Rapid crystal growth of bimetallic PdPt nanocrystals with surface atomic Pt cluster decoration provides promising oxygen reduction activity

Institute of Materials Science and Engineer Department of Mechanical and Biomedical E Kowloon, Hong Kong Department of Engineering and System S Hsinchu 30013, Taiwan. E-mail: chencaese 34271 Department of Electrical Engineering, Inst Tsing Hua University, Hsinchu 30013, Taiw Institute of Nuclear Engineering and Science 30013, Taiwan † Electronic supplementary information supported Pt catalysts before and aer A PdPt nanoparticle synthesized by using method, geometrical calculation r 10.1039/c7ra08405f Cite this: RSC Adv., 2017, 7, 55110


Introduction
Atomic conguration could be a dominant factor in the oxygen reduction reaction (ORR) activity of bimetallic nanocatalysts. Our previous studies and literature have proved that nanowires and nanorods possess enhanced ORR activity and chemical durability when compared with Pt NPs. 1 Such ORR enhancement on Pt-based nanomaterials could be attributed to their low roughness contact (i.e., reduced contact resistance) with the carbon surface, preferential exposure of crystal facets to reaction media, and properly arranged heteroatomic Pt-Pd bond pairs (i.e., facilitated oxygen reduction pathways) on crystal surfaces. 2 Apart from composition differences, recent studies have revealed that the electronic and chemical states of NPs can be manipulated via formation of 1-D structures with surface alloying, thus varying the surface oxophilicity and substantially enhancing the ORR performance. 3 An interesting approach to ORR optimization is to relocate the Pt cluster location on cathode NPs. Because of electron injection effects via local lattice strain and structure conguration, Pd NP with slight atomic Pt clusters decorated in nearsurface region shows outstanding ORR activity and thus promoting the development of low temperature fuel cells with low fabrication cost. 4 It has been reported that the formation of intra-particle heterogeneous interface (particularly cluster-incluster, surface modication, and core-shell structure) in NP can substantially improve their stability and catalytic activity as compared to that of homogeneous alloy. By proper decoration of atomic Pt clusters on Co core -Pd shell NP, the result from longterm accelerated degradation test (ADT) demonstrated that the numbers of ORR working cycles increase by 10 times and the device efficiency is improved by 55-65% as compared to the commercial Pt catalyst. 5 In such a NP, effects of local lattice strain near crystal surface are one of the most inuential factors in its electrochemical properties. 6 In this work, we demonstrate that ORR activity of PdPt NPs can be substantially improved by $29.4 times over Pt metal by decorating Pt clusters with proper atomic structure in the nearsurface region as compared to commercial Pt catalyst. Using robust wet chemical fabrication processes, such an NP with Pt metal loading less than 10 wt% could be a potential candidate of low-cost catalyst among fuel cell cathode materials. We also show that the ORR activity of PdPt NPs is optimized by controls of reaction methods. The effect of atomic structure and Pd alloying on ORR performance of PdPt NPs has been elucidated by crossreferencing results of electrochemical analyses (including linear sweep voltammetry (LSV), CO stripping, and CV sweeping curve), X-ray diffraction (XRD), X-ray absorption spectroscopy (XAS), X-ray photoemission spectroscopy (XPS), and high-resolution transmission electron microscopy (HRTEM) characterizations. For clarifying effects of atomic conguration on ORR reactivity, structures and electrochemical properties of control samples (CNT supported Pt and Pd NP prepared by identical method to that of PdPt-s) are compared in this study.

