Interactions between Low-Loading Pd Nanoparticles and Surface N-Functionalities and Their Effects on HCOOH Oxidation

Palladium (Pd)-based anode catalysts show greater activities than Pt catalysts for HCOOH oxidation. However, because of the ease of Pd oxidation, the electrocatalytic stability of Pd is not satisfactory. In the present study, 2.4-wt% Pd supported on N-functionalized carbon nanotubes (Pd/NCNTs) with high dispersion areas and narrow size distributions exhibited superior HCOOH-oxidation performance in comparison to Pd on charcoal (commercial Pd/C, 30-wt% Pd) and Pd on HNO3-treated CNTs (Pd/OCNTs). Surface characterization revealed that the interactions between Pd nanoparticles (PdNPs) and surface N-functionalities optimized local structures of the PdNPs and participated in the HCOOH oxidation by influencing the electronic properties of the PdNPs. © The Author(s) 2015. Published by ECS. This is an open access article distributed under the terms of the Creative Commons Attribution 4.0 License (CC BY, http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse of the work in any medium, provided the original work is properly cited. [DOI: 10.1149/2.0691512jes] All rights reserved.

Direct-formic-acid fuel cells (DFAFCs) have recently attracted attention because of their many advantages when compared to directmethanol fuel cells. [1][2][3] Despite high activities in the initial stage, dehydration may occur and produce CO as a reaction intermediate when noble metals are used as the anode material. 4,5 Resultant CO and carbonaceous deposits can be adsorbed on the metal surface and can inhibit further oxidation. 6,7 In order to avoid CO generation and to improve the electrocatalytic stability, surfaces of supported metal particles must be electronically and geometrically modified.
Surface and subsurface electronic properties of noble metals are responsible for the ease of adsorption and dissociation of reactant molecules during reactions. The electronic properties of a metal are determined by its local atomic structure. Nanosizing a metal decreases the coordination number of its surface atoms (creating steps, kinks, and corners), shrinks its dband, and shifts its center toward the Fermi level. [8][9][10] Moreover, the metal configuration can be modified via bonding with neighboring elements to form alloys, intermetallics, and strong metal-support interactions. [11][12][13] Functionalities on supporting materials can participate in catalysis by influencing the electronic structures of loaded metal particles and thereby altering the adsorbing strength toward reactive intermediates. In HCOOH oxidation, metal nanocrystals with optimal size, shape, and crystallinity may influence the gas diffusion, mass activity, and resistance to CO poisoning. 14 Nitrogen functionalities have been reported to influence electrocatalytic oxidation by improving the support conductivity and coupling with supported metal particles during reactions. 15 Compared to oxygen functionalities, however, the role of interactions between metal nanoparticles and N-functionalities in HCOOH oxidation has not been discussed in as much detail. Moreover, surface catalysis raises the question of whether 20-30 wt% loadings of noble metals are always necessary for meaningful HCOOHoxidation conversions and mechanism studies.
Herein, functionalization was first carried out on carbon nanotubes (CNTs) via oxidation in HNO 3 . Gaseous NH 3 was then flowed over oxidized CNTs (OCNTs) in a tubular quartz reactor at 600 • C for 4 h to introduce nitrogen-containing functional groups on the CNTs (NC-NTs). Using these functionalities as anchoring sites, palladium (Pd) ions were added to a solution of the functionalized CNTs for coordination. An appropriate hydrolysis treatment was applied to the impregnated materials to crystallize the precursors to achieve a 2.4 wt% loading [measured via inductively coupled plasma (ICP)] of Pd supported on the NCNTs (Pd/NCNTs) with a high dispersion scale. A reference sample of 3.2 wt% (ICP-measured) Pd supported on the z E-mail: lidongshao@outlook.com OCNTs (Pd/OCNTs) was prepared for comparison using the same impregnation method. Figure 1a shows the dispersion of Pd nanoparticles (PdNPs) as small as 1.42 nm, supported on the NCNTs. The inset shows the size distribution of the PdNPs. Scanning transmission electron microscopy (STEM)-energy-dispersive X-ray (EDX) analyses and elemental maps of individual PdNPs in line with nitrogen and oxygen distributions on the NCNTs indicated a homogeneous composition of PdNPs on the N-functionalized CNTs. Figure 1b shows an STEM image of dispersed PdNPs on the NCNTs with a higher magnification than that of Figure 1a. Figure 1c displays a high-resolution transmission electron microscopy (HRTEM) image of a metallic Pd particle with distinct edges. The dispersion profiles of the PdNPs on the Pd/OCNTs and the commercial Pd on charcoal (Pd/C,30 wt% Pd) are shown in the supporting materials. Pd had a similar size distribution on the OCNTs as it did when supported on the NCNTs, while clear aggregations were visible on the Pd/C with low dispersion areas (Figures S1 and S2).
