Reaching the Fundamental Limitation in CO2 Reduction to CO with Single Atom Catalysts

The electrochemical CO2 reduction reaction (CO2RR) to value‐added chemicals with renewable electricity is a promising method to decarbonize parts of the chemical industry. Recently, single metal atoms in nitrogen‐doped carbon (MNC) have emerged as potential electrocatalysts for CO2RR to CO with high activity and faradaic efficiency, although the reaction limitation for CO2RR to CO is unclear. To understand the comparison of intrinsic activity of different MNCs, two catalysts are synthesized through a decoupled two‐step synthesis approach of high temperature pyrolysis and low temperature metalation (Fe or Ni). The highly meso‐porous structure results in the highest reported electrochemical active site utilization based on in situ nitrite stripping; up to 59±6% for NiNC. Ex situ X‐ray absorption spectroscopy (XAS) confirms the penta‐coordinated nature of the active sites. The catalysts are amongst the most active in the literature for CO2 reduction to CO. The density functional theory calculations (DFT) show that their binding to the reaction intermediates approximates to that of Au surfaces. However, it is found that the turnover frequencies (TOFs) of the most active catalysts for CO evolution converge, suggesting a fundamental ceiling to the catalytic rates.


Introduction
Rising living standards, thriving industries, and increased transportation in emerging and developed economies have resulted DOI: 10.1002/adfm.202302468 in a record-high energy demand, which is mostly met by the combustion of fossil fuels.As a result, CO 2 emissions from fossil fuels are causing a severe environmental crisis and a threat to social development.Electrocatalytic CO 2 (CO 2 RR) reduction has emerged as a viable and potential transformative technology to close the carbon cycle.Although the production of hydrocarbons (methanol, ethanol, ethylene etc.) through CO 2 RR is still challenging in terms of selectivity and efficiency, [1][2][3] such products can be obtained through a tandem system where CO 2 is reduced to CO, and CO subsequently employed as a renewable feedstock for the Fischer-Tropsch reaction or further electrochemically converted to multicarbon products. [4,5]CO 2 reduction to CO involves only two electrons and two proton transfers; the state of the art catalysts for the reaction are based on Au and Ag, [6,7] first reported by Hori and coworkers. [8][13] Varela et al. first reported that Fe/Mn-N 4 sites embedded in graphene can reduce CO 2 to CO. [11] Since their seminal report, many strategies have been used to date for improving selectivity and current density in MNC materials, including tuning the intrinsic activity, enhancing the number of M-N x sites or controlling the rate of CO 2 diffusion. [14]While plenty of efforts have been dedicated to exploring different metal coordination environments in carbonbased supports, [15][16][17] the current synthetic protocols for MNC CO 2 electrocatalysts display several drawbacks that hinder intrinsic catalytic activity, selectivity, and active site utilization.MNC catalysts are often prepared by pyrolysis at high temperatures of a mixture comprising a metallic precursor and carbon and nitrogen containing monomers or polymers.With increased metal loadings, such synthetic protocols often lead to the formation of metal nanoparticles, metal oxides, and metal carbides, which are active towards hydrogen evolution. [18]Although, in principle, single atom MNC catalysts could exhibit 100% active site utilization (defined as the electrochemically accessible metal or metalnitrogen sites compared to the total amount [19,20] ), O 2 reduction literature shows that microporous FeNC catalysts often only utilize <10% of atomic active sites. [21,22]However, very recently we reported a synthetic protocol to prepare highly electrochemically accessible FeNC single atom catalysts (FeN x electrochemical utilization >50%) employing a decoupled two-step synthesis that entailed low temperature metalation of a templated micro-and mesoporous catalyst. [23]espite the continuous effort towards improving the catalyst material, the fundamental origin of the high overpotential in most CO evolution catalysts remains under debate. [24]Electroreduction of CO 2 to CO proceeds through activation of CO 2 and release of CO.[28] Nonetheless, the initial activation of CO 2 is easier at more carbophilic catalytic sites (e.g., can be measured as a stronger binding of the *COOH intermediates).However, the second step forming *CO and its subsequent desorption, is better catalysed by less carbophilic catalytic site (e.g., a weaker binding of *CO or *COOH).The scaling relationship between both the C-binding species (*CO and *COOH) leads to a Sabatier volcano.Even at the peak of the volcano, metal catalysts exhibit a substantial overpotential. [25]Furthermore, Nitopi et al. showed that when compared in terms of intrinsic activity, the most active metal-based catalysts tend to converge towards the activity of pure Cu. [29] In MNC catalysts, the activity and binding energies can be tuned through the modification of the metal and the coordination environment.Previous studies have reported the trend of CO adsorption energy on FeN x in the following order (from strong to weak) FeN 2 > FeN 3 > FeN 4 > FeN 5 . [30,31]The axial position of a planar Fe-N 4 can be coordinated to any nonmetal such as carbon, nitrogen, oxygen, or sulphur, thus breaking the site-symmetry and electronic distribution. [32]However, it is not clear how such ligands affect the reactivity of the MNC catalysts and their position on the volcano.
