Surface-modulated palladium-nickel icosahedra as high-performance non-platinum oxygen reduction electrocatalysts.

The search for high-performance non-platinum (Pt) electrocatalysts is the most challenging issue for fuel cell technology. Creating bimetallic non-Pt nanocrystals (NCs) with core/shell structures or alloy features has widely been explored as the most effective way for enhancing their electrochemical properties but still suffered from undesirable performance due to the limited interactions between the different components. By addressing the above issue, we report on a new class of active and stable bimetallic non-Pt electrocatalysts with palladium (Pd) icosahedra as the core and nickel (Ni) decorating the surface toward cathodic oxygen reduction reaction (ORR) under alkaline conditions. The optimized Pd6Ni icosahedra with unique interaction between an icosahedral Pd core and surface Ni yield the highest ORR activity with a mass activity of 0.22 A mgPd-1, which is better than those of the conventional Pd6Ni icosahedra with alloy surfaces or Pd-rich surfaces, and even two times higher than that of the commercial Pt/C (0.11 A mgPt-1), representing one of the best non-Pt electrocatalysts. Simulations reveal that the Pd icosahedra decorated with Ni atoms emerged in the subsurface can weaken the interaction between the adsorbed oxygen and Pd (111) facet and enhance the ORR activities due to an obvious shift of d-band center. More significantly, under electrochemical accelerated durability test, the Pd6Ni icosahedra can endure at least 10,000 cycles with negligible activity decay and structural change. The present work demonstrates an important advance in surface tuning of bimetallic NCs as high-performance non-Pt catalysts for catalysis, energy conversion, and beyond.

INTRODUCTION

Fuel cells that generate clean, sustainable, and efficient energy have been identified as ideal electrochemical energy conversion devices (, ). n class="Chemical">Oxygen reduction reaction (ORR), as the half reaction occurs at the cathode, has an intimate relationship with the development of fuel cell technologies, which highly requires effective electrocatalysts (, ). Among all studies, platinum (Pt) is the most efficient and indispensable element for ORR electrocatalysts (–). However, despite devoted efforts to commercialize Pt, its high cost and limited abundance have largely precluded the practical large-scale commercialization of fuel cells (–). Hence, developing a new alternative with largely reduced cost and competitive performance becomes mandatory.

Recently, the development of palladium (n class="Chemical">Pd)–based nanocrystals (NCs) provides a promising strategy because Pd is less costly and more abundant than Pt and exhibits great potential in catalyzing the reduction of oxygen, although its electrocatalytic activity and durability are inferior to Pt and thus there is still great room for improvement (, ). For the Pd-based NCs (–), their catalytic performances can be improved by the introduction of other elements (such as Co, Ni, etc.), which is likely caused by the modification of the electronic structure of Pd (–). One general way to achieve this goal is to alloy Pd with other metals, whereas the conventional Pd-based alloy only exhibits limited activity improvement (, ). Another common method is to build the core-shell construction by covering Pd as a shell on the proper non-noble metal core, which not only reduces the use of Pd but also enhances its catalytic properties (, ). However, the above core-shell structure with pure Pd surface inevitably suffers from gradual activity decay under harsh fuel cell operation conditions (). Fortunately, recent research has demonstrated that the surface modification of NCs is of great significance in ORR, since a balance between O2 absorption/desorption can be achieved via changing the electronic structure of a core metal (, , ). However, the surface design of NCs should counterbalance two completely opposing effects; they are the relatively strong absorption energy of O2 and reaction intermediates, as well as a relatively low coverage by unexpected oxygenated species and specifically absorbed anions (, , ). To be specific, insufficient amounts of new metals modified on the Pd NC surface might be ineffective in changing the electronic structure of Pd. Conversely, excessive introduction of new metals can block the active Pd sites. To reach an ideal equilibrium state, making a rational design of NCs with precise surface tuning is highly desirable but remains a formidable challenge.

