Computational Prediction of Graphdiyne-Supported Three-Atom Single-Cluster Catalysts

Open AccessCCS ChemistryRESEARCH ARTICLE22 Apr 2022Computational Prediction of Graphdiyne-Supported Three-Atom Single-Cluster Catalysts Jin-Cheng Liu, Hai Xiao, Xiao-Kun Zhao, Nan-Nan Zhang, Yuan Liu, Deng-Hui Xing, Xiaohu Yu, Han-Shi Hu and Jun Li Jin-Cheng Liu Department of Chemistry, Key Laboratory of Organic Optoelectronics and Molecular Engineering of the Ministry of Education, Tsinghua University, Beijing Google Scholar More articles by this author , Hai Xiao Department of Chemistry, Key Laboratory of Organic Optoelectronics and Molecular Engineering of the Ministry of Education, Tsinghua University, Beijing Google Scholar More articles by this author , Xiao-Kun Zhao Department of Chemistry, Key Laboratory of Organic Optoelectronics and Molecular Engineering of the Ministry of Education, Tsinghua University, Beijing Google Scholar More articles by this author , Nan-Nan Zhang Department of Chemistry, Key Laboratory of Organic Optoelectronics and Molecular Engineering of the Ministry of Education, Tsinghua University, Beijing Google Scholar More articles by this author , Yuan Liu Department of Chemistry, Key Laboratory of Organic Optoelectronics and Molecular Engineering of the Ministry of Education, Tsinghua University, Beijing Department of Chemistry, Southern University of Science and Technology, Shenzhen, Guangdong Google Scholar More articles by this author , Deng-Hui Xing Department of Chemistry, Key Laboratory of Organic Optoelectronics and Molecular Engineering of the Ministry of Education, Tsinghua University, Beijing Google Scholar More articles by this author , Xiaohu Yu Shaanxi Key Laboratory of Catalysis, School of Chemical and Environment Sciences, Shaanxi University of Technology, Hanzhong, Shaanxi Google Scholar More articles by this author , Han-Shi Hu Department of Chemistry, Key Laboratory of Organic Optoelectronics and Molecular Engineering of the Ministry of Education, Tsinghua University, Beijing Google Scholar More articles by this author and Jun Li *Corresponding author: E-mail Address: [email protected] Department of Chemistry, Key Laboratory of Organic Optoelectronics and Molecular Engineering of the Ministry of Education, Tsinghua University, Beijing Department of Chemistry, Southern University of Science and Technology, Shenzhen, Guangdong Google Scholar More articles by this author https://doi.org/10.31635/ccschem.022.202201796 SectionsSupplemental MaterialAboutAbstractPDF ToolsAdd to favoritesDownload CitationsTrack Citations ShareFacebookTwitterLinked InEmail While heterogeneous single-atom catalysts (SACs) have achieved great success in the past decade, their application is potentially limited by their simplistic single-atom active centers, which make single-cluster catalysts (SCCs) a natural extension in the domain of heterogeneous catalysis. SCCs with precise numbers of atoms and structural configurations possess SAC merits, yet have greater potential for catalyzing complex reactions and/or bulky reactants. Through systematic quantum-chemical studies and computational screening, we report here the rational design of transition metal three-atom clusters anchored on graphdiyne (GDY) as a novel kind of stable SCC with great promise for efficient and atomically precise heterogenous catalysis. By investigating their structure and catalytic performance for the oxygen reduction reaction, the hydrogen evolution reaction, and the CO2 reduction reaction, we have provided theoretical guidelines for their potential applications as heterogeneous catalysts. These GDY-supported three-atom SCCs provide an ideal benchmark scaffold for rational design of atomically precise heterogeneous catalysts for industrially important chemical reactions. Download figure Download PowerPoint Introduction In the past decade, single-atom catalysts (SACs) have emerged as the new frontier of heterogeneous catalysis, owing to their high performance with regard to atomic efficiency, selectivity, stability, and