Gold(III) chloride trihydrate (HAuCl₄ 3H₂O, 99.9%), sodium tetrachloropalladate(II) (Na₂PdCl₄, 98%), silver nitrate (AgNO₃, 99%), cetyltrimethylammonium chloride (CTAC, 97%), cetyltrimethylammonium bromide (CTAB, 99%) and ascorbic acid (AA, 99.0%) were all obtained from Aladdin Chemical (Shanghai, China) and used as received. Docoscyltrimethylammonium chloride (DTAC, 85%) was obtained from Degussa and used as received. Commercial Pd/C (10 wt.% Pd loading, matrix activated carbon support) was obtained from Sigma-Aldrich and used as received. Conductive carbon black was obtained from Cabot and used as received. In all experiments, we used deionized water with a resistivity of 18.2 MΩ cm, which was prepared using an ultrapure water system (Ulupure, China).

Concave cubic Au seeds were generated according to the standard procedure as described in our previous study (Zhang et al., 2020). Typically, one batch of Au concave nanocubes were purified and re-dispersed in 1 ml of water for further use (mass concentration: ∼0.15 mg_Au/mL). For the synthesis of Au@PdAg concave nanocubes (Procedure P1), aqueous solutions of DTAC (20 mM, 3 ml), as-prepared concave Au cubic seeds (0.5 ml), Na₂PdCl₄ (20 mM, 5 μL), AA (10 mM, 0.5 ml) were sequentially mixed in a 20-ml glass vial and aged for 10 min at room temperature, followed by the injection of AgNO₃ (2 mM, 0.5 ml) solution. The reaction proceeded for 2 h in a water bath set at 60°C and the products were collected via centrifugation and washed with water once prior to further use. See Table 1 for synthetic details for Au@PdAg core-shell concave nanocubes with different PdAg shell thickness (Procedure P1∼P5). For Au@Pd nanoparticles used in electrocatalysis, they were produced via the procedure P5, except that no AgNO₃ was added.

Transmission electron microscopy (TEM), high-resolution TEM (HRTEM), selected-area electron diffraction (SAED), high angle annular dark field-scanning transmission electron microscopy (HAADF-STEM) and EDX (energy dispersive X-ray)-STEM mapping images were obtained using a Talos F200X (FEI, USA) microscope operated at 200 kV accelerating voltage. Scanning electron microscopy (SEM) images were obtained using a Zeiss Ultra60 microscope operated at 12 kV. The crystalline structures were analyzed with a MiniFlex600 X-ray diffractometer (XRD, Rigaku). X-ray photoelectron spectroscopy (XPS) measurements were performed using a Thermo Fisher Scientific KALPHA XPS with monochromatic Al K_α radiation (hν = 1,486.6 eV). All extinction spectra were recorded using a T9 dual-beam UV-vis-NIR spectrometer (PERSEE, China). Inductively coupled plasma analysis was conducted on an ICAP-5000 inductively coupled plasma optical emission spectrometer (Focused Photonics Instrument, China). EOR measurement was carried out in a standard three-electrode system controlled by a CHI-760E potentiostat (CHInstruments, China). See supporting information for operational details.

Typical synthesis started with the preparation of concave Au nanocubes in high purity according to the method described in our previous study (Zhang et al., 2020), followed by purification and the use as seeding material. The growth process involved the use of DTAC and AA as the capping agent and reductant, respectively, as well as the sequential addition of Pd and Ag precursor, respectively (Figure 1A). TEM image of the Au seeds showed that they had the concave cubic shape in high uniformity, together with an average edge length of 37 nm (Figure 1B). The Au@Pd and Au@PdAg products obtained via the sequential addition, respectively, maintained the concave cubic morphology as seeds, indicating the successful conformal growth occurred (Figures 1C,D). It is worth noting that the absence of Pd precursor would lead to the formation of Au@Ag core-shell nanocubes with the truncations on the corners (Supplementary Figure S1A). The position of concave cubic Au seeds was not located in the right center of the resulting nanocube, which should be attributed to the fact that heterogenous nucleation and growth of Ag initially occurred to one or partial side faces of the concave cubic Au seeds (Zhu et al., 2012). Corresponding UV-vis extinction spectrum showed that the Au@Ag core-shell nanocubes exhibited the major peak located at 501 nm, together with a shoulder peak located at around 400 nm (Supplementary Figure S1B).

