Modelling of NO Reduction on CeO₂-Supported Pt and Pd Nanoclusters

ABSTRACT

The reduction mechanism of NO to N₂ on CeO₂-supported Pt and Pd nanoclusters (Pt/CeO₂ and Pd/CeO₂) is analyzed using density functional theory. Different NO decomposition and N₂ formation pathways are evaluated, and the thermodynamically preferable paths are identified. The energy barrier of NO decomposition on Pt/CeO₂ indicates that the rate of reaction via metallic cluster sites and the ceria oxygen vacancies are comparable. In contrast, the cluster metallic sites of the Pd/CeO₂ induce a lower energy barrier of NO decomposition relative to the oxygen CeO₂ vacancy sites. The N₂ formation on the Pt/CeO₂ preferentially occurs via the N–N association, whereas the N₂O deoxidation is energetically preferred on the Pd/CeO₂.

KEYWORDS:

• NO reduction
• ceria-based catalysts
• platinum
• palladium
• density functional theory

Introduction

Nitric oxides (NO_x) constitute a family of air pollutants captured by moisture to form acid rain. The NO_x discharge is strictly regulated in many countries because of its toxicity and ecological impact (Desantes et al. 2020; Duncan et al. 2016). Emissions of NO_x from combustion are primarily in the form of NO. Globally, power generation and automobile transportation contribute to most NO emissions (Bradley and Jones 2002). The purpose of the industrial detoxification of NO_x is to generate harmless nitrogen (N₂). Nowadays, two technologies are developed for NO_x emission control, namely selective catalytic reduction (SCR), and three-way catalytic conversion (TWC) (Gómez-García, Pitchon, and Kiennemann 2005). The SCR is widely used in power plants and heavy diesel vehicles (Bradley and Jones 2002), which catalyzes the NO_x reduction by introducing additional reductants (ammonia or urea) (Balland, Parmentier, and Schmitt 2014; Praveena and Martin 2018). The TWC is normally used in automobiles, which catalyzes the reaction of NO_x with unburnt hydrocarbons or carbon monoxide generated from vehicle engines (Koltsakis, Kandylas, and Stamatelos 1998). Ceria (CeO₂) based catalysts are generally used in both SCR (France et al. 2017; Hu et al. 2021; Yao et al. 2017) and TWC (Gu et al. 2019; Jin, Shen, and Zhu 2016; Kim 1982; Liu et al. 2011; Wang et al. 2008), and platinum group metals (PGMs) such as Pt and Pd (Bera et al. 2000; Cordatos and Gorte 1996; Li et al. 2023; Marei et al. 2022) are doped in the CeO₂ to boost the catalytic activity.

PGMs exhibit remarkable catalytic activity for NO_x reduction with hydrocarbons and CO. The reactivity of NO is higher on Pt, followed by Rh and Pd (Burch and Ramli 1998). The NO complete conversion temperature on Pt and Rh is lower than 350°C, with an N₂ yield exceeding 80% (Burch and Ramli 1998). The formation of N₂O and NH₃ as by-products is observed when PGMs catalyze the NO reduction (Adams et al. 2015; Auvray and Olsson 2015; Khosravi et al. 2014; Nováková 2001).

A complete conversion of NO to N₂ in the reaction with CO is observed at 400°C with CeO₂, indicating a remarkable N₂-selectivity (Liu et al. 2011). By doping the CeO₂ with other transient metals like copper (Liu et al. 2011), nickel (Wang et al. 2008), and manganese (Kijlstra et al. 1997; Yao et al. 2017), the reduction temperature of NO_x can be significantly decreased to below 300°C, with no loss of selectivity toward N₂. The reaction temperature of NO with CO can be reduced to 150 ~ 200°C by doping 1%wt of Pt or Pd in CeO₂ (Bera et al. 2000). Because of these properties, catalysts with PGM/CeO₂ have been developed for NO_x emission control (Harrison, Diwell, and Hallett 1988; Thomas et al. 2021; Törncrona et al. 1997; Wang et al. 2017; Zhang et al. 2021). These catalysts usually contain CeO₂ as substrate, and PGMs are doped on the CeO₂ surface as clusters.

