Dynamics of dislocation formations and their impacts on exsolution in Ru-doped perovskite oxide Authors: Sungwook Choi, Younghwan Lim, Puspendu Guha, Hayoung Kim, Jaeseung Kim, Sungwon Kim, Ross Harder, Wonsuk Cha, Hoyoung Suh, Jinseok Ryu, Sungeun Yang, Ho-Il Ji, Deok-Hwang Kwon, Hyunjung Kim Published: 2026 ABSTRACT AbstractSelf-assembled metal nanoparticles exsolved from host oxides have gained prominence in catalysis and electrochemistry owing to their exceptional activity and stability. Understanding the relation between dopant transport and exsolution is important, as the transport mechanism of dopants toward the surface of the host oxide directly influences exsolution sites, density, and dispersion, ultimately determining catalytic functionality. However, the pathways for dopant transport and their interactions with internal defects during exsolution remain unclear due to the complexity of defects hidden in the bulk. Here, we reveal the exsolution pathway mediated by dislocation evolution within a host oxide perovskite. By employing in situ Bragg coherent X-ray diffraction imaging and transmission electron microscopy, we show that dislocations nucleate in the bulk interior and propagate to the surface during the reduction of Ru-doped BaCe0.85Y0.1Ru0.05O3-δ. Moreover, we verify that the Ru dopant is specifically correlated with the formation of mixed dislocations, which act as mobile vehicles that dynamically carry Ru defects to the surface in tandem with dislocation propagation. These findings advance our understanding of dislocation dynamics and support the development of exsolved metal nanoparticles for next-generation catalysts.Subject terms: Fuel cells, Imaging techniques, Structural properties Introduction Dislocations and element segregation have been long-standing topics of study1–3, as they are prevalent in crystalline solids and significantly influence the mechanical and chemical properties of materials4,5. In particular, in ionic materials, such as ceramics, chemical modulations associated with dislocations can significantly alter the transport of electrons or ions. Despite their importance, the precise roles of dislocations and the associated chemical changes have remained unclear due to the complexity of these processes hidden in the bulk. Exsolution is a pivotal example of functional element segregation, in which dopant elements incorporated into a host matrix migrate to the surface and segregate as nanoparticles (NPs) driven by matrix reduction. Over the last few decades, exsolution has received significant attention as a controlled segregation process of dopant for creating highly active and stable NPs on the surface of functional oxide supports, which holds great promise for applications in thermo- and electro-catalysis6–8. Driven by the experimental recognition of its potential, extensive research on the exsolution mechanism has been actively pursued9–13. In particular, the role of structural defects—such as surface defects14, surface space charges15, interfacial strain16,17, and boundaries at pre-existing extended defects18–20—has been extensively discussed. Despite these efforts, exsolution exhibits significantly different behaviors depending on factors, such as the dopant and support oxide materials21, geometry of parent oxides12, and extrinsic control parameters, including temperature and gas atmospheres12,22. This variability presents a significant challenge for developing a consistent mechanistic framework23. Exsolution can occur with a wide variety of matrix–dopant combinations. Extensive studies have focused on Ni, Fe, and Co dopants in perovskite oxides due to their ease of exsolution at relatively low temperatures and their broad applicability in catalytic and electrochemical systems10,24–27. In particular, Ni exsolution from perovskite hosts, such as (La,Sr)(Ti,Fe)O₃, has been extensively investigated, with emphasis on its catalytic and electrochemical performance as well as the underlying formation mechanisms28–30. While noble metals exhibit exceptional catalytic activity, their high cost remains a critical limitation. The development of structurally exsolved noble metal nano-catalysts represents a promising strategy to address this issue. Nevertheless, noble metals are particularly challenging to exsolve due to their high reducibility, which often leads to decomposition and retention within the bulk of the support oxide24. Achieving a detailed understanding of the exsolution mechanism specific to noble metals is crucial for optimizing their exsolution, enabling precise control over nanoparticle size and distribution, and ultimately enhancing their catalytic performance. Here, we elucidate that the segregation of noble metal dopants is directed by dislocations. We investigate the BaCe0.85Ru0.05Y0.1O3-δ (Ru-BCY) particles as a model system by employing the deep probing capability of Bragg coherent diffraction imaging (BCDI) for three-dimensional (3D) strain imaging, combined with atomic-scale elemental mapping from transmission electron microscopy (TEM) near dislocations. Analyzing in situ 3D maps of the elastic response to exsolution and the elemental distribution near dislocations, we reveal the evolution mechanism of dislocations during bulk reduction and their intertwined role in redistributing noble metal dopants. Our findings show that Ru exsolution predominantly occurs at the ends of mixed dislocations, along which continuous Ru segregation and depletion develop. In addition, the dynamics of the strain field evolution indicate that mixed dislocations serve as interactive pathways for metal dopants during the exsolution process. Results Dislocation evolution associated with Ru doping To determine how dislocations relate to exsolution, Ru-BCY particles were used as a model system, which have been used in electrochemical energy conversion devices and chemical processes as catalysts24,31–33. Prior to the high-temperature reduction treatment, the formation of a single-phase perovskite without secondary phases was confirmed by X-ray diffraction (XRD) and scanning transmission electron microscopy/energy dispersive X-ray spectroscopy (STEM/EDS) analysis (Supplementary Fig. 1). To investigate the early stages