activity, as well as precisely tunable quantum states through support manipulation.1–5 However, SACs are not always an optimal design for complex reactions, such as those that require multistep redox reactions (e.g., in photosynthesis and nitrogenase), interaction among two or more adsorbed bulky reactant molecules, or multiple functional sites (e.g., in order to break the cumbersome scaling relations).6 Atomic clusters that contain only a small number of atoms can exhibit unique and often unexpected properties for catalytic reactions such that SAC may not work well.7–9 The term “cluster” was coined by F.A. Cotton in the early 1960s to refer specifically to compounds containing metal–metal bonds. Up to now, there are many synthetic strategies of supported atomic clusters for heterogeneous catalysis, for instance, gas-phase mass filters or so-called “soft landing,”10 the precursor-preselected strategy from the confined effect of zeolitic or MOF frameworks,11–13 the host–guest strategy,14 the wet chemical reduction,15 the dendrimer-based strategy, and so on.16 Recently, single-cluster catalysts (SCCs) with atomically precise active centers composed of well-defined stable clusters with constant atomic constitutions and structures have been proposed as a natural extension of SACs for optimal design of complicated heterogeneous catalysts.6,10,11,14,17–21 However, the stability and thus the synthesis of SCCs pose a grand challenge because a delicate balance is required to prevent both further aggregation of the clusters to form large size clusters or nanoparticles and their dispersion to form supported single atoms. To form robust SCCs, a prototype material with natural pores or defect-anchoring sites is necessary. The synthesis of graphdiyne (GDY) by Li et al.22–26 presents an ideal substrate for hosting both SACs and SCCs since GDY has natural 6-membered rings (6MRs) and 18-membered rings (18MRs). There have been reports of metal/GDY complexes with various applications,27,28 and the 18MR-hole of GDY has been shown to provide a suitable site for anchoring a metal (M) single atom (SA) or single cluster (SC) as a heterogeneous catalyst.29 However, most previous work on Mx/GDY has focused on SAs. For example, nonnoble metal Fe and Ni SAs anchored on GDY (denoted as Fe1/GDY and Ni1/GDY) have been shown to perform better in the hydrogen evolution reaction (HER) than in the commercial Pt/C.29 By first-principles calculations, the stability and electronic structures of M1/GDY with 3d metals (M = Sc − Zn) were systematically investigated in our group.30 Additionally, Mo1/GDY,31 Ir1/GDY,32 W1/N-doped GY,33 Pt1/GDY,34 Fe1/GDY,35 AM1/GDY,36 and TM1/GDY37 (AM = alkali metal and TM = transition metal) were theoretically proposed as good catalysts for reactions including nitrogen fixation, CO oxidation, HER, oxygen reduction reaction (ORR), and water splitting. However, these studies on M1/GDY did not explore the possibility that the SA form of metal on GDY might be less stable than its SC form, particularly when compared with the highly stable triatomic cluster form.38,39 Ma et al.40 reported diatomic cluster catalysts on GDY for nitrogen reduction reaction. Zhang et al.41 speculated that the triangular 18MR-hole of GDY can accommodate three Li atoms at the three symmetric corners with a unique triangular configuration, and the resulting Li3/GDY can be used as anode material for lithium ion batteries. Qi et al.42 investigated the performance of Pd clusters on GDY for catalytic reduction but did not characterize the structure of Pd clusters. Very recently, we showed that the Os3/GDY and its analogs are a class of potential catalyst for selective semihydrogenation of acetylene.43 Moreover, we predicted that the M3 form is indeed the most stable for Pt and Ni supported on GDY and suggested an efficient strategy based on the electrochemical potential window (EcPW) to prepare them via an electrochemical route.39 The metal trimer SCCs have been reported to deliver excellent performance. Based