Experimental
Preparation of Pt cluster decorated Pd nanoparticles and electrode for ORR activity test Atomic Pt clusters decorated Pd NP were prepared by using a wet chemical reduction method with sequential control. Before metallic NPs growth, carbon support (carbon nanotube "CNT from Cnano Technology Ltd."), was acid treated by immersing in aqua solution of 4.0 M sulphuric acid at 80 C for 6 h and then washed by distilled water till pH value of washing liquid was 6.0. Processes and schematic representation for growing PdPt-s NPs is shown in Fig. 1. In step 1, 50.0 mg of acid treated CNT powder was immersed in aqueous solution of Pd(IV) (Pd(NO 3 ) 4 , 99%, Sigma-Aldrich Co.) and stirred at 250 rpm at 25 C for 4 h resulting in Pd 4+/2+ chemisorption on CNT (CNT-Pd ads ). In this mixture (solution A), Pd/CNT ratio is 30 wt% (i.e., Pd ¼ 15 mg) and distilled water is 30 g. The Pd precursor solution was prepared by dissolving metal powder (Pd, 99%, Sigma-Aldrich Co.) in 1.0 M HCl(aq). Aer immersion, 3.0 g of solution B containing 10.0 mg (0.252 mmol) NaBH 4 (99%, Sigma-Aldrich Co.) was added to solution A and stirred at 250 rpm for 15 s in step 2. In step 3, Pt precursor solution containing 1.0 g distilled water and $6.42 mg ($0.032 mmol) of Pt ions was added in solution A with a molar ratio of Pt/Pd ¼ 0.1. Pt precursor, 17.05 mg of H 2 PtCl 6 $6H 2 O (Sigma-Aldrich, CAS number: 18497-13-7 206083) was dissolved in distilled water at room temperature. In step 3, atomic Pt clusters were decorated on the Pd NP surface via a galvanic replacement of Pd metal with Pt 4+ ion simultaneous reduction of residual metal ions (Pd 4+ and Pt 4+ ) with H: radicals. To complete metal reduction, additional reduction agent with identical amount of NaBH 4 to that in solution B was added to solution A in step 4 subsequent to the addition of Pt 4+ precursor and thus resulting in atomic Pt clusters decorated Pd NP (namely PdPt-s). Preparation of Pd-s followed similar processes as those used to prepare PdPt-s, except that only Pt (Pd) precursor was used in reaction system (i.e., only process to step 3 for preparing Pd-s). For preparing Pts, Pt 4+ precursor was used in step 1 and reaction proceeded to step 3.
As for PdPt alloy NPs (namely PdPt-p), polyol reduction method was employed. In reaction system, water solutions of Pt and Pd metal precursor (molar ratio of Pt/Pd ¼ 0.1) were mixed in 30 g of ethylene glycol (EG, 99 wt%, Sigma-Aldrich Co.) in the presence of 10 wt% (vs. EG) PVP polymer stabilizer (PVP-40, MW ¼ 40 000, 99%, Sigma-Aldrich Co.). Aer mixing at 25 C with a stirring speed of 250 rpm for 1 h, the mixture was heated at 160 C for 2 h in formation of polymer stabilized PdPt NP in EG (solution C). PdPt NP was extracted by a solvent exchange method to ethanol solution in the previous work. 7 The PdPt-p NPs was then mixed with acid treated CNT in ethanol (Pd/ CNT ratio is 30 wt%) at a stirring speed of 250 rpm for more than 12 h to form CNT supported PdPt-p. For stabilizing NC in CNT surface, PdPt-p powders were ltered, dried at 75 C for 3 h, and then annealed in Ar ambient at 270 C for 1 h.
Methods for preparing slurry samples on electrode in ORR activity test was adopted from previous studies and literature. 8 Details of slurry, electrode preparations and electrochemical characterizations are given in (ESI †).

Physical structure characterizations on PdPt nanoparticles
The physical properties of experimental NPs were determined by microscopic (HRTEM) and spectroscopic (X-ray absorption and X-ray photoemission) characterizations at electron microscopy center of National Sun Yat-Sen University and beamlines (BL-17C, 01C1, and BL-24A1) at National Synchrotron Radiation Research Center (NSRRC), Taiwan, respectively. The XRD patterns were collected at beamline of BL-12B2 at Spring-8, Japan.

Results and discussion
Crystal structure characterizations Surface morphology and crystal structure of experimental NPs are determined by TEM analysis (line histogram denoted by yellow double arrow, high resolution image, and low magnication image are shown in Fig. 2 and its inset). Shown in Fig. 2a, a slight angle (Ø 1) between the two (111) lattice fringes (with an interplanar spacing "d (111) " of 2.257 AE 0.03Å) in an absence of mirror reection images at two sides of selected axis (yellow double arrows) indicates the stacking of two Pt NPs in Pt-s. Denoted by yellow arrows, Pt-s possesses prevailing defects in NP surface. Such a surface morphology is attributed to the absence of stabilizers upon crystal growth and is revealed by aggregation of Pt NPs (Fig. 2a inset) and the random displacement of interatomic distances between Pt atoms (R1, R2, and R3 shown in line histogram). The distinct inter-particle boundaries and sharp surface morphology indicate the strong selfstabilization of Pt NP that reaches a critical size of heterogeneous crystallite. Above characteristics reveal the fatal issues of high surface free energy defect sites and a lack of heteroatomic bonding for Pt NPs in electrochemical reactions. Those issues cause easy corrosion and atomic restructure of edge atom on surface of Pt NPs (Fig. S1 †) in electrochemical characterizations. Therefore, it