Size-dependent catalytic activity over carbon-supported palladium nanoparticles in dehydrogenation of formic acid

Journal of Catalysis 352 (2017) 371–381

Contents lists available at ScienceDirect

Journal of Catalysis journal homepage: www.elsevier.com/locate/jcat

Size-dependent catalytic activity over carbon-supported palladium nanoparticles in dehydrogenation of formic acid Junjie Li a,1, Wei Chen a,1, Han Zhao a, Xusheng Zheng b, Lihui Wu b, Haibin Pan b, Junfa Zhu b, Yanxia Chen a, Junling Lu a,⇑ a Department of Chemical Physics, Hefei National Laboratory for Physical Sciences at the Microscale, Collaborative Innovation Center of Chemistry for Energy Materials (iChEM), CAS Key Laboratory of Materials for Energy Conversion, University of Science and Technology of China, Hefei, Anhui 230026, PR China b National Synchrotron Radiation Laboratory, University of Science and Technology of China, Hefei, Anhui 230029, PR China

a r t i c l e

i n f o

Article history: Received 5 September 2016 Revised 4 June 2017 Accepted 5 June 2017

Keywords: Formic acid Hydrogen generation Palladium catalyst Particle size effect, electronic effect

a b s t r a c t Hydrogen generation from formic acid (FA) under mild conditions has received significant attention, where Pd-based catalysts have been widely employed due to their superior activities. However, the Pd particle size effect has been much less systematically investigated. In this study, carbon-supported Pd nanoparticles (NPs) with five different Pd particle sizes, ranging from 2.1 to 4.5 nm were synthesized using sodium citrate as the stabilizing agent. The Pd particle sizes were determined by aberrationcorrected high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM). The FA dehydrogenation reaction was conducted in a FA-sodium formate (SF) aqueous solution in a batch reactor at room temperature. We found that decreasing the Pd particle size from 4.5 ± 0.5 to 2.1 ± 0.3 nm remarkably boosted the catalytic activity by about 3.6 times, resulting in a turnover frequency of 835 h
1, which is among the highest values for supported monometallic Pd catalysts in the literature. Our results suggest that both low- and high-coordination Pd surface atoms participated in the reaction. The remarkably higher activity of smaller Pd NPs was attributed to both higher Pd dispersion and the presence of a larger proportion of Pd species with positive charge, through which the Coulomb interaction between the positive Pd species and negative charged formate ions, the key reaction intermediate, is enhanced. Finally, the deactivation and regeneration of Pd/C catalysts were also discussed. Ó 2017 Published by Elsevier Inc.

1. Introduction Hydrogen, an environmentally clean energy carrier, has been regarded as one of the most promising candidates to meet the increasing demands for an efficient and clean energy supply [1]. Developing technologies for controlled storage and release of hydrogen in a safe manner are an urgent need for the utilization of H2 as transportation fuel, especially in low-temperature fuel cells. Compared with the classic pressurization or cryogenic liquefaction, using a proper chemical as the hydrogen carrier is currently the more desired approach [2,3]. Among the hydrogen storage chemicals investigated recently, formic acid (FA), which contains 4.4 wt% hydrogen, has attracted much attention because FA is a renewable, nontoxic source as a major product of biomass processes, and is stable at room temper⇑ Corresponding author. 1

E-mail address: [email protected] (J. Lu). These authors contributed equally.

http://dx.doi.org/10.1016/j.jcat.2017.06.007 0021-9517/Ó 2017 Published by Elsevier Inc.

ature [4–8]. Hydrogen stored in FA can be released using a highly efficient catalyst through the dehydrogenation pathway (HCOOH ? CO2 + H2) under a mild reaction condition. A number of experimental and theoretical studies suggested that formate (HCOO) is the key reaction intermediate [9–15]. Thus increasing the concentration of formate ion using sodium formate (SF) as an additive turned out to be an efficient way to accelerate H2 release [14,16–19]. On the other hand, HCOOH dehydration may also occur and generates low levels of the undesired product, CO (HCOOH ? CO + H2O) [3,5,20,21]. The worst scenario is that CO can even poison and deactivate the catalysts [22–27]. Hence, development of a catalyst to actively and selectively catalyze FA dehydrogenation to produce CO-free H2 at near ambient temperature is highly desirable [14,17,28–34]. Among other supported metal catalysts (e.g. Au, Rh, and Pt) [15,34–37], Pd-based heterogeneous catalysts have attracted significant attention due to their superior catalytic activities for FA decomposition [17,29–31,38–42]. Tremendous efforts have been devoted to improving the catalytic activities. One way is to import

372

J. Li et al. / Journal of Catalysis 352 (2017) 371–381

other elements to form alloys or core-shell structures. For instance, Tsang et al. reported that coating a thin layer of Pd atoms onto Ag nanoparticles (NPs) to form an Ag@Pd core-shell structure can enhance the activity by about 5 times at 20 °C compared with pure Pd or PdAg alloy nanoparticles in FA decomposition without any additive. The enhanced activity was attributed to both the presence of terrace sites in the Pd shell and electronic promotion by the Ag core [8]. Cai et al. reported that the borondoped Pd catalyst (Pd-B/C) was about 4 times more active than pure Pd, showing a turnover frequency (TOF) of 1184 h
1 from a FA-SF aqueous solution (FA:SF = 1.1:0.8) at 30 °C [17]. Xu et al. demonstrated that AuPd bimetallic nanoparticles immobilized in metal-organic framework had a much higher catalytic activity and a higher tolerance to CO poisoning than monometallic Au and Pd counterparts [42]. Chen et al. successfully synthesized carbon-supported AuPd alloy NPs with different sizes, and they found that the catalyst with smaller AuPd particle size had much higher catalytic activity than the larger ones in a FA-SF solution (FA:SF = 1:1) at 25 °C [43]. Jiang and coworkers showed that the Co0.3Au0.35Pd0.35/C trimetallic nanoalloy catalyst had significantly higher activity than that of monometallic (Pd/C, Au/C, and Co/C) and bimetallic (Au0.50Pd0.50/C, Co0.30Pd0.70/C, and Co0.30Au0.70/C) counterparts in FA decomposition without any extra additive at room temperature [39]. In case of monometallic Pd catalysts, the size of the Pd NPs can play an important role in catalyst activity. Xu et al. showed that highly dispersed Pd NPs (2.3 ± 0.4 nm) on nanoporous carbon MSC-30 had a much higher activity toward FA decomposition (a TOF of 750 h
1 at 25 °C) than a 3.6 ± 0.6 nm Pd catalyst on the same support in a FA-SF aqueous solution (FA:SF = 1:1) [44]. They further demonstrated in a later study that when the Pd particle size was reduced to 1.5 nm through potent alkalization of reduced graphene oxide with diamine, the TOF was also as high as 740 h
1 at 25 °C [14]. Independently, Jiang et al. also reported that decreasing the Pd size to about 2.8 nm with citric acid as dispersing agent significantly improved catalytic activity [19]. However, to our best knowledge, Yamashita and coworkers are the first ones to systematically investigate the Pd particle size effect by comparing the activities of carbon-supported Pd NPs ranging from 2.7 to 5.5 nm in FA decomposition without any additive. Therein, the Pd/C catalysts were synthesized by the polyol method using polyvinylpyrrolidone (PVP) as the protective agent [45]. They found that the 3.9 nm Pd/C catalyst had the highest activity and suggested that the reaction may primarily occur at the terrace Pd atoms with high coordination numbers based on a model of cuboctahedron-shape Pd NPs with a cubic closepacked structure. However, the use of PVP as a protective surfactant might be an issue, since the PVP capping agent can strongly bind to the surface of Pd NPs to largely reduce their activities. Indeed, such an issue has been reported in other studies [8,19]. For instance, Jiang et al. reported that the Pd/C catalyst with a Pd particle size of 2.3 nm, synthesized with PVP as a stabilizing agent, exhibited the lowest activity for the reaction. A dramatic enhancement in its activity was observed once the PVP was washed off [19]. In our present work, we synthesized Pd/C catalysts with five different Pd particle sizes in a range of 2.1–4.5 nm using a much weaker stabilizing agent of sodium citrate to eliminate the capping agent effect. The Pd particle sizes were determined by HAADFSTEM measurements. Evaluating their activities in FA decomposition from a FA-SF aqueous solution at room temperature, we found that reducing the Pd particle size could remarkably improve both catalytic activity and FA conversion. Herein, a TOF of 835 h
1 at 25 °C was observed on the 2.1 nm Pd/C catalyst, which is among the highest values for supported monometallic Pd catalysts reported in the literature.