In this work, we explore the fundamental limitations in CO production with MNC materials, with M being either Fe or Ni.We build on our previously reported synthetic protocol that provides MNC materials with optimum utilization that allows us to maximize TOF and reach the highest values reported in the state-of-the-art (6.8 e − site −1 s −1 at −0.59 V vs reversible hydrogen electrode, RHE). [23]The active site structure of the prepared catalysts is elucidated by means of XAS, the atomic dispersion by high angle annular dark field transmission electron microscopy (HAADF-TEM), and the effect of axial ligands in the binding energy of reaction intermediates is estimated by DFT calculations.Additionally, the stability of the catalyst is assessed through timeof-flight secondary ion mass spectrometry (ToF-SIMS) pre and post extended measurements.Finally, we compare the TOF of the best performing CO 2 -to-CO catalysts in the literature, showing the existence of similar fundamental limitation for MNC catalyst as for metals.

Results and Discussions
Nitrogen-doped carbon materials were prepared as recently shown by our group by pyrolysis of 2,4,6-triaminopyrimidine (TAP) in the presence of MgCl 2 .6H 2 O at 900 °C (Figure 1). [23]his process allows TAP to self-organize by hydrogen bonding between the amine groups and the water molecules of MgCl 2 .6H 2 O before polymerizing in a molten state.
The pyrolysis of TAP in the presence of MgCl 2 .6H 2 O at 900 °C resulted, after acidic washing, in a highly porous nitrogen-doped carbon material with a Brauner-Emmett-Teller specific surface area ≈3290 m 2 g −1 and bimodal pore size distribution centered at 0.8 and 2.1 nm.For further details regarding synthesis and characterization of the materials we would like to refer the reader to previous work. [23]We would like to note that we selected 900 °C as the pyrolysis temperature because it provides an adequate balance between nitrogen content and electrical conductivity.In our previous work, we observed that the pyrolysis of TAP in the presence of MgCl 2 .6H 2 O at 800 °C in 23.5 at% of N and allowed the coordination of 1.8 wt% of Fe single atoms.However, the electrocatalytic activity of the final material was negligible owing to the lack of electrical conductivity.TAP 900 on the other hand displays 4.5 at% of N and therefore higher electrical conductivity as confirmed by its low resistance.Fe and Ni were coordinated in the N moieties of TAP 900 employing the method first reported by Fellinger and co-workers for O 2 reduction single atom catalysts. [33,34]TAP 900 reaction in methanol reflux with either FeCl 2 or NiCl 2 .6H 2 O led to 0.520 wt% Fe and 0.265 wt% Ni, respectively, determined by inductively coupled plasma mass spectrometry (ICP-MS) (Figure 2a).While those values are lower than the state of the art for single atom catalysts, we would like to note that the focus of this study is to achieve accessible active sites rather than screening materials with different loadings that may lead to a lower utilization factor.Fe is detected in TAP 900 and TAP 900@Ni, as well as trace amounts of Cu in all TAP materials, likely arising from the low purity TAP precursor.Meanwhile, X-ray photoelectron spectroscopy analysis of the materials after metal coordination (Figure 2b; Figure S1, Supporting Information) further confirms the metal loading.N1s chemical states shows a chemical contribution standing for Nmetal coordination (arising from Mg in the case of bare TAP), as well as pyridinic, pyrrolic and graphitic, which remain consistent across the TAP materials.X-ray diffraction patterns suggests the absence of large aggregated metallic nanoparticles, as no sharp diffraction peaks are observed (Figure S2, Supporting Information), and confirm the amorphous nature of the nitrogendoped carbon supports as observed previously in ionothermal transformations. [35,36]The small broad peak in TAP 900@Fe and TAP 900@Ni at ≈13 o relates to graphite oxide (001), [37][38][39] while the ≈25 o (002) of graphite is not present in any TAP-derived material, suggesting a graphene-like structure.
