Direct Observation of Electron Donation onto the Reactants and a Transient Poisoning Mechanism During CO₂ Electroreduction on Ni Single Atom Catalysts

Abstract

Single atom catalysts (SACs) are an important class of materials that mediate chemical reduction reactions, a key subset of which is Ni within a carbon support for the electrochemical CO₂ reduction reaction (CO₂RR). However, how the metal atom/clusters and the carbon‐based support act in concert to catalyze CO₂RR is not well understood, with most reports attributing activity solely to the Ni‐N_x/C moieties. To address this gap, we have undertaken a mechanistic investigation, employing in situ X‐ray absorption spectroscopy (XAS) coupled with electrochemical studies and density functional theory (DFT) calculations to further understand how Ni single atoms work in conjunction with the nitrogen‐doped carbon matrix to promote CO₂RR to CO, and how the presence of impurities such as those present in CO₂‐containing waste flue gases (including NO_x, and CN⁻) changes the catalyst upon reduction. In contrast to previous works, we do not find strong evidence for a purely metal‐based reduction upon application of negative reductive potentials. Instead, we present evidence for an increase in the equatorial vs. axial splitting of Ni, consistent with electrons moving onto the reactants via the Ni single atom 3d_z ² orbital. In addition, we demonstrate a transient poisoning mechanism of the Ni SAC by nitrite and thiocyanate, explaining the recovery of activity during CO₂RR. These insights can aid the design of practical CO₂ valorization technologies.

In situ techniques reveal key insights into the electrochemical reduction of CO₂ on Ni single atom catalysts. Through poisoning studies, we reveal that the reduction of the system does not result in a metal‐based reduction, rather it is associated with donation of electrons onto a combination of the support and the reactants. We demonstrate the suitability of these catalysts for application to CO₂ waste streams which have unavoidable impurities.

Introduction

Metal‐organic and coordination compounds are well‐known catalysts for the CO₂ reduction reaction (CO₂RR). These can be found in nature as metalloproteins, for example as carbon monoxide dehydrogenase,[ ¹ , ² , ³ , ⁴ , ⁵ , ⁶ ] and are also known as molecular compounds such as iron porphyrin.[ ⁷ , ⁸ , ⁹ ] However, one of the significant challenges of synthetic coordination compounds in practical applications is the lack of stability under reaction conditions. ^[10] This can be overcome by incorporating single atoms in different host materials, i.e., the support. While single atom catalysts (SACs) are commonly considered as a single class of catalyst, in reality they consist of a range of different structures where metal atoms are distributed on various support materials ^[11] (with three distinct subtypes shown in Figure 1a–c), and they all may play a different role in CO₂RR catalysis. An important distinction within these structures relates to how the support interacts with the metal atom. For some systems, the support material is largely innocent, meaning it is not involved in the transfer of electrons either through bonds or two and from other molecules. In other systems, the support may play a key role in stabilizing the electronic structure of the metal in ways that are key for function, hence playing a critical role in the catalysis ^[11] For example, only certain configurations of atoms can stabilize and accommodate square planar Nickel (Figure 1a), as such, both the donor types and geometry must work in concert to achieve an observed electronic structure. ^[12] Further, some structures such as those based on delocalized p* systems have substantial capacity to both donate and accept electrons, which can be key for electrocatalysis.

Figure 1.

Simplified schematics of families of single atom catalysts, including (a) a support that plays a key role in stabilizing the metal active site, (b) a single atom weakly absorbed or associated with the support, ^[11] and (c) a substitutional or partially substituted single atom bound to the surface of the support; for example, there may be substitutional doping on the surface. (d) CO₂RR on a plausible atomic‐level depiction of a Ni SAC such as that described in this paper. Inset: Ligand field splitting of a 3d⁸ compound with a 3d_z ² orbital, presenting an optimal geometry to overlap with LUMO orbital of CO₂. Note: LUMO is the lowest unoccupied molecular orbital).