Herein, we present a facile method to synthesize a new class of bimetallic n class="Chemical">Pd-Ni icosahedra with tailored Ni on the surface of Pd, including Pd3Ni, Pd4Ni, Pd6Ni, and Pd8Ni icosahedra. After annealing at 400°C and 500°C under H2 atmosphere, the Pd6Ni icosahedra with Pd-Ni alloy surfaces and the Pd6Ni icosahedra with Pd-rich surfaces were also obtained, respectively. The ORR activities of Pd-Ni icosahedra under alkaline conditions show a volcano curve as a function of the Ni content, in which the Pd6Ni icosahedra yield the highest activity and durability under the alkaline condition with a mass activity of 0.22 A mgPd−1. This is two times higher than that of the commercial Pt/C (0.11 A mgPt−1), representing one of the best non-Pt electrocatalysts toward alkaline ORR. More significantly, the unique interaction between the surface Ni and Pd core of the Pd-Ni icosahedra plays a vital role on the excellent ORR activity, making their catalytic performance even better than those of the Pd-Ni icosahedra with Pd-Ni alloy surface and the Pd-Ni icosahedra with Pd-rich surface. The density functional theory (DFT) calculations confirmed a weak interaction between the adsorbed oxygen and Pd (111) facet and an obvious shift of the d-band center of Pd-Ni icosahedra due to the subsurface decoration of Ni atoms, making the optimized Pd-Ni icosahedra exhibit best ORR performance. The insights from the present work open significant opportunities for the design of high-performance non-Pt ORR catalysts by taking the advantages of surface modification.

RESULTS AND DISCUSSION

In a typical preparation of Pd6Ni icosahedra, n class="Chemical">Pd(acac)2 and Ni(HCO2)2·2H2O were used as precursors, ascorbic acid (AA) was selected as a reducing agent, and oleylamine (OAm) and 1-octadecene (ODE) were applied as solvents and stabilizers. All these chemicals were added into a glass vial (volume, 35 ml) and followed by 30-min ultrasonication. The resulting homogeneous mixture was then heated from room temperature to 160°C for around 0.5 hours and kept at this temperature for 5 hours in an oil bath. After cooling to room temperature, the colloidal products were collected by consecutive washing/centrifugation cycles.

The representative transmission electron microscopy (TEM) image (Fig. 1A) shows that the n class="Chemical">Pd6Ni icosahedra are fairly uniform in terms of size and shape. The size distribution shows that Pd6Ni icosahedra are 16 ± 0.8 nm from one apex to the opposite apex (fig. S1). The twin structures were observed from the contrast in the TEM image. The ratio of Pd to Ni was determined to be Pd/Ni = 85.8:14.2 by the energy-dispersive x-ray spectroscopy (EDX) (Fig. 1B), which is inconsistent with the inductively coupled plasma atomic emission spectroscopy (ICP-AES; Pd/Ni = 85.2:14.8). Powder x-ray diffraction (PXRD) was conducted to determine the phase of Pd6Ni icosahedra, where the icosahedra mainly display face-centered cubic (fcc) Pd with weak peaks assigned to the fcc (111) facet of Ni (Fig. 1C). The crystalline nature of Pd6Ni icosahedra was further confirmed by high-resolution TEM (HRTEM) (Fig. 1D). An interplanar spacing of 0.22 nm was observed, assigned to the (111) plane of Pd. Figure 1E, which displays the fast Fourier transform (FFT) image of the icosahedron nanoparticle in Fig. 1D, illustrates the lattice lines going outward radically from the center with a sixfold symmetry, as indicated by red circles. The three typical projections collectively confirmed the formation of well-defined Pd6Ni icosahedra (fig. S2). The high-angle annular dark-field scanning TEM (HAADF-STEM) image and the corresponding elemental mappings demonstrate that Pd distributes in the interior of the icosahedron, while Ni mainly locates on the outmost surface of Pd (Fig. 1, F and G). The HAADF-STEM line-scan analysis shows that the Ni is mainly on the surface of the Pd core (Fig. 1H).

Fig. 1

Morphological and structural characterization of Pd6Ni icosahedra.

(A) TEM image, (B) EDX, and (C) PXRD pattern of Pd6Ni icosahedra. (D) HRTEM (the threefold symmetry orientation was outlined by three yellow dotted lines) and (E) corresponding FFT image of Pd6Ni icosahedra (red circles indicated the diffraction spots). (F) HAADF-STEM image, (G) corresponding elemental mappings, and (H) line-scan analysis across the blue arrow in the inset of (F).

