H2 Reduction Annealing Induced Phase Transition and Improvements on Redox Durability of Pt Cluster-Decorated Cu@Pd Electrocatalysts in Oxygen Reduction Reaction.

Hierarchical structures in shell with transition metal underneath is a promising design for high-performance and low-cost heterogeneous nanocatalysts (NCs). Such a design enables the optimum extent of synergetic effects in NC surface. It facilitates intermediate reaction steps and, therefore, boosts activity of NC in oxygen reduction reaction (ORR). In this study, carbon nanotube (CNT)-supported ternary metallic NC comprising Cucluster-in-Pdcluster nanocrystal and surface decoration of atomic Pt clusters (14 wt %) is synthesized by using the wet chemical reduction method with sequence and reaction time controls. By annealing in H2 environment (H2/N2 = 9:1, 10 sccm) at 600 K for 2 h, specific activity of Cu@Pd/Pt is substantially improved by ∼2.0-fold as compared to that of the pristine sample and commercial Pt catalysts. By cross-referencing results of electron microscopic, X-ray spectroscopic, and electrochemical analyses, we demonstrated that reduction annealing turns ternary NC into complex of Cu3Pt alloy and Cu x Pd1-x alloy. Such a transition preserves Pt and Pd in metallic phases, therefore improving the activity by ∼29% and the stability of NC in an accelerated degradation test (ADT) as compared to those of pristine Cu@Pd/Pt in 36 000 cycles at 0.85 V (vs RHE). This study presents robust H2 annealing for structure stabilization of NC and systematic characterizations for rationalization of the corresponding mechanisms. These results provide promising scenarios for facilitation of heterogeneous NC in ORR applications.

Introduction

A burgeoning global energy crisis by the depletion of fossil fuels stimulates researchers to find effective, low-cost, and environmentally benign energy devices. Among energy conversion assessments, high efficiency, scalability of power densities, and ease of module assembly make fuel cells one of the most promising technologies to address the aforementioned problems.[1−7] These advantages make fuel cells attractive in terms of green and high-performance sectors; however, the wide-scale commercialization of fuel cells is stagnated by the lack of efficient and cost-effective cathodic nanocatalysts (NCs). Cathodic NCs are used to trigger the sluggish oxygen reduction reaction (ORR), which incurs the highest energy barrier (∼0.3–0.4 V) among components in a fuel cell module.[8−10] To achieve considerable efficiency, omnipotent cathodic NCs are made by high contents of noble metals (mostly Pt) with low overpotential toward ORR, leading to the highest capital cost among assembled components in a state-of-the-art fuel cell module.[11,12] Apart from material cost, chemical reactivity and structural reliability are long-standing kinetics bottlenecks that come into play, dragging down the performances of Pt-based NCs in ORR. As a consequence, developing a new class of catalysts consisting of lower dosage of Pt or even alternative metals with high structural reliability and ORR activity is of paramount importance to making fuel cells in practice.

The aforementioned issues can be addressed by modifying the surface structure of NCs or changing the heteroatomic intermixing.[13] Intensive efforts have been devoted to address such indexes and to the development of less expensive and more abundant materials in the form of intraparticle configurations (alloy,[14−16] nanowires,[17] onion,[18] core–shell,[19−22] etc.). However, efficiency and stability of current cathodic NCs remain far from the commercial standards. Among existing geometric configurations, core–shell structured heterogeneous NCs possess the strongest structural stabilization and chemical activity in electrochemical reactions.[23−25] These properties are enhanced by strain and electronegativity differences at the core–shell interface. More specifically, to optimize the interface effects and noble-metal utilization, a surface active Pt-shell with a core low electronegativity transition metal underneath is commonly designed in a core–shell-type NC. In such configurations, core crystal serves as a source, which injects electrons (or forms a negative potential field toward) to the shell crystal and thus improves the reduction activity of NCs in a redox system. Transition-metal additives (e.g., Cu, Ni, Co, Zn, Ru, and Sn) with high oxygen affinity are commonly employed as the core crystal. These elements provide low-energy pathways for allocation and recombination of radicals (i.e., O*, OH*, and H*) in H2O, which reduced the standing time of reactants on NC surfaces. Meanwhile, in a heterogeneous NC, redox kinetics is dominated by both geometric (lattice strain) and electronic (ligand effect) configurations along with ensemble effect. Ensemble effect arises when dissimilar surface atoms together turn on the bifunctional mechanism (selectivity and variety of sorption sites). Owing to differences of atomic arrangement between surface active atoms (Pt) and the transition-metal core crystal underneath, lattice strain arises on the heterogeneous binary interface. It can turn compressive or expansive in the surface layer, therefore either localizing or delocalizing electrons at the heterogeneous core–shell interface to enhance electronic properties at the surface. Ligand effects are caused by the atomic vicinity of two dissimilar surface metal atoms that induces electronic charge transfer between them and thus affects their electronic band structure (i.e., charge relocation to the shell region via heteroatomic electronegativity difference). Along with scientific advantages toward ORR activity due to these configurations, heterogeneous NCs are also highly sensitive to material degradation modes of cathodic NCs (Scheme S1).

Core–shell structured NCs with proper heterojunction components and minimized shell crystal thickness seem to be a perfect design for conducting ORR in fuel cells. However, noble-metal usage is limited by formation of monolayer shell thickness in such configurations. To further improve metal utilization and enhance electron relocation potential at the heterogeneous binary interface with a lower amount of Pt usage, surface decoration of strong electronegative and active sub-nanometer Pt clusters in cluster-in-cluster structured NCs is a possible strategy.[26] Recently, researchers also developed an effective strategy to control the depth of decorating clusters.[27] Those Pt clusters extract electrons from the heterogeneous binary interface of inner clusters by strong electronegative force and local lattice during redox reaction. As a result, ORR activity of NC is substantially enhanced by a couple of orders as compared to those in common structures (i.e., alloy, cluster-in-cluster, core–shell, etc.).