Recently highlighted in the literature, ^[6] a low spin 3d⁸ configuration associated with square planar Ni(II), ^[12] Co(I),[ ¹³ , ¹⁴ ] and Fe(0)[ ⁷ , ⁸ , ⁹ ] is particularly favorable for catalyzing CO₂RR, as the occupied 3d_z ² orbital that sits above the plane of the material has the optimal geometry to overlap or interact with an incoming reactant (Figure 1d). ^[15] An important feature of a catalytic system is that the reactant‐catalyst bonding is neither too strong nor too weak, as the former will result in poisoning or even decomposition of the catalytic sites, ^[16] while the latter will result in poor activity. ^[17] Square planar compounds leave an axial site free[ ¹⁸ , ¹⁹ ] to bond with an incoming reactant, and the occupied 3d_z ² orbital has the right symmetry to donate electrons into the lowest unoccupied molecular orbital (LUMO) of CO₂. These sites are often described as undercoordinated or unsaturated,[ ²⁰ , ²¹ ] however the square planar geometry common to these systems is stabilized by the 3d⁸ electron configuration.[ ⁷ , ⁸ , ⁹ , ²² ]

SACs are increasingly becoming one of the most important classes of transition metal catalysts, having undergone substantial innovation and development over the last decade in the field of energy conversion reactions[ ²³ , ²⁴ ] and renewable Power‐to‐X,[ ²⁵ , ²⁶ ] including CO₂RR to commodity products such as CO, syngas, and various liquid products. These materials have shown great versatility, and have opened up a new family of heterogeneous catalysts.[ ²⁷ , ²⁸ ] A prominent and interesting example of these materials which mediate CO₂RR are Ni SACs on a nitrogen‐doped carbon support, which are reported to display one of the best performances in terms of both faradaic efficiency (FE) and current density (j) for CO₂RR to CO. ^[29] For instance, Ni SACs exhibiting a Ni‐(N/C)₄ structure have been widely reported to achieve FEs for CO (FE_CO) and current densities of up to ca 100 % and −50 mA cm⁻² in standardized H‐cell systems, respectively.[ ³⁰ , ³¹ , ³² ] The performance of Ni SACs has been further scaled up in membrane electrolyzer assembly (MEA) systems utilizing a gas‐phase CO₂ input, attaining industrially relevant current densities of 300–500 mA cm⁻².[ ²⁹ , ³⁰ , ³³ ]

Importantly, efficient catalysis of CO₂ reduction by Ni SACs reported by us ^[29] and others ^[20] requires reduction of the catalyst either directly before or concomitantly with CO₂. ^[29] In our previous work, we reported on the extent to which competing reactants/impurities (e.g., NO₂ ⁻ or SCN⁻) can interfere with or compete with CO₂ reduction at varying Ni‐N−C moieties, affecting the activity and product selectivity. ^[29] This practical investigation was supported by ex situ XAS (identifying Ni single atoms) and theoretical calculations, correlating the adsorption of simulated impurity molecules with the Ni coordination structure (Ni‐N₄ and Ni‐N_4‐x‐C_x), determining that reduced free energy of the CO₂RR rate determining step could play a more significant role than impurity adsorption when comparing the differing Ni coordination structures.

However, the role of such impurities is also important from (i) a mechanistic perspective, where understanding reactants with known differences in bonding may provide key insights into catalyst mechanisms, and (ii) for practical CO₂RR, which requires the catalysts to maintain activity in the presence of impurities. If the latter is not provided, capital and operational expenditures associated with separating and purifying CO₂ from waste streams would significantly increase the cost of the process. As such, the application of in situ XAS measurements is critical to providing key insights into these key structure–activity questions and is the focus of the present work. Whilst some studies of in situ XAS on Ni SACs have been reported,[ ¹² , ³¹ , ³⁴ ] the data did not always show clear and systematic changes to Ni as a function of potential. Gaining further understanding of these changes may help us directly understand how Ni SAC materials facilitate the chemical reduction of CO₂.