By simply changing the reaction time with other conditions same as the synthesis of n class="Chemical">Pd6Ni icosahedra, the Pd8Ni, Pd4Ni, and Pd3Ni icosahedra were readily obtained. As shown in Fig. 2 (A, D, and G), all these NCs have uniform sizes and shapes. We also carried out detailed characterizations to reveal the crystalline nature and elemental distributions of these NCs. The EDX results show that the atomic ratios of Pd/Ni in Pd3Ni, Pd4Ni, and Pd8Ni icosahedra are 74.7/25.3, 79.8/20.2, and 89.1/10.8, respectively, which are in line with the ICP-AES results (fig. S3). The PXRD results show that all the NCs that are mainly in Pd phase with the shoulder peaks at around 44.6° become more and more distinct as Ni content increases, which assigned to the (111) plane of Ni (fig. S4). The crystalline nature of these NCs was further characterized by HRTEM (fig. S5). The lattice fringes of these NCs are about 0.22 nm and assigned to the (111) plane of Pd. The spatial distributions of Pd and Ni species in these NCs were further identified by HAADF-STEM and corresponding elemental mappings, all of which showed a similar feature with Pd6Ni icosahedron, where Pd concentrates in the core of icosahedron and Ni mainly modifies on the outmost surface. Meanwhile, it is also found that the Ni signal becomes weak as the Ni content decreased (Fig. 2, B, E, and H). The line-scan analysis offers more intuitive results to evaluate the Ni distribution. As shown in Fig. 2 (C, F, and I), as surface Ni content decreases, the Ni distribution of Pd3Ni, Pd4Ni, and Pd8Ni icosahedra is lessened. These results indicate the successful modification of different contents of Ni on the surface of Pd by simply controlling the reaction time. To obtain the surface properties of PdNi icosahedra, we carried out x-ray photoelectron spectroscopy (XPS) studies (fig. S6), where both the Pd and Ni are mainly in the presence of metallic states in those PdNi icosahedra.

Fig. 2

Morphological and structural characterizations of Pd3Ni, Pd4Ni, and Pd8Ni icosahedra.

(A, D, and G) TEM images, (B, E, and H) HAADF-STEM images and corresponding elemental mappings, and (C, F, and I) line-scan analysis of (A to C) Pd3Ni icosahedra, (D to F) Pd4Ni icosahedra, and (G to I) Pd8Ni icosahedra, respectively.

(A, D, and G) TEM images, (B, E, and H) HAADF-STEM images and corresponding elemental mappings, and (C, F, and I) line-scan analysis of (A to C) n class="Chemical">Pd3Ni icosahedra, (D to F) Pd4Ni icosahedra, and (G to I) Pd8Ni icosahedra, respectively.

To achieve a better control over n class="Chemical">Pd-Ni icosahedra, we optimized a variety of synthetic parameters. We found that proper precursors and reducing agent are essential for the preparation of well-defined Pd-Ni icosahedra (figs. S7 to S9). This synthetic method in the absence of Ni(HCO2)2·2H2O yielded aggregated Pd NCs (fig. S7, A and B), and only irregular NCs were observed when Ni(HCO2)2·2H2O was replaced by NiAc2 and Ni(acac)2 (fig. S9), suggesting that the use of Ni(HCO2)2·2H2O is essential for the formation of Pd-Ni icosahedra. To further reveal the growth mechanism of Pd-Ni icosahedra, we carried out time-tracking experiments and also performed characterizations for these intermediate products. As shown in fig. S10A, at the initial stage, only irregular NCs with a size of around 12 nm can be observed. The EDX result reveals that Pd is dominant in these NCs (fig. S10E). When the reaction time proceeded to 1 hour, we observed similar irregular morphology, and the ratio of Pd to Ni was 95.9:4.1 (fig. S10, B and E). After the reaction proceeded to 2 hours, the content of Ni was continuously increased, but the NCs were still irregular (fig. S10, C and E). When the time extended to 3 hours, we observed relatively regular NCs, and the ratio of Pd to Ni increased to 93.1:6.9 (fig. S10, D and E). It is noteworthy that Pd was always the dominant phase in these products from the PXRD results, likely due to the low content of Ni in the intermediates (fig. S10F). In this regard, the Ni content in the products is not in proportion to the reaction time although the reaction condition is not significantly changed. Therefore, it can be concluded that Pd initially formed as seeds, followed by the reduction and growth of Ni onto the Pd surface.