To further analyze their morphology and structure, a set of characterizations, including HRTEM, SAED, HAADF-STEM, EDX-STEM, and XPS, were conducted. As shown in Figure 2A, the TEM image of an individual particle clearly shows the concave feature on the particle surface. Curved projections could be observed at several different facets, projection angle was estimated to between 142^o, 151^o, 153^o, and 158^o, respectively. Considering these projection angles are the angles between the projection line of exposed facets, several typical high-index facets, such as {730}, {830}, {920}, and {720} facets, could be calculated and indexed (Yu et al., 2010; Zhang et al., 2015; Kuo et al., 2018; Luo et al., 2019). The HRTEM image (Figure 2B) shows lattice fringes at the corner region, where a clear boundary could be observed to differentiate the Au and PdAg part. The lattice spacing of 0.23 and 0.20 nm, can be ascribed to the (111) crystal plane of Au and PdAg, respectively. SAED patterns taken from two typical individual particles showed only one set of diffraction spots, respectively, suggesting their single-crystal nature (Figures 2C,D). HAADF-STEM (Figure 2E) and EDX-STEM (Figures 2F–I) images, as well as the line-scan profile (Supplementary Figure S2) confirmed the core-shell structure and the shell thickness was measured to 5.8 ± 1.9 nm. Despite the sequential addition of Pd and Ag precursor, both Pd and Ag was homogeneously distributed over the shell part, indicating the formation of alloy phase. Instead of showing a multi-layered structure (i.e., Au@Pd@Ag), the ease alloying of Pd and Ag under ambient condition could allow the shell to exhibit the homogenous distribution of both metals. The molar ratio of Au:Pd:Ag was determined by EDS as 81:13:6 (Supplementary Figure S3). XPS spectra showed typical doublets for zero-valent Au, Ag and Pd (Figures 2J–L). Taken together, we can conclude that the as-obtained products had the Au@PdAg core-shell structure and concave cubic morphology, as well as the single-crystal nature and curved surface.

As a major advantage of seeded growth, the PdAg shell thickness could be readily controlled by simply varying the amount of Pd precursor while maintaining other parameters unchanged. As shown in Figure 3, the HAADF-STEM and EDX-STEM images clearly showed that PdAg shell noticeably became thicker as more Pd precursor was added, while the overall particle kept the concave cubic morphology despite the change in PdAg shell thickness. In particular, the PdAg shell thickness was measured to be 7.8 ± 3.0, 8.8 ± 3.4, 9.9 ± 4.6, and 12.5 ± 4.9 nm, respectively, when Pd precursor with the amount of 0.2, 0.4, 1, and 2 μmol, respectively, was involved in the growth solution. Line-scan profiles of these products confirmed the core-shell structure and the concave cubic shape, where the Au signal was relatively stronger in the center of the particle and the signal of Pd and Ag mainly emerged at corner/edge regions (Supplementary Figures S4–S7). The molar ratio of Au:Pd:Ag for these products were determined using EDS, which were 75:16:9, 74:18:8, 55:30:15, and 40:44:16, respectively (Supplementary Figures S8–S11). It is worth noting that the amount of Ag precursor was not varied for all these procedures but the molar ratio of Pd to Ag was roughly maintained at 2:1 in all these products (Supplementary Figure S12). It suggested that not all the Ag precursor was reduced and grew on the Au seeds under current synthetic condition and the presence of more Pd precursor would facilitate the seed growth of Ag.

To confirm the presence of PdAg alloy phase, we also took these products with different shell thickness for XRD analyses. As shown in Figure 4, the products with the PdAg shell thickness of 5.8 ± 1.9, 7.8 ± 3.0, 8.8 ± 3.4 nm, mainly showed the diffraction peaks of face-centered cubic (fcc)-Au phase. The missing of PdAg diffraction peaks for these samples should be attributed to the limited amount of PdAg. As the PdAg shell thickness increased to 9.9 ± 4.6 nm, several diffraction peaks that were located at 39.5^o, 46.1^o, 67.6^o, and 81.4^o emerged, right between corresponding (111), (200), (220), and (311) diffraction peaks of Pd and Ag, suggested the formation of PdAg alloy phase. The diffraction peaks of PdAg turned dominant as the shell thickness reached 12.5 ± 4.9 nm, where the diffraction peak intensity of Au became dampened.