The atomic nature of both Fe and Ni was elucidated through HAADF-STEM.Both TAP 900@Fe and Ni are composed of atomically dispersed metals in a matrix of carbon and nitrogen, without visible presence of aggregates in several regions of the material (Figure 3; Figures S3 and S4, Supporting Information).Oxygen can be also observed owing to the remaining functionalities in the surface of the material (Figure S5, Supporting Information).We employed Raman spectroscopy to elucidate the number of layers in our materials as shown recently by Mehmood et al. (Note S1, Supporting Information). [22]We observed that the number of carbon layers is approximated to be 1, which is expected for a high surface area catalyst (Figures S6 and S7, Supporting Information), and which suggests a very high accessibility of the N-coordinated metallic single atoms in a graphene-like layer.
Cryo (5 K) X-band electron paramagnetic resonance (EPR) was chosen to probe unpaired electrons within the catalyst (Figure S8, Supporting Information).As shown in our previous work, TAP 900 shows a sharp g ≈ 2 signal, most likely from organic radicals, [13] which may arise from defect states of the carbon support.The reduced g ≈ 2 signal in TAP 900@Ni suggests the organic radicals are removed during the metalation and acid washing process with the slight background signal remaining arising from the EPR tube or resonator.Previous EPR studies on Ni have observed signals at g ≈ 2.2, assigned to the 3dx 2 -y 2 orbital of Ni(I) species. [40,41]However, TAP 900@Ni does not exhibit any signal at g ≈ 2.2, indicating that Ni (II) is the resting state in the assynthesized TAP 900@Ni catalyst.
The coordination environment around TAP 900@Fe and Ni was further evaluated through ex situ XAS.As discussed in previous work, the first derivative of the normalized X-ray absorption near edge structure (XANES) in TAP 900@Fe indicates a structural distortion (broken D 4h symmetry) and the presence of a penta-coordinated Fe site. [42,43]Conversely, XANES spectra, in addition to EPR, of TAP 900@Ni showed that a +2 oxidation state is predominant for Ni, and owing to its electronic configuration it is not expected to display a penta-coordinated environment (Figure 4c,d).Only one prominent peak is observed at 1.55 Å in the Fourier transformed (FT) extended X-ray absorption fine structure (EXAFS) spectrum (Figure S9a,b, Supporting Information), which is attributed to the M−N contribution.Moreover, no scattering peaks arising from M-M coordination are seen in TAP 900@M (where M = Ni or Fe).These results confirm the atomic distribution of Fe/Ni, consistent with the observations from STEM measurements as shown above.The low Ni content within TAP 900@Ni did not allow for an accurate EXAFS fitting, Figure 2. a) Metal loadings in TAP 900@Fe and TAP 900@Ni calculated from ICP-MS (see experimental section at Supporting Information for further details on ICP-MS measurements).Two independent measurements for TAP 900 and TAP 900@Ni and four independent measurements for TAP 900@Fe were conducted, with the error bars representing the standard deviation.b) N1s chemical states atomic percentage of TAP-derived materials before and after active metal coordination.but Ni-N coordination is indirectly confirmed by ToF-SIMS, discussed later).Meanwhile first-coordination EXAFS structural fitting was possible on TAP 900@Fe (Table S1, Supporting Information) which showed a coordination number (CN Fe-N ) ≈5.6, indicating the presence of a mixture of FeN 5 and FeN 6 active sites, consistent with our previous report via low-temperature Mössbauer spectroscopy. [23]avelet transform was employed to demonstrate the atomic positions of the Fe atom in the synthesized sample with respect to the Fe foil and Fe(III)PcCl references.As shown in Figure 4e, Fe foil has a strong signal at ≈2.3 Å that can be assigned to the Fe-Fe bond distance, while Fe(III)PcCl has a strong wavelet signal at 1.4 Å implying the presence of shorter Fe-N/O bonds.TAP 900@Fe exhibits a similar wavelet signal to Fe(III)PcCl due to the presence of Fe-N bonds.
