## Introduction

The selective catalytic reduction (SCR) of environmentally harmful nitric oxide (NO) with ammonia (NH3) is a well-known and established technology for the denitrification of exhaust gases from stationary (power plant) and mobile (e.g., lean-burn engines) sources1,2,3. However, the more stringent global legislations and the relatively low exhaust temperatures of more efficient engines and low-load engine operations require the search for more efficient catalytic systems. For example, In Euro 6 stage, the European Union legislative authorities have tightened the limits of nitrogen oxides being emitted from diesel cars (from 180 mg NOx/km in Euro 5 to 80 mg NOx/km in Euro 6)4. A wide variety of catalytic systems based on metal-containing zeolites and mixed metal oxides have been investigated in this reaction. The introduction of Cu-exchanged small pore molecular sieves such as Cu-SSZ-13 and Cu-SAPO-34 have been a revolutionary technology for SCR applications5 and have an optimal performance between 200–450 °C6,7,8. Among mixed metal oxides, V2O5-WO3-/TiO2 catalysts give >90% NO conversion at gas hourly space velocities (GHSV) of 60000–90000 h−1 between 250–400 °C9,10,11,12,13. However, all these systems fall short of providing sufficient performance at temperatures below 200 °C. Catalysts operating at lower temperatures are imperative in mobile applications due to engine cold start14 and new advances in low-temperature combustion15. In this respect, manganese-containing mixed metal oxides exhibit excellent catalytic activity in the NH3-SCR reaction operating at temperatures below 200 °C, and therefore is of particular interest as a potential low-temperature component in NH3-SCR16,17,18,19,20,21,22,23.

Typically, Mn-based catalysts are prepared by impregnation or homogeneous precipitation methods with other metal oxides, such as Ti and Ce oxides, that act as support, dopants, or promoters. During the last decades, the role of the different components on the catalytic activity and selectivity have been debated extensively3. Mn catalytic activity originates from its excellent redox ability at low temperatures. The importance of specific surface area, dispersion, and oxidation state of the different Mn oxides have been highlighted24,25,26. TiO2 is considered a metal oxide support providing optimal dispersion of Mn active species, surface area, thermal stability, and Lewis acid sites to adsorb NH327,28. For Ce and other transition metal, there is no clear consensus on their role on the catalytic reaction. The promotional effect is often explained by an improvement of the catalytic redox cycles by intimate contact of the active Mn oxides and the promotors29,30,31,32. Among the transition metals, Ce is widely used and probably one of the most promising promotors3. In binary MnCe systems, the addition of Ce was reported to improve the conversion levels compared to individual Mn oxides33,34. This promotional effect is generally explained by an enhancement of the redox functionality, which is proven by the easier reduction of Ce and/or Mn during temperature-programmed reduction experiments35. Baiker et al. also postulated that binary MnCe oxides have a higher adsorption of NO and NH3, which promotes catalytic activity36. In ternary MnCeTi oxides, the improvement of activity by Ce is also frequently explained by an increase of the Mn redox properties35,37,38,39. In contrast, other studies suggest that the MnCe electronic interaction decrease the activity of Mn oxide species for NO conversion40 by a reduction of the Mn4+/Mn3+ ratio. On the basis of the measured surface areas, binary MnCe33,34,36 and tertiary MnCeTi35,37,40,41,42 systems show better textural properties when Ce is added, but this is rarely discussed as a main promoting effect.

To resolve this unsettled dilemma, we studied the structure and catalytic performance of Mn, Ce, and Ti mixed-oxide catalysts with a wide range of metal-oxide compositions. The catalysts were prepared by a homogeneous precipitation method designed to obtain mixed oxides with amorphous structure and a homogeneous dispersion of the distinct metal oxides. More specifically, 30 catalysts were systematically synthesized with different Mn, Ce, and Ti compositions. To verify the formation of an amorphous and homogeneous mixed oxide, we have chosen a multi-technique approach combining XRD, electron microscopy and high-resolution X-ray photoelectron spectroscopy. The low-temperature NH3-SCR performance of all the catalysts was investigated under relevant conditions encountered in a car exhaust (see experimental details).

## Results

### Catalyst synthesis, structure, and metal-oxide spatial distribution

After synthesis of the catalysts, we aim to characterize their structure and chemical properties by a multi-pronged approach. The crystalline structure of the catalysts was studied by powder X-ray diffraction, and crystallite sizes were calculated from diffraction peaks integral breadth using the Scherrer equation. The results are summarized in the ternary diagram plotted in Fig. 1 and the detailed diffraction patterns and crystallite size analysis are in the Supplementary information (Supplementary Fig. 2, Supplementary Fig. 3, and Supplementary Table S2).

