Thus, in the case of Cu-rich LDH, diffraction peaks attributed to copper(II) hydroxide are presented, while Zn-rich LDH also exhibits broad and poorly defined diffraction peaks attributed to zinc hydroxycarbonate. 2θ (° ≈ −

**Figure 1.** XRD patterns of the layered double CuZnAl hydroxides. (Circles: zinc hydroxycarbonate), (diamonds: copper(II) hydroxide).

CuZnAl-based hydrotalcites have been previously studied by FTIR and Raman spectroscopies, demonstrating that the Cu/Zn molar ratio influences their microstructure. Thus, higher Cu/Zn ratios lead to more homogeneous CuZnAl hydrotalcites, while lower values give rise to less homogeneous structures due to its composition is closer to pure ZnAl and CuAl hydrotalcites [46]. This latter could provoke the asymmetric broadening of Bragg peak profiles of LDH for lower Cu/Zn molar ratio, also characteristic of the stacking disorder for Zn-rich compounds (Figure 1). Moreover, this Raman study also confirmed the presence of carbonate species, not only as counterion in the interlayer region of LDH, but also in partial segregated malachite for high Cu/Zn molar ratios, which is in agreement with XRD data [46]. Following the procedure described by Santos et al. for (Cu,Zn)1−xAlx(OH)2(CO3)x/2·mH2O LDHs [46], these layered materials were thermally treated at 300 ◦C for 4 h. In their XRD patterns (Figure 2), a broadening of the diffraction peak in comparison to their respective LDHs is noticeable, as a consequence of the stacking disorder of the layered structure due to decarbonation and dehydroxylation of the LDH, leading to the typical diffraction peaks of their mixed metal oxides [1]. Thus, XRD profiles reveal an evolution from the calcined Cu-rich (P-CuZn\_6) to the Zn-rich (P-CuZn\_0.2) sample. In this sense, the XRD pattern of calcined P-CuZn\_6 displays defined bands located at 2θ ( ◦ ) = 32.5, 35.5, 38.7, 48.8, 53.5, 58.3, 61.6, 66.2 and 68.1 assigned to CuO (PDF: 00-048-1548). The progressive increase in Zn content gives rise to a much more amorphous XRD pattern, being only a broad band observed at 2θ ( ◦ ) ≈ 33. These data suggest that calcined P-CuZn\_2.5, P-CuZn\_1 and P-CuZn\_0.4 display a more amorphous structure, or are formed by smaller particles than that the calcined P-CuZn\_6. For the calcined LDH with the highest Zn content (P-CuZn\_0.2), broad bands at 2θ ( ◦ ) = 31.8, 34.4, 36.2, 48.3, 56.5, 63.1 and 68.2, start to emerge. These peaks are assigned to hexagonal ZnO (PDF: 00-36-1451). On the other hand, no diffraction peaks could be ascribed to the presence of crystalline Al species, in such a way that these species must be amorphous, or dispersed in the other phases. − defined bands located at 2θ ( 2θ (°) ≈ 33. These data suggest that 2θ (°

° **Figure 2.** XRD profiles of the mixed oxides obtained after the calcination of LDHs at 300 ◦C.

→ → ° Once the catalyst precursors were synthesized after subsequently calcined, H2-TPR analysis was carried out to determine the appropriate reduction temperature to obtain the Cu 0 -based catalysts (Figure 3). It has been previously reported that usually it is not possible to discern the different stages of Cu reduction (Cu <sup>2</sup><sup>+</sup> → Cu <sup>+</sup> → Cu 0 ), so the different contributions observed are ascribed to Cu particles with different size and/or with different interaction with the support (ZnO/Al2O3) [26,30,34]. From the H2-TPR profiles, it can be inferred that the sample with the highest Cu content (CuZn\_6) is more easily reducible than the rest of catalysts. Its reduction curve shows two well-defined reduction steps located at 185 and 240 ◦C, which can be explained by the co-existence of CuO particles with different interaction with the support, or particles with different size. As the Cu content decreases and the Zn content concomitantly increases, the H2-TPR profile progressively changes. Thus, it seems clear that the hydrogen consumption at low temperature progressively decreased, as observed for CuZn\_2.5 and CuZn\_1. In the case of CuZn\_0.4, this low temperature band disappears, being only noticeable the presence of a single band whose maximum is located about 245 ◦C. Those catalysts with a lower proportion of Zn seem to be more reducible than the richer ones, thus indicating that the presence of ZnO would modify the electronic environment around CuO particles, as previously pointed out by other authors [47,48]. Previous research has reported that the reduction of Cu species in CuO/ZnO takes place about 210 ◦C [30], so Al species could exert an additional electronic promoter effect on the interaction between ZnO and CuO particles, in such a way that the reduction of Cu <sup>2</sup><sup>+</sup> species occurs at

higher temperature, as was indicated in the literature [49]. On the other hand, as the catalyst with a lower Zn content is easily reducible, it is expected that Al species possess a less pronounced promoter effect than ZnO.

**Figure 3.** H<sup>2</sup> -TPR profiles of the mixed metal oxides obtained after calcination of LDHs.

° H<sup>2</sup> consumption data of calcined P-CuZn\_X materials are always lower than theoretical values, so a fraction of Cu <sup>2</sup><sup>+</sup> species could not be reduced (Table 1). In this sense, several authors have reported that the reduction of segregated metal carbonates can take place at high temperature, about 550 ◦C [46]. However, other authors have noted that Cu <sup>2</sup><sup>+</sup> ions could also be embedded in the octahedral sites of Al2O3, ZnO, or even ZnAl2O4, whose reduction should require a higher temperature [50,51].


**Table 1.** Chemical composition and H<sup>2</sup> consumption of mixed metal oxides obtained by calcination of LDHs.

Taking into account the H2-TPR profiles, the precursors were reduced at 300 ◦C, maintaining this temperature for 1 h to ensure the complete reduction of Cu <sup>2</sup><sup>+</sup> species.

° 2θ (° XRD profiles of the CuZn\_X catalysts show the typical diffraction peaks of the metallic Cu <sup>0</sup> at 2θ ( ◦ ) = 43.3 and 50.4 (PDF: 00-85-1326) (Figure 4). The presence of Cu2O should be discarded since, despite the main peak at 2θ ( ◦ ) = 38.3 overlaps with a diffraction peak of ZnO, the secondary peaks of Cu2O do not appear, so partially reduced copper oxide (Cu2O) crystallites are not detected, or these are too small to be detected by XRD.

despite the main peak at 2θ (° at 2θ (° These signals are more defined than those corresponding to the hexagonal ZnO, which does not evolve to reduced Zn species. In the case of Al species, diffractograms do not reveal any characteristic peaks, so these Al species must be highly dispersed in the catalysts. The determination of the crystallite size for Cu <sup>0</sup> was carried out from the Williamson-Hall equation [52], using the main diffraction peak at 2θ ( ◦ ) = 43.3. The analysis of this (111) crystallographic plane reveals that the crystallinity of the Cu<sup>0</sup> particles increases directly with the Cu content, from 5.0 nm for the lowest content (CuZn\_0.2) to 28.8 nm for the highest (CuZn\_6). These value are slightly lower than those observed for Cu/ZnO or Cu/MgO synthesized by the co-precipitation method [26,30], so Al species, in addition of exerting an electronic promoter effect, seem to favor the dispersion of Cu nanoparticles [49,53].

**Figure 4.** XRD patterns of the CuZn\_X catalysts.

In order to elucidate the catalyst morphology, CuZn\_X catalysts were analyzed by TEM (Figure 5). In all cases, the micrographs allow to distinguish different morphologies, including layered structures probably due to the hydrotalcite structure that did not fully collapse after the thermal treatment. These data are in agreement with the literature, where a proportion of the layered structure was reported to be maintained after the thermal treatment [54]. In addition, it is noticeable the presence of pseudospherical and well-dispersed nanoparticles, mainly in the samples with a higher Cu content. In both cases, these particles are very small, below 15 nm.

**Figure 5.** TEM micrographs for CuZn\_0.2, CuZn\_1 and CuZn\_6. (Scale bar: 50 nm).

The analysis of these samples was also performed by Energy Dispersive X-Ray (EDX) (Figure 6). The images show that all elements (Cu, Zn, Al and O) are well dispersed in both lamellar and pseudospherical structures. In Figures 5 and 6, it can be seen the existence of interparticle voids between adjacent nanoparticles.

The textural properties of CuZn\_X catalysts were evaluated from their N<sup>2</sup> adsorption-desorption isotherms at −196 ◦C (Figure 7A). According to the International Union of Pure and Applied Chemistry (IUPAC) classification [55], these isotherms can be considered as Type II, which are typical of macroporous solids, as suggests the great growing of N<sup>2</sup> adsorbed at high relative pressure. The shape is similar for all CuZn\_X catalysts, and, consequently, the modification of the Cu/Zn molar ratio does not seem to affect the textural properties of catalysts (Table 2). Thus, SBET values hardly vary, being between 70 and 85 m<sup>2</sup> g −1 , while *t*-plot data indicate that the surface ascribed to the microporosity can be considered as negligible, since values are below 10 m<sup>2</sup> g −1 in all cases. In the same way, the pore volume is very similar for all catalysts, being in the range 0.474–0.679 cm<sup>3</sup> g −1 , while the micropore volume is very low in comparison to the total volume.

°

**perficial** 

**(μmol g**

**Figure 6.** EDX images of CuZn\_0.2, CuZn\_1 and CuZn\_6. (Scale bar: 200 nm).

−

**−**

– <sup>−</sup>

**(μmol g −**

−

− ° **Figure 7.** N<sup>2</sup> adsorption-desorption isotherms at −196 ◦C (**A**) and pore size distribution estimated by DFT method (**B**) for CuZn\_X catalysts.