2. Experimental 2.1. Catalyst preparation The Pd/C catalysts were prepared using sodium borohydride (NaBH4, 96%, Sinopharm Chemical Reagent Co. Ltd) as a reducing agent, and sodium citrate (Sinopharm Chemical Reagent Co. Ltd) as a stabilizing agent according to a procedure reported previously [40,46]. Carbon black (Vulcan XC72R, Carbot Corp.) was used as the catalyst support as received. The size of Pd nanoparticles was carefully controlled by varying the ratio of sodium citrate to palladium chloride (PdCl2, Aladdin) and by adjusting the reduction temperature by NaBH4. For instance, the procedure for the 2.1 nm Pd/C catalyst synthesis was as follows: 0.1 mmol PdCl2 (dissolved in 0.1 M HCl solution) and 0.8 mmol sodium citrate were dissolved into 150 mL water. 400 mg carbon black was then added. After stirring the mixture for 20 min, followed by 30 min sonication, 15 mL of 0.1 M NaBH4 solution was added into the suspension dropwise under vigorous stirring at 5 °C for 8 h. Next, the precipitate was filtered, and washed with deionized water several times to remove the weakly bonded sodium citrate agent. The obtained materials were then dried overnight in a vacuum oven at 25 °C to obtain the 2.1 nm Pd/C catalyst. For the 2.6 nm and 3.2 nm Pd/C catalysts, the NaBH4 reduction temperatures were 25 °C and 60 °C, respectively. The 3.8 nm Pd/C catalyst was obtained by changing the ratio of sodium citrate to PdCl2 to 2:1 and maintaining the reducing temperature at 25 °C. Finally, the 4.5 nm Pd/C catalyst was obtained by treating the 3.8 nm sample in 10% H2 in Ar at 300 °C for 60 min.

2.2. Catalyst characterization The Pd loadings and the possible B impurity were determined by an inductively coupled plasma atomic emission spectrometer (ICP-AES). The morphologies of the Pd/C catalysts were characterized on an aberration-corrected HAADF-STEM at 200 kV (JEOL2010F, University of Science and Technology of China (USTC)). The Pd particle size distribution was obtained by counting more than 200 particles from the HAADF-STEM images at different locations. Powder X-ray diffraction (XRD) measurements were carried out on a Philips X’Pert Pro Super diffractometer at the Structure Research Center at USTC, with a Cu-Ka radiation (k = 1.5418 Å), operated at 40 kV and 50 mA. Ex-situ X-ray photoelectron spectroscopy (XPS) measurements were taken on a Thermo-VG Scientific Escalab 250 spectrometer, equipped with an aluminum anode (Al Ka = 1486.6 eV) (Hefei University of Technology). In order to investigate the impact on the catalyst activity by the changes of oxidation states of Pd NPs through reduction at different temperatures, in situ XPS measurements were further carried out at the photoemission end-station at the beamline BL10B in the National Synchrotron Radiation Laboratory (NSRL) in Hefei, China. Briefly, the beamline is connected to a bending magnet and covers photon energies from 60 to 1000 eV. The end-station consists of four chambers, i.e., analysis chamber, preparation chamber, quick sample load-lock chamber and a high pressure reactor. The analysis chamber, with a base pressure of <2  10
10 torr, is connected to the beamline and equipped with a VG Scienta R3000 electron energy analyzer and a twin anode X-ray source. The samples were treated with different gases in the high pressure reactor. After sample treatment, the reactor can be pumped down to a high vacuum (<10
8 torr) for sample transfer. In the current work, each sample was first treated with the 10% H2 in Ar at 25 °C and 200 °C for 0.5 h in the high pressure reactor. Then the sample was transferred to the analysis chamber for XPS measurements without exposing to air. The binding ener-

J. Li et al. / Journal of Catalysis 352 (2017) 371–381

gies in all the XPS spectra were referenced to the C 1s binding energy of 284.8 eV. The metal dispersion was measured by the CO chemisorption method on a Micromeritics Autochem II 2920 instrument. 40 mg of the sample was pretreated in 10% O2/Ar at 100 °C for 30 min and 10% H2/Ar 100 °C for 30 min. Then the sample was cooled to room temperature and CO pulses were injected. A CO/Pd = 1/2 stoichiometry was taken in determining the Pd dispersion [47,48]. The particle sizes were calculated from the metal dispersion using the following equation [48]:

