B-site modulation tailors the size distribution of ex-solved nanoparticles for optimized active sites in perovskites

Abstract

Supported nanoparticles fabricated by an ex-solution process from perovskite scaffolds serve promising capabilities for various thermal and electrochemical reactions. Although the size distribution of the nanoparticles plays a crucial role in catalysis, it remains difficult to control in ex-solution due to the lack of understanding of their underlying energetic factors. Here, we investigate the impact of B-site doping in two different types of perovskite hosts (A²⁺B⁴⁺O₃ and A³⁺B³⁺O₃) on the size distribution of ex-solved nanoparticles. By incorporating non-reducible ions at the B-site, we observe the variations in nanoparticles’ sizes and densities using model cleaved surfaces of ceramic pellets. Density functional theory calculations help correlate the observed experimental results with the computed energetic descriptors, providing deeper insights into the particle ex-solution and enabling the description of particle density via a first-order nucleation model. Furthermore, we utilize the tailored ex-solution materials as catalysts for N₂O decomposition, and it exhibits a volcano-plot relationship between the catalytic activity and size distribution of ex-solution particles. It suggests that our findings can provide an alternative way to optimizing catalytic active sites for various environmental and energy applications.

The size distribution of nanoparticles strongly affects catalysis but controlling it during the ex-solution reaction remains challenging due to limited understanding of the underlying energetics. Here, the authors show how B-site doping in two perovskite hosts controls nanoparticle size distribution, offering a new strategy to optimize catalytic active sites for energy and environmental applications.

Introduction

Metal nanoparticles (NPs) distributed onto the surface of functional oxides (i.e., supported catalysts) have garnered significant interest as key materials for a wide range of chemical and energy conversion applications^1–6. Conventionally, such systems can be obtained by the external dispersion of metal salts to ceramic backbones (e.g., oxides, carbides, nitrides, etc.), followed by subsequent calcination and reduction processes. However, the supported catalysts yielded by the traditional deposition approach present weak interactions between NPs and the support frameworks. Thus, it usually faces the limited resistance to particle aggregation and carbon coking during material preparation and/or (electro)chemical reactions, causing catalytic performance degradation^7,8.

To overcome these issues, an innovative process, ex-solution phenomenon has been proposed as an alternative method for fabricating the supported catalyst over the past few decades^9–15. The ex-solution reaction can be described as follows: when the perovskite oxide system (ABO₃) is exposed to high temperatures under a reductive atmosphere (e.g., low oxygen partial pressure), catalytically active metal NPs are formed at the surface of the host oxide through the selective partial reduction. This reaction is highly rapid, cost-effective, and capable of achieving excellent particle dispersion rather than general deposition methods. Furthermore, the NPs produced through the ex-solution process exhibit superior thermal and chemical stability due to their partial embedding geometry in the surface of the host skeleton^16–20.

In general, it is widely considered that a smaller particle size and a higher NP dispersion on the oxide framework provide higher catalytic performance and more effective mass utilization. In this regard, extensive efforts have been made to date to tune the final characteristics of ex-solution particles, such as their size and number density^21–29. For this purpose, scientific approaches have primarily focused on controlling the compositions of perovskite materials and/or modifying the conditions or methods of the ex-solution reaction. As representative examples of adjusting the compositions of oxides, Gao et al. investigated the number density of ex-solution particles in relation to La doping in the Ni-doped SrTiO₃ system and revealed that La substitution at the Sr site decreases the surface density of ex-solved Ni particles³⁰. Lee et al. altered the Goldschmidt tolerance factor of perovskite oxide by changing the A-site component in Ni-doped stannate materials (i.e., Ni-doped ASnO₃, A=Ca, Sr, Ba) and found that a larger cation, which results in a higher Goldschmidt tolerance factor, produces large-sized ex-solved Ni particles with a low surface population compared to a smaller cation³¹. As other examples of controlling the ex-solution conditions or methods, Kim et al. observed that low oxygen partial pressure (pO₂) leads to a large number of ex-solution particles with a smaller size under limited amounts of the ex-solution ingredient source in Sr_0.98Ti_0.95Co_0.05O_3-δ thin-film framework²¹. Moreover, unlike traditional reductive annealing, which governs external conditions, novel approaches for ex-solution have been recently reported. Myung et al. triggered Ni ex-solution through applied potential on a La_0.43Ca_0.37Ni_0.06Ti_0.94O_3-γ electrode material, leading to a larger number of ex-solution particles compared to conventional gas-driven ex-solution¹⁴. Using the same perovskite material, Kyriakou and co-workers suggested a plasma-driven ex-solution process utilizing N₂ RF plasma. The authors reported that the particles formed by plasma treatment exhibit a higher particle count than those from conventional H₂ annealing at 650 °C³². In addition, thermal shock treatment and photothermal treatment approaches can also provide ultrafast ex-solution with a very high density of ex-solution particles^33,34. More detailed information is well summarized in several articles, including review papers^9,11,35.

