Recent advances in nanoscale engineering of Pd-based electrocatalysts for selective CO₂ electroreduction to formic acid/formate

Fundamental ECO₂RR mechanisms of Pd electrocatalysts

The CO₂ molecule is chemically inert and thermodynamically relatively stable^[42], so activating it on the catalyst surface needs a high activation energy which must be overcome by applying a much negative reduction potential^[43]. In an aqueous solution with a pH value of 7.0, the thermodynamic potential for the first electron transfer of CO₂ to form CO^2- is -1.90 V (vs. standard hydrogen electrode, SHE), which is adverse for the reaction^[44-46]. In contrast, proton-assisted multiple electron transfer processes can occur at lower negative potentials (-0.2~-0.6 V vs. SHE) on appropriate catalyst surfaces, which helps to stabilize reaction intermediates and subsequently leads to the formation of various CO₂ derivatives^[47].

As shown in Figure 1A, ECO₂RR, a complicated process^[48], contains multielectron and proton coupling steps, involving 2 to 12 electron reaction pathways [Figure 1B]^[49] and generating various reduction products^[50]. In general, the multi-step reaction includes three main steps: (1) CO₂ chemisorption and activation on the catalyst surface. This step involves the first electron transfer from the catalyst to CO₂ to form *CO₂^•- (*infers the adsorption site), along with the proton transfer from the electrolyte, forming *COOH or *OCHO. The *COOH or *OCHO intermediate is immediately stabilized on the metal surface by forming a metal-carbon or metal-oxygen bond; (2) Electron transfer from electrocatalyst to *COOH or *OCHO on the metal surface is accompanied by the proton transfer from the electrolyte, resulting in the formation of adsorbed key intermediates and their eventual product; and (3) Desorption of the products from the electrocatalyst surface and diffusion into the electrolyte^[5].

Figure 1. (A) Illustration of the electrochemical CO₂ reduction process and the possible products generated in the CO₂ electrolyzer^[48]. Reprinted with permission. Copyright 2018, WILEY^[48]. (B) The reduction potentials of the semi-reactions of the most commonly ECO₂RR products in aqueous media (pH 7, 1 atm and 25 °C, 1 M solute concentration)^[49]. Reprinted with permission. Copyright 2023, WILEY^[49].

The vast majority of ECO₂RR electrocatalysts studied at this stage are mainly based on various metal-based materials^[19,51-53]. The type and yield of the targeted product is controlled by tuning the binding ability of their surface and the adsorbed intermediates (*CO, *COOH, *CHO, *COH)^[54-57]. For example, for the C1 products, when the interaction between the electrocatalyst surface and the intermediates is not strong enough, HCOOH/HCOO^- and CO are the main reduction products [Figure 2]. As schematically illustrated in Figure 2, different adsorption configurations of *CO₂^•- radical anion on the electrode surface usually determine the final formation of HCOOH/HCOO^- or CO product. In brief, if O atoms in *CO₂^•- act as the binding sites via an oxygen coordination mode, the carbon atom in *CO₂^•- is protonated to create *OCHO, which is further reduced to HCOO^- after electron transfer^[58]. This is because the C-O bond cannot be dissociated in the case of weak interaction^[59]. The final liquid C1 product in the electrolyte can exist in the form of either formic acid or formate, which depends on the pH of the electrolyte. Alternatively, when the adsorption configuration is carbon coordination mode, *CO₂^•- binds on the electrode surface through the carbon atom, and the oxygen atom in *CO₂^•- radical intermediate is protonated to generate the carboxyl intermediate (*COOH), which is suggested as a key intermediate. *COOH is then reduced to *CO and desorbs from the electrode surface to generate CO product^[27,60]. However, when the interaction of *CO is strong, the intermediates will undergo a more complicated process, accompanied by more electron transfer and protonation, leading to the formation of different types of hydrocarbon and alcohols^[29,61,62].

Figure 2. Schematic diagram of proposed reaction pathways of electrochemical reduction CO₂ to HCOOH/ HCOO^- or CO product by forming different intermediates. (C atoms: black; O atoms: red; H atoms: white).

For CO₂-to-HCOOH conversion, metal Pd has received great consideration and is recognized as a promising catalyst due to its high selectivity at low overpotentials^[38,63]. Typically, Pd catalysts initially participate in ECO₂RR by preferentially adsorbing *H on the surface, forming an active hydrogen-adsorbed Pd surface (PdH_x phase)^[39]. The relationship between the active-phase transformation and the formate selectivity is highly associated with the applied potentials. As shown in Figure 3A, during the electrocatalytic reaction, the transformation from a pure hydrogen-absorbed PdH_x surface (above -0.2 V) to an almost bare metallic Pd surface (below -0.5 V) indicates a gradual decline in surface H coverage, which is controlled by the potentials and intermediates [Figure 3B]. At low potentials (> -0.2 V vs. reversible hydrogen electrode, RHE), the formation of a mixture of the α- and β-phases of a palladium-hydride core (α + β PdH_x@PdH_x) facilitates the *OCHO intermediates, favoring the production of formate. In the moderate potential range (-0.25~-0.45 V vs. RHE), the α + β PdH_x@PdH_x phase gradually transforms into the β-PdH_x@Pd phase. In this process, most of the surface PdH_x is covered by *CO intermediates and the number of active sites for *OCHO reduces, resulting in the formate pathway being suppressed and the competitive HER dominated. With the negative shift of bias voltage, the adsorbed *H gradually dissociates from the surface and destroys the initial mixed active phase. In the high potential region (< -0.45 V vs. RHE), the finally formed β-PdH_x phase selectively converts CO₂ into CO via *COOH intermediates dominated pathway, and CO subsequently dissociates and desorbs from the surface. In addition, the *CO intermediates coverage on the surface gradually increases as the potential shifts negatively, which makes the formation of PdH_x phases more favorable energetically, weakening the CO adsorption on the surface^[63]. This phenomenon further explains the high CO selectivity at high overpotentials on Pd nanoparticles (NPs).