is expectable to see the low electrochemical stability of Pt-s in an accelerated degradation test (ADT) of oxygen reduction reaction (ORR) (see Fig. 4). Details for structural and electrochemical correlations will be discussed in the latter section. HRTEM image and line histogram of CNT supported Pd NPs (Pd-s) are demonstrated in Fig. 2b. As demonstrated in Fig. 2b inset, Pd NPs are drop-like particles with indistinct inter-particle boundaries, revealing a severe interparticle collapse as a result of instability (easy oxidation) of Pd NPs (also proved by Pd K-edge XAS analysis). Consequently, it is inevitable for Pd-s to face a similar fate with Pt-s (low electrochemical performances) at both ORR and ADT (Fig. 4). Comparing to that of Pt-s, Pd-s has less interatomic displacement (R1 and R2) in line histogram. Such a feature can be explained by reduction of surface defects via atomic restructure upon interparticle collapse. Fig. 2c shows that PdPt-s NPs are grown in decahedron-like crystal; where d (111) is 2.295 AE 0.03Å and d (200) is 2.003 AE 0.03Å. Compared to that of metallic Pt,d (111) and d (200) of PdPt-s are expanded by 1.30% and 2.01%, respectively. Such an uneven lattice displacement explains the preference of Pt allocation at Pd {111} facets by minimization of total surface free energy in a NP system. 9 Given that atomic  Table 1. Accordingly, the three peaks A, B, and C at 39.91 , 46.33 , and 67.74 are diffraction lines of Pt/Pd (111), (200), and (220) planes with d (hkl)_Pt /coherent lengths (D (hkl)_Pt ) of 2.256Å/68.2Å, 1.957Å/51.2Å, and 1.382Å/47.4Å, respectively. The varying D (hkl)_Pt at different planes indicates an anisotropic crystal growth between the three facets in Pt-s. For PdPt-s, diffraction lines of the three facets are shied to lower angles, revealing lattice expansion by 1.52% to 2.15% as compared to that of metallic Pd planes due to formation of Pt to Pd alloy in NP. Uneven lattice strain indicates the preferential positioning of Pt atoms at Pd (111) and (220) facets with substantial displacement between Pt-Pd atoms (also proved by XAS characterization in latter section). As compared to metallic Pt lines, uneven shis in (111) and (200) lines reveal the preferential local alloying (by galvanic replacement between Pt 4+ ions and Pt metal atoms) in (200) planes of PdPt-s. In PdPt-s, D avg is 41.8Å for (111), 31.5Å for (200), and 29.4Å for (220) planes. The largest D (111) again revealing the preferential crystal growth in (111) planes. High local micro-compression strain and Pd-Pt alloy with small D avg together suggest that a considerable extent of local alloying between Pt 4+ ion and metallic Pd in (220) planes. Such a phenomenon is expectable considering the nature of surface free energy (g) minimization by placing Pt atoms on open facets (Pd (220)). 10 Given that identical carbon support is employed for all NPs, presence of diffusion scattering (denoted by Q and Q*) at two sides of main diffraction lines implies the high roughness NP surface in PdPt-s by Pt clusters decoration and will be consistently proved by XAS analysis. For XRD pattern of PdPt-p, downshi of diffraction lines to the same extent illustrates a formation of homogeneous PdPt alloy with a similar D avg in the three planes. Meanwhile, PdPt-p shows a strongest preferential growth at (111) facets among experiment NPs. Such a preferential growth is triggered by steric effect of pyrrolidone ligands 11 capping on opened facets for minimizing the system energy. 9a Size distribution and D avg of experimental NPs are further investigated by TEM analysis with consistent results shown in Fig. S3. † Position of Pt atoms and the nanostructure of PdPt NPs are further conrmed by XAS characterisation. Fig. 4a compares the normalized Pt L 3 -edge X-ray absorption near-edge spectra (XANES) of PdPt-s, PdPt-p, and Pt-s. In an L 3 -edge spectrum, position of inection point (arrow X) refers to a threshold energy (E 0 ) for 2p to 5d orbital transitions and is proportional to the oxidation state of transition metals. Intensity (h a ) and width (W a ) of near-edge absorption peak (peak A) correspond to the extent of empty states and splitting of 5d 5/2 orbitals. The width (W B ) and intensity (h B ) of oscillation hump in post-edge region elucidate the extent of local atomic structure ordering. Shown by similar inection position, the metallic characteristics of Pt atoms in all NPs are evident while a signicantly reduced h B denotes the presence of high density local structure defect (or disordering) around Pt atoms in PdPt-s. Compared to h a of Pt-s and PdPt-p, an increase of intensity by $7.7% features the prevailing oxygen chemisorption. Atomic structure of PdPt-s is consistently conrmed by extended X-ray absorption ne structure (EXAFS) spectrum ( Fig. 4b and tting results Table 2). As determined, Pt atoms are stacked as tri-Pt-atom clusters (CN Pt-Pt ¼ 2, peak E)    atop Pd NP surface. The substantial splitting between peaks E and E* is due to typical destructive interferences between signals from Pt-Pt and Pt-Pd bond pairs with similar amount of coordination numbers (CN Pt-Pt ¼ 2.09 and CN Pt-Pd ¼ 1.04), where oxygen CN (peak D) per Pt atom is $1.5.
