Size-Dependent Hydrogen Oxidation and Evolution Activities on Supported Palladium Nanoparticles in Acid and Base

Carbon-supported Pd nanoparticles with varying particle sizes were prepared by annealing the commercial 20 wt% Pd on Vulcan XC-72 (Pd/C, Premetek Co.) at different temperatures in an Ar/H₂ atmosphere in a tube furnace: the quartz tube was first purged with Ar for 30 min at room temperature to remove air. Furnace temperature was then raised to 100 °C at a rate of 10 °C/min in Ar and maintained at 100 °C for 30 min to remove adsorbed H₂O on the catalyst. Finally, H₂ was introduced into Ar (5 vol. % H₂) and the temperature was increased to 300 °C, 400 °C, 500 °C and 600 °C at a rate of 10 °C/min and maintained for 2 h.¹⁴ The obtained catalysts are denoted as Pd/C-300C, Pd/C-400C, Pd/C-500C and Pd/C-600C, respectively. Diameters of Pd nanoparticles were measured from transmission electron microscopy (TEM) images obtained on a JEOL JEM-2010F TEM. X-ray diffraction (XRD) patterns of Pd samples were measured using a Philips X'Pert X-ray diffractometer with Cu Kα radiation.

The electrochemical measurements were conducted in a three-electrode glass cell with a double-junction silver/silver chloride (Ag/AgCl) electrode as the reference electrode, a Pt wire as the counter electrode and a glassy carbon (5 mm diameter, PINE Ins.) as the working electrode. The ink dispersions of Pd/C samples were prepared by dispersing Pd/C (20 wt%) in 0.05 wt% Nafion isopropanol solution to a final concentration of 2 mg_Pd/C/mL with ultrasonication for 1 h. The thin-film electrodes were prepared by pipetting 2.5 μL of the ink (2 mg_Pd/C/mL) once, four times or six times to achieve a final loading of 5, 20 or 30 μg_Pd/cm²_disk. The Pd loading on the electrodes were 20 μg_Pd/cm²_disk for all the five Pd samples measured in 0.1 M KOH prepared from KOH tablet (85 wt%, 99.99% metal trace, Sigma Aldrich). For the samples measured in 0.1 M HClO₄ prepared by diluting 70 wt% HClO₄ (EMD) with DI water, the final loadings for Pd/C, Pd/C-300C, Pd/C-400C were lowered to 5 μg_Pd/cm²_disk so that the HOR/HER polarization curves will not overlap with the concentration overpotential curve, and those for Pd/C-500C and Pd/C-600C were 20 μg_Pd/cm²_disk and 30 μg_Pd/cm²_disk to maintain roughness factors greater than 1 due to their low electrochemical active surface areas (ECSAs).

Cyclic voltammetry experiments were performed in both Ar-saturated 0.1 M KOH and 0.1 M HClO₄ electrolytes at a scanning rate of 50 mV/s in the potential range of 0.07 to 1.25 V vs. reversible hydrogen electrode (RHE). All the potentials reported were converted to the RHE scale. The electrochemical surface areas of all the Pd/C samples were determined from PdO reduction peak from cyclic voltammograms (CVs) based on a charge density of 424 μC/cm²_Pd.^23,24

Activity measurements for hydrogen oxidation and evolution reactions (HOR/HER) on Pd catalysts were carried out in H₂-saturated 0.1 M KOH or 0.1 M HClO₄ at a scanning rate of 50 mV/s and a rotating speed of 1600 rpm by cycling the potential several times until a stable polarization curve was obtained. The scanning rate was then switched to 1 mV/ to minimize the contribution of capacitance current, and the positive scan of the first cycle at 1 mV/s was reported as the HOR/HER polarization curve. The internal resistance was determined by electrochemical impedance spectroscopy (EIS) measured from 300 kHz to 100 mHz at open circuit voltage right after the HOR/HER measurement, which was then used to correct the measured potential to iR-free potential based on the following equation,

where E is the measured potential, i is the corresponding current, R is the internal resistance, and E_{iR − free} is the internal-resistance free potential. The internal resistances in 0.1 M KOH and 0.1 M HClO₄ were determined to be about 40 and 20 Ω for, respectively.

