Identifying Key Descriptors for the Single-Atom Catalyzed CO Oxidation

Abstract

Open AccessCCS ChemistryRESEARCH ARTICLE3 Oct 2022Identifying Key Descriptors for the Single-Atom Catalyzed CO Oxidation Max J. Hülsey, Sambath Baskaran, Shipeng Ding, Sikai Wang, Hiroyuki Asakura, Shinya Furukawa, Shibo Xi, Qi Yu, Cong-Qiao Xu, Jun Li and Ning Yan Max J. Hülsey Department of Chemical and Biomolecular Engineering, National University of Singapore, Singapore 117585 Google Scholar More articles by this author , Sambath Baskaran Department of Chemistry, Southern University of Science and Technology, Shenzhen 518055 Google Scholar More articles by this author , Shipeng Ding Department of Chemical and Biomolecular Engineering, National University of Singapore, Singapore 117585 Google Scholar More articles by this author , Sikai Wang Department of Chemical and Biomolecular Engineering, National University of Singapore, Singapore 117585 Joint School of National University of Singapore, Tianjin University, International Campus of Tianjin University, Binhai New City, Fuzhou 350207 Google Scholar More articles by this author , Hiroyuki Asakura Department of Molecular Engineering, Graduate School of Engineering, Kyoto University, Kyoto 615-8510; 615-8245 Elements Strategy Initiative for Catalysts and Batteries (ESICB), Kyoto University, Kyoto 615-8245 Google Scholar More articles by this author , Shinya Furukawa Elements Strategy Initiative for Catalysts and Batteries (ESICB), Kyoto University, Kyoto 615-8245 Institute for Catalysis, Hokkaido University, Sapporo 001-0021 Google Scholar More articles by this author , Shibo Xi Institute for Chemical and Engineering Sciences, Agency for Science, Technology, and Research in Singapore, Singapore 138634 Google Scholar More articles by this author , Qi Yu Shaanxi Key Laboratory of Catalysis, Shaanxi University of Technology, Hangzhou 723001 Department of Chemistry, Key Laboratory of Organic Optoelectronics and Molecular Engineering of Ministry of Education, Tsinghua University, Beijing 100084 Google Scholar More articles by this author , Cong-Qiao Xu Department of Chemistry, Southern University of Science and Technology, Shenzhen 518055 Google Scholar More articles by this author , Jun Li *Corresponding authors: E-mail Address: [email protected] E-mail Address: [email protected] Department of Chemistry, Southern University of Science and Technology, Shenzhen 518055 Department of Chemistry, Key Laboratory of Organic Optoelectronics and Molecular Engineering of Ministry of Education, Tsinghua University, Beijing 100084 Google Scholar More articles by this author and Ning Yan *Corresponding authors: E-mail Address: [email protected] E-mail Address: [email protected] Department of Chemical and Biomolecular Engineering, National University of Singapore, Singapore 117585 Google Scholar More articles by this author https://doi.org/10.31635/ccschem.022.202201914 SectionsSupplemental MaterialAboutAbstractPDF ToolsAdd to favoritesDownload CitationsTrack Citations ShareFacebookTwitterLinked InEmail Fundamental knowledge of structure-activity correlations for heterogeneous single-atom catalysts (SACs) is essential in guiding catalytic design. While linear scaling relations are powerful for predicting the performance of traditional metal catalysts, they appear to fail with the involvement of SACs. Comparing the catalytic CO oxidation activity of different atomically dispersed metals (3d, 4d, and 5d) in conjunction with computational modeling enabled us to establish multiple scaling relations between the activity and simply calculated descriptors. Through these efforts, we found that the thermodynamic driving force for the oxygen vacancy formation needed to be considered in addition to the adsorption energies of substrates (in particular CO). Our approach was to reduce the computational requirements in determining better CO oxidation catalysts using a few key thermodynamic descriptors. This work presents one of the first successful approaches for re-establishing scaling relations for