Supported noble metal catalysts are widely used in the selective hydrogenation of organic compounds to give valuable derivatives [1]. Selective catalytic hydrogenation of benzaldehyde (BzH) is a green route to the production of benzyl alcohol (BA), which is an important feedstock for the synthesis of vitamins, pharmaceuticals, pesticides, fragrances and various esters [2]. Hydrogenation of BzH can give various products by either partial or complete reduction of the carbonyl group and/or benzene ring, and thus serves as a model reaction for probing the catalytic hydrogenation ability of specific catalysts [3-5]. Platinum and ruthenium catalysts tend to be more chemoselective for reduction of the carbonyl group than reduction of the C=C bonds, and their main drawback is further hydrodeoxygenation of BA into toluene (TOL) [5, 6]. Because only surface metal atoms are accessible to reactants and can act as active sites, the metal catalysts are dispersed as very small particles on support materials to give a high specific surface area. Unlike homogeneous catalysts, which are uniform complexes with defined central metal ions and ligands, supported metal particles usually contain exposed atoms in various environments, and the local coordination environments are not adequately characterized on the atomic level [7, 8]. In fact, the structures of the surface metals in a supported catalyst are complicated because of the various sizes of the metal particles and their varied locations on the support material. Each site coexisting on a metal particle provides its own activity and selectivity profile, and thus varies the overall catalytic performance according to its relative size and contribution.

Particle size diversity in supported metal catalysts is a major parameter affecting their catalytic performance. Most commercial supported metal catalysts are prepared by a conventional impregnation method, during which the precious metal atoms may form dissolved cations, cations adsorbed on the support, or highly dispersed or aggregated oxide and/or metallic species and/or particles. The metal particles and their distributions are generally characterized by high resolution transmission electron microscopy (HRTEM), and thus the relationship between the catalyst structure and the catalytic performance can be established. However, characterization of the catalysts with spherical aberration-corrected transmission electron microscopy (ACTEM), a new technique, has shown that, besides metal particles, single-atom metal species are also common, irrespective of the support material [9-13]. Such single-atom metal species are active structures in many reactions and, in some cases, even more active than their nanoparticle counterparts. They have therefore become the subject of research in the new field of "single-atom catalysis" [13-21]. Thus, it is necessary to reconsider the active sites of supported metal catalysts and re-evaluate their relative catalytic contributions. Among the reported Pt catalysts that contain single-atom Pt species, Pt/MgAl₂O₄, with proper modification, is particularly suitable as a model catalyst owing to its superior stability and activity [11, 22].

In this work, we report the preparation of MgAl₂O₄ spinel-supported single-atom to nano-sized Pt catalysts and their catalytic performance in the catalytic hydrogenation of BzH. We characterized the Pt/MgAl₂O₄ catalysts using BET, X-ray diffraction (XRD), ACTEM, inductively coupled plasma atomic emission spectroscopy (ICP-AES), CO chemisorption and Fourier transform infrared (FTIR) spectroscopy of the adsorbed CO species, and analyzed the relationship between the catalytic performance and the catalyst structure.

MgAl₂O₄ spinel support material was prepared by hydrolysis of aluminum isopropoxide with magnesium nitrate hexahydrate in ethanol, followed by calcination of the precipitates in ambient air at 800 ℃ for 5 h with a heating rate of 2 ℃/min. MgAl₂O₄-supported Pt catalysts were prepared by soaking 10 g of the MgAl₂O₄ support powders in 200 mL of an aqueous solution of chloroplatinic acid hexahydrate (Pt content > 37.5%, Tianjin Fengchuan Chemical Reagent Technologies Co., Ltd.) for 10 h under magnetic stirring. Appropriate amounts of Pt were used to give nominal loadings of 0.2 wt%, 0.5 wt% and 1 wt%. The powders were separated by filtration and washed three times with deionized water. The impregnated samples were dried at 60 ℃ overnight, calcined at 500 ℃ for 5 h and finally reduced at 300 ℃ in H₂ for 2 h. The catalysts are denoted 0.2Pt/MgAl₂O₄, 0.5Pt/MgAl₂O₄ and 1Pt/MgAl₂O₄, and the actual Pt loadings were determined by ICP-AES to be 0.19 wt%, 0.47 wt% and 0.92 wt%, respectively.