Our previous works have developed easy assessments on design and synthesis of ternary NCs comprising Cu core and Pd shell decorated with highly active sub-nanometer Pt clusters on the surface by controlling growth sequences of metal crystals. With such a unique structure, redox activity and stability of NC are improved by 2–3 orders as compared to commercial Pt catalysts. In this study, we demonstrate that specific activity (SA) of such type of NCs in alkaline-based ORR can be improved by more than 2-fold by phase transition of Cu@Pd–Pt NC into complexes of Cu3Pt and Pd1–Cu alloys via reduction annealing in the hydrogen environment (H2/N2 = 9/1, at 600 K for 2 h). Compared with pristine Cu@Pd–Pt and commercial J.M.-Pt/C NCs, specific activities (SAs) of postannealed Cu@Pd/Pt-H are 2.02- and 2.05-fold higher, respectively. Of special relevance is the fact that the mass activity (MA) of Cu@Pd/Pt-H is improved by 29% after 36 000 ADT cycles, whereas that of commercial J.M.-Pt/C is decreased by ∼32% after 31 000 ADT cycles. These results reveal the strong stability and activity of postannealed Cu@Pd/Pt-H and shed light on the development of high-performance and cost-effective cathodic NCs in alkaline fuel cells (AFCs).

Results and Discussion

Surface Morphology and Crystal Structure

The morphologies (including particle shape, crystal structure, and surface configuration) of experimental NCs were characterized using high-resolution transmission electron microscopy (HRTEM). Figure compares typical HRTEM images of pristine (Cu@Pd/Pt) and postannealed NCs (Cu@Pd/Pt-H). The corresponding inverse Fourier transformed (IFT) images (lower right corners) and line histogram of selected fringes are compared in insets. Variation in shape of postannealed NC as compared to pristine NC can be rationalized by phase segregation and atomic restructuring between NCs in reduction annealing. Such a phenomenon is triggered by reducing the surface Gibb’s free energy due to a thermal-induced atomic restructuring at the interparticle boundary and intraparticle interfaces of NCs. As shown in Figure a, the nanoparticles are grown into multifaceted spherical crystallites with twin boundaries (both sides of red arrow in IFT pattern) in Cu@Pd/Pt. In this case, the lattice space at twin facets is 2.13 Å. This value is decreased by 6.1% as compared to that of the ideal Pt(111) facet (2.26 Å), indicating the formation of a semicoherent core–shell interface with a strong compressive strain in the Pd shell region. Such a strong lattice strain is further probed directly by X-ray powder diffraction (XRD) analysis in a later section. A high defect density (denoted by green arrow) is formed due to the rapid crystal growth of Cu core and surface deposition of Pd atoms by interaction with a strong reducing agent.

Figure 1

HRTEM images of (a) pristine “Cu@Pd/Pt” and (b) postannealed “Cu@Pd/Pt-H” NCs. d-Spacing values of experimental NCs are calculated by using inverse Fourier transformed (IFT) images and their corresponding line histograms (insets). Fourier transformation patterns of selected areas in HRTEM images are shown in lower left corners. Low-magnification images of (c) pristine and (d) postannealed NCs are shown in the lower section.

Compared to Cu@Pd/Pt, Cu@Pd/Pt-H (Figure b) possesses less truncations and roughness in surface with larger particle size (Figure d) due to the formation of Cu3Pt and Cu1–Pd alloys. Such features are mainly attributed to restructuring among Pt, Pd, and Cu atoms between NCs. Such a hypothesis is confirmed by results of XRD and X-ray absorption spectroscopy (XAS) analyses. As compared to that of Cu@Pd/Pt, the lattice space is further compressed by 1.9% (d(111) = 2.09 Å). Significant distorted lattice fringes (denoted by yellow arrow) with domain size ranging from 1.5 to 2.0 nm reveal the lattice disordering of Cu3Pt by adjacent Cu1–Pd alloys and show a typical feature of the strong lattice strain at the heterogeneous interface (i.e., Cu@Pd binary interface). Fast Fourier transform (FFT) patterns (lower left corners) prove the atomic structure of experimental NCs. It is a qualitative interpretation of periodicity of the selected region in an image. In an HRTEM image, symmetry and profile of an FFT pattern refer to ordering of atomic arrangement. As shown in Figure a, nonsymmetrical twin bright spots with intensity difference indicate the presence of multiple twin boundaries, which confirms suppression of long-range ordering. Meanwhile, in the case of Cu@Pd/Pt-H, the presence of ring pattern is attributed to formation of polydispersed and locally disordered structure as compared to that of pristine NC. Low-magnification image in Figure c and histogram show that Cu@Pd/Pt is grown in a broad size distribution ranging from 2 to 6 nm with an average particle size of 3.8 nm, whereas in the case of Cu@Pd/Pt-H, particle size distribution range is noted from 7 to 11 nm owing to particle agglomeration and Cu3Pt/Cu1–Pd alloy formation. For comparison, crystal structural analyses of Cu@Pt and Pd@Pt NCs have been considered.[26]