Figure 3. (A) Schematic diagram of switchable reaction pathways for CO₂ electroreduction caused by electrochemically-driven active-phase transformations of Pd nanoparticles^[39]. Reprinted with permission. Copyright 2017, Springer^[39]. (B) Faraday efficiency and the Tafel slope of formate, H₂ and CO at different potentials^[39]. Reprinted with permission. Copyright 2017, Springer^[39].

In short, these studies provide important insights into the catalytic mechanism of Pd NPs in ECO₂RR. In order to obtain stable formic acid/formate generation during the ECO₂RR process, the design of efficient Pd-based catalysts focuses mainly on two aspects: (1) adjusting the surface oxygen affinity of catalyst to preferentially adsorb *OCHO rather than *COOH; and (2) appropriately tuning the electronic structure to weaken the binding strength of *CO at surface active sites. Next, several nanoscale design strategies for improving the ECO₂RR performance of Pd-based electrocatalysts and an exposition of their recent advancements are outlined. Therefore, it is worthwhile to start reviewing the recent progress on Pd-based electrocatalysts for electroreduction CO₂ to formic acid/formate.

Strategies to enhance ECO₂RR toward formic acid/formate of Pd-based electrocatalysts

As a precious metal catalyst, Pd-based catalysts have excellent catalytic activity and selectivity compared with other catalysts, which plays an irreplaceable role in many catalytic reactions, especially in ECO₂RR. Due to the formation of PdH_x in the reduction process, the variable active phases on the surface of Pd NPs determine the formation of *COOH or *OCHO intermediates^[39,64], leading to the production of CO₂ to CO at high potentials and formic acid/formate at low potentials with high FEs^[63]. In recent years, great efforts have been devoted to developing effective strategies on not only improving the formate activity and selectivity but also inhibiting CO poisoning and catalyst inactivation at high potentials. In the following sections, we will review nanoscale engineering strategies of Pd-based electrocatalysts for improving the catalytic performance toward formic acid/formate products, including size control, morphology and shape control, alloying, heteroatom doping, surface-strain engineering, and phase control. To aid readers in quickly capturing valuable and recapitulative information from related literature addressed in the following sections^{[31,34,40,65-74]}, we outlined the performances (potential, selectivity, current density, and stability) of various Pd-based related catalysts and their measurement conditions, including reactors and electrolyte solution, as shown in Table 1.

Size control

Size effect of metal NPs refers to the variation of catalytic performance with the particle size of the metal catalyst, which is intrinsically associated with the geometric and electronic properties^[75-77]. On the one hand, in terms of geometric properties, it is well-established that decreasing size of NPs would improve the proportion of catalytically active surface or corner sites of NPs, thereby altering the ratio and distribution of catalytic centers^[78,79]. On the other hand, from the perspective of electronic properties, the electron energy level of metal particles would be altered as the size changes, thus influencing the absorption energy of intermediates and determining the final products^[80,81]. Consequently, to improve the product selectivity and study the size effect for CO₂RR catalytic activities based on different-sized Pd-based nanocatalysts, rational utilization size effect is of great significance.

In general, with the usage of chemical additives in the synthetic systems to regulate nucleation and growth dynamics, the size control can be achieved. Rahaman et al. prepared three types of Pd NPs within the size range of 3.8-10.7 nm by adjusting the molar ratio of polyvinylpyrrolidone (PVP) to Pd precursor and reaction temperature with NaBH₄ as reducing agent and PVP as capping agent [Figure 4A]^[65]. As shown in Figure 4B, Pd NPs with the medium-sized 6.5 nm deliver a formate FE of 98% at the potential of -0.1 V (vs. RHE); however, the selectivity toward formate further decreases with the particle size further increases, suggesting a high size-dependence of formate selectivity. The major reason for the selectivity differences is the formation rates of β-PdH phase on various sized Pd NPs and the competitive HER. Later, Gao et al. also studied the impact of NP size on product production activity and selectivity by systematically evaluating the ECO₂RR performance over Pd NPs with different sizes ranging from 2.4 to 10.3 nm^[82]. These Pd NPs can efficiently reduce CO₂ to CO, and the FEs and catalytic current densities of CO were significantly improved with the decrease of particle size. The FEs for CO production over 2.4 and 3.7 nm Pd show the highest values among all the sizes, exceeding 80% across all the applied potential ranges [Figure 4C]^[82]. Based on three kinds of sites (terrace, edge, and corner) of Pd NPs, density functional theory (DFT) calculations reveal that the Gibbs free energies for the formation of COOH* on terrace (Pd (111)) are higher than those on stepped (Pd(221)) and corner (Pd55), indicating corner and edge sites are more active than terrace sites for the formation of COOH* intermediates [Figure 4D]^[82]. However, competitive HER has negligible differences on all these different active sites. Besides, LSV and CO stripping profiles over differently sized Pd NPs show a shift of peak potential of lower reduction potential of Pd(OH)₂ and weakened adsorption of CO with increasing particle size, which agrees well with the DFT calculations. Both the theoretical and experimental results suggest the nature of defected corner sites is beneficial to adsorb CO₂ and form the key reaction intermediates COOH* during CO₂ reduction. The optimal size having a high ratio of low-coordinated reaction sites significantly contributes to the high selectivity for CO production and current density.