d¼

6A

where d is the diameter of metal particle, D is the metal dispersion, A = Pd atomic mass (106.42 g/mol), q= density (12.02  106 g/m3), r= average surface area occupied by one Pd atom (0.7910
19 m2), and L = Avogadro’s constant. 2.3. Catalytic activity test The catalytic activities of the Pd/C catalysts in FA decomposition were evaluated in a gas generation setup, described elsewhere [49,50]. In brief, the Pd/C catalyst and the FA (98%, Sinopharm Chemical Reagent Co. Ltd)/SF (99.5%, Sinopharm Chemical Reagent Co. Ltd) aqueous solution were loaded into a two-necked roundbottomed flask (50 mL), which was placed in a water bath at a preset temperature (25–45 °C) under an ambient atmosphere. A gas burette filled with water was connected to the reaction flask to measure the volume of released gas. Here SF was used as an additive. Typically, 55 mg Pd/C catalyst was first loaded into the reaction flask, and 5 mL FA substrate solution containing FA (0.6 M) and SF (0.6 M) was then added while vigorously stirring. The volume of the generated gas was immediately monitored by recording the displacement of water in the gas burette with reaction time. In some cases, pure FA (3 mmol) was subsequently added into the reaction flask to test the durability of the Pd/C catalysts after the reaction in the first run was stopped. Additionally, regeneration of the used catalysts was carried out by filtering and washing with deionized water, followed by drying in a vacuum oven at 80 °C for 15 h. The performance of regenerated catalysts was again evaluated using the same procedure described above. The carbon balance in the above reactions was above 95%. Here FA conversion and general TOF values were calculated as the following equations [16–18]:

ngas  100%; 2nFA

ðbÞ

Here ngas is the mole of final generated gas (H2 + CO2), and nFA is the mole of charged FA reactant.

TOF ¼

2.5. Detection of possible CO formation in FA decomposition 2.5.1. Online gas chromatography measurements The gas burette was disconnected from the reaction flask and replaced by an argon (Ar) gas line. Meanwhile, a gas outlet from the flask was connected to an online gas chromatography instrument (GC, Shimadzu GC-2014) equipped with a FID-Methanator. After the reaction started, Ar was bubbled into the reaction solution at a flow rate of 15 mL/min, and the generated gas was carried to the GC for analysis. A reference mixture gas containing 1% CO and 1% CO2 in Ar was also recorded for comparison.

ðaÞ

qrLD

FA conv ersion ð%Þ ¼

373

ngas 1  2nPd t

ðcÞ

Here ngas is the mole of generated gas at a conversion below 20%, nPd is the mole number of Pd surface atoms, t is the reaction time. 2.4. Online mass spectroscopy measurements The gas burette was disconnected from the reaction flask and replaced by an argon (Ar) gas line. Meanwhile, the gas outlet from the flask was connected to an online mass spectroscopy instrument (MS, OmniStar, Pfeiffer). After the reaction started, Ar was bubbled through the reaction solution at a flow rate of 10 mL/min, and the generated gas was carried to the MS for analysis.

2.5.2. In situ ATR-IR measurements In situ attenuated total reflection-infrared spectroscopy (ATRIR) measurements were performed on a Varian FTS 7000e IR spectrometer with an MCT detector [51,52]. The configuration of the thin-layer flow cell used in this study has been described elsewhere in detail [52]. A catalyst ink was first prepared by mixing 1 mL ethanol with 10 lL Nafion solution (0.5%) and 5 mg Pd/C catalyst. The Pd/C catalyst ink was then dropped onto a silicon prism covered with a monolayer of graphene via a pipette. The experimental details about synthesizing the monolayer graphene can be found elsewhere [53,54]. After drying at room temperature, the silicon prism with the Pd/C catalyst was loaded into the flow cell, and a background was recorded in water at 25 °C. After injecting the FA-SF mixture solution, IR spectra were recorded with a time resolution of 60 s/spectrum and a spectral resolution of 4 cm
1. All spectra are presented in absorbance. 2.6. Pyridine titration As a control experiment, pyridine titration of the positive Pd species was further carried out. In this case, the 2.1 and 3.8 nm Pd catalysts were both first exposed to the pyridine vapor (denoted as P-2.1 and P-3.8 nm, respectively); then, the pyridine-titrated Pd samples were tested in the FA dehydrogenation reaction under identical conditions and compared to the corresponding fresh samples. 3. Results and discussion 3.1. Morphology of Pd/C catalysts The Pd loadings in all the Pd/C samples were 2.3%, as determined by ICP-AES. While the possible B impurities introduced during catalyst synthesis were not detected on these samples. Aberration-corrected HAADF-STEM was employed to characterize the morphologies of these samples. As shown in Fig. 1a and b, highly-dispersed Pd NPs with a size of 2.1 ± 0.3 nm were formed on the carbon support when the NaBH4 reduction temperature was set at 5 °C, and the ratio of sodium citrate to PdCl2 was kept at 8:1. By increasing the NaBH4 reduction temperature to 25 or 60 °C during catalyst synthesis, Pd NPs of 2.6 ± 0.4 nm and 3.2 ± 0.3 nm were synthesized, respectively. This temperature effect is likely due to more aggressive reduction of PdCl2 by NaBH4 at elevated temperatures (Fig. 1c–f). On the other hand, reducing the sodium citrate to PdCl2 ratio to 2:1 resulted in a more pronounced effect on the nucleation of Pd NPs, leading to formation of 3.8 ± 0.5 nm of Pd NPs. The Pd particle sizes in the above samples were slightly smaller than the ones reported in the literature using the same method, which is probably due to a significantly lower Pd loading in our work [46]. To obtain a Pd/C catalyst with a larger size, the 3.8 nm Pd/C catalyst was reduced at 300 °C in 10% H2 in Ar for 60 min. STEM measurements showed that the Pd particle size was about 4.5 ± 0.5 nm (Fig. 1i and j). The Pd dis-

374

J. Li et al. / Journal of Catalysis 352 (2017) 371–381

uted to the diffraction peak from the carbon black support. No other obvious diffraction peaks were observed from the 2.1 and 2.6 nm Pd/C samples, indicating the dominant presence of ultrafine Pd NPs. A weak diffraction peak at 39.7°, assigned to Pd(1 1 1), appeared on the 3.2 and 3.8 nm Pd/C samples, implying the formation of slightly larger Pd NPs. Sharp diffraction peaks at 40.1°, 46.8°, and at 68.1°, assigned to (1 1 1), (2 0 0) and (2 2 0) of Pd [46], were only observed on the 4.5 nm Pd/C sample, indicating the significant growth of Pd particle in size after high temperature reduction. These results are consistent very well with the STEM images shown in Fig. 1. 3.2. Electronic properties of Pd/C catalysts XPS was carried out to study the electronic properties of the Pd NPs in these samples in the Pd 3d region. As shown in Fig. 2a, the Pd 3d5/2 binding energy was mainly located at 336.3 eV, with a pronounced shoulder at 338.4 eV on the 2.1 nm Pd/C sample, indicating the presence of a large fraction of oxidized Pd species. Deconvolution of the XPS spectrum showed that there was about 28% Pd4+ and 26.5% Pd2+ [40]. This observation is consistent very well with the literature, in which Zhou showed about 50% of oxidized Pd species (PdO and PdOx) on a 2.7 nm Pd/C sample [55]. The higher binding energy shoulder at 338.4 eV was considerably reduced on the 2.6 nm Pd/C sample, and became nearly invisible on the 3.2, 3.8 and 4.5 nm Pd/C samples (Fig. 2b–e). Meanwhile, the main Pd 3d5/2 peak, assigned to metallic Pd species, gradually shifted to lower binding energies with increasing Pd particle size, and was 335.8 eV on the 4.5 nm Pd/C sample. Comparison of these five samples was shown in Fig. 2f. Obviously, the fraction of oxidized Pd species increased significantly with decreasing Pd particle size. The observed charge deficiency on smaller Pd nanoparticles is likely attributed to the particle size effect and the strong interaction between Pd and the carbon support [40,55–57]. 3.3. Catalytic activity