To this end, we mechanistically investigated a Ni single atom catalyst synthesized by a freeze‐drying approach[ ²⁸ , ²⁹ ] by ex situ and in situ XAS during CO₂RR, in both the absence and presence of electrolyte impurities (SCN⁻ and NO₂ ⁻), intended to mimic those of industrial CO₂‐containing flue gas. ^[29] Importantly, our experiments allowed us to compare reactant selectivity (i.e., the selectivity for CO₂ reduction compared to other reactants/impurities), revealing key differences that are only observable in situ, providing valuable insights into how these systems work and the competition between the CO₂ reactant and the impurities, NO₂ ⁻ and SCN^‐. By utilizing a system optimized for thin catalyst films and employing a continuously flowing electrolyte that maintained a consistent electrode‐electrolyte interface, ^[35] we obtain substantially improved XAS data and investigate changes that occur as function of both applied potential and the species present. In contrast to previous works, we do not find evidence for a clear metal‐based reduction of Ni(II) to Ni(I), but reveal that key reduction events occur via the donation of electrons to the reactant(s), rather than exclusively to the Ni(II) center (forming Ni(I)). This knowledge can be used to develop highly stable, active, and selective catalysts for CO₂RR to CO under realistic operating conditions.

Results and Discussion

The Ni SAC described herein was synthesized according to our freeze‐drying and annealing protocol (Methods).[ ²⁸ , ²⁹ ] As reported in our previous work, the catalyst demonstrated a high tolerance to the electrolyte impurities, ^[29] and as such, we investigate this catalyst (referred to as NiSAC from here on in) in the present work.

Key ex situ data before electrochemical activity testing are presented in Figure 2 and in Figure S1–S8, revealing the following: (1) annular dark‐field scanning transmission electron microscopy (ADF‐STEM) displays an absence of Ni nanoparticles and the existence of Ni single atoms (Figure 2a–c) distributed across a porous carbon matrix (Figure S1–S2). Each bright spot in the image represents a single Ni atom, evidenced by ~0.15 nm intensity profile (Figure S3) across the atom; (2) STEM‐energy electron loss spectroscopy (EELS) maps confirm fairly even distributed Ni atoms across the carbon matrix. Nitrogen is also evenly distributed, whereas the edge of the carbon matrix is more oxidized (Figure S4); while TEM images exhibit some apparently dense blocky components, the absence of any secondary phases in XAS (see below) is consistent with these features arising from thicker, overlaid and/or folded parts of the material under study (3) X‐ray diffraction (XRD) and Raman spectroscopy (Figure S5–S6) show clear peaks for graphitic carbon without the presence of detectable amounts of crystalline Ni phases. (4) Ni 2p X‐ray photoelectron spectra (XPS) shows a sharp peak at ~855 eV, similar to square planar Nickel bis(dimethylglyoximate), Ni(DMG)₂ (Figure S7), consistent with the assignment of low spin Ni(II), with the NiSAC consisting of ~3 wt % Ni (Table S1–S2), and (5) ex situ XANES (Figure 2d) energy lies between metallic nickel and nickel oxide, and is similar to reports for square planar nickel (Figure S8). The EXAFS of these materials is well fit by 4× Ni−N distances and is markedly distinct from that of both metallic nickel and nickel oxide. This is consistent with the fact that these systems are dominated by distinct metal‐nitrogen square planar species and are neither metals nor metal oxides, unlike those observed in systems based on other metals like Cu, Sn and Fe.[ ¹⁸ , ³⁶ ]

Figure 2.

(a–b) ADF‐STEM images of NiSAC, with single atoms evenly distributed across the carbon support (some single atoms are circled in red). (c) STEM‐EELS mapping of the C, O, and N K‐edges, and the Ni L₃‐edge. XAS data on the four materials before and after testing. (d) XANES spectra at Ni K‐edge for Ni foil, as‐synthesized NiSAC, and NiSAC after testing in impurity‐free electrolyte, NO₂ ^‐ impurity, and SCN⁻ impurity. (e) Fourier Transform of the EXAFS (the fit is given in Table S6, the main peak of the Fourier transform is well fit by 2x Ni−N @ 1.91±0.01 Å and 2x Ni_N at 2.07±0.03 Å, consistent with square planar nickel), and (f) EXAFS. (g) Dependence of the faradaic efficiency for CO and H₂ on applied potential for NiSAC (CO₂‐saturated 0.1 M KHCO₃ electrolyte). Variation in FE with the presence of (h) NO₂ ⁻, and (i) SCN⁻ in electrolyte, and upon removal of impurity. Corresponding chronoamperograms are shown in Figures S12–S13. Data points are presented as mean values and error bars indicate standard deviations obtained from triplicate (n=3) measurements.

