Regenerable Subnanometer Pd Clusters on Zirconia for Highly Selective Hydrogenation of Biomass-Derived Succinic Acid in Water

The size of metal particles is an important factor to determine the performance of the supported metal catalysts. In this work, we report subnanometer Pd clusters supported on zirconia by the microwave-assisted hydrothermal method. The presence of subnanometer Pd clusters on the zirconia surface was confirmed by two-dimensional Gaussian-function fits of the aberration-corrected high-angle annual dark-field images. These subnanometer Pd catalysts exhibit high catalytic performance for the hydrogenation of biomass-derived succinic acid to γ-butyrolactone in water and avoid the formation of overhydrogenated products, such as 1,4-butanediol and tetrahydrofuran. The catalyst with an ultra-low Pd loading of 0.2 wt. % demonstrated high selectivity (95%) for γ-butyrolactone using water as a solvent at 473 K and 10 MPa. Moreover, it can be reused at least six times without the loss of catalytic activity, illustrating high performance of the small Pd clusters.


Introduction
Supported Pd catalysts are widely used in many important commercial chemical processes such as oil refining and exhaust gas treatment [1][2][3][4].Subnanometer clusters with a few metal atoms in each cluster and a particle size of less than 2 nm as a catalytic species on solid support are found to play a major role in oxidation or hydrogenation reactions [5,6].However, the stability and recyclability remain a challenge because the isolated clusters or single atoms are mobile and readily sintered under high reaction temperatures [7].To the best of our knowledge, there are only a few contributions regarding the atomically dispersed Pd catalysts that are active and stable at high temperatures [8,9], while those of Au and Pt catalysts have been widely reported [10][11][12][13].Atomically dispersed Pd on graphene shows excellent durability against deactivation during 100 h of reaction time [8].Peterson et al. [9] prepared Pd species on La-doped γ-alumina supports that can be easily regenerated at 973 K through oxidation using air.Moreover, the addition of atomically dispersed La 3+ ions into the alumina support plays a key role in helping to stabilize and disperse isolated Pd atoms on the alumina surface.
As a promising alternative resource, biomass-derived molecules are widely available and renewable and can be converted into chemicals, medicines and fuels for vehicles.Succinic acid (SUC) as one of the most interesting biomass-derived platform chemicals can be generated via bioprocessing [14][15][16].The hydrogenation of SUC to produce valuable C4 compounds, including γ-butyrolactone (GBL), tetrahydrofuran (THF) and 1,4-butanediol (BDO), has attracted significant interest for the broad market of the downstream industries [15].In particular, GBL is a valuable intermediate for producing C4 chemicals, polymers, solvents, pesticides and so forth [17].Therefore, it is valuable to develop green and efficient processes for converting biomass-derived SUC into GBL.
Several supported catalysts have been reported to improve the efficiency of converting biomass-derived SUC into GBL [18,19].Palladium-based catalysts are considered one of the most efficient catalysts because of their high catalytic activity and selectivity [20][21][22].In addition, some factors such as support properties [23], the addition of a second metal [24] and the choice of solvent [25,26] were considered to improve the catalytic performance of Pd catalysts.However, these reported catalytic processes were either performed in the presence of organic solvents (1,4-dioxane or ethanol) or involved high Pd loadings (2.0%-5.0%),resulting in low Pd atom utilization efficiency and conflict with stringent environmental concerns.Such disadvantages make it important to design and cultivate catalysts with low Pd loading in green solvents for the liquid-phase hydrogenation of SUC.
As the fermentation broth contains over 95.0 wt.% water in the process of bio-fermentation for SUC generation [15], water as a solvent in the selective hydrogenation reaction of SUC can alleviate the risk of handling organic solvents and the expensive separation of SUC, which provides an economical process for SUC transformation.However, for the reaction taking place in water, the catalytic performance is poor as compared to organic solvents.Therefore, it remains a challenge to manufacture highly efficient catalysts in water for the hydrogenation of SUC [24,[27][28][29][30].
In this work, we prepared subnanometer Pd clusters supported on zirconia by a microwave hydrothermal method.The obtained subnanometer Pd clusters on zirconia were found to achieve high activity and stability for the SUC hydrogenation to GBL using water as a solvent.Furthermore, the catalyst is regenerable and can be cycled several times without loss of its catalytic activity.