Aside from tuning the catalyst coordination, the wettability of the electrode is critical for improving the CO 2 RR microenvironment and optimizing triple-phase interface of solid, liquid and gas. [44]Super hydrophilic electrode surfaces cause flooding of the gas diffusion electrode (GDE) and flow-channel, leading to unwanted hydrogen evolution, whereas superhydrophobic surfaces may not allow for enough contact between the electrolyte and the electrocatalyst. [45]As a result, the coexistence of hydrophilic and hydrophobic conditions is considered as an ideal condition. [46]Water vapor adsorption measurement (obtained at 25 °C) on TAP 900 is compared to hydrophilic and hydropho-bic carbons reported in previous literature in Figure S10 (Supporting Information), [11] which shows that the TAP 900 carbonframework is relatively hydrophobic in nature, thus facilitating the CO 2 RR .We would like to note that the values obtained for the carbons reported in the literature were obtained in the same conditions that the ones shown in this work, with a volumetric sorption analyzer following a standard sample degassing and water freeze-thawing protocol.Additionally all the measurements were performed at the same temperature, with the same volumetric vapor sorption principle ensuring the comparability of our results with the literature.
The prepared materials were employed as electrocatalysts for the CO 2 reduction in 0.5 m KHCO 3 .In an aqueous environment, CO 2 possesses a major mass-transport limitations due to its moderate solubility (33 × 10 −3 m at 1 atm and 25 °C).To circumvent this issue, we mixed the active catalyst with hydrophobic polytetrafluoroethylene (PTFE) and used a customized electrochemical cell (Figure S11, Supporting Information) to separate the gas flow channel and liquid compartment. [44]This ensured the formation of three-phase interface between gaseous CO 2 , solid active catalyst, and the aqueous electrolyte.Thus, the working electrode was prepared by spray-coating an ink comprising active catalyst and PTFE to maintain a catalyst loading of 0.75 mg cm −2 on a carbon paper with an active geometric surface area of 1 cm 2 .Scanning electron microscopy (SEM) image of the spray coated TAP 900@Ni electrode is shown in Figure S12 (Supporting Information) which shows the flaky nature of the synthesized catalyst together with PTFE particles.Prior to the electrochemical testing, the surface of the electrode was preconditioned by conducting cyclic voltammetry measurements in the potential range of 0.1 to +0.6 V versus RHE (Figure S13, Supporting Information) with Ag/AgCl and 40% Pt/C (spray coated on carbon paper with 40 wt% loading of Pt) as a reference and counter electrode, respectively.Additionally, Fumasep FAA-3-50, an anion exchange membrane (AEM), was employed to separate the catholyte and anolyte compartments (see Supporting Information for more information regarding the AEM).Impedance spectroscopic measurements were used to calculate the uncompensated ohmic resistances, which yielded the iR-corrected potentials (see electrochemical measurement section of Supporting Information for details).
Chronoamperometric (CA) runs were conducted for 40 min at each potential ranging from −0.3 to −0.6 V versus RHE.We would like to note that the catalysts were screened with the same starting potentials versus Ag/AgCl and such potential was then converted to V versus RHE and corrected for the solution resistance; therefore, the final potential depends on current density and solution resistance for each catalyst.We could observe that with both TAP 900@Fe and TAP 900@Ni electrodes, the sum of faradaic efficiency (FE), FE CO and FE H2 reached almost 100% in the potential range of −0.5 to −0.7 V (Figure S14, Supporting Information), indicating the absence or negligible formation of other products.[49] Selectivity towards CO varies depending on the choice of the transition metals (Fe or Ni).TAP 900@Fe showed good CO2RR activity at less negative potentials and approached a high CO selectivity (mean FE CO = 93.5 ± 3.7%) at −0.55 V versus RHE.In contrast, TAP 900@Ni exhibited low CO selectivity at less negative potentials and reached maximum CO  [51] Fe 3+ -NC, [52] Fe 3+ -NC-GDE, [52] Ni-NC, [53] Fe-NC [53] ) and c) Chronoamperometry curve and Faradaic efficiency of CO production by TAP 900@Ni at −0.57V versus RHE.
selectivity (mean FE CO = 95.3 ± 4.7%) at −0.59 V. Bare While bare TAP 900 displays 0.15 wt% of Fe, likely arising from the relatively low purity of the precursor, its selectivity towards CO remained very poor; indicating the activity in TAP 900@M resulting predominantly from the M-N x sites (Figure 5a).The partial current density of TAP 900@Fe and TAP 900@Ni to CO has been plotted and compared to some of the best reported catalysts (Figure 5b).TAP 900@Ni showed high stability as exemplified by the 10 h chronoamperometric test, where the FE remained >90% with 98% of the initial current for CO production and a stable current density of 15 mA cm −2 at −0.57V versus RHE (Figure 5c).While higher current densities can be obtained employing more sophisticated cell designs, such as a microflow cell or a membrane electrode assembly, [50] in our study we aim to provide insights into the intrinsic activity of single atom catalysts.SEM images of preand postmeasurement spray coated TAP 900@Ni (Figure S15, Supporting Information) demonstrated that the morphology of the electrode surface remains unchanged.
