Intermediate spillover-enabled cooperation in single-atom catalysts

Abstract

The linear scaling relationships between multiple intermediates constrain catalytic performance, especially in single-atom catalysts (SACs) with isolated sites. This review focuses on overcoming this limitation by constructing SACs-based cooperative systems that exploit the directional spillover of key intermediates between functionally distinct sites. We systematically summarize how hydrogen-, oxygen-, carbon-, and nitrogen intermediates spillover between four representative cooperation paradigms, including multi-center SACs, SACs-nanoparticle hybrids, SACs-support interaction, and single-atom alloys. Through functional decoupling and site-to-site relay, these strategies reduce kinetic barriers of rate-determining steps and reconstruct reaction pathways. Beyond mitigating activity-selectivity-stability trade-offs of SACs, intermediate spillover mechanism moves catalyst design from trial-and-error approaches toward mechanism-guided engineering.

The linear scaling relationships between multiple intermediates constrain catalytic performance, especially in single-atom catalysts (SACs) with isolated sites. Here, the authors review SAC-based cooperative systems that exploit the directional spillover of key intermediates between functionally distinct sites in order to overcome this limitation.

Introduction

Heterogeneous catalysis underpins modern chemical industries, spanning sustainable synthesis, energy conversion, and environmental remediation, requiring advanced catalysts that simultaneously deliver high efficiency, selectivity, and durability under increasingly stringent economic and environmental constraints^1–3. However, the variables affecting apparent performance of catalysts are complex, underscoring shortcuts that provide valuable insights in predicting reactivity trends without accounting complexity⁴. Scaling-based descriptors, particularly Brønsted-Evans-Polanyi relationship, formalize Sabatier principle and link activation barriers to reaction energies, which can be estimated from the binding energies of key intermediates via thermodynamic scaling relations^5,6. The best catalyst should strike a balance of intermediate binding strength: strong enough to drive the reaction, yet weak enough to release the products, requiring new catalyst platforms that enable rational design of active sites at atomic-level^7–9. Single-atom catalysts (SACs), featuring isolated metal center stabilized on solid supports, bridge the conceptual gap between homogeneous and heterogeneous catalysts^10–12. By combining maximal atom utilization with well-defined coordination geometry while retaining the recyclability, SACs is considered as model catalysts to elucidate structure-performance relationships^13,14. Correspondingly, various strategies have been developed to optimize static intermediate adsorption strengths on active sites within a limited range^15–17, including coordination environment engineering¹⁸, metal-support interactions¹⁹, strain modulation²⁰, ligand regulation²¹, and heteroatom doping²². However, in complex reactions involving multiple intermediates, SACs still remain fundamentally confined to performance envelopes governed by correlated adsorption behavior²³. Due to linear scaling relationships, strengthening one elementary step often compromises another, leading intrinsic trade-offs among activity, selectivity, and stability^24–26. These limitations underscore the need for new catalyst design principles.

Advances in operando spectroscopy and atomic-resolution microscopy reveal catalyst dynamics, including surface reconstruction, composition evolution, phase transformation, and valence state regulation^27–29. In particular, the dynamic migration of intermediates redirects mechanistic understanding from static active-site models to interfacial phenomena^30,31. Spillover, defined as the dynamic migration of adsorbed intermediate between spatially distinct surface sites, has been recognized as a long-established concept and transformative mechanism in heterogeneous catalysis³². By enabling distinct sites to preferentially stabilize, migrate, or desorb intermediates, spillover spatially optimize reaction steps independently, thereby overcoming scaling-induced limitations^33,34. Spillover was first experimental evidenced in 1964, demonstrating that H atoms generated on Pt particles migrate to adjacent oxide and promote reactions at separate sites³⁵. It is now evident that spillover is not restricted to hydrogen, as oxygen-, carbon-, and nitrogen- intermediates can also migrate directionally across interfaces, thereby elevating spillover from an auxiliary effect to a central design principle³⁶.

It is worth noting that SACs are not separate from heterogeneous catalysis, but rather represent a specific class of heterogeneous catalysts in which isolated metal atoms are stabilized on solid supports. Compared with conventional nanoparticle catalysts, SACs provide atomically dispersed active centers, well-defined coordination environments, reduced ensemble complexity, and more controllable donor-acceptor configurations, making SACs particularly suitable for elucidating spillover dynamics and engineering cooperative catalytic pathways³⁷. Kiriakou demonstrated that atomically dispersed Pd in Cu(111) decouples H₂ dissociation by enabling activation at Pd followed by H spillover to Cu, innovatively resolving dissociation-adsorption trade-off in SACs-based system³⁸. Moreover, spillover pathways can be engineered by integrating SACs-based systems, including multi-center SACs, SACs-nanoparticle hybrids, SACs-support interaction, and single-atom alloys (SAAs). These systems demonstrate enhanced performance by enabling intermediates spillover between sites, thereby each elementary step occurs at its most suitable sites and correlated adsorption trade-offs are avoided on a single site³⁹.

Despite rapid progress, intermediate spillover in SAC systems is often discussed as a reaction-specific phenomenon rather than a catalyst design principle, requiring a comprehensive framework that systematically classifies spillover intermediates, cooperative systems, and mechanistic evidence^40–42. This review aims to provide a mechanism-centered overview of intermediate spillover-enabled cooperation in SACs. We first summarize and classify spillover intermediates, including hydrogen-, oxygen-, carbon-, and nitrogen- species. The synergistic impact of partnerships (multi-center SACs, SACs-nanoparticle hybrids, SACs-support interaction, and SAAs) in SACs-based systems, including their role in functional decoupling, site-to-site relay mechanisms, reaction pathway reshape, and rate-determining steps change via intermediate spillover, is emphasized. By integrating operando characterization and theoretical simulations, the direct and indirect evidence supporting the spillover mechanism is systematically examined. Finally, the remaining challenges and future opportunities in controlling intermediate spillover are discussed, highlighting how this concept can guide the rational design of next-generation SACs. Beyond simply addressing the persistent activity-selectivity-stability trilemma of SACs, framing intermediate spillover as a designable mechanism rather than a reaction-specific curiosity can catalyze a shift from empirical trial-and-error optimization to mechanism-guided catalyst engineering, offering new conceptual and practical pathways to transcend long-standing trade-offs in heterogeneous catalysis.

Fundamental concepts of spillover mechanisms

In SAC-based cooperative systems, spillover is particularly valuable because it distributes elementary steps over differentiated sites, thereby reducing the need for a single isolated site to simultaneously satisfy conflicting adsorption requirements. Spillover generally requires two conditions: (i) an energetic preference for an intermediate to reside on one site relative to another, and (ii) a sufficiently low barrier for crossing the interfacial region. When two sites display distinct adsorption thermodynamics for the same intermediate, the difference in stabilization provides a chemical potential gradient that drives migration from a higher-energy state to a lower-energy state, provided that the migration barrier is accessible under operating conditions (Fig. 1a).

Fig. 1. Driving forces and pathways of the spillover.

a The adsorption energy-gradient-driven spillover; b functional separation bridged by hydrogen spillover; c interface maximization drives direct spillover; d ultrafast interfacial spillover pathways; e energy-level buffering chain; and f division and collaboration among different components in tandem catalysts.

Beyond this primary driving force, cooperative catalytic systems introduce additional, function-based directionality. In multicomponent architectures, different domains often play differentiated roles (generation versus transformation, activation versus desorption). As a result, spillover becomes the coupling mechanism that links the generation site to the consumption site, enabling ordered concatenation of elementary steps in space (Fig. 1b). This donor-acceptor division of labor is particularly compatible with SACs, where atomic dispersion yields high interfacial density and strong interfacial interactions with surrounding components, facilitating intermediate transfer (Fig. 1c). Notably, the coexistence of multiple sites alone does not guarantee spillover. The essential driving force lies in the adsorption-energy difference between nonequivalent sites, together with a sufficiently low interfacial migration barrier. The particular value of SACs is therefore not that they uniquely create spillover, but that they provide atomically defined donor-acceptor configurations that make spillover more controllable and mechanistically clearer.