Regardless of the types of approaches proposed in the aforementioned papers for tuning the size and density of ex-solved NPs, it is generally accepted, in principle, that the size and number density of ex-solution particles are determined by the combination of energetic driving force and kinetic nucleation/growth barrier during the partial reduction process of host oxide reservoirs. To utilize these backgrounds in more qualitative and quantitative descriptors, many researchers have defined and computed the segregation energy and reduction energy via density functional theory (DFT) calculations, because the ex-solution process involves the dopant surface segregation and reduction steps^{27–30,36–39}. Typically, the segregation energy is defined as the energy difference between the ex-solution ingredient at the surface and the bulk sites of the host oxide, and the reduction energy is calculated as the energy change upon oxygen removal. It is noted that various experimental results are well described by these descriptors. Therefore, the descriptors offer valuable insights into understanding the ex-solution mechanism, furthermore, open new opportunities to describe the size and surface density of ex-solved NPs through high-throughput screening with a wide range of material compositions and annealing conditions.

Despite numerous studies, attempts to directly correlate experimental findings with computational results remain limited. To the best of our knowledge, the effects of A-site substitution in perovskite oxides on the ex-solution process have been well explored; however, those of B-site doping remain relatively limited, despite the BO₆ octahedra serving as the skeleton of the perovskite structure. Moreover, while alkaline earth metals or lanthanide elements, which are mainly employed as potential A-site dopants, are thermodynamically non-reducible to the metal phase under common ex-solution conditions (e.g., 700–900 °C and pO₂ ~ 10⁻²⁰ atm), the reported candidates for B-site modulation in previous literature were typically selected as reducible elements such as Ni, Co, and Fe that can offer an additional degree of freedom for particle characteristics (i.e., NP composition). Thus, such reducible elements complicate the interpretation of B-site doping effect on the size and density of ex-solution particles^40,41. In addition, since most papers restrict their scope of investigation to a single perovskite host material, the applicability of the computational descriptors (e.g., segregation and reduction energies) for determining the size and surface density of ex-solved NPs across various types of host perovskites has not been systematically explored.

In this study, to address the aforementioned issues, we investigated the influence of B-site substitution with non-reducible elements under typical ex-solution conditions on the size distribution of ex-solved Co particles, using two different perovskite parents: Co-doped SrTiO₃ and Co-doped LaAlO₃, which are representatives of A²⁺B⁴⁺O₃ and A³⁺B³⁺O₃ perovskites, respectively. We first prepared the model dense surfaces by cleaving a sintered pellet and subsequently estimated the size and density of ex-solved particles after reductive annealing (e.g., 900 °C under 5 % H₂/Ar gases) as shown in Fig. 1a. Then, we calculated the segregation and reduction energy using DFT calculations and connected the results with experimental findings by employing a first-order nucleation model. Our observations and analyses figured out that (1) the type of host oxides predominantly determines the segregation energies of ex-solution ingredient (e.g., Co), providing an underlying explanation of differences in ex-solution particle distribution between the SrTiO₃ and LaAlO₃ hosts, (2) a B-site dopant which stabilizes perovskite materials under reducing atmosphere (i.e., leads a high value of reduction energy) produces larger particles with a lower population (Fig. 1b). Finally, to verify the potential impact of our suggested approach for tailoring the size distribution of ex-solution particles on chemical reactions, we tested the ex-solution materials as catalysts for N₂O decomposition and found a volcano-plot relationship between catalytic activity and size distribution of ex-solution particles. We believe that our results can provide valuable opportunities to tune the size and density of ex-solution particles for various environmental and energy applications.

Fig. 1. Size distribution control of ex-solution nanoparticles.

a Schematic illustration describing the overall processes of evaluating the size distribution of ex-solution nanoparticles (NPs) on model cleaved surfaces using a fragment of a sintered pellet. b Intuitive overview of the effects of host oxide types and B-site modulation on the size distribution of ex-solution NPs. Here, red and blue circles represent an oxygen ion and an ex-solution element (in this study, Co).

Results

Fabrication and Characterization of B-Site Modulated Perovskite Oxides

To explore the influence of B-site substitution by the various hard-to-reduce elements under a common ex-solution condition on the two different types of host oxides (e.g., A²⁺B⁴⁺O₃ and A³⁺B³⁺O₃) on the size distribution of ex-solution particles, we selected 10 at% Co-doped SrTiO₃ and LaAlO₃ (referred to a SrTi_0.9Co_0.1O_3-δ (STC) and a LaAl_0.9Co_0.1O_3-δ (LAC)) as benchmark materials, respectively. By employing these starting materials, we have conceived the perovskite derivatives in which 10 at% of B-site host ions (i.e., Ti⁴⁺ in STC and Al³⁺ in LAC) are replaced after considering the several factors of the Hume-Rothery rules (e.g., size and valence of cations) and thermodynamics. Specifically, we designed four variants from a STC case: SrTi_0.8Co_0.1Zr_0.1O_3-δ (STCZ), SrTi_0.8Co_0.1Hf_0.1O_3-δ (STCH), SrTi_0.8Co_0.1Nb_0.1O_3-δ (STCN), and SrTi_0.8Co_0.1Ta_0.1O_3-δ (STCT); and three variants from a LAC case: LaAl_0.8Co_0.1Ga_0.1O_3-δ (LACG), LaAl_0.8Co_0.1Sc_0.1O_3-δ (LACS), and LaAl_0.8Co_0.1Ti_0.1O_3-δ (LACT) (9 compositions in total). As shown in Supplementary Table 1 and Supplementary Fig. 1, the dopants used in this study exhibit similar charges and ionic sizes with the host ions, and show positive values of the Gibbs free energy of the reduction reactions (ΔG_red) at 900 °C (the target reduction temperature of this study).

