Improving the electrocatalytic properties of Pd-based catalyst for direct alcohol fuel cells: effect of solid solution

The tolerance of the electrode against the CO species absorbed upon the surface presents the biggest dilemma of the alcohol fuel cells. Here we report for the first time that the inclusion of (Zr, Ce)O2 solid solution as the supporting material can significantly improve the anti-CO-poisoning as well as the activity of Pd/C catalyst for ethylene glycol electro-oxidation in KOH medium. In particular, the physical origin of the improved electrocatalytic properties has been unraveled by first principle calculations. The 3D stereoscopic Pd cluster on the surface of (Zr, Ce)O2 solid solution leads to weaker Pd-C bonding and smaller CO desorption driving force. These results support that the Pd/ZrO2-CeO2/C composite catalyst could be used as a promising effective candidate for direct alcohol fuel cells application.

In this study, Pd/C, Pd/ZrO 2 /C, Pd/CeO 2 /C and Pd/ZrO 2 -CeO 2 /C composite catalysts have been prepared and characterized. The effect of the solid solution of ZrO 2 -CeO 2 binary oxides as support materials on the electrocatalytic performance of Pd-based catalysts has been inverstigated. The physical origin of the good anti-CO-poisoning property of Pd 4 /ZrO 2 -CeO 2 (111) composite catalyst has been unraveled based on first principle calculations as well.

Results
Structural characterization. The TEM images of different catalysts are shown in Fig. 1a-d. It is clearly that Pd nanoparticles of all catalysts are uniform and monodisperse. The average size of Pd is 4.8 nm for Pd/C, 4.1 nm for Pd/CeO 2 /C, 4.6 nm for Pd/ZrO 2 /C, and 4.0 nm for Pd/ZrO 2 -CeO 2 /C, respectively, as shown in Fig. 1a-d, consistent with that calculated on XRD data (in Fig. S1). It is noticed that the particle size of Pd changes more significantly with the addition of CeO 2 than ZrO 2 in this work. Similar results have been reported in Pt catalysts modified by Ce 0.6 Zr 0.4 O 2 31, 32 . It is observed that some grains with a d-spacing of 0.227 nm from the HRTEM images of all catalysts in Fig. 1e-h, which is assigned for the (111) plane of cubic Pd. And the d-spacing of some grains is 0.276 nm in Pd/CeO 2 /C (Fig. 1g), corresponding for the (200) plane of CeO 2 . While the d-spacing of some grains in Pd/ ZrO 2 -CeO 2 /C (Fig. 1h) turns out to be 0.274 nm, which is smaller than that of Pd/CeO 2 /C. This could be related to the replacement of Ce 4+ ions by Zr 4+ ions, which results in a shrinking of the lattice volume of the solid. Based on the HRTEM results, one can confirm that the addition of 50% ZrO 2 contributes to the formation of (Zr, Ce) O 2 solid solution in present work.
XPS was employed to investigate the nature of surface species for the catalysts. Figure 2 shows the XPS spectra of Zr 3d and Ce 3d in different species before and after Pd-loading. It is evident that the binding energy of Zr 3d 5/2 slightly increases from 182.3 eV for ZrO 2 /C to 182.5 eV for Pd/ZrO 2 /C (Fig. 2a); while the binding energy of Ce 3d 5/2 changes from 883.1 eV for CeO 2 /C to 883.6 eV for Pd/CeO 2 /C (Fig. 2b), which reveals the effect of Pd-loading on the electronic structure of Zr 3d or Ce 3d. Similar change in binding energy has been observed in Zr 3d 5/2 (Zr 3d 3/2 ) as well as Ce 3d 5/2 for ZrO 2 -CeO 2 /C after Pd-loading, as shown in Fig. 2c-d. The presence of ZrO 2 -CeO 2 in the vicinity of Pd nanoparticles could be beneficial for the EG and CO oxidation 36 .
