Insights into the mechanisms of NH3 inhibition on Cu-CHA SCR catalysts Authors: Dhruba J Deka, Mingyu Wan, Garam Lee, Eric Walter, Fanglin Che, Kenneth G Rappe, János Szanyi, Yong Wang Published: 2026 ABSTRACT AbstractAmmonia (NH3) inhibition during the selective catalytic reduction (SCR) of nitrogen oxides (NOx) over Cu-exchanged chabazite (Cu-CHA) catalysts limits low-temperature performance, yet its molecular origin remains unclear. Here we show that excess NH3 selectively suppresses the oxidation half-cycle of the SCR reaction while leaving the reduction half-cycle largely unaffected. By combining kinetic measurements, operando electron paramagnetic resonance spectroscopy, and density functional theory calculations, we identify hindered mobility of Cu+ ions as the key factor. Specifically, NH3 coordination increases the energy barrier for Cu+ diffusion, preventing the formation of reactive Cu2+-oxo dimer intermediates required for efficient oxidation. Spectroscopic measurements further reveal that the extent of inhibition depends strongly on temperature and copper loading. These insights provide a mechanistic basis for mitigating NH3 inhibition, suggesting that improved catalyst design and optimized operating conditions can enhance low-temperature SCR performance in practical emission control systems.Subject terms: Heterogeneous catalysis, Catalytic mechanisms, Chemical engineering Introduction The selective catalytic reduction (SCR) of nitrogen oxides (NOx) with ammonia (NH3) over small-pore Cu-zeolites (e.g., Cu-CHA, Cu-AEI, Cu-LTA) is a cornerstone technology for controlling NOx emissions from lean-burn engines1,2. These materials offer high activity, selectivity, and hydrothermal stability, rendering them indispensable in advanced emissions control systems. In Cu-CHA, the primary active sites are isolated CuII ions that exist in two distinct coordination environments: ZCuIIOH, where CuII is coordinated to a single framework Al and a hydroxyl group, and Z₂CuII, where CuII is stabilized by paired Al sites in the framework3–5. The NH3-SCR reaction (4NO + 4NH3 + O2 → 4N2 + 6H2O) proceeds via a redox mechanism involving a reduction half-cycle (RHC) and an oxidation half-cycle (OHC), with isolated Cu sites cycling between CuII and CuI states4–8. The RHC is achieved through the reduction of CuII to CuI by NO and NH3, forming linear [CuI(NH3)2]+ complexes. The solvation effect of NH3 on CuI sites decreases their electrostatic interaction with the negatively charged zeolite framework, granting mobility to these complexes9–11. The OHC involves the activation of O2 on CuI sites. Since O2 activation requires four electrons, and CuI can provide only one, the process is more facile when two CuI sites are involved which leads to the formation of O2-bridged dimeric CuII species4,5,12–14. The mobility of [CuI(NH3)2]+ complexes is crucial for the OHC as they need to diffuse between cages to co-locate with another [CuI(NH3)2]+ to form Cu dimers. This quasi-homogeneous process dominates below ~250 °C, while at higher temperatures Cu(I) ions become immobilized, and SCR proceeds via more heterogeneous pathways involving mononuclear Cu-nitrate/nitrite intermediates4,15–18. The mobility of Cu ions, especially at low temperatures, is influenced by several physicochemical properties, including zeolite topology, Si/Al ratio, Al siting, Cu loading, and Cu speciation5,9,19–21. For example, a higher Si/Al ratio reduces Brønsted acidity and Cu exchange capacity, impacting both reactivity and diffusion3. Spatial inhomogeneities in Al distribution can further constrain Cu mobility due to the difficulty of intercage migration between Al-rich and Al-lean regions9. Cu loading also strongly affects Cu-Cu pairing probability and, thus, OHC kinetics: higher loadings promote dimer formation, while lower loadings reduce Cu ion encounters and amplify diffusion limitations20. Reaction environment and solvation further modulate Cu mobility. NH3 and H2O solvation lowers mobility restrictions at low temperatures, but excess NH3 can occupy Brønsted sites and obstruct critical diffusion pathways (e.g., 8-membered ring windows), thereby impeding Cu transport3. Excess reactants such as NO, H2O, and O2 can further increase the energy required for [CuI(NH3)2]+ diffusion.22. Additionally, certain reactants, such as NO2 or O2, can destabilize mobile CuI species or oxidize them into framework-bound CuII, further limiting mobility and redox cycling23. These observations suggest that SCR performance is highly sensitive to local reaction conditions and the coordination environment of Cu sites. Reactant-induced inhibition is a broader catalytic phenomenon wherein excess reactants or intermediates impair activity by forming inactive complexes, competing for adsorption, or hindering diffusion23–25. Understanding these mechanisms and their impacts is essential for optimizing catalytic processes. The inhibition of SCR by NH3 is well-documented, particularly on vanadium-based catalysts, where NH3 competes with NO for active V2O5 sites26. NH3 inhibition of fast SCR reactions on Fe-zeolites, on the other hand, results from NH3 interacting with surface nitrates, preventing their reaction with NO27. However, despite growing interest in Cu-CHA, NH3 inhibition in standard SCR has received limited attention. Most existing insights derive from theoretical predictions, indirect experimental evidence, or demonstration under transient conditions19,22,28–31. The lack of fundamental studies can be attributed to the focus on high Cu-loadings, where inhibition is not easily observable in steady-state tests. In real-world applications, Cu sites deactivate with time, reducing effective loading and thereby exposing the catalyst to NH3 inhibition, especially under onboard excess dosing strategies aimed at avoiding parasitic NH3 oxidation losses. Furthermore, NH3 internal combustion engine exhaust may contain up to 2–3 times as much NH3 as NOx, raising the potential for NH3 inhibition even on relatively fresh Cu-CHA catalysts. In this study, we uncover the atomic-level mechanisms of NH3 inhibition on Cu-CHA catalysts by integrating steady-state kinetic measurements, RHC/OHC-specific modeling, operando EPR spectroscopy, and DFT calculations. We demonstrate that NH3 inhibition emerges above a Cu- and temperature-dependent critical NH3:NO ratio, where excess NH₃ impairs Cu ion mobility and selectively suppresses OHC kinetics. These findings provide actionable insights into catalyst design and operating conditions for mitigating NH3 inhibition in next-generation SCR systems. Results and discussion SCR kinetics and NH3-inhibition We first demonstrate the practical relevance of NH3 inhibition on Cu-CHA during SCR by examining the performance of commercial degreened (DG) and field-aged (FA) samples. Sample details are provided in Supplementary Table 1. As shown in