catalytic reactions by SACs with potentially broad implications for catalytic processes actively involving this support. Download figure Download PowerPoint Introduction The design of improved catalysts from the vast chemical space relies on the discovery of computationally inexpensive parameters that describe catalytic activity adequately.1,2 Among others, scaling relationships linking properties like adsorption and reaction energies with activation barriers have been developed for a range of reactions and surface morphologies but are so far predominantly focused on extended metal surfaces. While parameters like adsorption energies could be determined relatively quickly for large materials’ libraries with high accuracy and precision, activation energies are significantly more challenging to calculate. For the case of CO oxidation on extended fcc(111) surfaces and transition metal (TM) nanoparticles, linear correlations between the adsorption energies of O atoms, O2, and CO molecules, as well as the activation barriers, were predicted.3 For nanoparticles supported on reducible oxides, the CO oxidation commonly follows a Mars-van Krevelen (MvK) mechanism with the support actively contributing to the catalytic reaction by donating and accepting O atoms. For this reason, the CO oxidation activity was found to correlate well with the reducibility of the support.4,5 For metal oxides, scaling relations, in particular, for reactions involving C–H activation6–10 had been assessed previously. However, parameters at the interface of metal species and oxide supports are rarely discussed, although they are known to be essential for heterogeneous catalysis and are present in virtually all industrial catalysts.11–14 Single-atom catalysts (SACs) have recently been shown to be active and selective for a variety of oxidation, reduction, and coupling reactions.15–28 Density functional theory (DFT) calculations based on structures obtained via various tools such as X-ray absorption spectroscopy (XAS) serve as a powerful tool in revealing activity-structure correlations.15,16,27 Although CO oxidation is among the earliest and most thoroughly investigated reactions, performed experimentally and by computational modeling,29–33 clear guidelines for designing SACs are lacking. Obvious differences exist between catalytic surfaces and SACs because of the different electronic states, intimate metal-support interactions, the absence of adjacent TM atoms, and the absence of electronic bands forcing the treatment of scaling relations for SACs in alternative ways. In fact, many reports have shown that linear scaling relations do not apply to SACs supported on metal oxides or in host metals in so-called single-atom alloys.34–37 This has been explained previously by the unique electronic structure of SACs,36,38 geometric dynamics beyond what is achievable with nanoparticle catalysts,35 and ensemble effects with host metals in the case of single-atom alloys.34,37 Without discovering simple physical parameters of the catalysts governing catalytic activity, efficient screening of materials’ libraries is severely prohibited by computational expenses.39 Herein, we describe the synthesis of Cr, Mn, Fe, Co, Ni, Cu, Zn, Ru, Rh, Pd, Ag, and Pt SACs supported by phosphomolybdic acid (PMA). We present evidence for atomic dispersion by XAS and CO diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS), the uniform catalyst structure and morphology determined by N2 sorption, scanning electron microscopy (SEM), Raman, and IR spectroscopies, as well as the determination of identical active site structural arrangement of the exposed four-fold hollow site surface area by all Keggin-structured polyoxometalates (POMs) (Figure 1).40,41 Despite their structural similarities, the light-off temperatures for the CO oxidation reaction differed by as much as 400 °C for the SACs. DFT calculations further confirmed the stability of the active site structure and the dominance of the MvK mechanism of the reaction kinetics. It was further used to predict the