Having confirmed that the NiSAC ex situ XAS data are consistent with all of the features of the single atom catalyst described previously, ^[29] we then determined its catalytic activity for CO₂RR in a two compartment H‐cell system. Through cyclic voltammetry (CV), the catalyst displays a much higher current density in the CO₂‐saturated electrolyte compared to saturation with Ar (Figure S9a), demonstrating that NiSAC is active for CO₂RR. Through potential‐dependent chronoamperometric activity testing (Figure 2g), a FE_CO of >94 % was achieved within a potential range of −0.7 to −1.0 V vs. RHE, with a maximum FE_CO of ca 99 % at −0.9 V vs. RHE. This activity was maintained for 20 h of operation (Figure S9c). Gas chromatographic (GC) and ¹H nuclear magnetic resonance (NMR) analyses indicate CO and H₂ were the only gas‐phase products of the reaction, with no liquid products generated. The maximum FE_CO achieved (ca 99 % at −0.9 V vs RHE) is comparable to benchmarked Ni SACs for electrochemical CO₂RR in the literature (Table S3), however the j attained (ca 20 mA cm⁻²) at this potential falls short of the values reported in the literature for benchmarked Ni SACs (up to 70 mA cm⁻²), indicating that the number of active sites and/or the reactant access to active sites may be limited. A double layer capacitance value of (C_dl) of 1.03 mF cm⁻² was calculated for NiSAC (Figure S10), indicating an electrochemical surface area of ~49 cm² _ECSA per cm² _GEOMETRIC. The ECSA‐normalized j is shown in Figure S9b.

To evaluate the impurity tolerance of NiSAC, ^[29] we then undertook CO₂RR at selected potentials (Figure 2h ‐i) with the presence of 50 mM NO₂ ⁻ and 10 mM SCN⁻ impurities, respectively. The combined FE of CO and H₂ drops upon the addition of NO₂ ⁻ to the electrolyte during CO₂RR (Figure 2h). This is attributed to the diversion of charge towards the nitrite reduction reaction (NO₂RR) to NH₄ ⁺, confirmed via the Berthelot test. ^[37] The overall FE is close to 100 % indicating that no other minor products are likely formed. A faradaic efficiency for ammonium (FE_NH4 ⁺) of ca 6.5 %, 3.4 %, and 3.6 % was attained at −0.7, −0.8, and −0.9 V vs. RHE, respectively. This indicates that the catalyst displays some activity for the NO₂RR, in agreement with other reports on SACs for this reaction.[ ³⁸ , ³⁹ ] This is corroborated by the CVs recorded with the nitrite addition (Figure S11a‐b), in which a new reductive process commencing at approximately −0.3 V vs. RHE was detected and attributed to the NO₂RR, however, the chronoamperograms with each electrolyte condition are similar (Figure S12), indicating a comparable overall reaction rate with nitrite presence.

The effects of SCN⁻ impurity were then similarly investigated during CO₂RR (Figure 2i), as this anion can bind to and poison the single atom active sites.[ ²² , ³² , ⁴⁰ ] At −0.8 and −0.9 V vs. RHE in the presence of 10 mM SCN⁻, a slight decrease in the FE_CO was noted (from ca 87 % to 83 %, and from 88 % to 82 %, respectively), indicating that SCN⁻ may block the active sites, however, the activity is recovered when the electrode is then tested in fresh electrolyte, indicating that the poisoning effect is transient, unlike other materials such as Pt, Ag, and Cu surfaces where CO, SO_x, NO_x, CN⁻, and SCN⁻ can more permanently adsorb to or alter the surface.[ ⁴¹ , ⁴² , ⁴³ , ⁴⁴ ] The production of H₂ remains consistent as the HER could proceed on a range of defective carbon or C−N sites present in the catalyst, rather than at the Ni centre.[ ² , ³ , ⁴⁵ , ⁴⁶ ] As the overall FE slightly drops below 100 %, there may be some reduction of SCN⁻. The CV results comparing CO₂RR with and without the SCN⁻ impurity (Figure S11c–d) show a slight decrease in j, which may occur due to some adsorption of SCN⁻ to the active sites during reaction,[ ²² , ³² , ⁴⁰ ] which is corroborated by the similar chronoamperograms with and without SCN⁻ presence (Figure S13). In general, the minimal change in catalytic activity with SCN⁻ presence aligns with the results in our previous work. ^[29]