Supplementary Fig. 1a, both DG and FA samples showed a steady increase in NOₓ conversion at 200 °C with rising ammonia-to-NOx ratios (ANR), up to 1.0 for DG and 0.5 for FA. Beyond these points, conversion declines with further increases in ANR, indicating NH3-induced inhibition above a critical ANR that depends on catalyst state. The manifestation of inhibition at 200 °C, a temperature relevant to real-world exhaust conditions, highlights the operational significance of this phenomenon. Transient light-off and light-down tests (Supplementary Fig. 1b) further corroborate the presence of NH3 inhibition. To investigate its mechanistic origin, we synthesized a series of model Cu-CHA catalysts with Cu loadings of 0.48 wt%, 1.38 wt%, and 2.48 wt%, referred to as Cu-0.5, Cu-1.4, and Cu-2.5, respectively. The elemental compositions are detailed in Supplementary Table 1. All samples were prepared from the same parent zeolite support (Si/Al ratio of ~12). Recent studies have shown that differences in the spatial distribution of [AlO2]– centers in the zeolite framework can influence Cu ion mobility and SCR activity19. Thus, by using a single parent support, we eliminate the confounding effects of both Cu and Al ions to ensure that our observations are primarily influenced by the state of the Cu ions. Figure 1a presents the standard SCR NOₓ conversion as a function of temperature. As expected, NOₓ conversion at a given temperature increases with Cu loading. Cu-0.5 exhibits the well-known “seagull” shape behavior with a mid-temperature activity dip, which reflects the temperature-dependent shift from quasi-homogeneous catalysis (CuI mobility enabled by NH3 solvation) below 250 °C to heterogeneous SCR above 300 °C involving less mobile, framework-bound Cu sites4. The low-temperature SCR OHC involves oxygen activation on a dimeric oxygen-bridged copper species, requiring isolated CuI ions to migrate between cages to reside in proximity with another CuI ion. Negri and coworkers used in-situ X-ray absorption and infrared spectroscopies, along with DFT calculations, to reveal that these dimers were formed by two mobile linear [CuI(NH3)2]+ complexes, consisting of a side-on μ-η2,η2-peroxo diamino copper(II) structure ([CuII2(NH3)4(O2)]2+)12. At low Cu loading (Cu-0.5), the reduced likelihood of Cu–Cu encounters limits the formation of dimeric Cu-oxo species needed for efficient OHC, intensifying this kinetic bottleneck. This explains the lower activity and pronounced seagull dip in these samples. In contrast, Cu-1.4 shows only a subtle dip, while Cu-2.5 displays a monotonic increase, reflecting reduced kinetic constraints with higher Cu content. NH3 inhibition effects on NOx conversion over Cu-CHA catalysts. a NOx conversion during standard SCR as a function of temperature for three Cu-CHA catalysts with different copper loadings: Cu-0.5 (0.48 wt% Cu), Cu-1.4 (1.39 wt% Cu), and Cu-2.5 (2.48 wt% Cu), measured at NH3:NO = 1. b NOx conversion as a function of NH3:NO molar ratio (x-axis) for Cu-1.4 and Cu-0.5 at multiple temperatures: 220 °C, 200 °C, and 160 °C for Cu-1.4; and 220 °C, 200 °C, and 175 °C for Cu-0.5. Feed conditions: 350 ppm NO, 10% O2, 3% H2O and varying NH3 levels to achieve desired NH3-to-NOx ratios, at 150,000/h gas hourly space velocity (GHSV). Figure 1b presents the SCR NOₓ conversion on Cu-1.4 and Cu-0.5 as a function of ANR across temperatures ranging from 160 to 220 °C. Both samples exhibit a clear inhibition regime at elevated ANR. Two systematic trends emerge: (a) for a given catalyst, the critical ANR—defined as the ratio beyond which activity begins to decline − decreases with temperature (e.g., for Cu-1.4: 1.25 at 220 °C, 1.1 at 200 °C, 0.6 at 160 °C); and (b) at a given temperature, the critical ANR decreases with Cu loading (e.g., 1.25 for Cu-1.4 vs. 0.4 for Cu-0.5 at 220 °C). These trends highlight a strong interplay between Cu mobility and NH3 concentration. We selected Cu-0.5 for further kinetic evaluations as it provides a copious amount of data points in the kinetically limited region (<20% NOx conversion). It is essential to note that, while we require data under kinetic limitation for reliable rate parameter estimations (i.e., negligible reactant or product mass transfer limitation), the NH3 inhibition process is not only confined to low NOx conversions; Cu-1.4 exhibits inhibition at conversions as high as 85%, underscoring the practical importance of this effect. Further supporting this notion, the SCR light-off and light-down performance of Cu-2.5, shown in Supplementary Fig. 2 and discussed in Supplementary Note 2, illustrates significant NH3 inhibition below 220 °C during transient operations typical of real-world driving conditions. This sample has a high Cu loading similar to fresh commercial samples, again demonstrating the practical importance of understanding NH3 inhibition in SCR systems. To quantify the effect of NH3 on intrinsic SCR rates, we evaluated low-temperature turnover rates (TOR) for Cu-0.5 as a function of ANR (Fig. 2a). TORs decline monotonically with increasing ANR. Corresponding NOx conversion vs. temperature data in Supplementary Fig. 3a, b reveal two opposing effects of NH3 concentration: while higher ANR increases NOₓ conversion above 350 °C due to faster kinetics and complete NH₃ consumption, conversion drops markedly below 220 °C. Above 350 °C, the reasons behind this effect are clear: accelerated kinetics increase conversion rates for both NOx and NH3, and at ANR < 1.0, NH3 is entirely consumed. In addition to NH3 being the limiting reactant at these higher temperatures, there is also a contribution from non-selective NH3 oxidation; the latter is evident from continued NOx conversion increases when ANR increases from 1.0 to 1.25. Conversely, below 220 °C, a decreased NOx conversion with increasing ANR clearly demonstrates a strong NH3 inhibition effect; this is in-line with slow SCR kinetics following a quasi-homogeneous mechanism requiring mobile Cu ions. The low-temperature activity data in Fig. 2a were further analyzed utilizing a first order kinetic equation, r=FW(−ln1−X), where F is NOx flow rate (mol/s), W is catalyst amount (g) and X is fractional NOx conversion. The Arrhenius equation, k=r[NO]0=Ae−EaRT, was used to calculate apparent activation energy (Ea) and pre-exponential factor (A), where [NO]0 is the concentration of NO in the feed. Supplementary Fig. 3c presents the Arrhenius plot utilized to estimate SCR apparent activation energies and pre-exponential factors presented in Fig. 2b. The apparent activation energies exhibit a slight, gradual decrease with increasing ANR. These variations, however, are not large and suggest a consistent reaction mechanism and rate-determining step across all cases. Gao et al. demonstrated that the apparent SCR activation energy is typically around 80 kJ/mol for high Cu-loading Cu-CHA, decreasing