The binary catalysts have a certain degree of crystallinity and show reflection lines of fluorite CeO2 structure in the CeTi and MnCe oxides and rutile TiO2 in the MnTi oxides. MnOx crystalline phases, α-Mn2O3 (Bixbyite), Mn5O8, Mn3O4 (Hausmannite), and MnO(OH) (Groutite) are observed on the MnCe oxides, whereas binary MnTi samples display weak and broad reflections from MnOx phases, indicating nano-crystallites below 3–4 nm, most likely on the TiO2 surface. The existence of mono-component metal oxide crystalline phases is an indisputable proof that the preparation method is not successful for the synthesis of well-mixed oxides in binary systems. In sheer contrast to the binary systems, the diffraction patterns of most of the ternary systems show featureless diffraction patterns indicating the amorphous nature of the samples (Supplementary Fig. 3). Only a few samples with high Ti (Mn0.08Ce0.13Ti0.79) or Ce (Mn0.07Ce0.55Ti0.37 and Mn0.11Ce0.48Ti0.41) content display reflections from anatase TiO2 or fluorite CeO2, respectively (see Supplementary Fig. 4). This suggest that the third metal component facilitates the formation of a ternary amorphous phase, which is the first step on obtaining a homogeneous mixed oxide. Very importantly, the effect of Ce is crucial in the formation of an amorphous phase, as small amounts of such a component inhibits the formation of crystalline phases, as can be observed in the low edge of the ternary diagram in Fig. 1.

The bulk and surface composition of the catalysts were compared by using Inductive Coupled Plasma (ICP) and high-resolution X-ray Photoelectron Spectroscopy (XPS), respectively (Supplementary Table S1). The bulk chemical composition of the catalysts is comparable with the theoretical composition, with a slight deviation to lower Mn concentrations, which suggests that not all the Mn is precipitated during synthesis. One plausible explanation is the formation of soluble Mn complexes with ammonium ligands, which are less prone to hydrolysis-condensation reactions43. Those are expected to remain in the supernatant during the washing by centrifugation, thus decreasing the final Mn content on the catalyst. The bulk composition was compared to the surface composition of selected samples by XPS. For the MnTi binary, there is an enrichment of Mn on the surface, in line with the hypothesis that Mn is supported on TiO2. For the ternary systems, the results show that surface compositions are similar to bulk compositions, with some of the samples displaying a modest enrichment of Mn and Ce on the surface.

To further investigate the structure, metal oxide distribution, and local composition of the catalysts, we performed a high-resolution electron microscopy study. A representative transmission electron microscopy (TEM) image of a binary MnTi Sample (Mn0.37Ce0.00Ti0.63) is shown in Supplementary Fig. 5. We observed nanosized and crystalline TiO2 particles decorated with layers of amorphous material. High-angle annular dark-field (HAADF) STEM imaging and elemental mappings computed from energy-dispersive X-ray spectroscopy (EDX) data, presented in Supplementary Fig. 6, reveal that Mn is well dispersed on the TiO2 particles. In contrast, TEM images of a representative MnCeTi ternary system, in Fig. 2a, b, corroborate the existence of a purely amorphous system. The structural features resemble an agglomeration of shapeless nanoparticles forming a random porous structure. Annular dark-field (ADF) STEM imaging and elemental mapping computed from EELS data confirm the homogeneous distribution of Mn, Ce, and Ti metals over individual catalyst particles.

All the results strongly confirm that our synthesis route renders a homogeneous distribution of all the metal atoms in the ternary system and indicate that our synthesis method is effective for the preparation of ideal perfectly-mixed ternary metal oxides. From the characterization measurements conducted, we illustrate the structure of the MnTi binary and the MnCeTi ternary catalysts in Fig. 3.

### Impact of metal oxide composition on catalyst textural properties, oxide reducibility, and speciation

The specific surface areas of the discrete catalysts are presented in the ternary diagram in Fig. 4 and in Supplementary Fig. 7 and Supplementary Table S3.