Surface areas were measured on a Micromeritics ASAP 2010 apparatus using N₂ adsorption isotherms and Brunauer-Emmett-Teller (BET) analysis methods. All samples were degassed under vacuum at 300℃ for 5 h before the adsorption measurements. The content of Pt in the catalysts was measured on an ICP-AES 7300DV instrument. All samples were dissolved in aqua regia by microwave-assisted digestion at 900 W for 1 h. XRD patterns were recorded on a PW3040/60 X'Pert PRO (PANalytical) diffractometer equipped with a Cu Kα radiation source (λ = 0.15432 nm) operating at 40 kV and 40 mA. High-angle annular dark field scanning transmission electron microscopy (HAADF-STEM) images were obtained on a JEOL JEM-ARM200F microscope equipped with a CEOS probe corrector, with a guaranteed resolution of 0.08 nm. TEM specimens were prepared by depositing a suspension of the powdered sample on a lacey carbon-coated copper grid. Pt dispersions were measured by pulse adsorption of CO on a Micromeritics AutoChem Ⅱ 2920 automated characterization system at 50 ℃. Approximately 200 mg of the sample was loaded into a U-shaped quartz tube and heated to 300 ℃ for 30 min under pure H₂ (30 mL/min) and then cooled to 50 ℃ in pure He (30 mL/min). CO pulses (10%CO-90%He) were introduced by switching a six-way sampling valve. Pt dispersions were estimated by assuming the adsorption stoichiometry of CO to Pt to be equal to 1. In situ diffuse reflectance FTIR spectra of the adsorbed CO species were acquired with a Bruker Equinox 55 spectrometer equipped with a mercury cadmium telluride (MCT) detector at a spectral resolution of 4 cm^-1 and with 32 scans. The sample powders were loaded into the IR cell, pretreated in pure H₂ (6 mL/min) at 300 ℃ for 1 h at a heating rate of 10 ℃/min and then cooled to 50 ℃ in He (30 mL/min). The background spectrum was recorded under this condition. Then, CO/He (1 vol%, 30 mL/min) was introduced for 30 min and the sample was flushed with pure He (30 mL/min) before the spectra of chemisorbed CO were recorded.

Liquid-phase hydrogenation of BzH was conducted in a 10-mL Teflon-lined stainless steel autoclave reactor. In a typical hydrogenation reaction, 20 mg of the Pt/MgAl₂O₄ catalyst (or 1 μmol equivalent of Pt) was suspended in 5 mL ethanol containing 100 μL BzH (> 97%, AIKE Reagent, ~1 mmol). After the air was replaced three times with H₂, the autoclave was charged with H₂ (1.0 MPa) and maintained at various temperatures for various time under vigorous magnetic stirring. The absence of external diffusion limitations was verified by performing the reaction under various stirring rates in preliminary experiments. After the reaction, the autoclave was quenched with cold water. A reusability test was conducted with the 0.2Pt/MgAl₂O₄ catalyst, whereby the used catalyst was collected and recovered by filtration and washing with ethanol. The shortfall in the recovered catalyst was replenished with catalysts from other reactors that had undergone the same number of runs. The products were analyzed by gas chromatography (7890B, Agilent Technologies) using an HP-INNOWAX capillary column (30 m × 0.32 mm, Agilent Technologies) and a flame ionization detector.

The physico-chemical characteristics of the three Pt/MgAl₂O₄ catalysts are presented in Table 1. The surface areas, pore sizes and pore volumes of all three catalysts were 124-130 m²/g, 7.1-7.5 nm and 0.46-0.48 cm³/g, respectively, which indicates that the textural properties of the MgAl₂O₄ support materials changed little during the catalyst preparation process. The Pt loadings were accurately determined by ICP-AES to be 0.19 wt%, 0.47 wt% and 0.92 wt% for 0.2Pt/MgAl₂O₄, 0.5Pt/MgAl₂O₄ and 1Pt/MgAl₂O₄, respectively. The Pt dispersions were respectively estimated as 92.5%, 63.9% and 53.2% using CO chemisorption. The XRD patterns shown in Fig. 1 for the Pt/MgAl₂O₄ catalysts are identical to that of the MgAl₂O₄ support, which appears as a pure spinel crystal phase. The lack of detectable diffractions for Pt indicates that the mean Pt crystallite sizes are below the detection limit of XRD (~3 nm) in all samples. Representative HAADF-STEM images show that 0.2Pt/MgAl₂O₄ presents a large amount of single-atom Pt species as isolated bright dots on the MgAl₂O₄ substrate, and very occasionally Pt aggregates (Fig. 2(a)). Notably, the Pt aggregates appear to be amorphous, not crystallites with clear edges, which suggests that they do not contain enough atoms to form well-defined crystalline particles. Interestingly, 0.5Pt/MgAl₂O₄ displays many neighboring single Pt atoms in addition to the highly isolated Pt atoms and amorphous aggregates (Fig. 2(b)). This is reasonable because the surface density of Pt atoms on the MgAl₂O₄ substrate increases with increased Pt loading. In the 1Pt/MgAl₂O₄ sample, the amorphous Pt aggregates become dominant, and are accompanied by occasional isolated single Pt atoms and rare well-defined Pt crystallites of less than 2 nm (Fig. 2(c)). Clearly, with increased Pt loading, the packing of the Pt atoms on the MgAl₂O₄ surface becomes denser, and presents various Pt species, for example, isolated single Pt atoms, single-layered neighboring Pt atoms, amorphous aggregates and well-defined crystalline particles, depending on the local Pt atom density. These species include almost all possible transition forms, and thus illustrate the process for the formation of metal crystallites on a support substrate in slow motion. Similar processes are expected, and indeed evident albeit in a much faster mode, for Pt and other noble metals dispersed on various support materials [9, 10]. The Pt/MgAl₂O₄ catalysts retain much more of the single-atom and amorphous aggregate characteristics than conventional supported metal catalysts; this is presumably due to the markedly stronger metal/support interaction in this system.