Electron Microscopy
Aberration-corrected high-angle annual dark-filed scanning transmission electron microscopy (AC-HAADF-STEM) can recognize the atomically dispersed noble metal on the support, providing direct information about the surface structure of the catalysts [9,31,32].Here, it was employed to characterize the dispersion and identification of the Pd clusters in the materials.The typical AC-HAADF-STEM image of 0.2 Pd/ZrO 2 as well as the size distribution is shown in Figure 1a and additional representative images are present in Figure S1 in the Supplementary Materials (SM).In our previous works [33], the relationship between the cluster sizes and the numbers of Pd atoms in the clusters were determined by density functional theory calculation.Actually, the number of atoms is not directly relative to the horizontal projected area from the AC-HAADF-STEM images due to the difference in the layer stacking.From the reported EXAFS experiments on Pd clusters [34], the spherical Pd particles with a size less than 0.83 nm contain a maximum of 13 atoms.The major Pd species (>90%) represent subnanometer clusters based on statistical results over 400 Pd particles in 0.2 Pd/ZrO 2 catalyst with diameters <0.90 nm [13].The bright spots (marked 1 and 2 in Figure 1b,c) correspond to Pd particles, while the gray areas are zirconia nanospheres with a surface area of 216 m 2 ¨g´1 according to the Brunauer-Emmett-Teller (BET) method (Figure S2).In the selected and enlarged area (Figure 1b), the size of dispersed Pd species in the catalyst is about 0.45 nm.However, the varying background from the zirconia support limits this technology for locating the isolated Pd species in the catalyst.
The overlapped atom intensities are fitted using two-dimensional (2D) Gaussian functions by facile background subtraction so that the subnanometer Pd cluster can be recognized clearly in the AC-HAADF-STEM images (Figure 1c-e) [9].It is known that the contrast of atoms in AC-HAADF-STEM is generated from the nucleus by electron scattering.The resulting distinct intensity is relative directly to the square of the metal atomic number.The enlarged area in the original AC-HAADF-STEM image (Figure 1c) was colorized according to the intensity after 2D Gaussian function fits.The brightest part was colored in red while blue corresponds to the darkest part (Figure 1d).Then, the Pd particles can be easily identified in Figure 1e with the colorized intensity map.The histogram of intensity (Figure 1f) from Figure 1e shows a two-peak distribution of the two monodisperse species.The same intensity of the two peaks indicates the two dots (1 and 2) in Figure 1c only contain Pd species.
Catalysts 2016, 6, 100 3 of 12 intensity (Figure 1f) from Figure 1e shows a two-peak distribution of the two monodisperse species.The same intensity of the two peaks indicates the two dots (1 and 2) in Figure 1c only contain Pd species.In this histogram, the diameter of the Pd particles is about 0.50 nm, which is larger than that of the direct identification from the origin images (Figure 1b) owing to the background subtraction through the image analysis.Efremenko et al. [35] have reported that Pd clusters contain two to four atoms with a diameter less than 0.50 nm.So the Pd cluster marked 1 in Figure 1b with a size of 0.50 nm may contain two to four atoms.Our previous work [33] has also found similar results depending on the layer stacking.
In addition, the density of Pd atoms, calculated from the counted Pd atoms divided by the total surface area of ZrO2 in the AC-HAADF-STEM images, is about 0.048 Pd atoms•nm −2 for 0.2 Pd/ZrO2.It is very close to the actual Pd density (0.053 Pd atoms•nm −2 ), which is calculated from the number of supported Pd atoms in the catalyst over the total surface area of ZrO2 obtained by the BET method.The little difference between these two densities confirms there is only one layer of Pd atoms on the ZrO2 surface according to the radius of Pd.Thus, the observed Pd clusters are dispersed on the ZrO2 surface or the near subsurface [31].
Figure 2a shows the X-ray diffraction (XRD) patterns of a series of Pd/ZrO2 catalysts, all of which exhibit a tetragonal ZrO2 (JCPDF No. 81-1549, Figure S3) structure with high crystallinity.Besides, the peak at 28° in the sample prepared by the conventional hydrothermal method (ZrO2hydrothermal) is the characteristic peak of the monoclinic phase (JCPDF No. 72-1669).In this histogram, the diameter of the Pd particles is about 0.50 nm, which is larger than that of the direct identification from the origin images (Figure 1b) owing to the background subtraction through the image analysis.Efremenko et al. [35] have reported that Pd clusters contain two to four atoms with a diameter less than 0.50 nm.So the Pd cluster marked 1 in Figure 1b with a size of 0.50 nm may contain two to four atoms.Our previous work [33] has also found similar results depending on the layer stacking.