adsorption energies of the substrates and the product, along with the formation energy of the oxygen vacancy (OV). Scaling relations between each of the adsorption energies and the light-off temperatures did not well match the catalytic activity, but after including the OV formation energy, we understood the trends of the catalytic performance. Figure 1 | Schematic depiction of the SACs investigated in this study. Different atomically dispersed TMs (Cr, Mn, Fe, Co, Ni, Cu, Zn, Ru, Rh, Pd, Ag, and Pt) are anchored on POMs composed of a central PO4 unit with 12 surrounding edge-sharing MoO6 units. Download figure Download PowerPoint Experimental Methods Catalyst synthesis and characterization An aqueous solution of appropriate amounts of PMA in 30 mL of deionized water was cooled to 0 °C. Under stirring and cooling, a solution of 365 mg cesium nitrate (MilliporeSigma, Burlington, MA, USA) and the appropriate amount of chromium nitrate nonahydrate, manganese hydrate tetrahydrate, iron nitrate nonahydrate, cobalt nitrate hexahydrate, nickel nitrate hexahydrate, copper nitrate hemi(pentahydrate), zinc nitrate hexahydrate, ruthenium nitrosyl nitrate solution, rhodium nitrate, palladium nitrate dihydrate, and silver nitrate (all from MilliporeSigma, Burlington, MA, USA) dissolved in 30 mL deionized water with 5 drops of concentrated nitric acid was added gradually over 30 min period. The solution was aged for 5 h under constant stirring and ice cooling. The formed solid was then separated by centrifugation (5 min, 8000 g) and washed twice with 30 mL of deionized water. After freeze-drying overnight, the catalysts were used as obtained. Metal contents were determined using inductively coupled plasma-optical emission spectrometry (ICP-OES; iCAP 6000 series instrument; Thermo Fisher Scientific, Waltham, MA, USA) with calibration curves obtained from solutions with the pure metal salts. The catalysts were dissolved in aqua regia under heating at 80 °C for 4 h, and then the filtered solutions were diluted with an appropriate volume of deionized water before measurement. Attenuated total reflection (ATR-IR) spectroscopy was performed using a Thermo Scientific Nicolet iS50 FT-IR spectrometer (Thermo Fisher Scientific, Waltham, MA, USA) in the range of 525–4000 cm−1 with a spectral resolution of 4 cm−1. Raman spectromicroscopy was performed using a Horiba Yvon Modular Raman spectrometer (Horiba, Kyoto, Japan) with a 532 nm excitation laser, 1% filter, a grating with 1200 mm−1 and a x100 objective. Calibration was done using the 521 cm−1 vibrations of a silicon wafer. Field emission scanning electron microscopy (FESEM) was performed using a JEM-6700F (JEOL, Ltd., Tokyo, Japan) microscope. Computational modeling Vienna Ab initio simulation package (VASP; version 5.3.5)42–44 was used to perform spin-polarized DFT calculations. The projector augmented plane wave (PAW) pseudopotentials were employed to describe the electron-ion interaction.45 The generalized gradient approximation (GGA) with the Perdew–Wang 91 (PW91) exchange-correlation functional was used,46 and widely applied for single-atom supported polyoxometalate clusters by different research groups.47–49 The plane wave cutoff energy was set to 400 eV in all calculations. A 20 × 20 × 20 Å3 cubic box was used to avoid interactions between the periodic images, as reported by Macht et al.,47 Yu et al.,48 and Wang et al.49 The Γ point was used for Brillouin zone center sampling. The total energy was converged to 10−4 eV by applying the Gaussian smearing method with a width of 0.05 eV. All the ions were allowed to relax until the maximum atomic forces became less than 0.05 eV/Å. The dimer method50–52 was used to determine the transition states, followed by vibrational frequency calculations to confirm saddle points with one imaginary frequency. Following the convention of thermodynamics, the adsorption energy (Eads) of O2, CO, or CO2 on the [email protected] surface was calculated as E ads = E total − E TM @ PMA − E adsorbate where Etotal, E[email protected], and