The crystal structures of the as-synthesized perovskite oxides are identified by X-ray diffraction (XRD) as shown in Fig. 2a. As shown in this figure, we confirmed that all the materials show major perovskite diffraction patterns with the space group of Pm-3m without significant second phases. Apart from the selected oxides, we additionally tried to yield other compositions, Zr-doped LAC and In-doped LAC, however, the samples contain a large amount of different crystals such as La₂Zr₂O₇ and LaInO₃ (Supplementary Fig. 2). Thus, we excluded these compositions from the potential candidates for the further ex-solution experiments.

Fig. 2. Characterization of B-site modulation effects and the resulting ex-solution nanoparticles.

a High-resolution X-ray diffraction (XRD) patterns of B-site modulated STC and LAC series. b, c X-ray photoelectron spectroscopy (XPS) results for Co 2p and O 1s regions of B-site modulated STC and LAC series, respectively. d Scanning electron microscope (SEM) images of the cleaved surfaces of B-site modulated STC and LAC series after annealing at 900 °C with the 5% H₂/Ar gases for 10 h. e The number density and diameter of ex-solved Co particles for all compositions by evaluation of SEM images including (d). The gray line indicates the volume of ex-solution particles with the fitted value of 2.62 × 10⁵ nm³/μm_oxide². Error bars represent the standard deviation.

The chemical properties of B-site doped materials were also investigated using X-ray Photoelectron Spectroscopy (XPS) measurements (Supplementary Figs. 3, 4 and Fig. 2b, c). Figure 2b shows Co 2p XPS spectra after B-site substitution into each host framework. It is clearly shown that doping of the Zr⁴⁺ and Hf⁴⁺ (i.e., STCZ and STCH, respectively) which possess the same valence state with the host Ti⁴⁺ display the similar Co 2p_3/2 binding energy (BE) value around 780.8 eV. However, the BE of Co 2p_3/2 after the Nb⁵⁺ and Ta⁵⁺ (i.e., STCN and STCT, respectively) aliovalent substitution shifted to the lower binding energy value (BE ~ 780.3 eV), indicating the enrichment of electrons of Co. Similar trends were also observed in the case of the LaAlO₃ host material. The substitution of isovalent elements such as Ga³⁺ and Sc³⁺ with the host Al³⁺ cation shows almost equivalent Co 2p_3/2 BE value of 780.4 eV, but donor element, Ti⁴⁺, decreases the binding energy of Co species (i.e., 779.5 eV). The chemical states of the oxygen species were also checked as shown in Fig. 2c. All the O 1s XPS spectra of the materials could be deconvoluted by O_I and O_II which indicate the lattice oxygen and surface oxygen, respectively. It was found that the aliovalent donor substitution in both SrTiO₃ and LaAlO₃ systems decreases the ratios of surface-to-lattice oxygen (i.e., Area of O_II/O_I). According to the defect chemistry, the substitution of an electron donor element can be expressed as the Kröger-Vink notation⁴² as follows;

$${Nb}_2{O}_5{--------\to }^{Ti-siteinSrTi{O}_3}2{Nb}_{Ti}^{\mathrm{\bullet }}+2e^{\prime }+4O_O^{\times }+1/2O_2$$ 1

Here, we take a Nb⁵⁺ doping into the Ti⁴⁺ site in SrTiO₃ system as an example. ${Nb}_{Ti}^{\bullet }$ describes the Nb ion located at the Ti site with the +1 effective charge, $e^{\prime }$ is an electron, and $O_O^{\times }$ is a lattice oxygen, respectively. As shown in Eq. 1, we can find that donor substitution offers the additional electron (i.e., resulting in an electron-rich environment to Co) and lattice oxygen (i.e., lowering the O_II/O_I values), consistent with the results of XPS observations. Based on the XRD and XPS results, we have confirmed that the perovskite materials were well synthesized.

