Cinnamyl alcohol is an important fine chemical for high value applications, such as production of perfumes, flavors, and pharmaceuticals, and is typically manufactured by hydrogenation of cinnamaldehyde [1]. The hydrogenation of C=C in the C=C-C=O unit is more thermodynamically favorable than reduction of C=O. Thus, selective chemical reduction of the C=O group is a challenging synthetic transformation from both academic and industrial viewpoints [2]. Undesirable products such as hydrocinnamaldehyde or hydrocinnamyl alcohol are often formed in one-and two-step hydrogenation [3-5], leading to decreased selectivity of the unsaturated alcohol and increased purification costs. Thus, it is necessary to design more efficient catalysts that can facilitate selective hydrogenation of the C=O bond in the presence of other functionalities.

In general, supported catalysts based on transition metals (e.g., Pd, Pt, Ni, Cu or Au) are used for the hydrogenation of cinnamaldehyde [6-10]; however, the selectivity of such catalysts for the unsaturated alcohol is usually quite low. An alternative approach to reducing the C=O group by hydrogen transfer involves the use of a metal alkoxide as a catalyst and a secondary alcohol as a hydrogen donor. These processes are known as Meerwein-Ponndorf-Verley (MPV) reduction (Scheme 1) [11-17]. Traditional catalysts for MPV reduction are homogeneous catalysts based on aluminum, boron, and zirconium alkoxides [11-13]. However, these homogeneous catalysts feature problems, such as the need for a large amount of catalyst, environmentally unfriendly neutralization processes, and difficulty of isolating products and recycling/reusing the catalyst. These problems have lately been circumvented with the use of different heterogeneous catalysts. Many researchers have examined the performance of such catalysts, including metal oxides [18-20] or hydroxides [21], hydrotalcites [22-24], and various zeolites [25, 26]. However, these catalysts commonly show selectivity too low for practical use and also feature expensive post-synthesis separations. More efficient, selective, and stable heterogeneous catalysts for MPV reduction are highly desired.

Monoxides of metals such as Al₂O₃[27] and TiO₂[28] can be used as catalysts for MPV reduction; however, they are generally considered to be ineffective for intermolecular MPV reductions. Some recent reports have indicated that the selectivity for unsaturated alcohols can be improved with the use of partially reduced metal oxides, such as titanium oxide, because of the strong interaction between the carbonyl group and the positively charged catalytic center (TiO_x^δ+) [29]. In this work, a series of catalysts, based on titanium oxide dispersed on alumina, are prepared and their catalytic performances are tested for the MPV reduction of cinnamaldehyde to cinnamyl alcohol. A nanotubular alumina that enabled titanium oxide to be stably dispersed over the alumina surface in a low valance state (denoted as TiO_x/γ-Al₂O₃-nt) showed excellent performance for the title reaction.

Here, we used two types of Al₂O₃as substrates with different morphologies. We synthesized γ-Al₂O₃nanotubes by a hydrothermal method, with an alumina sol precursor. The precursor power (γ-AlOOH) was heated at a heating rate of 2 K/min to 873 K in air to obtain Al₂O₃ nanotubes. The as-prepared Al₂O₃nanotubes were denoted as γ-Al₂O₃-nt. Commercial Al₂O₃, exhibiting an amorphous morphology was purchased from Aluminum Corporation of China Limited, and heated at 873 K for 8 h in air before use. The commercial Al₂O₃ was denoted as γ-Al₂O₃-c.

The TiO_x/γ-Al₂O₃catalysts were prepared by a reactive deposition method, and TiO₂ nanoparticles were fabricated directly on the surface of the Al₂O₃. First, a certain amount of tetrabutyl titanate (TBOT) was dissolved in a 40-mL mixture of toluene (Aldrich) and ethanol (Aldrich) (V:V = 1:1) under magnetic stirring at room temperature. Then γ-Al₂O₃-nt or γ-Al₂O₃-c powder was slowly added to the mixture with magnetic stirring at room temperature for 24 h. The solid powder was then collected by centrifugation, washed several times with ethanol and dried in air at 353 K followed by grinding. The fine powder was heated to 673 K under an air atmosphere for 4 h and then heated under an Ar atmosphere at the same temperature for 2 h to remove organic pollutants and produce stoichiometric TiO₂ on the surface of γ-Al₂O₃. In this way, the TiO₂/γ-Al₂O₃-nt and TiO₂/γ-Al₂O₃-c composites were prepared (Fig. 1). The composite was subsequently treated in H₂ (50 mL/min) at 873 K with a heating rate of 10 K/min for 2 h and cooled to room temperature under same atmosphere. Finally, the TiO_x/γ-Al₂O₃-nt and TiO_x/γ-Al₂O₃-c catalysts were obtained.