The Pd nanoparticle size increases as the annealing temperature rises: the number-averaged particle diameter (d_n^TEM) calculated by grows from 3.2 nm for Pd/C to 33.6 nm for Pd/C-600C as shown by the TEM images and the corresponding histograms (Figure 1). The increase of standard deviation of d_n^TEM as particle size becomes larger (Figure 1f and Table I) is caused by the thermal annealing method, which is consistent with literature.^22,25 The volume/area averaged particle diameters (d_v/a^TEM) calculated by are larger than d_n^TEM (Table I). The d_v/a^TEM is considered a better measure of average particle diameter than d_n^TEM for calculating the specific surface area (S_v/a^TEM)²⁶ (Table I), according to

where ρ is the density of the metal (12.02 g/cm³ for Pd), d is the particle diameter in nm, S_v/a^TEM is in unit of m²/g.

Zoom In Zoom Out Reset image size

X-ray diffraction (XRD) patterns of all samples (Figure 2) exhibit diffraction peaks at 40.1°, 46.6°, 68.1° and 82.1° corresponding to (111), (200), (220) and (311) facets of Pd (JCPDS card no. 46–1043), respectively. The width of diffraction peaks width becomes narrower as the annealing temperature increases, suggesting an increase in the particle size, which is consistent with the TEM observations. The particle diameter determined from XRD, denoted as the volume-averaged particle diameter (d_v^XRD),²⁶ can be calculated using the Scherrer equation,

where λ is the wavelength of the X-ray wavelength (1.54 Å), △(2θ) is the full width of the half maximum (FWHM) of the diffraction peak and θ is the Bragg angle. Pd(111) peak was used in the particle size calculations. The d_v/a^TEM and d_v^XRD agree well for Pd/C, Pd/C-300C, Pd/C-400C and Pd/C-500C. However, the average particle diameter determined by TEM is significantly larger than that determined by XRD for Pd/C-600C (Table I), which can be caused by: 1) the width of XRD patterns reflects the primary crystalline size, while particles in TEM images could be polycrystalline, i.e., agglomerates of several primary crystals. The larger the particles are, the more likely the particles are polycrystalline. 2) The particle size distribution of Pd/C-600C is broad (Figure 1f), which could introduce errors in calculating the average particle size with by counting a finite number of particles. Specific surface areas of different Pd samples calculated from d_v^XRD using Eq. 2, denoted as S_v^XRD, are comparable with S_v/a^TEM (Table I).

Zoom In Zoom Out Reset image size

Specific surface areas of the supported Pd catalysts can also be measured from cyclic voltammograms (Figure 3). The lower limit was chosen to be 0.07 V to avoid the diffusion of H atom into the Pd lattice.¹¹ In 0.1 M KOH, the underpotential deposited hydrogen (H_upd) adsorption and desorption peaks in the potential region of 0.07–0.45 V are very asymmetric and the double layer regions are not well defined (Figure 3a), making it difficult to determine surface areas using the H_upd adsorption and desorption peaks. Additionally, the H_upd desorption peaks for all the samples in 0.1 M KOH are located at around 0.35 V (Figure 3a). By contrast, the H_upd adsorption and desorption peaks are more symmetric in 0.1 M HClO₄, with a well-defined H_upd desorption peak at 0.26 V and a shoulder at around 0.19 V (Figure 3b), both of which are at a lower potential than the desorption peak in base (0.35 V), and could represent different facets of Pd. When the current is normalized to the peak current after the double layer current deduction, the height of the H_upd desorption peak at about 0.26 V is fixed while that of the shoulder at about 0.19 V increases with increases of particle size (Figure 3c). The change of line shape of the H_upd desorption peak suggests a redistribution of facets in different Pd samples, which is consistent with a previous report on Pd catalysts.¹⁵ The PdO reduction peak shifts to a more positive value as Pd particle size increases (from Pd/C to Pd/C-600C sample) in both KOH and HClO₄ (Figures 3a and 3b), indicating a weaker Pd-O binding as the Pd particle size increases. In this study, we use integrated area of the PdO reduction peak to determine the electrochemical surface area (ECSA) of supported Pd catalysts. ECSAs calculated from CV conducted in KOH (S_Base^CV) and HClO₄ (S_Acid^CV) are comparable (Table I), in agreement with Henning et al.'s earlier observations.¹³ Consistent with the surface area determined by TEM and XRD, ECSA decreases with increasing particle size (Table I), however, ECSAs are roughly a factor of two smaller than S_v/a^TEM and S_v^XRD (Table I). One potential cause for this discrepancy is that a fraction of every particle is in contact with the support and thus is not electrochemically accessible. Alternative methods such as H_upd adsorption/desorption peaks and CO-stripping can also be used to determine surface area, all of which have advantages and drawbacks.¹³ Although the ECSAs derived from PdO reduction might not be the most accurate in absolute terms, the general trend is consistent with values obtained with other methods. Thus, it is still meaningful to compare activities normalized by ECSAs obtained with PdO reduction.