Ex situ XAS measurements (Figure 2d–f) indicate very similar spectra for all the electrodes, indicating that ex situ measurements cannot differentiate any changes to the catalyst arising post‐reaction. This result is further supported by post‐reaction XPS measurements (Figure S14–S16 and Table S4). Specifically, no notable differences between the N 1s or S 2p spectra are seen between catalysts tested in impurity‐free electrolyte compared to those with either NO₂ ⁻ or SCN⁻ present, indicating that the effect on electrocatalytic activity variation (Figure 2g–i) cannot be adequately probed by ex situ measurements. Importantly, in our previous work, we observed that annealing NiSAC at elevated temperatures (>1000 °C) with a high Ni loading might induce formation of metallic Ni, ^[29] which does not directly participate in the electrochemical CO₂RR. We have optimized the synthesis approach to strongly suppress the formation of this undesirable admixture to achieve >97 % NiSAC purity and undetectable contribution from other forms of nickel within the catalyst (see Figure S8 showing very high sensitivity of XAS to Ni and NiO, as well as Figure S29 which explores linear combination analysis and accompanying discussions in the Supporting Information). From the data presented herein, there is also no evidence for the formation of even small quantities of either NiO or metallic nickel in our materials post reaction (Figure 2). This high level of catalyst purity enables direct analysis of the NiSAC active sites by in situ XAS studies without interference from other nickel‐based species, as was the case in some of the previous studies. ^[47]

To understand the function of the catalyst and better understand the resilience to impurities, in situ XAS (Figures S17–S20) was conducted at the Ni K‐edge in the presence of competing impurities/reactants as a function of applied potential, in a flow cell we have previously described. ^[35] X‐ray Absorption Near Edge Structure (XANES) data taken in situ under four different experimental conditions (Ar‐saturated electrolyte, CO₂‐saturated electrolyte, and CO₂‐saturated electrolyte in the presence of NO₂ ⁻ or SCN⁻) are shown in Figure 3. Importantly, unlike the ex situ data (Figure 2d–f), notable differences between the materials are apparent in both the XANES (Figure 3a–d, Figure 4a) and the extended X‐ray absorption fine structure (EXAFS, discussed below). The XANES rising edge of the first transition series metals are dominated by two major sets of transitions: those to the unoccupied 3d orbital, and those to the unoccupied 4p orbitals. i.e., Ni 1s→3d and 1s→4p transitions.[ ⁴⁸ , ⁴⁹ ] In square planar Ni, (which is the case for the NiSAC system ^[29] ) the two contributions to the transitions arise from 1s→ 3d_x ²‐_y ² transition, this transition will shift in energy depending on ligand field strength. ^[50] In some systems, the 3d orbitals gain intensity from mixing with the 4p orbitals, however, this is not observed substantially in this work, indicating that there are not substantial changes in centrosymmetry around the Ni as a function of applied potential. One of the most striking features of the XANES of square planar Nickel is the lowering of the 4p_z concomitant with movement of the 4p_xy to higher energy, as was shown by angular dependent polarized XAS experiments on single crystals. ^[48] Our DFT calculations (detailed below) show that this lowering of the 4p_z orbital (and the states that arise) come from the stabilization of the 4p_z orbital (and the states that arise) by interaction with the occupied 3d_z ² orbitals (Figure 4 and Figures S31–S36). A scheme depicting these major sets of transitions is given in Figure 3e.[ ⁴⁸ , ⁴⁹ ] The XANES data on the Nickel SAC systems suggests (Figure 3a–d) that changes to this splitting pattern occur as a function of both the impurity present and the applied potential. It is striking that the XANES do not clearly shift to lower energy as a function of applied negative potential, as is seen in related Fe systems ^[51] and Cu(II,III) systems, ^[52] rather there is a substantial change in the shape of the XANES where it distributes to both higher and lower energy.

Figure 3.

XANES comparing in situ experiments with various electrolytes and applied potentials. (a) Ar‐saturated 0.1 M KHCO₃, (b) CO₂‐saturated 0.1 M KHCO₃, (c) CO₂‐saturated 0.1 M KHCO₃ with 50 mM NaNO₂, and (d) CO₂‐saturated 0.1 M KHCO₃ with 10 mM KSCN. (e) A schematic depicting the contributions to the Ni rising edge structure. (f) The major transition contributions affecting the XANES data of square planar Nickel. Note: CI; configuration interaction.