In situ FTIR spectroscopy of CO adsorbed on the Pt/MgAl₂O₄ catalysts was used to probe the possible differences in the coordination environments of the accessible Pt atoms in various Pt species. As shown in Fig. 3, 1Pt/MgAl₂O₄ presents four bands centered at 1820, 2030, 2073 and 2080 cm^-1. The 0.5Pt/MgAl₂O₄ catalyst displays two distinct bands centered at 1820 and 2073 cm^-1, and a shoulder peak at approximately 2030 cm^-1. However, 0.2Pt/MgAl₂O₄ presents only a broad asymmetric band at 2065 cm^-1 and a small band at 1820 cm^-1. According to empirical interpretations of the FTIR spectra of CO adsorbed on metals, bands in the low frequency range (1730-1900 cm^-1) are usually assigned to bridged carbonyl species on neighboring metal atoms, whereas those in the high frequency range (1950-2100 cm^-1) are assigned to linear carbonyl species on metals in various chemical environments [23, 24]. Thus, based on this expected blue shift in the stretching frequency with increasing coordination degree (or metallic nature) of the Pt atom, the CO adsorption bands at 2065, 2073 and 2080 cm^-1 in the spectra described above were assigned to the collective oscillation of CO linearly adsorbed on the single Pt atoms, amorphous aggregates and crystallites. The band at 2030 cm^-1 was assigned to the collective oscillation of CO linearly adsorbed on the highly under-coordinated Pt on the edges or corners of the Pt crystallites [23]. The CO adsorption band at 1820 cm^-1 was assigned to the bridged CO species adsorbed on the terraces of the well-defined Pt crystallites [24]. On the basis of the relative band areas of the linear and bridged carbonyl species, it is clear that the number of terrace atoms markedly increases with Pt loading, which is in line with the observed increase in the density of Pt crystallites shown in Fig. 2.

To explore the catalytic hydrogenation ability of the Pt/MgAl₂O₄ catalysts, we conducted hydrogenation of BzH under various reaction conditions and 100% BzH conversion. As shown in Table 2, three products were obtained: BA, TOL and benzaldehyde diethyl acetal (BDA). As illustrated in Scheme 1, BA is produced by the selective hydrogenation of BzH, TOL is generated through hydrodeoxygenation of the carbon-oxygen bonds in BzH or BA, and BDA is formed by aldol condensation of BzH and the ethanol solvent. The results are displayed as the molar compositions of the products. Under mild reaction conditions, with a temperature of 60 ℃, H₂ pressure of 1.0 MPa and time of 4 h, the 0.2Pt/MgAl₂O₄, 0.5Pt/MgAl₂O₄ and 1Pt/MgAl₂O₄ catalysts were highly active in the selective hydrogenation of BzH to BA, but much less active in the hydrodeoxygenation to TOL. The high selectivity for BA decreased slightly from 99.4% to 99.1% and 99.0% and the low selectivity for TOL increased slightly from 0.4% to 0.6% and 0.7% as the Pt loading was increased from 0.2 wt% to 0.5 wt% and 1 wt%, respectively. BDA was detected with similar selectivities and yields of (0.2-0.3)% for all three catalysts. A comparable yield of BDA was also obtained for 1Pt/MgAl₂O₄ under similar reaction conditions when the atmosphere was changed to N₂. This suggests that the formation of BDA is a parallel side reaction with a much slower rate. When the reaction temperature was elevated to 100 and 150 ℃, the 0.2Pt/MgAl₂O₄ catalyst presented a slightly decreased selectivity for BA (98.8% and 97.9%) and increased selectivity for BDA (0.7% and 1.7%), whereas the selectivity for TOL was almost unchanged (0.5% and 0.4%). The 0.5Pt/MgAl₂O₄ catalyst, however, showed a notable increase in the selectivity for TOL (to 0.8% and 1.9%) and a decrease in the selectivity for BA (to 98.4% and 96.4%) in reactions at 100and 150 ℃. Similarly, 1Pt/MgAl₂O₄ also showed increased selectivity for TOL (to 3.9% and 5.0%) and decreased selectivity for BA (to 95.4% and 93.1%). The selectivities for BDA appear to be comparable for all temperatures and catalysts. Through hydrogenation of BzH at 100% conversion and over a wide temperature range, the catalytic hydrogenation ability of the Pt/MgAl₂O₄ catalysts was evaluated. Selective hydrogenation of BzH to give BA is the predominant reaction over all three Pt/MgAl₂O₄ catalysts. Hydrodeoxygenation of the carbon-oxygen bonds in BzH or BA to give TOL is negligible for 0.2Pt/MgAl₂O₄, but becomes meaningful when the Pt loading is increased to 0.5 wt% and above.