on Ji et al.’s reported experimental work, the Ru3 cluster supported on N-doped carbon material was shown to be an efficient catalyst for selective oxidation of alcohols.11 The Ag3 cluster on alumina support was demonstrated with high activity and selectivity for direct propylene epoxidation.18 The [Cu3(μ-O)3]2+ cluster in mordenite was shown to exhibit high reactivity towards activation of inert C–H bonds in methane.21,44 By first-principles calculations, we predicted that the Fe3 cluster supported on Al2O3 leads to an associative mechanism for low-temperature ammonia synthesis with a high turnover frequency.20 However, in addition to the EcPW strategy, our group suggested efficient and specific ways to prepare stable metal trimer SCCs that are still lacking because of the delicate requirement for the interaction between metal and support. On the one hand, when the metal-support bonding is much stronger than the metal–metal bonding, the metal trimer SC will dissociate to form SAs. On the other hand, if the metal-support interaction is much weaker than the metal–metal bonding, the metal trimer SCs will aggregate into bigger clusters or nanoparticles. Thus, it is a prerequisite for a support for hosting the metal trimer SCC to balance the metal–metal and metal–support bonding strengths. In this work, we investigate the viability of GDY as a support for TM trimer SCCs from both thermodynamic and kinetic aspects by first-principles calculations. All in all, we have considered 13 late TM elements for M3/GDY. The geometries and electronic structures of M3/GDY have been further analyzed by taking Cu3/GDY and Pt3/GDY as two typical examples. Finally, we investigated the catalytic performance of these M3/GDY SCCs for three kinds of key reactions: ORR, HER, and CO2 reduction reaction (CO2RR). The computational results thus provide guidelines for their practical applications as heterogeneous catalysts. Computational Details All density functional theory (DFT) calculations were performed with the plane-wave basis sets of 400 eV cutoff kinetic energy to approximate the valence electron densities and projector-augmented wave method to account for the core–valence interaction,45 as implemented in the Vienna Ab initio Simulation Package (VASP) code.46,47 The spin-polarized Kohn–Sham formalism with gradient-corrected exchange and correlation functional of the Perdew–Burke–Ernzerhof (PBE) flavor was adopted.48 The Γ-point-only sampling was used for the Brillouin zone integration for the GDY(2 × 2)-based models, which were adopted for energy calculations. And a 3 × 3 × 1 Brillouin zone grid sampling was used for the GDY(1 × 1)-based models, which were adopted for electronic structure analysis. All atoms as well as the lattice parameters a and b were allowed to relax for geometry optimization. The optimized lattice constant for pristine GDY that we got is | a| = | b| = 9.46 Å, in good agreement with the previously reported value of 9.48 Å by Long et al.49 The convergence criteria were set to be 10−6 eV and 0.01 eV/Å for wavefunction and geometry optimization, respectively. Free energy correction for all species was performed for ORR, HER, and CO2RR reactions by VASPKIT.50 For free molecules, the ideal gas approximation was assumed. For adsorbates, the contributions from all degrees of freedom to the free energies were treated as vibrations under the harmonic approximation, with unphysically low frequencies reset to a threshold of 60 cm–1, which corresponds to the acoustic translational mode of the six-membered rings in water bulk.51,52 All electrochemical calculations were based on Computational Hydrogen Electrode (CHE) model.53 For ORR, we shifted the chemical potential of the electrons by the equilibrium potential of U = 1.23 eV [vs standard hydrogen electrode (SHE)], corresponding to the situation where the fuel cell has the maximum potential allowed by thermodynamics. Ab initial molecular dynamics (AIMD) simulations were carried out for all M3/GDY(2 × 2). The AIMD calculations were