Phase segregation, lattice strain, and average coherent length (Davg) of experimental NCs are further revealed by XRD analysis, and the corresponding structural parameters are summarized in Table . Composition for Pd phase in experimental NCs is estimated by using Vegard’s law, and the qualitative index for the preferential facet is determined by intensity ratios of H(111)/H( (H(111) and H(200) refer, respectively, to intensities for diffraction peaks of (111) and (200) facets). As indicated in Figure , peaks X1 and X2 centered at 17.524 and 20.221° refer to diffraction signals from (111) and (200) facets of metallic-phase Pt nanocrystals in Pt-CNT (green line). For XRD pattern of Pd-CNT, the two peaks (gray dashed line) are slightly shifted to the left (lower angle), suggesting its relatively larger lattice constant as compared to Pt-CNT. Such a lattice expansion could be attributed to formation of local disorder and possibly B dopant in Pd crystal due a rapid crystal growth by a strong reducing agent of NaBH4. In the meantime, Pd-CNT shows larger Davg, and this can be attributed to its higher surface free energy, which takes smaller specific surface area for stabilization as compared to that of Pt-CNT.

Table 1

XRD-Determined Structural Parameters of Pristine and Postannealing NCs Compared with Control Samples (Pd-CNT and Pt-CNT)

NC	facets (hkl)	2θ (deg)	d (Å)	D (nm)	H(111)/H(200)	Pd (%)a
Cu@Pd/Pt	(111)	17.515	2.262	3.86	2.08	95.5
	(200)	20.061	1.977	2.45		100.0
Cu@Pd/Pt-H	(111)	17.937	2.209	10.85	2.73	65.9
	(200)	20.721	1.915	8.03		65.2
Pt-CNT	(111)	17.524	2.261	4.79	1.89
	(200)	20.221	1.962	4.04
Pd-CNT	(111)	17.455	2.270	5.93	1.91	100
	(200)	20.080	1.976	3.72		100

Pd (%) is determined by using Vegard’s law. Reference crystal structure of Cu is refereed to mp-30 ID in Materials Project database. For Pd, lattice constant is estimated by XRD pattern of Pd-CNT.

Figure 2

XRD patterns of pristine (Cu@Pd/Pt) and postannealing (Cu@Pd/Pt-H) experimental NCs compared with control samples (Pd-CNT and Pt-CNT). Peaks “X1” and “X2” correspond to diffraction signals from (111) and (200) facets of NCs (Pt or Pd), whereas peak X* refers to (111) facet of Cu3Pt alloy. All of the spectra were measured under the incident X-ray of 18 keV.

For the case of Cu@Pd/Pt, a slight offset of diffraction peaks to the left indicates a lattice expansion of (111) and (200) facets as compared to those of Pt-CNT. The extent of Cu intermix is determined to be 4.5% for (111) and 0% for (200) facets. Considering that the extent of local disordering in Pd crystal is high (denoted by the high diffuse scattering background and high H(111)/H(200) ratio) and because of the uneven peak offsets, the changes of lattice spaces are assigned to the factors’ local expansion instead of alloy formation. On the other hand, for Co@Pd/Pt-H, a significant offset in diffraction peaks suggests the largest extent of Cu intermix (∼65%) in Pd crystal among experimental NCs. Meanwhile, a weak diffraction peak at 18.851° refers to a segregation of stoichiometric Cu3Pt alloy and the highest H(111)/H(200) value (2.74), indicating the strongest preference in (111) facet among experimental NCs. Those characteristics reveal a dramatic increase in the long-range ordering structure in Cu@Pd/Pt after H2 annealing. By cross-referencing HRTEM and XPS (later section) results, such heteroatomic intermix is attributed to segregation of the Cu to Pd region simultaneously with incorporation of Pt in the core region. Meanwhile, because of increased Pd content in the surface, easy intercalation of Pt atoms in the opened shell region took place and thus higher extent of Pd/Pt/Cu intermixing takes place. An even closer look reveals that for Cu@Pd/Pt-H because of Cu3Pt alloy formation, a higher index of lattice strain has been noted as compared to Cu@Pd/Pt. Meanwhile, significant increase in Davg after annealing is noted due to strong agglomeration in the absence of stabilizer (also shown in Figure d) between NCs. For comparison, details of structural interpretation of Cu@Pt and Pd@Pt NCs with XRD patterns are given in Figure S2.

XAS analysis was employed to analyze the local atomic and electronic structures of Pt atoms. Figure compares the normalized Pt L3 edge X-ray absorption near-edge spectra (XANES) and Fourier-transformed extended X-ray absorption fine spectra (EXAFS) of NCs under inspection. In a L3 edge spectrum, position of the inflection point (arrow X) refers to threshold energy (E0) for 2p to 5d electron transition and is linearly proportional to oxidation state of the target atom (particularly for transition metals). Intensity (HA) and width (WA) of near-edge absorption peak (white line) elucidate the relative extent of empty states and splitting of 5d5/2 orbital with the amount of surface oxygen chemisorption. Width (WB) and intensity (HB) of oscillation hump in the postedge region explain the extent of structure ordering around the target atom. It is evident from Figure S3 that metallic characteristics of Pt-CNT are similar to Pt foil as is evident from the similar position of inflection point and thus for postannealed Cu@Pd/Pt-H NC also (Figure a). The highest white-line intensity (HA) for Pt-CNT reflects the highest extent of oxygen adsorption at Pt atoms of Pt-CNT among all samples. On the other hand, significant suppressed HA for pristine Cu@Pd/Pt NC as compared to that of Pt-CNT reveals an inhibition of oxygen adsorption in Pt atoms or Pt oxidation. For Cu@Pd/Pt, the presence of strong electronegative Pt clusters on the surface reveals charge relocation to Pt clusters from neighboring atoms due to steric effects Pt clusters intercalating in the Pd shell. Charge relocation is further revealed by the downshift of inflection peak to low-energy sites (see Figure a inset). Pt cluster intercalation results in a local disordered structure as is revealed by a significant suppressed backscattering intensity in the postedge region (HB). As for Cu@Pd/Pt-H, incorporation of Pt atoms in the core followed by formation of Cu3Pt is consistently explained by reduction of Pt contents in NC surface and has been proved by XPS analysis in a later section. Compared to that of Pt-CNT, the comparable HA intensity indicates formation of crystal-phase Cu3Pt alloy, which consistently explains the results of XRD and XPS analyses (in a later section). In this event, compared to Cu@Pd/Pt, offset of inflection point close to that of Pt-CNT (Figure a, inset) suggests the typical feature of metallic Pt in Cu@Pd/Pt-H. This phenomenon indicates the absence of charge localization between Pt and neighboring atoms, again proving disassembling of intraparticle cluster-in-cluster interfaces due to phase segregation and formation of Cu3Pt and Cu1–Pd alloys in Cu@Pd/Pt-H.