**−1 −1 −1**


**Table 2.** Textural properties of the CuZn\_X catalysts.

<sup>1</sup> Micropore volume estimated from alpha-s-plot. <sup>2</sup> Quantification of acid sites by NH3-TPD. <sup>3</sup> Quantification of acid sites by CO2-TPD.

With regard to the pore size distribution, estimated by Density Functional Theory (DFT) [56] (Figure 7B), all CuZn\_X catalysts follow the same pattern. The microporosity is barely appreciated, and only a small contribution appears at 1.2 nm. However, these samples exhibit a wide pore size distribution, extending from 3 to 150 nm, in such a way that these catalysts can be labeled as mesoand mainly macroporous. This porosity is ascribed to the voids between adjacent particles, as was suggested previously from the TEM micrographs (Figures 5 and 6).

The quantification of surface Cu<sup>0</sup> species of CuZn\_X catalysts was carried out by N2O titration at 60 ◦C [57] (Table 3). The data indicate that dispersion decreases with the Cu content, going from 21% for CuZn\_0.2 to 7% for CuZn\_6. In the same way, the metallic surface area (m<sup>2</sup> Cu gCu −1 ) follows a similar trend, and the catalyst with the lowest Cu content (CuZn\_0.2) displays the highest value. This study also reveals that the metal particle size increases directly with the Cu loading, from 5 nm for CuZn\_0.2 to 19 nm for CuZn\_6, being these data slightly lower than those obtained from XRD by using the Williamson-Hall method [52].


**Table 3.** Metallic properties of CuZn\_X catalysts, as determined from N2O titration.

<sup>1</sup> By applying the Williamson-Hall equation to XRD data [52].

In order to determine the surface chemical composition of catalysts, X-ray photoelectron spectroscopy (XPS) analysis was carried out. Cu 2p core level spectra (Figure 8A) show that Cu 2p3/<sup>2</sup> consists of a single contribution located at 932.0 eV, ascribed to reduced Cu species [30,32], since the absence of the typical shake-up satellite of divalent metals about 942–943 eV would exclude the existence of Cu(II) [58].

However, from this contribution, it is not possible to discern between Cu<sup>+</sup> and Cu<sup>0</sup> species, so the Auger CuLMM signal is used to identify these oxidation states [30,32]. The broad Auger CuLMM band (Figure 8B) can be deconvoluted in two main contributions: 918.5 eV, which is ascribed to Cu<sup>0</sup> , and 917.0 eV due to Cu<sup>+</sup> [30]. The proportion of Cu<sup>0</sup> increases directly with the Cu content, being about 65–70% of the total reduced Cu species. The analysis of the Zn 2p region (Table 4) evidences a single band between 1021.6–1022.1 eV, which is assigned to ZnO. Some studies of Cu/ZnO catalysts have demonstrated that ZnO can be partially reduced, arising a new contribution located at lower binding energy. This new band is not observed in the Zn 2p core level spectra; however, the Auger ZnLMM line (Figure 8C) shows an asymmetric band that would confirm the existence of a small proportion of Zn partially reduced (Znδ+), or even a synergistic effect between Cu<sup>0</sup> and ZnO [59,60]. With regard to the Al 2p core level spectra, all CuZn\_X catalysts display a contribution about 74.3 eV, which can be attributed to Al2O<sup>3</sup> [61]. In the case of O 1s, all catalysts display a main contribution located about 531.0 eV, assigned to oxide species, and another less intense at 532.6 eV, that is assigned to hydroxyl and/or carbonate [61]. The existence of carbonates could be confirmed by the signal at a binding energy of 288–289 eV in the C 1s core level spectra [61], which could result from a low proportion of LDH, or a possible carbonation of the catalyst surface. –

**Figure 8.** Cu 2p core level (**A**), Auger CuLMM (**B**) and Auger ZnLMM (**C**) spectra of CuZn\_X catalysts.


**Table 4.** XPS data of CuZn\_X catalysts.

The atomic concentration data show that the surface Cu content progressive increases with the Cu loading, as expected (Table 4), although surface molar ratios seem to be lower than the theoretical values. This fact could be ascribed to both ZnO and Al2O<sup>3</sup> exhibit a smaller crystallite size than Cu<sup>0</sup> , in such a way that both metal oxides should be better dispersed in catalysts. This fact can lead to lower than expected surface Cu concentrations.

As both acid and basic sites could also influence on the gas-phase FUR hydrogenation reaction, NH3-TPD and CO2-TPD studies were performed (Table 2). The NH3-TPD data show a relatively low amount of acid sites, decreasing from 78 µmol g−<sup>1</sup> for CuZn\_0.2 to 60 µmol g−<sup>1</sup> for CuZn\_6. The presence of these acid sites is ascribed to Al2O<sup>3</sup> and Cu+, which provide Lewis acid sites, and even to ZnO due to its amphoteric character, similar to alumina. Several authors have reported that the modulation of the acidity plays a key role in FUR conversion, since a high number of acid sites can cause the polymerization of FUR, mainly in gas-phase, leading to the formation of a high proportion of carbonaceous deposits [29,37,62]. These polymerized FUR species interact strongly with the active sites, in such a way that catalysts tend to be deactivated relatively fast. However, the presence of weak acid sites could exert a beneficial effect on the catalytic behavior, since the interaction between FUR, or reaction products, and the catalyst is weakened in comparison with catalyst with higher acidity, thus favoring the desorption of products adsorbed on the catalyst surface. In this sense, several supports, such as SiO<sup>2</sup> or clay minerals, with similar amount of acid sites [29,37], have provided good catalytic activity in FUR hydrogenation, under similar experimental conditions.

Similarly, a basic support (MgO, CaO, ZnO or CeO2) [25,26,30,63,64] also seems to have a positive effect in this catalytic process. CuZn\_X catalysts display a small concentration of basic sites, between 15 and 39 µmol g−<sup>1</sup> , raising as the Cu content decreases. The presence of these basic sites is ascribed to the amphoteric character of both ZnO and Al2O3, as previously mentioned [65,66].

#### **3. Catalytic Results**

CuZn\_X catalysts were tested in the gas-phase FUR hydrogenation, using FUR dissolved in cyclopentyl methyl ether (CPME). This solvent was selected due to its interesting physico-chemical properties, such as low solubility in H2O in comparison to other ethereal solvents, low formation of peroxides, low boiling point (106 ◦C) or relative high stability under acid or basic conditions [67]. Considering these premises, the first study was to evaluate the stability of the solvent under similar experimental conditions to those used for the catalytic evaluation of CuZn\_X catalysts. At a reaction temperature of 190 ◦C, CPME was recovered without any modification, such as the cleavage of the –C–O–C– bond or isomerization.

#### *3.1. Influence of Cu Content*

between 15 and 39 μmol g<sup>−</sup>

Thereafter, CuZn\_X catalysts were studied in the FUR hydrogenation at 190 ◦C (Figure 9). All the catalysts are prone to suffer a progressive deactivation, which could be a consequence of the strong interaction of FUR, or FOL molecules, with Cu active sites [68]. The catalyst with the lowest Cu content (CuZn\_0.2) is also the least active catalyst, reaching a FUR conversion of 40% after 1 h of time-on-stream (TOS), which decreases until 26% after 5 h. By increasing the Cu content, the catalytic performance is clearly improved, since CuZn\_0.4 attains a FUR conversion of 43% after 5 h of TOS, while CuZn\_1 seems to be the most stable catalyst with a FUR conversion of 59%. However, this trend is not followed when the Cu content raises further, since both CuZn\_2.5 and CuZn\_6 catalysts exhibit a similar pattern, reaching a FUR conversion close to 50% after 5 h of TOS.

–

°

− − **Figure 9.** FUR conversion (**A**), FOL yield (**B**), and MF yield (**C**) in the FUR hydrogenation over CuZn\_X catalysts (Experimental conditions: Mass of catalyst = 0.15 g, Reaction temperature = 190 ◦C, Pressure = 0.1 MPa, H<sup>2</sup> flow = 10 mL min −1 , Fed flow = 2.3 mmol FUR h −1 ).

With regards to the selectivity pattern (Figure 9B,C), the modification of the Cu/Zn molar ratio hardly varies the ratio of the obtained products, since FOL always is the main compound, with a maximum yield of 55% for CuZn\_1, after 5 h of TOS. In all cases, MF was also detected, although this can be considered a minority because MF yield did not exceed 6% after 5 h of TOS. The catalytic data can be compared with other catalysts reported in the literature. Thus, Cu/ZnO and Cu/CeO<sup>2</sup> catalysts exhibited a high activity in the FUR hydrogenation, but the selectivity pattern is slightly different, since the amount of MF was higher in both cases [30,63]. Other authors have reported that the incorporation of Al2O<sup>3</sup> exerts a promoter effect in the catalytic behavior due to Al species weaken the Cu-ZnO interaction [44]. This data could be in agreement with those obtained in the H2-TPR analysis, where the catalyst with a lower ZnO content (CuZn\_6) is easily reducible. In earlier studies, Nagajara et al. established that co-precipitation is the most appropriate method to obtain a high dispersion of Cu species on a basic support like MgO [25]. These authors also reached a high activity in FUR conversion, with a high selectivity mainly towards FOL. In the same way, Dong et al. carried out a comparative study with Cu/ZnO, Cu/Al2O<sup>3</sup> and Cu/SiO<sup>2</sup> [29]. They concluded that the modification of the electronic density in the catalyst due to the Cu-support interaction has a determining role in the catalytic behavior, pointing out that the existence of weak acid sites favors MF formation, mainly in the case of Cu/SiO2, as a consequence of consecutive reactions (FUR → FOL → MF) [29]. These weak acid sites, associated to both the support and Cu <sup>+</sup> sites, could be involved in the hydrogenolysis of FOL to MF due to their electron-deficiency [29,69]. Similar results were obtained by Jiménez-Gomez et al., who synthesized Cu/SiO<sup>2</sup> catalyst by a complexation method to disperse Cu species [32]. They observed that, besides the presence of weak acid sites, it is necessary the existence of well-dispersed Cu nanoparticles, since the