The X-ray diffraction (XRD) patterns were measured with a Phillips X'Pro diffractometer using Cu K_α radiation (λ = 0.15418 nm) at 40 kV and 25 mA in the 2θ range 10°-90°. The texture of the solids was calculated from nitrogen adsorption-desorption isotherms at liquid nitrogen temperature, which were recorded on a Micromeritics ASAP-2020 instrument. The chemical composition of the samples was determined by X-ray fluorescence spectrometry (XRF) with the use of an ARL-9800 instrument. Samples were out gassed in vacuo at 573 K for 6 h prior to use. The specific surface areas were determined by the Brunauer-Emmett-Teller (BET) method. Transmission electron microscope (TEM) image was performed on a JEOL JEM-200CX at an accelerating voltage of 200 kV. The Raman spectra were recorded in the range of 150-800 cm^-1with a resolution of 4 cm^-1on a Renishaw Invia Raman microscope equipped with a Nd:YAG laser (λ = 532 nm).

For H/D exchange, the as-prepared samples (150 mg) were pretreated under Ar (40 mL/min) by heating to 873 K at 10 K/min and maintained at this temperature for 2 h. The sample was then cooled under the same atmosphere to room temperature. The exchange of protons for deuterium in the sample was measured by increasing the temperature to 873 K at a heating rate of 10 K/min. The signal of HD was monitored by mass spectrometry.

In situ X-ray photoelectron spectroscopy (XPS) measurements were conducted in a commercial XPS system (PHI 5000 Versa Probe) equipped with a hemispherical electron analyzer and monochromatic Al K_α X-ray excitation source. The sample was first evacuated at 393 K for 1 h, then cooled to room temperature, and the spectra were measured. Then the sample was reduced at 873 K under H₂(30 mL/min) for 2 h. After the sample was cooled to room temperature the spectra were measured again.

The acidity of the as-prepared catalysts was determined by temperature-programed desorption of ammonia (NH₃-TPD). The amount of NH₃chemisorbed on each catalyst was measured on a Micromeritics 2900 TPD/TPR analyzer. Prior to adsorption of the NH₃, the catalyst (100 mg) was heated at 873 K under a stream of argon (20 mL/min) for 2 h and cooled down in Ar to 303 K. The catalysts were then saturated with a 5% NH₃/95% Ar stream (40 mL/min) at 303 K for 1 h. Subsequently, a pure Ar stream (20 mL/min) was passed over the sample at 373 K for 1 h to remove physiosorbed molecules. Once a stable signal was obtained, chemisorbed NH₃ desorbed during heating from 373 to 840 K at 10 K/min. The final temperature was maintained for 20 min. Selected peaks were monitored throughout the process.

The MPV reduction was performed in a bath-type autoclave reactor. In a typical reaction, 0.3 mL of cinnamaldehyde and 30 mL of 2-propanol were mixed well in the autoclave at room temperature. Then 200 mg of catalyst was added to the reaction mixture. Helium was introduced into the reactor at room temperature and atmospheric pressure. The reactor was then heated to the reaction temperature and maintained at this temperature for a certain time. After the reaction, the reactor was allowed to cool to room temperature and the catalyst was filtered. The resulting solution was analyzed by GC, with the use of a GC-9560 instrument fitted with a HP-530 m × 0.32 mm column. For reuse tests, the catalyst was separated after the reaction and ultrasonically redispersed in a freshly prepared reaction solution for 15 min before the next run.