In addition, the density of Pd atoms, calculated from the counted Pd atoms divided by the total surface area of ZrO 2 in the AC-HAADF-STEM images, is about 0.048 Pd atoms¨nm ´2 for 0.2 Pd/ZrO 2 .It is very close to the actual Pd density (0.053 Pd atoms¨nm ´2), which is calculated from the number of supported Pd atoms in the catalyst over the total surface area of ZrO 2 obtained by the BET method.The little difference between these two densities confirms there is only one layer of Pd atoms on the ZrO 2 surface according to the radius of Pd.Thus, the observed Pd clusters are dispersed on the ZrO 2 surface or the near subsurface [31].
Figure 2a shows the X-ray diffraction (XRD) patterns of a series of Pd/ZrO 2 catalysts, all of which exhibit a tetragonal ZrO 2 (JCPDF No. 81-1549, Figure S3) structure with high crystallinity.Besides, the peak at 28 ˝in the sample prepared by the conventional hydrothermal method (ZrO 2 -hydrothermal) is the characteristic peak of the monoclinic phase (JCPDF No. 72-1669).Microwaves are electromagnetic waves containing electric and magnetic field components.Therefore, microwaves can provide fast heating, thus dramatically reducing reaction times and enhancing crystallized purity of the material produced [36,37].Further, no obvious peaks ascribed to Pd at 41°, 55° and 68° were observed, which indicates that the Pd species are well dispersed on the support.This phenomenon is attributed to the microwave heating for the hydrothermal synthesis of the noble metal clusters [38].Thus, it can be inferred that microwave heating leads to Pd species with a narrow size distribution and good dispersity on ZrO2 nanospheres.Moreover, energy-dispersive X-ray mapping measurements were conducted (Figure 2c,d) to examine the elemental distribution of Pd species in the 0.2 Pd/ZrO2 catalyst.The element mapping images indicate the bright dots in AC-HAADF images are Pd clusters (Figure 2b), which coincided well with the result of AC-HAADF images analysis.

Catalytic Activity of Pd/ZrO2
The heterogeneous Pd-based catalysts are efficient in the hydrogenation of SUC to GBL using organic solvents, e.g., 1,4-dioxane and ethanol [39][40][41].As the fermentation process for SUC generation is performed in water [15], it would be more economical to perform the SUC hydrogenation reaction directly in aqueous media with a heterogeneous catalyst.
The zirconia-supported subnanometer Pd clusters were utilized for the selective hydrogenation of SUC to GBL in water.In the reaction, GBL is the main product while propionic acid (PA) and butanoic acid (BA) are produced as byproducts.The catalytic results are shown in Table 1 and Figure S4.The support exhibits no catalytic activity for SUC hydrogenation (Table 1, entry 1).However, SUC Microwaves are electromagnetic waves containing electric and magnetic field components.Therefore, microwaves can provide fast heating, thus dramatically reducing reaction times and enhancing crystallized purity of the material produced [36,37].Further, no obvious peaks ascribed to Pd at 41 ˝, 55 ˝and 68 ˝were observed, which indicates that the Pd species are well dispersed on the support.This phenomenon is attributed to the microwave heating for the hydrothermal synthesis of the noble metal clusters [38].Thus, it can be inferred that microwave heating leads to Pd species with a narrow size distribution and good dispersity on ZrO 2 nanospheres.Moreover, energy-dispersive X-ray mapping measurements were conducted (Figure 2c,d) to examine the elemental distribution of Pd species in the 0.2 Pd/ZrO 2 catalyst.The element mapping images indicate the bright dots in AC-HAADF images are Pd clusters (Figure 2b), which coincided well with the result of AC-HAADF images analysis.

Catalytic Activity of Pd/ZrO 2
The heterogeneous Pd-based catalysts are efficient in the hydrogenation of SUC to GBL using organic solvents, e.g., 1,4-dioxane and ethanol [39][40][41].As the fermentation process for SUC generation is performed in water [15], it would be more economical to perform the SUC hydrogenation reaction directly in aqueous media with a heterogeneous catalyst.