Pd-based n class="Chemical">NCs are a class of active electrocatalysts for ORR under alkaline conditions (, , ). To evaluate the ORR performance of the Pd-Ni icosahedra, we loaded the Pd-Ni icosahedra on commercial carbon black by sonication and washed with acetic acid two times (fig. S11). The typical cyclic voltammogram (CV) curves for Pd-Ni/C and the commercial Pd/C (fig. S12) in 1 M KOH solution at a sweep rate of 100 mV s−1 are shown in fig. S13A. The 1.38 V peak is attributed to the oxidation of Ni (). As expected, the current intensity of oxidation peak of Ni considerably increased with the Ni increment of these catalysts. The electrochemically active surface area (ECSA), as an important parameter for assessing the active sites of catalysts, was calculated from the charge required for oxygen desorption (Fig. 3A) (, ). Herein, the ECSA values of Pd3Ni/C, Pd4Ni/C, Pd6Ni/C, Pd8Ni/C, and the commercial Pd/C are 31.3, 35.8, 41.5, 47.5, and 46.9 m2 gPd−1, respectively. As a comparison, the ECSA of the commercial Pt/C was calculated to be 62.5 m2gPt−1, based on the CV (fig. S13B).

Fig. 3

ORR performance of Pd-Ni/C, commercial Pd/C, and commercial Pt/C.

(A) CV curves of Pd-Ni/C and the commercial Pd/C recorded in 1 M KOH solution at a scan rate of 100 mV s−1. (B) ORR polarization curves of Pd-Ni/C and the commercial Pd/C recorded in O2-saturated 0.1 M KOH solution at a scan rate of 10 mV s−1 and a rotation rate of 1600 rpm. (C) Comparison of specific activities at 0.9 V versus RHE for these catalysts. (D) Comparison of mass activities at 0.9 V versus RHE for these catalysts and the commercial Pt/C. The activities were calculated on the basis of five independent measurements.

The ORR polarization curves of Pd-Ni/C and the n class="Chemical">commercial Pd/C are shown in Fig. 3B. Moreover, the specific activities and mass activities determined from Fig. 3B at 0.9 V versus reversible hydrogen electrode (RHE) for Pd-Ni/C were shown in Fig. 3 (C and D). The Pd6Ni/C (0.22 A mgPd−1 and 0.66 mA cmPd−2) exhibits the best ORR activity as compared with Pd3Ni/C (0.10 A mgPd−1 and 0.40 mA cmPd−2), Pd4Ni/C (0.16 A mgPd−1 and 0.56 mA cmPd−2), and Pd8Ni/C (0.09 A mgPd−1 and 0.24 mA cmPd−2), showing a volcano-type dependence on the surface Ni content. Significantly, it is clear that the mass activity of Pd6Ni/C is even two times higher than that of the commercial Pt/C (0.11 A mgPd−1) (Fig. 3D), showing that the Pd6Ni/C is a promising non-Pt electrocatalyst toward alkaline ORR (tables S1 and S2).

Considering that n class="Chemical">Pd-Ni/C have similar morphology, size, and surface structure, it is reasonably suggested that the highest mass activity of the Pd6Ni/C should be attributed to the suitable surface Ni decoration on the icosahedral Pd core, which is crucial to promote ORR activity and might markedly alter the surface structure of NCs compared with the conventional alloy surface. To confirm our assumption, we also prepared Pd6Ni icosahedra with alloy surface and Pd6Ni icosahedra with Pd-rich surface by annealing the synthesized Pd6Ni icosahedra at 400° and 500°C under H2 atmosphere for 4 hours. These are referred to as Pd6Ni/C-400°C and Pd6Ni/C-500°C, respectively (fig. S14). The specific surface structures of obtained products were characterized by elemental mappings and line-scan analysis. The HAADF-STEM images show the similar morphology of Pd6Ni/C-400°C and Pd6Ni/C-500°C with the size of around 17 nm, which is similar to that of Pd6Ni icosahedra (Fig. 4, A and D). The Pd6Ni/C-400°C exhibits homogeneous distributions of Pd and Ni in the whole NC with reduced Ni signals on the NC surface (Fig. 4B). This structure was further testified by line-scan analysis, where the signals of Pd and Ni nearly overlap with each other on the surface of the NC (Fig. 4C), suggesting the formation of a Pd-Ni alloy surface under this thermal annealing condition. This phenomenon is mainly caused by the different enhancements of the diffusion rates of Pd and Ni atoms by the thermal annealing under H2 atmosphere (, ). To this end, the Pd-Ni icosahedra with Pd-rich surface could be created as long as the annealing temperature increases to 500°C (Fig. 4D). The elemental mappings demonstrate that the Pd-Ni alloy core is surrounded by a Pd-rich shell (Fig. 4E). By comparing line profiles of Pd and Ni with that across Pd6Ni/C-500°C, we can confirm the successful formation of the Pd-rich shell after annealing at 500°C under H2 (Fig. 4F).