Figure 4. (A) Schematic diagram of the synthesis of Pd NPs with different sizes^[65]. (B) FE and yield of formic acid at different potentials^[65]. Reprinted with permission. Copyright 2017 WILEY^[65]. (C) Free energy of COOH* adsorption and CO₂ reduction to CO on different crystal planes^[82]. (D) Potential dependence of CO selectivity on different sized Pd NPs^[82]. Reprinted with permission. Copyright 2015, American Chemical Society^[82]. (E) Schematic diagram of the formation of hexagonal palladium NSs and (F) TEM images of palladium NSs with side lengths of 5.1 nm^[83]. (G) CO Faraday efficiency of palladium NSs with different side lengths^[83]. Reprinted with permission. Copyright 2018, WILEY^[83].

Apart from NPs, Zhu et al. also studied the size effect on ultrathin hexagonal Pd nanosheets (NSs) ranging from 5.1 to 52.6 nm [Figure 4E and F], which were obtained via a one-pot chemical reduction method with CO gas and halide ions as adsorbent^[83]. It is well known that ultrathin NSs with several atomic thicknesses process numerous low-coordinated edge sites^[84,85]. Therefore, adjusting the edge length of the uniform NSs to control the ratio of reactive sites is a straightforward strategy to promote CO₂ conversion and investigate the size-dependent catalytic activity and selectivity. Among the different-sized NSs, CO FE of 5.1 nm Pd NSs displays the highest value, reaching up to 90% at a bias of -0.5 V, whereas 52.6 nm Pd just shows its maximum round 30% at -0.9 V, showing a high dependence on the edge length [Figure 4G]^[83]. In addition, the partial current densities also show a declined trend for CO production with the edge length increasing. Adjusting the size of Pd NPs also helps to improve the formate selectivity at low potentials. In the range of low overpotentials (-0.4 to -0.2 V), it was found that the formate FE over 5.1 nm Pd NSs was 76.4% at -0.2 V, outperforming the other counterparts. Such improvement in CO and formate FEs could be attributed to abundant low-coordinated sites. According to the generalized coordinated numbers of atoms, the authors classified the active sites at different positions of the NS and constructed a series of atomic models based on hexagonal NSs. DFT calculations demonstrate that edge sites and side facets whose generalized coordination number is 5-7 are in favor of CO production. With decreasing the edge length from 52.3 to 5.1 nm, the proportion of atomic with generalized coordination number of 5-7 gradually increased from 2.4% to 17.9%, showing a high dependence on the edge length of NSs. The computational calculations confirm that the formation of *COOH and CO desorption is much easier on specific active sites on the edge of NSs, well explaining the high CO selectivity of 5.1 nm Pd NSs with a high percentage of coordination number 5-7 sites. It should be noted that even though both studies focused on the size effect of Pd catalysts on CO selectivity, they are instructive for the study on formic acid/formate products. The applied biased potentials for both are in the range of high overpotential, which is the CO competing dominated region. As such, the formation of *COOH intermediates is significantly facilitated within these potential ranges. Based on the fact that an electrochemically generated PdH_x surface has been an active site for formate generation, as mentioned previously, the applied potential dependence of formate FE is related to the PdH_x active phases. Even though experimental advances on size dependence of the selectivity towards formate, the real decisive factor for the formate selectivity enhancement over differently sized Pd NPs remains unclear. In this case, whether size-dependent electronic effects facilitate the formation of the PdH_x active phase needs further studies. Therefore, it is vital and significant to take DFT calculations to clarify the relationship between tunable ratios of active sites with different coordination numbers and Pd-H bond strength/catalytically active PdH_x phases to understand the size effect on CO₂RR.

Morphology and shape control

Except for the size-control strategy, morphology and shape control is a promising approach for improving the catalytic performance of Pd-based electrocatalysts^[66,67,86]. Of central importance for catalyst design is maximizing the number of active sites on the surface. To this end, fabricating Pd nanocrystals (NCs) with complex structures provides the opportunity to create more reaction sites due to their highly tunable morphology and shape, resulting in the modification of the electronic structures. Based on the fact that heterogeneous catalysis occurs on the surface of catalysts, from the perspective of morphology, it is meaningful to create porous or hollow architecture to increase the roughness and enlarge surface area of the catalyst, which can create a larger active surface and expose more active sites in the case of constant geometric area^[87]. Besides, it is well-known that different facets of NCs have unique coordination environments and intrinsic properties^[87]. Therefore, exposing specific facets with more active sites is one of the intrinsic regulatory strategies to reduce the activation barriers, thereby boosting the selectivity and activity toward target products. In principle, in a typical synthesis of well-defined Pd-based NCs, the decisive key for the shape control is modulating thermodynamic and kinetic control of NC growth by tuning a set of experimental parameters, such as reducing agent, solvent, ligand, capping agent, surfactant, and extraneous ions. These capping agents, ions, solvents, and so on are selectively bound to a specific facet on NCs, determining the surface energies and growth rate for various facets^[88]. As such, the anisotropic growth results in the formation of NCs with specific morphologies, shapes, or dimensions. The performance of catalysts is usually closely related to their morphology because different morphologies and structures would influence the exposure of crystal facets and the proportion of corner, edge, and terrace atoms of NPs. Hence, morphology and shape control also provides a broad new notion for obtaining highly active and selective Pd-based catalysts for ECO₂RR.