From a structural perspective, spillover in SAC-based systems can be enabled through several archetypal channels, which can be viewed as progressively engineered versions of the donor-channel-acceptor motif. In tightly coupled multi-site systems, intermediates can transfer rapidly through ultrashort interfacial pathways between neighboring functional centers. Here, site proximity and adsorption-energy gradients jointly define a natural migration outcome, and the spillover rate can approach the timescale of intermediate formation and consumption (Fig. 1d). When a direct transfer is kinetically hindered, constructing an “energy buffer chain” can create intermediate adsorption states that smooth the energetic transition, enabling continuous spillover along a sequence of sites with gradually tuned adsorption strengths. In practice, this can be realized by controlled gradients in coordination, defect density, or local electronic structure across the interface (Fig. 1e). While in tandem systems, distinct components specialize in intermediate generation, transport, and transformation. Spillover becomes the cooperative bridge that converts a set of individual catalysts into a coherent reaction route, frequently enabling higher selectivity by minimizing parasitic reactions at the generation site and maximizing coupling efficiency at the transformation site (Fig. 1f). Overall, these channels highlight the same core concept: spillover dynamically modulates surface active sites, enables spatial and functional decoupling of multistep reactions, and redistributes kinetic barriers across the pathway, thereby jointly improving activity, selectivity, and stability.

Spillover of different intermediate species

According to the chemical identity, spillover intermediates can be broadly classified as hydrogen, oxygen (*O, *OH, *OOH), carbon (*CO, *COOH, *CH_x), and nitrogen (*N and *NO₂) intermediates. Importantly, these intermediates serve dual roles during catalysis, acting as dynamic carriers of reactive equivalents and surface regulators that reshape local microenvironments through competitive adsorption, coverage redistribution, and interfacial charge modulation. In this view, spillover is not merely a transport event but an integrated mechanism that reorganizes the catalytic landscape in space and time.

Hydrogen spillover

Hydrogen spillover involves the movement of hydrogen species that are adsorbed between various surfaces or active sites, transferring from regions abundant in hydrogen to those lacking it^43,44. Hydrogen spillover has been widely exploited in heterogeneous hydrogenation and hydrogen storage applications⁴⁵.

In reactions involving H₂ activation or proton reduction, metal sites with strong hydrogen adsorption energies are generally favorable for hydrogen dissociation and activation, yet they are also prone to so-called “hydrogen poisoning” due to excessive hydrogen adsorption, which suppresses subsequent reaction steps. Hydrogen spillover can alleviate this limitation by transferring over-adsorbed hydrogen species away from the active sites to adjacent sites or supports, thereby enabling dynamic regeneration of active sites and restoring their reactivity. For instance, in the semi-hydrogenation of acetylene, hydrogen atoms dissociated on Pd sites can spill over to neighboring Mg sites. In this system, Mg acts as a hydrogen reservoir, effectively reducing hydrogen coverage on Pd surfaces and allowing Pd sites to continuously participate in reactant adsorption and hydrogenation, leading to significantly enhanced activity and selectivity⁴⁶. A similar mechanism operates in the Pt SA/m-WO_3-x system, where maximization of the Pt-W interfacial area creates ultrashort hydrogen spillover pathways. Strongly adsorbed hydrogen species Pt-H on Pt single-atom sites can efficiently spill over and inject into the WO_3-x support, forming hydrogen tungsten bronze (H_xWO₃). The intercalated hydrogen subsequently undergoes long-range migration within the WO₃ lattice in the form of H_xWO₃, thereby modulating the overall hydrogen insertion and extraction kinetics of the system. During hydrogen evolution reaction (HER), this process not only maintains continuous exposure of Pt active sites but also accelerates H₂ desorption, effectively alleviating the adsorption desorption trade-off described by the classical Sabatier principle. Comparative studies further reveal that, relative to Pt nanoparticle/WO₃ systems, Pt SA/m-WO_3-x exhibits markedly enhanced hydrogen insertion/extraction kinetics due to interfacial maximization and shortened spillover distances, with hydrogen transport dominated by direct interfacial spillover (Fig. 2a)⁴⁷.

Fig. 2. Spillover of hydrogen intermediates.

a Illustration the *H spillover in HER⁴⁷. Copyright 2019, Wiley. b Color change of hydrogen spillover effect⁴⁸. Copyright 2025, Elsevier. c Schematic of urea synthesis enhanced by hydrogen spillover⁴⁹. Copyright 2025, Elsevier. d In situ DRIFTS spectra of DMC adsorption⁵⁰. Copyright 2025, American Chemical Society. e Schematic illustration of the rational design of hybrid Ru catalyst⁵¹. Copyright 2025, Wiley. f H₂-TPD profiles⁵². Copyright 2025, American Chemical Society. g In situ EPR spectra⁵³. Copyright 2025, American Chemical Society. h TOF-SIMS signals and representative TOF-SIMS images of Ru_1+NPCe₂/GC in term of N₂H, N₂H₂, N₂H₃, and N₂H₄ positive ions⁵⁴. Copyright 2025, American Chemical Society. i In situ Raman spectra of Pt₁@PMo₁₂@PC⁵⁷. Copyright 2024, Elsevier.

Beyond interfacial structural effects, the driving force for hydrogen spillover in certain catalytic systems is also closely associated with local electronic effects. Mismatches in hydrogen adsorption energies arising from electronic structure differences between distinct metal sites can provide a thermodynamic driving force for directional hydrogen migration. In the semi-hydrogenation of alkynols, density functional theory calculations of hydrogen adsorption energies at different sites indicate that electronic effects between Pd single atoms and the Cu matrix generate a substantial adsorption energy gradient, which drives direct H* spillover from Pd sites to Cu sites. In this process, H₂O dissociates on Pd single-atom sites to generate active hydrogen, which subsequently migrates to Cu sites via hydrogen spillover, effectively preventing excessive H* accumulation on Pd. As a result, Pd sites remain continuously available for reactant adsorption and hydrogenation, leading to markedly improved efficiency and selectivity in propargyl alcohol semi-hydrogenation with obvious color change (Fig. 2b)⁴⁸.

Besides, the hydrogenation steps always limited by insufficient hydrogen supply because the reactant activation sites themselves exhibit weak hydrogen adsorption, dissociation, or delivery capabilities. Hydrogen spillover acts as a directional hydrogen supply mechanism, continuously transporting hydrogen species from hydrogen-rich generation sites to hydrogen-deficient reaction sites, thereby driving efficient catalytic turnover. A representative example is electrocatalytic urea synthesis over Ru single atoms anchored on ZrO₂ (Ru₁-ZrO₂/C), where hydrogen spillover constructs a sub-nanometer proton transport pathway described as “Ru to O to Zr”. Hydrogen atoms generated via H₂O dissociation on Ru single-atom sites migrate to adjacent Zr sites through spillover, providing a sustained hydrogen source for C-N coupling and accelerating subsequent hydrogenation steps, while simultaneously suppressing the competing hydrogen evolution reaction (Fig. 2c)⁴⁹. Similar directional hydrogen supply mechanisms have been applied in dimethyl carbonate (DMC) hydrogenation. In the Pt₁-Cu/WO₃ system, density functional theory calculations show that H₂ can spontaneously adsorb and dissociate on Pt₁ sites, with the generated H* subsequently spilling over and being temporarily stored in the WO₃ support. Because the adsorption energy of H* on Cu₅ clusters is substantially more favorable than that on Pt₁ sites, this adsorption energy difference provides a thermodynamic driving force for H* migration from the WO₃ surface to Cu clusters. Operando diffused reflectance infrared Fourier transform spectroscopy (DRIFTS) measurements further indicate the emergence of additional hydrogen-related surface species upon incorporation of Cu nanoparticles, ensuring sustained hydrogen supply even under low H₂ partial pressures and resulting in significantly enhanced catalytic activity and selectivity (Fig. 2d)⁵⁰. In photothermal CO₂ hydrogenation, H₂ dissociates on Ru nanoclusters, and the generated H* spills over to the hNCNC support before further migrating to Ru single-atom sites. This continuous hydrogen supply promotes CO₂ hydrogenation to CO while suppressing methanation, thereby alleviating the activity-selectivity trade-off (Fig. 2e)⁵¹. As shown in Fig. 2f, H₂ temperature programmed desorption (H₂-TPD) demonstrate that H₂ dissociates on Rh nanoparticles and the resulting H* spills over to the CeO₂ support in a dual-catalyst Rh-CeO₂/CeHY system for low-density polyethylene hydrogenolysis, where it hydrogenates olefinic intermediates adsorbed on Rh single-atom sites. This process accelerates key hydrogenation steps and enables efficient conversion of waste plastics into liquid fuels⁵².