A non-stoichiometry of oxygen CN can be attributed to coexistence of chemisorbed oxygen atoms in atop sites and edge bridge sites between Pt and neighbouring Pd atoms. In the absence of stabilizer agents in crystal growth, a low oxygen CN at Pd atoms (0.36) with a high oxygen CN at metallic Pt atoms ($1.5) are evidences to formation of island-like-growth Pt clusters as shielding layer capping in PdPt-s surface region. Such an island-like surface Pt-cluster is consistently revealed by the presence of diffusion scattering in XRD patterns and the suppressed post-edge scattering hump in XANES. Results of local atomic structure around Pd atoms complimentary verify such a hypothesis. As indicated in Table 2, CN Pd-Pd and CN Pd-Pt are 5.54 and 2.09, respectively. Compared with local structure around Pt atoms, similar preference of hetero-atomic intermix (c) is found both around Pt (c(Pd-Pt)) and Pd (c(Pt-Pd)) atoms. Therefore, a greater total CN of Pd than Pt atoms and high CN Pt-O elucidate the formation of Pd core crystal with high density of surface Pt cluster decoration. For EXAFS spectrum of PdPt-p, compared to signal of metallic Pt-Pt bond in Pt-s spectrum (peak C*), upshi of atomic bond pair signal (peak C) accounts for a slight out-of-phase interference between X-rays interacting with Pt-Pt (CN ¼ 7.35) and Pt-Pd (CN ¼ 0.99) bond pairs. In the meantime, CN is 7.31 for Pd-Pt and 1.13 for Pd-Pt bond pairs around Pd atoms. Above quantitative structural parameters suggest that Pt and Pd atoms tend to form homo-atomic clusters in PdPt-p. Such a hypothesis is consistently conrmed by results of XRD, EDX map images (Fig. S2 †), XPS (Fig. S6 † and  Table S1 †), and CO stripping (Fig. S7 †) analyses.
The decoration of Pt clusters on NP surface is further conrmed by cross-referencing results of local structure inspection ( Fig. 4c and d and Table 2), geometrical correlation ( Fig. S4 †), and electrochemical analyses ( Fig. 5 and S5 †). Fig. 4c compares the Pd K-edge XANES spectra of experimental NPs, where two sorption peaks A and A* are typical characteristics of metallic Pd atom. As indicated, PdPt-p shows similar spectra features to that of Pd foil which conrms majority of Pd atoms are in metallic state. Such a result is consistently revealed by its radial distribution prole in EXAFS spectra (Fig. 4d), where peaks B and B* are respectively the X-ray interferences from Pd-Pd and Pd-Pt bond pairs. For PdPt-s, the two radial peaks are shied to high radial site to different extents elucidating the extension of interatomic bond length and the local distortion between Pd and Pt atoms. Those characteristics are complimentary disclosed by uneven micro-strain and local alloying on different planes of PdPt-s (Table 1). It is possible to have mild solid solution of Pd-Pt in PdPt-s surcial region. Given that the reaction rate of metal ion reduction by H: radicals (nano second) is far higher than that of ion-metal exchange rate in galvanic replacement, surface decoration of Pt clusters could be the main structure in PdPt-s. This statement is consistently explained by high Pt-O CN and low Pt-Pt CN of PdPt-s. The surface decoration of Pt clusters is further revealed by CV, XAS, XPS (Fig. S6 †), and CO stripping (Fig. S7 †) analysis. More evidences are revealed by comparing RSF curves of PdPt-s and Pd-s NP at Pd K-edge. Shown in Fig. 4c, radial peak C centered at 1.3Å originates from interference of X-ray between Pd and O atoms chemisorbed in carbon supported Pd NP. Given that Pd-s is synthesized by using identical method to PdPt-s except the presence of Pt atoms, the strong Pd-O peak can be attributed to the surface oxidation by O chemisorption in Pd-s. Consequently, two important structural indexes prove that Pt atoms are preferentially located at defect sites as protection layer in NP  (111) and (200) facets of metallic Pt. The significant higher current density of peak c reveals the preferential grow of (111) facets in Pt-s. The counterpart responses of H + desorption appears at peaks e and e* in positive sweep curve.