The catalyst materials have a clear type IV isotherm, with an H2 or H3 loop characteristic of mesoporous coming from the agglomeration of the metal-oxide nanoparticles44. Clearly, Ce content has a significant impact on specific surface areas of the ternary catalysts. The addition of Ce plays a key role as structural promotor increasing the surface area from 108 to 245 m2/g when Ce content increases from 0 to 20 mol%. A further increase in Ce has a negative impact and surface areas drop down to 62 m2/g when Ce concentration reaches 55 mol%, similar to previous studies on CeTi45, MnCe46 binary systems and MnCeTi ternary systems47. Comparing the XRD and BET data, the increase of surface area is strongly correlated with the formation of amorphous structures.

To investigate the redox properties of the materials, temperature-programmed reduction (TPR) experiments with H2 were performed. Overall, the reduction peak at 200–450 °C is attributed to the reduction of Mn4+ and Mn3+ species to Mn2+ 48,49, whereas the main reduction peak at 550–650 °C is ascribed to the reduction of Ce4+ to Ce3+ in a mixed oxide phase50,51,52. The difference in the TPR of pure CeO2, plotted on supplementary Fig. 8, confirms that Ce is well-mixed and strongly interacting with Mn and Ti. The samples containing a low Mn/Ce molar ratio, plotted in Fig. 5a, show a monotonic decrease in the reduction temperature of Ce when Mn content increases. According to literature reports, the main parameters influencing the reduction temperature are specific surface area and the presence of other metals in intimate contact53. In this case, the surface area of the binary Ce-Ti system is higher than the Mn-contained samples shown in Fig. 5a. Therefore, changes in the reduction temperature are unrelated to specific surface area and can be rationalized as a consequence of the close proximity of Mn and Ce species, which improves the reducibility and thus the redox properties of Ce. The TPR of samples containing high amount of Mn are shown in Fig. 5b. For the binary MnTi samples, two main peaks from Mn reduction (289–306 °C and 341–408 °C) are observed. Addition of Ce in the catalyst formulation has a strong impact on the TPR profiles: the low-temperature peak strongly decreases with small amounts of Ce and the high-temperature contribution shifts to higher temperatures with increasing Ce content (see also Supplementary Fig. 9). We surmise that the changes in the TPR profiles probably result from the lower reducibility of the Mn species in close contact with Ce.

To gain more nuanced insight into the effect of Ce to the catalyst properties, a detailed high-resolution XPS study was performed. Oxidation states of Mn were rigorously fitted from a set of Gaussian-Lorentzian components per oxidation state, due to the multiplet splitting between the unpaired electrons in Mn 2+, 3+, and 4+. The set of components of the discrete oxidation states were obtained from measurements of pure MnO, Mn2O3, and MnO2 oxides and the results were compared with previously reported measurements54. More experimental details can be found in the supplementary information and Supplementary Fig. 10. The Mn (2p3/2) spectra from selected samples show a high amount of Mn3+ species in the binary MnTi catalysts, whereas a combination of Mn4+ and Mn2+ dominates the spectra of ternary samples (see Supplementary Table S5). There is no clear consensus in the literature on the effect of Ce on the Mn oxidation state. Whilst several authors found an increase in Mn3+ species with the addition of Ce35,39,40, Feng et al. observed a slight reduction in Mn3+ 42, and others found no clear correlation55,56. The origin of the discrepancy could be related to the complexity on the analysis of the Mn (2p3/2) spectra and the different structure of the prepared catalysts. In order to improve the confidence in our results, average Mn oxidation states were calculated from the XPS and TPR and gave comparable results (see Supplementary Table 7), which validates our XPS deconvolution method and confirms that the addition of Ce decreases the average Mn oxidation state. Raman spectra of selected binary MnTi and ternary MnCeTi were also collected and qualitative analysis is in line with the XPS and H2-TPR results (see supplementary information, Supplementary Fig. 11 and Supplementary Table S6). Based on our structural characterization results, we speculate that the reduction in the average Mn oxidation state is due to the existence of Ce species in close proximity to Mn. This might explain the shifting of TPR profile to high temperature after introducing Ce into the MnTi system.