Moreover, the interaction between Zr and Ce was determined by the XPS analysis. It is found the Zr 3d 5/2 peak of ZrO 2 -CeO 2 /C ( Fig. 2c) shifts negatively about 0.1 eV compared to that of ZrO 2 /C (Fig. 2a), and the Ce 3d 5/2 peak of ZrO 2 -CeO 2 /C (Fig. 2d) shifts negatively about 0.1 eV compared to that of CeO 2 /C (Fig. 2b). It indicates that there was an electronic interaction between the ZrO 2 and CeO 2 , which could play an important role in catalytic reaction and the activation of both dispersed metal and oxide matrix during electrode process 37 . The electron cloud density of Pd could be changed by ZrO 2 -CeO 2 /C supporting, and a synergetic effect could be yielded to improve the catalytic activity of Pd catalyst.
The XPS spectra of Pd 3d in different catalysts are shown in Fig. 3. It shows that the peaks corresponding for Pd 3d 5/2 (~335.7 eV) and Pd 3d 3/2 (~341.0 eV) core levels can be ascribed to Pd and PdO y (0 < y < 2) species, respectively 38,39 . It is obviously that both Pd metal and PdO y can be observed on the surface of prepared catalysts. The ratio of Pd: PdO y is respectively calculated to be 1.2, 1.4, 2.0 and 2.3 for Pd/C, Pd/CeO 2 /C, Pd/ZrO 2 /C and Pd/ZrO 2 -CeO 2 /C. As reported in the literature 33 , CeO 2 can act as an oxygen storage reservoir, which gives lattice oxygen to Pd and stabilizes Pd in an oxidized form. Therefore, the increased metallic Pd content can be mainly related to (Zr, Ce)O 2 solid solution.
In addition, the binding energy of Pd 3d 5/2 shifts positively from 335.65 eV for Pd/C to 335.70 eV for Pd/ CeO 2 /C, and to 335.74 eV for Pd/ZrO 2 -CeO 2 /C. Combining with the XPS results in Zr 3d, Ce 3d and Pd 3d (in Figs 2 and 3), it can conclude that strong interaction exists between Pd and oxides, especially for the addition of (Zr, Ce)O 2 solid solution. Moreover, the atom ratio of Pd: Zr: Ce is measured about 6.49: 0.98: 1.02 for Pd/ ZrO 2 -CeO 2 /C catalyst, while the values are about 6.51: 1.92: 0 and 6.45: 0: 2.01 for Pd/ZrO 2 /C and Pd/CeO 2 /C respectively, indicating that Pd has been sucessfully loaded in the oxide support and the atom ratio of Zr: Ce is nearly 1: 1 for ZrO 2 -CeO 2 solid solution.
Electrochemical activity and stability. Figure 4 shows the cyclic voltammograms of different catalysts in 1 M KOH solution. It is known that the position and size of anodic peak often change with the modification of Pd catalyst. In general, the anodic peak in the potential range between −0.9 and −0.7 V corresponds for the oxidation of the absorbed and adsorbed hydrogen, the peak from −0.6 to −0.4 V is for OH − adsorption, and the peak at about −0.2 V can be ascribed to Pd oxide formation, respectively. In addition, the cathodic peak ranging from 0.0 to −0.4 V is due to the reduction of Pd oxide, and the peak between -0.6 to −0.8 V relates to the hydrogen adsorption/absorption 40,41 . The reduction peak current density of PdO is 82 mA mg −1 for Pd/ZrO 2 -CeO 2 /C composite catalyst, which is significantly larger than other catalysts. This indicates the highest electrochemical activity of Pd/ZrO 2 -CeO 2 /C composite catalyst in present work. Figure 5 shows the cyclic voltammograms of different catalysts in 1 M KOH solution containing 1 M EG. There are two anodic peaks observed in the forward and reverse scans. The oxidation current increases with the potential increasing for EG oxidation in the forward scan, then decreases due to the reduced activity of Pd and CO poisoning 43 . The anodic peak appeared during the reverse scan could correspond to re-oxidation process of the intermediate product of EG 17,42 . It is noticed that the current density (i f ) of forward anodic peak of Pd/ZrO 2 -CeO 2 /C (3700 mA mg −1 ) is much greater than that of Pd/C (1695 mA mg −1 ) and Pd/CeO 2 /C (2448 mA mg −1 ). This implies that the Pd/ZrO 2 -CeO 2 /C composite catalyst has highest catalytic activity for EG electro-oxidation. It has been reported that the positive shifts of Pd 3d 5/2 peak of Pd-based catalysts indicates the decrease in the 3d electron density of Pd. The Pd with low 3d electron density is not easy to bind with the intermediate, and thus the intermediate coverage on Pd surface can be reduced 43 . Therefore, the highest catalytic activity of Pd/ZrO 2 -CeO 2 /C composite catalyst in present work can be attributed to the interaction between Pd and (Zr, Ce)O 2 solid solution (in Figs. 2 and 3) as well as the smallest particle size of Pd (Fig. 1).