electron-deficiency of the catalyst enhanced [29,32]. In the present work, in spite of CuZn\_X catalysts have acid sites, the amount of MF can be considered negligible in comparison to data reported in the literature [29,32]. In this sense, the catalysts highly selective towards MF display higher SBET values. This fact supposes a higher FUR-support interaction and consequently a longer residence time, which can favor consecutive reactions. CuZn\_X catalysts showed lower SBET values and reduced porosity (Table 2 and Figure 7), mainly ascribed to their meso-macroporous nature. As the pore size is not modulated to the dimensions of the FUR molecule, the interaction FUR-active sites should be weaker, although this fact also has some advantage related to the presence of more labile sites that favor an easier desorption of FUR and products. This trend was also observed for other catalysts synthesized by co-precipitation method, such as Cu/MgO, Cu/ZnO or Cu/CeO<sup>2</sup> [25,30,63]. The existence of Cu<sup>+</sup> sites, as determined by XPS, does not favor the hydrogenolysis process to form MF, so it could be inferred that textural properties have a more determining role in the hydrogenolysis reaction than the amount of acid sites. However, the hydrogenation FUR → FOL takes place on the metallic sites [28,29] and it does not seem to be affected by the textural properties of catalysts. This fact was confirmed in Cu-based catalysts supported on clay minerals [33,37,70]. Thus, those catalysts with very poor textural properties, like bentonite, or whose pore dimensions hinder the incorporation of Cu nanoparticles, like sepiolite, only dispersed metal nanoparticles on the catalyst surface [33]. In both cases, the catalysts were selective towards FOL. However, the use of a clay mineral with less crystallinity as support led to higher SBET values due to an increase in the micro- and mesoporosity, thus favoring a higher interaction between FUR and the catalyst and longer residence time, in such a way that the reaction can evolve toward consecutive reaction steps [70]. In this sense, the formation of MF is favored by the existence of weak acid sites associated to the aluminosilicate, used as support, (kerolite) and the presence of a small fraction of Cu<sup>+</sup> species. In these studies, the incorporation of promoters, such as ZnO, MgO or CeO2, modified the selectivity pattern, with an increase in the FOL yield due to milder interaction between support and Cu species, as indicated their H2-TPR data [29,37,70], which is in agreement with data reported in the present work.

#### *3.2. Turnover Frequency (TOF)*

In order to evaluate the activity of Cu sites, turnover frequency (TOF) data of catalysts were determined. It must be considered that most of catalysts are prone to suffer deactivation by the formation of carbonaceous deposits on the catalyst surface, and consequently the evaluation of the activity of these Cu sites only makes sense at zero time (t0), as long as these catalysts do not reach a total conversion at t0. In this sense, the linearization of the Figure 9A reveals that FUR conversion values vary between 43% for CuZn\_0.2 and 76% for CuZn\_2.5, at t0.

From the extrapolation to t<sup>0</sup> (Figure 10A), it is feasible to determine the corresponding TOF values. The obtained data reveal that CuZn\_0.2 displays the lowest TOF (28 h−<sup>1</sup> ), whereas an increase in the Cu content improves TOF, and CuZn\_0.4 and CuZn\_1 show TOF values of 43 and 44 h−<sup>1</sup> respectively. However, the use of larger Cu loadings worsens the efficiency of Cu active sites, since CuZn\_2.5 and CuZn\_6 reach TOF values of 36 and 37 h−<sup>1</sup> , respectively. In the same way, from the linearization of the data obtained in Figure 9A, it is possible to evaluate the amount of FUR converted per time unit and mass of catalyst at t<sup>0</sup> (Figure 10B). These data reveal that the highest FUR rate was 1.113·10−<sup>2</sup> mmol FUR g−<sup>1</sup> h −1 for CuZn\_1, while the poorest value was obtained for CuZn\_0.2, with a rate of 0.678·10−<sup>2</sup> mmol FUR g−<sup>1</sup> h −1 . In addition, from Figure 10B, an estimation of the catalyst lifetime can be made by extrapolation. These data reveal that the catalyst with longer life is CuZn\_1, which could be active for 15.15 h. In this time, the catalyst could convert 0.1537 mmol FUR g−<sup>1</sup> before exhaustion.

− − −

− −

−

−

**Figure 10.** Linearization of TOF values (**A**) and FUR consumption rate (**B**) with time-on-stream for CuZn\_X catalysts.

#### *3.3. Influence of Reaction Temperature*

° ° ° ° ° Considering that CuZn\_1 showed the longest lifetime, as well as the most efficient Cu active sites (Table 5), this catalyst was selected to evaluate the influence of the reaction temperature on the catalytic performance (Figure 11). FUR conversion clearly increases with the reaction temperature (Figure 11A), from 34% to 79% after raising the temperature from 170 to 230 ◦C, after 5 h of TOS. Several authors have reported that FUR hydrogenation is thermodynamically favored; however, the FUR conversion often displays a volcano shape with the reaction temperature, reaching a maximum conversion at 190–210 ◦C [27,62]. The worsening of the conversion at higher reaction temperature was attributed to FUR polymerization, due to its strong adsorption on the active sites involved in the hydrogenation process. This trend was not observed in the present work, which could be explained by the mesoand macroporosity of catalysts and the relatively low concentration of acid sites, in such a way that the FUR-active site interaction is weaker, so the desorption of FUR and reaction products should be easier [29,37,70]. With regard to the hydrogenation products (Figure 11B,C), in all cases, the main product was FOL, reaching a maximum yield of 65% after 5 h of TOS at 210 ◦C. MF yield values always were very low, although they increase slightly with the reaction temperature, attaining a maximum MF yield of 7% after 5 h of TOS at 230 ◦C [62]. Several authors have pointed out that the formation of MF is favored at higher reaction temperature (T > 190 ◦C) [25,30,62]. However, CuZn\_1 hardly modifies its selectivity pattern when the reaction temperature increases, probably due to the weaker interaction between the active sites and FUR molecules. Thus, this catalyst seems to be highly stable and selective toward FOL, along the TOS, in a wide range of reaction temperatures [15].



<sup>1</sup> Turnover frequency determined at t<sup>0</sup> (metallic surface data were obtained from Table 3). <sup>2</sup> Conversion of FUR at t0. <sup>3</sup> FUR rates at t0. <sup>4</sup> Total exhaustion time of catalysts. <sup>5</sup> Total FUR converted until catalyst exhaustion. <sup>6</sup> FUR converted between 1 and 5 h of TOS.

**−1 −1 −1**

**− − − − −**

**−1**

**– −1**

− − **Figure 11.** FUR conversion (**A**), FOL yield (**B**), and MF yield (**C**) in the FUR hydrogenation with CuZn\_6 (Experimental conditions: Mass of catalyst = 0.15 g, Pressure = 0.1 MPa, H<sup>2</sup> flow = 10 mL min −1 , Fed flow = 2.3 mmol FUR h −1 ).

#### *3.4. Influence of Aging Time in the Synthesis*

In the next study, the aging time of the LDHs was modified to evaluate its influence on the catalytic behavior. Thus, three CuZn\_1 catalysts with different aging time (1, 48 and 168 h) were compared. The catalytic data show that longer aging time in the synthesis of LDHs has an adverse effect on the catalytic activity (Figure 12A). Thus, that LDH whose aging time was only 1 h gave rise to a catalyst that hardly undergoes deactivation, since it maintains a FUR conversion of 84% after 5 h of TOS at 210 ◦C. However, a more prolonged aging treatment (168 h) worsens the catalytic activity, achieving a FUR conversion of only 37%, under similar experimental conditions. The selectivity (Figure 12B,C) follows the same trend than that observed previously for CuZn\_X catalysts, since FOL was the main product in all cases, attaining a maximum yield of 71%, after 5 h of TOS at 210 ◦C, with CuZn\_1 aged for 1 h. The amount of MF was very low in all cases, although a decrease in MF yield took place as aging time was increased.

– − − **Figure 12.** FUR conversion (**A**), FOL yield (**B**), and MF yield (**C**) in the FUR hydrogenation over CuZn\_1 catalysts synthesized with different aging time (1–168 h) (Experimental conditions: Mass of catalyst = 0.15 g, Pressure = 0.1 MPa, H<sup>2</sup> flow = 10 mL min −1 , Fed flow = 2.3 mmol FUR h −1 ).

In order to understand the changes of the catalytic behavior for the CuZn\_1 catalysts synthesized with different aging times, these catalysts were characterized. Their XRD patterns show clear differences in crystallinity (Figure 13A). Thus, the diffractogram of sample prepared with the shortest aging time (CuZn\_1(1h)) does not display any diffraction peak, which are clearly visible for catalysts obtained by increasing the aging time. Thus, CuZn\_1(168h) catalyst shows Cu crystallites of 33 nm. These data are in agreement with the literature, since the use of higher temperature and longer aging time favor the growing and better crystallization of layered double hydroxides [3,71]. This fact implies the formation of catalysts where nanoparticles of support and active phase are larger.

− −

–

**Figure 13***. Cont.* **Figure 13.** XRD (**A**) and H<sup>2</sup> -TPR (**B**) profiles of CuZn\_1 catalysts synthesized with different aging time.

The analysis of the H2-TPR profiles also reveals clear modifications depending on the aging time (Figure 13B). Considering that the three catalysts display the same chemical composition, it is expected that the profile differences should be attributable to differences in the size of their Cu nanoparticles. In this sense, a catalyst whose Cu crystallite sizes are smaller (CuZn\_1(1h)) is more easily reducible than those catalysts with bigger sizes, as a consequence of the weaker interaction between small CuO nanoparticles and the ZnO/Al2O<sup>3</sup> support.