This study aims to develop fundamental knowledge using density functional theory (DFT) calculations of the mechanism by which NO reduction occurs on the CeO₂-supported Pt or Pd nanoclusters. The minimum energy profiles for the decomposition of NO on metallic sites of the nanoclusters and the oxygen vacancies at the cluster-ceria boundary are obtained. The N₂ formation mechanism from the N–N association and the N₂O formation-deoxidation is examined. This will illustrate the origin of the high catalytic activity and N₂-selectivity of PGM/CeO₂ catalysts.

Methodology

Electronic energies were obtained with density functional theory (DFT) calculations, using periodic boundary conditions with the Perdew–Burke–Ernzerhof (PBE) functional as implemented in the Vienna Ab-initio Simulation Package (Kresse and Furthmüller 1996; Perdew, Burke, and Ernzerhof 1996). Bulk models of CeO₂, Pt, and Pd were minimized by fully relaxing the cell volume, shape, and atom positions. The convergence criteria were 1 × 10⁻⁵ eV/atom with linear-tetrahedron smearing with Blöchl correction, a k-spacing of 0.18 Å⁻¹, and wave functions truncated with an energy cutoff of 400 eV. The effect of 4f electron correlation was accounted for the Ce by considering on-site Coulomb and exchange interactions using the rotationally invariant formulation of the on-site model (Dudarev et al. 1998), with an effective correction factor ($U_{eff}$) of 5.0 eV on the Ce-f orbital. The van der Waals correction for weakly adsorbed species was simulated with the DFT-D3 zero-damping method (Grimme et al. 2010).

Surface p(4×4) slabs of Pt(111) and Pd(111) were created from the relaxed bulk models. All slabs are periodically repeated with a vacuum region of 15 Å between the surfaces to obviate interactions between slabs. These slabs contain four atomic layers, and the topmost two atomic layers are allowed to relax. Adsorbates were placed on the surface and optimized with a k-spacing of ~0.16 per Å along the x and y axis, together with a 0.2 eV first-order Methfessel-Paxton smearing.

The (111) termination for CeO₂ is the most thermodynamically stable (Branda et al. 2011; Nolan and Parker 2005; Vyas, Grimes, and Gay 1998; Wanklyn and Garrard 1984). Therefore, this termination is used to support nanoclusters. The CeO₂(111) surface contains six atomic layers, and it is terminated on both sides, with oxygen having cerium in the adjacent layer. A 10-atom nanocluster is placed on the CeO₂ surface to simulate the supported Pt or Pd cluster on the CeO₂. The cluster contains two atomic layers, with seven atoms aligned as a hexagon in the bottom and three atoms aligned as a triangle in the top layer. The surface slabs were relaxed with a k-spacing of ~0.48 per Å in the x and y-axis, together with a 0.05 eV Gaussian smearing, and the spin was allowed to relax in all CeO₂ slabs. Three atomic layers in the bottom of the slab are fixed during the relaxation.

The energy of gas-phase species ($E_{gas}$) was obtained in a cubic cell of side length 15 Å with spin polarization corrections, using a Γ-point spacing and the Gaussian smearing with a width of 0.001 eV.

The binding energy ($E_b$) of adsorbates was calculated using Eq. 1, where the $E_{slab}$ and $E_{sys}$ are, namely, the electronic energy of the bare surface slab and adsorbing system.

(1) $E_b=E_{gas}+E_{slab}-E_{sys}$(1)

Transition states were identified by the climbing image – nudge elastic band (CI-NEB) method (Henkelman, Uberuaga, and Jonsson 2000). The minimum energy path between the minimized initial and final structures was obtained using three or six intermediate images with a spring constant of 5 eV/Å. The image closest to the saddle point was allowed to climb up into the saddle point if the largest force on an atom was smaller than 1.0 eV/Å. The climbing process was terminated when the largest force was reduced to less than 0.05 eV/Å.

Results and discussion

The reduction of NO_(g) to N_2(g) on the metal clusters was analyzed by determining the reaction and the activation energies of the elementary steps as indicated in reactions R1 to R5. First, the binding characteristics of all species were evaluated on multiple binding sites, and the most stable configurations were selected. Second, the surface dissociation of NO to atomic nitrogen (N*) and oxygen (O*) and the recombination of N* to N_2(g) were evaluated (R1 ~ R3). Other possible pathways of NO* to form N₂O* and the subsequent formation of N_2(g) were also analyzed (R4 and R5).