Fig. 4

Structural characterizations and ORR performance of Pd6Ni/C-400°C and Pd6Ni/C-500°C.

(A and D) HAADF-STEM images, (B and E) corresponding elemental mappings from the area outlined by the red dotted line in (A) and (D), respectively, and (C and F) line-scan analysis across the blue arrow in the inset of (A) and (D), respectively, of (A to C) Pd6Ni-400°C and (D to F) Pd6Ni-500°C. (G) ORR polarization curves of Pd6Ni/C, Pd6Ni/C-400°C, and Pd6Ni/C-500°C recorded in 0.1 M KOH solution at a scan rate of 10 mV s−1. The activities were calculated on the basis of five independent measurements. (H) Electron transfer number of Pd6Ni icosahedra, Pd6Ni/C-400°C, Pd6Ni/C-500°C, and the commercial Pt/C.

(A and D) HAADF-STEM images, (B and E) corresponding elemental mappings from the area outlined by the red dotted line in (A) and (D), respectively, and (C and F) line-scan analysis across the blue arrow in the inset of (A) and (D), respectively, of (A to C) n class="Chemical">Pd6Ni-400°C and (D to F) Pd6Ni-500°C. (G) ORR polarization curves of Pd6Ni/C, Pd6Ni/C-400°C, and Pd6Ni/C-500°C recorded in 0.1 M KOH solution at a scan rate of 10 mV s−1. The activities were calculated on the basis of five independent measurements. (H) Electron transfer number of Pd6Ni icosahedra, Pd6Ni/C-400°C, Pd6Ni/C-500°C, and the commercial Pt/C.

The ORR tests were also carried out under the same condition for the n class="Chemical">Pd6Ni/C-400°C with alloy surface and Pd6Ni/C-500°C with Pd-rich surface. The ORR polarization curves of Pd6Ni/C-400°C and Pd6Ni/C-500°C are shown in Fig. 4G. The mass activities of Pd6Ni/C-400°C and Pd6Ni/C-500°C are 0.07 and 0.05 A mgPd−1, respectively. It is clear that both the Pd6Ni/C-400°C and Pd6Ni/C-500°C give significant activity reduction, as compared with Pd6Ni/C. These results demonstrate that the optimized Ni content modified on the Pd surface is more advantageous than the Pd-Ni alloy surfaces and Pd-rich surface for promoting the ORR activity. To verify the ORR behaviors of these NCs, electrocatalytic dynamic experiments were further performed by measuring a group of ORR polarization plots at different rotating speeds, from which the Koutecky-Levich (K-L) plots at different potentials could be obtained (fig. S15). On the basis of the slopes of the K-L plots, the electron transfer numbers (n) involved in the ORR process for these catalysts were calculated as ~4 at 0.50 V (Fig. 4H), indicating that the ORR mechanism for those catalysts follows the direct “4e−” pathway (O2 + 2H2O + 4e− = 4OH−) ().

In addition to catalytic activity, long-term durability is another important parameter that determines the practical application of an ORR electrocatalyst. Durability tests of Pd6Ni/C, n class="Chemical">Pd6Ni/C-400°C, Pd6Ni/C-500°C, and commercial Pt/C were evaluated by cycling the potential between 0.4 and 1.0 V versus RHE in O2-saturated 0.1 M KOH with a scan rate of 100 mV s−1 (Fig. 5). As illustrated by Fig. 5, after 10,000 continuous cycles, the commercial Pt/C exhibited 49.6% loss of its initial mass activity mainly due to the agglomeration of Pt nanoparticles during the potential cycling (Fig. 5A and fig. S16). In contrast, the Pd6Ni/C showed excellent cycling stability, as there was only a slight activity change after the long-term potential cycling. After the durability test, the Ni-riched surface structure and composition of the Pd6Ni/C are almost unchanged (Fig. 5B and figs. S17 and S18). In addition, both the chemical states of Pd and Ni of the Pd6Ni/C are largely maintained with a slight increase of oxidized Pd and oxidized Ni compared with the prepared Pd6Ni/C (fig. S19). However, the Pd6Ni/C-400°C and Pd6Ni/C-500°C showed more mass activity losses and agglomerations than those of the Pd6Ni/C (Fig. 5, C and D, and fig. S17). These results highlight that the Pd6Ni/C not only has excellent ORR activity but it also has superior stability among these catalysts.