In addition, it is also noticeable that both CuZn\_1(1h) and CuZn\_1(48 h) catalysts display two defined peaks, which could point out the formation of Cu nanoparticles with two different sizes. However, CuZn\_1(196h) only shows a peak at higher temperature. This fact could be ascribed to the use of longer aging time, which favors the formation of more homogenous structures. N2O titration data (Table 6) follow the trend observed by XRD and H2-TPR, since the catalyst with the lowest aging time (CuZn\_1(1h)) is also the catalyst with a higher amount of available surface Cu sites. This would confirm the formation of smaller particles and consequently a higher metallic surface area is obtained, which is in agreement with its better catalytic performance.

**(μmol g − − − Table 6.** Metallic characteristic of CuZn\_1 catalysts synthesized with different aging time. determined from N2O titration.


<sup>1</sup> Copper dispersion. <sup>2</sup> Crystal size estimated by Williamson-Hall equation from XRD data [52].

–

**(%)** 

– – Finally, the superficial analysis of these catalysts by XPS (Table 7) does not reveal any variation in core level spectra in comparison to CuZn\_X catalysts. It is only noteworthy a decrease in the surface Cu content with the aging time, which can be explained by the formation of larger Cu nanoparticles, thus limiting the amount of Cu detected by XPS due to this technique only allows to analyze 2–3 nm of depth (Table 6). With regard to the oxidation state of Cu species, Cu<sup>0</sup> is the main oxidation state, about a 70–75% of the total Cu, while the Cu<sup>+</sup> is about 25–30%.


**Table 7.** Surface atomic concentration of CuZn\_1 catalysts synthesized with different aging time.

#### **4. Materials and Methods**

#### *4.1. Preparation of Catalysts*

A series of Cu/ZnO/Al2O<sup>3</sup> catalysts has been synthesized by co-precipitation method, according to the methodology described by Santos et al. [46]. The LDH precursors were synthesized from aqueous solutions of Cu(NO3)2·2H2O (99%, Aldrich, Saint Louis, MI, USA), Zn(NO3)2·6H2O (99%, Aldrich) and Al(NO3)3·9H2O (99%, Aldrich) with a total metal concentration of 0.3 M. In all cases, (Cu2++Zn2+)/Al3<sup>+</sup> molar ratio was 3, whereas the Cu2+/Zn2<sup>+</sup> molar ratio was varied between 0.2 and 6. In the next step, an aqueous solution of Na2CO<sup>3</sup> (1.0 M) was slowly added to precipitate metal hydroxides and form the LDH structure. Later, the obtained gel was aged at room temperature for 48 h (in two different synthesis, the gel was aged 1 and 168 h). The obtained solid was filtered and washed with distilled water until reach a neutral pH to ensure the total removal of Na<sup>+</sup> ions. Finally, the solid was dried overnight at 90 ◦C and calcined at 300 ◦C for 4 h, using a ramp of 10 ◦C min−<sup>1</sup> .

#### *4.2. Characterization of Catalysts*

An X´Pert Pro automated diffractometer (PANanalytical, Bruker, Rheinstetten, Germany) was used to obtain powder X-ray diffraction patterns, this equipment are composed of a Ge (111) primary monochromator (Cu Kα1) and a X´Celerator detector with a step size of 0.017◦ (2θ), between 2θ = 10◦ and 70◦ with an equivalent counting time of 712 s per step. Williamson-Hall Equation (1) [52] was applied to calculate the crystallite size (D):

$$\mathbf{B}\cos\Theta = (\mathbf{K}\mathbf{\lambda}/\mathbf{D}) + (\mathbf{2}\,\varepsilon\sin\Theta)\tag{1}$$

where B is the full width at half maximum (FWHM) of the XRD peaks, θ the Bragg angle, B is, K is the Scherrer constant, λ is the wavelength of the X ray and ε is the lattice strain.

A FEI Talos F200X (Thermo Fisher Scientific, Waltham, MA, USA) system was used to study the catalyst morphology by transmission electron microscopy (TEM). The images obtained with this technique show a high resolution; moreover, this equipment allows the combination of STEM and TEM imaging and, therefore, a 3D characterization with chemical composition mapping. The samples were dispersed in isopropyl alcohol and each one of them was put on a carbon grid.

The temperature of catalysts reduction was evaluated by H2-TPR (hydrogen temperatureprogrammed reduction). To carry out the experiments, 0.080 g of sample has been used. First, the catalyst precursor is treated under a He flow (35 mL min−<sup>1</sup> ) at 100 ◦C for 30 min, after this, it is cooling to room temperature to start the analysis, where the H<sup>2</sup> consumption is monitored between 50 and 800 ◦C, by using an Ar/H<sup>2</sup> flow (48 mL min−<sup>1</sup> , 10 vol.% H2) with a heating rate of 10 ◦C min−<sup>1</sup> . An on-line thermal conductivity detector (TCD) (Shimadzu, Kioto, Japan) was used to carry out the H<sup>2</sup> quantification. It is necessary to trap the water formed in the process to avoid equipment contamination, so the outcoming flow is passed through a cold finger immersed in a liquid N2/isopropanol bath (−80 ◦C).

In previous research [30,37], N2O titration has been employed to determine the metallic surface area and the dispersion of Cu<sup>0</sup> species, where the superficial oxidation of Cu<sup>0</sup> under a N2O flow takes place according to the Equation (2):

$$\text{2Cu}^{0} + \text{N}\_{2}\text{O} \rightarrow \text{Cu}\_{2}\text{O} + \text{N}\_{2} \tag{2}$$

Prior to the analysis, the catalyst precursor was treated under a He flow (35 mL min−<sup>1</sup> ) at 100 ◦C for 30 min, followed by a reduction step under a 10 vol.% H2/Ar-flow (48 mL min−<sup>1</sup> ) at 300 ◦C for 1 h, with a heating rate of 5 ◦C min−<sup>1</sup> , being monitored the H<sup>2</sup> consumption by TCD. When the catalyst was reduced, it was cooled until 60 ◦C under a He flow to carry out the Cu<sup>0</sup> oxidation to Cu<sup>+</sup> performed by titration N2O (5 vol.% N2O/He), at 60 ◦C for 1 h. At last, a second sample reduction was carried out by heating from room temperature to 300 ◦C with a heating rate of 5 ◦C min−<sup>1</sup> , being also monitored its H<sup>2</sup> consumption by TCD.

The textural parameters were determined from the N<sup>2</sup> adsorption-desorption isotherms at <sup>−</sup><sup>196</sup> ◦<sup>C</sup> by using an automatic ASAP 2020 Micromeritics apparatus (Micrometrics, Norcross, GA, USA). The sample was previously outgassed at 200 ◦C and 10−<sup>4</sup> mbar for 12 h. The Brunauer-Emmet-Teller (BET) equation [72] was utilized to determine surface area taking a N<sup>2</sup> cross section of 16.2 Å<sup>2</sup> . Micropore volume was determined from alpha-s- method. Pore size distribution was determined using the density functional theory (DFT) [56].

X-ray photoelectron spectra were obtained using a Physical Electronic PHI 5700 spectrometer (Physical Electronics, Eden Prairie, MN, USA), equipped with an Electronics 80-365B multichannel hemispherical electron analyzer and an Mg Kα X-ray excitation source (300 W, 15 kV, hv = 1253.6 eV). High-resolution spectra were recorded by a concentric hemispherical analyzer in a 29.35 eV constant energy mode, using a 720 µm diameter analysis area, and the pressure in the analysis chamber was kept below 5 × 10−<sup>6</sup> Pa. Binding energies (BE) were determined to an accuracy of ± 0.1 eV, using the adventitious carbon C 1s signal at 284.8 eV as reference. PHI ACCESS ESCA-F V6 software (Eden Prairie, Minnesota, USA) was used for data acquisition and analysis. A Shirley-type background was subtracted from the signals. The recorded spectra were always analyzed with Gauss-Lorentz curves, in order to determine more precisely the binding energy of the atomic levels of the different elements. To avoid sample oxidation after reduction, they were stored in sealed vials with an inert solvent; moreover, samples were prepared in a dry box under a N<sup>2</sup> flow, where the solvent was evaporated prior to its introduction into the analysis chamber, and directly analyzed without previous treatment.

To calculate the total acid sites present in the samples, thermo-programmed desorption of ammonia (NH3-TPD) was carried out. In a typical procedure, 0.08 g of catalyst is placed in a U-shape quartz reactor and cleaned flowing He (40 mL·min−<sup>1</sup> ) up to 400 ◦C with a heating rate of 10 ◦C·min−<sup>1</sup> . Then, the sample is cooled until 100 ◦C under the same He flow, and once the temperature is stabilized at <sup>100</sup> ◦C, the sample is saturated with ammonia for 5 minutes and then physisorbed NH<sup>3</sup> was removed under He. Ammonia desorption is performed by heating the sample from 100 to 400 ◦C, with a rate of 10 ◦C·min−<sup>1</sup> , registering the signal using a GC-14B instrument (Shimadzu, Kioto, Japan) equipped with a thermal conductivity detector (TCD), previously calibrated with Ni(NH3)6Cl<sup>2</sup> (Aldrich) in order to quantify total acid sites.

Thermo-programmed desorption of CO<sup>2</sup> (CO2-TPD) was employed for the quantification of basic sites. In a typical procedure, 0.03 g of catalyst is pretreated under a He flow (40 mL min−<sup>1</sup> ) at 400 ◦C for 15 min (10 ◦C·min−<sup>1</sup> ). Later, the sample was cooled to 100 ◦C and a pure CO<sup>2</sup> stream (60 mL min−<sup>1</sup> ) was subsequently introduced into the reactor for 30 min. Finally, the amount of CO<sup>2</sup> evolved was analyzed using a TCD detector between 100 and 600 ◦C and a helium flow (10 ◦C min−<sup>1</sup> ).