The catalytic activities of the Pt/MgAl₂O₄ catalysts were further investigated at various temperatures and BzH conversions by using catalysts containing similar amounts of Pt (~1 μmol). As displayed in Table 3, 0.2Pt/MgAl₂O₄ catalyzed the hydrogenation of BzH at temperatures as low as 20 ℃ with 1.0 MPa of H₂. Under these conditions, the BzH conversion was 55.3% after 15 h, with BA, TOL and BDA selectivities of 98.8%, 0 and 1.2%, respectively. When the catalytic activity was normalized to the total number of Pt atoms and surface Pt atoms determined by CO chemisorption experiments, the catalytic reaction rate and turnover frequency (TOF) were calculated to be 284 mol/(mol·h) and 307 h^-1, respectively. After reaction at 40 ℃ for 2 h, the BzH conversion reached 30.1% and the BA, TOL and BDA selectivities were 99.3%, 0 and 0.7%, respectively. Additionally, the catalytic reaction rate and TOF increased to 1160 mol/(mol·h) and 1254 h^-1, respectively. After reaction at 60 ℃ for 0.3, 1 and 2 h, the BzH conversions increased from 22.5% to 53.6% and 94.9%, and the calculated rate and TOF decreased from 3430 to 2700 and 2388 mol/(mol·h) and from 3708 to 2919 and 2582 h^-1, respectively. For 1Pt/MgAl₂O₄, after reaction at 60 ℃ for 0.3, 1 and 2 h, the BzH conversions increased from 19.4% to 45.9% and 90.1%, and the calculated reaction rate and TOF decreased from 3055 to 2386 and 2072 mol/(mol·h) and from 5472 to 4485 and 3895 h^-1, respectively. The calculated reaction rates and TOFs are clearly related to the reaction time, although, according to the rate equation, the intrinsic factors affecting these parameters are the instantaneous concentrations of the reactants and products [25]. Thus, the catalytic activities that were obtained at the same temperature and with similar BzH conversions were compared among the Pt/MgAl₂O₄ catalysts. At BzH conversions of approximately 20%, the reaction rates for 0.2Pt/MgAl₂O₄ and 1Pt/MgAl₂O₄ were 3430 and 3055 mol/(mol·h), respectively, and the TOFs were 3708 and 5742 h^-1, respectively. The inverse change in the mass specific rates and TOFs suggests that the surface Pt atoms on the Pt nanocrystallites are highly active and can partially compensate for the loss of accessible Pt atoms inside the particles. It is notable that the selectivity of 1Pt/MgAl₂O₄ for BDA at 60℃ and low BzH conversions reached (10-20)%, and that BDA was eventually almost completely converted to BA by catalytic hydrogenation. This supports the reaction pathways illustrated in Scheme 1. A previously described Pt/MgAl₂O₄ catalyst prepared by colloidal deposition has been shown to have a TOF of 56 h^-1 at 40 ℃ and 0.1 MPa of H₂ [5]. A Pt/dendrimer core/shell catalyst has been reported to have a TOF of 91 h^-1 under the same conditions [26], and a Pt/TiO₂ catalyst displays a TOF of 2448 h^-1 at 80 ℃ and 0.1 MPa of H₂ [27]. Clearly, all of the Pt/MgAl₂O₄ catalysts in this work are considerably more active under similar reaction conditions, irrespective of Pt loading and dispersion degree; this is presumably due to the strong interaction between the Pt species and the MgAl₂O₄ spinel support.