Table 1 shows the chemical composition and morphological properties of as-prepared catalysts tested in MPV reduction. The specific surface areas for the catalysts ranged from 15 m²/g for TiO₂ to 240 m²/g for γ-Al₂O₃-c. The TiO_x/γ-Al₂O₃ catalysts that were prepared from γ-Al₂O₃ had a BET area similar to that of the support, which suggested that the incorporation of TiO_x species had no effect on the γ-Al₂O₃ pore network. The Ti content was determined by XRF and XPS. The XPS results indicated a greater Ti content than that from XRF measurements, which suggested that TiO_x species were mainly distributed on the surface of the Al₂O₃ support. The micropore volume and total pore volume of the catalysts are listed in Table 1, together with the morphological properties of the as-prepared catalysts. Fig. 2(a) shows the nitrogen adsorption-desorption isotherms of the catalysts. All samples showed typical Ⅳ isotherms according to the IUPAC classification, except for TiO₂, which showed a typical Ⅲ isotherm, indicating the mesoporous nature of this sample.

The XRD patterns of the five as-prepared catalysts are presented in Fig. 2(b). For all the γ-Al₂O₃ catalysts, diffraction peaks at 2θ = 37.4°, 45.9°, and 66.8°, respectively, corresponded to (311), (400), and (440) planes of γ-Al₂O₃ (JCPDS 29-0063). The two different alumina substrates showed differences in the FWHM and intensities of these peaks. The γ-Al₂O₃-nt sample showed a narrower FWHM and stronger diffraction peaks, which indicated a higher degree of crystallinity for γ-Al₂O₃-nt than that of γ-Al₂O₃-c. For pure TiO₂, peaks at 2θ = 25.3°, 35°, and 55°, respectively, corresponded to the (101), (200), and (211) planes of anatase TiO₂ (JCPDS 21-1272). However, for the TiO_x/γ-Al₂O₃ catalyst only diffraction peaks from γ-Al₂O₃ emerged (JCPDS 29-0063) and no diffraction peaks from TiO₂could be detected. This XRD result indicated that TiO_x was well dispersed on the Al₂O₃ and no characteristic peaks of TiO₂ could be detected at an appropriate loading. Thus, TiO_xwas present in a highly dispersed state on the surface of Al₂O₃.

Fig. 3 shows TEM images of the as-prepared γ-Al₂O₃supports and the TiO_x/γ-Al₂O₃ composite oxide. The as-prepared γ-Al₂O₃-ntsample (Fig. 3(a)) exhibited a nanotubular structure with different sizes, the tube lengths were in the range of 60-100 nm and their outer diameters were 7-8 nm. The morphology of the γ-Al₂O₃-c sample (Fig. 3(c)) exhibited a structure of agglomerated nanoparticles, which formed a porous structure. The TEM images clearly showed the different morphologies of the two supports, attributed to the different preparation methods. To investigate the morphology of the TiO_x/γ-Al₂O₃ composite oxides and the distribution of TiO_x on the Al₂O₃ surface, HRTEM images were also obtained. The as-prepared TiO_x/γ-Al₂O₃ sample (Fig. 3(b) and (d)) maintained the morphology of the pristine Al₂O₃ support and no agglomerations of TiO₂ particles were found, indicating that the TiO_x was highly dispersed on the Al₂O₃ surface. The surface free energy of TiO₂ is considerably less than that of Al₂O₃, thus, dispersion of TiO₂ on Al₂O₃ surfaces is favored. This result was also consistent with the XRD results. Raman spectra are shown in Fig. 4. For a sample of pure TiO₂, the bands at 443 and 605 cm^-1 are assigned to anatase TiO₂, consistent with the anatase diffraction peaks observed in the XRD pattern of the sample. For both the TiO_x/γ-Al₂O₃-nt and TiO_x/γ-Al₂O₃-c samples no Raman bands related to TiO₂ were observed. Thus, the surface titanium oxide was likely highly dispersed on the Al₂O₃as a nonstoichiometric titanium oxide.