Figure 4.

(a) A summary of the key interactions which contribute to the observed spectral shape. The effect of adding electrons to the system is indicated. (b) Density of states calculations showing the effect of the introduction of impurities and chemical reduction on the 3d and p manifolds, calculations for Ni pyridine (NiSACpy) are shown. Manifolds of Ni, in the presence and absence of CO₂, considered in the calculation as CO₂H, as this is needed for binding and the presence and absence of NO₂ ⁻ are shown. The SCN data is shown in Figure S36. (c) Experimental differences in splitting patterns observed for NO₂ ⁻ vs. CO₂ as a function of potential depicted at −0.2 and −1.1 V vs. RHE (See Figure S22 which includes an additional potential). The change is also observed in the density of states calculation, where we see the orbital formed between the occupied 3d_z ² and the LUMO on the NO₂ ⁻ become occupied before the equivalent on CO₂ (see green highlighted states green and red arrows Figure 4b).

Fundamentally, there are two contributions to the position and shape of the rising edge. First, the effective nuclear charge on the Ni, strongly associated with redox state, but not identical to it (Note that redox state and effective nuclear charge are not the same, because in redox state the orbital occupancy of metal matters). This will influence the average XANES position. Second, how the electronic structure or orbital occupancies and the interacting states changesdepends on what is bound to Ni, how it is bound, and how long it is bound over the time scale of the experiment (discussed below). Where there is a spherical bonding around the metals (i.e., an octahedral or a tetrahedral geometry), a metal‐based reduction would shift the XANES to lower energy as the p orbitals are influenced equally.[ ³⁵ , ⁵³ ] However, the square planar geometry of the NiSAC has a substantial effect on the shape of the XANES rising edge structure, and the changes seen in this transition provide important information for what is happening at the metal center.

A scheme depicting the major transition contributions to NiSAC is given in Figure 3f. Z_eff shifts the average XANES left (lower redox state) and right (higher redox state), with changes in axial vs. equatorial bonding changing the splitting between the 4p_xy/continuum and 4p_z orbitals, and hence the shape of the XANES. As a reductive potential is applied, the XANES of NiSAC splits more between the 4p_z and 4p_xy/continuum, rather than less, and the average XANES position does not move very significantly (Figures S21–S22). This suggests that the majority of the changes observed in the experimental data shown in Figure 3a–d are related to splitting (i.e., the shape of the spectra) rather than to the effective nuclear charge (the average shift of the spectra). This is clear experimentally from the in situ data, where the top part of the XANES spectra shifts to higher energy, and the lower energy part to lower energy (indicated by the arrows in Figure 3b–d). This phenomenon is also well reproduced by our DFT calculations, both as Bader charges and as density of states, which are described in detail below and in the Supporting Information.

Differences in the bonding at the Ni active sites between the different reactants/impurities can also be compared. The largest variation in the systems can be seen in Figure 3b (CO₂‐saturated electrolyte) and Figure 3d (CO₂ and SCN⁻ presence), where the most substantial change observed is with the SCN⁻ poisoned system, in which the splitting between the 4p_z and 4p_xy/continuum (depicted by the arrows in Figure 3d) increases more than with either CO₂ or NO₂ ⁻. In the case of NO₂ ⁻, the splitting of the orbitals is greater at −0.2 V vs. RHE compared to CO₂, however at −1.1 V vs. RHE their splitting is similar (Figure 4a). This is consistent with NiSAC binding NO₂ ⁻ and catalyzing its reduction at less negative potentials relative to CO₂, as shown by our experimental results (Figure 4a).