Eadsorbate correspond to the electronic energies of adsorbed species on the [email protected], [email protected], and free adsorbates (CO, O2, and CO2), respectively. The OV formation energy (Evac formation) of [email protected] was calculated as E vac formation = − ( E CO - TM / PMA + E CO 2 − E 2 CO - TM / PMA ) where ECO-TM/PMA, ECO2, and E2CO-TM/PMA correspond to the calculated electronic energies of adsorption structure with one CO molecule, gas-phase CO2, and adsorption structure with two CO molecules, respectively. CO DRIFTS Approximately 100 mg catalyst powder was loaded in a Harricks HV-DR2 (Harrick Scientific Products Inc., Pleasantville, New York, United States) reaction cell covered by a Praying Mantis high-temperature reaction chamber with ZnS windows in a Thermo Scientific Nicolet iS50 FT-IR spectrometer; Thermo Fisher Scientific, Waltham, MA, USA) with a mercury-telluride (MCT) detector. The chamber was closed gastight, and background scans were recorded under 40 mL/min nitrogen gas flow (Air Liquide purity, 99.9995%) at room temperature. 40 mL/min 5% carbon monoxide in argon gas (Air Liquide) were introduced at 50 °C for 30 min, whereupon the diluted carbon monoxide gas flow was replaced with a gas flow of 40 mL/min nitrogen gas. During the gas treatments, spectra were collected regularly. CO oxidation reaction For the CO oxidation reaction, appropriate amounts of the catalysts were loaded into a stainless-steel tubular plug flow reactor, fixed with quartz wool, and heated using a tube furnace with an external thermocouple. Before the reaction, catalysts were activated in a flow of 5% O2 (balance Ar) for 60 min at 250 °C with a heating rate of 5 °C min−1. Then 2.5% CO and 2.5% O2 (balance Ar) with a total flow rate of 80 mL min−1 were introduced into the reactor set at room temperature. Subsequently, the temperatures were increased, and the reactor efflux was analyzed with an Agilent 7890B gas chromatograph (Agilent Technologies, Santa Clara, CA, USA) with a thermal conductivity (TCD) detector. Conversions and yields were determined after 30 min when a steady-state was reached. X-ray absorption spectroscopy Self-supported undiluted pellets of 200 mg sample were pressed before measurements. X-ray absorption spectroscopy (XAS) measurements at the Cr, Mn, Fe, Co, Ni, and Cu K edges were performed using the X-ray absorption fine structure for catalysis (XAFCA) beamline detector at the Singapore Synchrotron Light Source (SSLS; Buona Vista, Singapore).53 The storage ring of the SSLS was operated at 0.7 GeV with a maximum current of 200 mA. Data were collected at fluorescence mode while the respective metal foils were measured in transmission mode simultaneously using a Si(111) double crystal monochromator. The average of 3–5 spectra was calculated to reduce noise. XAS measurements of the Ru, Rh, Pd-edges and Pt L3-edge were performed at BL37XU (Ru, Rh, Pd) and BL11XU (Pt) at SPring-8 (Japan Synchrotron Radiation Research Institute, Hyogo, Japan), operated at 8.0 GeV with a constant current of 99.5 mA. The XAS spectra were collected in fluorescence mode, while the standard samples were measured in transmission mode. For the Ru K-edge or Rh K-edge XAS measurement of the Ru/CsPMA or Rh/CsPMA, we exploited the fluorescence lines at Ru Kβ or Rh Kβ lines because the Mo Kβ1 lines (19606 and 19962 eV) significantly overlapped with the Ru Kα1 (19279 eV) or Rh Kα1,2 lines (20216, 20074 eV), measured with a Ge solid-state detector (Mirion Technologies, Inc., Atlanta, GA, USA). For the Pd K-edge XAS measurement of Pd/CsPMA, we measured the XAS spectra typically using the Pd Kα line. For the Pt L3-edge XAS measurement, we employed a high-energy resolution fluorescence detection (HERFD) method with a Si(733) spherically bent analyzer (XRS TECH LLC., Freehold, NJ, Unite States). The Pt L3-edge extended X-ray absorption spectroscopy (EXAFS) spectrum was broadened with a Gaussian function by 5 eV in width to make the EXAFS amplitude comparable to the EXAFS spectra measured in conventional transmission mode. We