Morphology and shape-controlled synthesis of metal NCs is crucial to creating effective catalysts and boosting the catalytic activity toward ECO₂RR^[87]. Lv et al. adopted a facile soft template method to synthesize single-crystalline mesoporous Pd nanocubes (denoted as s-mesoPd)^[66]. In the synthetic system, the quaternary ammonium salt was used as the surfactant template, and KCl worked as the facet-selective capping agent; both of them codetermine the precise control of the NC growth. The authors also explored diverse experimental parameters, including the carbon chain lengths of the surfactants, the extra Cl^- concentrations, and reaction times, and investigated their effects on structures and morphologies of the final products [Figure 5A]^[66]. Benefiting from large surface areas and rich mesoporous channels, s-mesoPdCu enabled a high formate selectivity over a broad potential window and achieved a FE approaching 100% at -0.2 V [Figure 5B]. Even the applied potential shifted as cathodic as -0.5 V, s-mesoPdCu still held its high selectivity and stability. The introduction of mesoporous greatly improves the charge transfer and dilutes CO coverage, which are responsible for enhanced CO tolerance, great stability and enlarged potential window [Figure 5C].

Figure 5. (A) STEM image of s-mesoPdCu nanocube^[66]. Formic acid selectivities (B) and stability test at -0.4 V (C) for s-mesoPdCu, s-mesoPd, p-mesoPd and Pd NPs^[66]. Reprinted with permission. Copyright 2018, Chinese Chemical Society^[66]. (D) Schematic diagram of formic acid production over Pd nanocrystals enclosed by different crystal facets^[67]. (E) Free energy of electrochemical reduction of CO₂ to formic acid on Pd(111) (red), Pd(100) (violet), Pd(110) (blue), Pd₁₉ clusters (green) and Pd(211) (yellow)^[67]. (F) HCOO* and *CO generate free energy on Pd nanocrystals with different crystal facets^[67]. Reprinted with permission. Copyright 2016, American Chemical Society^[67].

Surface engineering on specific shaped NCs has been proven to be a successful strategy to trigger high electrocatalytic activity for CO₂ electroreduction [Figure 5D]^[67]. It has been accepted that the surface atomic arrangement plays an important role in affecting the stabilization energies of active intermediates^[89]. High-index facets possess a high density of low-coordinated sites, such as surface steps and kinks^[90,91]. Such a feature offers abundant unsaturated dangling bonds on the surface, which are highly active for electrocatalytic reaction and affect the free energies required for the formation of intermediate and are usually very active for electrocatalytic reactions^[90]. To promote the formation of *OCHO and inhibit *CO poisoning on the Pd surface during the CO₂-to-formic acid/formate conversion process, surface engineering on shaped NCs to create a rough surface with a mixture of high-index crystal facets may be preferable. By carrying out DFT calculation, Klinkova et al. elucidated that the binding energy of the HCOO* intermediates for formate product is more negative on high index facets, suggesting high-indexed facets with rough surface favored the HCOO* intermediates^[67]. Besides, as the crystal index increases, the reaction energy barrier is obviously decreased, as shown by the free energy diagrams in Figure 5E^[67]. According to the free energies of formation for HCOO* and CO* intermediates on different indexed facets [Figure 5F], high-indexed surface of Pd much more favors the formation of HCOO* rather than stabilizing CO* intermediate. However, the formation of CO* on the low-indexed surface shows a contrary trend. Besides, the findings from Bader partial atomic charges show that the charge density on the Pd atom bound to *CO significantly decreases as the surface index increases. Lower charge density suggests a weaker *CO binding strength on the high-index surface compared with low-index facets, implying less susceptibleness to CO intermediates. Projected density of states (PDOS) results further indicate that Pd (211) has stronger binding of HCOO* than that of Pd (111), further demonstrating more favorable adsorption for HCOO* on high-index surface^[67]. All these results jointly support the argument that a high-indexed surface of Pd has a higher resistance to CO poisoning and simultaneously favors HCOOH production via HCOO*. The authors further prepared Pd nanocubes, rhombic dodecahedra, NPs with mixed low-index facets, and branched NPs enclosed by high-index facets to verify the catalytic behavior. Among these, branched NPs exhibit the highest current density of 22 mA cm^-2 and formate FE of 97% at a low overpotential of -0.2 V. Computational simulations combined with electrochemical measurements verified that design and shape control of high-index surfaces is of great significance for improving intrinsic activity and selectivity toward CO₂ electroreduction.

Alloying

Alloying with other metals has been considered as a successful strategy to develop highly active and stable Pd-based electrocatalysts^[56,92,93]. Generally, the Pd-M alloy strategy not only reduces Pd content and the overall cost of electrocatalyst but also regulates the electronic structure of Pd based on the synthetic effect among the active components. Introducing a foreign metal into the host metal to form an alloy could trigger strong electron transfer, optimize electronic structures, and create more active sites, thus resulting in the altered adsorption energy for intermediates^[94-96]. More importantly, in practical ECO₂RR, the modified electronic structure of Pd helps to improve the reaction kinetics and alter the surface atom configurations, thereby optimizing the bonding strength for reaction intermediates. These advantages can change the activity and product selectivity of electrocatalysts and thereby optimize catalytic performance.

To date, some progress has been made in improving the CO₂ reduction activity of Pd-based catalysts via a bimetallic alloy system. Compared with single-component Pd catalysts, the bimetallic palladium alloy showed better electrocatalytic activity and stability due to the alloy effect. Of note, the choice of secondary metal especially decides the variation in electronic structure of Pd and corresponding binding energy of intermediates. In recent years, researchers have devoted tremendous efforts to finely designing high-performance Pd-based electrocatalysts through alloying. A variety of bimetallic Pd alloys with other transition metals, such as tin (Sn)^[31,97], silver (Ag)^[33,40,68], gold (Au)^[73,98,99], copper (Cu)^[67,100,101], bismuth (Bi)^[34,74], lead (Pb)^[102], etc., have been synthesized and displayed enhanced catalytic performance for CO₂-to-formic acid/formate conversion.