Fig. 5

ORR performance of Pd6Ni/C, Pd6Ni/C-400°C, Pd6Ni/C-500°C, and the commercial Pt/C.

ORR polarization curves of (A) the commercial Pt/C, (B) Pd6Ni/C, (C) Pd6Ni/C-400°C, and (D) Pd6Ni/C-500°C before and after 10,000 cycles. The insets show corresponding mass activities at 0.9 V versus RHE before and after 10,000 cycles.

ORR polarization curves of (A) the commercial n class="Chemical">Pt/C, (B) Pd6Ni/C, (C) Pd6Ni/C-400°C, and (D) Pd6Ni/C-500°C before and after 10,000 cycles. The insets show corresponding mass activities at 0.9 V versus RHE before and after 10,000 cycles.

Previous reports have demonstrated that Pt-M alloy n class="Chemical">NCs (M = Fe, Co, and Ni) are beneficial to improve ORR performance by introducing another transition metal, which makes the d-band center of Pt downshift, weakens the adsorption of oxygen intermediates, and releases additional active sites of Pt (, , ). In our work, by the same token, the valence electron density of Pd can be markedly changed by desirable Ni modified on the surface of Pd. To understand the excellent ORR performance of PdNi icosahedra, we use DFT simulations to analyze the different surface features of PdNi icosahedra. A four-layer Pd (111) slab model decorated with different Ni ratios was constructed to simulate the PdNi icosahedra. It is found that the Ni atom tends to appear in the second and third layers with a lower total energy (0.2 to 0.3 eV) than that in the first layer. The adsorption energies of oxygen on Pd8Ni (−0.95 eV), Pd7.47Ni (−1.04 eV), Pd7.1Ni (−1.03 eV), and Pd6.2Ni (−1.05 eV) are weaker than that in Pd (111) (−1.31 eV), Pd3.36Ni (−1.65 eV), and Pd3Ni (−1.72 eV) (Fig. 6A). Lower negative adsorption energy of oxygen will be responsible for the improvement of ORR activity. As shown in Fig. 6B, the d-band center of Pd (111) is −2.24 eV, and it will increase to −2.16 eV for Pd8Ni, −2.14 eV for Pd7.47Ni, −2.12 for Pd7.1Ni, and −2.13 eV for Pd6.2Ni. With the change of decorated Ni atoms, the d-band center of a Pd atom connected with a Ni atom directly downshifts, reflecting a linear relationship between the Pd–O bonding and d-band center site. To verify the rationality of the model, we have further calculated the d-band center of these PdNi icosahedra on the basis of the XPS results, where the d-band center for these PdNi icosahedra increases with the increase of Ni content (fig. S20), matching well with the DFT calculation based on the model. That is the main reason why the Pd icosahedra–decorated Ni atom exhibits a better ORR activity. The Ni atom in the first layer of Pd (111) will introduce a surface-ligand effect and enhance the adsorption of oxygen. However, the lattice-strain effect from the Ni atom in the second layer will have an opposed influence on the Pd–O bonding according to the previous simulations (). Meanwhile, the Ni atom will lose electrons to adsorbed oxygen, enhancing the electron overlap between O and Pd and facilitating charge depletion.

Fig. 6

DFT simulations of the adsorption energy of adsorbed oxygen atom and the d-band center shift of PdNi (111) with different ratios of Ni.

(A) Adsorption energy of adsorbed oxygen atom with the change of Pd/Ni ratio. (B) Relationship between adsorption energy of oxygen and d-band center of Pd. The inset is the charge density difference of oxygen adsorbed on the surface of PdNi (blue and yellow isosurfaces indicate electron depletion and electron accumulation with the same isosurface values of 0.005 e/bohr3; the silver and orange balls represent Pd and Ni atoms, respectively).

(A) Adsorption energy of adsorbed n class="Chemical">oxygen atom with the change of Pd/Ni ratio. (B) Relationship between adsorption energy of oxygen and d-band center of Pd. The inset is the charge density difference of oxygen adsorbed on the surface of PdNi (blue and yellow isosurfaces indicate electron depletion and electron accumulation with the same isosurface values of 0.005 e/bohr3; the silver and orange balls represent Pd and Ni atoms, respectively).

