Tuning the electronic structure of Ag-Pd alloys to enhance performance for alkaline oxygen reduction

Alloying is a powerful tool that can improve the electrocatalytic performance and viability of diverse electrochemical renewable energy technologies. Herein, we enhance the activity of Pd-based electrocatalysts via Ag-Pd alloying while simultaneously lowering precious metal content in a broad-range compositional study focusing on highly comparable Ag-Pd thin films synthesized systematically via electron-beam physical vapor co-deposition. Cyclic voltammetry in 0.1 M KOH shows enhancements across a wide range of alloys; even slight alloying with Ag (e.g. Ag0.1Pd0.9) leads to intrinsic activity enhancements up to 5-fold at 0.9 V vs. RHE compared to pure Pd. Based on density functional theory and x-ray absorption, we hypothesize that these enhancements arise mainly from ligand effects that optimize adsorbate–metal binding energies with enhanced Ag-Pd hybridization. This work shows the versatility of coupled experimental-theoretical methods in designing materials with specific and tunable properties and aids the development of highly active electrocatalysts with decreased precious-metal content.

x-ray diffractograms of as-prepared Ag, Pd, and mixed Ag1-xPd1-x thin films (deposited on standard microscope glass slides). Full 2θ range for data shown in Figure 1a. (b) GI x-ray diffractograms of the (111) peak for Ag, Pd, and Ag-Pd thin films of different composition after ORR testing, with Ag (38.17°) and Pd (40.23°) peak locations shown by dotted lines, and (c) Vegard's Law analysis of the (111) peak showing Pd content in alloy versus nominal Pd composition both pre-and post-electrochemical ORR activity measurements. All post-electrochemistry characterization was done with thin films deposited on glassy carbon substrates. Grazing incidence XRD measurements on the thin films after electrochemical testing (Supplementary Figure   1b), along with the corresponding Vegard's Law analysis (Supplementary Figure 1c), indicates that the structure and alloy composition of the thin films did not change significantly after electrochemical experiments. The minor differences in alloy composition given by Vegard's Law before and after electrochemistry are likely due to the low signal to noise ratio in the post-ORR testing diffractograms of the thin films taken on glassy carbon substrates.
The absolute value differences in 2θ (111) peak position of the as-deposited versus the after electrocatalysis seen in Supplementary Table 1 are likely due to the different substrates used, as well as the different grazing incidence angle needed to obtain a measurable signal when using glassy carbon substrates.
Supplementary Table 1   Similarly to in GI-XRD, a strong (111) peak is also present in all diffractograms taken in a symmetric scan (ω = half of 2θ range) geometry (Supplementary Figure 2), consistent with a preferential 〈111〉 out of plane A compositional analysis from XPS measurements (Figures 1c, 1d x-ray absorption near edge spectra for Pd, Ag0.1Pd0.9, Ag0.2Pd0.8, Ag0.5Pd0.5, Ag0.8Pd0.2. All samples measured on glassy carbon substrates. Full energy-range XAS spectra for data shown in Figure 1e. The filling in the Pd 4d-band results in an increase in the incident energy needed to excite Pd electrons from the 2p to 4d band, which is illustrated by the positive energy shifts (ranging between 0.5 -1 eV) in the white-line position with increased Ag content. 5,6 As explained in the main manuscript, our Pd L3 XANES agrees wells with literature, and therefore we hypothesize Flatness/smoothness in AFM topographical images is represented by color homogeneity, not the actual color being displayed. For example, for Ag0.7Pd0.3, the "post" image is more dark-red in color compared to the "as deposited" image, but both images are highly homogeneous in their own color distribution, meaning that they are both highly smooth/flat. The difference in the main/most noticeable color among images can be due to differences in the magnitude of the interactions between the cantilever and the sample surface, as well as the scanning settings, which change sample to sample.
AFM was used to determine the surface roughness and topography of the thin films before and after electrochemical testing. AFM imaging shows a lack of topographical features on the thin films both before and after electrochemical measurements (Supplementary Figure 6). The surface roughness factors (RF) of the thin films were estimated by AFM imaging (Supplementary   corresponding Koutecký-Levich (KL) selectivity analysis (inverse mass transport (MT) limited current density as a function of the inverse square root of rotation rate) for Ag0.1Pd0.9. For this experiment the glassy carbon substrate/disk was slightly taller than the change disk insert cavity, causing the edges of the disk, which had catalyst material on it, to be exposed, and hence slightly increasing the overall catalyst surface area relative to the geometric electrode area (0.196 cm 2 ) leading to mass transport limited current densities slightly more negative than expected given 4eselectivity. We used F = 96485.333 C mol -1 , A = 0.196 cm -2 , D = 1.93 x10 -5 cm 2 s -1 , = 1.09 x10 -2 cm 2 s -1 , and CO2 = 1.26 x10 -6 mol cm -3 . 9 In addition to suggesting full 4eselectivity, the KL measurements in Supplementary Figure 9 have overlapping kinetically controlled regions, indicating that mass transport limitations were not problematic.