#### *4.3. Catalytic Tests*

The furfural hydrogenation was carried out at atmospheric pressure in a tubular quartz reactor with an internal diameter of 6.35 mm. For this, 150 mg of pelletized catalyst (325–400 µm) were placed inside the reactor between two layers of quartz wool, which is placed inside a programmable temperature tubular furnace, controlled with a thermocouple. Before starting the reaction, catalysts were reduced in-situ with a hydrogen flow (99.99%, Airgas, Paris, France) of 60 mL min−<sup>1</sup> for 1 h at the reduction temperature deduced from their corresponding H2-TPR profiles (350 ◦C). Subsequently, the desired reaction temperature is set and once the system is stable, the reaction is carried out under a flow of H<sup>2</sup> that ranges between 10 and 60 ml min−<sup>1</sup> and a feed flow of 3.87 mL h−<sup>1</sup> of a furfural solution in cyclopentyl methyl ether (CPME, 5 vol.%), which is introduced with the help of an HPLC piston pump, model 307 SC-10 (Gilson, Middleton, WI, USA).

To avoid problems, such as blockage of the lines in the equipment, furfural was dissolved in cyclopentyl methyl ether (CPME). This solvent is environmentally friendly and has been used in different organic reactions [30]. Reaction samples were collected every hour, dissolved with a chloroform and o-xylene solution (internal standard), stored in sealed vials to be subsequently analyzed by gas chromatography, using a Shimazu GB-14A chromatograph equipped with a flame ionization detector and a CP-Wax 52 CB capillary column.

In a preliminary test, the CuZn\_1 catalyst was chosen to evaluate the stability of this solvent at 190 ◦C, after 5 h of time-on-stream (TOS), but in the absence of furfural. The analysis of the collected samples confirmed the full recovery of CPME; without any products different from this ether, thus demonstrating its stability. The conversion, selectivity and yield values were determined using the following expressions:

$$\text{Conversion } (\%) = \frac{\text{mol of furfural converted}}{\text{mol of furfural fed}} \times 100\tag{3}$$

$$\text{Selectivity } (\%) = \frac{\text{mol of the product}}{\text{mol of further converted}} \times 100\tag{4}$$

$$\text{Yield } (\%) = \frac{\text{mol of the product}}{\text{mol of furtheral feed}} \times 100\tag{5}$$

The turnover frequency was calculated as follows (6):

$$\text{TOF} = -\frac{\ln\left(1 - \chi\right)}{\left(\frac{W}{\mathcal{F}}\right) \times \mathbf{M}}\tag{6}$$

where F is the molar rate of furfural (mol h−<sup>1</sup> ), W is the catalyst weight (g), X is the conversion and M is the mole of sites loaded (mol g−<sup>1</sup> ). This equation, in which –ln (1 − X) substitutes for X assumes a pseudo first-order reaction which may be justified by the excess of hydrogen [63].

#### **5. Conclusions**

Several LDHs with chemical composition (Cu,Zn)1−xAlx(OH)2(CO3)x/2·mH2O have been synthesized by the co-precipitation method, maintaining a (Cu2<sup>+</sup> + Zn2+)/Al3<sup>+</sup> molar ratio of 3, and varying the Cu2+/Zn2<sup>+</sup> molar ratio between 0.2 and 6.0. After calcination and reduction steps, Cu/ZnO/Al2O<sup>3</sup> catalysts were obtained. The physico-chemical characterization has revealed the formation of small Cu nanoparticles highly dispersed in the ZnO-Al2O<sup>3</sup> structure.

All the catalysts were active in the gas-phase hydrogenation of furfural, with a high selectivity toward furfuryl alcohol, which is considered a valuable product due to its use for manufacturing resins. The presence of a low proportion of acid sites seems to disfavor consecutive reactions, being the formation of MF very limited in all cases. In addition, the textural properties of catalysts are typical of meso- and mainly macroporous solids, and the weaker Cu<sup>0</sup> -FUR interaction, in comparison to other

supports with stronger acidity, favors the lability of adsorbed reaction products and consequently shorter residence times, minimizing the deactivation by the formation of carbonaceous deposits.

In additional studies, the influence of the aging time in the synthesis of LDHs was also evaluated. The catalytic results have demonstrated that a shorter aging time leads to the formation of catalyst with a lower crystallinity, which favors the formation of Cu<sup>0</sup> -rich surfaces. This provides more active and stable catalysts, highly selective to FOL, obtaining a maximum FOL yield of 71% after 5 h of TOS at 210 ◦C for the CuZn\_1 catalyst that was aged for 1 h.

**Author Contributions:** Conceptualization: J.A.C. and P.M.-T.; methodology: C.P.J.-G. and J.A.C.; validation: C.P.J.-G, J.A.C. and P.M.-T; formal analysis: G.R.B. and C.P.J.-G.; investigation: G.R.B. and C.P.J.-G.; resources: P.M.-T; data curation: C.P.J.-G. and J.A.C.; writing—original draft preparation: J.A.C.; writing—review and editing: C.P.J.-G., J.A.C. and P.M.-T.; visualization: C.P.J.-G. and J.A.C.; supervision: J.A.C. and P.M.-T.; project administration: P.M.-T.; funding acquisition: P.M.-T. All authors have read and agreed to the published version of the manuscript.

**Funding:** The authors are grateful to financial support from the Spanish Ministry of Innovation, Science and Universities (Project RTI2018- 094918-B-C44) and FEDER (European Union) funds.

**Acknowledgments:** J.A.C. and C.P.J.-G. thank University of Malaga for contracts of PhD incorporation.

**Conflicts of Interest:** The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Hierarchical PtIn**/**Mg(Al)O Derived from Reconstructed PtIn-hydrotalcite-like Compounds for Highly E**ffi**cient Propane Dehydrogenation**

#### **Jiaxin Li, Ming Zhang, Zhen Song, Shuo Liu, Jiameng Wang and Lihong Zhang \***

Department of Catalysis Science and Technology and Tianjin Key Laboratory of Applied Catalysis Science & Technology, School of Chemical Engineering and Technology, Tianjin University, Tianjin 300350, China; xgclone@163.com (J.L.); minne\_zhang@hotmail.com (M.Z.); songzhens12345@163.com (Z.S.); shuoecho@outlook.com (S.L.); wangjiameng0216@163.com (J.W.)

**\*** Correspondence: zlh\_224@163.com or zlh\_224@tju.edu.cn

Received: 8 August 2019; Accepted: 9 September 2019; Published: 12 September 2019

**Abstract:** The challenges facing propane dehydrogenation are to solve the Pt sintering and carbon deposition. This paper provides a new way to disperse and stabilize Pt species and resist carbon deposition. Highly dispersed Pt species were topologically transformed from reconstructed PtIn-hydrotalcite-like precursors in a flower-like hierarchical microstructure. The lattice confinement of reconstructed hydrotalcite-like precursor is in favor of stabilizing the highly dispersed Pt species, and the hierarchical microstructure is an important factor to prolong its lifetime by enhancing tolerance to carbon deposition. In propane dehydrogenation, the propene selectivity decreases in the sequences of catalyst in flower-like > single-plate > block mass with small, flakeys. A propene selectivity of >97% with a conversion of 48% at 600 ◦C has been achieved over a flower-like PtIn/Mg(Al)O catalyst. Additionally, no visible Pt sintering can even be observed on this catalyst after a reaction time of 190 h. This strategy provides an effective and feasible alternative for the facile preparation of highly dispersed metal catalysts.

**Keywords:** propane dehydrogenation; hierarchical microstructure; reconstruction; high selectivity; excellent durability

#### **1. Introduction**

Industrially, propene is a vital chemical material with a huge demand [1,2]. Catalytic propane dehydrogenation (PDH) is considered an efficient technique to obtain propene [3,4]. In this process, catalyst Pt/Al2O<sup>3</sup> is widely used [5,6].

However, there are several open issues for this Pt-based catalyst under reaction conditions [3,7,8], such as Pt sintering, carbon deposition and propane cracking. Extensive research efforts have been made to inhibit deactivation and improve propene selectivity of Pt-based catalysts by doping promoters of Sn [9,10], In [11,12], Ga [13,14] and Cu [15] to form bimetallic catalysts. For example, Pt–In [16] and Pt–Ga [17,18] bimetallic nanoparticles (NPs) supported on calcined hydrotalcites (HT) show higher activity and selectivity in PDH, which has been attributed to the increased Pt dispersion and electron transfer from In or Ga to Pt [11,18]. Besides promoters, another method is to modify Al2O<sup>3</sup> support with ZnO [19], La2O<sup>3</sup> [20] or TiO<sup>2</sup> [21] or replace Al2O<sup>3</sup> support with ZnAl2O<sup>4</sup> [22], ZSM-5 [23] or Mg(Al)O oxide [8,24]. Additionally, it has been verified that the single Pt atom sites on Pt-Sn/γ-Al2O<sup>3</sup> with a "sandwich structure" are more favorable for PDH reaction than Pt ensembles [25,26]. Gong et al. [15] also reported that the PtCu single atom alloys can enhance the propane conversion and propene selectivity simultaneously in PDH.

Whether as promoters or supports, an important aim of using them is to obtain highly and stably dispersed Pt-based catalysts so that improving PDH performance. However, it is still difficult to

break through these challenges for PDH reactions [27], and the exact Pt-M-support relationship is still unclear. Therefore, it is extremely necessary to find more efficient methods to solve the problems mentioned above.

In fact, the hierarchicalmorphology and porous structure playimportant rolesinmany heterogeneous catalytic reactions. However, the effect of hierarchical microstructure is seldom reported.

Additionally, calcined HT-like compounds are regarded as a good way to obtained supported metal catalysts with high surface area and atomic-scale uniform distribution of metal species due to the topotactic transformation [28]. These features, the changeable composition and the acid-base property make it is possible to improve PDH performance by optimizing Pt-based catalysts [8,24,29].

However, the most of HT-like compounds are prepared by co-precipitation method; as a result, the morphologies of as-synthetized products are usually low crystalline NPs consisting of a great number of small and thick plates [3,8]. That must be detrimental to fully utilizing the active metal species in bulk. Fortunately, the calcined HT can be reconstructed due to the memory effect of HT. During the process, some metal ions have the chance to be introduced into the layer of reconstructed HT [30,31]. The lattice-confined metal species can be converted into highly dispersed metal NPs and even single metal atom sites.