Additional information about the chemistry and structure of Mn on selected samples was obtained by X-ray absorption spectroscopy (XAS). Mn K-edge spectra were recorded ex situ for selected binary (Mn0.35Ce0.00Ti0.65) and ternary (Mn0.37Ce0.04Ti0.60 and Mn0.30Ce0.19Ti0.51) samples to investigate the effect of Ce in the Mn oxidation state and local structure. All XANES spectra (Fig. 5c) display a weak peak in the pre-edge region (inset) due to the 1 s → 3d quadrupolar transitions and a white line characteristic of an oxidized state of Mn sitting mostly in octahedral environments57,58. Depending of the Ce concentration, the position of the 1 s → 3d transitions are located at different energies indicating different manganese oxidation states. Without Ce, Mn is found in its most oxidized state (1 s → 3d: 6542.0 eV), while adding Ce reduces the average Mn oxidation state. The same findings are observed regarding the absorption edge energies at 6551.6, 6549.2, 6550.0 eV respectively for Mn0.35Ce0.00Ti0.65, Mn0.37Ce0.04Ti0.60 and Mn0.30Ce0.19Ti0.51 compositions. An estimate of the average Mn oxidation state was performed by comparison with manganese oxide standards. More details about the analysis can be found in the supplementary information. Summarizing, the average oxidation states (in Supplementary Table S7) are in agreement with the XPS and H2-TPR data.

Subsequent qualitative assessment of the EXAFS spectra in Fig. 5d, e shows a main peak at about 1.42 Å attributed to Mn-O scattering paths and several peaks between 2.0 and 5 Å assigned mainly to Mn-Mn paths with some minor contributions from Mn-O scattering and some multiple scattering processes. The peak area corresponding to the Mn-Mn scattering paths are larger for the binary Mn0.35Ce0.00Ti0.65 indicating that the later materials has the best long-range order among the three composition (while still appearing amorphous by XRD). The ternary samples display the strongest disorder, supporting than such a system is a homogeneous mixture of mixed oxides.

### Role of Mn and Ce on NO reduction at low temperature

Next, we inspected the activity of the different samples in the selective catalytic reduction of NO with NH3 at 150 °C. During the activity measurements, there was no NO2 formation, and the only products were N2 and N2O. Activity plot with the amount of Mn shows a close-to-zero intercept of the ordinate, indicating that Mn oxides are the most important species in catalytic performance (Supplementary Fig. 13). Up to 60% of Mn, there is a modest correlation of the increase of activity with the increase in Mn content. However, the results revealed that no evident trend is observed when Mn is above 60% and strongly suggest that Mn content is not the only factor determining catalytic activity.

To unravel the effect of the distinct metal oxides, we re-examined the catalytic performance by surface area normalization of the activity. Specific activities (ml of NO converted per m2 and per min) were plotted as a function of Mn content in Fig. 6. For the individual Mn and the binary MnTi samples, a clear linear dependency of the measured specific activities with the Mn content is observed. The MnCeTi ternary systems has also a pseudolinear correlation with the Mn content but with a lower slope. The Mn surface enrichment for the MnTi catalysts observed by XPS (Supplementary Table S1) can insufficiently explain the almost twofold increase in specific activity. Therefore, we infer that MnOx species in the individual Mn and binary MnTi catalysts have similar activity, and those are more active than in the MnCeTi ternary system. The results are also in line with the higher reduction temperature of manganese species observed in the TPR data of MnCeTi ternary systems.

To assess the implication of the different MnOx species on catalytic performance, the normalized activities with the distinct Mn surface species (Mn4+, Mn3+, Mn2+, and total Mn) were constructed in Supplementary Fig. 14. The plots show an increase in activity with the amount of Mn4+ and Mn3+ species, however there are no direct evidences that the Mn2+ species are promoting catalyst activity, in line with previous reports59. Hence, the increase in the Mn2+ content in the samples containing Ce could, to a certain extent, explain the lower specific activity of those samples. In addition, the substantial effect of Ce on the TPR profiles, shifting reduction peaks of Mn4+ and Mn3+ to higher temperatures, indicates that other interactions of Mn4+ and Mn3+ with Ce species can be responsible for the lower specific activity.

We further investigated the nature of the oxygen species in the solid catalyst, which has been suggested to have a significant impact on SCR activity, more specifically on the oxidation reactions60,61. The reactivity of oxygen species on the catalyst was investigated by monitoring the formation of N2O during NH3-TPD experiments. In these experiments, N2O is formed in the range of 140–40 °C via the oxidation of ammonia with active oxygen species on the catalyst surface. The plots in supplementary Fig. 15. shows that NO activities at 150 °C are correlated to the amount of N2O evolved during the NH3-TPD experiments, which points to the direct role of the active oxygen in the overall reaction mechanism at low temperature. Since the NO activity is related to Mn content, we can infer that those active oxygen species are related to Mn species.