of exsolution, experiments were conducted at 800 °C, near the exsolution onset temperature, as validated by extended X-ray absorption fine structure (EXAFS) and TEM analyses (Supplementary Figs. 2 and 3). Notably, picometer-level lattice deformation is detectable in BCDI34,35, enabling real-time monitoring of the lattice deformation in the host oxide particle during exsolution. The in situ measurements were conducted under a reducing condition (3.5–5% H2 in N2 at 800 °C), replicating the operational environment of solid oxide electrochemical cells near the onset temperature where exsolution occurs. Combined with in situ BCDI, the relation between dislocation lines and Ru exsolution is further investigated by TEM. Figure 1 shows the dislocation topologies identified by in situ BCDI upon reduction of the Ru-BCY particle, as well as TEM results, comparing the doped and non-doped BCY matrices after reduction. Note that the apparent particle morphologies differ between BCDI and TEM due to different projection directions and imaging principles (3D coherent diffraction vs. 2D transmission). Figure 1a shows a three-dimensional (3D) image of a Ru-BCY crystal, exhibiting a dislocation core that developed after 194 min of reduction. The dislocation core appears as a spiral singularity on the lattice displacement field (Fig. 1b) and exhibits a low Bragg electron density (hollow regions in Fig. 1b, c), indicating a highly deformed structure4,36. A black solid line indicates a 20% iso-level, and opaque regions indicate a 10% iso-level with low Bragg electron density near the dislocation core. Tracking these singular points in 3D space, the branched and jointed displacement spirals along the slice height (h) represent each dislocation segment, constituting a loop-like topology37 (Fig. 1a). Different spiral-handedness indicates that two dislocation segments have opposite signs, leading to significant strain sign inversion across the dislocation loop (Fig. 1c and Supplementary Fig. 4a–c; all observed dislocations are labeled from #1 to #5). By taking the radial distribution of the lattice displacement located 20 nm from the dislocation core, the Burgers vectors (b) of the dislocations are deduced. In Fig. 1d, the radial plot for the two singular spirals in Fig. 1b (red: anticlockwise and blue: clockwise) represents the displacement field singularity ranging from −1.56 to 1.56 Å (Supplementary Fig. 4d, e). Given that the radial-line integral of displacement around the dislocation core represents the projection of b projected onto Q = [110], most dislocations (except dislocation #5) exhibit a net displacement equivalent to d110 = 3.11 Å (interplanar spacing of the (110) plane), indicating that b = a2[110] (a is the lattice constant of Ru-BCY)38,39, which is parallel to Q (Supplementary Fig. 5). Exsolution-driven changes in the host perovskite oxide. a 3D dislocation topology after exsolution (3.5% H2/N2 at T = 800 °C for 194 min). The semi-transparent surface represents a 20 % iso-surface, and the cyan region indicates the dislocation core. b Displacement field and c strain field along Q = [110] direction are mapped on three slices at different positions. To illustrate the low Bragg electron density near the dislocation core, the iso-level slices at 20% and 10% are shown as solid lines and hollow regions, respectively. d Magnitude of displacement field along the radial coordinates at 20 nm from two singular spirals (red: anti-clockwise and blue: clockwise). e In situ strain evolution as a function of time. The Q-including slices are taken at the center (red frame) with 20% and 10% iso-level slices represented by solid lines and hollow regions, respectively. f Normalized strain energy for Ru-doped and non-doped NP. Note that each strain energy is normalized with the average strain energy at the pristine state. g–j Electron microscopy analysis of Ru-doped BCY after reduction, showing the formation of exsolved Ru NP and dislocations. g BF-TEM image of Ru-doped BCY after exsolution, h along with enlarged HR-TEM and i EDS map images for the exsolved Ru NP (red square region in g) are presented. j NP exsolution in proximity to the dislocations after reduction is depicted in the WBDF-TEM image. Arrows labeled g and b vectors indicate the reflection vector of the WBDF-TEM image and the Burgers vector of the marked dislocation line, respectively. k BF-TEM image and l WBDF-TEM image of non-doped BCY after reduction. Figure 1e represents strain projection along the [110] direction as a function of reduction time in Ru-BCY. To underline the low Bragg electron density in highly deformed regions (Supplementary Fig. 6), which is not due to morphological changes of the nanoparticles, iso-level slices at 10% are shown as semi-transparent regions. Homogeneously distributed strain fields in the pristine state become inhomogeneous, highly concentrated tensile strain under the reducing atmosphere until 120 min. At 120 min of reduction, dislocation nucleation is initially observed as the transformation of a tensile region into a compressive region, inducing a tensile-compressive strain distribution adjacent to the dislocation core4,40 (Supplementary Fig. 7a). The normalized strain energy (Fig. 1f and Supplementary Fig. 7b) increases linearly with the reduction time of Ru-BCY, whereas no elastic response is observed for non-doped BCY (Supplementary Fig. 8). Given that the strain response followed by dislocation nucleation is observed only in the Ru-doped case, Ru doping must play a crucial role in initiating dislocation formation. The continuous increase of strain energy in Ru-doped BCY also indicates that dislocations accumulate rather than dissipate within our experimental time window. One may think that Ce is also an easily reducible element in the system; its reduction and subsequent chemical expansion might induce the formation of dislocation nuclei. However, in situ BCDI results in Fig. 1e show not the global, but the localized strain field development in the bulk, which would not have been expected if chemical expansion of Ce, the dominant element in BCY, were the dominant mechanism. The direct effects of Ru dopants on dislocation formation were verified by bright-field (BF)-TEM, high-resolution (HR)-TEM, EDS29, and weak-beam dark-field (WBDF)-TEM imaging in Ru-doped (Fig. 1g–j) and non-doped (Fig. 1k, l) BCY NPs. After reduction, the protruded exsolved Ru NPs were observed on the surface of the BCY host oxide in the BF-TEM images (Fig. 1g). The exsolved Ru NPs were confirmed by the lattice fringes determined in the HR-TEM images (Fig. 1h) and the identification of the Ru element in the EDS compositional maps (Fig. 1i). Notably, exsolved Ru NPs were observed at the ends of the dislocation lines, visualized as bright lines in the WBDF-TEM images (Fig. 1j). In contrast, the non-doped BCY showed no dislocations after the same reduction treatment. No visible bright lines were observed in the WBDF-TEM images (Fig. 1k, l). These observations imply that the Ru dopant triggers dislocation formation, which is consistent with BCDI results, and that the exsolved Ru NPs form at the ends of dislocation lines. Exsolution dependence on the dislocation type To examine the correlation between dislocation evolution and Ru exsolution, we investigated 60 regions, identifying 53 dislocations and 28 exsolved Ru NPs (Fig. 2). To carefully count the number of Ru exsolutions, Ru EDS signals or fringes from high-resolution images are used to cross-confirm the exsolved Ru particles as seen in Fig. 1, eliminating any miscounting (Supplementary Fig. 9). Surprisingly, the exsolved Ru NPs exhibited strong selectivity, preferentially forming at the end of specific dislocations. Dislocation types, such as edge, screw, and mixed, are identified by the angle between the dislocation line (i.e., line-sense vector) and the Burgers vector. The direction of Burgers vectors was determined under two-beam WBDF-TEM imaging conditions using the g⋅b rule (see Supplementary Fig. 10 and “Methods” for the detailed determination of dislocation types). The dislocation types noted here are defined by the character of segments near the surface rather than those in the bulk. As seen in Fig. 2 and Supplementary Fig. 10, the dislocation lines are generally curved, such that the local character of a given dislocation can vary along its length. Notably, we observed a statistically significant correlation when the character of dislocation segments near the surface was taken into account. This is physically reasonable because exsolution involves nucleation and growth at the oxide–gas interface, making the dislocation character at or very near the surface intersection most relevant. Survey of Ru exsolution events associated with dislocations. a–d WBDF-TEM and EDS surface map images of Ru-BCY matrices with Ru nanoparticles exsolved at the terminations of mixed dislocation lines. e–h WBDF-TEM and EDS surface map images of other Ru-BCY matrices with no Ru nanoparticles exsolved at the terminations of edge dislocation lines. Insets in (a, c, e, g) show selected area electron diffraction patterns with (200) reflections at the [001] zone axis under two-beam conditions. Arrows labeled b indicate the Burgers vectors of the dislocation lines, positioned at their respective dislocation terminations. i Statistics of TEM analysis showing the percentage distribution and number of different dislocation types associated with exsolution and no-exsolution cases. j Schematic illustration summarizing the observed high selectivity of Ru exsolution at mixed dislocations. Dislocations related to exsolution are only observed for the mixed dislocation type, with no instances of pure edge or screw dislocations involved. Note that all dislocations were identified from a particular [001] zone axis orientation, and that a spherical particle schematic is used to represent our observations, while the particles have a certain shape. We observed 20 edge dislocations, five screw dislocations, and 28 mixed dislocations—the mixed and edge types were predominant. Representative WBDF-TEM images are presented in Fig. 2a, c, e, g, revealing a few bright white lines corresponding to mixed dislocations (Fig. 2a, c) and edge dislocations (Fig. 2e, g). The surface EDS mappings of Ru and Ba are also presented, demonstrating a correlation between dislocation type and exsolution behavior—Ru exsolutions were predominantly found at the mixed dislocations (Fig. 2a–d), not at the edge dislocations (Fig. 2e–h). Statistical analysis of WBDF-TEM and EDS results is summarized in Fig. 2i, showing that 75% of exsolved Ru NPs were found at the ends of mixed dislocations, whereas the other 25% were not involved with any dislocations. To visualize the dislocation-selective exsolution phenomenon, a schematic representation is provided in Fig. 2j. Note that our dislocation analysis is mainly taken from the [001] zone axis; therefore, the exsolved NPs identified without visible dislocation (~11%) may still be correlated with mixed dislocations oriented in different crystallographic directions. Mixed dislocations without Ru exsolution (~11%) are also explicitly counted (green sector). These cases likely reflect insufficient local driving force from incomplete redistribution of Ru at the time of observation. Also, some of the NPs exsolved at mixed dislocation terminations may be obscured by the host matrices. It should also be noted that low-angle grain boundaries, typically characterized by periodic arrays of dislocations, were rarely observed. While this may be related to the exsolution process, the scarcity of grain-boundary features precluded their consideration in the present study. Our statistical analysis explicitly accounts for these distinctions. To further investigate the chemical composition of Ru exsolution at dislocation terminations, we performed detailed EDS analysis. Two representative cases were demonstrated in Fig. 3, exhibiting the WBDF-TEM and Ru elemental map images from two Ru-BCY particles with and without exsolved Ru nanoparticles. Note that STEM/EDS mappings were taken from the same top-view direction as WBDF-TEM imaging, visualizing the area where dislocations are embedded within the matrix along the beam direction rather than from a side-view. Consequently, EDS provides projected chemical information for the corresponding area along the beam direction. The positions of dislocation lines identified in WBDF-TEM