The zirconia-supported subnanometer Pd clusters were utilized for the selective hydrogenation of SUC to GBL in water.In the reaction, GBL is the main product while propionic acid (PA) and butanoic acid (BA) are produced as byproducts.The catalytic results are shown in Table 1 and Figure S4.The support exhibits no catalytic activity for SUC hydrogenation (Table 1, entry 1).However, SUC conversion reaches 34% with GBL selectivity as high as 99% for 0.1 Pd/ZrO 2 .In contrast, SUC conversion increases from 63% to 99% and GBL selectivity drops from 95% to 85% with increased Pd loadings (0.2 wt.% to 1.0 wt.%).The byproducts are acids with 2.7 mol. % and 9.6 mol.% of BA, 1.6 mol.% and 4.5 mol.% of PA over 0.2 Pd/ZrO 2 and 1.0 Pd/ZrO 2 , respectively.40.0 mL water at 473 K, 10 MPa and 600 rpm, reaction time is 33 h unless otherwise mentioned; b selectivity to GBL; c Pd dispersion and average size are determined by H 2 chemisorption and AC-HAADF-STEM; d turnover frequencies (TOFs) were calculated based on moles of SUC converted at an initial reaction time of 0.25 h per mole of Pd per hour; e 2.0 g SUC in 40.0 mL water; f 4.0 g SUC in 40.0 mL water; g reaction time is 24 h; h reaction conditions: 6.0 g SUC and 1.0 g catalyst in 114.0 g water at 433 K and 15 MPa, reaction time is 48 h [22]; i TOFs were calculated at the reaction time of 1 h; j reaction conditions: 5.0 g SUC and 1.0 g catalyst in 95.0 g water at 433 K and 15 MPa, reaction time is 48 h [24]; k reaction conditions: 15.0 g SUC and 1.0 g catalyst in 85.0 g water at 453 K and 15 MPa, reaction time is 35 h [29].
The effects of the SUC concentration on the catalytic performance are listed in Table 1 (entries 3-5).When the SUC concentration is increased from 1.0 wt.% (0.4 g SUC in 40 mL H 2 O) to 5.0 wt.% (2.0 g SUC in 40 mL H 2 O), the SUC conversion and GBL selectivity over 0.2 Pd/ZrO 2 remain basically constant within the range of experimental error in aqueous solution.The SUC conversion is lower while the selectivity of GBL remains unchanged when the SUC concentration increases to 10.0 wt.% (4.0 g SUC in 40 mL H 2 O).The atomically dispersed noble metal particles with small size are generally favored for the formation of a specific product using a different reaction route [5].Thus, the subnanometer Pd clusters possess similar activity with uniform size and have specific selectivity to GBL (~95%) with the enhanced SUC concentration.Moreover, the SUC over 0.2 Pd/ZrO 2 can be completely converted over a long reaction time of 58 h (Figure S4a).GBL is still the main product and overhydrogenated products (THF and BDO) have yet not been detected.
The supported metal catalysts generally have types of metal particles with broad size distributions and each metal particle may possess multiple active species, which leads to reduced selectivity to the target product [5].The catalytic performance, depending on the sizes of noble metals, can also be elucidated in Table 1 including comparative literatures [22,24,29].The selectivity to GBL is reduced from 99% to 85% (entries 2-7) with an average particle size raised from 0.8 nm to 2.5 nm, owing to the increased multiple active species.With the similar conversion of about 54% (Figure S4b), the selectivities to GBL over 0.2 Pd/ZrO 2 , 0.5 Pd/ZrO 2 and 1.0 Pd/ZrO 2 are 96%, 87% and 86% (Figure S4c), respectively, which presents a similar trend shown in Table 1 (entries 3-7).The well-defined subnanometer Au clusters as hydrogenation catalysts were previously reported to control the selectivity for α,β-unsaturated alcohols for the C=O bond hydrogenation of α,β-unsaturated ketones (or aldehyde).The α,β-unsaturated alcohol as the intermediate could not be continuously hydrogenated into saturated alcohol [42].SUC is mainly transformed into GBL from the first hydrogenation step, and subsequently converted into small quantities of THF and BDO.In our work, the subnanometer Pd clusters are highly specific for the hydrogenation of SUC to GBL and avoid the further hydrogenation of GBL (Scheme 1).On the other hand, the reported Pd catalysts with a size of 2~10 nm can catalyze the subsequent hydrogenation of GBL into BDO and THF [21,22].Accordingly, the selectivity is extremely sensitive to the metal nanoparticle size [43].Further, the formation of PA or BA does not originate from the C-C bond breaking of BDO and it occurs via the C-C cracking of succinic acid due to the catalytic activity of hydrogenolysis for the Pd-base catalysts [40,41].Therefore, the main (formation of GBL) and side reactions (formation of PA and BA) occur independently.Moreover, the reported catalysts for the hydrogenation of SUC in water under different reaction conditions (temperature and H2 pressure) are also listed in Table 1, entries 8-10.As a result, the turnover frequencies (TOFs) for low loadings of Pd/ZrO2 (0.1 wt.% and 0.2 wt.%) are ~50 times higher than the reported 2.1% Pd/TiO2 based on the hydrogenation of SUC reaction with higher H2 pressure (15 MPa) and lower reaction temperature (433 K).For the high-loading Pd samples (0.5 wt.% and 1.0 wt.%), the TOFs are also ~30 times more active than the 2.1% Pd/TiO2.These results demonstrate the high catalytic performance of the subnanometer Pd cluster catalysts.Similar results have been reported for the small Pd clusters on La-stabilized alumina, which shows higher reactivity in the CO oxidation than the large particles [44].Moreover, both SUC conversion and the selectivity to GBL on the reported Ru/C are lower than those on the Pd/ZrO2 catalysts (entry 10).Further, the Ru catalyst can catalyze GBL into THF and BDO with severe carbon loss of as high as 63% in water solvent [29].