reported some of the details in a previous paper.54 Data analysis was done with the Demeter software package.55 σ2 values were determined from commercial metal oxide samples with a known crystal structure. X-ray absorption near-edge spectra (XANES) simulations were performed using the CASTEP ab initio quantum mechanical program code56 with the PW91 exchange-correlation functional46 based on the GGA. The plane-wave basis set was truncated at a kinetic energy of 400 eV. Wavelet transformation analysis was performed using the Morlet wavelet transform procedure with values for κ and σ of 5 and 1, respectively.57 Results and Discussion Active site structure of M1/CsPMA The single-atom M1/CsPMA catalysts were synthesized by a coprecipitation method starting with cesium nitrate, metal nitrate salts, and PMA, as reported previously.40,58 Then the dried catalysts were used for analysis and catalytic testing without further modification. Atomic dispersion of the TMs was confirmed by XAS, as shown in Figure 2a. XANES revealed that all SACs were highly charged (2+, 3+, or 4+), and EXAFS confirmed that none of the catalysts exhibited metal-metal bond scattering, with all of them showing significant metal-oxygen scattering contributions (Figure 2b, Table 1, Supporting Information Table S1, and Figures S1–10). EXAFS fitting was consistent with the formation of the active site, which displayed metal atoms in the four-fold hollow sites of the Keggin structure with adsorption of one O2 molecule, except for Cu1/CsPMA and Pd1/CsPMA, which showed no adsorbed O2. This was consistent with DFT calculated O2 adsorption energies (vide infra). For 4d elements, the XAS analysis was more complicated due to the strong X-ray fluorescence of Mo close to the absorption edges of Ru, Rh, and Pd. Nevertheless, Morlet wavelet transformation analysis suggested that the SAC only exhibited first shell coordination to light scattering atoms like oxygen and none to heavier atoms like TMs (Figure 2c and Supporting Information Figures S11–S19). To develop an appropriate catalyst model for atomistic simulations, we ensured that all the catalysts exhibited the same known active site structure. XANES simulations based on DFT-calculated catalyst structures could help identify the exact local coordination environment. Metal oxide samples were used as a reference to confirm the validity of the employed level of theory ( Supporting Information Figure S20). For the predicted active site structures, an almost ideal overlap between simulated and experimental spectra confirmed the formation of active site structures, as discussed above (Figure 2d and Supporting Information Figure S20). This is in accordance with our previous findings, where the active sites of POM-supported Rh1 and Pd1 catalysts were elucidated in detail.58,59 Figure 2 | Confirmation of the SAC structure of M1/CsPMA. (a) K edge XANES spectra of ten different M1/CsPMA catalysts and their respective metallic and oxide reference materials. (b) R-space EXAFS spectra of different M1/CsPMA catalysts (colored, solid lines) and their respective metal foils (black, dashed lines). (c) Morlet wavelet transformation analysis of Cr foil, Cr2O3, and Cr1/CsPMA. (d) Experimental and simulated XANES spectra for different M1/CsPMA catalysts; the active site structure considered for XANES simulations are shown. Additional data are shown in the supporting information. Download figure Download PowerPoint Table 1 | EXAFS Fitting Parameters of Some Representative M1/CsPMA Materials. Additional Details Can Be Found in the Supporting Information Coordination Shell CN R (Å) σ2 CoO Co–O 6 2.10 ± 0.04 0.017 ± 0.008 Co1/CsPMA 6.0 ± 1.9 2.03 ± 0.04 0.009 ± 0.007 Rh2O3 Rh–O 6 2.04 ± 0.01 0.003 ± 0.001 Rh1/CsPMA 6.8 ± 0.9 2.02 ± 0.01 0.003 ± 0.002 PtO2 Pt–O 6 2.14 ± 0.02 0.001 ± 0.002 Pt1/CsPMA 3.9 ± 0.7 2.13 ± 0.01 0.001 ± 0.002 We employed the DRIFTS technique as another approach to confirm the single-atom identity using CO as a probe molecule. We observed that even CsPMA without additional metal adsorbs CO with two