images were cross-confirmed with complementary STEM images obtained at multiple collection angles (Supplementary Fig. 11), ensuring accurate spatial correlation with EDS mappings as well. Element analysis at mixed and edge dislocation cores. a–d EDS analysis at a mixed dislocation core. a WBDF-TEM image of a single matrix viewed along the [001] zone axis. b Ru elemental net intensity map acquired from the surface region indicated by the red dashed box in (a). c Three-dimensional waterfall plot of line profiles at 1 nm intervals across the dislocation core and d averaged line profile across the 12 nm region for Ru, showing the depletion at the dislocation core and slight segregation on either side. e–h EDS analysis at an edge dislocation core, presented in the same manner as (a–d). e WBDF-TEM image of a single matrix at the [001] zone axis. f Ru elemental net intensity map acquired from the surface region indicated by the red dashed box in (e). g Three-dimensional waterfall plot of line profiles at 1 nm intervals across the dislocation core and h averaged line profile across the 12 nm region for Ru, showing randomly fluctuating Ru distribution. In (a, e), the directions of g and b vectors are indicated by white and yellow arrows, respectively. The locations of the dislocation cores are indicated by orange markers. The regions where the line profiles are extracted are indicated by yellow dashed boxes in (b, f). The minimum and maximum values of the color scale in (c, g) correspond to the Ru atomic fraction range of each plot. Figure 3c shows line profiles of Ru concentration, demonstrating consistent depletion along dislocation cores and slight side segregation. The averaged profile in Fig. 3d further confirms the contrast between the dislocation core and the surrounding matrix. Notably, no such features are observed in nearby matrix regions (see Supplementary Fig. 12). These features are also depicted in the line profiles shown in Supplementary Fig. 13. No other significant elemental changes were observed, including the concentration of yttrium. In contrast, for the edge dislocation, a decoupled distribution of the Ru signal from the dislocation line was observed (Fig. 3e–h and Supplementary Fig. 14). These observations suggest that Ru not only plays a key role in forming dislocations, but its redistribution and exsolution are correlated with mixed-type dislocations. Quantification of EDS mapping shows relatively small changes in concentration compared to the matrix average. While EDS signals may fluctuate at the pixel level due to the stochastic nature of spectroscopy, the recurring Ru enrichment and depletion profiles across the dislocation core—coherently aligned along mixed dislocation lines (i.e., extended 1D depletion lines)—indicate a genuine compositional variation within the matrix. These features are consistently observed in mixed dislocations, in contrast to the absence of discernible patterns or directionality in edge dislocations (see Supplementary Fig. 15). Since the interaction volume of the dislocation core is confined to a few nanometers, whereas the matrix thickness extends over tens of nanometers, the actual concentration changes at the dislocation core would be underestimated. Thus, the extent of segregation and depletion could be even more significant than observed. Across numerous cases, the selective formation of Ru exsolution at the ends of mixed dislocations suggests that dislocations may serve as nucleation sites for Ru exsolution. While one could hypothesize that Ru exsolution induces dislocations, this would not naturally explain why only mixed dislocations are formed, while edge and screw dislocations are absent, nor the systematic Ru redistribution observed along them. More importantly, in situ BCDI demonstrated that no dislocations propagated from the surface toward the bulk; instead, they emerged from the interior. TEM results also corroborate these observations taken at 800 °C with a shorter exposure time of 1 h (Supplementary Fig. 16). Nucleation and evolution of dislocations The evolution of the dislocation topologies as a function of reduction time is shown in Fig. 4. We aimed to determine where dislocation nucleation begins and how its evolution differs by type. Figure 4 shows that similar strain fields were developed at the initial stage of nucleation; however, during successive dislocation propagation, the strain was released differently depending on the dislocation types, reflecting their different growth mechanisms. Importantly, in situ BCDI resolves dislocations as three-dimensional loop structures, directly visualizing their nucleation and propagation toward the surface. While dislocations appear as line-like features in WBDF-TEM due to the projection-based and diffraction-selective nature of TEM imaging, closer inspection reveals loop-derived segments (Supplementary Fig. 17). Near-surface dislocations often manifest as partial-loop features where the three-dimensional loop intersects the free surface, whereas segments located deeper within the particle exhibit noticeably weaker contrast because of thickness-dependent dynamical diffraction effects. These complementary observations establish that the surface-intersecting dislocation segments governing Ru exsolution originate from three-dimensional loops that form in the particle interior and propagate toward the surface. In situ dislocation dynamics during the exsolution. a, b Evolution of dislocation topology during exsolution for NP 1 (a) and NP 2 (b), showing the [110] direction and each line-sense for screw and edge components of b = a2[110] dislocation. c Enlarged view of dislocation topology and strain field changes at the transient nucleation and the final stage for four b = a2[110] dislocations. Four dislocations are sorted along the dislocation type. Strain cross-sections are defined by the initial loop-containing plane at the transient nucleation state, with angle inclinations between this plane and Burgers vectors (∠loop,b) indicated. Figure 4a, b shows dislocations with intermediate stages in two NPs. The reference line-sense vectors for the edge (tedge) and screw (tscrew) dislocations are marked as normal and parallel directions to the b = a2[110], respectively. Note that pure screw dislocations were not observed in BCDI measurements, consistent with the TEM analysis trend in Fig. 2. Most dislocations nucleate in bulk with a loop topology in the transient nucleation stage instead of immediately propagating to the surface. Subsequently, they propagate toward the surface by expanding the loop upon reduction, wherein several dislocations ultimately connect to the surface at their terminations. Figure 4c shows four different dislocation types, ranging from edge-dominant to mixed-dominant, and their related strain field evolutions from transient nucleation to the final stage for the Burgers vector b = a2[110]. The dominant types of four dislocations are distinguished based on the inclination angle of the loop-containing plane (∠loop,b) from the Burgers vector in the transient nucleation stage (see “Methods” for details). By comparing the two stages, we found that (i) in the nucleation stage, the compressive and tensile strains are built up in the interior and exterior of the dislocation loops, respectively, and (ii) in the propagation stage, mixed-dominant dislocations (#3 and #4 in Fig. 4c) release these built-up strains more easily, whereas edge-dominant dislocations (#1 and #2 in Fig. 4c) retain compressive strain inside the loop throughout the evolution period. In particular, the mixed-dominant dislocations with relatively higher screw components exhibit nonplanar topological evolutions that escape from the initial loop-containing plane and finally spread the loop topology; concurrently, the internal compressive strain within the loop is released, and no strong compressive strain remains in the final stage compared with the post-nucleated stage. The observed dislocation and strain-field evolution can be understood in terms of dislocation motion and defect chemistry. When dislocation loops nucleate, an internal compressive strain field forms for all types of dislocations, suggesting that defects precipitate in the lattice as nuclei4. As these loops propagate, their net length increases through interactions with point defects in a process known as climb4,41, which facilitates the incorporation of neighboring defects into the loop, thereby extending the loop plane. Whereas climb-based dislocation movement is facile for the edge component of dislocations, it becomes difficult for the screw component when the dislocation loop becomes too large, as the screw component relies on glissile atomic rearrangement42,43. While both types of dislocation loops initially take in defects and facilitate expansion, mixed dislocations, upon reaching the mobility limit of their screw components, undergo a transition in which further intake is suppressed. Instead, screw components facilitate Ru redistribution, thereby releasing a strain field around the mixed dislocation. In the following sections, we first establish the temporal sequence of dislocation nucleation, propagation, and Ru exsolution by combining in situ BCDI and ex situ TEM observations. We then explain the mechanistic origin of Ru redistribution and the selective role of mixed dislocations, distinguishing this dynamic process from classical pipe-diffusion models and reconciling it with prior literature. Mechanistic framework for dislocation-mediated exsolution By combining in situ BCDI and TEM results, we can frame a plausible sequence in which dislocation nucleation and evolution precede Ru exsolution at the surface, occurring when a dislocation loop terminates at the surface with a mixed segment. In situ BCDI results demonstrate that dislocations nucleate in the particle interior and propagate outward toward the surface. The strong spatial correlation between mixed-dislocation terminations and Ru nanoparticles disfavors scenarios in which dislocations migrate after Ru exsolution or in which Ru exsolved elsewhere subsequently relocates to dislocation terminations. The effectively immobilized dislocation migration may involve kinetically limited or locally constrained motion within our experimental timeframe and temperature range. The latter scenario would further require an implausible assumption that mixed dislocations exhibit unequivocally higher surface energy than other dislocation types to account for such exclusive selectivity. Taken together, these observations indicate that Ru exsolution occurs after dislocation loops propagate to the surface and terminate with a mixed segment. Following this sequence, the dynamics of defect evolution under reducing conditions can be considered. Point defects would migrate to modify the dislocation core structure, driving loop expansion. Our findings (Fig. 3 and Supplementary Fig. 15) highlight Ru as the dominant element in migration and redistribution within mixed dislocations. In contrast, Ce and Ba show minimal changes, as confirmed by stable elemental distributions in EDS profiles (Supplementary Fig. 13). Although direct oxygen ion quantification is challenging due to its low atomic number, its high mobility at elevated temperatures, such as 800 °C, suggests redistribution in response to cationic dynamics, further contributing to modifications in the localized dislocation structure. Taken together, under reducing conditions, Ru defects and their corresponding oxygen defect counterparts are likely generated as Schottky defect pairs and migrate cooperatively into the dislocation loop, thereby contributing to its formation and expansion. From a kinetic perspective on exsolution, this coupled mechanism, Ru migration and dislocation loop propagation, offers a shorter route for Ru redistribution, circumventing major bottleneck stages involving cation diffusion inside the bulk22,44, where cation diffusion is typically slower than oxygen vacancy diffusion. In conventional bulk diffusion-driven exsolution, metal cations must traverse the entire oxide lattice, which requires prolonged high-temperature annealing to overcome diffusion length limitations. In contrast, our dislocation-mediated mechanism confines Ru redistribution to nanometer-scale regions near the dislocation core. As mixed dislocations propagate, sites of redistributed (segregated/depleted) Ru systematically connect to the surface. Since