In addition, the catalytic activities of the supported metal catalysts could be related to metalsupport interactions and the distributions of metal species on support.The mutual effect between the noble metal and the support makes metal clusters behave differently from their nanoparticle counterparts [45,46].The smaller clusters may create more active sites around the metal-support interface.Conversely, the growth of particles likely loses the metal-support interface that gradually inhibits the ability of the adsorbed species to turn over at the metal-support boundary [47].Furthermore, very small metal clusters with electronic quantum size effects, in other words the highest occupied molecular orbital-lowest unoccupied molecular orbital (HOMO-LUMO) energy gap, lead to lower activation barriers of the reaction, which have a decisive role in the increase of reactivity [48].Therefore, the very small metal clusters, especially isolated single atoms, anchored on the supports represent a new type of catalyst with a superior catalytic performance for various reactions [5,[8][9][10].Thus, the 0.2 Pd/ZrO2 with an average Pd cluster size of 1.0 nm shows higher catalytic activities for SUC hydrogenation than those of 0.5 Pd/ZrO2 and 1.0 Pd/ZrO2 with Pd cluster sizes of 2.1 nm and 2.5 nm, respectively.
Arrhenius-type plots of the selective hydrogenation of SUC over the Pd/ZrO2 catalysts are shown in Figure 3 and the activation energy (Ea) was calculated based on initial reaction rates measured at different temperatures.The obtained Ea of 25.0 ± 1.5 kJ•mol −1 is similar for all the catalysts.Moreover, when the Pd loading is increased from 0.05 wt.% to 0.2 wt.%, the additional 0.15 wt.% Moreover, the reported catalysts for the hydrogenation of SUC in water under different reaction conditions (temperature and H 2 pressure) are also listed in Table 1, entries 8-10.As a result, the turnover frequencies (TOFs) for low loadings of Pd/ZrO 2 (0.1 wt.% and 0.2 wt.%) are ~50 times higher than the reported 2.1% Pd/TiO 2 based on the hydrogenation of SUC reaction with higher H 2 pressure (15 MPa) and lower reaction temperature (433 K).For the high-loading Pd samples (0.5 wt.% and 1.0 wt.%), the TOFs are also ~30 times more active than the 2.1% Pd/TiO 2 .These results demonstrate the high catalytic performance of the subnanometer Pd cluster catalysts.Similar results have been reported for the small Pd clusters on La-stabilized alumina, which shows higher reactivity in the CO oxidation than the large particles [44].Moreover, both SUC conversion and the selectivity to GBL on the reported Ru/C are lower than those on the Pd/ZrO 2 catalysts (entry 10).Further, the Ru catalyst can catalyze GBL into THF and BDO with severe carbon loss of as high as 63% in water solvent [29].