started with the optimized configurations with lattice parameters fixed and were performed for more than 15 ps with a time step of 1 fs. The canonical (NVT) ensemble and Nosé-Hoover thermostats were used with the temperature set to 300 K.54,55 AIMD annealing was performed for each Mx/GDY (M = Fe, Co, Ni, Cu, Ru, Rh, Pd, Ag, Os, Ir, Pt, Au; x = 1–20, 30) structure. Mx clusters were deposited onto the GDY surface by simulated annealing from 1000 to 100 K in 10 ps, followed by structure optimization. This modeling process was carried out to mimic any experimental deposition procedure that does not necessarily impose precise control on the chemical potential for the target metal element, such as the traditional impregnation and coprecipitation methods. Phonon dispersion was calculated using density-functional perturbation theory,56 as implemented in the VASP and analyzed by interfacing with the Phonopy code.57 Band structures of GDY, Cu3/GDY, and Pt3/GDY were computed along the special line of Γ (0, 0, 0) → M (0.5, 0.5, 0) → X (0, 0.5, 0) → Γ (0, 0, 0) at both PBE and Heyd–Scuseria–Ernzerhof (HSE06) levels.58 Band decomposed charge densities were calculated at the Γ point for both the conduction band minimum (CBM) and valence band maximum (VBM) with degenerate bands summed up. Real space wavefunctions of CBM and VBM at the Γ point were extracted by VASPKIT.50 The crystal orbital Hamilton population (COHP) analysis was performed with the LOBSTER 3.1.0 package, which reconstructs the orbital-resolved wavefunctions via projection of the delocalized plane waves to localized atomic-like basis sets.59,60 The fragment molecular orbital (MO) analysis was performed using spin-restricted DFT with PBE and Slater basis sets of triple-zeta with two polarization functions (TZ2P) as implemented in the Amsterdam Density Functional (ADF) program.61 The frozen core approximation was applied to C[1s2] and Cu[1s2–2p6]. Relativistic effects were introduced by the zero-order regular approximation method. The optimized Cu3/GDY molecular counterpart and its corresponding fragments (Cu3 and GDY) were constrained to the D3h symmetry. Results and Discussion Stability of M3/GDY Scheme 1a shows the structure of pristine GDY (the details are listed in Supporting Information Table S1), in which the length of the Cc≡Cd bond (and its symmetric equivalents) is 1.227 Å, close to that in acetylene (1.20 Å), indicating a typical triple-bond character. There are six C≡C bonds bordering the 18MR-hole of GDY, but only two of them are involved in coordination with metal SA in reported M1/GDY cases (Scheme 1b).30 Meanwhile the remaining space of an 18MR-hole may accommodate two additional metal atoms coordinated by the rest of the four triple bonds to form M3/GDY (Scheme 1c). Thus, we investigated the stability of M3/GDY by comparing the average binding energies Ebind (Ebind = [E(Mx/GDY) − x · E(M, bulk) − E(GDY)]/x, where x is the number of atoms composing the anchored cluster of metal clusters on GDY, with M covering groups VIII and IB TM elements since they are usually considered as good catalyst candidates. Scheme 1 | Schematic illustrations of (a) GDY, (b) M1/GDY, and (c) M3/GDY. Download figure Download PowerPoint M1/GDY can hardly be the most stable case. Instead, it is most unstable for Fe, Co, Cu, Ru, Rh, Os, Ir, Pt, and Au on GDY when compared with other-sized clusters (Figure 1 and Supporting Information Figures S6, S7, S14, and S15). As expected, M3/GDY is the most stable case for Mx/GDY (x = 1 − 10) cases except Ag, Os, and Au. Taking Cu as an example, the Ebind of Cu1 on GDY is 1.57 eV, but it decreases dramatically to the lowest value of 0.81 eV for Cu3. When adding one more Cu atom to Cu3/GDY, Ebind increases to 1.06 eV. Ebind peaks at Cu5 with a value of 1.18 eV and starts to decrease to 0.90 eV at Cu20, due to the formation of more metal–metal bonds. Thus, x = 3 becomes a magic number for the thermodynamic stability of these GDY-supported metal clusters. To determine the thermodynamical difference between each