a) b)
Koutecký-Levich Analysis The Koutecký-Levich equation 19 can be expressed as: where, is the experimentally measured current, ! is the kinetic current from the specific electrochemical reaction, and "# is the experimentally measured mass transport limited current density. and "# can be found using cyclic voltammetry, and then used to solve for ! . Kinetic current density, jk, can be obtained by normalizing ! by a representative catalyst surface area, and thus be used as a measure of specific (intrinsic) activity.  x = 90%

Tafel Analysis
Ag 1-x Pd x Ag 1-x Pd x

Performance Summary
Supplementary Table 3. Summary of Ag1-xPdx ORR activity performance. The kinetic current densities are normalized by the roughness factors determined by AFM (Supplementary Table 2), and are representative of specific activity. The Tafel Slopes are for the low current density (lcd) regions. Exchange current densities were calculated by setting potential equal to 1.23 V vs. RHE and using the lcd region Tafel linear fit (Potential = Tafel Slope * log10(-jk) + constant). All average (AVG) data points are representative of the average value from the cathodic and anodic scan of the third CV cycle from measurements on separate samples (n = 2 samples for 0-30 at% Pd, and n = 3 for 40-100 at% Pd). Relevant for Figure 2c in the main text, and Supplementary   Note that = 45 ∘ is the standard/default collection angle for non-AR XPS measurements, with an estimated probe depth of 3.1 nm. 20 AR-XPS measurements were taken using a standard collection aperture because no peak signal was observed using the narrow collection aperture (slit cover on detector). The estimated probe depth at = 10 ∘ is ~ 0.8 nm 20 . We estimated probe (95% of electrons probed) depth as d95% = 3 sin , based on the estimated Ag0.1Pd0.9 inelastic mean free path ( %& !.# '( !.$ ), and defined %& !.# '( !.$ = 0.1 %& + 0.9 '( ≈ 1.449 nm, where %& ≈ 1.53 nm and '( ≈ 1.44 nm at an electron kinetic energy of 1096 eV (approximately equivalent to the binding energies corresponding to the Ag and Pd 3d peaks). 20 (d) Pd content (at%) as a function of cumulative sputtered depth pre-(blue, filled; separate sample from the same synthesis batch) and post-(orange, unfilled) stability measurements (CA + CVs) determined by XPS depth profiling (Ar + sputtering, sputtering rate calibrated versus SiO2).