Insights into the surface fragments of the materials were obtained through ToF-SIMS, which has emerged as a powerful tool to elucidate active site composition in single and dual atom electrocatalysts. [15,54,55]Here, ToF-SIMS was used to help confirm the NiN x C y moieties in TAP 900@Ni (Figure S16, Supporting Information), as previously carried out by Koshy et al. for their Ni-N-C CO2RR catalyst. [54]In addition we report for the first time using ToF-SIMS to probe the stability of these moieties by measurement prior and post-CA testing (45 min at −0.55 V vs RHE).TAP 900@Ni was recorded in negative polarity of SIMS due to the previously reported higher ionization yields of NiN x C y fragments with the primary ion beam. [54]Prior to ToF-SIMS measurements the surface was lightly sputtered with an Ar beam to cleanse the surface of contaminants such as small organics.In contrast to other reports on FeNC materials, [56,57] we could not detect FeN x C y + in TAP 900@Fe due to a possible combination of lower FeN x C y ionization and sputter yields (compared to NiN x C y ) and organic species masking FeN x C y + peaks.Previously reported FeNC materials used a specific carbon precursor (perylene tetracarboxylic dianhydride) which reduced organic species signals, therefore assisting detection of FeN x C y + ions. [56,57]ocusing on ToF-SIMS of TAP 900@Ni, confirmation of peak identification and correction is necessary since many peaks can-not be deconvoluted due to the equivalent masses of possible isotopic fragments.Peak identification and validation are discussed in Note S2 in the Supporting Information.Due to the low Ni content in TAP 900@Ni, many Ni ion fragments counts are low; therefore, focus is made on Ni fragments which exhibit clearly defined peaks in the mass spectrum.As mentioned earlier, EXAFS was not possible due to the low Ni content, however the identified fragments in Figure 6 supports the identification of isolated Ni single atoms in N-doped C from combined HAADF-STEM and elemental mapping EDX.The most abundant NiN x C y fragment post-CA, NiNC 3 − , was also found by Koshy et al. [54] to be the highest normalized count fragment for their fresh Ni-N-C catalyst with similar Ni loading of 0.2 wt%.We found no detectable peaks in m/z values matching those of possible dual Ni atom fragments (Ni 2 N x C y − ), confirming a purely single NiN x atom nature To cleanse the catalyst surface, the samples were first non-interlaced sputtered with a 10 keV Ar + cluster ion beam at ≈10 nA current, until reaching a dose density of 1 × 10 15 ion cm −2 .For measurements a Bi 3 + primary ion beam with 25 keV and 0.5 pA beam current was used for 25 min.
of the TAP 900@Ni catalyst.Interestingly, in Figure 6, normalized counts are lower prior to electrochemistry for all Ni fragments identified.Meanwhile normalized counts of ion fragments assigned for calibration remained the same pre-and post-CA, within error, except for C 5 − (Figure S17a, Supporting Information).Under reaction conditions either more Ni sites are exposed over time, or species formed on the catalyst surface change the chemical state of the surface and lead to a matrix effect which enhances NiN x C y − ionization yields.This matrix effect could be caused by precipitation of salt such as KHCO 3 on the electrode surface (Figure S17b, Supporting Information).Nevertheless, the homogeneous dispersion of Ni fragments across the post-CA TAP 900@Ni is confirmed by imaging from ToF-SIMS (Figure S18, Supporting Information).