to approximately 35 kJ/mol in low Cu-loading scenarios where the oxidation half-cycle (OHC) becomes rate-limiting20. In our case, the apparent activation energies at all ANRs range between 30 and 40 kJ/mol, suggesting that SCR on Cu-0.5 is already in the OHC-limited regime. The slight decrease in activation energy with rising ANR may indicate further increases in OHC limitations, a notion we will discuss in more detail shortly. In contrast, the pre-exponential factors show a rapid decline with increasing ANR, dropping by two orders of magnitude as this ratio increases from 0.2 to 1.25. This decline indicates a significantly greater entropic barrier to reaction with increased ANR and suggests significant impedance of collisions between active Cu sites and reactant molecules at higher NH3 concentrations. Therefore, the observed NH3 inhibition of the SCR reaction is not due to a significant change in the reaction mechanism, but rather a decrease in the efficient utilization of active Cu sites. Kinetic parameters for standard SCR over Cu-CHA (Cu-0.5). a NO Turnover rates (TOR), and b apparent activation energy and pre-exponential factors for NOx conversion derived from temperature-dependent SCR measurements on the Cu-0.5 catalyst (0.48 wt% Cu). Kinetic parameters were obtained by fitting NOx conversion data under varying NH3:NOx ratios (ANR). Feed conditions: 350 ppm NO, 10% O2, 3% H2O, and balance gas, with NH3 adjusted to achieve the desired ammonia-to-NOx ratio (ANR), at a gas hourly space velocity (GHSV) of ~150,000 h⁻¹. The interplay between the reduction half-cycle (RHC) and the oxidation half-cycle (OHC) determines the overall SCR kinetics. Different feed NH3 concentrations can impact either one or both redox half-cycles. To further resolve the influence of NH3 on the individual redox half-cycles, we adopted a method recently introduced by Krishna and coworkers14,19,32. This approach models steady-state SCR rates as a function of O2 partial pressure using a Langmuir-type rate law to decouple the apparent rate constants for Cu oxidation and reduction. Here, kox and kred represent apparent rate constants for Cu oxidation (CuI → CuII) and reduction (CuII → CuI), respectively, and pO2 in the oxygen partial pressure. At the extremes of this analysis, high pO2 approaches −rNOCutot=kred and low pO2 approaches −rNOCutot=koxpO2. Supplementary Fig. 4 presents the steady state SCR TOR at 200 °C as a function of oxygen partial pressure at various ANRs, alongside their model fits. The extracted rate constants are depicted in Fig. 3. Interestingly, the Cu reduction rate constant (kRed) is largely insensitive to ANR. In contrast, the Cu oxidation rate constant (kOx) is ANR dependent: kOx initially increases as the ANR rises from 0.1 to 0.2, then consistently decreases with further increases in ANR. The critical ANR at 200 °C, observed to be between 0.2 and 0.4 (as seen in Fig. 1b), suggests that the initial rise in kOx with ANR enhances SCR activity. A modest NH₃ concentration enhances OHC kinetics up to a critical ANR (~0.4), beyond which excessive NH₃ suppresses Cu oxidation. These trends reinforce the view that NH3 inhibition primarily stems from impairment of the OHC, while the reduction process remains relatively unaffected. Ideally, this analysis would be conducted at different temperatures to estimate the activation energy of each half-cycle. However, this would require accurate resolution of NO concentrations as low as 1–2 ppm, which is outside of the capability of our MKS FTIR gas analyzer. Therefore, we adopted an alternate approach described below involving transient response experiments with associated kinetic models. Oxidation and reduction rate constants for Cu-CHA (Cu-0.5). Rate constants for Cu oxidation (kox) and reduction (kRed) determined for the Cu-0.5 catalyst (0.48 wt% Cu) using Langmuir-type regression of standard SCR reaction rates as a function of O2 partial pressure (data shown in Supplementary Fig. 4). Error bars represent R2 error from the regression. In addition, the Langmuirian-type expression applied to steady-state analysis is not meant to represent the elementary CuI kinetics. Its purpose is to describe how the observed macroscopic SCR rate depends on O2 pressure under steady-state conditions. Krishna et al. have shown that for samples with low Cu-loading, like the ones used in our study, the apparent O2 reaction order is close to unity14. On the contrary, the elementary Cu(I) kinetics has a second-order dependence on Cu, as shown by O2-pulse experiments described in Supplementary Note 3 and Fig. S5. To quantify the temperature dependence of these rate constants, we conducted transient kinetic experiments inspired by prior work from Deka et al.33 and Nasello et al.34. Supplementary Note 4 (and Supplementary Fig. 6) provides a detailed description of the methodology. Briefly, the RHC was probed by initiating NO flow over a NH3-saturated oxidized Cu-0.5 sample. Since the RHC reaction can be expressed as NO + NH3 + ZCuIIOH → N2 + 2H2O + ZCuI, each mole of NO consumed corresponds to one mole of CuII reduced33,35. The NO transient obtained from these experiments therefore provides kinetic information on CuII reduction. We used a kinetic model assuming the plug flow reactor as 20 CSTRs in series and used it to fit NO transients and estimate RHC rate constants (kRHC). After completing each RHC NO transient, O2 was introduced to perform SCR. The steady state NOx concentration achieved during such an SCR, along with the amount of CuI re-oxidized during RHC → SCR transition, was used to estimate OHC rate constant (kOHC) (further details in Supplementary Note 4). Figure 4a presents an example RHC NO transient and its model fit at ANR = 0.50 and 150 °C, and Supplementary Fig. 7 shows all these transients across varying ANRs and temperatures. Additionally, Supplementary Fig. 8 shows the subsequent SCR NO transients. The calculated kRHC and kOHC, presented as Arrhenius plots in Fig. 4b, c, respectively, reveal a minor decrease in kRHC with increasing ANR, but a significantly pronounced decrease in kOHC, consistent with the trends in Fig. 3. The apparent activation energies and pre-exponential factors for RHC and OHC, derived from these rate constants, are summarized in Table 1. These activation energies align with those reported by Jones et al. for RHC and OHC under stoichiometric NH3/NO conditions on Cu-CHA with a Cu density of less than 0.1 per cage, akin to Cu-0.5 in this study14. The RHC activation energy remains relatively unchanged with increasing ANR, and the pre-exponential factor decreases modestly by a factor of ~4, which is insufficient to explain the ~100× decrease observed during steady-state SCR. Conversely, the OHC pre-exponential factors exhibit ~100× decrease when ANR is increased, clearly indicating that NH3 primarily inhibits SCR by slowing down OHC kinetics. The