broad vibration bands at 1930, 2042 and a sharper one at 2122 cm−1, probably assignable to O–CO and Mo–CO species. None of the SACs based on early TMs showed additional CO adsorption bands, which was in line with their generally low CO affinity; another possible explanation might be the lower extinction coefficient of CO adsorbed on undercoordinated or charged metal sites such as those presented by our SACs.60 For some of the heavier elements, however, strong CO adsorption could be observed, in particular, for Rh, Ag, and Pt. For Rh1/CsPMA, two CO molecules were adsorbed simultaneously, and both the symmetric and asymmetric vibrations were visible at 2027 and 2099 cm−1, similar to previous reports on Rh SACs.58,61 Ag1/CsPMA adsorbs CO with an IR vibration at 2182 cm−1, slightly higher than gas-phase CO, indicating the presence of a non-classical carbonyl as commonly observed for cationic closed-shell d10 systems like Ag+.62 In the CO DRIFT spectrum of Pt1/CsPMA, a dominant peak at 2122 cm−1 was observed, indicative of a highly positively charged Pt in a 4+ oxidation state. A shoulder at 2099 cm−1 was probably related to Pt2+ species.32,63,64 Furthermore, no additional peaks in the range of 1750–2080 cm−1 were observed, suggesting the absence of metallic nanoparticles in all the catalysts (Figure 3).63,65 A summary of approximate oxidation states for the different M1/CsPMA catalysts based on CO DRIFTS and XANES is provided in Table 2. Figure 3 | CO DRIFTS spectra for CsPMA and different M1/CsPMA catalysts under a 5% CO atmosphere. Download figure Download PowerPoint Table 2 | Approximate Oxidation State of M1/CsPMA Materials Based on CO DRIFTS and XAS Cr Mn Fe Co Ni Cu Ru Rh Pd Ag Pt Oxidation state based on CO DRIFTS N.D. N.D. N.D. N.D. N.D. N.D. N.D. 3+ N.D. 1+ 4+/2+ Oxidation state based on EXAFS 3+ 2+ 3+ 2+ 2+ 2+ 4+ 3+ 2+ N.D. 2+ For the electrostatic stabilization of SACs, the POM structure must be preserved during the synthesis and catalytic reaction. The stability and persistence of the Keggin structure after the adsorption of TMs were confirmed by Raman and infrared spectroscopies. For both CsPMA and the M1/CsPMA materials, the vibrational modes typical for PMA were apparent, whereas no overlap existed with the spectrum of MoO3 or other molybdenum oxides (Figures 4a and 4b).66 The surface areas determined ranged from 89–132 m2 g−1 by N2 physisorption and Brunauer–Emmett–Teller (BET) analysis. Those values were typical for precipitated POMs and were not significantly affected by the type of atomically dispersed metal. FESEM images of the different M1/CsPMA catalysts showed a consistent morphology featuring spherical particles predominantly, with radii between 100–300 nm (Figure 4c and Supporting Information Figure S21). At higher magnification, the surface of those particles appeared to be rough with featured a few nanometers in size ( Supporting Information Figure S22). Figure 4 | Characterization of the POM structure and morphology of M1/CsPMA. (a) Raman spectra of M1/CsPMA with labels of the surface areas determined by BET analysis of N2 adsorption isotherms, (b) Infrared spectra of M1/CsPMA, (c) FESEM images of different catalysts; the scale bar has a length of 2 μm. Download figure Download PowerPoint Scaling relations for the CO oxidation activity of M1/CsPMA CO oxidation light-off curves were measured with a plug-flow reactor at temperatures up to 500 °C. When the T20 temperatures for the CO oxidation reaction were compared on different SACs, significant differences were visible, especially between earlier TMs, some of which exhibited T20 values above 400 °C, identified as noble metals. Rh constituted the most active SAC with a T20 value of ∼88 °C, followed by Pd, Pt, and Ru with T20 values of 173, 244, and 308 °C, respectively (Figure 5a). The differences were much less pronounced for earlier TMs, although Co SACs appeared to be particularly active for CO oxidation with a T20 temperature of 330 °C, almost paralleling Ru1/CsPMA. Other metals analyzed were Cr, Mn, Cu, Fe, Zn, and Ni, with