Moreover, the capabilities of reduction of the B-site modulated oxides were also explored via hydrogen temperature-programmed-reduction (H₂ TPR) analysis as shown in Supplementary Fig. 5. As shown in this figure, we can observe two kinds of peaks in the range of 380 – 530 °C (peak I) and 700 – 850 °C (peak II) which might be attributed to the reduction of surface Co and bulk Co species, respectively. In the case of LACG, LACS, and LACT, the peak II was not observed. Considering the two facts that (1) the peak I of LAC is located at the lowest temperature (i.e., high reducibility) among the LaAlO₃ series, and (2) the peak II of LAC was detected at around 800 °C, it is expected that the peak II of LACG, LACS, and LACT might be identified at a higher temperature than the current measurement conditions. Therefore, we examined the reducibility of the various perovskites based on the position of peak I. In the case of the STC series, it was confirmed that STCZ and STCH possess similar reducibility with STC; however, STCN and STCT, which the aliovalent-doped materials are able to readily reduce compared to the others (Supplementary Fig. 5a). We additionally explored the reducibility of LAC materials, and it can be confirmed that the degree of reducibility follows the order: LAC > LACS > LACG > LACT as shown in Supplementary Fig. 5b.

Quantitative Analysis of Ex-Solution Particles from B-Site Modulated Perovskite Oxides

Now, we produced the sintered pellets utilizing the as-synthesized B-site doped materials to quantitatively analyse the ex-solution characteristics (e.g., size and number density). Typically, a native surface generated by a high temperature annealing process offers a terracing structure rather than smooth surface, thus, we created the cleaved surface by fracturing the sintered pellets and choose it as a platform to obtain a reliable size distribution of ex-solution particles (Fig. 1a). Then, as-prepared pellet fragments were annealed under 5 % H₂/Ar atmosphere at 900 °C for 10 h, and consequently, ex-solved NPs were yielded as shown in Fig. 2d. As a comparison, ex-solved particles were also observed in the powder samples (Supplementary Figs. 6 and 7). Furthermore, through scanning transmission electron microscopy combined with energy-dispersive X-ray spectroscopy (STEM-EDS) elemental mapping, we confirmed that the component of the ex-solved NP is Co (Supplementary Figs. 8 and 9). After the reductive heat treatment, the crystal structures of the materials were examined, and we confirmed that the oxides preserved the main perovskite lattice (Supplementary Fig. 10). Subsequent to the yielding of the ex-solution particles, we first determined the particle density by counting the particles located on a specific oxide surface area (i.e., the number of NPs per oxide area). Then, we marked the particles and estimated the marked area to obtain the radius (or diameter) by assuming the particles were spheres. The representative of this approach is described in Supplementary Fig. 11. To secure the precision, approximately 50 – 100 NPs were statistically analysed. The histograms of particle size were presented in Supplementary Figs. 12 and 13. The average diameter and number density of Co ex-solution particles from all the perovskite derivatives were summarized in Fig. 2e. As shown in this figure, it can be found that the size and density exhibit an inversely proportional relationship. Moreover, the ex-solution particles show the size in a range of 7.9 – 45.6 nm (Fig. 2d, Supplementary Figs. 6 and 7) and the number density in a range of 60 – 840 #/μm_oxide². By using the obtained values of particles’ size and density, we calculated the amount of ex-solution (i.e., volume of ex-solution particles) by applying the Eq. 2 as follows;

$$Particledensity\left(\mathrm{\#}/\mu m_{oxide}^2\right)=\frac{6\times Volume\left(n{m}^3/\mu m_{oxide}^2\right)}{\pi \times {Diameter\left(nm\right)}^3}$$ 2

The gray line in Fig. 2e indicates the volume of ex-solution particles with the fitted value of 2.62 ×10⁵ nm³/μm_oxide², and it is shown that the amount of ex-solution is almost similar regardless of the types of host oxides and/or B-site dopants. One notable point is that doping with aliovalent cations (e.g., STCN, STCT, and LACT) results in sizes and densities that deviate significantly from those of the starting materials, STC and LAC, respectively. For example, STCN and STCT exhibit the smallest particle size and the highest surface density, while LACT shows the largest particle size and the lowest particle density. In addition to this, we also try to find a connecting point between the reducibility of B-site modulated materials (which is obtained by H₂ TPR measurements) and ex-solution particles (Supplementary Fig. 14). It is revealed that a B-site dopant which is able to stabilize the perovskite materials under a reductive condition leads to larger-sized ex-solution particles in both STC and LAC systems (i.e., a positive relationship between H₂ TPR peak position and ex-solution particle size).

Correlation between the Size Distribution of Ex-Solved Particles and Atomistic Energies

To rationalize the experimental trends in the size distribution of ex-solution particles, we have performed DFT calculations. The atomistic models we used in this study are shown in Fig. 3a. To model the doped structures, we substituted one atom in the middle of the slab. Additional details regarding the modeling and simulations are provided in the experimental section. Then, we calculated the segregation (more precisely, the segregation of cobalt in this study) and reduction energies to describe the ex-solution phenomena, following previous studies (Fig. 3b)^{18,24,28,38,43}. The particle density was then fitted to the equations below, following ref. ²⁸;