activation energies for OHC, estimated from these transient experiments, also show a decrease with increasing NH3/NO, a trend similarly observed for SCR activation energies. Transient analysis of reduction and oxidation kinetics over Cu-CHA (Cu-0.5). a Experimental NO transient response (circles) obtained during NO + NH3 titration, along with the corresponding model fit (solid line). b Reduction half-cycle (RHC) rate constants derived from model fits to NO transient data at ammonia-to-NOx ratios (ANR) of 0.5 and 1.25 (transients shown in Supplementary Fig. 7). c Oxidation half-cycle (OHC) rate constants obtained from steady-state SCR measurements and post-SCR RHC transient analysis (data shown in Supplementary Fig. 8) for the Cu-0.5 catalyst (0.48 wt% Cu). Error bars represent R2 error from the regression. Table 1Apparent activation energies (Ea) and pre-exponential factors (A) for the reduction half-cycle (RHC) and oxidation half-cycle (OHC) determined for the Cu-0.5 catalyst (0.48 wt% Cu)Rate parameterReduction half cycleOxidation half cycleNH3/NO0.50NH3/NO1.25NH3/NO0.50NH3/NO1.25Ea (kJ/mol)39354531A2.7 × 1036.5 × 1024.9 × 1034.3 × 101 Apparent activation energies (Ea) and pre-exponential factors (A) for the reduction half-cycle (RHC) and oxidation half-cycle (OHC) determined for the Cu-0.5 catalyst (0.48 wt% Cu) Our findings so far suggest that NH3 inhibition of SCR is primarily caused by impaired OHC kinetics. The following sections elaborate on the molecular-level origins of this effect. As previously noted, the OHC proceeds via two steps: (a) inter-cage diffusion of linear [CuI(NH3)2]+ complexes to pair up with another [CuI(NH3)2]+ complex in the same cage, (b) two [CuI(NH3)2]+ complexes in close proximity activate an O2 molecule to form a [CuII2(NH3)4(O2)]2+ dimeric species. Both CuI and N have atomic radii >0.5 Å, and since Cu-N and N-H bond lengths are approximately 2 Å and 1 Å, respectively, the linear [CuI(NH3)2]+ complexes need to be at an appropriate orientation to be able to pass through the ~3.8 Å openings of cha cages4. Given that such orientation is random, transport of these ions, particularly at low Cu loading, can be slow. In such cases, the intercage transfer of [CuI(NH3)2]+ becomes entropically costly, and hence the dimeric Cu moiety formation rather than O2 activation determines the SCR kinetics5,22,36. Consequently, slow OHC kinetics at high ANR suggest a higher barrier to [CuI(NH3)2]+ diffusion. Evidence from Supplementary Fig. 9, which shows NOx and NH3 conversion data on Cu-0.5 at 220 °C across varying ANRs, reveals that the onset of inhibition coincides with excess gas-phase NH3 (i.e., NH3 conversion <100%). This unreacted NH3 likely saturates the CHA cages and 8-membered rings, physically hindering CuI migration. Additionally, NH3-TPD results (Supplementary Fig. 10 and Supplementary Note. 5) confirm that NH3 adsorption on Cu sites increases substantially at high ANRs. These findings motivated our use of DFT calculations to further elucidate the effect of excess NH3 on Cu ion mobility. Density functional theory To probe the molecular origins of NH3 inhibition, we performed DFT calculations focused on how excess NH3 affects the mobility of Cu(I) species in Cu-CHA. NH3 can interact with Cu ions in two primary configurations: (1) coordination (ligated NH3), which makes a chemical bond with the Cu ions to form [CuI(NH3)x]+, and (2) weak physisorption within the zeolite cage as free NH3. To evaluate the impact on diffusion, we computed energy barriers for [CuI(NH3)x]+ migration across an 8-membered ring (8-MR) window separating adjacent CHA cages, both with and without the presence of excess NH3. Figure 5a presents the initial, transition, and final geometries for the following cases: (I)[CuI(NH3)2]+, [CuI(NH3)2]+, (II)[CuI(NH3)2]+ with 1 free NH3, [CuI(NH3)2]+ with 1 free NH3, (III)[CuI(NH3)3]+, [CuI(NH3)3]+, (IV)[CuI(NH3)3]+ with 1 free NH3. [CuI(NH3)3]+ with 1 free NH3. DFT insights into NH3-assisted CuI(NH3)x (x = 2 or 3) diffusion in Cu-CHA. a Optimized geometries of the initial state, transition state, and final state for CuI(NH3)x diffusion in Cu-CHA at 473 K with varying NH₃ coordination environments: I, two ligated NH3; II, two ligated NH3 with one additional free NH3; III, three ligated NH3; IV, three ligated NH3 with one additional free NH3. b Corresponding diffusion barriers for CuI ions as a function of the number of ligated and free NH₃ molecules. Atomic color scheme: Cu (orange), N (blue), H (white), O (red), Si (yellow), Al (pink). Figure 5b and Supplementary Figs. 11 and 12 present the diffusion barriers in each of these cases. For the base case of [CuI(NH3)2]+, the calculated diffusion barrier is ~45−55 kJ/mol, close to the measured SCR activation energy range shown in Fig. 2, suggesting that [CuI(NH3)2]+ diffusion is the rate-limiting step in this process. When an additional NH3 molecule enters the CHA cage, it may either remain free or weakly bind to the Cu ion. This was validated through NH3-temperature programmed desorption (NH3-TPD) on a reduced Cu-0.5 sample and the corresponding H-CHA support. In these experiments, NO and NH3 were flowed over the catalyst at 180 °C to reduce all CuII sites to CuI while simultaneously adsorbing NH3. The subsequent NH3-desorption phase included a 3-h isothermal step followed by a ramp up to 550 °C under an inert N2 flow. As detailed in Supplementary Note 6 and Supplementary Fig. 13, we estimated the NH3:Cu molar ratio for chemisorbed NH3 on CuI sites to be approximately 2:1 while that for weakly-bound (physisorbed) NH3 was approximately 0.75:1, giving a total NH3:Cu ratio of ~2.75:1. This indicates that under excess NH3 conditions, roughly 25% of the reduced Cu sites are present as [CuI(NH3)2]+ and the remaining 75% as [CuI(NH3)3]+. It is worth noting that the third NH3 ligand on [CuI(NH3)3]+ is weakly bound and only present under conditions with excess NH3, which occurs when the ANR exceeds the critical value. In both scenarios, the DFT-computed diffusion barrier increases relative to the base [CuI(NH3)2]+ case, suggesting that excess NH3 hinders Cu ion mobility. Notably, the diffusion barrier for [CuI(NH3)3]+ is ~45 kJ/mol higher than that of [CuI(NH3)2]+ with 1 free NH3, likely due to the increased bulk of the Cu complex, which reduces its diffusivity. The highest diffusion barrier, approximately 173.7 kJ/mol, occurs when 3 ligated NH3 are present along with a fourth free NH3, rendering [CuI(NH3)3]+ diffusion energetically unfavorable. These findings clearly indicate that excess NH3 significantly impedes CuI ion diffusion across the CHA cages, thereby limiting the formation of reactive Cu dimers essential for OHC and suppressing overall SCR activity. The O2-driven oxidation half-cycle requires two [CuI(NH3)2]+ species to come together and form an O2-bridged dimer5,37. [CuI(NH3)2]+ oxidation probed