T20 values of 382, 413, 425, 430, 465, and 466 °C, respectively. Ag1/CsPMA appeared to be a special case where the coordination of CO in the non-classical carbonyl seemed to have inhibited the background activity of the support ( Supporting Information Figure S23). As the T20 value exceeded 500 °C, the Ag-based SAC was omitted from further analysis of scaling relations. Raman and IR spectroscopy confirmed the stability of the catalyst support after the reaction had reached up to 500 °C ( Supporting Information Figure S24). This is in accordance with our previous studies in which the stability of the POM-supported Rh SACs at high reaction temperatures was demonstrated.58,67 Figure 5 | CO oxidation activity and reaction descriptors of different M1/CsPMA catalysts. (a) Light-off curves of select M1/CsPMA catalysts (see Supporting Information Figure S23 for additional catalysts). The horizontal line at 20% CO conversion serves to guide the eye. (b) Calculated CO and O2 adsorption and OV formation energies on M1/CsPMA (M = 4d TM). The more stable of the two different O2 adsorption modes (O2,s η2 side-on, and O2,e η1 end-on) were considered. (c) O2 reaction order for different M1/CsPMA catalysts at different temperatures. Download figure Download PowerPoint In order to correlate the catalytic CO oxidation activity with simple physicochemical properties of the SACs, we carried out DFT calculations on PMA-supported TM. To confirm the validity of the catalyst model used for the simulations, analyses of the assumed reaction mechanism for Rh1/CsPMA were also performed. The predicted mechanism matched the MvK mechanism previously determined by spectroscopic and kinetic analyses with support reoxidation as a rate-determining step ( Supporting Information Figures S23a and S23b).58 We expected other SACs to have other reaction barriers or rate-determining steps. For high-throughput computational screening of catalyst materials, it was beneficial to identify simple thermodynamic parameters that could be determined with minimum computational efforts. Often, scaling relationships are established by correlating between adsorption energies of different substrates and the activity. SACs were naturally different from extended metal surfaces because they did not possess adjacent metal active sites; thus, coverage effects and lateral interactions between adsorbates did not play a role. Furthermore, metal centers in SACs interacted solely with the support, highlighting the importance of considering the participation of the support during the catalytic reaction in modeling. Screening of the adsorption energies of CO and O2 ( Supporting Information Table S2 and Figure S25) on all TM revealed clear trends throughout the periodic table ( Supporting Information Figure S26). Generally, trends among different rows were comparable with slightly larger adsorption energies for heavier TMs due to increased nd orbital radii. Stark differences were evident when groups of TMs with low CO but high O2 binding energies were compared in groups 3–6; generally, low adsorption energies for both substrates in groups 10–12 and balanced O2 adsorption strength coupled with high CO adsorption energies in groups 7–9. The formation of oxygen vacancies was endothermic for groups 3–7 (except Mn) and then monotonously increased in energy up to group 11 but was slightly less favorable for group 12. From these trends, it became obvious that early TMs were too oxophilic, preventing the OV formation and hampering CO adsorption, while late TMs did not bind CO strongly, and their high OV formation energies likely led to thermodynamically unbalanced reaction energy profiles. It appeared that only catalysts with balanced adsorption strengths and OV formation energies were effective in the oxidation of CO (Figure 5b and Supporting Information Figure S26). Although O2 adsorption occurred in the oxygen vacant metal-support interface in the MvK mechanism, we found that these adsorption energies were almost congruent with those on the metal site alone ( Supporting