$$\mathrm{\Delta }{G}_{seg\_\_fit}=c_1+c_2\mathrm{\Delta }{E}_{seg}^{DFT}$$ 3

$$\mathrm{\Delta }{G}_{red\_\_fit}=c_3+c_4\mathrm{\Delta }{E}_{red}^{DFT}$$ 4

$$Particledensity\left(\mathrm{\#}/\mu m_{oxide}^2\right)=\frac{\exp \left(-\frac{\mathrm{\Delta }{G}_{seg\_\_fit}}{k_{B}T}\right)}{1+\exp \left(-\frac{\mathrm{\Delta }{G}_{seg\_\_fit}}{k_{B}T}\right)}\exp \left(-\frac{\mathrm{\Delta }{G}_{red\_\_fit}}{k_{B}T}\right)$$ 5

where $\mathrm{\Delta }{E}_{seg}^{DFT}$ and $\mathrm{\Delta }{E}_{red}^{DFT}$ are the segregation and reduction energies from DFT calculations, respectively, $k_B$ is the Boltzmann constant, and T is the temperature. $c_1$-$c_4$ are fitting parameters (0.216, 2.422, 0.099, and −2.777, respectively) used to fit DFT energies of the simple model to the realistic values of free energies. Figure 3c and Supplementary Table 2 show the fitted results for the STC and LAC series. The strong agreement between experimental measurements and theoretical calculations indicates that the ex-solution behaviors in both STC and LAC systems can be consistently explained by the competition between segregation and reduction, even across different host materials. We also performed calculations at different doping concentrations and found that the density values of STCN and STCT vary substantially with doping concentration, whereas those of the other systems show relatively smaller changes, consistent with the trend described below (see Supplementary Fig. 15). Notably, the reduction energies show strong agreement with H₂ TPR measurements (Supplementary Fig. 16), which also shows that the materials doped with B-site elements resulting in high reduction energies have a tendency to be difficult to reduce in each host framework. These agreements support the reliability of our simulations in explaining the experimental results.

Fig. 3. Computational investigation of the size distribution of ex-solution nanoparticles.

a Atomic models of STC and LAC, with Nb-doped STC (STCN) and Ti-doped LAC (LACT). b Atomic models of dopant segregation and perovskite reduction processes. c Comparison of the number density of ex-solution particles obtained from experiments and calculations across all compositions in this study. The inset graph displays the linearity between experimental and calculated values. d Segregation energies ($\mathrm{\Delta }{E}_{seg}^{DFT}$) of B-site modulated STC and LAC series. e, f Reduction energies ($\mathrm{\Delta }{E}_{red}^{DFT}$) and those effects on size and density of ex-solution particles in B-site modulated STC and LAC series, respectively. Error bars represent the standard deviation.

To gain further chemical insight into the ex-solution trends, we analyzed how the two key descriptors, segregation and reduction energies, depend on the material compositions. Figure 3d presents the segregation energies for different host and dopant types. We found that the segregation energies of STC series are generally lower than those of LAC series, regardless of dopant types. Additionally, Fig. 3e, f show the particle density and diameter as functions of the reduction energy for STC and LAC, respectively. These figures prove that the reduction energy exhibits a linear relationship with both density and diameter. Specifically, a higher reduction energy value leads to larger and fewer ex-solution particles. To further quantify these trends, we evaluated the relative contributions of segregation and reduction energies to the density values using Eq. 5, confirming that the density values of STC series are mainly governed by segregation energy, whereas those of LAC series are mainly governed by the reduction energy (Supplementary Note 1 and Supplementary Fig. 17). By integrating these outcomes, it can be deduced that the host material composition has a greater influence on segregation, while B-site modulations primarily affect reduction in the ex-solution process. This can be rationalized by considering that the segregation energy of the ex-solved element (i.e., Co in this study) is determined by the local chemical bonding environment of the host material, as the segregation process of ex-solution elements involves breaking numerous bonds between host oxide ions and the ex-solvable ions. In contrast, the reduction process is governed by the excess electron density of the system, which generally depends on the compositions of the B-site of a perovskite system.

Interestingly, we find that the density trends of ex-solution particles upon n-doping (e.g., aliovalent doping) exhibit opposite behaviors for STC and LAC systems. The particle densities of STCN and STCT are higher than that of STC, whereas the density of LACT is lower than that of LAC. Correspondingly, the reduction energies of STCN and STCT are lower than that of STC, while for LACT, the reduction energy is higher than that of LAC (Fig. 3e and f). According to conventional understanding, if the excess electron from n-doping localizes on surface oxygen, the reduction energy should decrease, as observed in STC. However, the unique behavior of LAC suggests that the excess electron occupies states other than surface states. To investigate this further, we plotted the density of states (DOS) for pristine and reduced STCN and LACT in Supplementary Fig. 18. The results show that, in LACT, the excess electron occupies bulk Co sites in addition to surface sites, whereas in STC, the electron primarily localizes at surface sites. Bader charge analysis (Supplementary Table 3) further supports this trend, revealing a significant increase in bulk Co electron occupancy in LAC upon Ti doping. This explains the opposite trend of reduction behavior in LAC upon n-doping.

Overall, these findings indicate that the number density and size of ex-solution NPs can be effectively tailored by (1) the types of host materials and (2) the compositions of B-site elements. Moreover, it is clear that, at the atomic level, these factors influence segregation and reduction energetics, which can be accurately described by theoretical calculations.