Given that the small Pt NPs and hierarchical microstructure may provide more stable and easily accessible active sites for PDH reaction, a promising design for obtaining the desired size of active Pt and hierarchical pore morphology was proposed. In this paper, PtIn/Mg(Al)O catalyst in a flower-like nanosheet array (PtInHT-FR) was fabricated by reduction—followed by calcination of the flower-like PtIn-HT precursor (PtInHT-F) with a hierarchical microstructure, which produced from a sodium dodecyl sulfate-assisted hydrothermal synthesis, followed by thermal decomposition and structural reconstruction. Although there is no an additional support for this catalyst with low specific surface area, a stable and excellent propene selectivity of >97% and a high conversion of 48% still can be achieved at 600 ◦C. As contrasting samples, the single-plate (PtInHT-PR) and block mass (PtInHT-MR) PtIn/Mg(Al)O catalysts were derived from the corresponding high crystalline single-plate HT (PtInHT-P) and low crystalline block mass HT (PtInHT-M) with small, flakeys, which were synthesized by the hydrothermal synthesis and co-precipitation methods, respectively. The corresponding calcined samples were marked as PtInHT-XC (X represents morphology of F—flower-like, P—single-plate and M—block mass). After PDH reaction, the catalysts used were labeled as PtInHT-XU. During the preparation process, the HT-X, HT-XC, InHT-X and InHT-XC samples were also prepared. The structure and physic-chemical properties of catalysts and precursors were studied. The promoted effects of HT reconstruction on the microstructure and morphology of the 3D oxide nanosheet array catalyst, and the subsequently superior PDH performance, were profoundly discussed.

#### **2. Results and Discussion**

#### *2.1. Phase Structure and Reconstruction Study*

In order to ascertain the crystal phase structure of all samples in details, the XRD patterns of precursors, calcined and reduced samples are shown together in Figure 1, and the analyzing results of XRD patterns of all samples before calcination are given in Table 1.

It can be seen that the XRD patterns of HT-F, HT-P and HT-M exhibit strong characteristic diffraction peaks in the HT phase [32], although their preparation processes are different. After calcination, followed by the impregnation of HT-FC, HT-PC and HT-MC with In3<sup>+</sup> ions, the characteristic diffraction peaks of the HT phase appear again in the XRD patterns of InHT-F, InHT-P and InHT-M, without other diffraction peaks. That suggests that the impregnation condition is beneficial to reconstructing the HT phase. A similar HT recrystallization phenomenon happened again after the impregnation of In-based metal oxides with Pt4<sup>+</sup> ions. For HT-M, the broad and well-defined diffraction peaks indicate that the obtained single HT phase was poorly crystallized. With the introduction of In3<sup>+</sup> ions and further Pt4<sup>+</sup> ions, the relative intensity of diffraction peaks of HT phase gives a decreasing tendency. From HT-F

to InHT-F and HT-P to InHT-P, the shrinkage of peak intensities is more considerable than that from HT-M to InHT-M. This should be related to the decrease of HT crystallinity. At the same time, the average crystallite size in the *c* direction (*Lc*) also decreases with reconstruction and even drops from PtInHT-M to PtInHT-P and to PtInHT-F (see Table 1), which can be estimated by means of the Scherrer equation [33]. This indicates that the HT layer becomes thinner and thinner.

**Figure 1.** XRD patterns of HT precursors and derived samples in different morphology.


**Table 1.** Lattice parameters of HT phase in all dried samples.

a *a* = 2*d*110. <sup>b</sup> Average value calculated from (003), (006), and (009) reflections. <sup>c</sup> Average crystallite size in c direction calculated from the Scherrer equation using the FWHM (Full Width at Half Maximum) of (003) and (006) reflections [33].

Additionally, the decreasing M2+/(M3<sup>+</sup> + M4+) ratio can cause the increase of positive charge density in the layers, while the big radii of In3<sup>+</sup> (0.080 nm, the Shannon ionic radis [34]) and Pt4<sup>+</sup> (0.063 nm, Shannon ionic radius [34]) ions can broaden the distance of metal ions in the layers. These can make the octahedral cell shrink or expand and then influence the value of the lattice parameter *a* and *c* of HT phase, as shown in Table 1. The change of lattice parameters *a* and *c* of the HT phase (Table 1) demonstrates that In3<sup>+</sup> and Pt4<sup>+</sup> ions have been introduced into the HT layer. It signifies that

the Pt and in species could be highly dispersed on the final metal oxides with close contact among them, and support [10].

After calcination and reduction, the HT phase disappears along with the occurrence of MgO (JCPDS file number 45-0946) and meixnerite (Mg4Al2(OH)14·3H2O, JCPDS file number 35-0964) diffraction peaks in Figure 1. At the same time, there are not any metal phase diffraction peaks can be found in the XRD patterns, which means that Pt and/or in metal phases are exactly evenly-dispersed on the surface of HT-derived catalysts.

From TEM images of PtInHT-FR, PtInHT-PR and PtInHT-MR in Figure 2, it is difficult to find metal particles except thin nanosheets. That indicates that the metal particles are so small that the TEM cannot identify them. Therefore, it is reasonable to infer that metal atom clusters or single metal atoms are possibly formed on the surface of metal oxide nanosheets due to the low metal loading amount and the HT lattice confinement effect [10]. The EDS element-mapping images of PtInHT-FR further demonstrate the uniform distribution of Pt and in elements without aggregation, and support the probability of metal atoms or atom clusters type distributions [35].

**Figure 2.** TEM images of reduced samples: PtInHT-FR, PtInHT-PR, PtInHT-MR and EDS element-mapping analysis of PtInHT-FR. In the abbreviations: F—flower-like, P—single-plate, and M—block mass and R—reduced sample.

#### *2.2. Morphology, Texture and Pt–In Interaction Analysis*

Figure 3 shows SEM images of three series of samples. It clearly depicts that the HT-F and corresponding derivative samples are constructed by huge ball-flower-like microstructure. For fresh HT-P and its derivative samples, perfect single-plate NPs are displayed in the SEM images. On the contrary, small, flakes are disorderly stacked into big block masses for HT-M and its derivative products. The calcination makes the nanosheets become rough and the impregnation-induced reconstruction makes them become smooth again. The loose surface for the calcined samples is related to the newly formed pores, which is the result of the decomposition and removal of interlayer anions and hydroxyl groups. It is clear that a large amount of small and ultrathin nanosheets grow in an orderly manner with *ab*-planes interdigitated perpendicular to the two lateral surfaces of recrystallizing original HT

nanosheets for InHT-F and InHT-P. According to the XRD results in Figure 1, the reconstruction should also take place in InHT-M, although it is difficult to distinguish the small new nanosheets from the original flakes. In other words, this microstructure is indistinctive in the reconstructed HT-M series and instead severely aggregated nanoplates without any regular shape leave over. This microstructure should be the result of abundant hydroxyl groups on the edge sites and the basal plane of calcined HT plates [36]. The similar morphology can be seen in the SEM images of corresponding PtIn-based samples. In any case, there must be a strong interaction between the newly formed nanoplates and the initial huge templates, originating from the shared Al and other elements [36]. The strong interaction is beneficial to stabilize and disperse metal active species. The formation of ultrathin nanosheets is bound to decrease the average crystallite size in the *c* direction (Table 1) and weaken the diffraction peaks of HT phase in initial huge HT-F and HT-P template, which is consistent with the XRD results in Figure 1. As for the reduced samples, the basic morphology feature still can be maintained, but some broken particles can be found on the surface of single-plate PtInHT-PR sample. That indicates that the fully opening surface is detrimental to protect the newly formed nanosheets from destruction by an external force.

**Figure 3.** *Cont*.

**Figure 3.** SEM images of HT precursor, calcined, reconstructed and reduced samples with different morphologies.

It can be seen that the slit-shaped mesopores are the common features of all samples. As for the recrystallizing single-plate samples, including InHT-P, PtInHT-P and corresponding calcined and reduced products, the reconstruction not only results in the formation of narrow slit-like pores due to the nanoplates' delaminating, but also leads to the appearance of honeycomb-like pores arising from the growth of ultrathin nanosheets perpendicular to the surface of the primary large single-plate nanoplates. Especially in the InHT-F and PtInHT-F series, the big wedge-shaped pore channels appear in the huge ball-flower-like microstructure, except the honeycomb-like pores. However, the similar size between the original flakes and newly formed nanosheets make the pore structures of the HT-M and reconstructed samples no different. Therefore, the obvious hierarchical pore feature can be assigned to recrystallizing flower-like and single-plate samples; and the pore shape and size of recrystallizing flower-like samples are more complicated than those of single-plate samples.

Low-temperature N<sup>2</sup> adsorption-desorption isotherms and pore-size distribution (PSD) curves were conducted to evaluate the textural properties of PtInHT-FC, PtInHT-PC and PtInHT-MC. Their plots are shown in Figure 4 and the corresponding *S*BET, *V*<sup>p</sup> and *d*<sup>p</sup> are given in Table 2. According to the isotherm classification, all samples exhibit a type IV isotherm with H3 hysteresis loops, implying the presence of typical slit-shaped mesopores [8,24]. The PSD curves and *d*<sup>p</sup> values show that the pore size increases in the following order, PtInHT-FC < PtInHT-PC < PtInHT-MC, and the same trend can be found for their *S*BET values.

**Figure 4.** (**A**) Low temperature N<sup>2</sup> adsorption-desorption isotherms and (**B**) PSD curves of different catalysts.



<sup>a</sup> BET specific surface area. <sup>b</sup> Total pore volume. <sup>c</sup> The most probable pore size determined by the BJH method. <sup>d</sup> Pt dispersion of PtInHT-XR determined by the pulse chemisorption of CO.