Ru segregation and depletion occur within a few nanometers of the dislocation core, migration is confined to short distances, eliminating the need for Ru defects to diffuse across the entire bulk over hundreds of nanometers. Instead of bulk diffusion, locally redistributed Ru is transported to the surface via mixed dislocation propagation, with the dislocation acting not as a static channel but as a mobile vehicle. Although this causes an imbalance in Schottky defect pairs to a certain extent, the disparity can be bypassed by choosing a shorter migration path to the adjacent mixed dislocations rather than to the surface, thereby accommodating more Schottky defects and fostering the formation of segregated Ru along the dislocation. This dislocation-mediated transport fundamentally differs from bulk diffusion, offering accelerated kinetics by decoupling Ru redistribution from long-range lattice diffusion. Overall, the segregated Ru induces dislocation nucleation and uses the mixed dislocation core as effective redistribution sites, carrying Ru to the surface through dislocation propagation rather than through bulk diffusion, ultimately facilitating exsolution at the surface. Critically, when a propagating dislocation arrives at the surface—particularly when exhibiting a mixed character—this terminal configuration is determinative. The termination region of mixed dislocations creates a locally favorable environment for final Ru redistribution and subsequent exsolution at the oxide–gas interface. We emphasize that the near-surface dislocation character provides the driving force for exsolution nucleation—the critical step that determines where and whether exsolution occurs. The selective role of mixed dislocations in Ru redistribution and exsolution can be rationalized through atomistic insights into the structure and functionality of different dislocation types. The observation that neither pure edge nor pure screw dislocations exhibit Ru exsolution indicates that exsolution requires the combined functionalities of both components present in mixed dislocations. In ABO perovskite oxides, including SrTiO345,46 and La1-xSrMnO347,48, edge dislocations are prone to reduction, highly probable to offer the preferable sites for Ru reduction. Once reduced Ru ions form nuclei, a B-site deficiency would develop in the neighboring area, exposing an AO plane at the dislocation core. In addition, the screw component introduces a lattice phase shift along the dislocation line, enabling the AO-plane-terminated core to transition into a BO-plane-exposed anti-phase boundary (APB)49. Such APBs, composed of rearranged edge-sharing [BO₆] octahedra, can provide a favorable structural environment for B-site cation migration. Consequently, the depletion and adjacent segregation of Ru along the mixed dislocation lines would have developed (Fig. 3a–d). Recently, APB has been proposed as a preferred site for B-site cation segregation and migration, such as Ni exsolution in perovskite oxides18. In short, neither type of dislocation alone can provide the structural conditions necessary for Ru redistribution. While edge dislocations may provide favorable sites for reduced Ru elements, they lack a mechanism to drive Ru migration without an associated screw component. Conversely, screw dislocations alone cannot function effectively because, despite their role in modifying the dislocation core structure, they would not inherently provide sites for reduced Ru incorporation. Thus, only mixed dislocations, which integrate both edge and screw components, would enable Ru redistribution and ultimately serve as nucleation sites for exsolution upon reaching the surface. Discussion We emphasize that our mechanism differs fundamentally from classical pipe diffusion models. Dislocations do not act as static, pre-existing rapid diffusion pathways. Instead, mixed dislocations are dynamically coupled with Ru redistribution and deliver redistributed Ru to surfaces. Ru accumulates preferentially near mixed dislocations due to their structural characteristics—APB formation accommodating Ru and oxygen defects. The accumulated Ru is then transported toward the surface in tandem with dislocation propagation. In other words, Ru redistribution actively participates in the propagation of mixed dislocations. This cooperative mechanism is distinguished from passive diffusion along preformed dislocation cores, which explains the localized exsolution observed exclusively at mixed dislocation termination points. In a broader context, transport along dislocations has long been debated, with studies reporting contradictory results, showing both hindered2,46,50 and enhanced51–54 defect transport. While these conflicting results vary across material systems55,56, even within the same material, transport properties can differ significantly depending on the dislocation structure or experimental conditions, such as heat treatment profiles57. In our case, the possibility of pipe diffusion cannot be excluded once the mixed dislocation segments reach the surface. Although fast pipe diffusion along dislocations remains unclear, accumulated Ru may still serve as a diffusion pathway supplying Ru atoms for exsolutions from a certain extent of depth. In addition, a recent study by Weber et al.20 reported that pre-introduced dislocations facilitate Ni exsolution in thin films. Our findings, along with this independent evidence, suggest a broader role for dislocations in exsolution processes. Notably, edge dislocations are often assumed to act as fast diffusion pathways; however, numerous studies report counterexamples in which edge dislocations do not enhance—and may even hinder—defect transport2,46,50. Our results are consistent with these observations, as edge-dominant dislocations do not facilitate Ru redistribution or exsolution. It should be noted that edge dislocations have been more extensively studied40,58–61, as they are more accessible for systematic investigations with computational modeling. In some cases, the properties of edge dislocations have been overstated as representative of all dislocations, leading to overgeneralization. However, screw and mixed dislocations remain less explored due to characterization