Demonstration as Catalytic Materials for N₂O Decomposition for Greenhouse Gas Mitigation

Next, to explore the applicability of the methodology on tailoring the size and density of ex-solution particles in the fields of heterogeneous catalysis, we selected the direct nitrous oxide (N₂O) decomposition as a model reaction ($N_{2}O\to N_2+1/2O_2$). N₂O gas is the third largest emitter of greenhouse gases, following CO₂ and CH₄, according to international regulations, and it causes approximately 310 times the greenhouse gas effect compared to CO₂ gas (i.e., CO_2-eq of N₂O gas is 310)^44–47. Despite its negative impact on global warming, the demand and use of N₂O have increased, as it is often used as a processing gas in semiconductor and display manufacturing processes, including those of Samsung companies. In addition, research related to N₂O decomposition is being conducted in academic circles as well. Hinokuma et al. reported that there is no NO_x (NO and NO₂) selectivity during N₂O decomposition reaction using Ru/SnO₂ catalyst, indicating that only N₂ and O₂ are produced⁴⁸. Moreover, according to previous papers⁴⁴, various types of metal oxides were employed as catalysts, such as simple binary oxides, spinel, and perovskite structures that contain Co, Fe, and Cu elements. Nevertheless, to the best of our knowledge, it is very limited to apply the ex-solution technology for N₂O decomposition reaction. Considering these factors, we try to demonstrate the feasibility of tailored ex-solution catalysts for N₂O decomposition reaction.

The light-off curves of N₂O conversion (decomposition) for various B-site modulated STC and LAC series after annealing at 900 °C and with the 5 % H₂/Ar gases for 10 h are shown in Fig. 4a. For clarity, STC and LAC series are presented as square symbols with solid lines and circle symbols with dashed lines, respectively. As shown in this figure, it is found that the LAC series shows high N₂O conversion activities compared to those of the STC series, indicating that LaAlO₃ host materials are more active compared to SrTiO₃ framework for N₂O decomposition. We also checked the kinetic data of prepared samples (Fig. 4b). It is confirmed that the same oxide hosts exhibit similar activation energy values (i.e., STC series; 78 kJ/mol and LAC series; 108 kJ/mol), which suggests that the N₂O decomposition mechanisms in each host framework are comparable, while the differences arise due to variations in the frequency factor (k₀) in Arrhenius equation of reaction rate constant (k), which is proportional to the number of active sites. In addition, to provide a more comprehensive comparison of catalytic activity, we surveyed previously reported Co-containing perovskite catalysts for N₂O decomposition (Supplementary Table 4). As shown in the Supplementary Table 4, although the reaction conditions vary slightly among studies, the T₅₀ values are generally observed in the range of 450–700 °C in most cases. Therefore, the catalytic activity measured in this study is considered to be comparable to that of typical Co-containing perovskite catalysts for N₂O decomposition. Notably, while reported catalysts typically contain approximately 10–20 wt% Co, the Co content in our catalyst is only 2.71 wt%, which is roughly an order of magnitude lower. From the perspective of mass utilization, our catalyst proposed in this study demonstrates relatively excellent N₂O decomposition activity despite the significantly reduced Co loading.

Fig. 4. Catalytic performance of ex-solution materials.

a Light-off curves pertaining to N₂O conversion (N₂O decomposition) with various B-site modulated STC and LAC series after annealing at 900 °C and with the 5% H₂/Ar gases for 10 h, b Arrhenius plot (kinetic data) for N₂O conversion, c A plot of reaction rate of N₂O decomposition vs. the number of surface Co atoms in ex-solved NPs on B-site modulated STC and LAC series with three different temperatures; 700 °C (empty symbol with dashed line), 750 °C (half-filled symbol with short-dashed line) and 800 °C (full-filled symbol with solid line). Here, the notation “900 R” means a reduction process was carried out at 900 °C. The shade region indicates an optimal ex-solution region for N₂O decomposition reaction among the tailored ex-solution catalysts.

We also conducted durability tests on the two catalysts, LAC 900 R and LACG 900 R (i.e., LAC and LACG after 900 °C reduction with 5% H₂/Ar gases, respectively), which exhibited high N₂O decomposition efficiency. The LAC 900 R and LACG 900 R catalysts stably operated with N₂O conversion values of 96.3 % and 88.8 %, respectively, for 100 h at 800 °C (Supplementary Fig. 19). The surface morphology of the samples before and after the long-term stability test was also examined using SEM. As shown in Supplementary Fig. 20, both catalysts were not degraded under long-term operation, demonstrating a robust and sustainable performance of ex-solution catalysts.