The broad PSD was in the mesopore range; big *d*<sup>p</sup> and *S*BET values of PtInHT-MC are related to unordered aggregation of small, flakey-like particles. As expected, the big, ordered nanosheets make PtInHT-FC and PtInHT-PC present the narrow PSD feature, and small *d*<sup>p</sup> and *S*BET values. The *V*p, *d*<sup>p</sup> and *S*BET values of PtInHT-FC are far lower than that of PtInHT-PC especially, which is attributed to the fully open surface of PtInHT-PC relative to that of PtInHT-FC.

XPS analysis was applied to investigate the surface chemical state and relative concentration of In element on the reduced samples. The broad In3d5/<sup>2</sup> and In3d3/<sup>2</sup> peaks of catalysts can be deconvoluted into two peaks, respectively, in Figure 5. The deconvolution results of the corresponding spectra are listed in Table 3. The low bind energy (BE) contribution is assigned to metallic state In (In<sup>0</sup> ), and the high one is related to oxidation state In (In3+). A slightly decreasing BE value of In<sup>0</sup> from PtInHT-FR to PtInHT-PR and to PtInHT-MR reflects the weakening electrons' transfer from metallic In to Pt species and the Pt–In interaction. Additionally, a slightly increased ratio of In3+/In<sup>0</sup> can be seen in Table 3, in accordance with the order of PtInHT-MR, PtInHT-PR and PtInHT-FR. This indicates that the flower-like hierarchical structure with porous channels is in favor of maintaining the In element in In3<sup>+</sup> ions rather than In<sup>0</sup> states, while the In3<sup>+</sup> ions in PtInHT-MR, with small, flakey NPs, can easily be reduced.

It is worth mentioning that PtInHT-FC exhibits a higher surface Pt density than that of PtInHT-PC and PtInHT-MC, due to its low *S*BET and high *D*Pt, as listed in Table 2. In addition, the relative content of surface, metallic-state In increases in the order of PtInHT-FR, PtInHT-PR and PtInHT-MR (see Figure 5 and Table 3). Therefore, it can be inferred that the most of the surface In species have a chance to make close contact with Pt species and form plenty of strongly interacting Pt–In centers over PtInHT-FR. Accordingly, the Pt–In interaction over PtInHT-PR and PtInHT-MR should be decreased and weakened due to their low *D*Pt and high *S*BET. Additionally, too much In over PtInHT-MR could block Pt sites, thus cause a disadvantageous influence to the PDH reaction [8,24].

On the basis of the aforementioned results and preparation conditions, the formation of a ball-flower-like and single-plate hierarchical microstructure is tentatively proposed in Scheme 1.

**Figure 5.** XPS spectra of the In 3D region of reduced catalysts.



**Scheme 1.** Schematic illustrations of the formation of the ball-flower-like and single-plate hierarchical microstructure.

#### *2.3. Catalytic Performance and Structure–Activity Relationship Discussion*

To investigate the performance, three catalysts were evaluated in the PDH reaction. Figure 6 presents the propane conversion and propene selectivity against time for these catalysts. The homogeneous reaction is also tested in the blank tube and the propane conversion and propene selectivity as functions of time are shown in Figure S1. It was found that the conversion of propane is only around 1%, with a propene selectivity of around 65%. This indicates that the contribution of homogeneous reaction can be ignored. As shown in Figure S2, the selectivities to by-products (methane, ethane and ethene) are very low and can be ignored. In the PDH reaction, the initial propane conversions for PtInHT-FR, PtInHT-PR and PtInHT-MR are 6%, 44% and 28%, respectively. After a long induction period of 20 h, the propane conversion and propene selectivity of PtInHT-FR can be stabilized at above 40% and 97% with a maximum conversion of 48%. Until upon 190 h of time on stream, the conversion of PtInHT-FR only decreases to 36%. Despite a high initial conversion with a short induction period of around 5 h, a smooth stable period for around 50 h can be achieved by PtInHT-PR. A decreased propene selectivity cannot be ignored during the whole reaction process. As for PtInHT-MR, the reaction only can run for 20 h with moderate initial conversion; in the same period, the propene selectivity decreases severely with the rising propane conversion.

**Figure 6.** Propane conversion and propene selectivity as functions of time for different catalysts (reaction conditions: T <sup>=</sup> <sup>600</sup> ◦C, H<sup>2</sup> :C3H<sup>8</sup> :N<sup>2</sup> = 7:8:35 (molar ratio), weight hourly space velocity (WHSV) = 3 h−<sup>1</sup> , mcat = 0.4 g).

−

After the reaction, the XRD patterns of used samples are shown in Figure 7. It can be found that the meixnerite phase is still retained in the used PtInHT-FU, which is supposed to be favorable for the PDH reaction [7]. Additionally, the change occurs from fresh PtInHT-FR having more MgO phase (see Figures 1 and 7) while the used one has more meixnerite phase. This phase transfer of support could be the one cause of induction of performance during reaction. In addition, a new diffraction peak appears at around 26.0◦ in all the used samples, which belongs to the carbon deposits (JCPD file number 44-1644). The diffraction peak intensity of carbon deposits increases following the order: PtInHT-PU > PtInHT-MU > PtInHT-FU, which must be consistent with the amount of deposited carbon.

**Figure 7.** XRD patterns of used catalysts with different reaction time.

As shown in Figure 8, a large amount of carbon deposits with different sizes and morphologies, including granule, stick, loop and so on, are formed on the surface of used catalysts after reactions. Moreover, even the morphology features of PtInHT-PU and PtInHT-MU disappear, due to the surface coverage and pore blockage by carbon deposits. However, the flower-like particles of PtInHT-FU are still well dispersed and recognizable after a long reaction time, reflecting the fact that the amount of carbon deposits on PtInHT-FU number fewer than those on PtInHT-PU and PtInHT-MU. The most important thing is that the carbon is only deposited on the partial pore channel mouth of PtInHT-FU, meaning the active sites in the pore channels have no chance to be covered by the deposited carbon.

**Figure 8.** SEM images of used catalysts with different reaction time.

In order to determine the degree of carbon deposition during PDH reaction, TG experiments were conducted over the used catalysts. As shown in Figure 9, the mass loss of all samples must be related to the removal of deposited carbon. The amount of deposited carbon on the catalysts used increases following the order of PtInHT-FU, to PtInHT-MU and finally to PtInHT-PU. The low carbon amount is the most important factor to extend the lifetime of the flower-like PtInHT-FU.

After 190 h of reaction, it is still difficult to pick out metallic particles from the TEM images of PtInHT-FU in Figure 10. However, the Pt particles can be found on the surface of PtInHT-PU and PtInHT-MU. According to the particle size distribution of PtInHT-MU, the average particle size is higher for PtInHT-MU (5.5 nm) than that for PtInHT-PU (3.2 nm). That implies that the metallic particles on PtInHT-FU are more stable than those on PtInHT-PU and PtInHT-MU under a high temperature reaction, which has a vital influence for the performance of PDH. Certainly, the carbon deposits also can be easily found on the surface of all samples.

**Figure 9.** TG profiles of used catalysts with different reaction times.

(**A**)

(**B**) **Figure 10.** *Cont*.

(**C**)

**Figure 10.** TEM images, HR-TEM and the statistics of particle size distributions in the insets of the used catalysts with different reaction times. (**A**) 190 h, (**B**) 50 h and (**C**) 20 h.

The activity induction should be closely related with the partial coverage of the active metal surface by In2O<sup>3</sup> species [7]. We propose that the covered, active Pt particles tend to slowly migrate from the support onto In2O<sup>3</sup> particles to build highly dispersed "sandwich structure" of the Pt–In2O3–support under the reaction atmosphere [25,26]. Therefore the duration of induction should be determined by the coverage degree of In2O<sup>3</sup> species.

− Although the maximum conversion is up to 50%, the PDH reaction cannot be carried out after 50 h for PtInHT-PR and 20 h for PtInHT-MR due to the reactor blockage by a large amount of deposited carbon (see Figures 7–10), and the deposited carbon particles fully block the honeycomb-like pores of PtInHT-PR and cover the block mass particles of PtInHT-MR. Fortunately, the deposited carbon over PtInHT-FR after reacting for 190 h is much less than that over others. It is interesting to note that the carbon is mainly deposited on the pore mouth of ball-flower-like PtInHT-FR, which protects the metallic sites on the interior honeycomb-like surface of wedge-shaped pores.

Obviously, the deposited carbon amount is inversely correlated to the corresponding propene selectivity. The ball-flower-like multi-level hierarchical microstructure with abundant wedge-shaped pores, and short-channel honeycomb-like pores, but not the slit-like pores, is in favor of propane and propene diffusion. Furthermore, the low *S*BET of PtInHT-FR facilitates the propene desorption and then decreases the likelihood of deep dehydrogenation and hydrogenolysis on metal sites. These can be used to explain the high selectivity and low carbon deposits of PtInHT-FR. On the contrary, plenty of carbon deposits and the sharply reduced selectivities of PtInHT-PR and PtInHT-MR arise from the propene concentrating on the surface with a high *S*BET value, which triggers carbon formation and accumulation.

Except for the influences of texture and morphology, the propene selectivity and catalytic stability can be improved, by partially transferring electrons from IN to Pt [11]. PtInHT-FR presents a high surface In3+/In<sup>0</sup> ratio and BE value of In<sup>0</sup> , indicating that a high electron density of Pt and strong Pt–In interaction can be obtained (see Figure 5 and Table 3), which facilitates the desorption of propene and the migration of carbon precursors from the metal surface to the support [7,11]. Moreover, the suitable In<sup>0</sup> species can strengthen the Pt–In-support interaction, improve dispersion and prevent sintering of Pt particles (see Figures 2 and 10), stabilizing the catalytic performance of PtInHT-FR. Meanwhile, the overabundant In<sup>0</sup> species could bring about a disadvantageous influence to the PDH reaction. The high content In<sup>0</sup> species in PtInHT-PR and PtInHT-MR can not only block Pt active sites [8], but also weaken Pt–In interactions and lead to the Pt sintering (see Figures 2, 8 and 10). This is associated with the fact that the carbon deposits can easily form on the large-size Pt sites over PtInHT-PR and PtInHT-MR [37].