Figure 3

Comparative Pt L3 edge (a) XANES and (b) k3 weighted EXAFS spectra of Cu@Pd/Pt and Cu@Pd/Pt-H with Pt-CNT.

Figure b compares Fourier-transformed Pt L3 edge EXAFS spectra (i.e., radial structure functions, RSFs) of experimental NCs and Pt-CNT. In an RSF spectrum, position and intensity of radial peak correspond, respectively, to the distances and number of backscatterers (neighboring atoms) around targeting atoms. For Pt-CNT, the radial peak (A) across 1.7–3.2 Å accounts for contribution of outgoing X-ray interferences in the metallic Pt–Pt bond. For the case of Cu@Pd/Pt, as compared to that of Pt-CNT, the radial peak is split into two peaks, and the intensity is substantially suppressed by ∼67%. These characteristics can be attributed to the destructive X-ray interferences of Pt–Pt, Pt–Pd, and Pd–Cu bond pairs around Pt atoms, which prove the high content of heteroatomic intermix in Cu@Pd/Pt. By adopting H2 annealing treatment, as compared to RSF of pristine Cu@Pd/Pt, the radial peak is shifted to the left and is enhanced by 2.66-fold. These characteristics indicate a substantially reduced metallic Pt–M bond length, which is consistent with the result of XRD, proving the formation of Cu3Pt alloy in Cu@Pd/Pt-H.

To figure out the effects of heteroatomic intermix and alloy formation, quantitative structure parameters of experimental NCs and control sample (Pt-CNT) are determined by model analysis on Pt L3 edge EXAFS spectra (Figure b), and the corresponding results are shown in Table . For Pt-CNT, the radial peak in the RSF results from the X-ray interference with the Pt–Pt bond pair at peak A (2.752 Å) with a coordination number (CNPt–Pt) of 6.45. Such a small CN is understandable due to the presence of high-density surface defects in rodlike fine NC in Pt-CNT in which most Pt atoms are located (Figure S1). Such a characteristic is consistently revealed by the higher intensity ratio of (111) to (200) peaks and broadened X-ray diffraction peaks as compared to that of the bulk Pt crystal (Figure ). For Cu@Pd/Pt, the radial peak of the Pt–M bond pair (peak A) is split into peaks C and D, which comprise contributions from backscattering interferences of CNPt–Pt (3.92), CNPt–Pd (2.86), and CNPt–Cu (0.57) with the heteroatomic intermixes (χ) of 38.9% for Pd and 7.7% for Cu around the Pt atom. With a total CN of 7.35, a stronger radial peak is expected but has been significantly suppressed by 67% as compared to that of Pt atom in Pt-CNT. Such a phenomenon seems controversial; however, it can be rationalized by destructive interferences between both in-phase (Pt–Pt) and out-of-phase (i.e., mainly Pt–Pd and partially Pt–Cu) backscattering X-rays[28] and the presence of high-density local defects at intraparticle interfaces. This scenario is expected in a ternary nanoparticle synthesized by rapid crystal growth processes in the presence of a strong reducing agent with sequential controls on metal ion reduction. In this event, as consistently revealed by suppressed scattering hump in the postedge region, Pt atoms tend to form discrete atomic Pt clusters in the shell region or the interface between Pd and Cu in the NC surface.

Table 2

Quantitative Results of X-ray Absorption Spectroscopy Model Analysis at Pt L3 Edge of Pristine (Cu@Pd/Pt) and Postannealed (Cu@Pd/Pt–H) Nanocatalysts Compared with Pt-CNTa

	Pt L3 edge
sample	bond pair	CN	R	χ (%)
Cu@Pd/Pt	Pt–Pt	3.92	2.711	53.3
	Pt–Pd	2.86	2.682	38.9
	Pt–Cu	0.57	2.562	7.7
Cu@Pd/Pt-H	Pt–Pt	1.33	2.674	28.3
	Pt–Pd	1.01	2.679	21.5
	Pt–Cu	2.36	2.576	50.2
Pt-CNT	Pt–Pt	6.45	2.752	100

χ stands for extent of heteroatomic intermix for M atoms around Pt in a bond pair of Pt–M, that is, χ for Pt–Pd refers to extent of Pd atoms among total coordination numbers around the Pt atom. For optimization of structure parameters, sigma square is determined to be 0.004 Å2 by fitting the spectrum of Pt foil and is adopted for fitting all experimental spectra.