In summary, the Pd-Ni icosahedra with n class="Chemical">controlled surface structure (that is, Pd-Ni icosahedra with Ni surface, Pd-Ni icosahedra with Pd-Ni alloy surface, and Pd-Ni icosahedra with Pd-rich surface) were created through the combination of wet-chemical synthesis and thermal annealing. Among these Pd-Ni icosahedra with various modifications of Ni on the surface, the optimized Pd6Ni/C reaches an ideal equilibrium state and exhibits the highest activity and superior stability for ORR under alkaline conditions. More significantly, the Pd6Ni/C decorated with desirable Ni displays the best ORR activity, which is much better than those of the conventional Pd6Ni icosahedra with alloy surfaces and the Pd6Ni icosahedra with Pd-rich surfaces. The theoretical investigation revealed that the enhanced ORR activities of Pd-Ni icosahedra was mainly attributed to subsurface decoration of Ni atoms, resulting in a weak interaction between the adsorbed oxygen and Pd (111) facet and an obvious shift of the d-band center. Our study provides an attractive strategy for the design of high-performance non-Pt electrocatalysts through the combined surface composition tuning and surface structure controlling.

MATERIALS AND METHODS

Chemicals

Bis(acetylacetonate)palladium(II) [n class="Chemical">Pd(acac)2, 99%], nickel(II) formate dihydrate [Ni(HCO2)2·2H2O, 99%], ascorbic acid (AA; C6H8O6; reagent grade, 99%), oleylamine [OAm; CH3(CH2)7CH═CH(CH2)7CH2NH2, >70%], 1-octadecene [ODE; CH2═CH(CH2)15CH3; technical grade, 90%], commercial Pd/C, and Nafion (5%) were all purchased from Sigma-Aldrich. Commercial Pt/C (20 weight %, 2- to 5-nm Pt nanoparticles) was obtained from Johnson Matthey Corporation. Cyclohexane (C6H12; analytical reagent, >99.5%), ethanol (CH3CH2OH; analytical reagent, >99.7%), and perchloric acid (HClO4; analytical reagent, 70 to 72%) were purchased from Sinopharm Chemical Reagent Co. Ltd. All the chemicals were used as received without further purification. The water (18 megohms/cm) used in all experiments was obtained by passing through an ultrapure purification system (Aqua Solutions).

Synthesis of of Pd-Ni icosahedra

In a typical preparation of Pd6Ni icosahedra, n class="Chemical">Pd(acac)2 (3.5 mg), Ni(HCO2)2·2H2O (4.0 mg), AA (36.0 mg), OAm (2.5 ml), and ODE (2.5 ml) were added into a vial (volume, 35 ml). After the vial had been capped, the mixture was ultrasonicated for around 0.5 hours. The resulting homogeneous mixture was then heated from room temperature to 160°C in around 0.5 hours and kept at 160°C for 5 hours in an oil bath. After cooling to room temperature, the colloidal products were collected by centrifugation and washed with cyclohexane/ethanol (v/v = 1:9) mixture three times. For the syntheses of Pd8Ni icosahedra, Pd4Ni icosahedra, and Pd3Ni icosahedra, all the conditions were similar to those of Pd6Ni icosahedra except by altering reaction time to 4, 6, and 7 hours, respectively.

Synthesis of Pd6Ni/C-400°C and Pd6Ni/C-500°C

The Pd6Ni icosahedra dispersed in 5 ml of n class="Chemical">cyclohexane and Vulcan XC-72 carbon dispersed in 15 ml of ethanol were mixed and sonicated for 1 hour to make Pd6Ni/C. The Pd6Ni/C were collected by centrifugation, washed with ethanol two times, and then dried under ambient conditions. Finally, the Pd6Ni/C were annealed at 400° and 500°C under H2 atmosphere (named as Pd6Ni/C-400°C and Pd6Ni/C-500°C, respectively) for 4 hours to prepare the catalysts with Pd-Ni alloy surface and Pd-rich surface, respectively.