The regeneration and activation test for PtInHT-FU is shown in Figure 11. After each regeneration, the activity can be restored up to the level before regeneration, but decreases rapidly; almost no change in the selectivity can be observed. Although, the deposited carbon can be removed by a simple oxidation process (not the industrial condition), but the regeneration conditions can cause Pt sintering and decrease the Pt dispersion to a certain extent, which can result in the decrease of initial activity with cycles of regeneration [38,39].

− **Figure 11.** Propane conversion and propene selectivity as function of time for PtInHT-FU after regeneration and activation (regeneration conditions: T = 600 ◦C, oxidation in air for 2 h, sequencing purge with N<sup>2</sup> for 30 min and then reduction in 5 vol% H<sup>2</sup> /N<sup>2</sup> for 2 h; reaction conditions: T = 600 ◦C, H<sup>2</sup> :C3H<sup>8</sup> :N<sup>2</sup> <sup>=</sup> 7:8:35 (molar ratio) and WHSV <sup>=</sup> 3 h−<sup>1</sup> and mcat <sup>=</sup> 0.4 g).

#### **3. Materials and Methods**

#### *3.1. Materials*

Mg(NO3)2·6H2O, Al(NO3)3·9H2O, urea and sodium dodecyl sulfate (SDS) were all of analytical purity. H2PtCl6·6H2O powder was at 99.6% purity. All the chemicals were purchased from Fuchen (Tianjin) chemical reagent co. LTD (Tianjin, China).

#### *3.2. Preparation*

The flower-like precursor (HT-F) was prepared by a hydrothermal method. 0.82 g Mg(NO3)2·6H2O, 0.6 g Al(NO3)3·9H2O, 0.2 g SDS and 6 g urea were dissolved in 65 mL deionized water, then moved into a 100 mL Teflon autoclave. After aging at 100 ◦C for 20 h, the sample was filtered, washed with deionized water to neutrality and dried in air at 100 ◦C overnight.

Single-plate precursor (HT-P) was prepared using the same method as HT-F, except with no addition of SDS.

Block mass precursor (HT-M) was prepared by coprecipitation method. First, 17.09 g Mg(NO3)2·6H2O and 12.50 g Al(NO3)3·9H2O were dissolved in 100 mL deionized water. Another solution was prepared with 8 g NaOH and 10.6 g Na2CO<sup>3</sup> dissolved in 100 mL deionized water. Then, two solutions were mixed drop-wise, with strong stirring at room temperature for 2 h. The pH value was controlled at about 10. Subsequently, the mixed solution was aged at 65 ◦C for 12 h. The resulting suspension was filtered, washed with deionized water to neutrality and dried in air at 100 ◦C overnight.

All the precursors were calcined for 4 h at 600 ◦C to obtain the corresponding calcined samples, named HT-XC (X represents morphology, either F, P or M).

The corresponding In-based precursors, InHT-X, were acquired by incipient wetness impregnation method for calcined samples HT-XC with In(NO3)3·*x*H2O aqueous solution at room temperature for 6 h. Next, PtInHT-XC catalysts were prepared by successive incipient impregnation method. Firstly, HT-XC was impregnated with an In(NO3)3·*x*H2O aqueous solution at room temperature for 6 h and later dried at 120 ◦C for 12 h to obtain In-based precursors InHT-X, and then these solids were calcined at 550 ◦C for 4 h; corresponding products were labeled InHT-XC. Finally, the same procedures and conditions were performed on InHT-XC impregnated with H2PtCl6·6H2O, except that the time of impregnation was 2 h. After drying and calcination, the resulting solids were defined as PtInHT-X and PtInHT-XC, respectively. The loadings of Pt and In were 0.48 wt% and 1.38 wt%, respectively. After the reduction treatment under 5 vol% H2/N<sup>2</sup> at 600 ◦C for 2.5 h, these samples were labeled PtInHT-XR. And the used catalysts were marked as PtInHT-XU.

#### *3.3. Characterizations*

X-ray diffraction (XRD) characterization was carried on a Bruker D8-Focus X-ray diffractometer (BRUKER AXS GMBH, Karlsruhe, Germany) by Ni-filtered Cu Kα radiation (λ = 0.15406 nm) with a scan speed of 2θ = 8 ◦ ·min−<sup>1</sup> .

Low-temperature N<sup>2</sup> adsorption-desorption was performed at 77 K on a TriStar 3000 micromeritics apparatus (Micromeritics, Norcross, GA, USA) to collect the textural properties. Prior to measurements, the samples were outgassed under vacuum at 300 ◦C for 4 h, and then the specific surface area of samples was calculated by the Brunauer–Emmett–Teller (BET) method; the pore size distribution (PSD) was determined by the Barrett–Joyner–Halenda (BJH) method, used upon the adsorption branch of the isotherms.

Field emission scanning electron microscopy (FESEM) was carried out on a Hitachi S-4800 instrument (Tokyo, Japan) using a 3.0 kV electron beam.

Transmission electron microscopy and energy dispersive spectroscopy (TEM-EDS) were performed on a JEM-2100F field-emission transmission electron microscope (JEOL, Tokyo, Japan) with an accelerating voltage of 200 kV.

The X-ray photoelectron spectra (XPS) of catalysts were recorded on a Thermo Scientific Escalab 250Xi X-ray photoelectron spectrometer (Thermo Fisher Scientific, Waltham, MA, USA) using Al *K*α (*hv* = 1253.6 eV) radiation. The BE was calibrated using the C1s level at 284.8 as an internal standard.

The dispersion degree of Pt (*D*Pt) was calculated by CO chemisorption. CO chemisorption was conducted on AutoChem II 2920 analyzer (Micromeritics, Norcross, USA). A total of 100 mg of the samples were reduced under 10 vol% H2/Ar atmosphere at 600 ◦C for 2 h with a heating rate of 10 ◦C·min−<sup>1</sup> . After reduction, the catalysts were swept with He, at 600 ◦C for 30 min, to remove H2, and then cooled down to 50 ◦C. Subsequently, the CO pulse chemisorption was carried out at 50 ◦C by injecting pulses of 5 vol% CO/He until CO was adsorbed to saturation.

Thermogravimetric analysis (TG) was carried on a DTG-50/50H thermal analyzer (PerkinElmer, Waltham, MA, USA) to determine the carbon amount over the used catalysts with a heating rate of 10 ◦C·min−<sup>1</sup> from room temperature to 800 ◦C in air.

#### *3.4. Catalytic Tests*

Propane dehydrogenation (PDH) reaction was carried out in a fixed-bed reactor. The calcined sample (0.4 g) was placed into the stainless-steel tube and reduced to a flow of 5 vol% H2/N<sup>2</sup> at 600 ◦<sup>C</sup> for 2.5 h with a rising rate of 5 ◦C·min−<sup>1</sup> . Afterwards, the PDH reaction was performed in a mixture of <sup>H</sup>2, N<sup>2</sup> and C3H<sup>8</sup> (H2:C3H8:N<sup>2</sup> <sup>=</sup> 7:8:35 (molar ratio) and WHSV <sup>=</sup> 3 h−<sup>1</sup> ) at 600 ◦C. The regeneration test was performed at 600 ◦C, including oxidation in air for 2 h, sequencing purge with N<sup>2</sup> for 30 min and then reduction in 5 vol% H2/N<sup>2</sup> for 2 h. The propane and gas products were analyzed by an online gas chromatograph equipped with a FID detector and an Al2O<sup>3</sup> column. The propane conversion and propene selectivity were calculated as follows:

$$\text{Propane conversion}(\%) = \frac{\text{propane, in - propane, out}}{\text{propane, in}} \times 100\% \tag{1}$$

$$\text{Propene selectionity} \left( \% \right) = \frac{\text{propene, out}}{\text{propane, in -} \text{propane, out}} \times 100\% \tag{2}$$

#### **4. Conclusions**

In conclusion, the ball-flower-like, single-plate and block mass PtIn/Mg(Al)O catalysts were obtained by topological transformation and the reconstruction of HT precursors. The diversified morphology of catalysts was dependent on the conditions of synthesis. The metal ions of Pt4<sup>+</sup> and In3<sup>+</sup> can be introduced into the HT layer during the reconstruction process and the metallic species can be highly dispersed on the surface of Mg(Al)O. The ball-flower-like PtIn/Mg(Al)O catalyst exhibited high activity, excellent propene selectivity and superior durability as well as excellent resistance to the carbon deposition and Pt-sintering in the propane dehydrogenation to propene, which were mainly arising from its multi-level hierarchical microstructure with low specific surface area, and a large amount of strong and stable Pt–In interactions. It will be of great importance to revisit these catalysts and research the essential reasons for different catalytic performance in the PDH reaction by using more methods.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4344/9/9/767/s1, Figure S1: Propane conversion and propene selectivity as function of time for blank tube, Figure S2: The selectivity of methane, ethane and ethene as function of time for different catalysts.

**Author Contributions:** J.L. wrote the manuscript, performed the experiments and analyzed data; M.Z. and Z.S. collected references and characterized the physico-chemical properties of materials; S.L. and J.W. made the figures and tables, collected and checked data; L.Z. provided research ideas, analyzed data and revised the manuscript.

**Funding:** This research was funded by the National Natural Science Foundation of China (21776214).

**Acknowledgments:** This work was supported by the National Natural Science Foundation of China (21776214) and State Key Laboratory of Chemical Resource Engineering.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Catalysts* Editorial Office E-mail: catalysts@mdpi.com www.mdpi.com/journal/catalysts

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18

www.mdpi.com ISBN 978-3-0365-3551-7