In tandem catalytic architectures, hydrogen generation and consumption are assigned to distinct active sites, thereby enabling spatial separation of elementary reaction steps. For example, in the Al₂O₃/Al₁-O-C tandem catalyst, in situ electron spin resonance demonstrated that Al₂O₃ domains act as hydrogen generation and supply centers, while Al single-atom sites serve as the catalytic centers for H₂O₂ formation. By confining hydrogen production and H₂O₂ synthesis to different locations, this architecture alleviates hydrogen supply limitations under high current densities, leading to markedly enhanced H₂O₂ yield, Faradaic efficiency and catalyst stability (Fig. 2g)⁵³. For mild-condition ammonia synthesis, a tandem catalyst composed of Ru nanoparticles and Ru single atoms (Ru₁) enables hydrogen intermediates to migrate from RuNP sites to Ru₁ sites, effectively separating H₂ and N₂ activation. With increasing hydrogen coverage proved by Time of flight secondary ion mass spectrometry (TOF-SIMS, Fig. 2h), the hydrogen adsorption strength on Ru clusters decreases and the corresponding spillover barrier is reduced, which relieves hydrogen poisoning on RuNP sites and promotes N₂ dissociation as well as subsequent hydrogenation on Ru₁ sites⁵⁴. An analogous design concept has also been extended to Ru-Pt tandem catalysts, in which H₂ dissociation and H* desorption during the hydrogen oxidation reaction are spatially decoupled, resulting in substantially accelerated reaction kinetics⁵⁵. A single-atom iron catalyst of Fe-N₃S₁, sulfur-induced asymmetric charge distribution substantially accelerates proton spillover, alleviating the sluggish water dissociation kinetics characteristic of conventional Fe-N₄-C structures. This effect promotes rapid formation of OOH* intermediates and facilitates H₂O₂ desorption, leading to significantly improved efficiency and selectivity in electrochemical H₂O₂ synthesis⁵⁶.

Constructing multilevel buffer chains or ultrashort migration pathways represents an effective strategy to decompose a single high-energy migration step into multiple low-barrier processes. Such structural engineering approaches provide powerful means to accelerate cross-scale hydrogen transport. In the Pt@POMs@PC system, a multi-shell hydrogen buffer chain with a continuous spillover pathway described as “Pt to O_br to O₃H to Mo/W to O_c to PC_sub-1 nm” was revealed by in situ Raman spectra. This design decomposes a high single-step migration barrier into multiple steps with barriers below 0.2 eV, substantially accelerating overall hydrogen transport and enhancing HER activity and stability (Fig. 2i)⁵⁷. Similarly, in ordered intermetallic PtZn, introduction of local disorder creates an ultrashort spillover pathway (Pt site to Pt-Zn bridge to Zn site), where the thermoneutral Pt-Zn bridge acts as a migration buffer, reducing the hydrogen migration barrier to approximately 0.1 eV and effectively mitigating the intrinsically high hydrogen overpotential of Zn components⁵⁸.

Oxygen-based intermediate spillover

Beyond hydrogen, a diverse range of oxygen-containing intermediates is involved in catalysis, including adsorbed atomic oxygen (*O), hydroxyl (*OH), hydroperoxyl (*OOH), and electronically enriched active oxygen species (such as O^-/O₂^-). Oxygen intermediate spillover plays a crucial role and directly participate in key reaction steps in catalysis, arising from the migration and redistribution of reactive oxygen species across interfaces.

In the oxygen evolution reaction (OER), oxygen-intermediate spillover has been demonstrated to significantly enhance both the activity and stability of catalysts. For example, in a RuO₂/MoO₃ catalytic system, a new OER reaction pathway under acidic conditions has been proposed (Fig. 3a). Water molecules adsorbed on the RuO₂ surface undergo stepwise deprotonation to form *OH and then to *O species. The *O then migrates to MoO₃, where it participates in lattice-oxygen-related O-O coupling to generate O₂. Meanwhile, another water molecule occupies the newly formed oxygen vacancy and regenerates *OH, which is further deprotonated to replenish lattice oxygen, thereby enabling a sustained catalytic cycle. This process effectively suppresses the over-oxidation and dissolution of Ru species, markedly reducing the formation of high-valence Ru states and consequently improving both catalytic activity and durability.

Fig. 3. Spillover of oxygen-based intermediates.

a Schematic illustration of the proposed oxygen spillover in the RuO₂/MoO₃ catalysts. Green, blue, red and white balls represent Ru, Mo, O and H atoms, respectively. b Chronopotentiometry tests of RuO₂/MoO₃, RuO₂, and c-RuO₂⁵⁹. Copyright 2024, Royal Society of Chemistry. c Proposed photocatalytic reaction pathway illustrate synergistic charge transfer and active-site mechanisms⁶⁰. Copyright 2024, Elsevier. The Fourier transformed frequency-resolved complex conductivity from the THZ time domain spectroscopy of d pure SnO₂ and e as-prepared SnO₂/Pd_atom, the solid lines correspond to the fittings with Drude-Smith models⁶¹. Copyright 2025, Springer Nature. f SEIRA spectra of Zn₁/NOC in Ar-saturated 1.0 M KOH to show molecular-level proof of *OH enrichment⁶² Copyright 2025, Wiley. In situ Raman spectra of g Ru-Ni(OH)₂ and h R-Ru-Ni(OH)₂ in 1 M KOH⁶⁴. Copyright 2025, American Chemical Society. i Illustration of spillover break the linear scaling limit⁶⁶. Copyright 2025, Springer Nature. j Operando Raman spectra peak intensities⁶⁷. Copyright 2025, Wiley.

A comparison of oxygen-vacancy concentrations before and after stability tests indicates that the vacancy level remains essentially unchanged during OER operation (Fig. 3b). Cyclic voltammetry reveals two oxidation peaks at approximately 0.62 and 1.23 V for RuO₂ and c-RuO₂, corresponding to the Ru³⁺/Ru⁴⁺ and Ru⁴⁺/Ru⁶⁺ redox couples, respectively, which involve *OH and *O intermediates. Upon incorporation of MoO₃, the Ru⁴⁺/Ru⁶⁺ oxidation peak in the RuO₂/MoO₃ system is significantly attenuated, further demonstrating that the migration of *O-related intermediates effectively prevent excessive oxidation of Ru species to high-valence states during OER⁵⁹.

In addition, oxygen-based intermediate spillover can directly participate in the activation of inert C-H bonds. For instance, in the oxidative coupling of methane, a dual-site-engineered Au/Sr–ZnO photocatalyst enables efficient lattice-oxygen activation and regulation of methyl transfer. This system follows a Mars-van Krevelen mechanism, in which the dynamic formation and replenishment of oxygen vacancies promote the participation of lattice oxygen in C-H bond cleavage and enhance CH₄ adsorption, significantly improve ethane yield and catalytic durability (Fig. 3c)⁶⁰. For gas sensing, the migration of reactive oxygen species (O^-) is likewise recognized as a key mechanism for enhancing sensor performance. Studies have shown that during H₂ detection using SnO₂/Pd_atom materials, O^- species generated at Pd single-atom sites can migrate onto the SnO₂ surface. The coupling and hybridization between Pd 4d orbitals and O 2p orbitals facilitate electron transfer from Pd 4d to O 2p states can increase the concentration of reactive O^- species on the sensor surface. The migrated O^- reacts with H₂ to form water while releasing electrons, which substantially increases the electron density in the material and lowers the device resistance. Moreover, the terahertz time domain spectroscopy (THz TDS) measurements prove that an extra energy band near Fermi level appeared in SnO₂/Pd_atom, and Pd SAs doped on SnO₂ enhance signally concentration of free carrier in SnO₂/Pd_atom (Fig. 3d, e). As a result, SnO₂/Pd_atom sensors exhibit ultrafast response times, excellent sensitivity and selectivity⁶¹.