via O2-pulse experiments presented in Supplementary Fig. 14 shows that excess NH3 leads to a decreased O2 uptake by the CuI sites. Because both the dimer formation and the subsequent O2 activation may be rate-limiting, we next investigate these two elementary steps after the two Cu ions diffuse into the same CHA cage (Fig. 6). For [CuI(NH3)2]+(Case I), the reaction barrier for O2 binding and O-O activation are 8.7 and 17.1 kJ/mol, respectively. When an additional free NH3 is present (Case II), the corresponding barriers are 8.4 and 15.2 kJ/mol, indicating that excess free NH3 has a negligible effect on O2 binding and O-O activation on [CuI(NH3)2]+. In contrast, for [CuI(NH3)3]+ (Case III), the barriers for O2 binding and O-O activation increase to 17.2 and 35.7 kJ/mol, respectively, demonstrating that excess ligated NH3 inhibits both processes. When an additional free NH3 molecule is introduced (Case IV), the barriers further increase to 22.3 and 48.2 kJ/mol, indicating that excess free NH3 suppresses O2 binding and O-O activation on [CuI(NH3)3]+. However, as shown in the preceding analysis (Fig. 5b), the diffusion barriers for [CuI(NH3)2]+ and [CuI(NH3)3]+ across the eight-membered ring are 55 and 123 kJ/mol, respectively, which are significantly higher than the barriers associated with O2 binding and O-O activation, suggesting that CuI ion diffusion is the rate-controlling step of the overall reaction. Effect of NH3 coordination on O2 activation energetics over Cu-CHA. a DFT-optimized geometries along the O2 activation pathway on CuI(NH3)x (x = 2 or 3) sites at 473 K, including the initial state (IS), O2 binding transition state (TS1), intermediate, O–O activation transition state (TS2), and final state (FS), for different NH3 coordination environments: I, two ligated NH3; II, two ligated NH3 with one additional free NH3; III, three ligated NH3; IV, three ligated NH3 with one additional free NH3. b Corresponding free energy barriers for O2 binding (TS1) and O–O activation (TS2) as a function of the number of ligated and free NH3 molecules. Atomic color scheme: Cu (orange), N (blue), H (white), O (red), Si (yellow), Al (pink). In summary, excess NH3 in [CuI(NH3)3]+ modestly increases the barriers for O2 binding and O-O activation, but these barriers remain substantially lower than the corresponding diffusion barriers across the 8-membered ring. This indicates that CuI ion diffusion, which is strongly hindered by excess NH3 through increased ligation and steric bulk, is the dominant rate-controlling step governing Cu dimer formation and, consequently, the overall SCR activity. Operando EPR spectroscopy The state of CuII ions in Cu-0.5, including coordination environment and oxidation state, was investigated via operando electron paramagnetic resonance (EPR) under SCR conditions, across a range of temperatures and ANRs. Figure 7 shows EPR spectra collected at fixed ANR = 1.0 and temperature range of 100–350 °C. In the high field region, three overlapping signals are observed at 3240 G, 3300 G and 3330 G. The 3330 G signal is an instrumental artifact, and hence is not associated to any catalytically relevant process (see details in Supplementary Note 7 and Fig. 15). The 3300 G EPR signal is assigned to CuII-ions with high Cu–N coordination, such as CuII(NH3)4 or CuIIOH(NH3)3, while the 3240 G signal arises from mobility-restricted CuII-ions with mixed Cu–N/Cu–OL coordination. Cu ions are readily solvated by NH3 at low temperatures as evidenced by the high Cu–N coordination signature at 3300 G. As the temperature increases, a gradual decrease in Cu–N coordination and an increase in Cu–OL coordination (OL = lattice oxygen atom) is apparent. This trend reflects the progressive immobilization of CuII ions described by the following reaction: CuII(NH3)n ⇔ CuII(OL)(NH3)n-1 + NH338–40. The contribution from C-N coordination at 3300 G decreases with increasing temperature, persisting through ~220 °C, and the contribution from Cu–OL coordination at 3240 G increases above ~220 °C. This is consistent with the activity trends in Fig. 1. The high field region in the EPR spectra also displays superhyperfine features between 160 and 300 °C, which we will discuss shortly. Temperature-dependent operando EPR of Cu-CHA under standard SCR conditions. Operando EPR spectra collected on the Cu-0.5 catalyst (0.48 wt% Cu) under steady-state standard SCR at an ammonia-to-NOₓ ratio (ANR) of 1 over a temperature range of 100–350 °C (100, 160, 180, 200, 220, 250, 300, and 350 °C). Spectral features correspond to Cu²⁺ species present under reaction conditions. Feed conditions: 500 ppm NO, 500 ppm NH₃, 10% O₂, 3% H₂O, balance gas, at a gas hourly space velocity (GHSV) of 400,000 h⁻¹. Mobility-restricted CuII-ions have anisotropic characteristics that give rise to hyperfine splitting features in x and y directions with varied tensor values based on changes in their coordination environment. These features offer further insights into Cu coordination environments. At 100 °C, the observed features with g// = 2.25 and A// = 170 G are attributable to CuII(NH3)n (n ≤ 5) complexes40. With increasing temperature, the hyperfine features gradually shift to lower fields. Consequently, spectra above 160 °C consist of two overlapping hyperfine signals, with tensor values of g// = 2.26, A// = 155 G and g// = 2.33, A// = 148 G, with the latter becoming more prevalent at higher temperatures. The feature at g// = 2.26, A// = 155 G comes from largely anisotropic CuII-ions that contain mixed Cu–N/Cu–OL coordination20. The tensor values g// = 2.33, A// = 148 G are close to those reported for para-Z2Cu (coordinated to Al-pair with Al-Si-Si-Al linkage) in previous studies, indicating that the associated CuII-ions primarily contain OL ligands39,41–43. These features mainly arise at 350 °C, i.e., above the NH3 solvation region. To examine NH₃ effects more directly, we collected EPR spectra under varying ANRs (0.125–2.0) at two temperatures: 220 °C (Fig. 8a, b) and 180 °C (Fig. 8c, d). At 220 °C (Fig. 8a), the low field region displays two overlapping hyperfine features with spin-Hamiltonian parameters g// ≈ 2.26, A// ≈ 155 G and g// ≈ 2.36, A// ≈ 135 G. These two features were used to fit each EPR spectrum, and the detailed method and results are described in Supplementary Note 8 and Supplementary Figs. 16–18. The fitting results show that the first feature dominates at higher ANR, while the second is more prominent at the lowest ANR. We have already assigned the first hyperfine feature to CuII-ions with mixed Cu–N/Cu–OL coordination. The second feature has similar tensor values to those seen in ex-situ spectra of Cu-0.5 shown in Supplementary Note 9 and Fig. 19, which suggests that CuII-ions at low ANRs (as low as 0.125) have negligible NH3 ligands—at least within the EPR’s detection limit. This explains the lack of NH3 inhibition at this condition since (i) NH3 conversion is close to 100% and (ii) NH3 likely reacts with NO on Cu ions with a short residence time that EPR cannot detect, resulting in a low inventory of unreacted NH3 on the catalyst. Effect of NH3:NOx ratio on Cu speciation from operando EPR. a, b Operando EPR spectra collected on the Cu-0.5 catalyst (0.48 wt% Cu) under steady-state standard SCR at 220 °C with varying ammonia-to-NOx ratios (ANR). c, d Corresponding spectra acquired at 180 °C under identical conditions. Changes in spectral features reflect variations in Cu2+ speciation with NH3:NOx ratio and temperature. 