The effect of the adsorption of NO and NO + O2 on catalytic performance was also investigated by temperature-programmed techniques in selected samples with similar Mn content (Supplementary Fig. 16). Quantification during NO experiments shows a higher adsorption of NO for the binary MnTi sample. In the case of the NO + O2 adsorption measurements, the sample with a small amount of Ce has a larger amount of NO adsorbed. The lack of a clear trend evidences that the capacity of NO and NO + O2 adsorption is unable to fully describe the catalytic performance.

The effect of Ce on the Mn activity has been previously reported in the literature with inconsistent findings. For example, Liu et al. proposed a positive effect of Ce and Ti in activity by a dual redox cycle consisting of Mn4+ + Ce3+ ↔ Mn3+ + Ce4+ and Mn4+ + Ti3+ ↔ Mn3+ + Ti4+ cycles35. A different explanation of the beneficial effect of Ce was proposed by Yang et al., based on a mechanistic study using in-situ FTIR spectroscopy. The authors suggested that MnOx species show a faster rate for the conversion of NO to nitrate or nitrite, whereas CeO2 mainly provides adsorption sites resulting in nitrites species33. In line with our results, Wu et al. also observed a negative effect of Ce in the activity in ternary MnCeTi compared to binary MnTi. However, these authors explained the negative effect of Ce by a reduction in the Mn4+/Mn3+ ratio40. In order to resolve the origin of these contradicting findings, our results were contrasted with previously reported ones after surface-area normalization of the activity data. These data are shown in Supplementary Table S11. The analysis shows that when the activities are surface-area normalized, there is no positive effect of the addition of Ce on the specific Mn activity, which validates our experimental results. Although this analysis is incomplete due to our lack of knowledge on several parameters, such as Mn surface composition, oxidation states, degree of interaction between the different oxide species, etc., there is no apparent trend pointing to an increase in activity by Ce addition. Therefore, we propose that Ce is not improving the intrinsic catalytic properties of Mn species at low reaction temperatures and that the only promotional effect of Ce is purely structural due to an increase in catalyst surface area.

When looking at the catalytic performance at higher temperatures, the role of Ce is clear. Supplementary Fig. 17 shows NOx conversion of selected samples with increasing the amount of Ce. The activity data of all catalyst samples can be found in Supplementary Table S8. The binary MnTi catalysts has the highest activity at low temperatures, but the catalytic performance drastically drops at temperatures >250 °C due to the unselective oxidation of NH3 to NOx. The addition of Ce drops the conversion at low temperatures, but promotes NOx conversion at temperatures >250 °C, widening the operational temperature window of the catalyst materials. Understanding this effect lies beyond the scope of our investigations as other parameters, such as close proximity of the redox and acidic functions, may govern the reaction at high temperatures62.

### Role of Mn and Ce on N2O selectivity at low temperature

Finally, the N2O formation at low temperature was investigated on all the samples. Possible effects of conversion on N2O selectivity were ruled out by inspecting the N2O selectivity plot as function of conversion in Supplementary Fig. 18. In general, the N2O formed over MnOx and binary MnTi catalysts is significantly higher than in the MnCeTi ternary systems. For a deeper analysis of the selectivity results, the surface-specific formation of N2O of selected samples was plotted in Fig. 7 as function of the Mn surface density calculated from the XPS measurements. The results clearly show an exponential increase of the N2O formation at 150 °C with the Mn surface density, indicating that N2O formation obeys a higher-than-one-order dependence in Mn. Our characterization and NOx activity measurements pointed to an interaction of Ce with Mn, leading to a decrease in the Mn activity. This can definitely play a role in the lower N2O activity but does not fully describe the Mn order dependence. A plausible explanation of this behavior is that more than one Mn active species in close proximity must participate in the kinetic formation of N2O. According to literature, two main reaction mechanisms explain the formation of N2O at low temperature63: one base on a Langmuir-Hinshelwood mechanism where NO is oxidized to NO3 species that react with NH4+ to give N2O, and the other is an Eley-Rideal mechanism resulting from the oxidative dehydrogenation of NH3 to NH species that react with NO to yield N2O. A mechanistic study on the formation of N2O is beyond the goal of this work, but both proposed mechanisms are based on multistep oxidation of NO and/or NH3. Therefore, we postulate that the formation of N2O requires the participation of at least two Mn active sites and that is promoted by neighboring Mn active sites. The homogeneous composition of the ternary MnCeTi dilutes the Mn species on the catalyst surface and inhibits subsequent oxidation steps by breaking up the MnOx ensembles. This mechanism is schematically depicted in Fig. 8. Besides the dilution effect, the addition of Ce could also decrease the activity of surface oxygen observed by a suppressed N2O formation in the NH3-TPD.