Supplementary Figure S13 shows the UV-vis extinction spectra of Au@PdAg concave nanocubes with different shell thickness. For the product with PdAg shell thickness of 5.8 nm, it showed major peak located at 396 nm and a shoulder peak at 508 nm. As the shell became thicker (i.e., 7.8 and 8.8 nm), the major peak shifted to between 500 and 600 nm. Further increase in PdAg shell (i.e., 9.9 and 12.5 nm) caused a broad absorption centered at 616 nm. These results suggested that the plasmonic properties of Au@PdAg core-shell nanocrystals can be tuned by varying the shell thickness and the major absorbance peak can still be observed in the UV-vis extinction spectrum if the thickness was controlled to sub-10 nm range.

To the elucidate the formation mechanism of PdAg shell with a unique morphology of concave cube, control experiments were conducted to investigated the effect of capping agent and reaction kinetics on product morphology. As shown in Figures 5A,B, when the capping agent was replaced by CTAC, the resultant products still exhibited the concave feature but the cubic shape largely lost. It showed that alkyl chain length of quateryammonium surfactant played a more crucial role in determining the shape of final products. Contrarily, the use of CTAB led to the formation of products anchored with lots of tiny particles on the surface. Such variation could be attributed to the stronger affinity of the Br⁻ ions to the Au seeds surface (Zhu et al., 2013), which made it difficult for PdAg atoms to deposit via the layer-by-layer mode (Frank-van de Merwe mode). Considering the lattice mismatch between Au and PdAg, the PdAg atoms may tend to form tiny particles on the seed surface to minimize the interfacial energy (Vollmer-Weber mode) (Peng and Hong, 2009).

In addition to the capping agent, the effect of reaction kinetics was also investigated by increasing the amount of reductant in the growth solution. As shown in Figures 5C,D, as the amount of AA increased from 5 to 20 μmol and 100 μmol, respectively, the quantity of self-nucleated particles increased in the final products. It could be attributed to the competition between heterogenous and homogeneous nucleation and growth. It is more likely for heterogenous nucleation and growth to occur when the concentration of metallic atoms was controlled to a low level. This is because for homogeneous nucleation, the concentration of the Pd/Ag atoms should go beyond the supersaturation point (e.g., critical nucleation concentration) to cross the energy barrier (Hsia et al., 2018; Gamler et al., 2020; Liu and Zhang, 2020). Typically, the use of excessive reductant can cause the significant increase in the concentration of newly-formed zero-valent atoms in the growth solution (Zhang et al., 2018). To this end, the generation of Pd/Ag atoms in large quantity within a short time period would make the nucleation and growth dominated by the homogenous route considering the lattice mismatch existed. Without the regulation of seeds, the resulting products would take random shapes and crystallinities. In short, the use of adequate reductant and Br⁻-free quateryammonium surfactant contributed to the conformal growth of PdAg over concave cubic Au seeds in the current study.

Due to the combined advantage in both structure and elemental composition, the current Au@PdAg core-shell concave nanocubes are expected to find important use as electrocatalysts for EOR in alkaline media. In particular, the product with the 12.5-nm shell thickness was chosen for the dominant phase of PdAg. They were collected via centrifugation and loaded on carbon black to construct Au@PdAg/C electrocatalysts. Prior to EOR measurements, the catalysts were washed with water once to remove the residual surfactant for surface purification. For comparison, carbon-supported Au@Pd core-shell nanoparticles and commercial Pd/C electrocatalysts were employed as the reference materials.