Having probed the nature and stability of the active site, we then went on to quantify the number of active sites, which in turn allows us to accurately determine the TOF, the number of product molecules produced or reactants consumed per unit time on a single active site.In a bulk or nanostructured catalyst, most of the active sites are inaccessible, therefore the CO 2 RR intermediate is adsorbed and interacts exclusively with the surface atoms/active sites during a catalytic process.This lowers the catalyst's atomic utilization efficiency.However, due to their atomic dispersion in the carbon matrix, in theory MNCs could achieve 100% active site utilization.In practice, MNC active sites are inaccessible due to a lack of porosity, particularly mesoporosity; therefore, utilization typically remain under 10%. [21]We determined the number of electrochemical active sites per cm 2 (N Nitrite ) by means of in situ electrochemical nitrite stripping experiments (Figure S19, Supporting Information) [20] and compared it to the bulk active site density per cm 2 obtained by ICP (N ICP ) (Equations S4 and S5), to obtain the electrochemical active site utilization efficiency (Utilization Nitrite/ICP ) (Equation S6, Supporting Information). [58]or TAP 900@Fe and TAP 900@Ni, we obtain N Nitrite of 1.9 and 1.2×10 19 sites cm −2 (Table S4, Supporting Information).Considering measured N ICP , we obtain Utilization Nitrite/ICP of 45±14% and 59±6% for TAP 900@Fe and TAP 900@Ni, respectively (Table S4, Supporting Information), the highest reported values to date for MNC catalysts.We would like to note that these values are based on nitrite stripping with a determined 5 e − process; [59] however, if a 3e − process to NH 2 OH takes place, as previously suggested, [59,60] then Utilization Nitrite/ICP for TAP 900@Fe and TAP 900@Ni would be 75±14% and 98±6%, respectively.The Utilization Nitrite/ICP for TAP 900@Fe is likely lower due to inactive Fe contamination within TAP 900 (Figure 2a), which results in an increased Fe N ICP .Nevertheless, these remarkably high values arise from the combined micro-and meso-porosity of the TAPderived materials, which allows for sufficient accessibility to the active sites as well as from the metal-coordination step, which hinders the aggregation and results in atomic dispersion within the C-N scaffold.
To illustrate the performance of the TAP based catalyst, we collate multiple experimental data in Figure 7, where Figure 7a is focused on only CO 2 to CO and Figure 7b is focused on comparison with Cu-based catalyst.In these figures we use the TOF of CO 2 reduced molecules where we defined TOF as the number of CO 2 molecules reacted per active site per second.TOF min and TOF max are defined as the TOF normalized with respect to the number of sites calculated using ICP and nitrite stripping (Equations S7 and S8, Supporting Information), respectively, and a comparison is shown in Figure S20 in the Supporting Information.State-ofthe-art TOF are achieved for both TAP 900@Fe and TAP 900@Ni.TAP 900@Fe and TAP 900@Ni display TOFs of 4.9 and 6.8 e − site −1 s −1 , respectively, at −0.59 V versus RHE (Figure 7a; Table S5, Supporting Information), which are over two orders of magnitude higher than the TOFs of other FeNCs, nanoporous Ag electrode (np-Ag) [58] and carbon black supported Au nanowires with a length of 500 nm (C-Au-500). [61]In Figures S21 and S22 (Supporting Information), we show the TOF of Ni and Fe based MNC, respectively.We observe that the prepared materials show a TOF equal to the highest reported materials of the same class, in particular TAP 900@Ni shows a similar TOF at −0.59 V versus RHE to that reported by Liu et al. with a Ni-CNT composite.In terms of FeNC materials, TAP 900@Fe shows the highest TOF at −0.59 V versus RHE observed in the literature for FeNC materials.Intriguingly comparing MNC materials with metal catalysts in terms of number of CO 2 molecules reduced per site per second, in Figure 7a shows a single fundamental limitation, depicted by a collation of the best experimental data all fall on the same line.The limitation can be simplified to the fact that CO 2 reduction is carbon chemistry; where carbon is activated and subsequently needs to get off the surface.Regardless of the exact CO 2 activation step, whether it is cation-induced, electron-induced or PCET to form *COOH, this study suggests that discussion of the absolute activation step is scientifically interesting, but only societally relevant if it allows research to overcome the barrier shown in the analysis of Figure 7a.