cluster from a metal particle, we add a dashed line as reference to M30/GDY, which represents a typical ∼1 nm nanoparticle, in each subpanel of Figure 1. The thermodynamically stable M3/GDY systems lie lower than the dashed line, whereas the metastable or unstable M3/GDY are above this dashed line. Indeed, Fe, Co, Ni, Cu, Rh, and Pt, are relatively stable, but Ru, Pd, Ag, Os, Ir, and Au are not as stable as supported large particles such as M30/GDY. The bulk limit of adding an atom of a large metal particle is 0 eV as shown in Figure 1 because the equation of Ebind refers to the average energy per atom in bulk metal, E(M, bulk). Figure 1 | The average binding energies (Ebind) relative to bulk limit, Ebind = [E(Mx/GDY) − x · E(M, bulk) − E(GDY)]/x, of metal clusters with 1–20 and 30 atoms on GDY. E(Mx/GDY) and E(GDY) is the energy of GDY-supported metal clusters and GDY respectively. x is the number of atoms composing the anchored cluster. E(M, bulk) is the average energy per atom in bulk metal. The black curve shows the Ebind changes from M1 to M30. The points for M3/GDY are marked in red circles, and their structures are shown in the insets. The dashed line is the reference Ebind of M30/GDY, which represents typical ∼1 nm nanoparticle. The results show that Fe, Co, Ni, Cu, Rh, and Pt are relatively stable whereas Ru, Pd, Ag, Os, Ir, and Au are not as stable as supported large particles such as M30/GDY. Download figure Download PowerPoint AIMD simulations and phonon dispersions of all M3/GDY cases further characterize their kinetic stability ( Supporting Information Figures S1–S5 and S14). In the 15 ps AIMD trajectories, all M3 clusters remain at the 18MR, except for Ag3/GDY and Au3/GDY. The root-mean-square deviations of M3 clusters are all at low levels with small fluctuations, indicating that no diffusion and decomposition occur within 15 ps. The phonon spectrum, showing no imaginary frequency, further confirms the kinetic stability of Cu3/GDY ( Supporting Information Figure S5). The outstanding stability of M3/GDY enables the possibility of synthesizing them with the reported EcPW strategy, due to the presence of chemical potential windows that distinguish the trimer cluster form from the other-sized forms anchored on the GDY support.39 The stability of M3/GDY originates from the specific interactions between the metal trimer clusters and the GDY support, which can be analyzed and elucidated from geometries and electronic structures discussed in the following sections. Geometries and electronic structures of M3/GDY Only with M3 in the plane of GDY may M3/GDY retain the D3h symmetry, but most M3 cannot fit into the 18MR-hole. The optimized M3/GDY structures show that only Cu3/GDY is of D3h symmetry, and the rest of the M3/GDY structures have distortions in both the GDY substrate and M3 cluster. We summarize the three types of distortions as shown in Figure 2. Type I distortion is with the rotation of both C≡C bonds and M3 within the GDY plane. Type II distortion is with the out-of-plane rotation of M3. Type III distortion is also with M3 moving out of plane but with one metal atom detached from GDY. The cases of Mn, Fe, Ni, Co, Pd, and Pt belong to Type I distortion, which is of C3h symmetry. All type I structures locate in the stable region as shown in Figure 1. This distortion is due to either the mismatch between the 18MR-hole and M3 or the requirement for specific coordination. The cases of Ru, Rh, Os, and Ir belong to type II distortion, which has no local symmetry. The cases of Ag and Au belong to type III distortion, in which M3 is out of the GDY plane with only two atoms coordinated to GDY and the remaining atom pushed out and M3 almost vertical to the GDY plane. The M–M bond lengths of Au3 and Ag3 are 2.688 and 2.669 Å, respectively, which are too large to fit in the 18MR. For type II and III structures, except Rh3/GDY, all of them are relatively unstable. Compared with type I distortion, types II and III are more mismatched between the 18MR-hole and M3 and are more