500 ppm NO, varying NH3 to achieve the desired ANR, 10% O2, 3% H2O, balance gas, at a space velocity (SV) of 400,000 h-1. The high-field region of the 220 °C spectra (Fig. 8b) reveals “spikes” indicating superhyperfine features at ANR ≥ 0.5. These features arise from the coupling between the unpaired electron in CuII and the nuclear spin of 14N in NH3 ligands, leading to (2nI + 1) superhyperfine lines (where n is the number of NH3 ligands, I is the nuclear spin, I = 1 for 14N)20,40,44. For ANR ≥ 0.5, we observe at least seven such signals, suggesting that CuII ions exist as Cu(OH)(NH3)3 (i.e., n = 3 NH3 ligands). Although seven superhyperfine lines are visible, it is possible that there are more, with their intensity being too low, especially at the extremes of the high-field signal. For example, Wu et al. observed 9 such features on low Cu-CHA samples with low Cu-loading (similar to Cu-0.5 in this work)41. Thus, it is plausible that some Cu ions could exist as Cu(NH3)4 (i.e., n = 4 with 9 superhyperfine signals) at the highest ANRs. However, our fundamental conclusions outlined below remain unchanged irrespective of whether CuIIions are present as Cu(NH3)4 or Cu(OH)(NH3)3. Moreover, prior studies have established that Cu(NH3)4 complexes are not redox active on their own and need to hydrolyze to form Cu(OH)(NH3)3 complexes before they can participate in subsequent RHC3,4,7. The high-field spectra at ANR = 0.125 do not display any superhyperfine signals, and only three to five discernible features are seen at ANR = 0.25 (indicating 1 or 2 NH3 ligands on Cu). These observations clearly demonstrate that the number of NH3 ligands on CuII-sites is negligible at ANR = 0.125 but increases with higher NH3 levels in the feed. An NH3 coordination = 3 of three on CuII-sites is achieved at ANR = 0.5 and remains constant with further increases in ANR. These trends of Cu-N/Cu-O coordination as a function of ANR as inferred from superhyperfine features agree well with those obtained from hyperfine parameters and fit results shown in Fig. 8a and Supplementary Figs. 16–18. Interestingly, at 220 °C, where all these spectra in Fig. 8a, b were collected, the SCR activity results shown in Fig. 1 indicate that the critical ANR is 0.5—which is close to where EPR spectra reveal maximized NH3 coordination on CuII ions. When the ANR is <0.5, CuII ions exhibit fewer NH3 ligands, reducing the likelihood of forming mobile CuI(NH3)2 complexes. Consequently, fewer CuI ions occupy the same cage, thereby limiting OHC kinetics. This observation aligns well with the data in Figs. 1 and 3, showing that both SCR activity and OHC rate constant increase with rising ANR up to the critical value of 0.5. Beyond this critical ANR, although the NH3 coordination of CuII ions remains maximized, the excess NH3 begins to impede the mobility of CuI ions, as demonstrated by our DFT calculations. As a result, OHC kinetics slow, SCR NOx conversion drops, and the observed NH3 inhibition of SCR takes effect. At 180 °C, the low field region (Fig. 8c) has only one discernible hyperfine feature that appears at all ANRs. This feature is characterized by spin-Hamiltonian values of g// = 2.26 and A// = 155 G, attributed to CuII-ions with mixed Cu–N/Cu–OL coordination. The absence of the feature at g// = 2.36 and A// = 135 G, associated with CuII-ions void of NH3 ligands (e.g., as observed at 220 °C), indicates higher Cu–N coordination at 180 °C. This higher Cu–N coordination is further supported by the high-field region shown in Fig. 8d, where at least 7 superhyperfine signals are visible at ANR as low as 0.25. The SCR activity data in Fig. 1 show that critical ANR decreases with decreasing temperature, which aligns with the EPR data, indicating that maximized NH3 coordination of Cu is achieved at a lower ANR at lower temperatures. The double integration of the EPR signal is directly proportional to the number of EPR-visible Cu ions, making it a valuable tool for quantifying these sites20. Since SCR is not active at 100 °C on Cu-0.5 (e.g., as shown in Supplementary Fig. 3), we can assume that all Cu sites are in the CuII state at this temperature. Once SCR becomes active, a fraction of CuII sites reduces to CuI. With this assumption, we can calculate the fractional composition of CuII as CuII/CuTotal from EPR as a function of reaction temperature and ANR, where CuTotal is the total EPR signal at 100 °C. Supplementary Note 10 and Fig. 20 present CuII/CuTotal values at 180 °C and 220 °C across ANRs ranging from 0.125 to 2.0. At both temperatures, the CuII/CuTotal decreases as the ANR increases. At a given reaction condition, the fractions of CuI and CuII are determined by the relative kinetics of RHC & OHC and, generally speaking, a higher relative OHC kinetics (to RHC) will lead to greater fractional population of CuII ions5,20. Hence, our observation of decreasing CuII/CuTotal with increasing ANR suggests that OHC kinetics are relatively slower at elevated NH3 levels. Supplementary Fig. 20 also presents CuII distribution calculated for a high-Cu sample (Cu-2.5) under SCR conditions at 180 °C. The relevant operando EPR spectra and discussion are provided in Supplementary Note 11 and Supplementary Fig. 21. Interestingly, unlike Cu-0.5, the CuII fraction on Cu-2.5 does not decrease with increasing ANR. Wu et al. recently elucidated the degree of rate-control from NH3-solvated Cu ion transfer and redox kinetics on Cu-CHA catalysts with varying Cu content, and they demonstrated that Cu-transfer is not the rate-limiting factor for high Cu containing samples20,41. The low kinetic relevance of [CuI(NH3)2]+ diffusion on SCR makes Cu-2.5 less susceptible to NH3-inhibition. All these findings agree well with our SCR kinetics, individual RHC and OHC rate parameters, and DFT calculations. At the beginning of Section “SCR kinetics and NH3-inhibition”, we discussed how the field-aged (FA) catalyst exhibits a higher extent of NH3 inhibition than the degreened (DG) sample, with inhibition appearing at a substantially lower ANR. Field aging modifies Cu-CHA catalysts via two primary pathways: partial loss of active Cu through CuOₓ cluster formation and depletion of framework Al via dealumination, both of which restrict Cu ion mobility. The effect of reduced active