NO Adsorption $N{O}_{\left(g\right)}+\ast \rightleftharpoons N{O}^{\ast }$(R1)

NO Decomposition $N{O}^{\ast }+\ast \rightleftharpoons N^{\ast }+O^{\ast }$(R2)

N–N Association $N^{\ast }+N^{\ast }\rightleftharpoons N_{2\left(g\right)}$(R3)

N₂O Formation $N{O}^{\ast }+N^{\ast }\rightleftharpoons N_2{O}^{\ast }$(R4)

N₂O Deoxidation $N_2{O}^{\ast }\rightleftharpoons N_{2\left(g\right)}+O^{\ast }$(R5)

Binding characteristics of nitrogen species

Figure 1 shows the binding energies ($E_b$) of N*, NO*, N₂*, and N₂O* on the most stable sites of Pt/CeO₂, Pd/CeO₂, Pt(111), and Pd(111), and their corresponding molecular configurations. As can be seen, adsorbed product gas-phase species have lower binding energies (N₂*, and N₂O*) than intermediate species (NO* and N*).

Figure 1. (a) binding energies and (b) molecular configurations of N*, NO*, N₂*, and N₂O* on the Pt/CeO₂, Pd/CeO₂, Pt(111), and Pd(111) surfaces. Numbers in (b) are the bond lengths (in Å) of N-O or N-N.

The N* has the largest $E_b$ of all species, which is reasonable given its free atomic radical nature. As seen in Figure 1a, there is a considerable structural deformation of the clusters upon the N binding on the Pd/CeO₂ cluster, leading to lower $E_b\left(N^{\ast }\right)$ compared to the other surfaces. The NO binds N-down on the flat (111) surfaces and the clusters, with only slight deformation of the atomic cluster positions. The $E_b\left(N{O}^{\ast }\right)$ on the clusters are stronger than the flat surfaces. Both N₂* and N₂O* are only weakly bound to the flat surfaces, whereas the $E_b$ of these two species on the clusters is marginally stronger.

NO decomposition

The NO decomposition proceeds through two pathways: On metal surfaces, the O species in the adsorbed NO* molecule is extracted by nearby metallic sites (NO* + * ⇌ N* + O*) (Gonzalez et al. 2018; Mavrikakis et al. 1999). On the CeO₂ surface, oxygen vacancies ([O_v]) can extract the O from the NO molecule to fill the lattice (NO* + [O_v] ⇌ N* + [O]) (Wang et al. 2017). The CeO₂-supported nanoclusters can provide both metallic sites and surface oxygen vacancies, so both pathways are evaluated and compared. Figures 2 and 3 show the energy profiles and transition state structures of these two pathways.

Figure 2. Energy profiles and transition state structures of NO decomposition with metallic sites on supported nanoclusters and flat metal surfaces. Orange numbers in the structure diagrams are the distance between the N and O atoms ($d_{\mathrm{N}-\mathrm{O}}$). Energies are relative to NO_(g) and the clean surface.

Figure 3. Energy profiles and transition state structures of NO decomposition with metallic sites on supported nanoclusters and flat metal surfaces. Orange numbers in the structure diagrams are the distance between the N and O atoms ($d_{\mathrm{N}-\mathrm{O}}$). Energies are relative to NO_(g) and the clean surface with an oxygen vacancy.

In Figure 2, the step from the first to second stage refers to the binding energy of NO*; the energy difference between the third and second stages is the activation energy ($E_a$); the difference from second to fourth stage is the reaction energy referred to the co-adsorption state ($\mathrm{\Delta }{E}_r$); and the energy difference between the fifth stage and the fourth stage is the co-adsorption energy of N* and O* on the same cluster ($E_{co-ad}$).

As shown in Figure 2, the NO decomposition is endothermic on Pt/CeO₂ and Pd/CeO₂. Either $E_a$ or $\mathrm{\Delta }E$ on the Pt/CeO₂ is slightly lower than on the Pd/CeO₂. As indicated in the figure, the $d_{\mathrm{N}-\mathrm{O}}$ at the transition state on the Pt/CeO₂ is marginally shorter than on the Pd/CeO₂. The N atom at the transition state on the Pd/CeO₂ causes a slight deformation to the cluster, which leads to the relatively low stability of this transition state. Also, the negative $E_{co-ad}$ at the final state indicates a stable co-adsorption structure of N* and O* on the clusters, which can benefit the NO decomposition.