For the case of Cu@Pd/Pt-H, as compared to radial peak A, the two radial peaks are merged in one (peak B) and shifted to the left. Those features refer to the constructive interference of backscattering X-rays from local coordination neighbors of the Pt atom. As consistently revealed by the enhanced HB/WB ratio in the postedge region and the presence of peak X* in the XRD pattern, Pt atoms are transformed from local disordering atomic clusters (probed by significantly suppressed HB/WB ratio in the postedge region) to a local ordering structure in a Cu3Pt crystal. Quantitative structure parameters around the Pt atom further confirm this scenario (Table ). Accordingly, CNs are determined to be 1.33 for Pt–Pt, 1.01 for Pt–Pd, and 2.36 for Pt–Cu bond pairs and ending up with the total CN of Pt atoms being 4.7 in Cu@Pd/Pt-H. This value is far smaller than that of the atom in the metallic crystal surface (6–9), which suggests mainly that Pt atoms are located in interfaceted corners or edges of NC. In this event, as compared to that of Cu@Pd/Pt, a substantially enhanced radial peak is a consequence of constructive interferences of backscattering X-rays in the local environment. It reveals the reduction of interfacial defect density due to the atomic restructuring followed by segregation of Pd1–Cu and Cu3Pt phases by thermal energy in Cu@Pd/Pt-H.

X-ray photoemission spectroscopy (XPS) analysis at Pt 4f and Pd 3d orbitals was performed to investigate surface chemical compositions and binding energies (BEs) of elements in experimental NCs. In this study, the incident X-ray with 485 eV excitation energy with probing depth of ∼2.6 nm is used. Figure shows the fitted XPS spectra in the Pt 4f and Pd 3d regions for the experimental NCs. In a Pt 4f spectrum, doublet peaks at 71.2 and 74.3 eV, respectively, emerge the photoelectron emission from Pt 4f7/2 and Pt 4f5/2 orbitals. However, for the Pd 3d spectrum, doublet peaks around 336 and 341 eV are photoemission lines from Pd 3d3/2 and Pd 3d5/2 orbitals, respectively. The peaks are further deconvoluted for determination of the signals from different oxidation states, and the corresponding results are summarized in Table . Accordingly, compared to Pt and Pd emission lines of metallic Pt and Pd, BEs are decreased to 71.06 eV at Pt 4f7/2 and 335.58 eV at Pd 3d5/2 orbitals in Cu@Pd/Pt. These decrements are attributed to charge relocation from Pd to Pt and steric shielding effects of Pt atoms in defect sites that protect Pd from oxidation. Compared to Cu@Pd/Pt, BEs of Pt 4f7/2 and Pd 3d5/2 are further decreased to 70.97 and 335.43 eV, respectively. Such a phenomenon can be attributed to the reduction of defect sites among Pt, Pd, and Cu clusters and in NC surface followed by phase segregation and formation of Cu3Pt and Pd1–Cu alloys in Cu@Pd/Pt-H during H2 annealing. Surface compositions of experimental NCs are summarized in Table . Accordingly, Cu@Pd/Pt comprise 53 atom % of Cu, 23 atom % of Pd, and 24 atom % of Pt. After H2 annealing, the three elements redistribute to 52, 27, and 21 atom %, respectively, in Cu@Pd/Pt-H. Given that probing depth of XPS is less than 2 nm, a slight difference for the composition change is understandable between NCs in the size ranging from 3 to 10 nm. In this event, as compared to that of Cu@Pd/Pt, a slight increase of 4 atom % depicts the surface segregation of Pd as consistently revealed by the increasing oxidation of Pd contents by 6 atom % in Cu@Pd/Pt-H.

Figure 4

XPS spectra of pristine and postannealing experimental NCs at (a, c) Pt 4f/Cu 3d and (b, d) Pd 3d orbitals.

Table 3

XPS-Determined Binding Energies of Experimental NCs Compared with Commercial J.M.-Pt/C

	binding energy (eV)
NCs	CuO	Cu(OH)2	Pd(0)	Pd(II)	Pd(IV)	Pt(0)	Pt(II)	Pt(IV)
J.M.-Pt/C	NA	NA	NA	NA	NA	71.36	72.51	74.06
Cu@Pd/Pt	74.96	77.81	335.58	336.79	338.24	71.06	71.99	73.02
Cu@Pd/Pt–H	74.97	77.69	335.43	336.52	337.79	70.97	71.90	72.93
Pd metal[29]	NA	NA	335.60	NA	NA	NA	NA	NA
Pt metal[30]	NA	NA	NA	NA	NA	71.20	NA	NA

Table 4

Comparative XPS-Determined Composition Ratios of Experimental NCs and Commercial J.M.-Pt/C

	oxide ratio
NCs	CuO	Cu(OH)2	Pd(0)	Pd(II)	Pd(IV)	Pt(0) (%)	Pt(II) (%)	Pt(IV) (%)	Cu	Pd	Pt
J.M.-Pt/C	NA	NA	NA	NA	NA	74.42	19	8.58	NA	NA	NA
Cu@Pd/Pt	61%	39%	75%	16%	9%	66	24	11	53%	23%	24%
Cu@Pd/Pt-H	65%	35%	68%	22%	10%	70	20	9	52%	27%	21%