For ammonia oxidation, a bifunctional catalyst was designed in which atomically dispersed Zn₁/NOC sites serve as efficient *OH adsorption centers. In situ surface-enhanced infrared absorption (SEIRA) spectroscopy reveals that the intensity of the O-H stretching vibration (~3400 cm⁻¹) on the Zn-carbon catalyst increases by approximately 120% between 0.4 and 0.7 V vs. RHE, indicating progressively enhanced *OH adsorption at more positive potentials (Fig. 3f). The *OH species adsorbed on Zn₁ sites subsequently migrate to adjacent Pt nanoparticles, facilitating NH₃ dehydrogenation while simultaneously alleviating the poisoning of Pt active sites by toxic intermediates⁶². In alkaline hydrogen evolution reactions, *OH species generated from water dissociation can severely poison active sites if not promptly removed. By constructing heterostructures such as Ru-Sn-O/SnO₂, *OH species can spontaneously spill over from metal active sites to oxophilic supports, thereby regenerating active sites and enabling continuous water dissociation⁶³. In Ru-Ni(OH)₂ core shell structures, hydroxyl species transfer from the Ni(OH)₂ shell to Ru active sites to form Ru-OH_ad. The Ru-OH_ad species directly participate in the reaction via hydroxyl-interfacial water proton exchange mechanism, enhancing electron transfer at Ru active sites, optimizing reaction pathways, and lowering the high energy barrier associated with *OH desorption (Fig. 3g, h)⁶⁴. Recent studies also revealed the formation and stabilization of oxygen vacancies in the near-surface region of CeO₂ supports and highlighted their critical role in regulating the anchoring and spreading behavior of Ru nanoparticles. Vacancy-induced local electron enrichment weakens surface O-H bond strength, thereby facilitating hydroxyl-related spillover and redistribution on the CeO₂ surface⁶⁵.

The adsorption state and interfacial behavior of *OOH intermediates play a decisive role in governing the selectivity and kinetics of oxygen reduction reactions (ORR). In conventional 4e^- ORR pathways, excessively strong *OOH adsorption can hinder subsequent O-O bond cleavage and proton-electron transfer steps, thereby slowing reaction kinetics and even inducing undesired H₂O₂ formation via side pathways. Through spillover effects, *OOH intermediates can migrate from strong-binding sites (e.g., Pd) to adjacent weak-binding sites (e.g., Ag), facilitating their regulated transformation or desorption. This mechanism enables Pd/Ag heterostructures to overcome the intrinsic volcano-plot limitations of Pd and achieve superior ORR performance (Fig. 3i)⁶⁶. In contrast, in the 2e^- ORR pathway, *OOH serves as the direct precursor of the target product H₂O₂, making its formation and stabilization critically important. In F-Cu-MOF systems, water spillover provides an abundant local proton source to promote *OOH generation, while strong F-Cu interactions effectively stabilize the *OOH intermediate. Raman spectroscopy reveals relatively weak OOH-related vibrational signals, indicative of moderated OOH binding strength, which favors efficient desorption and accumulation of H₂O₂ (Fig. 3j). As a result, high Faradaic efficiencies for H₂O₂ production can be achieved over a wide current density range⁶⁷. Beyond activity and selectivity modulation, spillover of oxygenated species also plays a critical role in electrocatalytic stability. In PtFe/Fe-N-C composite catalysts, H₂O₂ generated at Fe-N-C sites can spontaneously spill over to neighboring PtFe alloy nanoparticles, where it is rapidly decomposed. This spillover-enabled internal purification effectively removes harmful oxygenated intermediates from vulnerable Fe-N-C sites, fundamentally suppressing Fenton reactions and significantly enhancing the long-term durability of fuel cell catalysts⁶⁸.

Carbon-based intermediate spillover

The migration of key carbon-based intermediates (e.g., *CO) is widely regarded as one of the important physical foundations for efficient electrochemical CO₂ reduction reaction (CO₂RR). This intersite intermediate transfer allows CO₂ activation and subsequent C-C coupling to be spatially decoupled while being functionally integrated at the system level, ultimately promoting the formation of C₂₊ products.

The *CO spillover mechanism is centered on the efficient reduction of CO₂ to *CO at donor sites, followed by the migration of the generated *CO intermediates to adjacent acceptor sites, where subsequent C-C coupling reactions take place. For example, Chala et al. developed a Co-Cu dual-SAC, in which Co sites are responsible for efficient CO generation, while neighboring Cu sites serve as the active centers for C-C coupling (Fig. 4a). In this system, *CO migrates from Co sites to adjacent Cu sites, enabling a high ethanol Faradaic efficiency of up to 71% at −0.8 V vs. RHE (Fig. 4b). Mechanistic investigations revealed that the high local CO concentration established at the Co sites plays a crucial role in driving *CO transfer toward Cu sites and ultimately facilitating ethanol formation⁶⁹.

Fig. 4. Spillover of carbon-based intermediates.

a Proposed CO spillover mechanism for the efficient ethanol formation at a heterobimetallic CoCu electrocatalyst, b corresponding FEs and product distribution at different applied potentials⁶⁹. Copyright 2024, American Chemical Society. c A proposed *CO spillover catalytic mechanism of Cu NPs+Ni-SOD/NC⁷¹. Copyright 2023, Wiley. d In situ SERS spectra of Cu-CB, Co-CB, and CuCo-DSAC as a function of potential. e The following schematic illustrates the CO₂ reduction pathways for the copper (yellow rails) and cobalt sites (purple rails)⁷². Copyright 2025, Wiley. f Asymmetric *CH₂-CO coupling process to form ethanol at a high local CO concentration on the layered Cu/Ag tandem catalyst⁷³. Copyright 2024, American Chemical Society. g Illustration of reaction pathway for covalorization of CH₄ and N₂O⁷⁴. Copyright 2025, American Chemical Society.

Similarly, Zheng et al. combined atomically precise Cu₃₂ nanoclusters with a nickel phthalocyanine covalent organic framework (NiPc-COF). As shown in Fig. 4c, the high CO concentration generated within the pores of the NiPc-COF promotes *CO migration toward the Cu₃₂ nanoclusters, leading to an increase in the total Faradaic efficiency for C₂₊ products from 24.4% on pristine Cu₃₂ to 57.1% in the hybrid system⁷⁰. In another representative example, Chen et al. constructed a tandem catalyst composed of Ni single atoms (NiN₃) and Cu nanoparticles. The NiN₃ sites exhibit excellent CO production activity (FE_CO > 96.5%), and the abundant CO generated is subsequently transferred to the Cu nanoparticle surfaces, resulting in an enhanced ethylene Faradaic efficiency of 62.5% and a substantially broadened potential window for efficient ethylene production⁷¹. Collectively, these studies demonstrate that *CO spillover represents an effective strategy for achieving highly selective C₂₊ product formation under high current density and industrially relevant CO₂RR conditions.