The NO* decomposition on the Pt(111) and Pd(111) are evaluated, resulting in the $E_a$ of 203 kJ·mol⁻¹ on Pt and 226 kJ·mol⁻¹ on Pd, which are consistent with literature data (Gonzalez et al. 2018; Mavrikakis et al. 1999). The reaction is slightly endothermic on both (111) surfaces, with a $\mathrm{\Delta }E$ of 10 kJ·mol⁻¹ on Pt and 59 kJ·mol⁻¹ on Pd. Compared to metallic surfaces, the CeO₂-supported clusters show remarkably low $E_a$ for the NO decomposition.

The reaction of NO with the [O_v] is evaluated at the cluster boundary as shown in Figure 3. A boundary lattice oxygen [O] is pre-removed at the initial state of the reaction, and the slab containing NO* and boundary [O_v] sites is re-optimized. The [O_v] only slightly reduces the binding energy of NO* compared to the non-defected surface. During the reaction, the O species moves in to the [O_v] and finally becomes lattice oxygen on the CeO₂ surface, leaving N* at the cluster. The $E_a$ of this reaction on the Pt/CeO₂ is significantly lower than on the Pd/CeO₂, and the reaction is more exothermic on the Pt/CeO₂ as well. As can be seen in Figure 3, the Pd cluster shows a considerable deformation at the transition state, while the Pt cluster is stable during the reaction.

In comparison, the NO decomposition on the Pt/CeO₂ has similar $E_a$ via the NO* + * and NO* + [O_v] pathways, while on the Pd/CeO₂, the NO* + * pathway shows a significantly lower barrier. Both catalysts have better activity than the pure metal surfaces, and the Pt/CeO₂ tends to be more active.

N₂ formation

Two pathways of N₂ formation are investigated, namely the association of two N* (R3), and the deoxidation of N₂O (R4 and R5).

As indicated in Figure 4, during the N–N association, two N atoms are first adsorbed on a supported cluster to form N₂* molecule. The co-adsorption of 2 N* is exothermic on both Pt and Pd clusters. Therefore, the adsorption of multiple N atoms on one cluster is energetically preferable. The $E_a$ of the N–N association is lower on Pt/CeO₂ than on Pd/CeO₂. The transition state structures indicate that the $d_{\mathrm{N}-\mathrm{N}}$ on the Pt/CeO₂ is shorter, which improves its stability. The reaction on both clusters is exothermic, where the Pd/CeO₂ has a larger $\mathrm{\Delta }E$.

Figure 4. Energy profiles and transition state structures of N-N association on supported nanoclusters and flat metal surfaces. Green numbers in the structure diagrams are the distance between the N and N atoms ($d_{\mathrm{N}-\mathrm{N}}$). Energies are relative to N_2(g) and the clean surface.

Compared to the flat (111) surfaces, the supported clusters provide better catalytic activity for the N–N association. The $E_{co-ad}$ of 2 N* is 85 kJ·mol⁻¹ on Pt(111) and 46 kJ·mol⁻¹ on Pd(111). The $E_a$ for the N–N association of 2 N* is 176 kJ·mol⁻¹ on Pt(111) and 179 kJ·mol⁻¹ on Pd(111), which are consistent with literature data (Gonzalez et al. 2018; Mavrikakis et al. 1999). Therefore, on the flat (111) surfaces, the N–N association requires higher $E_a$ than the supported clusters.

The N₂ can also be produced through the N₂O formation-deoxidation pathway (R4 and R5). As compared in Figure 5, the co-adsorption of N* + NO* on the Pt/CeO₂ is marginally endothermic, while on the Pd/CeO₂ is highly exothermic. The $E_a$ from the co-adsorbed state on the Pt/CeO₂ is lower than on the Pd/CeO₂. The transition state structures in Figure 5 shows that the N₂O formation only involves the interaction between the two N atoms, while the N-O bond is not affected. The N–N bond length ($d_{\mathrm{N}-\mathrm{N}}$) at the transition state on the Pt/CeO₂ is close to the N₂O_(g) molecule (1.40 Å), resulting in a similar energy. On the contrary, the $d_{\mathrm{N}-\mathrm{N}}$ at the transition state on the Pd/CeO₂ is relatively longer, so the transition state is less stable and the $E_a$ is also higher on the Pd cluster.