As shown in Supplementary Figure S14, CV curves collected in 1 M KOH exhibited a pronounced cathode peak between −0.5 and −0.1 V during the backward sweep corresponding to the reduction of surface PdO to metallic Pd. Based on the integrated charge associated with this cathodic peak, we estimated that the electrochemically active surface area (ECSA) of Au@PdAg/C, Au@Pd/C, and Pd/C was 28.50, 30.67, and 161.7 m² g_Pd ⁻¹, respectively. Compared to the commercial Pd/C, the relatively small ECSA values of Au@PdAg/C and Au@Pd/C electrocatalysts could be attributed to the large overall particle size. With the presence of ethanol, all CV curves were drastically changed to have a strong peak in the forward scan due to EOR electrocatalysis (Figure 6A). The mass activity and specific activity of Au@PdAg/C was calculated to be 600.21 mA mg_Pd ⁻¹ and 19.57 A m⁻², which were both larger than that of Au@Pd/C (317.49 mA mg_Pd ⁻¹ and 11.32 A m⁻²), and Pd/C (367.67 mA mg_Pd ⁻¹ and 2.27 A m⁻²). Furthermore, the onset potential (E_s) for the EOR over Au@PdAg/C was −768 mV, which were more negative than that of the Au@Pd/C (−734 mV) and Pd/C (−678 mV) electrocatalyst. It indicated that the electroxidation of ethanol can occur more easily on the Au@PdAg electrocatalyst.

The electrocatalytic kinetics of EOR were also compared to further explore the oxidation process on the surface of the different catalysts. As shown in Figures 6A,B linear relationship between the square root of the scan rate (v^1/2) (30–200 mV s⁻¹) and the forward peak current density could be observed in all three nanomaterials (Supplementary Figure S15), indicating a diffusion-controlled process for the EOR (Sun et al., 2012). In addition, the Au@PdAg/C exhibited the enhanced electrocatalytic kinetics as can be concluded from the larger slope value when compared to the Au@Pd/C and Pd/C. All these values imply that the current Au@PdAg/C electrocatalysts have a higher intrinsic electrocatalytic activity for the EOR as compared to Au@Pd/C and Pd/C. In addition, the long-term durability was evaluated via three successive cycles of current-time (i-t) chronoamperometry measurements at a potential of −0.2 V vs. SCE for 3,000 s. The Au@PdAg/C electrocatalysts manifested a larger residual activity for each cycle as compared to the other two electrocatalysts (Figure 6C). All the as-mentioned data was summarized in Table 2 for comparison.

The morphology and structure of Au@PdAg/C electrocatalyst after the CA measurements were characterized using electron microscope. As shown in Supplementary Figure S16, the electrocatalyst particles were still isolated from each other and no significant agglomeration was noticed (Supplementary Figure S16A). The morphology of the PdAg shell over the Au seed varied with some branching feature. Some of the PdAg shells became disengaged from the Au seed surface but their curved surface maintained (Supplementary Figure S16B). By conducting analysis using HAADF-STEM (Supplementary Figure S16C) and EDX-STEM (Supplementary Figure S16D), despite of the reservation of core-shell structure, the alloying between Au and PdAg was observed. The change in morphology and structure of Au@PdAg/C electrocatalysts should be caused by the long-term cycling test, which also explained the decline in ending-point specific activity for each test.

To evaluate the reusability of catalysts, we also conducted three consecutive CA measurements of EOR using Au@PdAg/C electrocatalyst, where the fresh aqueous solution of 1M KOH+1M ethanol was used as the electrolyte for each cycle. As shown in S upplementary Figure S17, the specific activity maintained at 3.1, 2.5, and 2.1 A m⁻² for the consecutive three cycles, respectively, demonstrating the acceptable reusability of the current product for EOR electrocatalysis. In short, owing to the synergistic structural (i.e., stepped facets on the surface) and compositional advantages (i.e., containing both Pd and Ag), the current Au@PdAg nanocrystals exhibited remarkable specific activity, long-term stability, and enhanced electrocatalytic kinetics for EOR in alkaline media.

Compared to commercial Pd/C electrocatalyst, the principal drawbacks of the current products could be mainly attributed to the relatively higher j_r/j_f value (e.g., 0.72 for Pd/C and 1.02 for Au@PdAg/C, Table 2), suggesting their weaker ability to poisoning-resistance of the electrodes for ethanol oxidation. It should be caused by its elemental composition, where the doping of Pd with Ag changed the electronic state and the surface affinity towards reaction intermediates. This disadvantage has also been noted in previous studies on PdAg-based EOR electrocatalysts (Li et al., 2014; Fu et al., 2015).