To understand the changes in the XAS of these systems, density functional theory (DFT) calculations were undertaken, using three different models for NiSAC, a pyridine model, a pyrrole model, and Ni tetraphenyl porphyrin (Nitpp), Figure S23. ^[12] With the different carbon backbone and models for NiSAC, these calculations result in slightly different Ni−N distances (from 1.87–2.02 Å), due to the different electronic and steric effects of the carbon support model. However, and most importantly, the trends in the change in the electronic structure of the nickel metal site as electrons are added to the system, are the same for all models. These calculations are detailed below and in the Supporting Information, and were undertaken in order to understand the likely metal‐support interactions in NiSAC and their expected trends in the key observations in the XAS; the effective nuclear charge and the energies of the states arising from the orbitals of Ni. For simplicity, our discussions focus on the pyridine SAC model (identified as Ni(SACpy) here), however calculations on the pyrrole and porphyrin models of NiSAC showed the same trends Table S8. [NiSACpy] and [NiSACpy]²⁻ are used to model the catalyst without any reactant present, and are compared to [Ni(SACpy)CO₂H], [Ni(SACpy)CO₂H]²⁻, [Ni(SACpy)NO₂]⁻, [Ni(SA‐Cpy)NO₂]³⁻, and [Ni(SACpy)SCN]⁻, to model the catalyst with CO₂ or impurities present (Figure 4b and Figure S33). The changes that take place in both the geometric and electronic structure of these compounds were compared systematically. The density of 3d and 3p/4p states observed in the calculations (Figure 4b) can also be coupled to the observations made from the XANES. As noted above, one of the clearest observations that can be made from this data for these systems is that the XANES do not simply shift to lower energy as a negative potential is applied, which would be expected if a metal based reduction of Ni(II) to Ni(I) occurred. Rather, we note that the splitting associated with the 4p orbitals increases, consistent with changes to the way the 4p_z and 3d_z ² orbital interact and is directly due to the strong donation from the 3d_z ² orbital into an unoccupied CO₂ orbital (Figure 1d). ^[6] The DFT calculations (shown as density of states, Figure 4b) clearly show both of these effects, and are consistent with the XANES observations. The 3d density of states shows that even when electrons are added to the system, the LUMO orbital of [NiSACpy]²⁻, and [NiSACpy CO₂H]²⁻ are all 3d_x ²‐_y ²; if a reduction event had been localized on Ni, this orbital would become occupied, i.e., even when electrons are added to the system it remains as Ni(II) and does not reduce to Ni(I). Further, when CO₂ or NO₂ ⁻ are added to the system, the donation of the 3d_z ² orbital into the unoccupied reactant molecule is clear (represented by red lines in Figure 4b), which increases as more electrons are added to the system. This splitting occurs as a result of the interaction between the 3d_z ² orbital based on the metal and LUMO orbital of CO₂/NO₂ ⁻ _. It is evident from the XAS results that this interaction increases with applied potential, and that the splitting between the 4p_z and the 4p_xy/continuum orbitals increases with applied potential and is different for NO₂ ⁻ vs CO₂, with NO₂ ⁻ splitting at lower over potential (Figure 4c). While there have been other in situ studies of Ni SAC systems, some of which assigned a nickel based redox event Ni, it is clear from the more resolved XANES observed here and the subtle changes in XANES that occur as a function of both the reactant/impurity and the potential, that the redox state of the metal does not reduce substantially, rather as the system is reduced electrons are mostly donated to a combination of the reactant(s) and the support.[ ¹² , ³¹ , ³⁴ ]

Systematic and consistent changes are also seen in the EXAFS as a function of potential, that can be compared to the DFT calculations. As shown in Figure 5 and Figures S24–S26, examination of the EXAFS reveals systematic changes as a more negative potential is applied. Most clear is the shift in the phase of the EXAFS evident around k=4 Å⁻¹, which is observed for all systems. While the real systems are likely a distribution of slightly different but highly related sites, because the catalyst is the same in each case, the DFT calculations on the pyridine Nickel SAC model (NiSACpy) can be used to examine likely geometric trends that could occur as a function of potential. ^[54] From a geometric perspective, as electrons are added to the system there is an increase in the average distances of the second coordination sphere in the system, and little change in the first coordination sphere of Ni SACpy (Figure S23 and Tables S9–S10).

Figure 5.

EXAFS and Fourier Transform data for (a, c) CO₂‐saturated 0.1 M KHCO₃, and (b, d) CO₂‐saturated 0.1 M KHCO₃ with 10 mM KSCN. Fits to this data are given in Table S8. Scheme depicting how XAS sees a process with different timescales, for example between two reactants/impurities that are (e) always observed, and (f) observed only 20 % of the time.