Figure 5. Energy profiles and transition state structures of N₂O formation-deoxidation on supported nanoclusters and flat metal surfaces. Green numbers in the structure diagrams are the distance between N and N atoms ($d_{\mathrm{N}-\mathrm{N}}$), and oranges are between N and O atoms ($d_{\mathrm{N}-\mathrm{O}}$). Energies are relative to N₂O_(g) and the clean surfaces.

The $E_a$ of the N₂O deoxidation on the Pt/CeO₂ is lower than that of Pd/CeO₂. At the transition state shown in Figure 5, the $d_{\mathrm{N}-\mathrm{O}}$ on the Pt/CeO₂ is longer than on the Pd/CeO₂, while the $d_{\mathrm{N}-\mathrm{N}}$ is shorter, resulting in a more stable N₂* molecule on the Pt. The energy barriers of the N₂O deoxidation are generally lower than the N₂O formation in these two clusters, so the N₂O formation is a rate-limiting step for N₂ formation through this path. Furthermore, on both clusters, the $E_a$ are lower than the $E_b\left(N_2{O}^{\ast }\right)$, which leads to the formation of N₂ energetically preferable than the N₂O desorption.

The co-adsorption of N* and NO* on Pt(111) and Pd(111) surfaces is also endothermic, with $E_{co-ad}$ of 43 kJ·mol⁻¹ and 38 kJ·mol⁻¹ on Pt and Pd, respectively. The $E_a$ for N₂O formation on the Pt(111) is 101 kJ·mol⁻¹, which is considerably lower than the Pd(111) (198 kJ·mol⁻¹). For the N₂O deoxidation, the $E_a$ is 137 kJ·mol⁻¹ and 68 kJ·mol⁻¹ on Pt(111) and Pd(111), respectively, and both are higher than the $E_b\left(N_2{O}^{\ast }\right)$ on these surfaces. As a result, the N₂O desorption tends to be more thermodynamically preferable on either Pt or Pd, which leads to the formation of N₂O_(g) as observed in previous experimental works (Novakova and Kubelkova 1997; Tanikawa and Egawa 2011). Therefore, the CeO₂-supported nanocluster can provide a higher N₂-selectivity and avoid the generation of unwanted product N₂O.

Based on the differences in energy barriers and reaction energies, the N₂ formation on the Pt/CeO₂ prefers the N–N association pathway, whereas the N₂O pathway is energetically preferable on the Pd/CeO₂. Both supported clusters can improve the selectivity N₂ during the NO reduction because of the stable adsorption of the by-product N₂O.

Conclusion

The catalytic mechanism of NO reduction to N₂ on the CeO₂-supported Pt and Pd nanoclusters is investigated using DFT. The elementary pathways including NO decomposition, N–N association, N₂O formation, and N₂O deoxidation are assessed.

The supported clusters show remarkable activity for NO decomposition compared to the pure metal (111) surfaces. The activation energy is 90 kJ·mol⁻¹ on Pt/CeO₂ and 131 kJ·mol⁻¹ on Pd/CeO₂. Also, the NO decomposition with the oxygen vacancy at the cluster-ceria boundary is evaluated, which results in a barrier of 102 kJ·mol⁻¹ on the Pt/CeO₂ and 192 kJ·mol⁻¹ on the Pd/CeO₂. Therefore, the metal sites on both clusters are active for NO decomposition, but only the vacancy at Pt/CeO₂ boundary shows considerable activity.

Two pathways of N₂ formation are exanimated, namely the direct association of atomic nitrogen and the by-path through the formation and deoxidation of N₂O. The N–N association barrier is 56 kJ·mol⁻¹ on the Pt/CeO₂, and 186 kJ·mol⁻¹ on the Pd/CeO₂. Both $E_a$ are lower than the pure metal (111) surfaces.

The Pd/CeO₂ shows better catalytic activity for N₂ formation through the N₂O pathway because of its lower barrier for the N₂O formation compared to the Pt/CeO₂. The barriers for the N₂O deoxidation are generally lower than the N₂O formation, so it would not limit the N₂ formation rate through the N₂O pathway. On the other hand, the relatively strong binding energy of N₂O on the supported Pt and Pd clusters can avoid the desorption of N₂O, which is one of the major advantages of such catalysts compared to the pure Pt and Pd.

Acknowledgements

The research reported in this work was funded by the Office of Sponsored Research at King Abdullah University of Science and Technology (KAUST) under the Competitive Research Grant OSR-CRG2018-3042.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

The work was supported by the King Abdullah University of Science and Technology [OSR-CRG2018-3042].