PdSn systems have been developed because Sn is a formate-selective metal and favorable to the formation of HCOO* intermediates^[103]. Bai et al. have prepared Pd-Sn alloy NPs by varying the feeding ratio of the Pd and Sn salts^[31]. In these series of Pd-Sn alloy NPs with different metal atomic ratios, the optimal sample PdSn/C with an equal molar ratio of Pd and Sn delivered an outstanding formic acid FE approaching 100% at the lowest overpotential of -0.26 V [Figure 6A and B]. A higher or lower Pd concentration in alloy leads to significant hydrogen evolution and thus lowers the formic acid selectivity and the catalytic activity. The relationship between the relative intensity ratio of Pd⁰/Pd^II and the mole ratio of Pd/Sn exhibits a valley-shaped profile. On the surface of metal NPs, PdSn/C has the lowest Pd⁰/Pd^II relative intensity ratio, implying a surface Pd-Sn-O layer derived from a certain amount of oxygen affinities on the PdSn alloy surface. DFT calculations revealed that it is favorable to form key reaction intermediate HCOO* and the production of HCOOH over the surface of PdSnO₂ whose Pd/Sn ratio is 1:1 [Figure 6C], which was matched well with the experimental results. The exclusive formation of formic acid can be attributed to optimized surface electronic structures induced with proper Pd-Sn-O layer. This study claimed that the PdSn alloy catalyst and its appropriate surface metal oxide configurations contributed to the favorable intermediates and CO₂-to-HCOOH pathway.

Figure 6. (A) Morphology characterization and size distribution of PdSn/C^[31]. (B) Product FEs and overpotentials over Pd/C, Sn/C and alloy Pd_xSn/C under applied potentials^[31]. (C) Energy calculation of CO₂ electroreduction to CO (top) and HCOO (bottom) on the surface of PdSnO₂^[31]. Reprinted with permission. Copyright 2017, WILEY^[31]. Formic acid selectivity (D), stability (E) and anodic oxidation peaks in LSV (F) of PdAg alloy with different chemical compositions^[40]. Reprinted with permission. Copyright 2020, WILEY^[40]. (G) XPS and XANES spectra of Pd₄Ag and pure Pd^[68]. (H) PDOS of Pd₄Ag and pure Pd^[68]. Reprinted with permission. Copyright 2020, WILEY^[68].

Ag, a well-known CO-generating metal, has an electron-rich state and intrinsic weak CO binding^[104]. During ECO₂RR, at high overpotentials, CO is inevitably generated and strongly adsorbs on the Pd surface, causing Pd deactivation and consequently affecting the activity and product selectivity. Therefore, it is essential to introduce a secondary metal on the catalyst that can weaken the CO binding strength to maintain the formate production. As such, PdAg alloy would be an ideal case. Thus, to alleviate CO poisoning, Zhou et al. prepared mesoporous PdAg nanospheres with controlled atomic percentages by using cationic surfactant^[40]. As seen in Figure 6D, among the samples, the best alloy candidate PdAg_2 composed of 62 wt% Pd and 38 wt% Ag exhibited impressive performance of superb formate FE close to 100% at zero overpotential and excellent stability. The improved CO tolerance of PdAg_2 nanospheres was reflected by long-time amperometric stability, CO stripping and anodic oxidation peaks in LSV [Figure 6E and F]. The electron redistribution of Pd and Ag atoms upon alloying well explained weakened CO surface affinity and the resultant high selectivity and excellent stability toward CO₂-to-formate conversion. Thereafter, Han et al. further developed composition-controlled PdAg alloy nanowires with high aspect ratios and defected surfaces^[68]. Benefiting from the electronic structure coupling with the one-dimensional structure with rough surface, the prepared Pd₄Ag nanowires achieved a high formate selectivity (90% FE at ≈ 0 V) and maintained the current density stable within 15,000 s at -0.43 V vs. RHE. Because of the higher electronegativity and lower work function of Ag relative to Pd, there is an electron transfer from Ag to Pd within the alloy, which was verified by X-ray photoelectron spectroscopy (XPS) and X-ray absorption near edge structure (XANES) spectra [Figure 6G]. In general, the d-band center is identified as a descriptor for electronic structure. With the cooperation of Ag, a down-shift of the d-band center occurs. The projected partial density of states (PDOS) of Pd₄Ag and pure Pd demonstrated the downshift of the d-band center with alloying with Ag, which was consistent with the experimental results [Figure 6H]. According to the DFT calculations, the production of the optimal d-band center mainly leads to the weakening of the binding of CO intermediates and Pd sites, thus promoting the formation of formate.

In addition, Pd-based electrocatalysts face a major challenge of a relatively narrow potential window for the selective production of formate. Cu is a metal that has a unique electronic property with moderate weakly *CO binding energy and smaller functions^[105,106], which are lower than those of Pd. What is more, it is a high oxyphilic metal that is readily oxidized to form the surface metal oxide configuration, generating the new oxygen-containing sites. From this perspective, in our recent work, we alloyed Cu into the parent metal Pd to fabricate PdCu bimetallic dendritic NPs [Figure 7A]^[69]. In this case, the dendritic architectures, the ensemble effect between Pd and Cu atoms, and the surface oxygen-containing features synergistically contribute to the overall catalytic performance. Among all the Pd_xCu_y alloy catalysts, Pd₈₂Cu₁₈ displayed the highest formate FE of 96.0% at -0.3 V vs. RHE, with a broad potential window of 600 mV for the high selectivity (> 90%) [Figure 7B and C]. Here, the dendritic structure in our work principally provides a structural platform to improve all of the Pd-based catalysts. The formate FE and potential range over mono-component Pd catalysts are also superior to those of other reported Pd electrocatalysts. The alloy effect from the modified electronic structure and the optimal surface O-containing configuration seems to be the dominant reason for the enhancement in the formate selectivity and potential window.