*COOH is a key intermediate in CO formation, which is typically generated via consecutive proton-electron transfer (PECT) steps following the adsorption of CO₂ on the catalyst surface. In situ surface-enhanced Raman spectroscopy provides strong evidence for the formation of this intermediate. Subsequently, *COOH undergoes further proton-electron transfer accompanied by C-O bond cleavage to yield CO, which then desorbs from the catalyst surface (Fig. 4d). In the CuCo-DSAC catalyst, analysis of the free-energy evolution of reaction intermediates during CO₂RR further supports this reaction pathway. The calculations indicate that Co sites preferentially stabilize the *COOH intermediate, while adjacent Cu sites reduce the energy barrier of subsequent reaction steps through interfacial charge redistribution, thereby facilitating CO formation and desorption. Such site-to-site cooperative regulation accelerates the conversion of *COOH and prevents its excessive accumulation at the donor sites, effectively alleviating active-site blocking while maintaining high CO selectivity (Fig. 4e)⁷².

Beyond CO₂RR, carbon-based intermediates also play crucial roles in a broad range of reactions. For instance, in Cu/Ag tandem catalysts, Cu(111) facets preferentially generate reduced carbonaceous intermediates such as *CH₂, while the Ag component provides a high local CO coverage. The interfacial synergy between these two functionalities promotes asymmetric *CH₂-CO coupling in the vicinity of the Cu-Ag interface, thereby facilitating ethanol formation (Fig. 4f)⁷³. In another representative system, Ir-Gd₂O₃ SACs have been employed to investigate the covalent coupling of methane with N₂O. Methane is activated at isolated Ir sites and undergoes stepwise dehydrogenation to form key carbon-based intermediates, including CH₃, CH₂, and CH₂O species. Among them, *CH₂O has been identified by in situ DRIFTS as a critical reaction intermediate. This species can undergo interfacial migration or transfer from the Ir sites to the Gd₂O₃ support surface, where C-O bond formation and subsequent CO generation and desorption occur (Fig. 4g)⁷⁴. Moreover, Zhang et al. demonstrated the presence of Mo single atoms on the MWCNTs does not alter the electronic state of Pt nanoparticles but enhances their catalytic efficiency by increasing benzene coverage via benzene spillover from Mo single atoms to Pt nanoparticles, leading to a higher energy barrier for benzene hydrogenation⁷⁵.

Nitrogen-based intermediate spillover

Nitrogen-related intermediates are ubiquitous in catalytic processes mainly involving N₂ activation, nitrogen conversion, and the synthesis of nitrogen-containing compounds. In plasma-assisted ammonia synthesis, the reaction mechanism involves a complex interplay between gas-phase plasma chemistry and surface catalytic processes, where reactive nitrogen (*N) and hydrogen species undergo adsorption, dissociation, and recombination on catalyst surfaces⁷⁶.

A representative example of N-intermediate spillover was recently reported for an Fe-Al dual-site catalyst in photocatalytic ammonia synthesis. Fe serves as the N₂ adsorption and dissociation site, while adjacent Al sites, which bind nitrogen weaker, accept the dissociated *N through spillover. Operando Raman (Fig. 5a) and NAP-XPS (Fig. 5b) spectra showed the decrease of Fe-N species and the appearance of Al-N species under illumination, supporting *N spillover from Fe to Al. Because *N is less strongly bound on Al sites, the subsequent NH_x formation and NH₃ desorption are facilitated, thereby spatially decoupling N₂ cleavage from product release⁷⁷. In addition, studies on Pt@Cu₃(HHTP)₂ thin films have revealed that *NO₂ molecules can migrate from ultrasmall Pt nanoparticles to the Cu₃(HHTP)₂ framework, where they are further activated, thereby accelerating reaction kinetics and improving sensing performance (Fig. 5c)⁷⁸. In fluorine- and phosphorus-co-doped cobalt catalysts, the introduction of F and P generates adjacent active sites with distinct functionalities, forming heteronuclear dual-center motifs for tandem catalysis. The F-Co sites preferentially adsorb and activate NO₃^-, facilitating its deoxygenation to NO₂^- intermediates, while neighboring P–Co sites promote subsequent hydrogenation to NH₃ and simultaneously enhance water dissociation to ensure sufficient hydrogen supply (Fig. 5d). This cooperative mechanism integrates the complementary advantages of F and P, enabling efficient conversion of each intermediate step and optimizing both catalytic activity and selectivity (Fig. 5e, f)⁷⁹.

Fig. 5. Spillover of nitrogen-based intermediates.

a Raman spectra and b Al 2p XPS spectra of N₂ pretreated AlFe/Al under dark and light conditions⁷⁷. Copyright 2024, Springer Nature. c Calculated adsorption rate constants of the sensors⁷⁸. Copyright 2021, Springer Nature. d Schematic illustrations of the NO₃RR pathways of direct conversion mechanism and *NO₂ spillover mechanism, and the calculated Gibbs free energy diagrams for NO₃RR on the F, P-Co. e LSV curves of F, P-Co, P-Co, F-Co and Co in 0.1 M K₂SO₄ + 10 mM KNO₃ electrolyte. f Stability test of F, P-Co⁷⁹. Copyright 2025, Wiley.

SACs-based spillover cooperation systems

The construction of SACs-based catalytic cooperation systems to overcome the intrinsic limitations imposed by scaling relationships by intermediate spillover were systematically discussed, including four representative synergistic paradigms: (i) multi-center configurations within SACs (Fig. 6a), (ii) hybrid systems composed of SACs and nanoparticles or clusters (Fig. 6b), (iii) synergistic interactions between SACs and metal sites or vacancies in the support (Fig. 6c), and (iv) SAAs systems (Fig. 6d). It is worth noting that SACs and bulk materials are co-existed in system ii to iv. The key distinction lies in the spatial and compositional relationships. In systems ii, SACs and nanoparticles/clusters are all anchored on a support; in system iii, spillover occurred between SACs and support; while system iv is atomically dispersed heterometallic alloys within a host metal matrix. Through the reconstruction of the local reaction microenvironment and the decoupling of adsorption energetics, such synergistic systems ultimately enable the concurrent enhancement of catalytic activity, selectivity, and stability.

Fig. 6. Spillover in SACs-based cooperation paradigms.

a multi-center SACs, b SACs-nanoparticle hybrids, c SACs-support interaction, and d single-atom alloys.

multi-center SACs

Constructing multi-center architectures, in which active sites cooperate through functional specialization and site coupling, has emerged as an effective strategy to enable reaction pathway reorganization and reaction barrier decomposition, leading to enhanced performance⁸⁰. This cooperation is realized by engineering microenvironments of SAC, where chemically distinct atoms jointly participate in intermediate behavior. Multicenter microenvironments are not restricted to strictly bonded dual-atom motifs, but broadly encompass composite atomic configurations formed by multiple proximate active sites within the local coordination sphere of SACs. By modulating microenvironments, the superior catalytic activity is enabled through spillover-mediated synergy (Fig. 6a).

Engineering heteronuclear configurations provides effective route to promote intermediate spillover. A related strategy was demonstrated in a dual-metal SAC in which Cu and Mg atoms were simultaneously anchored on a nitrogen-doped carbon support to form a quasi-heteronuclear configuration, as proved by high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM). Atomically dispersed Mg sites not only modulate the electronic structure of Cu sites to lower the overall energy barrier of the CO₂RR to CH₄, but also efficiently activate CO₂ to form *CO intermediates, which subsequently migrate to adjacent Cu sites to further hydrogenate into CH₄ with high Faradaic efficiency of 78.3% (Fig. 7a, b)⁸¹.

Fig. 7. Multi-center SACs systems.

a Dual-site (CuN₂-MgN₂) for CO spillover during CO₂ reduction to CH₄. b product Faradaic efficiency of CuN₂-MgN₂ at different potentials⁸¹. Copyright 2025, Wiley. c Spillover pathway and d atomic-resolution HAADF-STEM images of Ni₁-S₆/Ni₁-Mo₂ catalysts⁸². Copyright 2025, American Chemical Society. e Atomic-scale EDS mappings show the dispersion of Cu and Co atoms in CuCo-DSAC⁷². Copyright 2025, Wiley. f AC-HAADF-STEM images and corresponding intensity maps of PtRu@W₂N_DF@NC. XANES and EXAFS spectra of g Pt L-edge and h Ru L-edge of PtRu@W₂N_DF@NC. i Top and side views of the HER mechanism on PtRu@W₂N_DF@NC, including H₂O adsorption, H₂O dissociation, and H₂ formation⁸³. Copyright 2025, Wiley.