To rule out the effect of the amount of ammonia adsorbed on the catalysts on N2O selectivity, NH3-TPD measurements were performed and the total number of acid sites, obtained from the total amount of ammonia adsorbed, was plotted as a function of Ce content (Supplementary Fig. 19). The total acidity of all the samples was also plotted in the ternary diagram in Supplementary Fig. 20. The total acidity is around 1.1 μmol/m2 up to 20% Ce and then monotonically increases up to around 2.5 μmol/m2. Owing to the fact that the inhibiting of N2O formation is observed in catalysts with Ce content below 20%, we can exclude the role of the number of acid sites on the N2O formation.

Concluding, our research results add new insight into our understanding of Mn catalysts for low-temperature NH3-SCR applications and provides a direction toward settling the ongoing debate over the effect of Ce on the Mn active species. We postulate that the activity of the Mn active sites is not positively affected by Ce species in intimate contact. In fact, our results suggest that Ce is decreasing the average oxidation state and activity of Mn active species and is just a structural promotor, increasing catalyst surface area. On the other hand, addition of Ce is increasing N2 selectivity as it suppresses second step oxidation and thus N2O formation by a dilution effect on the MnOx active sites. The latter still makes Ce an attractive additive for MnTi systems.

## Methods

### Catalyst preparation

Titanium(IV) sulfate solution (Ti(SO4)2, Pfaltz & Bauer., 30% in H2SO4), cerium(III) nitrate hexahydrate (Ce(NO3)3·6H2O, Sigma-Aldrich, 99.999% trace metals basis), manganese(II) nitrate hydrate (Mn(NO3)2·xH2O, Sigma-Aldrich, 99.999% trace metals basis), ammonium hydroxide (NH4OH, Alfa Aesar, ACS grade, 28.0–30.0%) were used as received, without further purification.

A series of individual, binary and ternary materials with different molar concentrations were prepared by a controlled co-precipitation method as described in the Supplementary information and Supplementary Fig. 1. Our method is a novel and highly efficient approach, where the aim is to precipitate all metals at the same pH level to obtain a homogeneously well-mixed metal oxide system. This is done by dual dosing of NH3 and salt solution at a constant predetermined volumetric ratio. In most literature, the salt solution is added dropwise to an NH3 solution, but this gives a pH change over time (from high to final low pH) and could lead to a suboptimal co-precipitation of the elements. First manganese nitrate hydrate (Mn(NO3)2·xH2O), and cerium nitrate (Ce(NO3)3.6H2O) were dissolved in deionized water and stirred for 10 min. Then, a 30% titanium sulfate solution [Ti(SO4)2 in H2SO4] was added to the salt solution. These solutions were mixed under magnetic stirring at a constant speed (400 rpm) for 30 min, leading to a perfectly mixed metal salt solution. The mixed metal salt solution (loaded in a syringe pump) was injected simultaneously along with 14.7 M solution of ammonium hydroxide with a Gilson pump (NH4OH, Sigma-Aldrich, 97%) to a recipient containing 20 ml of mother solution that is already at the target pH of 10.5. During this simultaneous injection of the metal precursor and base, the resulting suspension was continuously stirred. This procedure allows to operate at a constant pH of around 10.5 by adding the same amount of hydroxide consumed during the catalyst precipitation reaction. The metal oxides will precipitate at the same time, rendering a very high level of homogeneity. Then, the precipitated solution was stirred for 30 min at 400 rpm. The sample was centrifuged at 7500 x g and washed several times with milli-Q water until the conductivity of the supernatant reached to 50 μs.cm−1. Then the samples were dried overnight at 100 °C, followed by calcination at 500 °C for 6 h. The list of samples and compositions is shown in Supplementary Table S1.

### Catalyst characterization

$${NO}\,{conversion}\,(\%)=\frac{{[{NO}]}_{{in}}-[{{NO}}]_{{out}}}{[{{NO}}]_{{out}}}\times 100$$
$${N}_{2}O\,{selectivity}\,(\%)=\frac{2[{{N}_{2}O}]_{{out}}}{[{N{O}_{x}}]_{{in}}+[{N{H}_{3}}]_{{in}}-[{N{O}_{x}}]_{{out}}-[{N{H}_{3}}]_{{out}}}\times 100$$