Cu-sites has been already discussed with the help of Fig. 1. We independently probed the role of framework Al density by synthesizing a 0.5 wt% Cu-CHA catalyst with a higher Si/Al ratio (Si/Al = 36), compared to the Si/Al ≈ 12 samples used throughout this study. Increasing the Si/Al ratio lowers the density of charge-compensating Al sites and increases their average separation, thereby reducing Cu mobility. As shown in Supplementary Fig. 22, the Si/Al = 36 catalyst displays significantly stronger NH3 inhibition, with the critical ANR shifting to markedly lower values relative to the Si/Al ≈ 12 Cu-0.5 sample. Together with the FA-DG comparison, these results demonstrate that NH3 inhibition becomes more severe as Cu mobility decreases, whether mobility is restricted by lower Cu loading, reduced Al density, or combined aging effects. This observation reinforces the central conclusion that NH3 inhibition is governed by mobility-limited oxidation half-cycle kinetics. The insights gained from this study inform several strategies to mitigate NH₃ inhibition of SCR on Cu-CHA catalysts. A key approach is maintaining a high Cu loading to avoid OHC limitations. Although fresh commercial catalysts often start with high Cu content (2–3 wt%), real-world aging decreases active Cu content, increasing NH3 inhibition risk over time. Thus, rational material design should also address other factors that alleviate [CuI(NH3)]+ diffusion limitations, such as optimizing the Si/Al ratio, distribution of Al site density and Al-Al distances. For instance, recent work demonstrated that the stability of Cu pairs varies significantly with Al distribution, differing by up to 0.55 eV, with the most favorable Cu-pair formation occurring when the Al–Al distance is approximately 7.5 Å36. Along with material design considerations, control over operating conditions is also important. A high temperature, when feasible, helps reduce inhibition. Moreover, controlling the NH3:NO feed ratio is crucial; ratios above the critical point impair catalyst performance and lead to unwanted NH3 emissions and increased costs. These findings emphasize the importance of advanced urea dosing strategies, such as segmented NH3 dosing, which spatially distributes NH3 injection across multiple zones in the SCR system to balance reactivity and minimize localized inhibition. When combined with closed-loop feedback control, such dosing strategies can dynamically adjust NH3 feed under transient driving conditions. This is especially important for lean-burn and hybrid powertrains that operate at lower exhaust temperatures, where NH3 inhibition risks are amplified. Altogether, these strategies can help extend catalyst durability, ensure compliance with emissions regulations, and support next-generation low-carbon mobility. In summary, this study reveals the mechanistic origin of NH3 inhibition during standard SCR over Cu-CHA catalysts and identifies practical strategies for its mitigation. We demonstrate that NH3 inhibition arises primarily from suppression of the oxidation half-cycle (OHC), while the reduction half-cycle (RHC) remains largely unaffected. Kinetic modeling of both steady-state and transient experiments shows that excess NH3 significantly reduces the pre-exponential factor for OHC, indicating a strong entropic or diffusional barrier under NH3-rich conditions. DFT calculations confirm that additional ligated or free NH3 increases the diffusion barrier for [CuI(NH3)2]⁺ complexes, hindering the formation of Cu-oxo dimers essential for O2 activation. Operando EPR spectroscopy reveals a direct correlation between NH3 coordination on Cu sites and SCR activity, with excessive NH3 solvation restricting Cu mobility and shifting the redox balance toward CuI. Together, these findings provide a comprehensive molecular picture of NH3 inhibition and highlight actionable pathways to mitigate its effects – namely, by optimizing Cu content, enhancing framework design to promote Cu mobility, controlling operating temperatures, and employing advanced NH3 dosing strategies such as segmented injection and feedback control. These insights are particularly relevant for low-temperature applications and aged catalysts, offering a foundation for robust and durable SCR systems in future lean-burn and hybrid vehicles. Methods Catalyst preparation Three Cu-SSZ-13 catalysts with varying Cu content were prepared using the following steps: (1) hydrothermal synthesis of Na-SSZ-13 zeolite support with Si/Al ratio of 12 and (2) series of ion-exchange processes (Na-SSZ-13 to NH4+-SSZ-13 and NH4+-SSZ-13 to Cu-SSZ-13). To synthesize Na-SSZ-13 with Si/Al ratio of ~12, 0.8 g of NaOH (Sigma-Aldrich) was dissolved in 38 g of deionized water. While stirring the aqueous solution, 17 g of N,N,N-trimethyl-1- adamantyl ammonium hydroxide (TMAda-OH, Sachem Inc., 25 wt% in H2O) was added. Then, under constant stirring, 1.6 g of Al(OH)3 (Aldrich, containing ~54% Al2O3) was slowly added to form a homogeneous mixture. After the mixing, 40 g of colloidal SiO2 (AS-30, Sigma-Aldrich, 30 wt% suspension in H2O) was added slowly into the mixture and left under constant stirring for 2 h. The resulting mixture was transferred to a 125 mL Teflon-lined stainless-steel autoclave with a magnetic stirrer. The hydrothermal synthesis was carried out in a sand bath at 165 °C for 96 h under constant stirring. After the synthesis, the resulting white solid zeolite was recovered via centrifugation, followed by a three-time washing with DI water and subsequent centrifugation. The solid powder obtained this way was dried overnight in a vacuum oven at 70 °C and subsequently calcined under stagnant air at 650 °C for 5 h. The Si/Al ratio of ~12 was confirmed via inductively coupled plasma atomic emission spectroscopy (ICP-AES) performed at Galbraith Laboratories, Knoxville, Tennessee (Supplementary Table 1). The parent Na-SSZ-13 zeolite support was first transformed into NH4+-SSZ-13 via two-time aqueous phase ion exchange with 0.1 M NH4NO3 (Sigma-Aldrich) solution at 80 °C for 2 h under constant stirring. The solid was recovered via three-time centrifugation and washing with DI water. Then NH4+-SSZ-13 was transformed into Cu-SSZ-13 with three different Cu loadings of 0.48 %, 1.39 % and 2.48 % (termed as Cu-0.5, Cu-1.4 and Cu-2.5, respectively) via aqueous ion exchange with Cu(NO3)2·2.5H2O (Sigma-Aldrich) solution at 80 °C for 2 h under constant stirring. The variation in Cu wt% was achieved by the following variations in the ion exchange procedure: one-time exchange with 0.02 M Cu2+ solution for Cu-0.5, two-time exchange with 0.02 M Cu2+ for Cu-1.4, and one-time exchange with 0.1 M Cu2+ for Cu-2.5. The obtained Cu-SSZ-13 solid was recovered with three-time centrifugation and washing with DI water, which was then dried overnight in a vacuum oven at 70 °C and subsequently calcined under stagnant air at 550 °C for 