Beyond regulating pathways, multicenter can indirectly promote spillover. For example, a dual-coordination single-atom nickel catalyst, denoted as Ni₁-S₆/Ni₁-Mo₂, achieves the simultaneous enhancement of activity, selectivity, and stability in selective acetylene hydrogenation. In this system, synergistic dual-center cooperation is established: the Ni₁-Mo₂ sites exhibit stronger H₂ adsorption, activation, and dissociation, continuously supplying hydrogen species for downstream hydrogenation steps. The generated hydrogen atoms subsequently spill over via sulfur species to acetylene adsorbed on Ni₁-S₆ sites, where over-hydrogenation and undesired C-C coupling are effectively suppressed (Fig. 7c, d)⁸².

Furthermore, theoretical studies on Co-Cu dual-center catalysts reveal that electronic interactions between Co and Cu modulate their respective d-band centers, rendering Co sites favorable for stabilizing COOH* intermediates while weakening CO adsorption on Cu sites (Fig. 7e). This electronic differentiation provides the fundamental driving force for COOH* spillover from Co to Cu, underpinning the observed synergistic catalytic behavior⁷². Lu et al. developed a space guarding strategy to precisely place Pt-Ru dual atom pair sites confined in W₂N lattice for efficient hydrogen evolution with high mass activity of 72.7 A mg_(Pt+Ru)^-1 in H₂SO₄. The Pt-Ru pair was first proved by AC-HAADF, where Pt atom shows a higher Z-contrast and Ru atom shows a lower Z-contrast compared to the neighboring W elements (Fig. 7f). The X-ray absorption near edge structure (XANES) and extended X-ray absorption fine structure (EXAFS) also suggest that the Pt-Ru pairs tend to load in the W₂N lattice defects rather than on the NC support, much less form clusters (Fig. 7g, h). Mechanism demonstrated that the overall hydrogen evolution pathway proceeds along three steps: fast H₂O adsorption on W site, facile H* and OH* respective migration from W site to Pt and Ru sites via its distinct electronic flows, and favorable H₂ desorption on Pt site (Fig. 7i)⁸³.

A recent study on an out-of-plane Fe-Co dual-atom catalyst provides a more explicit example of how oxygen-intermediate spillover can relax the conventional single-site ORR limitation. The curved Fe-Co dual-site configuration enhances the d-p orbital hybridization between the metal centers and oxygen-containing intermediates, thereby strengthening *OOH adsorption and accelerating the first proton-coupled electron-transfer step. As a consequence, the rate-limiting step shifts from *O₂ + H₂O + e^- → *OOH + OH^- to *OH desorption. The accumulated *OH then spontaneously spillover from Fe sites to adjacent Co sites due to thermodynamic favorability, where it undergoes further reduction and desorption. Thus, *OOH formation and *OH removal is spatially decoupled and no longer forced to proceed on the same site. In situ ATR-SEIRAS revealed a sharper *OOH signal and stronger *OH features than in the single-site reference catalyst, while DFT calculations confirmed that the *OH spillover pathway is kinetically more favorable than direct desorption from Fe sites. As a result, the c-FeCoDAC catalyst achieved a half-wave potential of 0.85 V, outperforming Pt/C, which clearly illustrates how intermediate spillover can redistribute the energetic penalties of multistep ORR across adjacent atomically defined sites⁸⁴.

SACs-nanoparticle hybrids

The rational construction of hybrid catalytic systems in which SACs coexist with nanoparticles or metal clusters lie in precisely tuning the metal dispersion and the supporting matrix, thereby forming a composite catalytic architecture (Fig. 6b). Although nanoparticle-only catalysts can also show spillover, introducing SACs into hybrid systems enables clearer functional separation and more precise regulation of directional intermediate transfer.

In Pd single atom-Pd nanoparticle systems supported on C₃N₄ (Fig. 8a), Pd₁ sites activated CO₂, while Pd_NPs sites boosted water dissociation for increased H* coverage. The H* produced by Pd_NPs migrate to the Pd₁ sites to promote multiple proton-electron coupling transfer via hydrogen spillover (Fig. 8b)⁸⁵. Similarly, in Pt₁/Ni@C catalysts, Ni@C nanoparticles with strong H₂ activation and hydrogen diffusion capabilities act as hydrogen donors, while Pt single atoms regulate the local concentration and diffusion kinetics of spillover hydrogen. This spatial decoupling of hydrogen activation and hydrogenation sites effectively lowers reaction energy barriers and significantly enhances catalytic performance. DFT simulations further reveal that a higher H adsorption energy associated with Ru SACs predominantly display inhibitory effects on hydrogenation, while a lower H adsorption energy on Pd SACs was not involved the principal kinetic barrier (Fig. 8c–e), leading the pivotal of co-adsorbed hydrogen density⁸⁶. Oxygen spillover is positioned as the key mechanistic bridge between spatially separated Ag nanoclusters and Ag single atoms on oxygen-vacancy rich CeO₂. The Ag nanoclusters preferentially adsorb and activate O₂, and promote O-O bond dissociation via a nearby surface oxygen vacancy, after which the resulting O* migrates from the vacancy region toward the Ag SACs. Ab initio molecular dynamics tracks this transfer as a progressive shortening of O-Ag₁ distance after dissociation, supporting a spillover route that supplies reactive oxygen to the SACs and enables controlled oxidative C-C cleavage (Fig. 8f)⁸⁷. In double-shelled hollow Ni-SA@Cu-NP systems, the inner Ni single-atom layer serves as a localized CO generation center, increasing the local CO concentration and promoting its outward spillover to the outer Cu nanoparticle shell. The Cu nanoparticles subsequently catalyze C-C coupling reactions, converting CO into C₂₊ products such as ethanol, thereby achieving efficient spatial separation and coupling of reaction pathways⁸⁸.

Fig. 8. SACs-nanoparticle hybrids systems.

a Aberration-corrected HAADF-STEM image of Pd_1+NPs/C₃N₄. b The reaction mechanism for photocatalytic conversion of CO₂ to CH₄ over the synergetic catalyst⁸⁵. Copyright 2022, Wiley. Adsorption energies of spillover hydrogen with different densities on c Pt, d Pd, and e Ru single-atom sites, respectively⁸⁶. Copyright 2025, Wiley. f MD investigation on the competitive adsorption on Ag₁ or Ag_n of O₂⁸⁷. Copyright 2025, AAAS.

Li et al. provide an atom + nanoparticle spillover analog using Pd single atoms embedded in Cu nanoparticles to dissociate H₂ and drive hydrogen spillover toward a separate CoMn component⁸⁹. Likewise, in Cd-SACs/Cu catalysts, Cd single-atom sites effectively adsorb and activate CO while simultaneously modulating the electronic structure of Cu nanoparticles, synergistically promoting C-C coupling and significantly improving catalytic performance and product selectivity⁹⁰. Furthermore, in Ce-Pd/WO₃ sensing systems, Ce single atoms modulate the electronic structure of Pd nanoparticles and reduce the energy barrier for hydrogen spillover, functioning as electronic bridges that indirectly enhance sensing performance⁹¹.

SACs-support interaction

The catalytic performance of SACs extends beyond the intrinsic activity of isolated metal sites and is critically governed by their synergistic interactions with the support. Emerging studies reveal that secondary metal species or intrinsic vacancy defects within the support can cooperatively integrate with SACs to form functionally complementary catalytic architectures (Fig. 6c).