Moreover, the chronoamperometry curves of different catalysts were collected in 1 M KOH containing 1 M EG as shown in Fig. 6. It is clear that the Pd/ZrO 2 -CeO 2 /C catalyst has a highest initial anodic oxidation current density because of its best electrocatalytic activity in EG. Moreover, there are rapid decays of the current densities observed for all catalysts, due to the formation of CO-like intermediates during EG oxidation 13 . Then the current decreases slowly and reaches a pseudo-steady state after 4000 s. It is worth noting that the Pd/ZrO 2 -CeO 2 /C composite catalyst exhibits a highest steady state current density during the EG oxidation measurement, indicating its best stability among all catalysts in present work. The current density of Pd/ZrO 2 -CeO 2 /C catalyst is about 1.7 times as that of Pd/CeO 2 /C catalyst (295.5 vs. 177.8 mA. mg −1 ) after 5000 s.
It is known that CO species are the main poisoning intermediate during the electro-oxidation process, thus a good catalyst for EG electro-oxidation should possess excellent CO electro-oxidizing ability. Figure 7 shows the CO-stripping voltammograms of different catalysts in 1 M KOH. It shows that the CO-striping onset potential of the Pd/ZrO 2 -CeO 2 /C composite catalyst is similar with all the catalysts studied, and only the peak areas are different, which is related to the electrochemical surface area.
Therefore, the electrochemical active surface area (EAS) is calculated by following equation 44,45 = EAS Q/mC (1) where Q is the charge for CO desorption electro-oxidation, m is the amount of Pd loaded, and C (420 μC cm −2 ) is the charge needed for the adsorption of a CO monolayer. It is obvious that the EAS of the Pd/ZrO 2 -CeO 2 /C catalyst is larger than that of Pd/C (117.9 vs. 101.1 m 2 g −1 ), which further supports the promoted CO-stripping  To clarify the catalytic effect of ZrO 2 -CeO 2 , the specific activity of catalysts has been calculated by normalizing the cyclic voltammograms results (in Fig. 5) to EAS result. It is also worth noting that the specific activity obviously increases from 1.68 mA cm −2 for Pd/C to 3.14 mA cm −2 for Pd/ZrO 2 -CeO 2 /C catalyst. Therefore, the inclusion of (Zr, Ce)O 2 solid solution enhances the stability as well as activity of Pd/C catalyst, which could be related to the strong interaction between Pd and (Zr, Ce)O 2 solid solution and the weaker adsorption strength between the Pd surface and CO as discussed for XPS result in Fig. 3.
The TEM images and particle size distribution of different catalysts after 900 cycles in the 1 M KOH solution containing 1 M EG are displayed in Fig. 8. It shows that the average diameter of Pd nanoparticles after 900 cycles   increases to 6.5, 5.5, 4.9 and 4.5 nm for Pd/C, Pd/ZrO 2 /C, Pd/CeO 2 /C and Pd/ZrO 2 -CeO 2 /C, respectively. The smaller increasement in diameter of Pd particles for Pd/ZrO 2 -CeO 2 /C has been obtained compared with that of catalysts before test as shown in Fig. 1, which indicates its better stability in the alkaline solution. This result can be attributed to protective decoration of Pd/C by (Zr, Ce)O 2 solid solution which can prevent Pd from dissolution and agglomeration during cycling 46 . It is also observed that the d-spacing of 0.227 which belongs to the (111) plane of cubic Pd, from the HRTEM images of all catalysts in Fig. 9e-h, and the corresponding d-spacing of ZrO 2 , CeO 2 and ZrO 2 -CeO 2 have been measured and confirmed, which is similar as the TEM analyses of the catalysts before cycling in Fig. 1.