Next, we calculated the TOF for the number of CO 2 molecules that reacted on the Cu sites and on the MNC catalysts in order to evaluate how well the Cu catalysts compare with the best-reported MNC catalysts (Figure 7b).The fact that Cu catalysts exhibit identical intrinsic activity despite having different morphologies is particularly intriguing. [29,62]Although morphology and nanostructuring may enhance geometric current density, they have little impact on intrinsic activity. [29,63]It is also clear that the intrinsic activity of CO2RR to CO on Au-based and MNC catalysts are higher than on the surface of Cu metal-based samples.We attribute this to the fact that Cu binds *CO to enable "beyond-CO" products, however, *CO in this case can also be considered as a poison limiting the TOF of CO2RR.An optimum exists through the scaling of carbon intermediate, via an activation of CO 2 and desorption of *CO to CO. [25] To obtain insights into the catalytic activity of the prepared materials, DFT calculations were carried out.As CO 2 reduction to CO is a two proton-electron reaction, the simplistic picture is a Sabatier volcano as a function of one adsorption energy, in this case *COOH, which corresponds to the proton-electron coupled activated intermediate of CO 2 .Figure 8a demonstrates this volcano with a strong binding leg (reaction: *COOH + (H + + e − ) → CO(g) + H 2 O) and weak (reaction: CO 2 + (H + + e − ) → *COOH) binding leg.Notably, such a construction is similar to the hydrogen evolution volcano (HER), which is also a two protonelectron reaction, where *H controls the reaction.However, the two proton-electron volcano is a too simplistic picture in the case of CO 2 to CO reduction, which needs to include three elements: (i) activation of CO 2 , (ii) competition with HER, and (iii) desorption of CO.The first element is included directly by using the *COOH intermediate, while the two other elements can be added  [51] CoPc2, [64] TAP 900@Ni (this work), TAP 900@Fe (this work), FePGH-H, [65] Oxide-derived (OD) Au, [66] Fe 2+ -NC, [52] Fe 3+ -NC, [52] Fe 3+ -NC-GDE, [52] C-Au-500, [61] np-Ag, [58] Co-N/NPCNSs, [67] Ni-N/NPCNSs, [67] Co-N-Ni/NPCNSs, [67] Fe-NC, and Fe0.5d. [18]b) Turnover frequency (TOF) of reported metal-doped carbon catalysts listing the number of CO 2 molecules reacted per site per second.A-Ni-NSG, [51] CoPc2, [64] TAP 900@Ni, and TAP 900@Fe (this work), FePGH-H, [65] Fe 3+ -NC-GDE, [52] OD Cu foam, [68] OD Cu foil, [69] polycrystalline Cu, [70] and OD Cu nanocubes. [71] using the scaling relations.Competition with HER (*H) is included by dashed vertical lines added from Figure S23 (Supporting Information), which show the scaling relationship between the adsorption energy of *COOH plotted against the *H adsorption energy for metal (111) and penta-coordinated FeN 4 X surfaces.The scaling relationship, due to this difference in binding (on-top vs hollow sites), gives two different intercepts for metals and MNCs.Finally, the CO desorption can also be included by using the scaling between *CO and *COOH from metals of (similar to that of Hansen et al. [25] ), as the *CO and *COOH scaling on MNCs is poor.We are unaware if this is real or an artifact from the electronic structure on MN 4 Cs.If we were to assume that the reaction controlled by the adsorption and desorption of *COOH, the most optimum catalyst would be at the peak of the volcano, where *COOH adsorption is thermoneutral.However, this is not the case, as competing reactions can occur: the HER, further reaction of CO and the self-poisoning of the reaction by CO. [72] Hence, on the weak binding side, Au and Ag occupy an optimal position close to ΔG = 0 for CO (g) → *CO, showing that these catalysts need energy to activate CO 2 , but are not limited by CO desorption.However, Au and Ag clearly do not sit at the top of the *COOH binding volcano, illustrating the fundamental constraint in CO 2 to CO catalysis.Interestingly, while Cu is closer to the top of the volcano, its CO 2 RR TOF is lower than the high TOF for OD-Au as seen in Figure 7a.This reflects that it is important not to have a CO 2 RR catalyst which is too reactive and hence is limited by the desorption of *CO.
To understand the effect of the fifth axial coordination on the   111) surface and FeN 4 -X where X is the ligand attached.MN 4 adsorption data were extracted from the previous reference. [73]The red line represents that the reaction is limited by the strong binding of *COOH and the green line represents that the reaction is limited by the weak binding of COOH. to CO catalysts, TAP 900@Fe and Au have similar TOF as seen in Figure 7a and thus, we can assume they lie around ΔG = 0 line for CO (g) → *CO.Consequently, our data point towards the existence of a fundamental limitation -or ceiling -that prevents the attainment of higher CO 2 to CO activity than Au towards; this limitation is due to the scaling relation between the stability of the activated form of CO 2 (whether it is *COOH or *CO 2 stabilized by a cation) and *CO.To the best of our knowledge, there is no evidence that any solid electrocatalyst, including, the MNC catalyst we report therein, has not been subject to this scaling relation and exhibited better TOF.