In addition, the catalytic activities of the supported metal catalysts could be related to metal-support interactions and the distributions of metal species on support.The mutual effect between the noble metal and the support makes metal clusters behave differently from their nanoparticle counterparts [45,46].The smaller clusters may create more active sites around the metal-support interface.Conversely, the growth of particles likely loses the metal-support interface that gradually inhibits the ability of the adsorbed species to turn over at the metal-support boundary [47].Furthermore, very small metal clusters with electronic quantum size effects, in other words the highest occupied molecular orbital-lowest unoccupied molecular orbital (HOMO-LUMO) energy gap, lead to lower activation barriers of the reaction, which have a decisive role in the increase of reactivity [48].Therefore, the very small metal clusters, especially isolated single atoms, anchored on the supports represent a new type of catalyst with a superior catalytic performance for various reactions [5,[8][9][10].Thus, the 0.2 Pd/ZrO 2 with an average Pd cluster size of 1.0 nm shows higher catalytic activities for Catalysts 2016, 6, 100 7 of 13 SUC hydrogenation than those of 0.5 Pd/ZrO 2 and 1.0 Pd/ZrO 2 with Pd cluster sizes of 2.1 nm and 2.5 nm, respectively.
Arrhenius-type plots of the selective hydrogenation of SUC over the Pd/ZrO 2 catalysts are shown in Figure 3 and the activation energy (Ea) was calculated based on initial reaction rates measured at different temperatures.The obtained Ea of 25.0 ˘1.5 kJ¨mol ´1 is similar for all the catalysts.Moreover, when the Pd loading is increased from 0.05 wt.% to 0.2 wt.%, the additional 0.15 wt.% Pd species significantly enhances the reaction rate; however, it does not change the Ea.On the other hand, both the rate and Ea remain constant when the Pd loading is enhanced from 0.2 wt.% to 1.0 wt.%.The results indicate that the presence of additional Pd (0.8 wt.%) in 1.0 Pd/ZrO 2 neither increases the rate nor changes the Ea of the reaction, which is in agreement with previous reports [49].It is also reported that only the atomically dispersed Au rather than nanoparticles is associated with the active species for the water-gas shift reaction over Au/CeO 2 catalysts and the excess Au neither increases the reaction rate nor changes the Ea [50].  Figure 4 presents the changes in TOFs over Pd loadings and dispersion during the initial period (Figure S4) of Pd/ZrO2 catalysts.It is interesting to note a non-linear relationship between TOFs and Pd loadings (or dispersion).The TOFs in Figure 4 are almost unchanged at low Pd loadings (≤0.2 wt.%) or high Pd dispersion (>50%), and then decrease with increased Pd loadings [51].When the size of the Pd particles decreases to the atomic dispersion level, the TOFs remain unchanged.This phenomenon indeed reveals that the catalytic active species are subnanometer Pd clusters.The similar TOFs for the low Au loadings (≤0.08 wt.%) have been reported on ZrO2-supported isolated Au 3+ ions which are atomically dispersed and isolated from each other for the chemoselective hydrogenation of unsaturated 1,3-butadiene [51].The TOFs decrease sharply with increased Pd loadings (Figure 4a) and decreased Pd dispersion (Figure 4b) which is associated with the transition of palladium from atomical dispersion to metallic nanoparticles.Similar results have also been reported for the oxidation and hydrogenation reactions using atomically dispersed Au and Pd catalysts [32,52].Thus, these subnanometer Pd clusters supported on zirconia for SUC hydrogenation with low loadings or high dispersion of Pd and high selectivity to GBL indicate promising applicability.Figure 4 presents the changes in TOFs over Pd loadings and dispersion during the initial period (Figure S4) of Pd/ZrO 2 catalysts.It is interesting to note a non-linear relationship between TOFs and Pd loadings (or dispersion).The TOFs in Figure 4 are almost unchanged at low Pd loadings (ď0.2 wt.%) or high Pd dispersion (>50%), and then decrease with increased Pd loadings [51].When the size of the Pd particles decreases to the atomic dispersion level, the TOFs remain unchanged.This phenomenon indeed reveals that the catalytic active species are subnanometer Pd clusters.The similar TOFs for the low Au loadings (ď0.08 wt.%) have been reported on ZrO 2 -supported isolated Au 3+ ions which are atomically dispersed and isolated from each other for the chemoselective hydrogenation of unsaturated 1,3-butadiene [51].The TOFs decrease sharply with increased Pd loadings (Figure 4a) and decreased Pd dispersion (Figure 4b) which is associated with the transition of palladium from atomical dispersion to metallic nanoparticles.Similar results have also been reported for the oxidation and hydrogenation reactions using atomically dispersed Au and Pd catalysts [32,52].Thus, these subnanometer Pd clusters supported on zirconia for SUC hydrogenation with low loadings or high dispersion of Pd and high selectivity to GBL indicate promising applicability.unsaturated 1,3-butadiene [51].The TOFs decrease sharply with increased Pd loadings (Figure 4a) and decreased Pd dispersion (Figure 4b) which is associated with the transition of palladium from atomical dispersion to metallic nanoparticles.Similar results have also been reported for the oxidation and hydrogenation reactions using atomically dispersed Au and Pd catalysts [32,52].Thus, these subnanometer Pd clusters supported on zirconia for SUC hydrogenation with low loadings or high dispersion of Pd and high selectivity to GBL indicate promising applicability.