surface. (1) In the absence of polymer stabilizer, Pd-O CN in PdPt-s NPs substantially decreases by reducing Pt atoms on Pd core surface by NaBH 4 . (2) A dramatic suppression of peak B (Pd-Pd and Pd-Pt) intensity because of destructive interference of backscattering waves as a result of Pd-Pt local alloy in NP surface. Taking structural information, geometric correlation, and intrinsic properties of selected metals into considerations, nanostructure of experimental NPs can be conrmed by crossreferencing the local structure and oxidation states of Pt and Pd atoms (Fig. S4 †).
Hypothetical mechanisms for crystal growth in self aligned wet chemical method (Pt-s and PdPt-s) and polyol reduction method (PdPt-p) are proposed by cross-referencing experimental results, solid (liquid) state ions diffusion, and chemical reaction kinetics. As demonstrated in Scheme 1, NP in both the two methods are grown in heterogeneous nucleationcrystal pathways in three stages including (a) metallic ion incubation and embryo formation in substrate, (b) crystal nucleation, and (c) heterogeneous crystal growth followed by atomic cluster decoration in crystal surface (in bimetallic PdPt-s NP). In Pt-s NP, the three stages are (a) chelation between Pt 4+ ions and dangled (C 5 :) or thiolated defect sites (C 5 -C 2 H 6 SO) to form embryos on carbon support, (b) reduction of Pt 4+ ions on embryos surface by H: radicals (provided by sodium borohydride) to form Pt nuclei, and (c) Pt NP formation with preferential growth at (111) facets. In stage (c), strong bonding between thiol groups and Pt atoms forms a steric barrier to protect NP from oxidation. Nucleation rate in stage (b) is dominated by kinetics of three reaction pathways and result in the nal atomic congurations of NP. The reaction kinetics of three pathways are (1) K ads-Pt : chemisorption of Pt 4+ atop embryo surface heqn (1)i followed by its reduction by electrons [e] donated by H: radicals heqn (2)i, 12 (2) K ads-C : chemisorption heqn (1)i and reduction of chemisorbed Pt 4+ (Pt ads-C 4+ ) into Pt 0 at C 5 : or C 5 -C 2 H 6 SO sites by [e] heqn (2)i, 13 (3) diffusion heqn (3)i and allocation of Pt 0 atoms or Pt ads-C 4+ from carbon to embryos surface.
where d(Pt 4+ )/dt refers to amount of Pt 4+ chemisorbed at time t and a and a are constants related to the chemical potentials and vibrational frequency (amplitude) of sorption sites at Pt embryo. The kinetics of Pt 4+ chemisorption at carbon (i.e., C5: and C 5 -C 2 H 6 SO sites) are dened as K ads-C in pathway (2).
where D 0 is an intrinsic parameter of material (both carbon support and Pt ads-C 4+ ) and is independent of temperature. E d is the activation energy for Pt ads-C 4+ diffusion which includes the energy barrier for Pt ads-C 4+ exchange between sorption sites due to chemical potential forces and momentum kinetics (vibrations of carbon in directions normal "3" and horizontal "h" to Pt ads-C 4+ diffusion pathways). Given that kinetics of H: radical reduction ($10 À7 to 10 À9 second per effective reaction) is far higher than that of both K ads-Pt (1 to 10 3 seconds per effective sorption) and D (10 À3 to 10 À6 second per effective diffusion), growth of Pt-s is dominated by eqn (2) and NP size is controlled by local depletion of H: radicals in carbon support. Crystal growth of PdPt-s goes on the same pathways as Pt-s except that Pd 4+ and Pt 4+ ions are coexisting in the second pathway. Therefore, K a of second step is affected by concentrations of H: radical [e], Pt ads-C 4+ and Pd ads-C 4+ in reaction system (eqn (4)) ] ¼ 0.1), formation of Pd nuclei in core crystal followed by decoration of atomic Pt clusters are expectable. In PdPt-s, crystal growth is stopped by dispersing end product into distilled water with 10 times the volume aer mixing reduction agent for 15 second.