To further examine the TiO₂ dispersed on the surface of the Al₂O₃ support and the surface hydrogen of the as-prepared catalysts, exchange of gaseous D₂ with surface hydroxy groups of the catalyst was performed. Fig. 5 shows the HD evolution with increasing sample temperature. The γ-Al₂O₃-nt catalyst showed two main evolution peaks at 620 and 740 K. The first peak, i.e., peak α, derived mainly from HO-Al hydroxyl groups; the second peak was attributed to additional hydroxyl (β species) of the alumina support. For the γ-Al₂O₃-c, only a single HD evolution peak was observed. The peak γ was observed at 615 K and attributed to surface hydroxyl groups of γ-Al₂O₃-c. These results indicate that the surface hydroxyl groups of these two different alumina morphologies were quite different. For pure anatase TiO₂, the peak temperature for H/D exchange occurred at 693 K. This peak related to δ species, and indicated the lower acidity of surface hydroxyl groups on TiO₂ than those on alumina. Hydrogen in hydroxyl groups was tightly bound and thus required high temperatures to dissociated and exchange with D₂. After encapsulation of Al₂O₃ with TiO₂, TiO_x/γ-Al₂O₃-c featured a prominent peak temperature for H/D exchange at 695 K, which was similar to the behavior of surface hydroxyl groups from pure TiO₂(peak δ); however, a small peak at 578 K remained, which derived from surface hydroxyl groups of γ-Al₂O₃-c (peak γ). The TiO_x/γ-Al₂O₃-nt showed similar H/D exchange results to those of pure TiO₂(peak δ) with a prominent peak at 681 K and two small peaks at 570 and 778 K, which were characteristic peaks of γ-Al₂O₃-nt (α and β peaks, respectively). The H/D exchange results showed that both TiO_x/γ-Al₂O₃-nt and TiO_x/γ-Al₂O₃-c featured a main characteristic peak from surface hydroxyl groups of pure TiO₂. Thus, the surface of the as-prepared catalyst shared structural features with TiO₂. This result further confirms that TiO_x was present in a highly disperse state on the surface of γ-Al₂O₃.

The catalytic performances of the catalysts for MPV reduction of cinnamaldehyde, including the TiO_x/γ-Al₂O₃ composite oxide and pure metal oxide, are summarized in Table 2. The selectivity for cinnamyl alcohol (COL) and conversion of cinnamaldehyde (CAL) showed marked differences among the catalysts. The selectivity for COL can be expected to deceased slightly as the conversion rate increases, because subsequent hydrogenation reactions can take place simultaneously. Therefore, a high selectivity for COL at a high conversion of CAL is difficult to achieve. Notably the TiO_x/γ-Al₂O₃-nt sample showed the best catalytic performance with a conversion of 96.1% and was completely selective for COL at 150 ℃ (entry 2). This result is comparable to the performance of the current best reported catalysts [30]. For comparison, TiO_x/γ-Al₂O₃-c was used as catalyst in the same system. The activity of TiO_x/γ-Al₂O₃-c was much lower than that of TiO_x/γ-Al₂O₃-nt. The TiO_x/γ-Al₂O₃-c catalyst (entry 7) gave a much lower conversion (75.3%) and inferior selectivity (89.3%). Clearly, in the TiO_x/γ-Al₂O₃-nt catalysts, the γ-Al₂O₃-nt support contributed to their superior catalytic performance. This result indicates that the morphology of the alumina support, as shown in Fig. 3, and different surface structures can alter the dispersion of titanium oxide leading to very different catalytic properties under similar reaction conditions. The activity of the pure supports (γ-Al₂O₃-nt, γ-Al₂O₃-c and TiO₂) was also examined (entries 8-10). The bare alumina supports also showed inferior catalytic performance compared with that of the TiO₂/γ-Al₂O₃composite oxide.

The effects of temperature and reaction time over the TiO_x/γ-Al₂O₃-nt catalyst on the MPV reduction reactions were also examined over a temperature range of 120-180 ℃ (entries 1-3) and reaction time of 4-14 h (entries 2 and 4-6). The conversion of CAL increased with increasing temperature and reaction time. Importantly, the selectivity towards COL remained higher than 99% in all cases. These results indicate that increasing temperature and reaction time enhanced the reaction rate without affecting the selectivity. Furthermore, the TiO_x/γ-Al₂O₃-nt catalyst was highly stable under the reaction conditions and could be recycled several times without any notable loss of catalytic activity (Fig. 6).