Zoom In Zoom Out Reset image size

The HOR/HER polarization curve on Pd/C measured in 0.1 M KOH before (black line) and after (red line) iR correction deviate significantly from the concentration overpotential curve (grey-dash line) and reach the HOR limiting current at overpotentials above 0.3 V (Figure 4a). The concentration overpotential curve can be calculated using the following equation,

where I_d is the diffusion limited current, I_l is the maximum HOR limiting current, F is the Faraday constant, η is overpotential, R is the gas constant, and T is temperature in Kelvin. The HOR/HER polarization curve on Pd/C measured in 0.1 M HClO₄ reaches the HOR limiting current at overpotential as low as 0.07 V and is close to the concentration overpotential curve (Figure 4b), indicating faster HOR/HER kinetics in acid than in base. The polarization curve after iR-correction (red line in Figure 4b) still show deviation from the concentration overpotential curve, which enables the quantification of kinetic currents. The limiting current densities (i_l) - limiting current (I_l) normalized by the geometric area of the electrode (A) - in both 0.1 M KOH and 0.1 M HClO₄ increase as the rotation speed increases from 400 to 2500 rpm (Figures 4c and 4d), which are proportional to the square root of the rotation speed (ω^−1/2) according to the Levich equation i_l = 0.62nFD^2/3ν^{− 1/6}c₀ω^1/2 = Bc₀ω^1/2, where n is the number of electron transferred, D is the diffusivity of H₂ in the electrolyte, ν is the kinematic viscosity of the electrolyte, c₀ is the solubility of H₂ in the electrolyte. The Koutecky-Levich plots in both base and acid, which show the inverse of limiting current density (i_l) as a function of the inverse of ω^1/2, can be fitted into straight lines, yielding Bc₀ values of 0.070 and 0.071 mA/(cm²_diskrpm^1/2), in reasonable agreement with the calculated Bc₀ value (0.068 mA/(cm²_diskrpm^1/2)).

Zoom In Zoom Out Reset image size

The HOR/HER polarization curves were measured for all the Pd samples in H₂-saturated 0.1 M KOH and 0.1 M HClO₄ at 1600 rpm and 1 mV/s (Figures 5a and 5b). The HOR limiting current densities for all the polarization curves in base are about 2.9 mA/cm²_disk (Figure 5a) which agrees with the theoretical values within experimental errors. Similar HOR limiting currents of about 2.9 mA/cm²_disk were observed in acid as in base (Figure 5b). Atomic hydrogen is known to be able to diffuse into the Pd lattice (H_bulk), therefore the HOR/HER process might be accompanied by hydrogen absorption/desorption reaction (Eq. 5), which could appear as an extra oxidation peak in the polarization curve.^13,27

Smaller or no extra peaks associated with oxidation of absorbed hydrogen were observed in the polarization curves in 0.1 M KOH (Figure 5a), which might be attributed to the slow HOR reaction and the slow scanning rate used in the measurement to allow the absorbed hydrogen to diffuse back to the surface to be reacted as H_upd. In 0.1 M HClO₄, there indeed exists absorbed hydrogen oxidation peaks in the polarization curves for Pd/C-300C, Pd/C-400C, Pd/C-500C and Pd/C-600C at about 0.069 V, 0.079 V 0.086 V and 0.134 V, respectively (Figure 5b).