unstable. Figure 2 | Type I, II, and III distortions of structures of M3/GDY. Type I is the rotation in the GYD plane with local C3h symmetry (M = Mn, Fe, Ni, Co, Pd, and Pt). Type II trimers break the mirror symmetry but also bond with 18MR-hole (M = Ru, Rh, Os, and Ir). Type III is out of the GDY plane configuration (M = Ag and Au). Download figure Download PowerPoint We further investigate the electronic structures of GDY, Cu3/GDY, and Pt3/GDY as representative cases. The electron localization function (ELF) and electrostatic potential maps of GDY (Figures 3a–3c) show that the six triple bonds of 18MR-hole provide localized π electrons (in the GDY plane) at the border, offering perfect anchor sites for TM atoms. In Cu3/GDY (Figures 3d–3f), each Cu atom indeed forms two d-π coordination bonds with the localized π systems of C≡C bonds in the GDY plane, leaving the π systems perpendicular to the GDY plane intact. And the three 4s orbitals of Cu3 form a three-center two-electron (3c-2e) bond with the bonding electrons localized above the center of Cu3 (Figure 3f), which renders the Cu3 center as a potential nucleophilic SCC. For Pt3/GDY, in addition to the d-π coordination, ELF shows a localized red region between Pt and Cc (Figure 3g), implying a typical d-σ covalent bond. Meanwhile, the hybridization of Cc changes from sp to sp2-like. ∠ ( C d – C e – C f ) decreases from 180° to 145.7° with Ce going away from the Pt3 cluster (Figures 3h and 3i). Figure 3 | The 2D contour plots of ELF for (a and b) GDY, (d and e) Cu3/GDY, and (g and h) Pt3/GDY at the GDY plane and 1.5 Å above it, respectively. The red/blue region represents electrons accumulation/dispersal. (c, f, and i) The electrostatic-potential-colored charge density isosurfaces at 0.02 |e|/bohr3 for (c) GDY, (f) Cu3/GDY, and (i) Pt3/GDY, where the red region is dominated by electrons, and the blue region is dominated by nuclei. Download figure Download PowerPoint The band structures (Figures 4a–4i and Supporting Information Figure S9) show that the direct band gap of GDY is calculated to be 0.49 and 0.93 eV at PBE and HSE06 levels, respectively, in agreement with previous work.49,62 The VBM of GDY is simply composed of the conjugated π system formed by the C≡C bonds and benzene rings, with the corresponding π* orbitals composing the CBM. In Cu3/GDY, two new bands appear between the original VBM and CBM, which are composed of the 3c-2e bond by Cu 4s orbitals and the d-π anti-bonding orbitals. The rest of the 3d bands of Cu3 are very narrow and localized between −2.0 to −3.2 eV below the Fermi level. The original CBM bands of GDY are partially occupied, due to the charge transfer from Cu3 to GDY, indicating the metallic character of Cu3/GDY. The Bader charge of Cu3 is +1.21 |e|, indicating that one 4s electron of Cu3 is donated to GDY. Thus, the Cu3 cluster is of +1 oxidation state, and the left two 4s bonding electrons forming the 3c-2e bond satisfy the Hückel [4n + 2] electron-counting rule, which is a typical aromatic feature. The analysis of MO interactions between GDY and Cu3 is presented in Supporting Information Figure S10, which leads to similar conclusions. Figure 4 | The band structures (at Γ-M-K-Γ symmetry points) and corresponding projected densities of states at PBE level for (a and b) GDY, (d and e) Cu3/GDY, and (g and h) Pt3/GDY. The red curves are contributed by C, and the blue curves are contributed by Pt. The isosurface of partial charge density contours of (c) VBM and CBM of GDY, (f) VB (1-6) of Cu3/GDY, and (i) VBM and CBM of Pt3/GDY. Download figure Download PowerPoint For Pt3/GDY, the direct band gap is 0.56 and 0.63 eV at PBE and HSE06 levels, respectively. Different from Cu3/GDY, there is no new band between the VBM and CBM of GDY. Instead, the 5d orbitals of Pt3 are mixed into the GDY bands, as shown in the projected band structure and density of states. Such mixing leads to charge transfer from Pt3 to GDY and strong bonding between Pt and C, reducing the C≡C bonds of GDY into double bonds. The