Cyclic voltammetry (CV) analysis was employed for determining the electrochemical surface area (ECSA) and surface chemical states of experimental NCs and the control sample (J.M./Pt). The CV curves (Figure a) were recorded in N2-purged 0.1 M KOH solutions at 20 mV/s under room temperature, and the corresponding electrochemical parameters are summarized in Table . In a CV curve, the three distinctive potential regions are related to underpotential deposition of hydrogen atoms between 0 < E < 0.38 V, the double-layer region in between, and the chemisorption of oxygen species (OH) over 0.6 V vs RHE. The position and width of the current peak are susceptible to chemical composition and structure in the NC surface. For instance, the current peak at ∼0.8 V (vs RHE) in the forward sweep refers to the redox response by the formation of α Pt oxide (EOads) and that at ∼0.7 V (vs RHE) in the backward sweep is the contribution of oxygen desorption from the Pt surface of J.M./Pt (EOdes). Meanwhile, the broad peak width refers to the large variety of oxygen adsorption energies in different reaction sites of the Pt surface. For Cu@Pd/Pt, as compared to J.M./Pt, EOads is shifted to the right by 0.1 V, indicating the increasing barrier for surface adsorption of O. With the same metal loading in the electrode, there was a substantially increased peak intensity again showing the formation of highly dispersed Pt clusters (i.e., high electrochemical surface area). Notice that, compared to that of J.M./Pt, the EOdes in backward sweep is shifted to the right, suggesting the weaker oxygen adsorption in the Cu@Pd/Pt surface. These evidences explain its substantial improvement of mass activity (MA) and onset potential as compared to J.M./Pt.

Figure 5

(a) CV and (b) linear sweep voltammetry (LSV) spectra of pristine and postannealing experimental NCs compared with commercial J.M.-Pt/C catalysts.

Table 5

Comparative Electrochemical Performances of Experimental NCs in ORR

NCs	n	ECSA cm2 mgPt+Pd–1 a	Eonset V vs RHE	JK 0.85 mA cm–2	S.A.0.85 mA cm–2	MA0.85 mA mgPt–1
Cu@Pd/Pt	3.9	546.7	0.909	21.6	0.265	408.0
Cu@Pd/Pt–H	3.7	204.0	0.898	16.3	0.536	308.2
J.M.–Pt/C	4.0	257.0	0.915	5.3	0.261	67.0

ESCA is calculated by using the oxygen reduction region. The corresponding parameters are given in the following section (preparation of the electrode and the method for ORR activity experiment).

Apparently, compared to Cu@Pd/Pt, intensities of the EOads and EOdes peaks are, respectively, attenuated by ΔH1 and ΔH2 for Cu@Pd/Pt-H. Both the characteristics consistently revealed the reduction of ECSA as a result of the agglomeration between NCs. Meanwhile, position of the EOdes peak of Cu@Pd/Pt-H is shifted to the left, which depicts the higher energy barrier of oxygen desorption as compared to that of Cu@Pd/Pt. Those characteristics can be rationalized by restructuring of highly active surface Pt clusters into Cu3Pt accompanied by the formation of Pd1–Cu alloys. In these two phases, the presence of Cu CN increases oxygen adsorption energy. It suppresses ORR activities of Pt and Pd domains; as a result, Eonset of NC is reduced. In the same mechanism, reducing of local defects by alloy formation improves stability of Cu@Pd/Pt-H and will be discussed by results of the accelerated degradation test (ADT) later (Figure ).

Figure 6

(a) CV and (b) LSV curves of the pristine and postannealing experimental NCs in ORR for the initial state (#0 cycle) and the final state. (c) CV and (d) LSV curves of Cu@Pd/Pt-H NC for the selected ADT cycles. (e) Normalized initial and final current densities for experimental NCs compared with J.M.-Pt/C after different ADT cycles.

The ORR measurements were carried out in O2-saturated 0.1 M KOH solutions with rotation rates of working electrode at 400–3600 rpm. Examples of LSV curves at 1600 rpm are given in Figure b. The oxygen reduction is under a mixed kinetic and diffusion control from 0.95 to 0.80 V (vs RHE), followed by a mainly diffusion-controlled region. Through extrapolation, onset potential (Eonset) of experimental NCs follows the order Cu@Pd/Pt-H (0.898 V) < Cu@Pd/Pt (0.909 V) < J.M.-Pt/C (0.915 V). In an LSV curve, Eonset refers to threshold voltage for initiating ORR and is affected by trade-offs among the extent of surface oxidation (steric shielding effect), heteroatomic intermix (bifunctional mechanism), and configurational identity of heterogeneous clusters in NC. This explains that higher Eonset refers to the lower energy barrier for redox cycles in fuel cells.