X-ray photoelectron spectroscopy (XPS) was applied to study the surface and subsurface regions of the Pd/NCNTs. The N 1s spectrum in Figure 2a (N content: 4.54%) could be deconvoluted into the following four peaks: N1 (398.7 eV), N2 (400.1 eV), N3 (401.2 eV), and N4 (403.3 eV), which were assigned to pyridinic nitrogen, amino nitrogen, pyrrolic nitrogen, and graphitic or quaternary nitrogen, respectively. 16,17 (The N-functionalities have been discussed previously as anchoring sites for embedding metal nanoparticles. [18][19][20] Oxygen species were represented by peaks at binding energies around 531.1, 532.2, 533.3, 534.3, and 536 eV. Figure S3 shows the O1s profile of the Pd/NCNTs. Compared to the O1s profile of the Pd/OCNTs in Figure S4 (O content: 13.28%), the oxygen functionalities on the Pd/NCNTs (O content: 9.06%) were clearly reduced due to the significant reduction effect during ammonolysis. Figure 2b shows the fitted Pd3d core-level spectra for the Pd/NCNTs. The binding energy (BE) of monometallic Pd (Pd 0 ) was located at 336.0 eV. As for the peak profile at 337.6 eV in Figure 2b, this result was reportedly due to the formation of a Pd-O or Pd-N bond. 21,22 Owing to the partial replacement of O-functionalities with N-functionalities on the Pd/NCNTs, the preserved peak profile at 337.6 eV indicated the formation of Pd-N bonds in the present Pd/NCNTs. The peak contribution at 338.5 eV was attributed to either raised, divalent, polynuclear Pd-hydroxo complexes that resulted from the incompletely reduced Pd-salt precursor or dissolved and reprecipitated species in the used samples. 23 Moreover, all the peaks of the Pd/NCNTs were 0.3-0.6 eV more positive than those of the Pd/OCNTs (Figure 2c) and the Pd/C (Figure 2d). These findings resulted from a decrease in the 3d-electron density of Pd, indicating a strong interaction between the support and the PdNPs.   Table I, the anodic-peak current density of the formic-acid oxidation on the Pd/NCNTs electrocatalyst was 550.50 A/g at the peak potential, which was 1.70 and 2.60 times greater than the values measured for the Pd/OCNTs and Pd/C, respectively. The electrocatalytic specific activity of the Pd/NCNTs ( Figure S5) was better (7.55 A/m 2 ) than those of the Pd/OCNTs (5.79 A/m 2 ) and the Pd/C (6.41 A/m 2 ), indicating the improved catalytic activity of the Pd/NCNTs toward the HCOOH oxidation. To further evaluate the stability of the catalysts, chronoamperometry tests were conducted in 0.5 M H 2 SO 4 and 0.5 M HCOOH at 0.1 V for 3600s. As shown in Figure 3b, in the first 1000s, the mass activity of the Pd/NCNTs was 76.86 A/g, while the mass activities of the Pd/OCNTs and Pd/C were both around 35.7 A/g. After another 1000s, the mass activity of the Pd/NCNTs was 45.60 A/g, which was still a greater value than those of the Pd/OCNTs and the Pd/C.