Because 0.2Pt/MgAl₂O₄ represents a novel catalyst and has the highest Pt availability and BA selectivity, we performed recycling experiments and evaluated the reusability of the catalyst. In these experiments, we were more interested in the stability of the catalyst and the BA selectivity than in the kinetic activity, so we conducted the reaction under conditions that gave total BzH conversion. As shown in Fig. 4, the BzH conversions remained at 100% and the BA selectivity at about 99.5% over at least six runs. Even though the 0.2Pt/MgAl₂O₄ catalyst was not able to provide high kinetic activity, the excellent stability of this catalyst in terms of BA selectivity was demonstrated, and indicates that the highly dispersed nature of the Pt species was unchanged during the recycling experiments.

The dispersion of Pt on an MgAl₂O₄ spinel support can result in the formation of isolated or neighboring single Pt atoms, amorphous Pt aggregates and well-defined crystalline particles; the species formed depend on the local Pt atom density. With a typical impregnation method, the coexistence of multiple Pt species seems to be inevitable, and it is currently impossible to quantify the relative amounts of the species in the mixture. Fortunately, the dispersion of the surface Pt atoms, which are potential active sites accessible to reactants, can be estimated empirically by CO chemisorption experiments, irrespective of whether they are present on the surface of the MgAl₂O₄ support or as Pt aggregates/crystallites. Accordingly, we obtained the Pt dispersions and calculated the TOFs using the empirical assumption that the stoichiometry of CO adsorption on Pt is equal to 1. This is appropriate for comparison with results reported in the literature that use a similar assumption. However, the detection of bridged carbonyl species suggests that the CO/Pt stoichiometry needs to be corrected by deconvolution of the contribution of the linear and bridged carbonyls using a CO/Pt ratio of 1 or 2, respectively. As shown in Table 4, the ratios of surface Pt with linear (Pt_L) and bridged (Pt_B) carbonyl species were calculated to be 90/10 and 44/56 for 0.2Pt/MgAl₂O₄ and 1Pt/MgAl₂O₄, respectively. Accordingly, the Pt dispersions were calibrated to be approximately 100% and 73.9%, respectively. As a result, the corresponding TOFs were determined to be 3430 and 4134 h^-1 at approximately 20% BzH conversion at 60 ℃. Such amendments to the overall TOFs, however, still do not reflect the intrinsic activity of specific active sites. As we have revealed above, different Pt species present various FTIR bands for chemisorbed CO owing to the different local coordination environments. In principle, each type of Pt species could be distinguished and quantified by deconvoluting the characteristic band; however, there is a high risk of over-interpreting the FTIR spectra if we try to deconvolute the broad IR bands into many specific bands with slightly different shifts. Here we consider only the relative contributions of Pt_L and Pt_B atoms, even though both include several types of Pt species. Thus, the average TOFs of Pt_L and Pt_B were calculated to be 3277 and 4807 h^-1, respectively, without further considering the differences in the detailed structures of the Pt_L and Pt_B atoms in the 0.2Pt/MgAl₂O₄ and 1Pt/MgAl₂O₄ catalysts. Thus, it can be concluded that the Pt_B atoms are considerably more active than the Pt_L atoms. Considering that some of the Pt_L atoms exist in the Pt crystallites, we can estimate that the activity of the crystallite surface atoms is higher than that of the single atoms on the MgAl₂O₄ support. The additional ability of the 1Pt/MgAl₂O₄ catalyst, which is rich in terrace Pt atoms, to hydrogenate BA to TOL suggests that the origin of this increased activity could be the ability to form bridged carbonyl groups. This is in line with the observation that TOL formation usually occurs on well-crystallized nanocatalysts [5, 28, 29].

We have demonstrated that Pt can be readily dispersed as single atoms on an MgAl₂O₄ spinel support and forms various dispersed Pt species, such as isolated or neighboring single Pt atoms, amorphous aggregates and crystallites, according to the local surface density of the Pt atoms. The single-atom Pt species located directly on the MgAl₂O₄ spinel surface are highly active in the selective catalytic hydrogenation of BzH to BA over a wide temperature range. The terrace atoms on the Pt crystallites are more active but less selective, and have the ability to further hydrodeoxygenate BA to TOL. These MgAl₂O₄ spinel-supported Pt single-atom catalysts represent a novel type of catalyst with high stability and selectivity for carbonyl hydrogenation.