Zoom In Zoom Out Reset image size

To determine the exchange current densities of HOR/HER, the kinetic currents (I_k) were first calculated according to the Koutecky-Levich (Eq. 6),^20,22

where I is the measured current and I_d is the diffusion limited current defined in Eq. 4. The kinetic currents are then normalized by their corresponding Pd surface area (which are referred to as specific kinetic current densities, i_k) and plotted as a function of overpotential. A linear relationship was observed between i_k and η in the micropolarization regions (Figures 5c and 5d) where i_k and η can be fitted into the linearized Butler-Volmer equation with the assumption that the summation of anodic and cathodic transfer coefficient equals to 1 (α_a + α_c = 1) (Eq. 7) to obtain the exchange current density (i₀).

Alternatively, the kinetic currents can also be fitted into the Butler-Volmer equation (Eq. 8) to get the exchange current densities and transfer coefficients (Figures 5e and 5f).

The specific exchange current density determined using linear fitting for Pd/C in 0.1 M KOH at 293 K is 0.052 ± 0.002 mA/cm²_Pd (Table II), which is higher than the value obtained on Pd/C reported (0.06 ± 0.02 mA/cm²_Pd at 313 K corresponding to 0.02 mA/cm²_Pd at 293 K).¹ The Butler-Volmer fitting with α_a + α_c = 1 yields a very similar exchange current density value of 0.051 ± 0.002 mA/cm²_Pd and a α_a value of 0.54 ± 0.02 (Table III) which corresponds to an anodic Tafel slope about 108 mV/dec. This suggests that the Volmer-step (H_ad↔H⁺ + e) is the rate-determining step, in agreement with Durst et al.'s report.¹ Without constraining α_a + α_c = 1, the best fitting into the Butler-Volmer equation generates α_a value of 0.45 ± 0.01 and α_c value of 0.38 ± 0.02 (Table III), and a slightly higher exchange current density of 0.060 ± 0.003 mA/cm²_Pd. The HOR/HER activity on Pd/C is much higher in 0.1 M HClO₄ than in 0.1 M KOH – the specific exchange current densities determined from linear fitting and Butler-Volmer fitting with α_a + α_c = 1 on Pd/C are 2.56 ± 0.12 and 2.70 ± 0.04 mA/cm²_Pd respectively, which represent a 50-fold increase of that in 0.1 M KOH. The exchange current density determined using a H₂-pump in a proton exchange membrane fuel cell configuration (which resembles that in electrolyte with pH = 0) is 5.2 ± 1.2 mA/cm²_Pd at 313 K,¹ which can be converted to 2.3 ± 0.5 mA/cm²_Pd at 293 K according to Arrhenius equation with an activation energy of 31 kJ/mol.²⁷ The excellent agreement in the exchange current densities measured using RDE in this work and H₂-pump indicates the validity of RDE method to characterize HOR/HER activity on Pd/C in acid.

The specific exchange current density of HOR/HER increases as the Pd particle size increases from 3 to 19 nm and then levels off afterward in both acid and base (Table II and Figure 6a). The specific exchange current density in base (from linear fitting) reaches about 0.12 mA/cm²_Pd as the particle size exceeds 20 nm. Sheng et al. reported an exchange current density value of 0.13 mA/cm²_Pd on polycrystalline bulk Pd disk electrode in 0.1 M KOH by extrapolating from the Tafel plot in the HER region to equilibrium potential (0 V).¹¹ Alia et al. obtained an exchange current density of 0.18 mA/cm²_Pd by fitting the kinetic current from HOR/HER polarization curve into the Butler-Volmer equation.¹² The specific exchange current density on Pd nanoparticles larger than 20 nm approaches that of a bulk Pd electrode. Specific exchange current densities determined from the following three methods show good agreement: 1) linear fitting (black squares in Figure 6a); 2) Butler-Volmer fitting with α_a + α_c = 1 (red circles in Figure 6a); and 3) Butler-Volmer fitting with unconstrained α_a + α_c (blue triangles in Figure 6a). The mass exchange current density shows a general decreasing trend with the growing particle size (Figure 6b), due to the decrease of ECSA.