Pd-based electrocatalysts suffer from poor durability for formicacid oxidation in DFAFCs, although they demonstrate greater catalytic activity. Accordingly, the stability of the catalysts was further evaluated via continuously scanning the CV curves to follow the changes in the anodic-peak current density as a function of the cycle number, as shown in Figure 4. Figure 4d shows the changes in the anodic current density for formic-acid oxidation on the three catalysts with 50 cycle numbers. As can be seen in this figure, the Pd/NCNTs catalyst showed better stability than either the Pd/OCNTs or the commercial Pd/C. CO-stripping measurements are normally used to estimate the EASA of Pd-based catalysts. For this purpose, the specific EASA was calculated according to the following equation: where Q is the charge of the CO-desorption electrooxidation, m is the loaded amount of Pd, and C (420 μm/cm 2 ) is the charge required for the adsorption of a CO monolayer. Figure S6 shows the CO-stripping measurements of the Pd/NCNTs, Pd/OCNTs, and Pd/C before and after 50 CV cycles. The initial EASA values were 55.96 m 2 /g for the Pd/OCNTs, 32.76 m 2 /g for the Pd/C, and 73.15 m 2 /g for the Pd/NCNTs ( Figure S6). After 50 continuous CV cycles, the EASA value of the Pd/NCNTs remained about 34.99% of the original EASA value, while those of the Pd/OCNTs and Pd/C were 27.19% and 22.40%, respectively. Table I lists the main electrocatalytic properties of the Pd/NCNTs, Pd/OCNTs, and Pd/C. These results again indicated the enhanced activity and durability of the Pd/NCNTs. Ammonolysis treatment modified both the structural and surface aspects of the OCNTs. High-temperature annealing in a reducing atmosphere upgraded the graphitic degree by healing the previously created defects on the OCNT walls. In turn, the conductivity of the carbon networks was enhanced, and the O-functionalities were reduced. Meanwhile, N-functionalities were introduced on the CNT surfaces. Previously, HCOOH reactivity in electrocatalysis has been correlated with the particle-size effect. 24,25 It has been reported that the bond strength of the COOH-intermediate adsorption on the nanosized Pd surfaces could be weakened, resulting in an increase in the rate of formic acid oxidation to CO 2 (and in the associated oxidation current). 24 In the present work, the high density of surface functionalities revealed via XPS gave rise to a strong anchoring effect toward salt precursors. After metal ions coordinated with oxygen and nitrogen functional groups, hydrolysis was applied to crystallize the precursors nucleated on the oxidized carbon supports to form nanocrystals. The temperature and reduction time could tune the size, shape, and crystallinity of the nanocrystals. Therefore, we concluded that the PdNPs  with non-equilibrium shapes and a large number of low-coordination sites (shown in Figure 1) were crucial for HCOOH oxidation with high activity.
In electrocatalysis, high activities have been reported to appear in association with low stabilities in HCOOH oxidation due to the high bond strength of the COOH-intermediate adsorption and the increased oxidation rate to CO. When we considered the metallic Pd 0 as the only active phase, it should have been quickly deactivated due to its high activity, as was the case for Pd/C. In contrast, we observed high activity with improved stability on the Pd/NCNTs. When we compared the Pd/NCNTs to the Pd/OCNTs via XPS analyses, the phases of the Pd 0 and the Pd-salt precursors on the Pd/NCNTs were similar to those of the Pd/OCNTs. Moreover, the size distribution and structural information of the PdNPs on the NCNTs were similar to those of the PdNPs on the OCNTs. Therefore, the XPS observation of the Pd-N bonds at 337.6 eV and the shift of the peaks on the Pd/NCNTs might have explained the obtained catalytic differences that the bonded PdNPs on the NCNTs had been a part of in the active    Experimental CPR24-ps carbon nanotubes were purchased from Pyrograf Products Inc. (OH, USA). An FEI aberration-corrected Titan 80-300 transmission electron microscope was employed to conduct structural and chemical investigations in both the TEM and STEM modes. EDXspectrum elemental maps were acquired in the STEM mode. XPS analyses were performed on a Thermo ESCALAB 250 instrument using monochromatic Al K α radiation with 1486.6 eV, operating at 150 W. The Pd content of each sample was measured with an inductively coupled plasma optical emission spectrometer (ICP-OES, PerkinElmer Optima 8000). For detailed experimental information, please see the supporting material.

Conclusions
Palladium with a 2.4 wt% loading was supported on NCNTs with large dispersion areas and narrow particle-size distributions. The optimal sample configuration, in turn, provided us opportunities to determine the active phases during the HCOOH oxidation due to its improved homogeneity. Conventional high loadings with large structural varieties may compromise applied characterization approaches and hinder reaction-pathway studies. Sample handling in terms of functionalization, impregnation, and reduction is crucial to establish the metal-support interactions to bridge the gap between catalyst configuration and catalytic performance. In the present work, N-functionalized CNTs not only provided a platform for anchoring PdNPs with optimal structures during sample preparation but also participated in the catalysis by influencing the electronic properties and stabilities due to the interaction between the functionalities and the PdNPs.