Stability of Pd/ZrO 2 Catalysts
The active metal leaching or sintering of the active species with a loss in catalytic activity during reaction often occurs over the heterogeneous catalysts.Contributions to the stabilization of precious metal catalysts for the hydrogenation of SUC using water as a solvent are very scarce [29].Therefore, it is necessary to investigate the stability of the supported subnanometer Pd clusters, especially under high temperature, for practical applications.
The thermogravimetric analysis (Figure S5) was conducted to study the thermal stability of the catalyst for the hydrogenation of SUC.The first weight loss of 8% at about 373 K resulted from desorption of physically adsorbed H 2 O.The anchored ethoxy groups thermally decomposed on the surface of the sample generates the fast weight loss of 28% between 540 K and 550 K.These species have been observed to be associated with the decomposition of zirconia precipitates prepared in organic medium, where a small quantity of alkyl groups bound to the surface oxygen atoms of the support have been detected [53,54].Moreover, no weight loss over 550 K indicates the catalysts are fully crystallized under the calcinated temperature of 573 K and should be thermally stable during the reaction process.Moreover, the TGA results of multi-used catalyst (dashed line in Figure S5) further confirm the thermal stability of the 0.2 Pd/ZrO 2 catalyst.
Accordingly, the reusability tests of this catalyst are shown in Figure 5.A slight fluctuation of SUC conversion (~65%) and GBL selectivity (~95%) was observed.In addition, inductively coupled plasma atomic emission spectroscopy characterization on the fresh catalysts and the used catalyst after 33 h hydrogenation was performed to determine the Pd leaching during the reaction process (Table S1).The similar loading between the fresh (0.22 wt.%) and used (0.21 wt.%) 0.2 Pd/ZrO 2 catalysts indicate no leaching of Pd species.Moreover, SA conversion remains unchanged in a hot-filtered solution without solid catalyst after 10 h under the same reaction conditions.These results mean no leaching of Pd species and are consistent with recycling experiments.Besides, no collapse of the pores in the support occurs under the reaction conditions that can be determined from the BET surface area of 0.2 Pd/ZrO 2 before and after the catalytic reaction (Table S1).Further, the XRD pattern of the sixth recycled catalyst (Figure S3) shows the same peaks as the fresh catalysts, which confirms the stability of the catalyst.Therefore, the stable and reusable Pd/ZrO 2 catalysts could serve as promising catalysts in the hydrogenation of SUC in water.
of the pores in the support occurs under the reaction conditions that can be determined from the BET surface area of 0.2 Pd/ZrO2 before and after the catalytic reaction (Table S1).Further, the XRD pattern of the sixth recycled catalyst (Figure S3) shows the same peaks as the fresh catalysts, which confirms the stability of the catalyst.Therefore, the stable and reusable Pd/ZrO2 catalysts could serve as promising catalysts in the hydrogenation of SUC in water.

Materials and Methods
Atomically dispersed Pd supported on zirconia were prepared by a microwave-assisted hydrothermal method using K2PdCl4 (>99.9%,Aladdin Chemistry Co. Ltd., Shanghai, China) as

Materials and Methods
Atomically dispersed Pd supported on zirconia were prepared by a microwave-assisted hydrothermal method using K 2 PdCl 4 (>99.9%,Aladdin Chemistry Co. Ltd., Shanghai, China) as palladium source and zirconium oxynitrate hydrate (>99.5%,Aladdin Chemistry Co. Ltd., Shanghai, China) as zirconium source.Prior to microwave hydrothermal irradiation, the zirconium oxynitrate hydrate was dissolved in ethanol and K 2 PdCl 4 solution was added to the mixture based on the Pd loading.After stirring for 30 min, the homogeneous mixture was transferred into a high-pressure vessel with a magnetic stirrer.Next, the vessel was placed in a microwave reactor (Mars 450, CEM Corp., Matthews, NC, USA), which was set at 2.45 GHz with a power of 400 W and maintained at 393 K for 120 min with stirring.Finally, the products were dried at 333 K for 12 h followed by calcination at 573 K for 3 h.The catalysts prepared were identified as m Pd/ZrO 2 , where m represents the weight percentage of Pd based on the total catalyst.Prior to activity tests and characterization, all catalysts were reduced in a tube furnace in 10% H 2 /He atmosphere at 573 K for 3 h and stored under vacuum in the dark.