In polyol reduction method, PdPt-p NP is grown through pathways including (4) decomposition of ethylene glycol into metastable radical pairs of HOCH 2 -CHO:-H:, (5) diffusion of radicals to solid surface in polymer matrix, (6) diffusion of reduced metal atom to NP surface in polymer matrix, and (7) reduction of metal ions to solid phases surface by CHO: radicals. Kinetics of pathway (4), is controlled by reaction temperature in Arrhenius correlation. Kinetics of pathways (5) is controlled by a slow diffusion of CHO: radicals in polymer phases. Kinetics of pathway (6) shows the similar form but suffered from higher value of E d as compared to that of pathway (3) due to the diffusion of Pt 4+ and Pd 4+ ions in highly exible carbon matrix with substantial vibration amplitude in all directions. Reaction kinetics of pathway (7) is controlled by thermal reduction of polymer chelated metal ions. In this pathway, metal ion reduction can be described by a reaction of "aPt (L) 4+R + bPd (L) 4+D + cCHO: (L) + dPd (h,k,l)(S) 4 ePd 0 1Àx Pt 0 x @Pd (h,k,l)(S) "; where Pt 4+R , Pd 4+D , and Pd (h,k,l)(S) refer to the residual Pt 4+ and dissolved Pd 4+ ions aer nucleation reaction, and the molar volume of the solid material (i.e., number density of reaction sites (area) at Pd crystallites, [Pd (S) ]). Hereby, kinetics of NP growth with an approximation of reagent concentrations can be shown by eqn (5): where, k is the rate constant and x is the crystal growth rate dominating by diffusion kinetics of atoms at crystal surface. 14 Since the CHO: radical is decomposed from ethylene glycol molecules and the number density of solid phases (i.e., concentrations of "Pd 1Àx Pt x(S) " and "Pd (S) " crystallites) remain unchanged, these values can be constants that without affecting reaction kinetics. In the polyol reaction system, Pt 4+ ions possess a stronger chelation with pyrrolidone ligands (Pt 4+ -P) than that of Pd 4+ -P. Such feature is generally shown by formation of PVP-blended Pt NP with preferential growth facets. Therefore, Pt 4+ -P would tend to interact with CHO: to form Scheme 1 Schematic representation for atomic structure relocation in reaction pathways of experimental NCs upon crystal growth. In selfaligned wet chemical process, three steps correspond to (a) metal-ligand (carbon) complex formation, (b) core crystal nucleation, (c) sorption of Pt metal ion followed formation of atomic cluster in core crystal surface. In polyol process, three steps correspond to (a) PdPt nucleation (PdPt ni ) by success reactions of chelation of Pd and Pt ions (PVP-Pd or PVP-Pt) by pyrrolidone ligands and their interaction with CHO:, (b) heterogeneous crystal growth by subsequent sorption of PVP-Pt/PVP-Pd and reduction in PdPt ni surface, (c) atomic relaxation formation of PdPt crystal. Reaction pathways refer to (1) Pt 4+ diffusion, sorption, and interaction with H:/e À in Pt ni surface, (2) reduction of Pt 4+ or Pd 4+ ion by H:/e À followed by their accommodation in CNT surface, (3) metal/metal ion diffusion in CNT surface, where h c and 3 c denote the chemical resistances of defect sites parallel and perpendicular to metal/metal ion diffusion vectors, diffusion of C 2 H 6 O 2 : in (4) ethylene glycol and in (5) PVP molecules followed by reduction of PVP-Pt 4+ /Pd 4+ , (6) diffusion of metal/metal ions in PVP molecules where h p and 3 p denote the resistances of friction coefficient and chemical strength of pyrrolidone ligands parallel and perpendicular to metal/metal ion diffusion vectors in PVP molecule, (7) reduction of Pd 4+ /Pt 4+ into metal nuclei by thermal activated CHO:.