The long-term durability of experimental NCs was tested using the accelerated degradation test (ADT) with repetitive potential CV cycles. The comparative CV and LSV curves of pristine and postannealing NCs for initial and post-ADT cycles are illustrated in Figure a,b, respectively, and the corresponding parameters are summarized in Table . Furthermore, to investigate the pathways regarding the ORR performances of Cu@Pd/Pt-H, changes in the surface chemical environment are revealed by CV analysis at selected ADT cycles (Figure c). According to Figure a, the intensity of the EOads peak progressively decreases with the ADT cycle and is attenuated by ΔH1 after 36k ADT cycles (please put ΔH1 in Figure a). Such a suppression of the EOads peak can be attributed to removal of oxide species from the NC surface. It reduces oxygen adsorption energy as consistently proved by increase of Eonset by 2% and decrease of the Pt/Pd-oxide reduction peak EOdes by ΔH2 as compared to the initial state (Figure d). In addition, the position of the EOdes peak is shifted to the right by 0.04 V. The two observations integrally show the suppression of oxygen species in Cu@Pd/Pt by H2 annealing. To figure out the stability of experimental NCs in ORR, residual currents (JR) at 0.85 V (vs RHE) at selected cycles are compared in Figure e. Accordingly, the JR of Cu@Pd/Pt-H is improved by 29.3% after ADT for 36 000 cycles and exhibits the strongest resistance to oxidative corrosion or OH passivation compared to J.M.-Pt/C (JR decreased by 37.2% after 31 000 ADT cycles). In Figure e, the residual current is normalized by the current density in the first cycle. For Cu@Pd/Pt, the abnormal current fluctuation of the ADT curve is due to surface oxidation of NCs during the sample exposure to ambient storage between each 5000 cycles, and the measurement was stopped at 40 000 cycles due to detachment of the NC slurry film from the rotation disk electrode. These results are consistent with our previous discovery that the decoration of Pt clusters can keep ORR performance of Cu@Pd NC. However, in the same measurement scenarios, Cu@Pd/Pt-H showed a stable JR (±7%) during ADT, which demonstrates its outstanding stability in ORR by reduction of local defects in the near surface of NC accompanied by the formation of Cu3Pt and Pd1–Cu. For further discussions of experimental results, comparative electrochemical and long-term stability analyses of Cu@Pd/Pt-H with bimetallic (Cu@Pt and Pd@Pt) NCs are compared in Figure S4. Meanwhile, CV sweeping and LSV curves of pristine and postannealed NCs for different ADT cycles are given in Figure S5. For evaluating stability in ORR, changes on nanostructures of experimental NCs before and after ADT are investigated by using TEM (Figure S6) and DEX (Table S2) analyses. Accordingly, Pd corrosion is suppressed by H2 annealing, which could be the reason for the improved ORR stability for Cu@Pd/Pt-H as compared to that of Cu@Pd/Pt.

Table 6

Comparative Durability of Experimental NCs in ORR for Different ADT Cycles

		current density at 0.85 V JR0.85 (mA cm–2)
NCs	ADT cycles	before ADT	after ADT	change (%)
Cu@Pd/Pt	40 000	3.13	2.90	–7.3
Cu@Pd/Pt–H	36 000	3.38	4.37	+29.3
J.M.-Pt/C	31 000	4.17	2.62	–37.2

Conclusions

In this study, ternary Cu@Pd/Pt NC was designed with Cucluster-in-Pdcluster and Pt cluster decoration on the surface for ORR application in alkaline electrolyte. Such synthesized NCs were subjected to reduction annealing in a H2 environment. The as-obtained NC showed excellent electrocatalytic activity as well as durability toward ORR in alkaline medium compared to the commercial J.M.-Pt/C catalyst. Most importantly, Cu@Pd/Pt-H NC showed SA (0.536 mA cm–2), which is almost twice as that of pristine (i.e., for Cu@Pd/Pt: 0.265 mA cm–2) and J.M.-Pt/C (0.261 mA cm–2). An exceptional stability in a long-term ADT for different cycles is achieved. An even impressive result is that, for Cu@Pd/Pt-H, after 36 000 ADT cycles, residual current density is increased by 29% as compared to its original value. By cross-referencing the results of local and crystal structural parameters, we demonstrate that the improvements of SA and JR in ORR can be attributed to the formation of Cu3Pt and Pd1–Cu alloys accompanied by the reduction of local interface defects by H2 annealing on Cu@Pd/Pt. This reduction annealing is conventionally available, which can be an effective assessment the development of long-stability cathodic NCs in alkaline fuel cells.

Experimental Section

Synthesis of Cucluster-in-Pdcluster (Cu@Pd) Nanocrystals

Cu@Pd NC was synthesized by using a sequential wet chemical reduction method. Reaction steps for synthesis methodology are presented in Scheme (steps 1–3). In the first step, carbon nanotube support (CNT, Cnano Technology Ltd.) is acid-treated in 4.0 M H2SO4 at 80 °C for 6 h for the sake of strengthening the attachment of metallic crystals on their surface. After that, CNT powder is washed by distilled water until the pH value of rinsing water is 6.0. The CNT is then mixed with Cu precursor solution, which is made up of 0.023 mmol copper(II) sulfate pentahydrate (CuSO4·5H2O, 99%, Sigma-Aldrich Co.) and 30 g of distilled water. The mixture is stirred at 200 rpm at 25 °C for 6 h. After mixing, 3.0 g of water solution containing 10 mg of sodium borohydride (99%, Sigma-Aldrich Co.) is added in the mixture to reduce Cu ions (solution A). In this step, excess molecular ratio of NaBH4 (0.252 mmol) was added to ensure complete metal reduction and stabilize metastable metallic Cu cluster in the reaction system. After stirring at 200 rpm for 10 s, 1.0 g water solution of Pd precursor (containing ∼24.2 mg, ∼0.023 mmol, of Pd ions) is added to solution A and form solution B. In this step, Pd ions are reduced by the excessive amount of NaBH4 added in the first step and deposit on Cu surfaces. The Pd precursor solution was prepared by dissolving powder of Pd metal (99%, Sigma-Aldrich Co.) in the solution of DI water and HCl with the weight ratio of 9:1 (1.0 M HCl(aq) solution). The Pd powder is sub-micrometer in diameter with a specific surface area of 40–60 m2/g (CAS number 7440-05-3). Such a sub-micrometer particle is chemically active and could be dissolved in HCl solution. More specifically, 1.5 g of palladium chloride (PdCl2) is mixed in a 100 mL sample vial with 10 times diluted aqueous hydrochloric acid solution (22 g) + 84.6 g DI water. The atomic ratio of Pd/Cu is 1.0 in solution B.