Fig. 1. HAADF-STEM images of the (a) 2.1, (c) 2.6, (e) 3.2, (g) 3.8, and (i) 4.5 nm Pd/C catalysts and their corresponding Pd particle size distribution (b, d, f, h, and j).

persions of these Pd/C samples were measured using the CO chemisorption method. As shown in Table 1, the Pd dispersions were 68, 45, 31, 29 and 28% for the 2.1 2.6, 3.2, 3.8, 4.5 nm Pd samples, respectively. The calculated Pd particle sizes based on the metal dispersions were consistent well with the ones obtained by HAADF-STEM, strongly indicating the surface of Pd NPs in the Pd/C catalysts was clean. XRD measurements were also performed on these samples. As shown in Fig. S1 in the supporting information, a broad diffraction peak at 43° was observed on all the samples, which can be attrib-

The activities of these Pd/C catalysts were evaluated in FA dehydrogenation in a FA-SF aqueous solution at 25 °C. As shown in Fig. 3a, we found that the 2.1 nm Pd/C catalyst showed the highest activity, generating approximately 132 mL gas (CO2 + H2) in 45 min, which was about 92% FA conversion. Upon increasing the Pd particle size, both the activity and volume of released gas significantly decreased. On the 4.5 nm Pd/C catalyst, it only released less than 80 mL gas (a FA conversion of 56%) in about 160 min. The observation of continuous decrease in both activity and FA conversion with increasing Pd particle size in our work was in sharp contrast with the previous study where Yamashita and coworkers reported that the 3.9 nm Pd/C catalyst showed the highest activity compared with others with a Pd particle size ranging from 2.7 to 5.5 nm [45]. To note that there was no any mass transfer issue during reaction according to the Madon-Boudart test, where the plot of gas generation rates as a function of the Pd concentration yielded a straight line (Fig. S2) [58–60]. However, we found that the apparent activation energies (Ea) for FA dehydrogenation at these samples were essentially similar, about 43 to 48 kJ/mol (Fig. 3b and Fig. S3 and Table S1), implying that a similar type of active site is present in these samples. Besides the above, it is worth noting that the gas generation rate remained nearly constant on the 2.1 nm catalyst during the first 30 min, but it gradually decreased on Pd/C catalysts with larger sizes with reaction time. The gradual decrease in gas generation rates suggests the gradual deactivation of Pd/C catalysts during FA dehydrogenation. Indeed, the catalytic activities of the 2.1 and 4.5 nm samples both dropped significantly in the second run, when 3 mmol of pure FA was added into the reaction flask after the first

375

J. Li et al. / Journal of Catalysis 352 (2017) 371–381 Table 1 Dispersions and sizes of Pd NPs in these five Pd/C catalysts.

a b

Samples (nm)

CO uptake (lmol/g)

Dispersion (%)

dchem (nm)a

dTEM (nm)b

2.1 2.6 3.2 3.8 4.5

70.8 48.6 34.1 29.6 26.1

62 45 31 29 28

1.8 2.4 3.6 3.8 4.1

2.1 2.6 3.2 3.8 4.5

The average Pd particle diameter calculated based on the Pd dispersion. The average Pd particle diameter determined by STEM.

Fig. 2. Ex-situ XPS spectra of the (a) 2.1, (b) 2.6, (c) 3.2, (d) 3.8, and (e) 4.5 nm Pd/C catalysts in the Pd 3d region. (f) Comparison of the proportion of Pd4+, Pd2+ and Pd0 species in these five Pd/C catalysts.

run (Fig. 3c and d). We will discuss the factors causing the catalyst deactivation in detail in the later sections. As an additive, SF can significantly accelerate the FA dehydrogenation reaction [14,16,18,19]. In our work, we also found that

the higher SF concentration would yield a higher reaction rate, but SF itself does not produce any gas generation since the solution has to be electronically neutral during reaction (Fig. S4a). Apparently, all generated gases (CO2 + H2) were produced by FA, validat-

376

J. Li et al. / Journal of Catalysis 352 (2017) 371–381

Fig. 3. (a) Plots of the volume of the generated gas (CO2 + H2) versus time from 5 mL solution of 0.6 M FA + 0.6 M SF in the presence of 55 mg of Pd/C (2.3 wt% Pd) catalysts with five different particle sizes at 25 °C under ambient atmosphere. (b) Arrhenius plots of these five Pd/C catalysts. Durability tests over the 2.1 (c) and 4.5 nm (d) Pd/C catalysts by adding additional 3 mmol of pure FA into the reaction flask after the first run.

ing the calculation of FA conversion using the equation (b) [16–18]. On the contrary, varying the FA concentration while keeping the SF concentration constant would have a similar initial reaction rate (Fig. S4b). Clearly, formate ion is the key reaction intermediate [11–14,61]. While it should be noted that the catalyst reacts formate ion from formic acid and sodium formate indistinguishably. In the FA-SF aqueous solution, the SF additive provides a constant concentration of formate ion, thus the ‘‘straight-line” curve in catalytic performance was observed on the 2.1 nm Pd/C catalyst (Fig. 3a). In addition, we also noticed that the pH values of the FA-SF aqueous solution before and after reaction were 2 and 8, respectively, which suggests that the pH of the FA-SF aqueous solution gradually increased as increasing the FA conversion. The ‘‘straight-line” reaction curve on the 2.1 nm Pd/catalyst (Fig. 3a) indicates that the reaction rate was independent on the pH values, and the pH of the solution might play a less important role in catalytic activity, compared to the concentration of formate ions. 3.4. Particle size effect It is well known that the surface of metal NPs is composed of different types of sites, classified as high-coordination terrace atoms and low-coordination atoms (edges and corners). The relative proportion of these sites varies with the metal particle size. In the case of Pd NPs, a cuboctahedron-shaped particle model with a cubic close-packed structure (the inset of Fig. 4) is often employed to

Fig. 4. Calculated TOFLS, TOFHS, TOFTS and TOFChem as a function of the Pd particles size over these five Pd/C catalysts. The inset is a schematic illustration of highcoordinated surface atoms (HS) and low-coordinated surface atoms (LS) on a cuboctahedron-shaped Pd NP.

understand their roles in catalytic reactions [45,62–65]. Here, we adopted this model and calculated the initial activities of these samples based on the number of high-coordinated surface atoms

377

J. Li et al. / Journal of Catalysis 352 (2017) 371–381

Fig. 5. In situ XPS spectra of the 2.1 and 3.8 nm Pd/C samples in the Pd 3d region after reduction at different temperatures: The 2.1 (a) and 3.8 nm (c) samples reduced at 25 °C; the 2.1 (c) and 3.8 nm (d) samples reduced at 200 °C.