Figure 7. (A) Schematic diagram of formate production from a tunable composition PdCu bimetallic alloy catalyst based on dendrite structure^[69]. (B) Potential dependence of formate FE on Pd₈₂Cu₁₈, Pd₆₀Cu₄₀, Pd₅₁Cu₄₉ and pure Pd^[69]. (C) Comparison of potential windows of these four electrocatalysts with FE_formate above 90% and 80%, respectively^[69]. Reprinted with permission. Copyright 2022, American Chemical Society^[69]. (D) Schematic diagram of formic acid production on nanoporous Pd alloy (np-PdX) formed by electrochemical dealloying of Pd₁₅X₈₅ alloy (X = Co, Ni, Cu and Ag)^[70]. (E) Binding energies (eV) for CO combined with clean surfaces (CO*-1), CO with H*-containing surfaces (CO*-2), CO with 2H*-containing surfaces (CO*-3), H with clean surfaces (H*-1), H with H*-containing surfaces (H*-2), H with CO* containing surfaces (H*-3)^[70]. (F) Side view of representative models of the Pd-skinned Pd₃X alloy used for calculations^[70]. Reprinted with permission. Copyright 2019, American Chemical Society^[70].

The binding strength of *CO of Pd-based alloys is highly corrected with the selection of the secondary metal. To study this component-dependent catalytic behavior, Chatterjee et al. prepared free-standing, nanoporous Pd-X (np-PdX) alloys (X = Ni, Co, Ag, Cu) by electrochemical dealloying of Pd₁₅X₈₅ alloys^[70]. After dealloying, the formed core-shell structure is composed of Pd skin and Pd-X alloy bulk [Figure 7D]. Here, the secondary metal components near the Pd-skin surface not only modify the electronic property but also modulate H-solubility. Ideal active Pd electrocatalysts have both weak H and CO bindings, simultaneously promoting CO₂ hydrogenations and CO destabilization. Accordingly, to seek optimal surface with the lowest H and CO binding strength, the authors performed DFT calculations based on six kinds of models, including CO* on clean surfaces, CO* on hydrogenated surfaces (*H and 2 *H), H* on clean surfaces, and H* on hydrogenated surfaces, and H* on a preadsorbed *CO surface. As shown in Figure 7E and F, in the low-energy surface configuration, the Pd-skin/Pd₃Co system exhibits the lowest binding energies of H* and *CO, which is coincident with the experimental trend of deactivation rate: np-PdAg > np-PdCu > Pd/C > np-PdNi > np-PdCo. Owing to the relatively weak *H and *CO binding, the Pd-skinned np-PdCo and np PdNi displayed a high formate selectivity and CO tolerance. According to both experimental and calculated results, the secondary metal component in np-PdX alloys influences the CO binding strength and H-sorption capacity in the near-surface, which jointly affect the observed deactivation behavior of catalyst under electrolysis.

Heteroatom doping

Heteroatom doping, which refers to some nonmetal elements (such as H, B, N, and P) inserted into the lattice of host matrix and forms, has been extensively used to improve catalytic performance^[71,107,108]. In the aforementioned section of Pd-based alloy, the secondary metal atoms generally substitute some parent metal atoms and occupy the location of Pd atoms. However, unlike the traditional bimetallic alloys, the foreign heteroatoms of an appropriate concentration occupy the interstices of the host metal lattice due to the small atomic radii of these light elements. In this case, heteroatom interstices could tune the lattice parameters and modify the electronic structure. Moreover, because of the high electronegativity of the light elements, the heteroatoms can induce the polarization of the surrounding atoms and alter the band centers of the active sites. So far, such an approach of interstitial heteroatom doping in Pd bulk has been adopted in various electrocatalysis applications, including oxygen reduction reaction, methanol oxidation reaction, HER, and ECO₂RR^[71,109-111]. In particular, for electroreduction of CO₂ to formic acid/formate, designing heteroatom-doped Pd catalyst is a promising strategy to boost the electrocatalytic performance.