The catalysts were further characterized to explore the relationships between their structure and performance. The surface titanium species were checked by in situ XPS and the results are shown in Fig. 7. Before the XPS measurement, the samples were reduced at 873 K in H₂ for 2 h in a catalytic cell connected to the XPS chamber and the samples were protected from oxidation. Notably, peaks of Ti(Ⅲ) species appeared in all the TiO_x/γ-Al₂O₃ catalysts. However, Ti(Ⅲ)species were not detected in TiO₂ reduced under the same conditions. These results indicate that the Al₂O₃ support contributed to the formation of Ti(Ⅲ)species. In addition, the three catalysts examined showed different contents of Ti(Ⅲ) species. The maximum Ti(Ⅲ)content was found for TiO_x/γ-Al₂O₃-nt, while only a small amount of Ti(Ⅲ) species were detected in pure TiO₂. This result indicates the regular morphology of the γ-Al₂O₃-nt with homogeneous surface sites and titanium oxide uniformly dispersed over the surface was more easily reduced and featured a high Ti(Ⅲ) content. As described in previous reports [29, 31], Ti(Ⅲ)and Ti(Ⅱ) defects can affect the adsorption modes of CAL with a preference for the vertical mode, which might explain the high selectivity for COL. The high selectivity of the TiO_x/γ-Al₂O₃-ntcatalyst can be attributed to the high content of Ti(Ⅲ) defects, which enabled vertical adsorption of CAL through terminal C=O groups onto the catalyst surface. Thus, Ti(Ⅲ) defects controlled the adsorption modes of CAL. The adsorption modes of CAL on TiO_x/γ-Al₂O₃-c and the metal oxides (γ-Al₂O₃or TiO₂) likely involved a mixture of both horizontal (through C=C groups) and vertical modes (through C=O groups), which led to poor selectivity for COL.

The surface acid properties of the catalysts were examined by temperature-programmed desorption of pre-adsorbed probe molecules NH₃ (Fig. 8). Before the adsorption of NH₃, the as-prepared samples were treated at high temperature (873 K), to dehydrate surface hydroxyl groups onto the surface of the catalysts and remove strong Br nsted acid sites. Thus, the amount of acid determined by NH₃-TPD analysis mainly related to Lewis acid sites. Pure TiO₂ showed the lowest amount of acid sites among the as-prepared catalysts. For the pure Al₂O₃ supports, γ-Al₂O₃-nt showed more acid sites than γ-Al₂O₃-c. Furthermore, a high temperature desorption peak (at 652 K) of NH₃emerged, which further confirmed the stronger Lewis acidity of γ-Al₂O₃-nt. This stronger acidity may be attributed to the regular morphology of the γ-Al₂O₃-nt with homogeneous surface sites, which resulted in strong surface acid properties. The higher Lewis acidity of the catalysts was beneficial for activating carbonyl groups and thus increased the overall reaction rate. As shown in Table 2, γ-Al₂O₃-nt was more active than γ-Al₂O₃-c and TiO₂ in MPV reduction of cinnamaldehyde. These results further confirmed the relationship between the Lewis acidity of the catalysts and their activity in the MPV reaction. The composite oxide TiO_x/γ-Al₂O₃ showed a slightly different TPD profile to that of the γ-Al₂O₃support, but retained the inherent acidity of the γ-Al₂O₃ support.

On the basis of our results, the high conversion and selectivity of the TiO_x/γ-Al₂O₃-nt catalyst can be explained by the following. The high Lewis acidity of the catalyst was beneficial for the MPV reduction, and the modified TiO_x/γ-Al₂O₃-nt catalyst maintained the high Lewis acidity of the γ-Al₂O₃-nt support. The acidity was related to the regular morphology of γ-Al₂O₃-nt. The product selectivity in the MPV reduction for CAL hydrogenation can be attributed to the substrate adsorption modes. For the MPV reduction over the pure support, CAL adsorbed randomly to the surface of the catalyst via both horizontal and vertical modes. Consequently, the adsorption modes of CAL resulted in a poor selectivity for COL. However, the Ti(Ⅲ) defects on the modified catalyst surface showed preferential vertical adsorption of CAL through terminal C=O groups onto the surface of catalyst. This adsorption mode resulted in the high selectivity of COL. Thus, the high conversion and selectivity for COL over the TiO_x/γ-Al₂O₃-nt catalyst could be attributed to the regular morphology, high Lewis acidity, and large amount of Ti(Ⅲ) defects.

A high-performance catalyst, TiO_x/γ-Al₂O₃-nt, was obtained by dispersing titanium oxide on an alumina support for MPV reduction of cinnamaldehyde to cinnamyl alcohol with isopropanol. Our characterization results indicated that the regular morphology of the alumina support and the well dispersed surface titanium(Ⅲ) oxide contributed to the excellent catalytic performance. We believe that this highly efficient and easily prepared catalyst has great potential for applications to other α, β-unsaturated aldehyde MPV reductions and other Lewis acid catalyzed reactions.