Table 2 The changes of Pd dispersion, electronic properties, and catalytic activity of the 2.1 and 3.8 nm Pd/C catalysts after reducing at 25 and 200 °C in FA dehydrogenation. Samples (nm)

Reduction temp (°C)

Dispersion (%)

Pd 2.1 2.1 3.8 3.8 a

25 200 25 200

– 68 – 25

TOF (h
1)a

Fraction of Pd species (%) 4+

13 0 12 0

Pd

2+

27 14 19 15

Pd

0

60 86 69 85

835 253 405 285

TOF values were calculated based on the metal dispersion. Reaction conditions: 5 mL solution of 0.6 M FA + 0.6 M SF in the presence of 55 mg of Pd/C at 25 °C.

(NHS), low-coordinated surface atoms (NLS), and total number of surface atoms (NTS, NTS = NHS + NLS), respectively, as shown in Fig. 4. The average Pd particle size was used to calculate NHS, NLS, and NTS for each Pd/C sample. The calculation equations can be found elsewhere [45,62–65]. We found that the calculated TOFs based on NHS (TOFHS) and NLS (TOFLS) were both strongly dependent on the particle size, while the TOFs normalized to NTS (TOFTS) were more close to be independent of the Pd particle size (Fig. 4). The results imply that all the Pd surface atoms act as active sites and participate in the FA decomposition reaction, in contrast to the previously-suggested mechanism in which the reaction primarily occurs at the terrace Pd atoms with high coordination numbers [45]. TOFs on these five Pd/C catalysts were also calculated based on the number of Pd surface atoms determined by CO chemisorption (TOFChem). We found that these TOFChem values showed a very similar trend with TOFTS as a function of Pd particle size. It is worth noting that the TOF of 835 h
1 on the 2.1 nm Pd/C sample based on

the Pd surface atoms at 25 °C is among the highest values for monometallic Pd catalysts ever reported in the literature [14,17,19,44,45]. 3.5. Electronic effect Remarkable improvements in catalytic activity via electronic promotion have often been observed in bimetallic or trimetallic catalyst systems [8,39,66–70]. In our work, higher TOFs achieved on smaller Pd particles here (Fig. 4), are likely due to the presence of a larger fraction of positively charged Pd species, as indicated by the ex-situ XPS results shown in Fig. 2. To further address the electronic promotion on the Pd/C catalyst activity, in situ XPS measurements of the 2.1 and 3.8 nm Pd/C samples were carried out after 25 and 200 °C reduction in 10% H2/Ar. After reduction at 25 °C, a significant amount of positive Pd species (40%) was still present on the 2.1 nm Pd/C sample, although the fraction of Pd4+ species

378

J. Li et al. / Journal of Catalysis 352 (2017) 371–381

Fig. 6. Plots of the volume of the generated gases (CO2 + H2) versus time from 5 mL solution of 0.6 M FA + 0.6 M SF at 25 °C on the 2.1 (a) and 3.8 nm (b) Pd/C samples, which were pre-reduced at 25 and 200 °C, respectively.

was remarkably reduced (comparing Fig. 2a with Fig. 5a). However, the 3.8 nm Pd/C sample showed a similar fraction of positive Pd species (31%) before and after reduction at 25 °C (comparing Fig. 2d with Fig. 5b). Here the reduction condition at 25 °C in H2 is expected to be a more reducing environment than the aqueous solution reaction condition, although a large amount of H2 is generated. Therefore, the above results strongly suggested that it is certainly possible to preserve the positively charged Pd species under reaction conditions. Nevertheless, operando characterizations are still highly desirable to further reveal the electronic state of Pd species in the FA-SF solution in the near feature. After reduction at 200 °C, the fraction of positive Pd species decreased significantly from 40% to 14% on the 2.1 sample and 31% to 15% on the 3.8 nm sample, respectively, (Fig. 5c and d). However, we did not see considerable changes in the Pd dispersions (Table 2). When these two pre-reduced samples were evaluated again in the FA dehydrogenation reaction in a FA-SF aqueous solution at 25 °C. We found that the activities of these two samples both largely decreased after reduction at 200 °C (Fig. 6). Obviously, the above activity changes induced by reduction at 200 °C provide strong evidence that the presence of a larger proportion of Pd species with positive charge indeed has a significant impact on the activity enhancement. Meanwhile, we also noticed that the 2.1 and 3.8 nm Pd/C samples reduced at 200 °C had similar fractions of positively charged Pd species of 14% and 15%, resulting similar TOFs of 253 and 285 h
1 (Table 2). Therefore, these results again strongly suggest that all the Pd surface atoms are the active sites