Many studies using Co species reported that Co acts as an active site in direct N₂O decomposition, where redox reactions play a crucial role^44–46. Taking this into account, we confirmed that ex-solved Co nanoparticles in LAC and STC series after the ex-solution process are the active sites of N₂O catalytic conversion through control experiments. The LaAlO₃ (LAO) and SrTiO₃ (STO) exhibited significantly lower N₂O conversion compared to that of the LAC 900 R and STC 900 R samples, respectively (Supplementary Fig. 21). These results proved that active sites of the catalysts are ex-solved Co nanoparticles rather than the host materials. According to previous literature, N₂O is considered to adsorb a Co active site on the surface, followed by cleavage of the N-O bond. Afterward, two adsorbed oxygen atoms are combined by the Langmuir-Hinshelwood mechanism and desorbed as gaseous O₂, as shown in Eqs. 6 and 7^45,49.

$$N_{2}O+{Co}^{*}\to \left({Co}^{*}-O\right)+N_2\left(g\right)$$ 6

$$2\left({Co}^{*}-O\right)\to {2Co}^{*}+O_2\left(g\right)$$ 7

In this regard, we specifically investigated the relationship between N₂O decomposition activity and the number of Co atoms on the surface of ex-solved Co NPs formed with various size distribution as shown in Fig. 4c. As shown in this figure, both LAC and STC series present a volcano-plot relationship, indicating the presence of an optimal region for the reaction (see highlighted region) at three different reaction temperatures (e.g., 700, 750, and 800 °C). Specifically, the STC series exhibited the highest catalytic performance when the number of Co atoms on the NP’s surface was approximately 2.1 × 10⁻⁷ mol, while the LAC series achieved peak performance at around 3.1 × 10⁻⁷ mol.

Consequently, such an alternative approach opens up various possibilities for enhancing catalytic activity through the selection of host materials, B-site modulation, and changes in the size distribution of ex-solved NPs, which affect the number of active sites in catalytic reactions.

Discussion

We modulated the type of host frameworks (e.g., SrTiO₃ and LaAlO₃) and B-site composition of perovskites, and investigated the relationship between their physicochemical properties and the size distribution of NPs formed via the ex-solution process. Combining experimental observations with a computational approach (i.e., DFT calculations) provided a more comprehensive understanding of the material behavior, linking theoretical descriptions with experimental outcomes. In all experimental groups used in this study, the size and surface density of the ex-solved particles showed an inverse relationship. Notably, the particle density could be determined by the interplay of two atomic-level descriptors: segregation and reduction energies of materials. More specifically, the segregation energy is predominantly influenced by the type of host perovskites, while the reduction energy is determined by the composition of the B-site of perovskites. Within the same host material, less reducible perovskites tend to form fewer but larger ex-solved particles. Lastly, by incorporating such surface-tailored ex-solution materials, we measured the catalytic activity of the N₂O decomposition reaction and identified the optimal number of catalytic active sites in the ex-solution systems. This work advances our capability to design tailored ex-solution catalysts for environmental and energy applications, optimizing material properties for specific catalytic functions.

Methods

Precursors and sample preparation

SrTi_0.9Co_0.1O_3-δ (STC), SrTi_0.8Co_0.1Zr_0.1O_3-δ (STCZ), SrTi_0.8Co_0.1Hf_0.1O_3-δ (STCH), SrTi_0.8Co_0.1Nb_0.1O_3-δ (STCN), SrTi_0.8Co_0.1Ta_0.1O_3-δ (STCT), LaAl_0.9Co_0.1O_3-δ (LAC), LaAl_0.8Co_0.1Ga_0.1O_3-δ (LACG), LaAl_0.8Co_0.1Sc_0.1O_3-δ (LACS) and LaAl_0.8Co_0.1Ti_0.1O_3-δ (LACT) (9 compositions in total) materials were synthesized by a conventional solid-state reaction method. Stoichiometric amounts of carbonate and oxide precursors, such as SrCO₃ (Sigma-Aldrich, 99.9 %), TiO₂ (Sigma-Aldrich, 99.7 %), Co₃O₄ (Sigma-Aldrich, 99.7 %), ZrO₂ (Sigma-Aldrich, 99 %), HfO₂ (Sigma-Aldrich, 99.95 %), Nb₂O₅ (Sigma-Aldrich, 99.99 %), Ta₂O₅ (Sigma-Aldrich, 99.99 %), La₂O₃ (Sigma-Aldrich, 99.9 %), Al₂O₃ (Sigma-Aldrich, 99 %), Ga₂O₃ (Sigma-Aldrich, 99.99 %), In₂O₃ (Sigma-Aldrich, 99.99 %, not included in main experiment due to 2^nd phase) and Sc₂O₃ (Alfa Aesar, 99.9 %) were mixed in ethanol or acetone media with zirconia balls and ball-milled overnight. The mixtures were vigorously stirred and dried on a hot plate; subsequently, the perovskite powders were obtained after annealing under ambient air. Detailed annealing conditions are follows: STC, STCZ, STCH, STCN, and STCT: 1200 °C for 8 h with a ramping rate of 4 °C/min, afterward, 1400 °C for 8 h with a ramping rate of 4 °C/min, LAC and LACG: 1300 °C for 24 h with a ramping rate of 5 °C/min, LACS and LACT: 1400 °C for 24 h with a ramping rate of 5 °C/min. To obtain cleaved surfaces for ex-solution analysis, the powders were shaped into a cylindrical form (diameter ~ 10 mm and height ~ 2 mm) by uniaxial press and each perovskite material was sintered under the same conditions as the final calcination condition. Each annealing condition was selected to yield optimal phase quality and morphology depending on the material composition. The as-sintered pellets were cleaved into four pieces and then subjected to a reduction heat treatment at 900 °C for 10 h in a 5 % H₂/Ar atmosphere with a ramping rate of 5 °C/min to obtain Co ex-solution particles.