Aberration-corrected high-angle annual dark-filed images were obtained on a ChemiSTEM using a scanning transmission electron microscopy (Titan G 2 80-200, FEI Corp., Hillsboro, CA, USA) at 200 kV with probe size of ~1.3 Å.The energy-dispersive X-ray analysis element mappings were also conducted under HAADF mode from the same microscope.The HAADF images were formed through collecting transmitted electrons with an annular detector from an angle of 22 mrad.
X-ray diffraction characterization was performed on the X'Pert PRO diffractometer (PW3040/60, PANalytical Corp., Egham, England ) using a CuKα radiation source (λ = 1.54056Å).XRD patterns were recorded in the 2θ range from 10 ˝to 80 ˝with scanning speed of 10 ˝¨min ´1 in continuous mode.Inductively coupled plasma atomic emission spectroscopy were employed to determine the Pd loadings in the catalysts on an XSP instrument (IRIS intrepid II, Thermo Electron Corp., Beverly, MA, USA).
The dispersion of Pd exposed on ZrO 2 surface and average Pd particle sizes for Pd/ZrO2 catalysts were determined by H 2 chemisorption on a chemisorption analyzer (Autochem 2920, Micromeritics Instrument Corp., Norcross, GA, USA).The dried sample (about 0.1 g) was firstly reduced in a stream of 10% H 2 /Ar with 30 cm 3 ¨min ´1 for 2 h at 753 K and purged with Ar for 1 h at 753 K to remove the absorbed H 2 .Then, the catalyst was cooled to 318 K and H 2 pulses were injected through a on-line valve to record the absorption isotherm.Then, the sample was evacuated and the weak adsorption isotherm was ascertained with a second adsorption isotherm.
Adsorption and desorption isotherms of N 2 were measured at 77 K using physisorption analyzer (ASAP 2020, Micromeritics Instrument Corp., Norcross, GA, USA).Before measurement, the samples

Figure 1 .
Figure 1.The typical Aberration-corrected high-angle annual dark-filed scanning transmission electron microscopy (AC-HAADF-STEM) image and size distribution: (a) a selected and enlarged region from (a); (b) an enlarged area from (b); (c) a colorized intensity map from (c); (d) 2D Gaussian function fits to the intensity maps (e) and normalized intensity distributions received from the image (e) along arrow direction (f) of the 0.2 Pd/ZrO2 catalyst.The size distribution (inset in (a)) was based on over 400 Pd clusters from the high-magnification images.The two Pd clusters in (b) were designated 1 and 2, respectively.

Figure 1 .
Figure 1.The typical Aberration-corrected high-angle annual dark-filed scanning transmission electron microscopy (AC-HAADF-STEM) image and size distribution: (a) a selected and enlarged region from (a); (b) an enlarged area from (b); (c) a colorized intensity map from (c); (d) 2D Gaussian function fits to the intensity maps (e) and normalized intensity distributions received from the image (e) along arrow direction (f) of the 0.2 Pd/ZrO 2 catalyst.The size distribution (inset in (a)) was based on over 400 Pd clusters from the high-magnification images.The two Pd clusters in (b) were designated 1 and 2, respectively.
wt. % to 1.0 wt.%.The results indicate that the presence of additional Pd (0.8 wt.%) in 1.0 Pd/ZrO2 neither increases the rate nor changes the Ea of the reaction, which is in agreement with previous reports[49].It is also reported that only the atomically dispersed Au rather than nanoparticles is associated with the active species for the water-gas shift reaction over Au/CeO2 catalysts and the excess Au neither increases the reaction rate nor changes the Ea[50].

Figure 3 .
Figure 3. Arrhenius plots of reaction rates over Pd/ZrO2 catalysts with different Pd loadings.

Figure 3 .
Figure 3. Arrhenius plots of reaction rates over Pd/ZrO 2 catalysts with different Pd loadings.

Figure 4 .
Figure 4. Turnover frequencies as function of Pd Loadings (a) and dispersion (b) for Pd/ZrO 2 catalysts.

Table 1 .
Catalytic activity of Pd/ZrO 2 and reported Pd catalysts for the hydrogenation of succinic acid.