in the FA decomposition reaction, in a good agreement with our analysis based on the cuboctahedron-shaped particle model shown in Fig. 4. To note that electron withdrawing oxygen species were observed on the XC72R carbon support (Fig. S5). We speculate that these interfacial oxygen species between Pd NPs and the carbon support are likely the reason for the formation of positive Pd species in our Pd/C samples. When positive Pd species are present in the Pd/C samples, the cationic Pd can behave as a Lewis acid site, which could strongly bond to a base molecule, such as pyridine. Therefore, a control experiment of pyridine titration was further carried out, as shown in Fig. 7. We found that the initial TOF decreased about 57% from 835 to 360 h
1 on the 2.1 nm sample by pyridine titration, while the decrease of the initial TOF of the 3.8 nm sample was only about 24% (from 405 to 310 h
1). The larger activity decrease in the 2.1 nm sample by pyridine titration is attributed to a stronger interaction between pyridine and the higher fraction of positive Pd species in the 2.1 nm Pd/C sample as observed by XPS (Figs. 2 and 5). Therefore, the results of pyridine titration again provide evidence that a larger fraction of positively charged Pd species were present in the 2.1 nm Pd/C sample. After reaction for 10–15 min, the reaction rates were accelerated on both samples (Fig. 7a, b), which were even higher than the fresh samples. The unexpected result is likely caused by two factors: Firstly, Pyridine is prone to dissolve into water, thereby evoking the Pd catalytic activity; Secondly, the dissolved pyridine in water could act as a base, which can further facilitate dissociation of the OAH bond in FA, and enhance the catalytic activity at the later stage [28,61]. To further discriminate whether the positive Pd species are PdOx or not, an additional control experiment using an online MS instrument was conducted to analyze the composition of the generated gases. Here the as-prepared and oxidized 2.1 nm Pd/C catalysts were examined in the FA decomposition reaction under identical conditions, where the oxidized catalyst was obtained by calcination in 10% O2 in Ar at 200 °C for 30 min before the reaction. On the as-prepared 2.1 nm Pd/C sample, the intensities of CO2 (m/ z = 44) and H2 (m/z = 2) signals were nearly identical during the entire reaction (Fig. 8a). While on the oxidized 2.1 nm Pd/C catalyst, tremendous amount of CO2 was first generated in the initial reaction period, and much less amount of H2 was only observed after the reaction for 500 s (Fig. 8b). The CO2 to H2 ratio was much higher than the one obtained on the fresh catalyst. Obviously, the positive Pd species on the fresh 2.1 nm Pd/C sample are not PdOx species, but rather the electron deficient species by electron transfer from Pd to the carbon support through the metal-support interaction, similar to the literature [71]. The negative charged formate ion is known to be the key reaction intermediate in the reaction [11–14,61], thus the enhanced Coulomb interaction between positive charged Pd species and negative formate ion might be the major reason for promoting the reaction. On the oxidized Pd surface, we speculated that the oxidative dehydrogenation reaction (HCOOH + O⁄ ? CO2 (gas) + H2O (liquid)) likely occurred, thus yielding CO2 as the dominate gaseous product detected by online MS in the initial reaction period (the asterisk designates a surface species). Besides above, we also noticed that the intensity of m/z = 28 signal (either CO or N2 from air leaking) was trivial, compared to the CO2 intensity. This observation consists with the considerably high reaction barriers for the dehydration path by theoretical calculations in the literature [72,73]. For instance, Tian and coworkers reported that the reaction barrier for direct CO formation is as high as 1.97 eV in the decomposition of HCOOH on a Pd7 cluster in solution [73]. Moreover, adsorption of hydroxyls from the aqueous solution onto the Pd surface might also hinder the CO formation.

J. Li et al. / Journal of Catalysis 352 (2017) 371–381

379

Fig. 7. Pyridine titration effect on the Pd/C catalysts. Volume of the generated gas (CO2 + H2) versus time for the dehydrogenation of FA at 25 °C under ambient atmosphere over the 2.1 (a) and 3.8 nm (b) Pd/C catalysts with and without pyridine titration. (c) The enlargements of the detailed data in the dotted boxes in (a) and (b). (d) Calculated initial TOFs over 2.1, 3.8, P-2.1 and P-3.8 nm Pd/C catalysts.

Fig. 8. Online Mass Spectrum of generated gas from the FA/SF solution over the as-prepared (a) and pretreated in 10% O2/Ar at 200 °C for 0.5 h (b) over 2.1 nm Pd/C catalysts.

Therefore, the FA dehydrogenation is the dominate reaction pathway on these Pd catalysts, and the dehydration path to CO formation at the low-coordinated Pd sites suggested by Tsang et al. was less pronounced [74].

3.6. Catalyst deactivation and regeneration To further verify whether the CO product of the dehydration pathway was released as a gas during the reaction, in situ monitor-

380

J. Li et al. / Journal of Catalysis 352 (2017) 371–381

might have weakened the adsorption of such strong adsorbates on Pd. Nonetheless, the detailed deactivation mechanism requires further investigations. 4. Conclusions We showed that carbon-supported Pd NPs catalysts with different Pd particle sizes can be controlled in a range of 2.1–4.5 nm. These Pd/C catalysts were highly active in FA dehydrogenation from the FA-SF aqueous solution at 25 °C. The 2.1 nm Pd/C catalyst showed a TOF of 835 h
1, which is among the highest values for monometallic Pd catalysts reported in the literature. Besides that, we showed that the Pd particle size can greatly influence the catalytic activity in FA decomposition. By adopting the cuboctahedron-shape model for Pd NPs, our results suggest that both low- and high-coordination Pd surface atoms are the active sites in FA dehydrogenation. The significantly higher activity of smaller Pd NPs relative to the larger ones was attributed to both the higher dispersion and the presence of a larger proportion of Pd species with positive charge, according to our XPS measurements. The enhanced Coulomb interaction between positive Pd species and negative charged formate ion might be the major reason for the promotion. It is noteworthy that the deactivation of Pd/C catalysts can be fully regenerated by removal of chemisorbed intermediate species. Finally, our work suggests that downsizing the Pd particle size could be an efficient way to improve catalytic activities for hydrogen generation from FA dehydrogenation. Acknowledgments

Fig. 9. Plots of the volume of the generated gas (CO2 + H2) versus time from a 5 mL solution of 0.6 M FA + 0.6 M SF over the as-prepared and regenerated 2.1 and 4.5 nm Pd/C catalysts at 25 °C under ambient atmosphere. Pd/C catalyst regeneration was performed via vacuum drying at 80 °C.

ing of the composition of the released gas by an online GC was also carried out on the 2.1 and 4.5 nm Pd/C catalysts. As shown in Fig. S6, the CO2 peak generated by FA decomposition over the 2.1 nm Pd/C was much larger than that obtained for the 4.5 nm Pd/C, again suggesting higher activity of 2.1 nm Pd/C. However, according to the reference mixture gas, CO was below the detection limit during reaction in either case, consistent with our online MS results shown in Fig. 8. We also tempted to evaluate the possibility of formation of a low level of chemisorbed CO on the Pd NPs surface using in situ ATR-IR [17]. However, we found that the sensitivity was rather low, due to the strong absorbance of the carbon support (Fig. S7). The presence of a trace amount of CO on the Pd surface still cannot be totally ruled out. Hu and coworkers reported that the continuous loss of Pd/C catalyst activity can be also caused by the occupation of active sites by proton, CO2, H2O and HCOO intermediates [75]. To evaluate this possibility, the 2.1 nm and 4.5 nm Pd/C catalysts after the first run were collected and dried in a vacuum oven at 80 °C for 15 h to remove the possible chemisorbed intermediate species. The second run of the FA decomposition was then carried out on these two samples. As shown in Fig. 9, the catalytic activities of the two samples were totally recovered, in good agreement with the previous work [75]. These results suggest that the deactivation of Pd/C catalysts was mainly caused by the blockage of active sites by certain adsorbates, such as CO, while aggregation of Pd NPs nanoparticles was fairly impossible. Remarkable improvements in FA conversions with reducing Pd particle size, as shown in Fig. 3a, imply that the positive charge of Pd species present on the smaller Pd NPs