challenges62,63. Our systematic investigation directly addresses this gap by explicitly distinguishing dislocation types under identical experimental conditions. Critically, we find that edge dislocations do not facilitate Ru transport; instead, they exhibit strong strain fields which may hinder diffusion in oxides2. In stark contrast, only mixed dislocations exhibit distinct transport behavior, underscoring the need to differentiate dislocation types when evaluating transport mechanisms. Lastly, it is worth additionally noting that several factors influence how exsolution behavior changes with the size of the parent oxide particle. In general, as the particle size decreases, the diffusion length shortens. Simultaneously, higher surface energy and increased strain may promote exsolution. However, due to the higher Peierls stress in nanoparticles, the nucleation and propagation of dislocations are relatively suppressed, potentially limiting dislocation-mediated exsolution. Determining which factor dominates requires systematic experiments and detailed analyses using well-controlled particles of various sizes, which will be crucial for future practical applications. Combining BCDI and TEM results, we determine the stages of dislocation growth and dopant exsolution in perovskite oxide, particularly the nucleation and propagation of dislocations and their relationship with Ru exsolution. In situ strain-field evolution and associated chemical changes reveal that dislocation formation is induced by dopants in the bulk and extends toward the surface. In particular, mixed dislocations are specifically correlated with Ru exsolution. The combined features of edge and screw dislocations—ease of reduction and phase shift in a lattice—are considered to provide a unique structure around the mixed dislocation core, enabling Ru transport driven by dislocation propagation toward the surface, a dynamic process distinct from static pipe diffusion. Our results provide valuable insights into the role of dislocations in oxides, catalysis, and solid-state ionics. Methods Materials synthesis The glycine nitrate method was used to synthesize the Ru-doped BaCe0.85Y0.1O3-δ matrix. For the reagent, barium nitrate (Ba(NO3)2, 99%, Sigma-Aldrich), cerium nitrate (Ce(NO3)36H2O, 99%, Sigma-Aldrich), yttrium nitrate (Y(NO3)34H2O, 99.8%, Sigma-Aldrich), and ruthenium(III) chloride trihydrate (RuCl33H2O, Sigma-Aldrich) were used. Ru-doped-BCY (BaCe0.85Y0.1Ru0.05O3-δ) was stoichiometrically dissolved in distilled water and stirred overnight. The solution was completely dissolved, and glycine (CH5NO2, 99%, neat) was added to adjust the molar ratio of glycine to nitrate to 0.55. The solution was heated at 100 °C for 2 h to ensure the evaporation of excess water. The solution was gradually poured into a preheated quartz pot at 300 °C, leading to auto-ignition64. The obtained powder was calcined at 1300 °C for 2 h. To avoid secondary-phase formation such as Ba(OH)2 and BaCO3, the synthesized powders were stored in the Ar-filled glove box maintained at H2O < 5 ppm, O2 < 3 ppm. During sample preparation for TEM/BCDI, they were stored in a vacuum desiccator. XRD confirms that the Ru-doped BaCe0.85Y0.1O3-δ particle has an orthorhombic structure (Pnma, space group 62). The minute peak near 33° may correspond to trace CeO2, but its intensity is extremely low, indicating a negligible amount (if any). To simplify peak indexing throughout the manuscript, we used cubic notation for convenience65. The exsolution process was carried out by reducing the resulting powder at a high temperature (800 °C) under a reducing atmosphere (5% H2, in Ar-balanced) for 4 h. Meanwhile, BaCe0.9Y0.1O3-δ (BCY) without Ru doping was synthesized using the same GNP process and oxidative heat treatment followed by a similar reduction process. Bragg coherent X-ray diffraction imaging In situ BCDI experiments were conducted at the 34-ID-C beamline in the Advanced Photon Source, USA. Drop-casted Ru-doped BCY NPs were illuminated by coherent X-ray with a beam energy of 9 keV and a focused beam size of approximately 600×600 nm2. CXD patterns in the vicinity of isolated Ru-doped BCY (110) Bragg peaks (i.e., Q = [110]) were collected by a Timepix detector (Amsterdam Scientific Instruments) with 55 × 55 μm2 pixel size. To acquire 3D CXD patterns, the NPs on the goniometer were rotated along the goniometer’s normal direction with increments of 0.01°, a scan range of ±0.5° or ±0.6°, and an exposure time of around 2.5 s. The projected image of 3D CXD patterns along reduction was shown in Supplementary Fig. 18. The sample to detector distance varied from 0.6 to 0.8 m to optimize for each NP size. The entire sample was placed in a gas flow cell (pure N2 at T = 800 °C, 100 ml/min continuous flow) before starting in situ measurements. During in situ measurements, the chosen NP was successively measured in a reductive environment (3.5% H2 in N2 at T = 800 °C, 100 ml/min continuous flow). For continuous 3D measurements, the lateral and angular positions of the NP were maintained by automatically aligning scans. The time intervals between two 3D measurements were approximately 5–8 min, including scan alignment. All scan parameters were maintained during successive measurements for a single NP. The reconstruction of 3D images via phase-retrieval66 was performed using the error-reduction (ER) and hybrid-input-output (HIO) algorithms. Starting with an initial support (square with a random phase), a combination of ER and HIO with a 20:180 sequence was applied for a total of 620 iterations. Guided analysis (GA) was employed to minimize uncertainty from phase retrieval (3 generations, 5 populations with the best chi-square metric). The entire process was executed on a GPU (NVIDIA TESLA K20). Reconstructed 3D images were rendered using Paraview software. The identification of dislocations was performed using the method presented by Clark et al.40. Different types of dislocations are defined by the angle between the tangential direction of the dislocation line (i.e., line-sense vector) and the Burgers vector (b = a2[110]). The line-sense vectors are determined from the loop-containing plane at the transient nucleation state, and the