Zoom In Zoom Out Reset image size

Transfer coefficient (α) and Tafel slope (TS) are related according to TS = 2.303 RT/αF, which are important parameters to reveal the reaction mechanism. For HOR/HER on different-sized Pd particles in 0.1 M KOH, the α_a obtained from the Butler-Volmer fitting decreases from 0.54 on Pd/C (3 nm) to 0.45 on Pd/C-600C (42 nm) when constraining α_a + α_c = 1 (Table III), which corresponds to a change of anodic TS from 108 to 129 mV/dec. However, the R² value for the regression is relatively low (0.991). By contrast, a much better Butler-Volmer fitting can be achieved by removing the constraining of α_a + α_c = 1 (R² = 0.999). Still though the α_a decreases with increase of Pd particle size (from 0.45 to 0.32) which translates to a TS change from 129 to 182 mV/dec, while the α_c's are about 0.38 (TS = 153 mV/dec) and barely change with particle size (Table III). In 0.1 M HClO₄, the α_a value ranges from 0.57 to 0.71 when the fitting was conducted with α_a + α_c = 1 (Table III). Only currents in the narrow potential regions (– 20 mV to 20 mV) can be used for data analysis in acid, thus introducing a large degree of uncertainty in the curve fitting. As a result, we did not calculate α_a and α_c without constraining α_a + α_c = 1. Sheng et al. reported a TS value of 210 mV/dec (or α_c = 0.28) for HER on bulk Pd in 0.1 M KOH at 293 K,¹¹ and Durst et al. reported an α_a value of 0.43 ± 0.07 in 0.1 M NaOH and 0.35 ± 0.1 in H₂-pump with a PEMFC configuration at 313 K.¹ The discrepancy in the transfer coefficient values between our work and the literature might be resulted from the possible diffusion of atomic hydrogen into the Pd lattice, which introduces uncertainty in the transfer coefficient. To mitigate the H absorption into Pd, Henning et al. deposited Pd on polycrystalline Au electrode and measured their HOR/HER activity in 0.1 M NaOH to obtain Pd surface coverage - dependent TS: surfaces with low Pd coverage yield HOR TS of about 240 mV/dec while surface with high Pd coverage generates a HOR TS of 150 mV/dec.¹³ In 0.1 M HClO₄, Schmidt et al. obtained a HOR TS of about 210 mV/dec on Pd/Au(111) at 293 K.²⁸

The particle size effect of HOR/HER activity on Pd/C could be attributed to the redistribution of surface facets among different-sized Pd nanoparticles (known as geometric effect), as revealed by the change in peak ratios in the H_upd desorption peak profiles (Figure 3c), together with the possible structure sensitivity of the reaction activity on different facets. Studies on single crystalline Pt showed that the exchange current density of HOR/HER increases in the order of Pt(111) < Pt(100) < Pt(110) in both acidic²⁹ and alkaline³⁰ electrolytes. HOR/HER activities on single crystalline Pd surfaces have not been reported, which is likely also due to diffusion of hydrogen atoms into the bulk of single crystal samples. It is reasonable to conclude that HOR/HER on Pd is structure sensitive based on the particle size dependence of HOR/HER activities in this study. The particle size effect of HOR/HER on carbon supported Ir nanoparticle in alkaline electrolyte also suggests the possibility of structure sensitivity of HOR/HER on Ir.²² Additionally, the H_upd desorption peak potential is related to hydrogen binding energy (HBE) as in Eq. 9, which has been proposed to be a dominating descriptor for HOR/HER activities.²²