Conclusions
In this work, we developed single atom MNC CO2RR catalysts through a decoupled synthesis approach that entailed the selective coordination of metal sites in highly porous N-doped carbon, leading to a record electrochemically active site utilization.The high accessibility of the active sites enabled by the mesoporosity of the substrate as well as the presence of an axial ligand that favorably shifts the binding energy of reaction intermediates.These features enabled a CO 2 to CO electrocatalytic performance equal to the highest TOF in the literature value for MNC materi-als, equivalent, but not surpassing that of Au.Our experimental observation that the most active materials converge in terms of intrinsic activity for CO 2 to CO supports the theoretical notion that scaling relations exists between the stability of the carboncontaining intermediates, showing there is a fundamental limitation to the intrinsic activity.Moving forwards, there are two routes to improving beyond the state-of-the art (a) finding compounds that preferentially accelerate the CO 2 activation step relative to *CO desorption, for instance by manipulating the spin state of MNC catalysts [74] (b) or emulating enzymes such as nitrogenase so that we engineer the catalysts to yield more energy rich C2+ products. [23,75]

Figure 1 .
Figure 1.Synthetic pathway for the preparation of pyrolyzed TAP 900 and the subsequent low temperature metal coordination.M represents Ni or Fe.

Figure 4 .
Figure 4. a) Ex situ XANES and b) normalized first-derivative spectra of as received TAP 900@Fe compared with reference XANES spectra of Fe(II)Pc, Fe(III)PcCl, Fe 2 O 3 and Fe-foil.c) XANES and d) normalized first-derivative spectra of TAP 900@Ni compared with reference XANES spectra of Ni foil, Ni(OH) 2 , and NiO.e) Wavelet transform of the k 2 weighted EXAFS data of Fe foil, Fe(III)PcCl, and TAP 900@Fe.Dotted line represents the position of Fe-Fe and Fe-N/O radial bond distance.

Figure 6 .
Figure 6.ToF-SIMS of Ni ion fragments pre-and postchronoamperometry of TAP 900@Ni in the negative spectrum.Measurements were repeated four times in separate locations and corrected counts normalized to the total ion counts, with the error bars representing the standard deviation.To cleanse the catalyst surface, the samples were first non-interlaced sputtered with a 10 keV Ar + cluster ion beam at ≈10 nA current, until reaching a dose density of 1 × 10 15 ion cm −2 .For measurements a Bi 3 + primary ion beam with 25 keV and 0.5 pA beam current was used for 25 min.

Figure 8 .
Figure8.The limiting potential volcano for CO 2 reduction to CO on a) metal (111) surface and FeN 4 -X where X is the ligand attached.MN 4 adsorption data were extracted from the previous reference.[73]The red line represents that the reaction is limited by the strong binding of *COOH and the green line represents that the reaction is limited by the weak binding of COOH.The horizontal black line at −0.11 V RHE represents the equilibrium potential.Vertical dashed lines represent the ΔG = 0 for the respective reactions.ΔG = 0 for HER is different for Metal and MN 4 surface while it is similar for *CO adsorption/desorption.b) Structures of FeN 4 -Pyridine and FeN 4 -Pyrrole with different adsorbed molecules (*NH 2 , *NH 3 , *CH 3 , *Pyridine, and *Pyrrole) at an axial position.
Figure8.The limiting potential volcano for CO 2 reduction to CO on a) metal (111) surface and FeN 4 -X where X is the ligand attached.MN 4 adsorption data were extracted from the previous reference.[73]The red line represents that the reaction is limited by the strong binding of *COOH and the green line represents that the reaction is limited by the weak binding of COOH.The horizontal black line at −0.11 V RHE represents the equilibrium potential.Vertical dashed lines represent the ΔG = 0 for the respective reactions.ΔG = 0 for HER is different for Metal and MN 4 surface while it is similar for *CO adsorption/desorption.b) Structures of FeN 4 -Pyridine and FeN 4 -Pyrrole with different adsorbed molecules (*NH 2 , *NH 3 , *CH 3 , *Pyridine, and *Pyrrole) at an axial position.