Scheme 1

Schematic Representation for Structure Evolutions in Reaction Steps of Crystal Growth for Cu@Pd/Pt-H Ternary NC

Synthesis of Atomic Pt Cluster-Decorated Cucluster-in-Pdcluster Nanocrystals and Reduction Annealing

Pt precursor solution was prepared in advance by diluting ∼13.3 mg of H2PtCl6·6H2O (99%, Sigma-Aldrich Co.) to 500 mg with distilled water. Once the reaction was complete in solution B, Pt precursor solution was added into solution B. In this stage, the Pt clusters are intercalated in the Pd shell region by a galvanic replacement of Pt4+ to Pd followed by the reduction of residual metal ions by a reducing agent (NaBH4). The resulting powder is washed several times with acetone, centrifuged, and then dried at 70 °C. Henceforth, the Pt cluster-deposited Cu@Pd NC is called Cu@Pd/Pt. As-prepared Cu@Pd/Pt NC was annealed at 600 K in a hydrogen environment (H2/N2::9:1) for 2 h followed by natural cooling at room temperature. In this study, postannealed Cu@Pd/Pt NC is called Cu@Pd/Pt-H.

Preparation of the Electrode and the Method for ORR Activity Experiment

The slurry sample for the ORR experiment was made by dispersing 5.0 mg of CNT-supported catalyst in 1.0 mL of isopropanol containing 50 μL of Nafion-117 (99%, Sigma-Aldrich Co.). The mixture was ultrasonicated for 30 min prior to the ORR test. For conducting the ORR test, 10.0 μL of catalyst slurry was casted and air-dried on a glassy carbon rotating disk electrode (0.196 cm2 area) as the working electrode. Hg/HgCl2 (the voltage was calibrated by 0.242 V, in alignment with that of RHE) electrode saturated in KCL aqueous solution and a platinum wire were used as the reference electrode and the counter electrode, respectively. The electrochemical surface areas (ECSAs) of the experimental catalysts were calculated by acquiring the Coulombic charge for reduction of Pt or Pd oxides after integration and double-layer correction using the following equationwhere Qref denotes the charge required for reduction of monolayer oxide from the bright Pt surface (i.e., 0.405 mC cm–2), m stands for metal loading, and QPt represents the charge required for oxygen desorption, which is calculated by the following equationHere, ν is the scan rate for cyclic voltametry (CV) analysis and integral parts refer to the area under the Pt/Pd oxide reduction peak in CV curves.

The kinetic current density (JK) was calculated based on the following equation:where J, JK, and JL are the experimentally measured, mass transport free kinetic, and diffusion-limited current densities, respectively. For each NC, the MA and SA were obtained when JK was normalized to the Pt loading and ECSA, respectively. Slopes of LSV curve linear fits at half current potential (normally at inflection point) are adopted to the Koutecky–Levich (K–L) equation (eqs and 3) to calculate the number of electrons transferred in ORR.where J is the measured current density, JK and JL are the kinetic and diffusion-limiting current densities, respectively, ω is the angular velocity, n is transferred electron number, F is the Faraday constant, Co is the bulk concentration of O2, Do is the diffusion coefficient, v is the kinematic viscosity of the electrolyte, and k is the electron-transfer rate constant.

Characterization of Experimental NCs

The surface morphologies of as-prepared NCs were analyzed by using high-resolution transmission electron microscopy (HRTEM) operated at a voltage of 200 kV in the Electron Microscopy Center at National Sun Yat-Sen University. Average coherent length of experimental NCs is calculated from XRD peak broadening of (111) facets using the Scherrer equation. For XRD analysis, the incident X-ray of wavelength 0.688 Å (18.0 keV) was used at beamline of BL-01C2 at National Synchrotron Radiation Research Center (NSRRC), Taiwan. The X-ray photoelectron spectroscopy (XPS) analysis was employed to identify the surface chemical states of the experimental NCs at BL-24A of NSRRC (Hsinchu, Taiwan). The XAS spectra of experimental NCs were measured in the fluorescence mode at the BL01C1 and 17C beamlines at NSRRC (Taiwan) and at the beamline of BL-12B2 at Spring-8 (Japan). A Si monochromator was employed to adequately select the energy with a resolution ΔE/E better than 10–4 at the Pt L3 edge (11 564 eV). All catalysts were dispersed uniformly on the tape and prepared as thin pellets with an appropriate absorption thickness (μx = 1.0, where μ is the X-ray attenuation coefficient at the absorption edge and x is the thickness of the sample) to attain a proper edge jump step at the absorption edge region. To acquire acceptable quality spectra with good quality, each measurement was repeated at least twice and averaged for successive comparison. For the EXAFS analysis, the backgrounds of the pre-edge and the postedge were subtracted and normalized with respect to the edge jump step from the XAS spectra. The normalized spectra were transformed from energy to k-space and further weighted by k3 to distinguish the contributions of backscattering interferences from different coordination shells.

The electrochemical measurements were carried out at room temperature using a potentiostat (CH Instruments model 600B, CHI 600B) equipped with a three-electrode system. Cyclic voltammetry (CV) and linear sweep voltammetry (LSV) data were measured at the voltage scan rates of 0.02 and 0.001 V s–1 and potential ranges of 0.1–1.3 V (V vs RHE) and 0.4–1.1 V (V vs RHE), respectively, in an aqueous alkaline electrolyte solution of 0.1 M KOH (pH 13). The rotation rate of 1600 rpm was used for LSV. N2 and O2 atmospheres were used for CV and LSV, respectively. Electrochemical stability of the ternary NCs was characterized using an accelerated durability test (ADT) in the potential range of 0.5–1.0 V (V vs RHE) with the applied scan rate of 0.05 V s–1 in O2 atmosphere for different ADT cycles.