Discussions
It is well established that the addition of oxides into Pd-based catalysts increases alcohol oxidation and the removal of adsorbed CO on the surface of catalysts. The higher catalytic activity of Pd/ZrO 2 -CeO 2 /C compared with that of Pd/CeO 2 /C in present work could be related to the addition of ZrO 2 to CeO 2 . According to the bi-functional mechanism, the addition of ZrO 2 probably more easily activates H 2 O to form oxygen-containing species (OH ads ) at lower potential. Then the CO-like intermediate species could be reacted with these oxygen-containing species on the surface of Pd surface to release the active sites for further alcohol oxidation 30 . It has also been reported that the synergistic effect can be obtained if intimate mixing (solid solution) is achieved, then the electrochemical properties could be modified 47 . Therefore, the improved electrocatalytic activity and stability of Pd/ZrO 2 -CeO 2 /C composite in present work can be mainly attributed to the strong interaction between Pd and (Zr, Ce)O 2 solid solution as well as the synergistic effect between CeO 2 and ZrO 2 .
First principle calculations were performed to further unravel the physical origin of the improved electrocatalytic activity and stability of Pd/ZrO 2 -CeO 2 /C composite. Since the CeO 2 (111) surface is more stable than CeO 2 (110) surface 48 , we have proposed a catalyst model with 3 × 3 CeO 2 p(111) surface slab and a Pd 4-atom cluster, referred to as Pd 4 /CeO 2 (111) hereafter. According to our previously analysis, we have obtained (Zr, Ce)O 2 solid solution with CeO 2 structure by replacing 50% Ce atoms with Zr atoms. Herein, the special quasi-random structure (SQS) approach 49 has been introduced to model the (Zr, Ce)O 2 solid solution. Corresponding catalyst model refers to as Pd 4 /ZrO 2 -CeO 2 (111) hereafter. Figure 9 illustrates the relaxed structures of Pd 4 /CeO 2 (111) and Pd 4 /ZrO 2 -CeO 2 (111) catalyst models. Very interestingly, the Pd cluster behaviors is very different on top of the CeO 2 and (Zr, Ce)O 2 solid solution surface. As can be seen in Fig. 9a, for the Pd 4 /CeO 2 (111) catalyst model, the Pd atoms spread on top of the CeO 2 surface, forming a 2D plane structure. On the other hand, the four Pd atoms gather into a tilted tetrahedron on top of the surface of (Zr, Ce)O 2 solid solution, forming a 3D stereoscopic cluster in Fig. 9b. The corresponding charge transfer between the Pd 4 cluster and oxide surface ρ chargetransfer illustrated in Fig. 8c and d are analyzed by:  Fig. 9c is stronger then on the ZrO 2 -CeO 2 (111) surface, indicating that the chemical environment on the ZrO 2 -CeO 2 (111) surface is more homogeneous. We performed the Bader charge 50 analysis to quantify the transfer of charge density. The results show 0.592 e and 0.427 e net charge transferred from the Pd 4 cluster to the CeO 2 (111) and ZrO 2 -CeO 2 (111) surface, respectively, indicating the 3D stereoscopic cluster is more stable than the 2D plane structure. We assumed such structure difference leads to different anti-CO-poisoning properties. The required driving forces of CO desorption are evaluated by the following equation:    Fig. 7).
To further exploring the anti-CO-poisoning properties, the charge transfers from absorbed CO molecule to Pd cluster was studied by analyzing the charge density differences ρ diff , which is defined as:  (111) catalyst. The violet isosurfaces present the charge depletion ρ diff < 0, while the cyan isosurfaces show the charge accumulation ρ diff > 0. As seen in Fig. 10a, the absorption of CO on top of the Pd 4 /CeO 2 (111) catalyst will lead to charge depletion in the middle of the CO molecule and the Pd 2D flat cluster, which contributes to strong Pd-C bonding. Meanwhile, for the case of CO absorbed on the Pd 4 /ZrO 2 -CeO 2 (111) catalyst as shown in Fig. 10b, the charge accumulation between the CO molecule and the Pd 3D stereoscopic cluster increases the repulsive force between the C and Pd atom, resulting in a weaker Pd-C bonding. Combining with our previously analysis, the 3D stereoscopic Pd cluster on (Zr, Ce)O 2 solid solution contributes to the good anti-CO-poisoning properties of Pd 4 /ZrO 2 -CeO 2 (111) catalyst.