a core component in PdPt-p. The total growth rate of synthesized NP by polyol reduction depends on identity and concentrations of metal ion precursors (polymer stabilizer) and reaction temperature. A Pt based bimetallic crystal growth rate was estimated to be 60-80 Â 10 À12 m s À1 by in situ SAXS/WAXS characterization. It means that the growth of PdPt-p in $70Å takes $15 minutes to go. Surface chemical composition of experimental NPs is analysed by using CV analysis (Fig. 5a). For Pt-s, the two redox reaction peaks c (0.23 V vs. RHE) and c* (0.38 V vs. RHE) at a potential more negative to approximately 0.46 V vs. RHE (hydrogen underpotential deposition region, i.e., UPD H) refer to current responses of protons (H + ) adsorption at (111) and (200) facets of Pt metal during the negative sweep, and the coupling signals of H + desorption at peaks e (0.235 V vs. RHE) and e* (0.405 V vs. RHE) during the positive sweep, respectively. Accordingly, both adsorption and desorption proles clearly reveal the substantially weakened H + interaction on (200) facets, therefore proving the preferential grown of {111} facets for Pt/C. To clarify the extent of local alloying between Pt and Pd atoms in NP surface, CV curve of Pd/C is compared as reference. In this curve, H + adsorption peak during negative sweep at 0.17 V (vs. RHE) together with the absence of desorption peak in positive sweep show the typical characteristics of strong H affinity of Pd metal. Consequently, H + desorption peak should be located between peaks a and c for cases of NP with considerable extents of surface alloy between Pt and Pd atoms. For PdPt-s, preferential (111) growth is evidently shown by substantially enhanced (111) current peak (peak d) with the absence of (200) current. The broadened redox peaks (d at 0.238 V vs. RHE and d* at 0.415 V vs. RHE) indicate the various chemical states (i.e., a slight extent of local surface PdPt alloy and diffused interfaceted regions) on its surface. Given that the potential of H + desorption peak (peak d) is higher than that of Pt-s (peak c), surface local alloy between Pt and Pd atoms should be a minor factor in the electrochemical properties of PdPt-s. Therefore, complexity of chemical states can be rationalized by heteroatomic intermixing between atomic Pt clusters and Pd regions in NP surface; where sorption sites of H + including atop, bridge, and hollow sites in Pt clusters, Pd crystal, and on Pd-Pt interfaces. In the CV curve of PtPd-p, potential of H + sorption peak (peak b) is 0.2 V vs. RHE. It locates between those of Pd-s and Pts, therefore, indicating the formation of Pd-Pt alloy. The indistinct difference between current density of (111) and (200) facets in H UPD reveals the isotropic crystal growth at (111) and (200) facets. Therefore, by cross-referencing results of XRD and CV analyses, one can note that PdPt-s is grown with a metallic Pd core decorated by Pt clusters in NP surface. The shi of diffraction peaks A and B in opposite sides suggests that Pt atoms tend to alloy on opened (200) facets. In PdPt-p, homogeneous Pd-Pt alloy occurs with a structure of Pt clusters on Pd crystal NP. Such a structure is revealed by presence of homoatomic Pt-Pt bond pairs in EXAFS and upshi of A and B peaks from that of metallic Pd (111) and (200) in diffraction pattern of PtPd-p. Comparison between surface to bulk ratio and extent of oxidation further reveals location of Pt atoms in NPs; where correlations of spherical NPs can be estimated by geometric calculations and are discussed as simplied interpretations. As shown in Fig. S4, † surface to bulk ratios (h) are 24.7% for PdPt-s (4.2 nm), 16.1% for Pt-s (5.5 nm), and 20.1% for PdPd-p (7.3 nm). Considering that all PdPt NPs are grown in homogeneous alloy, the oxidation extent of Pt (determined by tting corresponding XANES spectra using linear combination tting method) should be determined to a similar value to their h (PdPt-p). A slightly lower Pt oxidation to h (also found in Pt-s) could be attributed to the protection of Pt from oxidation by carbon support at interface. However, extent of Pt oxidation is determined to be 38.2% for PdPt-s. This value is far higher than that is expected in homogeneous alloy and can only be rationalized by exposing Pt clusters to NP surface (i.e., decoration of Pt atoms in NP surface). Fig. 5b and Table 3 show the linear sweep voltammetry (LSV) spectra and corresponding parameters of Pt-s, PdPt-s, and PdPt-p NPs measured in a O 2 -saturated KOH electrolyte (0.1 M) on a rotational working electrode at 1600 rpm. As shown, the onset potential (V oc vs. RHE) of NP follows the trend: PdPt-s (c ¼ 0.941 V) > Pt/C (b ¼ 0.935 V) > PdPt-p (a ¼ 0.928 V) and is inversely proportional to their Pd-Pt intermix "c(Pd-Pt) at Pt L 3 -edge" ( Table 2). Moreover, their mass activity (MA, current densities vs. Pt metal loading) possesses the same sequence at 0.85 V (Table 3). With a 4-electron pathway (n $ 4.0 AE 0.05), results mentioned above demonstrate the facilitation of oxygen reduction by a superior reaction kinetic of PdPt-s (J k ¼ 20.08 mA cm À2 at 0.85 V vs. RHE) among NPs under investigation. Such a promotion on reaction kinetics can be attributed to a substantial electron relocation to Pt sites with a local distortion in their neighbouring  and c(Pt-Pd)) and local distortion comparing to that of PdPt-s. Consequently, a substantially reduced MA to 734.0 mA mg Pt À1 is expectable due to the presence of low density Pt active sites (also indicated by a reduction of n to 3.7-3.8) and slight local distortion between Pt and Pd atoms in NP surface exposed to reaction media. For clarifying the effects of Pd-Pt intermix and surface conguration to NP ORR activity, electrochemical and structural characterization results of Pt-s are compared as reference. As clearly indicated in Table 3, J k and MA of Pt-s are improved by 3-4 times as compared to that of J.M.