5 h. On top of lab-synthesized Cu-SSZ-13 catalysts, commercial degreened and real-world field-aged (used for 710,000 miles) Cu-SSZ-13 catalysts provided by Cummins Inc. were investigated in this study. Details of all samples, including lab-synthesized and commercial catalysts, are provided in Supplementary Table 1. Catalyst testing All catalytic tests were performed in a fixed-bed reactor system composed of a vertically mounted quartz tube, a K-type thermocouple placed directly above the catalyst bed, a tubular furnace (Applied Test Systems) with a PID controller (Omega CN3251), a set of Brooks 5850E series mass flow controllers, a Perma Pure MH™-Series humidifier (MH-110-12-S-4) for supplying H2O, and a MultiGas™ 2030 FTIR gas analyzer (MKS Instrument). Typically, ~100 mg of catalyst powder (sieved in 40–60 mesh) was loaded into a quartz tube. In standard steady-state SCR with varying ANR, the simulated exhaust feed was composed of 350 ppm NO, varying NH3, 10% O2, and 3% H2O in N2 balance with GHSV of ~150,000/h. NOx conversion was calculated using the following equation: Cu oxidation and reduction kinetic parameters on Cu-0.5 were estimated using two different methods: (1) Langmuirian regression of standard SCR as a function of O2 partial pressure introduced by Krishna and coworkers19,32, and (2) transient response methodology used by Deka et al.33 and Nasello et al.34. For Langmuirian regression method, steady state SCR rate was measured for ANR of 0.1, 0.2, 0.5, 1.0, and 1.25 with varying O2 partial pressure from 0 kPa to 187 kPa at 200 °C. For the transient response method, oxidized Cu-0.5 was first saturated under 175 ppm (for ANR of 0.5) or 437 ppm (for ANR of 1.25) NH3, 10% O2, and 3% H2O in N2 balance at 150, 175, and 200 °C. After the saturation, O2 was removed and subsequently the RHC was performed by introducing 350 ppm NO while constantly flowing NH3 and 3% H2O in N2 balance. This allowed the measurement of RHC NO transient under two different ANRs and three different temperatures. After the RHC, 10% O2 was introduced for creating standard SCR environment. RHC was performed once more at the end of standard SCR. The RHC and SCR transients were fitted with kinetic models as described in Supplementary Note 3. Density functional theory computational setup DFT calculations were performed using the Vienna Ab-initio Simulation Package (VASP) package (version 5.4.4)45,46. The Perdew–Burke–Ernzerhof (PBE) functional47 and projected augmented wave (PAW) potentials were adopted for all the calculations. The convergence criteria of energy and force are 0.03 eV/Å and 10−4 eV for relaxing structures, respectively. A Monkhorst–Pack mesh48 k-point grids of (3 × 3 × 3) and an energy cutoff for the valence plane waves of 400 eV were used in this work. To measure the diffusion barrier of Cu ions with NH3, we built a dual-rhombohedral chabazite unit cell with a = 18.72 Å, b = 9.48 Å, and c = 9.36 Å, including 24 tetrahedral sites (22 Si, 2 Al) and 48 O atoms, where the Si/Al ratio was 11. Initially, two Cu ions with NH3 were placed in two CHA cages separately, with a Cu-Cu distance of ~9 Å. A Cu ion, coordinated with NH3, migrated through the 8-MR window and settled in the same cage as the other Cu ion, with a Cu-Cu distance of approximately 4.5 Å. We captured 9 images throughout the migration process. The image with the highest energy was identified as the transition state, and the energy difference between this transition state and the initial state was used to determine the diffusion barrier. While calculating the energy barriers, all DFT energies were corrected by including zero-point energy (ZPE), internal energy (∆U) and entropy (S) contribution, and the Gibbs free energy (G) was obtained by G = EDFT + ZPE + ∆U – TS. The thermodynamic corrections were performed using the VASPKIT package49. The climbing image nudged elastic band (CI-NEB) and dimer methods were used to search for the transition state of O2 binding and O-O activation, which was further verified by calculating the vibrational frequencies, confirming that a single imaginary frequency was obtained. NH3-temperature programmed desorption (NH3-TPD) NH3-TPD was performed on ~100 mg Cu-0.5 sample in the same fixed-bed reactor system used for the catalytic tests. For NH3-TPD on reduced Cu-SSZ-13 catalyst, the catalyst was saturated with NH3 under a gas flow of 350 ppm NH3 and 350 ppm NO in N2 balance at 180 °C. The system was then purged under constant N2 flow for 3 h. Subsequently, TPD was performed under the same N2 flow with 2 °C/min ramp to 550 °C. For NH3-TPD on oxidized Cu-SSZ-13 catalyst, NH3 saturation was first performed under 350 ppm NH3 and 10% O2 in N2 balance at 180 °C. Then, the system was purged, and TPD was performed as described above. NH3-TPD on H-SSZ-13 with Si/Al ratio of 12 was performed by first saturating H-SSZ-13 under 350 ppm NH3 in balance N2 at 180 °C, then purging the system under constant N2 flow, and subsequently performing desorption with 2 °C/min ramp to 550 °C. Additional details can be found in Supplementary Note 5. Operando electron paramagnetic resonance (EPR) EPR spectra were collected with a Bruker 580 Elexsys equipped with a SHQE resonator. The sample temperature was maintained with a Bruker ER4131VT nitrogen temperature control system. Typically, the microwave frequency was 9.3 GHz at 0.2 mW power. The field was swept 1200 G in 84 s with 5 G field modulation and 82 ms of time constant. For ex situ measurements, the samples were contained in 4 mm OD quartz tubes (Wilmad). The EPR sample holder used for operando experiments was a coaxial plug flow reactor consisting of an inner 3 mm tube that contained the sample (~15 mg) between plugs of quartz wool. An outer 5 mm tube carried the gas in a cross-flow manner to preheat the reactant gases to the desired temperature before contacting the sample. Total gas flow was 100 ml/min, with the mixture controlled by four Brooks computer-controlled flow meters. Feed gases (raised to the desired temperature by an electrically heated coil) enter through the outer tube, and then flow through the aforementioned pinhole and leave the system through the open end of the sample holder (which is connected and sealed with a custom-built adapter). Around 20–30 mg of Cu-0.5 samples was loaded into the inner tube, supported with quartz wool at the top and bottom, and reactant gases flow through this bed. EPR spectra were collected under steady-state SCR at 100–350 °C temperature range and ANR of 0.125–2.0 with 500 ppm NO, 500 ppm NH3, 10% O2, and 3% H2O with a space velocity of ~400,000/h. The EPR spectra were collected throughout the experiment. Changes in ANRs and temperatures were made only when three stable subsequent spectra were obtained at a specific ANR and temperature combination. Supplementary information Supplementary Information Transparent Peer Review file