Hydrogen elements have the smallest atomic radii in the periodic table. Permeating hydrogen atoms into host Pd metal to form hydride could alter the electron density of Pd and thus change the binding strength of Pd with the reaction intermediates^[112]. Accordingly, absorption of interstitial hydrogen atoms is identified to be a highly attractive method to trigger catalytic performance toward formate production in ECO₂RR. Previous studies have suggested that absorption of interstitial hydrogen atoms can weaken *CO adsorption, inhibit the CO poisoning, and facilitate the formation of formic acid/formate^[38,39]. For example, Guo et al.synthesized a hydrogen-rich Pd hydride by adopting a one-step solvothermal synthesis route^[71]. After inserting in situ hydrogen originating from the decomposition of organic solvent on Pd surface, a Pd hydride catalyst with a hydrogen content of around 0.5 was produced. The hydrogen content is lower than that of β-hydride (PdH_x, x~0.7), suggesting a mixture of the α- and β-phases in PdH_0.5/C. Because of the strong interactions between Pd and H, hydrogen atoms were directly inserted into the interstitial sites of the Pd lattice, resulting in the lattice expansion. High-resolution transmission electron microscopy (HRTEM) images and X-ray diffraction (XRD) pattern results confirmed this phenomenon. Hence, the enlarged interplanar distance contributes to the downshift of the d-band center of Pd and the enhanced activity. Notably, PdH_0.5/C showed a formate FE of 93.1% at -0.4 V, far exceeding Pd/C catalyst. More importantly, CO poisoning study, CO stripping measurement, and stability tests certified that PdH_0.5/C had an outstanding CO tolerance, as shown in Figure 8A and B^[71]. Using isotopic analysis, the authors experimentally confirmed that lattice hydrogen participated in the formate generation pathway and elucidated the crucial roles of lattice hydrogen and surface adsorbed H* species in the HCOO^- formation pathway over PdH_0.5/C [Figure 8C]^[71]. In addition, the authors also proposed that the mutual diffusion behavior between the lattice hydrogen and the surface H* species derived from the proton species in electrolyte favored keeping a stable hydrogen-rich equilibrium composition during CO₂ electroreduction. This lattice hydrogen participated mechanism well explained the broadened electrochemical window for formate generation and suppressed formation toward CO.

Figure 8. (A) CO stripping measurement of PdH_0.5/C and Pd/C^[71]. (B) CO tolerance tests for PdH_0.5/C at -0.4 V vs. RHE in CO₂-saturated 0.1 M KHCO₃ electrolyte^[71]. (C) Reaction mechanism for HCOO^- generation on PdH_0.5/C catalysts^[71]. Reprinted with permission. Copyright 2022, Elsevier^[71]. (D) Calculated surface adsorption energies (eV) of HCOO* and *COOH on Pd(111) and Pd(111)-4B with and without H adsorption^[72]. (E) Potential dependence of mass normalized formate concentration on Pd-B/C catalysts^[72]. Gibbs free energy diagrams for the intermediates in formate pathway on Pd-B/C (red) and Pd/C (black) with (F) and without (G) hydrogen adsorption^[72]. Reprinted with permission. Copyright 2018, American Chemical Society^[72].

In addition, Boron (B) insertion in the Pd lattice to form Pd-B alloy becomes an attractive way to promote ECO₂RR with a gratifying catalytic performance^[113,114]. Boron, the lightest metalloid element in the IIIA group, has a relatively small electronegativity compared with other nonmetal elements. It has been demonstrated that B doping has a great beneficial effect on enhancing catalytic performance and hampering CO accumulation on Pd surface^[109,115]. Taking advantage of the unique properties of B elements, Jiang et al. synthesized a boron-doped Pd catalyst (Pd-B/C) via a treatment method^[72]. XRD patterns of Pd-B/C catalyst shifted negatively relative to those for Pd/C, indicative of a lattice expansion after doping B into Pd. XPS analysis revealed the electron transfer between Pd and B, which resulted in the downshift of the d-band center of Pd. With the synergy of optimized electronic structure by B, B-Pd interstitial NCs achieved a formate selectivity of 70% and a formate concentration of ca. 234 mM mg^-1 Pd after 2 h of electrolysis at the potential of -0.5 V, outperforming Pd/C without B doping. The enhancement in formate selectivity and yield was ascribed to the B-doping effect, which was further confirmed by DFT calculations. It was found that, compared with Pd/C, B-doped Pd made the adsorption energy of HCOO* shift negatively and the Gibbs free energy for intermediates reduced [Figure 8D]^[72]. This suggests that the presence of B is more favorable for the HCOO* intermediates and formate pathway than the *COOH and CO pathway. On the basis of experimental observations and DFT calculations with and without B doping, the authors demonstrated the positive influence of the B doping on the formate pathway of CO₂RR [Figure 8E-G]^[72]. It is worth noting that Pd-B/C is a promising bifunctional catalyst for formic acid-CO₂ interconversion.

Strain effect

Surface strain widely exists in heterogeneous catalysis, and it is an effective method to modulate the electronic structure and thus influence its catalytic performance^[116-118]. The strain effect, a long-range effect arising from lattice mismatch, reflects the lattice distortion and the change in the metal bond. According to the differences in lattice constants of metals in the shell/surface and in the bulk form, it can be classified into tensible and compressive strain. A tensible shell with an enlarged interatomic distance is produced when introducing a foreign metal with a larger lattice constant than that of Pd^[119]. On the contrary, introducing a metal with a smaller lattice constant would generate a compressive-strained shell with a shortened interatomic distance^[120]. The interatomic distance displays a direct impact on the d-band center of host metal. It is well established that the d-band center is widely used to describe the change of electronic structure of catalyst^[121]. The above-mentioned subtle changes in atomic distance of the metals result in the surface atoms being either enlarged or compressed, leading to a shift in the d-band center and, thus, a variation in binding strength toward intermediates. The former causes an upshift of the d-band center, strengthening the binding strength, whereas the latter induces a downshift of the d-band center, weakening the interaction. Encouraged by the strain effect, constructing effective Pd nanocatalysts by altering the atomic spacing of surface atoms to tune the binding strength toward intermediates is welcome for promoting ECO₂RR^[122].