where E_peak is the H_upd desorption peak potential, and S⁰_H2 is the standard entropy of H₂. Since Pd belongs to the strongly-binding branch in the "volcano plot" (HOR/HER exchange current densities vs. HBE),^11,31 sites with weaker HBE will possess higher HOR/HER activity. Zhou et al. deconvoluted the H_upd desorption in the cyclic voltammogram of Pd/C obtained in 0.1 M HClO₄, and attributed the peaks to low-index Pd facets (111), (100) and (110).¹⁵ In the current study, the sites associated with the H_upd desorption peak centered at around 0.19 V on Pd samples in 0.1 M HClO₄ bind to hydrogen more weakly than those associated with peak centered at around 0.24 V, and thus should have higher HOR/HER activity. The increased ratio of the sites with lower HBE with rising particle size is responsible for the improved HOR/HER activity from untreated Pd/C (about 3 nm) to Pd/C-500C (about 19 nm). Similar result has been observed on Ir/C where the exchange current density and the fraction of sites with lowest HBE increase with particle size and a good correlation between exchange current density and the fraction of sites with lowest HBE is obtained.²² Meanwhile, it is generally accepted that larger particles contain less defect sites compared with smaller particles. The lower ratio of the defect sites with low coordination number and a weaker binding strength to adsorbates (Pd-H binding in this case) in larger nanoparticles could lead to a higher HOR/HER activity. Therefore, Pd nanostructures with extended surfaces and less defect sites such as Pd nanowires (PdNWs) or Pd nanotubes (PdNTs) will have higher activity toward HOR/HER than Pd nanoparticles. Indeed, a much higher exchange current density of 0.96 mA/cm²_Pd was obtained on PdNTs synthesized via galvanic displacement reaction using copper nanowires (CuNWs) as templates, which is even about 5 times of bulk Pd possibly attributed to the small amount of Cu left in Pd lattice.¹²

The HOR/HER activities in 0.1 M HClO₄ are approximately a factor of 50 higher than those in 0.1 M KOH on all Pd samples (Table II and Figure 6a). Durst et al. reported that the exchange current density on Pd/C (about 3 nm) measured using H₂ pump in a PEMFC configuration (equivalent to pH = 0) is about 90 times of that obtained in 0.1 M NaOH using RDE method at 313 K,¹ which is consistent with our results. Consensus has not been reached regarding the root cause for the higher HOR/HER activity in acid than in base on Pt or Pt group metals so far: whether HBE is the sole and unique descriptor for HOR/HER^1,22,32,33 or there is a change in reaction mechanism from acid to base and adsorption OH will facilitate HOR/HER in base.^34,35 The H_upd adsorption/desorption peaks on Pd samples in CVs measured in 0.1 M HClO₄ are located at more negative potentials compared with those measured in 0.1 M KOH (Figures 3a and 3b), suggesting a weaker Pd-H binding energy in acid than in base. Therefore, HOR/HER should have a smaller activation energy in acid according to the Brønsted-Evans-Polanyi (BEP) principle. In this regard, we measured HOR/HER polarization curves on Pd/C in both 0.1 M HClO₄ and 0.1 M KOH at temperatures from 275 to 313 K, and determined the exchange current densities at each temperature to generate the Arrhenius plots (Figure 7). The apparent activation energy of HOR/HER on Pd/C in 0.1 M HClO₄ is 32.3 ± 0.7 kJ/mol, which is similar to that reported when the exchange current densities were determined using H₂-pump (31 ± 2 kJ/mol).²⁷ A higher activation energy is obtained in 0.1 M KOH (38.9 ± 3.0 kJ/mol) than in 0.1 M HClO₄, which confirms our prediction. In addition, the activation energies of HOR/HER on Pt/C, Pd/C and Ir/C in 0.1 M KOH are in the sequence of Pt/C (29.5 ± 4.0 kJ/mol)⁸ < Ir/C (32.9 ± 1.5 kJ/mol)²² < Pd/C (38.9 ± 3.0 kJ/mol, this work), which correlates well to the sequence of HOR/HER activity in base in the order of Pt/C (0.57 ± 0.07 mA/cm²_Pt)⁸ > Ir/C (0.21 ± 0.02 mA/cm²_Ir)²² > Pd/C (0.052 ± 0.002 mA/cm²_Pd, this work). Similar correlation between the activation energy and the exchange current density on Pt/C, Pd/C and Ir/C is also valid in acid.²⁷ Therefore, the higher Pd-H binding energy in base is likely responsible for the lower HOR/HER activity.

Zoom In Zoom Out Reset image size