integrated crystal orbital Hamilton population (ICOHP) of C≡C bonds is reduced from −15.39 to −12.84 eV, and the ICOHP of Cd–Ce single bonds increases from −11.70 to −12.72 eV ( Supporting Information Figure S11). Thus, the bond strengths along the linkage become uniform, which is consistent with the bond length and angle analysis ( Supporting Information Figure S8). Catalytic performance Because of low oxidation states of the metal atoms, the metal clusters of M3/GDY are electron-rich and thus are good candidate catalysts for reduction reactions ( Supporting Information Figure S12). In addition, the GDY substrate can serve as an electron reservoir to buffer the oxidation state change of the M3 cluster during the catalytic reactions.63 Here, we investigate the ORR, HER, and CO2RR to demonstrate the catalytic capability of M3/GDY as a novel series of SCCs. ORR There are two distinct pathways for ORR.53,64 One is the associative mechanism, where O2 is first reduced to *OOH. The other is the dissociative mechanism, where O2 dissociates first. The associative mechanism : O 2 + H + + e − → OOH * The dissociative mechanism : O 2 → 2 O * , followed by 2 O * + H + + e − → OH * + O * We find that on all M3/GDY, the dissociative adsorption of O2 is thermodynamically favored over the molecular adsorption, and the *OOH species is thermodynamically unstable with respect to *O + *OH. Thus, the dissociated mechanism dominates on M3/GDY, and the four electrochemical steps are: ( 1 ) O 2 + H + + e − → OH * + O * ( 2 ) OH * + O * + H + + e − → O * + H 2 O ( 3 ) O * + H + + e − → OH * ( 4 ) OH * + H + + e − → H 2 O Note that the dissociation of O2, which is a chemical step, is integrated into step (1) because we only consider the electrochemical steps here, based on the method by Nørskov et al.53 Figure 5b summarizes the linear relationship between the reaction-free energy ΔG for each electrochemical step and ΔG*O. For step (1), its ΔG increases as the ΔG*O increases. But for the rest of the steps, their ΔG’s have negative slopes with respect to ΔG*O. Figure 5 | (a) The free energy diagrams for ORR at U = 1.23 V (vs SHE) on M3/GDY. (b) The linear relationship between ΔGO* and ΔG for each electrochemical step on M3/GDY. ΔG is the energy change of each ORR elementary electrochemical steps at U = 0 V (vs SHE). Download figure Download PowerPoint Step (3) on all M3/GDY has the highest ΔG as shown in Figure 5a, and thus it is always the potential limiting step (PLS) that determines the overpotential. Only the overpotential of Au3/GDY, Ag3/GDY, Cu3/GDY, Pt3/GDY, and Pd3/GDY are lower than 1 V. Considering that Au3/GDY and Ag3/GDY are not stable, only Cu3/GDY, Pt3/GDY, and Pd3/GDY are most likely to be the potential catalysts for ORR with an overpotential of 0.99, 0.94, and 0.70 V, respectively. Then we compare the Cu3/GDY, Pt3/GDY, and Pd3/GDY with Cu(111), Pt(111), and Pd(111) metal surfaces to further investigate the catalytic difference between M3/GDY and their metal surfaces ( Supporting Information Figure S13). The ORR overpotential of Cu(111), Pt(111), and Pd(111) are 0.88, 0.36, and 0.76 V, respectively. It is worth noting that the atomic utilization is 100% for M3/GDY, but its ORR performance is not as good as the metal surface. The reason is that all M3 clusters on GDY are nearly charge-neutral, which will easily reduce O2 and be poisoned in the first electrochemical step. Such oxygen poison results of ORR on all M3/GDY is on the same side of the volcano plot. This is different from the corresponding series of metal surfaces, among which the Au and Ag surfaces are on the opposite side of the volcano curve to the other metal surfaces, with Pt and Pd close to the top of the volcano.53 HER There are two distinct mechanisms for HER, the Volmer-Heyrovsky mechanism and the Volmer–Tafel mechanism.65,66 The first elementary reaction of both mechanisms provides the key intermediate *H via the Volmer step, H + + e − → H * ( the Volmer step ) in which the proton source H+ can be either the solvated proton (e.g