Recently, Bok et al. used tensile strain effect from the confined AuPd NPs to boost CO electroreduction to formate^[73]. They synthesized atomically dispersed Au on tensile-strained Pd NPs anchored in the confined space of metal-organic frameworks (MOFs) via a confined reduction method [Figure 9A and B]. By analyzing the lattice spacing from aberration-corrected transmission electron microscope (Cs-TEM) image, the authors confirmed the existence of tensile strain and lattice distortions in M-AuPd(x), where x represents the ratio of Au. The tensile strain induced in the Au atom-doped Pd NPs originates from the metal-MOF interaction, as shown in Figure 9C and D. With the precise adjustment of strain effect, a remarkable enhanced catalytic activity was found over M-AuPd(20), reaching an outstanding formate FE over 99% at -0.25 V vs. RHE [Figure 9E]. Besides, the current density and mass activity of M-AuPd(20) was increased by 26 and 31 times compared with Pd/C. DFT calculations enclosed the origin for the enhanced electrocatalytic performance for formate production. The synergy between tensile strain effect in Pd surface and electronic effect from atomically dispersed Au incorporation contributes to the improved electrocatalytic performance in terms of activity, formate selectivity and stability. In the electrocatalysis process, these two effects generally function together synchronously to break the scaling relationship of COOH* vs. CO* intermediates and, thus, ultimately determine the catalytic behavior. On the one hand, the generating tensile strain in Pd surface stabilizes the intermediate species COOH* and HCOO* and facilitates CO₂ activation. On the other hand, dispersed Au atoms destabilize CO* to suppress CO poisoning without affecting other intermediates. This work highlights the important role of strain effect in promoting electrocatalysis for formate production from CO₂.

Figure 9. (A) Tensile strain effects generated on AuPd alloy nanoparticle through MOF-confined reduction^[73]. (B) Schematic diagram of the preparation of M-AuPd(x) catalyst^[73]. (C) Representative Cs-TEM image of M-AuPd(20)^[73]. (D) Intensity profile measured from the area indicated by the colored rectangular box from (C)^[73]. (E) Comparison of formate Faraday efficiency of Pd/C, M-Pd and M-AuPd(20)^[73]. Reprinted with permission. Copyright 2021, American Chemical Society^[73].

Phase control

For Pd-based electrocatalysts, phase control has recently emerged as a promising tactic to optimize their catalytic performance. Solid-solution alloy with disordered atomic arrangement is the most common type of metal alloy^[123,124]. Compared to disordered solid solution alloy, ordered intermetallic compound has a specific atom stoichiometric ratio and long-range atomic ordering, leading to strong interatomic interaction^[125]. Even though they have the same stoichiometry as random alloys, the physico-chemical properties of intermetallic compounds are significantly distinct. Profiting from the features, intermetallic compounds exhibit unique properties and great potential in various applications^[91,126,127]. Consequently, it is particularly appealing for ECO₂RR to construct robust Pd-based multimetallic NCs through phase engineering. More importantly, in-depth study of phase-dependent activity and selectivity of catalyst is of great significance for electrocatalyst design.

Jia et al. synthesized Pd₃Bi intermetallic alloys (IMAs) with a fixed stoichiometry of 3/1 and uniform sizes by adopting a facile solvothermal method^[74]. It is well accepted that a solvothermal method provides a versatile approach to synthesizing intermetallic compounds. The ordered NCs can be transformed into a random solid solution alloy (Pd₃Bi-SSA) through thermal annealing at 500 °C, as illustrated in Figure 10A. By manipulating the atomic ordering, the phase-performance relationship was investigated. Impressively, electrochemical measurement for ECO₂RR of Pd₃Bi IMAs holds an excellent formate selectivity of approximately 100% and high stability even at a bias of < -0.35 V, exceeding vast majority of Pd-based materials reported in the literature^[128,129]. By contrast, Pd₃Bi-SSA is greatly inferior to Pd₃Bi IMAs in terms of formate selectivity and CO poisoning resistance. To reveal the influence of atomic ordering on the catalytic performance toward formate generation, DFT simulations were conducted. According to the PDOS analysis, adsorption energies of key adsorbates, and energetic trend, the favorable formation and stabilized adsorption of *OCHO on Pd₃Bi IMAs was validated. Experimental observations, coupled with theoretical calculation, corroborate that ordered phase endows PdBi alloy with enhanced CO₂RR performance. This work emphasizes the important role of phase engineering in influencing catalytic behavior and electrochemical CO₂-to-formate activity.

Figure 10. (A) Schematic diagram of the synthesis of Pd₃Bi-IMA and Pd₃Bi-SSA^[74]. Reprinted with permission. Copyright 2021, WILEY^[74]. (B) Schematic diagram of the synthesis of fully ordered (o-), partially ordered (p-) and disordered (d-) PdBi NS^[34]. (C) Formate selectivity at different applied potentials^[34]. Reprinted with permission Copyright 2022, Elsevier^[34].

Likewise, taking advantage of phase engineering, Xie et al. reported intermetallic PdBi NSs with ordered hexagonal phase as electrocatalysts for ECO₂RR^[34]. In the synthetic process of PdBi NSs, the ordering degree of Pd and Bi atoms in crystallographic structures can be finely tuned by controlling the synthesizing temperature. To accurately evaluate the structure-ECO₂RR performance relationship of PdBi NSs, they prepared three types of Pd-based NSs for comparison, as shown in Figure 10B. PdBi alloy with completed ordered phase was found to be exceptionally selective and active for producing formate. Compared to disordered PdBi NSs, partially ordered PdBi NSs, and pure Pd NSs, intermetallic PdBi NSs delivered a high formate FE of 91.9% at -1.0 V and a wide potential window [Figure 10C]. DFT calculations indicated that a free energy barrier for *OCHO adsorption on the Bi surface in fully ordered PdBi NSs is 0.13 eV, lower than that in disordered PdBi NS, which is mainly responsible for the enhanced catalytic performance.