The major shifts observed in the EXAFS for Ni(SACpy)CO₂H are well described by a fairly consistent first coordination sphere and a slightly increased second coordination sphere with multiple Ni..C interactions (Table S8), consistent with our insights drawn from DFT calculations, where the reduction event is localized on the carbon support. In addition to this phase shift, the EXAFS in the presence of SCN⁻ increases in intensity, as indicated by the arrows in Figure 5. While intensity changes in EXAFS can be due to changes in disorder or coordination number, as the system is identical otherwise the increase in the EXAFS intensity is likely due to a longer residence time of SCN‐ binding relative to CO₂, as depicted schematically in Figure 5e–f and reproduced in the EXAFS fits, Figures S25–S26. The XAS experiments probe the system averaged over time; as catalysts are dynamic systems, with competitive bonding we see the average or a “snapshot” of all processes in the system. In a system where two species are competing (i.e., CO₂ vs. SCN⁻, or CO₂ vs. NO₂ ⁻), we see a time average of the system. Therefore, if a molecule is bound to the catalyst 20 % of the time, it will contribute only 20 % of the observed XAS data. From the data we observe herein, the SCN⁻ stays bound to the Ni over a longer time than CO₂, and as such, the EXAFS is stronger. The timescales and lifetimes of bound states are also important in this scenario. A very short‐lived active state may never be seen in XAS. ^[35] Similarly, if a catalyst is poisoned 95 % of the time, an XAS experiment will see 95% of the poisoned form of the material. The impact of these scenarios in the EXAFS is illustrated schematically in Figure 5e (where the impurity/reactant is always observed), and Figure 5f (where the impurity/reactant is observed only 20 % of time). These Schemes are simplified from the “real” experimental scenarios, as multiple reactants and bonds are present in dynamic systems.

Conclusions

In summary, we carried out a mechanistic study on the dynamic poisoning of a Ni single atom catalyst. As a result of the unique geometric and electronic structure of square planar Ni, key insights can be gained through subtle changes arising from the sensitive in situ XANES data. In contrast to the common assumption that key redox processes are associated with a reduction to Ni(I), in this study we reveal that the reduction of the system does not solely result in a metal‐based reduction, rather it is associated with a concomitant donation of electrons onto a combination of the support and the reactants(s), where the metal electronic structure is relatively consistent as low spin Ni(II). This electronic structure is the reason that poisoning of these systems is only observed through in situ measurements. The transient nature of the poisoning via electrolyte impurities is notably different to those reported in other metal‐based systems, highlighting the suitability of these catalysts for application to CO₂ waste streams which have unavoidable impurities.

Additional Information

See Supporting Information for: SEM images, TEM images, XRD, Raman, XPS, electrocatalytic data, EXAFS fitting, activity comparisons, DFT methods and results, and supplementary tables summarizing ICP measurements, XPS measurements, ICP measurements, activity comparisons, and XAS measurements.

Author Contributions

R.D., R.A., and R.K.H conceived the concept, designed, and directed the research. J.L. conducted material synthesis, characterization, and electrochemical measurements. J.L., G.B., T.T−P., and R.K.H. conducted XAS experiments, analyzed the results, and performed the fitting. J. Y., P.K., and R.K.H. designed and carried out the DFT measurements and analysis. H.W. and S.L.Y.C. undertook TEM imaging and analysis. B.J. and D.S. assisted in XAS experiments. A.N.S, A.T., and L.D. contributed to the data interpretation and analysis. J.L., R.K.H., R.D., G.B, R.A., J.Y., P.K. co‐wrote the manuscript. All authors discussed the results and contributed to the final version of the manuscript.

Conflict of Interests

The authors declare no competing interests.

Supporting information

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supporting Information

Leverett J., Baghestani G., Tran-Phu T., Yuwono J. A., Kumar P., Johannessen B., Simondson D., Wen H., Chang S. L. Y., Tricoli A., Simonov A. N., Dai L., Amal R., Daiyan R., Hocking R. K., Angew. Chem. Int. Ed. 2025, 64, e202424087. 10.1002/anie.202424087

Data Availability Statement

All data that support the findings of this study are available from the corresponding author with reasonable request.