Characterization

The samples were prepared by dropping cyclohexane dispersion of samples on n class="Chemical">carbon-coated copper TEM grids using pipettes and dried under ambient conditions. Low-magnification TEM was conducted on a Hitachi HT7700 transmission electron microscope at an acceleration voltage of 120 kV. High-magnification TEM, STEM, and EDX (HAADF-STEM-EDX) spectroscopy were conducted on an FEI Tecnai F20 transmission electron microscope at an acceleration voltage of 200 kV. A PXRD pattern was collected on X’Pert-Pro MPD diffractometer (PANalytical) with a Cu Kα x-ray source (λ = 1.540598 Å). Low-resolution EDX was performed on a scanning electron microscopy (SEM; Hitachi, S-4700). The concentration of catalyst was determined by ICP-AES (Varian 710-ES).

Electrochemical measurements

Before the electrochemical tests, the Pd-Ni icosahedra were collected by centrifugation and washed three times with an ethanol/cyclohexane mixture. The Pd-Ni icosahedra dispersed in 5 ml of cyclohexane and Vulcan XC-72 carbon dispersed in 10 ml of ethanol were then mixed and sonicated for 1 hour to make Pd-Ni/C. The Pd-Ni/C was collected by centrifugation and washed with acetic acid two times. The products were dried under ambient conditions. The final catalysts were redispersed in the mixture containing isopropanol and Nafion (5%) (v/v = 1:0.005) to form the homogeneous catalyst ink by sonicating for 30 min. The concentration of Pd or Pt was controlled to be 0.25 mgPt or Pd ml−1 based on ICP-AES measurement. Five microliters of the dispersion (1.25 mgcatalyst ml−1) was transferred onto the glassy carbon electrode with the loading amount of Pt or Pd at 1.25 μg.

The ORR measurements were performed by using a glassy carbon rotating disk electrn class="Chemical">ode (RDE; Pine Research Instrumentation; diameter, 5 mm; area, 0.196 cm2) connected to an installation of rotating electrode speed control (Pine Research Instrumentation; model, AFMSRCE). A leak-free saturated calomel electrode was used as the reference electrode, and a Pt wire was used as the counter electrode. The electrolyte was 0.1 M KOH solution. The loading amounts of Pd or Pt for Pd-Ni/C, Pd6Ni/C-400°C, Pd6Ni/C-500°C, and the commercial Pt/C (Johnson Matthey Corporation) catalysts were all kept at 6.4 μg cm−2. The potential scan rate was 100 mV s−1 for the CV measurements in 1 M KOH solution. The ECSA (m2·gPd−1) of the catalysts was estimated according to the equation ECSA = Q/(0.405 × WPd), where WPd represents Pd loading (mg·cm−2) on the electrode, Q is the coulombic charge by integrating the peak area of the reduction of PdO (mC), and 0.405 represents the charge required for the reduction of a PdO monolayer (mC·cm−2Pd). ORR measurements were conducted in 0.1 M KOH solutions purged with saturated O2 during the measurements. The scan rate and rotation rate for ORR measurements were 10 mV s−1 and 1600 rpm, respectively. In the ORR polarization curves, the current densities were normalized in reference to the geometric area of the glassy carbon RDE (0.196 cm2). For each catalyst, the kinetic current was normalized to the loading amount of Pt or Pd to generate mass activities. The accelerated durability tests were performed at room temperature in 0.1 M KOH solutions by applying the cyclic potential sweeps between 0.4 and 1.0 V versus RHE at a sweep rate of 100 mV s−1 for 10,000 cycles.

DFT models and calculations

The projector-augmented wave with the exchange functional of generalized gradient approximation presented by Perdew-Burke-Ernzerhof (GGA-PBE) was chosen to optimize the geometrical structures and electronic properties into the ground state until the convergence of force and energy reaches to 0.01 eV/Å and 10−4 eV, respectively (, ) Meanwhile, the empirical DFT-D3 correction method was adopted to take the weak van der Waals interaction between the adsorbed oxygen and Pd (111) into consideration (). The cutoff energy was set to 420 eV with γ-centered K-points of 4 × 4 × 1. A four-layer Pd (111) slab model was constructed, and the bottom layer was constrained to simulate the bulk during the calculations. All the calculations were done in the Vienna ab initio simulation package (, ). The adsorption energy of oxygen (E*O) was calculated by: E*O = E(M − *O) − E(M) − 1/2E(O2), where E(M − *O), E(M), and E(O2) are the total energy of Pd (111) after the adsorption of oxygen, the total energy of Pd (111) before the adsorption of oxygen, and the total energy of the oxygen molecules, respectively.