Vacancies enable an integrated catalytic framework through modulating intermediate behavior. Bai et al. developed an electrocatalysts with atomically dispersed Ru sites anchored on lattice vacancies of La_0.9NiO₃ for furfural hydrodimerization. Mechanistic investigations reveal that Ru single atoms serve as active sites for the initial hydrogenation to form the FF-CHOH* intermediate with almost 100% selectivity, thereby suppressing the side reaction of hydrogen evolution. Subsequently, the increased FF-CHOH* intermediate undergoes spillover to adjacent Ni sites, decreasing the energy barrier for the subsequent C–C coupling step to Hydrofuroin (Fig. 9a)⁹². Similar cooperative effect has been observed in Cu-SACs/Ti₃C₂T_x catalysts, where the dual Cu⁺ sites with hydrogen-spillover facilitated by the MXene significantly enhance water dissociation to produce more absorbed active hydrogen (Fig. 9b)⁹³. Moreover, Zhang et al. developed single-atom coordination engineering in tailoring atomic-level hydrogen transfer dynamics via a four-oxygen-coordinated Rh single-atom electrode, which may directly modulates the relative position of the Rh d-band center to the Fermi level, thereby regulating H* adsorption on Rh and the RHS efficiency between titanium oxide’s surface oxygen and the Rh single atom on the titanium foam electrode (Fig. 9c)⁹⁴.

Fig. 9. SACs-support interaction systems.

a The mechanism of intermediate spillover for electrocatalytic FF hydrodimerization on Ru-LaNiO₃⁹². Copyright 2025, Elsevier. b Online DEMS measurement of NO₃^- reduction using Cu catalyst⁹³. Copyright 2025, Wiley. c Illustration of H spillover from support to SACs in Rh SACs⁹⁴. Copyright 2025, American Chemical Society. d The HD and H₂ signals of Ni_xMg_1-xO/CeO₂ detected by mass spectrometry⁹⁵. Copyright 2025, Wiley. e Schematic illustration of advantageous features of Fe-N_v-C SAC for ORR compared to Fe-N-C SAC⁹⁶. Copyright 2025, Wiley. f The simulated HER pathway on Pt/QDs/Co3O4. g KIE values of Pt/Co₃O₄ and Pt/QDs/Co₃O₄⁹⁸. Copyright 2025, KeAi.

Vacancies also act as electronic regulation centers that modulate the electronic states of SACs. In Ni_xMg_1-xO/CeO₂ catalysts, Ni single atoms anchored on CeO₂ experience pronounced electronic coupling between Ni d orbitals and CeO₂ p orbitals. This interaction alters the electronic structure of Ni, enhancing its ability to activate N-H bonds during ammonia decomposition (Fig. 9d)⁹⁵. Conversely, the FeN₄-C system provides an instructive counterexample, in which pyridinic nitrogen vacancies adjacent to Fe centers promote O₂ pre-adsorption while inducing a downward shift of the Fe d-band center. This shift suppresses electron spillover from Fe to OH intermediates, optimizes OH adsorption strength and ultimately leads to a substantial enhancement in ORR activity (Fig. 9e)⁹⁶. Besides, in systems composed of Sc single atoms coupled with p-block heteroatoms, Sc serves as the primary site for H₂ adsorption and activation, while the introduced B, N, O, or S atoms function as electron donors or acceptors⁹⁷.

The metal-support electronic interactions can also directly determine whether hydrogen spillover is thermodynamically favorable and kinetically accessible. A recent work on Pt/QDs/Co₃O₄ symbiotic oxide support showed that interface redistributes the Pt 5d electronic structure and generates two electronically distinct Pt environments with different functions: one preferentially promotes H₂O dissociation, whereas the other favors H adsorption and desorption (Fig. 9f). Kinetic isotope measurements (Fig. 9g) further supported the migration of dissociated H* between these sites, indicating that support-induced electronic asymmetry can build both the adsorption-energy gradient and the short transfer pathway required for hydrogen spillover⁹⁸. As a consequence, SAC-support interactions can accelerate catalytic turnover, alleviate H* or OH* accumulation on individual sites, and, in more complex reactions, direct hydrogen delivery toward the desired transformation pathway, thereby jointly influencing activity and selectivity.

Single-atom alloys

In SAAs, electronic interactions between dissimilar metal atoms play a critical role in determining catalytic behavior (Fig. 6d). In the Co₁Zn SAAs, Co atoms serve as the active centers for H₂ dissociation and activation, while Zn atoms cooperate with Co by modifying its electronic structure (Fig. 10a). This synergistic interaction stabilizes the atomic dispersion of Co and simultaneously optimizes hydrogen adsorption and spillover behavior, leading to enhanced Zn-NO₃^- battery performance (Fig. 10b)⁹⁹. In AgCu SAAs catalysts, Ag SAAs promote the reduction of CO₂ to CO intermediates during CO₂RR. The surface structure of Ag favors the adsorption and migration of CO intermediates, which subsequently undergo C-C coupling at Cu single-atom active sites to form multi-carbon products. It is worth noting that the generated CO is directly transferred to the adjacent Cu site rather than desorbing into the electrolyte (Fig. 10c). Through this synergistic spillover mechanism, the selectivity toward C₂₊ products in CO₂RR is significantly improved¹⁰⁰. Similarly, in PtCu SAAs, Pt single atoms are stabilized by Cu nanoparticles and efficiently activate the C-H bond of formic acid molecules, while Cu nanoparticles facilitate hydrogen spillover and promote H₂ release¹⁰¹. Xiong et al. propose a continuous intermediates spillover strategy to boost nitrate reduction by designing dual SAA. The intermediates continuously spill over back and forth on the atomically dispersed Mo and Fe sites in Pd lattice, which adaptively experiences low energy barrier for each elementary step in nitrate conversion (Fig. 10d, e)¹⁰².

Fig. 10. Single-atom alloy systems.

a Mechanism diagram of NO₃RR performance enhanced by hydrogen spillover over Co₁Zn NSs. b Digital photos of the electronic timer powered by the Co₁Zn NSs-based Zn-NO₃^- battery⁹⁹. Copyright 2025, Wiley. c Schematic diagram of CO migration between the cascade catalysts with out-migration routes and inter-migration routes¹⁰⁰. Copyright 2025, American Chemical Society. d The Gibbs free energy diagram for NO₃RR on Mo₁Fe₁Pd along the optimal pathway. e The structural models of reaction intermediates of NO₃RR on Mo₁Fe₁Pd, and the demonstration of dynamic intermediates spillover between Mo and Fe sites¹⁰². Copyright 2025, Wiley.

Perspective and outlook

Intermediate spillover in SAC-based systems has emerged as a promising strategy with the potential to redefine traditional catalytic scaling relations. Growing studies have revealed that the rational design of spillover pathways, interfacial architectures, and the spatial distribution of active sites offers a new paradigm to overcome the long-standing trade-offs among efficiency, selectivity, and stability in SACs. Nevertheless, translating this strategy from lab-scale conceptual demonstrations and case-specific designs into a generally applicable catalyst design principle, and ultimately into industrially viable systems with sufficient robustness, controllability, and scalability, remains a formidable challenge requiring further fundamental and engineering advances (Fig. 11).

Fig. 11. Challenges and prospects of spillover effects.

a Detection of Spillover Species, b Design and Synthesis of Catalysts, c Interface Engineering, d Integrated Applications.

Mechanistic understanding: toward operando, dynamic, and quantitative descriptions of spillover

Current understanding of intermediate spillover still relies predominantly on indirect characterizations. For example, in situ infrared spectroscopy and Raman spectroscopy identify intermediate “fingerprints” and thus enable direct tracking of surface intermediates, including their identity, surface coverage, and dynamic evolution^103,104. Operando X-ray absorption spectroscopy (XAS) and near ambient pressure X-ray photoelectron spectroscopy (NAP-XPS) probe the electronic structure and local coordination environment of active sites under reaction conditions, providing a structural and electronic foundation for understanding spillover mechanisms⁶⁸. Electron paramagnetic resonance (EPR) and differential electrochemical mass spectrometry (DEMS) exhibit exceptional sensitivity toward short-lived reactive species, providing some of the most direct evidence for the existence of spillover intermediates^64,105. Temperature programmed reduction, when combined with spillover sensitive probe materials such as WO₃, enables semi-quantitative comparison of hydrogen spillover efficiency¹⁰⁶. In situ STEM captured the full evolution of MoO₃ nanosheets driven by hydrogen spillover¹⁰⁷. Although many characterizations were developed to illustrate intermediate spillover, direct experimental evidence capturing the full lifecycle of spillover species under realistic reaction conditions remains scarce.