-Pt. In Pt-s, the lower extent of local lattice distortion (Tables 1 and 2) and absence of Pd-Pt intermix explains the reduction of J k and the substantially suppression of MA compared with that of PdPt-s. Electrochemical durability of NPs in ORR is illustrated by results of ADT (Fig. 5c) complementary with CV analysis (Fig. S5 †) at corresponding test cycles. Retention current (J R ) vs. initial current (J 0 ) of NPs (J R vs. J 0 ) with increasing ADT cycles is shown in Fig. 5c. As revealed, J R vs. J 0 is progressively decreased by 14.0% for PdPt-s at 30000 cycles, 31.5% for PdPtp at 5000 cycles, and 24.0% for Pt-s at 12 000 cycles (Table 3). Those results show that PdPt-s has the highest structural and electrochemical durability among experimental NPs. It could be accounted to a structural reinforcement by high ratio of heteroatomic contact. 15 Current uctuation of PdPt-s (denoted by black arrows) can be attributed to simultaneous dissolution and re-deposition of Pt/Pd atoms in NP during potential sweeping cycles. Such a statement is revealed by uctuation of peak potential at $0.727 to 0.750 V vs. RHE (shown by red solid arrow) with uneven changes of redox peak intensity as affected by the changes of chemisorption states for oxygen adsorption in PtPd-s surface. For Pt-s, intensity of CV peaks is progressively faded away without signicant changes in shape with ADT cycles. The fading of CV prole suggests a dissolution induced atomic restructure between facets of Pt NP. Considering that Pd is a noble metal, total MA (MA total ) is compared to verify the industrial value of experimental NPs. As shown in Table 3, MA total of NPs follows the trend of PdPt-s (329.9 mA mg À1 ) > Pt-s (271.9 mA mg À1 ) > PdPt-p (118.1 mA mg À1 ) > J.M.-Pt (67.1 mA mg À1 ). Compared to Pts, a substantially reduced MA total of PdPt-p can be rationalized by the surface coverage of high oxygen/metal affinity ligands (pyrrolidone in our case). Such a capping layer hinders the diffusion kinetics of oxygen molecules and increases the interface resistance for charge exchange in NP surface. Both the two factors reduce the J k and V oc as compared to that of Pts. Meanwhile, pyrrolidone ligand possesses a high metal chelation affinity. It forms a metastable metal-complex and thus reduces the re-deposition rate of dissolved metal ions in NP surface during potential cycles. Such characteristics explain the highest J R degradation among experimental NPs in ADT. Charge transfer number (n) is another important index for NPs in ORR application. For cases of NPs with n $ 4.0 (i.e., complete charge transition), all electrons are contributed from the reduction of chemisorption oxygen in surface reaction sites. In those cases, current degradation is caused by reduction of surface reaction site which is because of inter-particle sintering during ADT. For Pd-s, n is determined to be 5.5. This value is substantially higher than 4.0 indicating the presence of exceed electrons from Pd oxidation. For PdPt-p, n is determined to be 3.8 implying the formation of intermediate hydrogen peroxide. Such a feature could be attributed to the strong chelation effect of PVP-ligands that shielding (suppressing) the surface Pt sites from ORR.

Conclusions
A robust method on growing bimetallic PdPt NP with high extent of surface alloy, high density of Pt cluster on Pd surface, and low Pt metal loading ($10 wt%) is developed based on selfaligned wet chemical reduction method with precise control on reaction sequences and time. Structural characterization results combining heterogeneous nucleation crystal growth theory suggest the nanostructure of bimetallic PdPt NP is controlled by reaction time and precursor inlet sequence. We demonstrate that such PdPt NP performs a 29.4-times the value of MA as compared to that of commercial Pt catalyst. The structural and electrochemical characterizations show that such impressive ORR performances are attributed to the local electron relocation and atomic structure distortion in the presence of high activity tr-Pt-atom clusters capping atop Pd NP surface. In such a unique surface conguration, Pt clusters with high Pd intermix perform high oxygen splitting kinetics and the neighbouring edge sharing Pd CN with high local distortion provide low activation energy open sites for conducting subsequent oxygen radical reduction steps. By combining the two factors, PdPt-s gains an impressive MA of 2040.6 mA mg Pt À1 .

Conflicts of interest
There are no conicts to declare.