This work was supported by the One Thousand Young Talents Program under the Recruitment Program of Global Experts, the National Natural Science Foundation of China (21673215, 21473169 and 51402283), and the Fundamental Research Funds for the Central Universities (WK2060190026, and WK2060030017), and the startup funds from University of Science and Technology of China. We also thank one of the reviewers for suggesting us to consider the role of formate ion. We also thank Prof. Yanwu Zhu and Mr. Zhuchen Tao for their help in the preparation of the graphene coated Si prism. Finally, the authors also gratefully thank the BL10B beamline at National Synchrotron Radiation Laboratory (NSRL), China. Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.jcat.2017.06.007. References [1] S. Young, Nature 414 (2001) 487–488. [2] A. Boddien, B. Loges, F. Gartner, C. Torborg, K. Fumino, H. Junge, R. Ludwig, M. Beller, J. Am. Chem. Soc. 132 (2010) 8924–8934. [3] M. Yadav, Q. Xu, Energy Environ. Sci. 5 (2012) 9698–9725. [4] D.T. Whipple, P.J.A. Kenis, J. Phys. Chem. Lett. 1 (2010) 3451–3458. [5] M. Grasemann, G. Laurenczy, Energy Environ. Sci. 5 (2012) 8171–8181. [6] W.H. Wang, M.G. Niu, Y.C. Hou, W.Z. Wu, Z.Y. Liu, Q.Y. Liu, S.H. Ren, K.N. Marsh, Green Chem. 16 (2014) 2614–2618. [7] A. Boddien, F. Gärtner, C. Federsel, P. Sponholz, D. Mellmann, R. Jackstell, H. Junge, M. Beller, Angew. Chem. Int. Ed. 50 (2011) 6411–6414. [8] K. Tedsree, T. Li, S. Jones, C.W.A. Chan, K.M.K. Yu, P.A.J. Bagot, E.A. Marquis, G.D. W. Smith, S.C.E. Tsang, Nat. Nanotechnol. 6 (2011) 302–307. [9] N. He, Z.H. Li, Phys. Chem. Chem. Phys. 18 (2016) 10005–10017. [10] P. Wang, S.N. Steinmann, G. Fu, C. Michel, P. Sautet, ACS Catal. 7 (2017) 1955– 1959. [11] L. Jia, D.A. Bulushev, S. Beloshapkin, J.R.H. Ross, Appl. Catal. B- Environ. 160– 161 (2014) 35–43. [12] L. Jia, D.A. Bulushev, J.R.H. Ross, Catal. Today 259 (2016) 453–459. [13] D.A. Bulushev, L. Jia, S. Beloshapkin, J.R. Ross, Chem. Commun. 48 (2012) 4184– 4186.

J. Li et al. / Journal of Catalysis 352 (2017) 371–381 [14] F.-Z. Song, Q.-L. Zhu, N. Tsumori, Q. Xu, ACS Catal. 5 (2015) 5141–5144. [15] Q.G. Liu, X.F. Yang, Y.Q. Huang, S.T. Xu, X. Su, X.L. Pan, J.M. Xu, A.Q. Wang, C.H. Liang, X.K. Wang, T. Zhang, Energ, Environ. Sci. 8 (2015) 3204–3207. [16] Y. Chen, Q.L. Zhu, N. Tsumori, Q. Xu, J. Am. Chem. Soc. 137 (2015) 106–109. [17] K. Jiang, K. Xu, S. Zou, W.B. Cai, J Am Chem Soc 136 (2014) 4861–4864. [18] Q.L. Zhu, N. Tsumori, Q. Xu, J. Am. Chem. Soc. 137 (2015) 11743–11748. [19] Z.L. Wang, J.M. Yan, H.L. Wang, Y. Ping, Q. Jiang, Sci. Rep. 2 (2012) 598. [20] F. Joo, Chemsuschem 1 (2008) 805–808. [21] B. Loges, A. Boddien, H. Junge, M. Beller, Angew. Chem. Int. Ed. 47 (2008) 3962– 3965. [22] Z.L. Liu, L. Hong, M.P. Tham, T.H. Lim, H.X. Jiang, J. Power Sources 161 (2006) 831–835. [23] P. Gruene, A. Fielicke, G. Meijer, D.M. Rayner, Phys. Chem. Chem. Phys. 10 (2008) 6144–6149. [24] C. Lamy, A. Lima, V. LeRhun, F. Delime, C. Coutanceau, J.M. Leger, J. Power Sources 105 (2002) 283–296. [25] V. Bambagioni, C. Bianchini, A. Marchionni, J. Filippi, F. Vizza, J. Teddy, P. Serp, M. Zhiani, J. Power Sources 190 (2009) 241–251. [26] Y. Xu, A.V. Ruban, M. Mavrikakis, J. Am. Chem. Soc. 126 (2004) 4717–4725. [27] A. Boddien, B. Loges, H. Junge, M. Beller, Chemsuschem 1 (2008) 751–758. [28] Y. Karatas, A. Bulut, M. Yurderi, I.E. Ertas, O. Alal, M. Gulcan, M. Celebi, H. Kivrak, M. Kaya, M. Zahmakiran, Appl. Catal. B- Environ. 180 (2016) 586–595. [29] C. Hu, X. Mu, J. Fan, H. Ma, X. Zhao, G. Chen, Z. Zhou, N. Zheng, ChemNanoMat 2 (2016) 28–32. [30] S.-J. Li, Y. Ping, J.-M. Yan, H.-L. Wang, M. Wu, Q. Jiang, J. Mater. Chem. A 3 (2015) 14535–14538. [31] N. Cao, S. Tan, W. Luo, K. Hu, G. Cheng, Catal. Lett. 146 (2015) 518–524. [32] D.W. Lee, M.H. Jin, Y.J. Lee, J.H. Park, C.B. Lee, J.S. Park, Sci. Rep. 6 (2016) 26474. [33] A. Boddien, D. Mellmann, F. Gartner, R. Jackstell, H. Junge, P.J. Dyson, G. Laurenczy, R. Ludwig, M. Beller, Science 333 (2011) 1733–1736. [34] Q.Y. Bi, X.L. Du, Y.M. Liu, Y. Cao, H.Y. He, K.N. Fan, J. Am. Chem. Soc. 134 (2012) 8926–8933. [35] J.K. Sun, W.W. Zhan, T. Akita, Q. Xu, J. Am. Chem. Soc. 137 (2015) 7063–7066. [36] N. Yi, H. Saltsburg, M. Flytzani-Stephanopoulos, Chemsuschem 6 (2013) 816– 819. [37] F. Solymosi, A. Koos, N. Liliom, I. Ugrai, J. Catal. 279 (2011) 213–219. [38] J.S. Yoo, Z.-J. Zhao, J.K. Nørskov, F. Studt, ACS Catal. 5 (2015) 6579–6586. [39] Z.L. Wang, J.M. Yan, Y. Ping, H.L. Wang, W.T. Zheng, Q. Jiang, Angew. Chem. Int. Ed. 52 (2013) 4406–4409. [40] X. Zhou, Y. Huang, W. Xing, C. Liu, J. Liao, T. Lu, Chem. Commun. 30 (2008) 3540–3542. [41] Y.J. Huang, X.C. Zhou, M. Yin, C.P. Liu, W. Xing, Chem. Mater. 22 (2010) 5122– 5128. [42] X.J. Gu, Z.H. Lu, H.L. Jiang, T. Akita, Q. Xu, J. Am. Chem. Soc. 133 (2011) (1825) 11822–11825. [43] J. Cheng, X.J. Gu, X.L. Sheng, P.L. Liu, H.Q. Su, J. Mater. Chem. A 4 (2016) 1887– 1894. [44] Q.-L. Zhu, N. Tsumori, Q. Xu, Chem. Sci. 5 (2014) 195–199. [45] M. Navlani-García, K. Mori, A. Nozaki, Y. Kuwahara, H. Yamashita, ChemistrySelect 1 (2016) 1879–1886. [46] D.F. Gao, H. Zhou, J. Wang, S. Miao, F. Yang, G.X. Wang, J.G. Wang, X.H. Bao, J. Am. Chem. Soc. 137 (2015) 4288–4291.