Conclusions
In present work, the effect of (Zr, Ce)O 2 solid solution on the structures and properties of Pd-based catalysts has been clearly demonstrated. The inclusion of (Zr, Ce)O 2 solid solution reduces the particle size and increases the EAS of Pd nanoparticles, resulting in enhanced electrocatalytic activity of the composite catalyst for EG electro-oxidation in KOH solution. In addition, the strong interaction between Pd and (Zr, Ce)O 2 solution decreases the Pd 3d electron density, which reduces the intermediate coverage on Pd surface and releases more active sites for EG electro-oxidation. The synergistic effect in (Zr, Ce)O 2 also significantly increases the concentration of the OH ads species on Pd surface and thus improves the stability and anti-CO-poisoning of Pd-based composite catalysts. Furthermore, the 3D stereoscopic Pd cluster on the surface of (Zr, Ce)O 2 solid solution were explored based on first principle calculations, which leads to weaker Pd-C bonding and smaller CO desorption driving force. The findings on the good electrocatalytic properties of Pd/ZrO 2 -CeO 2 /C composite catalyst and the related mechanism will promote the development of stable catalyst for DAFCs application.

Methods
Sample preparation. 26  The catalysts were prepared through a borohydride reduction approach 43,51 . 20 mg of PdCl 2 (Shanghai Jiuyue Chemical Reagent Co., Ltd, AR) was sonicated in ultrapure water for 2 h. The prepared CeO 2 /C was sonicated in ultrapure water for 1 h. Two solutions were then mixed and stirred for 1 h, to obtain a solution with a nominal Pd-loading of 20 wt. %. The pH of this solution was adjusted to 9 using 1 M NaOH. NaBH 4 was then added to the solution as a reduction agent, with a molar ratio of 1:8 for Pd and NaBH 4 . The resulting precipitate was filtered and washed with ultrapure water and ethanol, and then dried at 60 °C. The catalysts are designated as Pd/ZrO 2 /C, Pd/ZrO 2 -CeO 2 /C and Pd/CeO 2 /C, respectively. The Pd/C (Pd:C = 20:80 wt.%) was also prepared for comparison.

Materials characterizations.
The high resolution transmission electron microscopy (HRTEM) measurement was conducted in an Electron Microscope system (Tecnai G2 F20 S-TWIN) at 200 kV. Samples were prepared by transferring the catalyst suspension to a copper grid. The catalysts after 900 cycles in 1 M KOH containing 1 M EG were also subjected to TEM measurement.
The electronic structures of Zr 3d and Ce 3d, in different species before and after Pd-loading, were investigated by X-ray photoelectron spectroscopy (ESCALAB 250, Thermo Scientific, Inc.) with a monochromatic Al Kα source (10 mA, 15 kV), respectively. The electronic structure of Pd 3d in different catalysts was also measured by XPS. The spectra were fitted by the Gaussian-Lorentzian method, with a background subtraction by Shirley's method 52 .
Electrochemical measurements. 3 mg of catalysts was added in 0.6 mL of Nafion solution (0.5 wt. %) and stirred for 30 min. 4 μL resulting ink was then transferred onto the electrode surface (glassy carbon disk, 3 mm in diameter), and dried at 70 °C for 10 min.
The electrochemical properties of catalysts were measured using an electrochemical workstation (CHI660D, Chenhua Inc., Shanghai, China) in a three-electrode system. The Pt plate (1 cm 2 ) was served as the counter electrode, a Hg/HgO electrode for the reference electrode, and the glassy carbon disk electrode for the working electrode. All measurements were conducted in the water bath of 25 ± 1 °C. The measured Pd loading in the catalysts was 0.057 mg cm −2 , by an inductively coupled plasma equipped with with atomic emission spectroscopy (ICP-AES).
The chronoamperometry was performed at a potential of −0.25 V (1 M KOH containing 1 M EG). In the electrochemical CO-stripping measurements, CO was bubbled into the solution for 15 min with a fixed the catalyst potential (0 V vs. Hg/HgO). The residual CO in the solution was removed by N 2 (99.9%). The cyclic voltammograms (CVs) were conducted in the potential ranging from −0.9 V to 0.4 V vs. Hg/HgO with a sweep rate of 50 mV s −1 (1 M KOH and 1 M KOH containing 1 M EG).