c) d)
Representative Stability Testing of Ag0.1Pd0.9 For our best performing composition, Ag0.1Pd0.9, we performed chronoamperometry (CA) at ~ 0.8 V vs. RHE for 2 hrs (Supplementary Figure 14a) and saw a decrease in current of about 40% (interestingly, it is seen to decrease linearly after around 15 min). We hypothesize that the decrease in performance is due to reversible adventitious carbon uptake due to carbon corrosion from the graphite rod counter electrode, as well as possible Pd and/or Ag partial surface oxidation; the latter is in fact the main source of decreased performance in similarly high performing Pt-based ORR catalysts. ORR CVs pre-and post-CA testing (Supplementary Figure 14b) support this hypothesis, as 100% of the initial (pre-CA) performance was recovered by cycling the working electrode up to around -0.2 V vs. RHE to reductively clean off the adventitious carbon and any surface oxide on the film's surface. Additionally, AR-XPS (Supplementary Figure 14c) and XPS depth profiling (Supplementary Figure 14d) indicate that the composition of the film did not significantly change after stability testing. Altogether, our stability testing indicates that there is no intrinsic material or performance degradation in the timescale of or experiments.
Limited Ag1-xPdx Stability Testing Ag1-xPdx thin film mechanical instabilities (not related to intrinsic activity) made our system not conducive for long-term durability testing. When cycling and chronoamperometric experiments were attempted the thin films partially or totally delaminated within three hours of testing. For all samples tested, we collected 3 ORR CVs as activity measurements. The first CV serves as a "cleaning" cycle, and the second and third serve as duplicate activity measurements for each sample tested (2 or 3 samples at each for all samples tested composition). These last 2 CVs were very reproducible for each sample and we report all our results using the third ORR CV cycle for each sample.
To assess the material response to electrochemical testing, we characterized the structural, morphological, and compositional surface changes after electrocatalysis (3 CVs in O2-saturated 0.1 M KOH, followed by 2 CVs in N2-saturated electrolyte, followed by 5 CVs at different scan rates in N2-saturated electrolyte, all at 1600 rpm) via XPS, GI-XRD, and AFM, respectively. Respectively, XPS (Supplementary Figure 4) Table 4. Comparison to a selection of high-performing ORR catalysts in 0.1 M KOH (or similar) using a RDE setup. We only report representative catalysts among the most active found in the literature, where kinetic current density is given or easily calculated and explicitly normalized by a physically representative exposed catalyst (or active site) surface area ("cat") (also referred to as electrochemically active surface area (ECSA)). Examples of ECSA techniques are AFM surface roughness for thin films and flat surfaces, electrochemical stripping (e.g of CO or Cu), hydrogen under-potential deposition, metal-oxide layer reduction, and scan rate potential-cycling. 21,22 In the absence of AFM, geometric electrode area was taken as an acceptable approximate ECSA only for well-defined single crystal catalysts. Outside of this work, references are for linear sweep voltammetry data. For our Ag-Pd thin films in this table, we report the average (from the cathodic and anodic cycle) specific activity; see Supplementary Table 3 for the individual activity at each cycle leg (our anodic/forward cycle shows slightly higher activities than the average we report in this table  Figures 4, 14c, and 14d) and GI-XRD (Figure 1b and Supplementary Figure 1c) data suggest that, within technique sensitivity (top several nanometers), the Ag-Pd alloy thin films did not undergo major surface rearrangements during ORR experiments.
Interestingly, the limiting potential for the ORR on Pdatom−Ag (−5%) and Pdlayer−Ag (111)  This peak could result from the ineffective mixing of the electron densities of the Pd single atom and the Ag (111) surface and the effective tensile strain on the Pd single atom due to large lattice of Ag(111). A lower d-band edge relative to the Fermi level is observed for Pd single atom in -5% compressively strained Ag(111), Pdatom-Ag(-5%), compared to Pdatom-Ag and Pdatom-Ag(5%) indicating a weaker oxygen adsorption on Pd atom. This results in increase in ORR overpotential and therefore increased activity. A similar d-projected density of states of Ag atom in Ag(111) and Pdatom-Ag(111) surfaces with different strains is observed (Supplementary Figure 17b) indicating that the ORR activity of these Ag atoms are not affected.
Supplementary Table 5. DFT adsorption free energies of OH* and OOH*, and thermodynamic limiting potentials and steps for considered Ag−Pd active site models in Figure 3a. Note that all the considered surfaces are (111) facets. (This is the data plotted on Figure 3b in the main text). use the ISPIN = 2 tag to turn on the spin and then define the magnetic moment using the MAGMOM tag as implemented in VASP. Because no drastic change in observed activity trends is expected due to valence spin orbits, in this study we do not consider valence spin orbits. 37 We do not account for dispersion interactions because they are expected to only have a minor effect and generally calculations for chemisorbed species using RPBE and BEEF-vdW (which accounts for dispersion interactions) follow similar trends. 38 While adsorbate-adsorbate interactions could play a role in the apparent catalyst activity, we anticipate similar adsorbate−adsorbate interactions across all the metal surfaces considered in this study due to the similarity in surface atoms ordering (all considered surfaces are (111) facets) and hence similarity in the resulting adsorbate−adsorbate distances.
Therefore, adsorbate−adsorbate interactions would likely not change the calculated activity trends. The aim of this work is to look at activity trends across the Ag-Pd compositional spectrum and therefore a detailed analysis considering adsorbate−adsorbate interactions for all the individual surfaces is beyond the scope of this study.