381

[47] P. Canton, G. Fagherazzi, M. Battagliarin, F. Menegazzo, F. Pinna, N. Pernicone, Langmuir 18 (2002) 6530–6535. [48] T. Lear, R. Marshall, J.A. Lopez-Sanchez, S.D. Jackson, T.M. Klapotke, M. Baumer, G. Rupprechter, H.J. Freund, D. Lennon, J. Chem. Phys. 123 (2005). [49] Y. Yamada, K. Yano, S. Fukuzumi, Energy Environ. Sci. 5 (2012) 5356–5363. [50] H.L. Jiang, Q. Xu, Catal. Today 170 (2011) 56–63. [51] S.X. Liu, L.W. Liao, Q. Tao, Y.X. Chen, S. Ye, Phys. Chem. Chem. Phys. 13 (2011) 9725–9735. [52] Y.X. Chen, M. Heinen, Z. Jusys, R.J. Behm, Angew. Chem. Int. Ed. 45 (2006) 981– 985. [53] Y. Yao, W. Chen, Y.X. Du, Z.C. Tao, Y.W. Zhu, Y.X. Chen, J. Phys. Chem. C 119 (2015) 22452–22459. [54] S. Bae, H. Kim, Y. Lee, X.F. Xu, J.S. Park, Y. Zheng, J. Balakrishnan, T. Lei, H.R. Kim, Y.I. Song, Y.J. Kim, K.S. Kim, B. Ozyilmaz, J.H. Ahn, B.H. Hong, S. Iijima, Nat. Nanotechnol. 5 (2010) 574–578. [55] W. Zhou, M. Li, O.L. Ding, S.H. Chan, L. Zhang, Y. Xue, Int. J. Hydrogen Energy 39 (2014) 6433–6442. [56] W.P. Zhou, A. Lewera, R. Larsen, R.I. Masel, P.S. Bagus, A. Wieckowski, J. Phys. Chem. B 110 (2006) 13393–13398. [57] A.L.D. Ramos, P.D. Alves, D.A.G. Aranda, M. Schmal, Appl. Catal. A-Gen. 277 (2004) 71–81. [58] R.J. Madon, M. Boudart, Ind. Eng. Chem. Fund 21 (1982) 438–447. [59] R. Miao, J.K. He, S. Sahoo, Z. Luo, W. Zhong, S.Y. Chen, C. Guild, T. Jafari, B. Dutta, S.A. Cetegen, M.C. Wang, S.P. Alpay, S.L. Suib, ACS Catal. 7 (2017) 819–832. [60] J. Ohyama, Y. Hayashi, K. Ueda, Y. Yamamoto, S. Arai, A. Satsuma, J. Phys. Chem. C 120 (2016) 15129–15136. [61] K. Mori, M. Dojo, H. Yamashita, Acs Catal. 3 (2013) 1114–1119. [62] K. Mori, T. Hara, T. Mizugaki, K. Ebitani, K. Kaneda, J. Am. Chem. Soc. 126 (2004) 10657–10666. [63] R.E. Benfield, J. Chem. Soc. Faraday T 88 (1992) 1107–1110. [64] B. Veisz, Z. Kiraly, L. Toth, B. Pecz, Chem. Mater. 14 (2002) 2882–2888. [65] D. Ferri, C. Mondelli, F. Krumeich, A. Baiker, J. Phys. Chem. B 110 (2006) 22982– 22986. [66] H.W. Wang, C.L. Wang, H. Yan, H. Yi, J.L. Lu, J. Catal. 324 (2015) 59–68. [67] F. Maroun, F. Ozanam, O.M. Magnussen, R.J. Behm, Science 293 (2001) 1811– 1814. [68] Z. Guo, B. Liu, Q.H. Zhang, W.P. Deng, Y. Wang, Y.H. Yang, Chem. Soc. Rev. 43 (2014) 3480–3524. [69] J. Greeley, T.F. Jaramillo, J. Bonde, I.B. Chorkendorff, J.K. Norskov, Nat. Mater. 5 (2006) 909–913. [70] F. Gao, D.W. Goodman, Chem. Soc. Rev. 41 (2012) 8009–8020. [71] R. Arrigo, M.E. Schuster, Z. Xie, Y. Yi, G. Wowsnick, L.L. Sun, K.E. Hermann, M. Friedrich, P. Kast, M. Hävecker, A. Knop-Gericke, R. Schlögl, ACS Catal. 5 (2015) 2740–2753. [72] Y. Wang, Y. Qi, D. Zhang, C. Liu, J. Phys. Chem. C 118 (2014) 2067–2076. [73] S.J. Li, X. Zhou, W.Q. Tian, J. Phys. Chem. A 116 (2012) 11745–11752. [74] S. Jones, S.M. Fairclough, M. Gordon-Brown, W. Zheng, A. Kolpin, B. Pang, W.C. Kuo, J.M. Smith, S.C. Tsang, Chem. Commun. 51 (2015) 46–49. [75] C.Q. Hu, J.K. Pulleri, S.W. Ting, K.Y. Chan, Int. J. Hydrogen Energy 39 (2014) 381–390.