Physical and chemical characterization

The crystal structures of the synthesized perovskite materials were analysed using a X-ray diffraction (XRD, D8 Advance, Bruker) with Cu Kα radiation beam source. X-ray photoelectron spectroscopy (XPS, Quantera II, ULVAC-PHI) measurements were performed to obtain the surface chemical states of the perovskites using Al Kα radiation source with 55 eV of pass energy and 0.1 eV step size. The microstructures of the cleaved surfaces after ex-solution reaction were observed by scanning electron microscopy (UHR SEM, SU-9000, Hitachi). By using SEM images, the size distributions of ex-solved nanoparticles were evaluated (the detailed method was described in supporting information). Reducibility of the perovskites were evaluated by H₂ temperature programmed reduction (H₂ TPR) measurement (AutoChem 2950 HP, micromeritics). The temperature was elevated from 50 °C to 800 °C with a ramping rate of 10 °C/min under 3% H₂/Ar atmosphere.

DFT calculations

All DFT simulations were performed using the Vienna Ab Initio Simulation Package (VASP 5.4.4 version)⁵⁰ with projector augmented-wave (PAW) pseudopotentials⁵¹. The Perdew-Burke-Ernzerhof (PBE) exchange-correlation functional was employed⁵², and a Hubbard U correction⁵³ was applied to the Co 3 d states, with U–J set to 3.3 eV. To model the surface slab, the (110) facet was considered. The surface areas of the ab planes for STO and LAO were set to 11.14 Å × 7.87 Å and 10.82 Å × 7.63 Å, respectively (see Fig. 3a for details). The cutoff energy and k-point grids were determined through convergence tests using criteria of 10 meV/atom for energy, 0.1 eV/Å for force, and 10 kbar for stress. A 1 × 2 × 1 k-point mesh was used to sample the Brillouin zone (BZ) for both materials. The plane-wave cutoff energy was set to 500 eV. The initial magnetic moments of Co atoms were set at 3.0–5.0, and were adjusted such that the resulting spin moments of Co atoms were approximately 2.6–2.7. The surface terminations of STO and LAO were determined following the method described in ref. ²⁸ (Supplementary Fig. 22 for details). For STO, the most stable termination was identified as SrTiO₂, consistent with the previous literature. In the case of LAO, the LaAlO_1.5 termination was found to be the most stable. The dipole correction was applied to avoid spurious periodic interactions⁵⁴. The structural relaxation was performed until the maximum force was reduced to 0.05 eV/Å.

Catalytic test for N₂O decomposition

N₂O decomposition was tested in situ with a Fourier transform infrared spectroscopy (FT-IR, I4001-E, MIDAC Corporation) equipped with a gas cell in a fixed-bed Inconel reactor. In the catalytic bed, quartz wool was placed at the bottom to prevent catalyst displacement, and 100 mg of the catalytic material was packed. Prior to the measurement, the powders were heat-treated in synthetic air (100 sccm) at 180 °C for 1 hour to remove any surface adsorbates. For N₂O decomposition test, 100 sccm in total reactant gas was injected into the reactor in a composition of 3000 ppm N₂O with He balance gas corresponding to weight hourly space velocity (WHSV) of 60,000 mL g_cat⁻¹ h⁻¹. The reaction temperature was elevated from 25 °C to 800 °C with a ramping rate of 4 °C/min. The N₂O conversion (%) is defined as a percent ratio of (${\mathrm{mol}}_{N_2\mathrm{O,in}}$ – ${\mathrm{mol}}_{N_2\mathrm{O,out}}$)/${\mathrm{mol}}_{N_2\mathrm{O,in}}$. The number of surface Co atoms in ex-solution particles were estimated as 2r²/r_Co², where r means radius of an ex-solution particle and r_Co means atomic radius of Co atom (i.e., 0.167 nm).

Author contributions

J.K.K., S.K.K., and J.-M.L. performed the synthesis and characterization of the materials. S.K. performed the DFT calculations. J.K.K., S.K.K., S.K., J.-M.L., and H.J.K. analyzed the data and wrote the manuscript. J.K.K. and H.J.K. conceived the idea, designed the experiments, and wrote the paper. All authors contributed to the draft and revision of the manuscript.

Peer review

Peer review information

Nature Communications thanks Wangcheng Zhan and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

All authors disclose support for the research of this work from Samsung Electronics Co., Ltd.

Data availability

Source data are provided with this paper. Data are available from the corresponding authors upon request.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-74102-z.