inclination angle from the Burgers vector of the initial loop-containing (∠loop,b) was examined. Note that the post-nucleation stage is defined as when dislocation loops form and exhibit singular points in the host matrix, and the final stage is the last measurement was taken. Transmission electron microscopy Conventional TEM and weak-beam dark field images were obtained in a field-emission transmission electron microscope (TalosTM F200X, Thermo Fisher Scientific) operated at 200 kV. The WBDF-TEM imaging was performed under a conventional g/3g condition. The directions of reflection vectors (g) and Burgers vectors (b) are indicated by white and yellow arrows in each WBDF-TEM image, respectively. The specific aperture configuration for WBDF-TEM imaging is as follows: C1 aperture (2000 μm), C2 aperture (150 μm), selected-area aperture (40 μm, corresponding to ~800 nm in the image plane), and objective aperture (10 μm, corresponding to ~2.5 mrad in the diffraction plane). For cross-confirmation of dislocation line positions, complementary STEM imaging at multiple collection angles was conducted independently (Supplementary Fig. 11). Chemical mapping with energy-dispersive X-ray spectroscopy (EDS) in scanning transmission electron microscopy (STEM) imaging mode was carried out in a Spectra Ultra microscope (Thermo Fisher Scientific) equipped with six integrated EDS detectors, as well as in a TalosTM F200X and a Themis Z (Thermo Fisher Scientific), each equipped with four integrated EDS detectors. EDS mappings were acquired with an acquisition time of 20–40 min, using a probe current of 200–400 pA (for Talos and Spectra Ultra) and 150 pA (for Themis Z). The mapping conditions in the Spectra Ultra were a dwell time of 5 μs, a pixel size of 0.1 nm, and a pixel resolution of 512 × 512. When attenuation-induced artifacts arise from shadowing effects of the grid mesh, signals from affected detector segments were excluded to preserve quantification quality67. During analysis, pre-filtering (average filter ~4 pixels) and post-filtering (Gaussian filter ~3) were applied. For line profile analysis (Supplementary Figs. 12 and 13), each step in the profiles was integrated across a nanometer-scale width, ensuring reliable quantification across the dislocation core. An averaged line profile was also obtained by combining all profiles acquired at 1 nm intervals across the 12 nm region, representing the overall compositional variation. Burgers vectors were determined using the g·b rule by examining the dislocation visibility in WBDF-TEM images in various g vectors (the detailed analyses in Supplementary Fig. 10). Note that the type of dislocation is determined in proximity to the surface, as these segments are more significant than other parts of the dislocation lines. EXAFS X-ray absorption spectroscopy and EXAFS analyses were conducted with Ru-BCY samples treated at different reduction temperatures. The experiments were conducted at 10 C and 7D beamlines of the PLS-II, Pohang, Korea. Ru K-edge spectra were obtained in the fluorescence mode. The Ru mesh was concurrently measured with the samples for calibration. Ruthenium oxide, RuO2, and powder (Ruthenium (IV) oxide, 99.9%, Sigma-Aldrich) were measured and used as a reference to obtain the S02 value to calculate the coordination number. The X-ray absorption spectra were processed and fitted using ATHENA and ARTEMIS software. Supplementary information Supplementary Information Transparent Peer Review File Source data Source Data Acknowledgements This work was supported by the National Research Foundation of Korea grant RS-2021-NR059920 and Samsung Electronics (S.C., J.K., S.K., and Hyunjung K.), the National Research Council of Science & Technology (NST) grant by the Korea government (MSIT) No. GTL24052-100 (S.Y., H.-I.J., and D.-H.K.), and Korea Institute of Science and Technology (KIST) internal research grant 26E0292 (Y.L., P.G., Hayoung K., S.Y., H.-I.J., and D.-H.K.). Experiments using PLS-II (EXAFS) were supported in part by the Ministry of Science and IT. Use of the Advanced Photon Source (34-ID-C, BCDI) was supported by the Office of Basic Energy Science under the Office of Science of the US Department of Energy (contract no. DE-AC02-06CH11357). Author contributions H.-I.J., D.-H.K., and Hyunjung K. conceived and supervised this work. S.C., J.K., S.K., R.H., W.C., and Hyunjung K. performed Bragg coherent X-ray diffraction imaging experiments, and analysis was performed by S.C. Transmission electron microscopy measurements were performed by Y.L., P.G., and D.-H.K. and analyzed by Y.L., P.G., H.S., J.R., S.C., and D.-H.K. Samples were synthesized and characterized by Hayoung K. under the supervision of S.Y. S.C., Y.L., P.G., S.Y., H.-I.J., D.-H.K., and Hyunjung K. wrote the manuscript. All authors discussed the results and commented on the manuscript. Peer review Peer review information Nature Communications thanks Christian Lenser, Moritz Weber, and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available. Data availability Raw data were generated at the Advanced Photon Source synchrotron radiation facility (34-ID-C beamline). All data supporting the findings of this study are available from the corresponding authors upon request. Source data are provided with this paper. Code availability All analysis codes are available from the corresponding authors upon request. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. These authors contributed equally: Sungwook Choi, Younghwan Lim. Contributor Information Ho-Il Ji, Email: hiji@kist.re.kr. Deok-Hwang Kwon, Email: dkwon@kist.re.kr. Hyunjung Kim, Email: hkim@sogang.ac.kr. Supplementary information The online version contains Supplementary material available at 10.1038/s41467-026-73457-7. References Associated Data Supplementary Materials Supplementary Information Transparent Peer Review File Source Data Data Availability Statement Raw data were generated at the Advanced Photon Source synchrotron radiation facility (34-ID-C beamline). All data supporting the findings of this study are available from the corresponding authors upon request. Source data are provided with this paper. All analysis codes are available from the corresponding authors upon request.