Some techniques may show great potential to visualize the spillover pathways, for example: Electrochemical tip-enhanced Raman spectroscopy (EC-TERS) is especially promising because it combines chemical fingerprinting with nanometer-scale spatial resolution under electrochemical bias, making it attractive for tracking short-lived interfacial intermediates locally. Scanning electrochemical probe microscopies such as SECM and SECCM can spatially resolve local activity, selectivity, and product flux, which may help identify where intermediates are generated and where they are subsequently consumed. In the future, the most convincing mechanistic evidence will probably come not from a single characterization method, but from multimodal operando approaches that combine nanoscale imaging, chemical fingerprinting, isotope labeling, and time-resolved kinetic analysis to directly follow the location, lifetime, and fate of migrating intermediates.

Moreover, theoretical simulations are becoming indispensable tools for elucidating the microscopic nature of spillover mechanisms at the atomic and electronic levels. However, theoretical simulations should integrate with experimental results instead of standalone, thus, mutual validation between calculated descriptors and experimentally observed signals are enabled. The integration of density functional theory, molecular dynamics, and machine learning offers a systematic and versatile approach to understanding spillover phenomena (Fig. 12).

Fig. 12. Theoretical simulation.

a Theoretical modeling framework, b DFT calculation, c Molecular dynamics simulations, d Machine learning (ML).

Consequently, spillover is often treated as a phenomenological concept rather than a rigorously validated elementary step, limiting its incorporation into microkinetic models. This limitation is compounded by the strong coupling of spillover with adsorption-desorption equilibria, surface diffusion, local coverage effects, and dynamic interfacial restructuring. Under operating conditions, spillover frequently coexists with electronic modulation and interfacial field effects, making it difficult to unambiguously distinguish true physical migration from electronic or thermodynamic contributions. As a result, similar catalytic behaviors can often be rationalized by multiple competing mechanisms, weakening the predictive power of spillover-based interpretations.

Beyond experimental advances, establishing a systematic “spillover species map” is essential. Integrating quantitative descriptors of migration length, activation barriers, and adsorption-energy gradients with operando measurements and microkinetic modeling is critical for elevating spillover from a qualitative concept to a predictive design parameter¹⁰⁸. Comparative analyses of diverse intermediates from H and O to C, and N-containing species could reveal universal trends and system-specific regulation principles. Moreover, a very recent study indicates the metal species can also spillover during catalyst. In Cu/SiO₂-CeO₂ systems, Cu species migrate from SiO₂ to CeO₂ and form highly dispersed Cu entities. Surface hydroxyl groups on the support act as formation sites for Cu-OH intermediates, facilitating the detachment and hydroxylation of Cu atoms and enhancing their surface mobility. Meanwhile, the water layer present on the support surface serves as an effective diffusion channel that lowers the energy barrier for Cu-OH migration across surfaces and interfaces. This vacancy- and hydroxyl-assisted migration pathway enables long-range transport of Cu species, suppresses high-temperature sintering, and significantly improves catalyst stability¹⁰⁹. This blurs the conceptual boundary between the catalyst and the intermediate.

Another important unresolved issue is the mechanistic boundary between intermediate spillover and tandem catalysis. In some systems, this distinction can be conceptually straightforward. When a highly reactive species is unlikely to exist as a stable free molecule under operating conditions, its transfer between neighboring sites is more reasonably interpreted as intermediate spillover. By contrast, when a species can desorb as a stable molecule and subsequently reabsorb at another site, or even in another reactor, the process is more appropriately categorized as tandem catalysis. However, between these two limiting cases lies a substantial gray zone. For intermediates such as CO and OH, which may behave either as a surface bound intermediate or as a desorbed molecular species under catalytic conditions, currently available operando techniques often cannot unambiguously determine whether the dominant pathway is true surface bound migration or desorption and reabsorption mediated transfer. Therefore, the distinction between spillover and tandem catalysis should, in some cases, be regarded as an open mechanistic question rather than a settled classification. Resolving this ambiguity will require future development of spatiotemporally resolved operando characterization capable of directly tracking the fate, location, and lifetime of reactive intermediates under realistic working conditions.

Structural design and stability: toward robust and adaptive spillover pathways

Beyond mechanistic identification, a central challenge for intermediate spillover lies in its structural context. In many systems, spillover depends on engineered proximity between single-atom sites, secondary metal centers, or support defects, often realized through complex tandem catalyst designs. However, such proximity is commonly treated as static, whereas under realistic reaction conditions the local structure surrounding single atoms is inherently dynamic. SACs are particularly sensitive to structural evolution because isolated active centers lack the spatial redundancy of extended surfaces or nanoparticles. Subtle changes in coordination environment, defect distribution, or adsorption layers can disrupt the continuity of spillover pathways. Effective spillover in SAC systems therefore depends not only on the coexistence of multiple sites, but on whether their spatial connectivity is preserved under dynamic operating conditions.

At the same time, structural dynamics should not be viewed solely as a destabilizing factor. Reaction-induced reconstruction of supports or redistribution of surface species can, in some cases, generate new migration channels that facilitate intermediate transfer toward isolated metal atoms. Such behavior suggests that effective spillover pathways may emerge in situ, rather than being fully predetermined by the as-synthesized structure. This realization also exposes a critical gap between static theoretical models and dynamic, multiscale experimental observations, which currently limits the predictive power of theory in guiding rational catalyst synthesis.

These observations point to the need for a more flexible design philosophy. Instead of pursuing rigid structural precision alone, future SAC spillover systems may benefit from architectures that tolerate, or even exploit, controlled structural evolution. One promising direction is the development of a closed-loop framework that integrates theoretical prediction, high-throughput screening, and precision synthesis. Machine learning assisted exploration of large chemical spaces could rapidly identify optimal combinations of active sites and supports, while advanced techniques such as modular self-assembly, atomic layer deposition, and microfluidic synthesis enable scalable and controllable catalyst fabrication¹¹⁰. Continuous experimental feedback would, in turn, refine theoretical models, forming a “design, predict, synthesize, validate, and optimize” loop. Expanding support landscape beyond conventional carbons and metal oxides to include materials such as covalent organic frameworks, porous polymers, and two-dimensional materials may further enrich spillover design.

Interface engineering and adaptive regulation: from static junctions to dynamic spillover interfaces

Catalysts under working conditions are inherently dynamic, and this is particularly true for systems involving intermediate spillover. Spillover processes themselves can actively reshape interfacial structures. Experimental observations of reconstruction or oxygen encapsulation on graphene surfaces highlight how spillover can both originate from and feed back into interface evolution¹¹¹, yet predictive control over such dynamics remains limited. In the context of single-atom catalysis, interfacial dynamics are especially consequential. Operating at the limit of atomic dispersion, SACs are highly sensitive to subtle changes in interfacial electronic structure or defect populations, which can shift the balance between direct and spillover-mediated reaction pathways. Interfacial energetics, including work-function alignment, local electric fields, defect-induced charge redistribution, serving as primary control parameters governing spillover efficiency. These challenges motivate the concept of intelligent, self-adaptive interfaces. Rather than designing interfaces that are merely stable, future systems may be engineered to undergo controlled and reversible reconstruction: rapidly activating during the initial reaction stages while maintaining an optimal configuration under steady-state conditions. Reducible supports with dynamically adjustable oxygen vacancies or acid-base sites could act as environment-responsive spillover channels, adapting to reaction atmospheres in real time.

Precise regulation of interfacial energetics provides a direct route to modulate spillover at its energetic origin. Systematic tuning of work-function differences between donor and acceptor components, through rare-earth single-atom modulation of Pd/WO₃ interfaces can accelerate hydrogen spillover and electron transfer by directly governing adsorption strengths and migration barriers. Extending spillover concepts into photo- and electro-driven regimes further expands the design space, as demonstrated by molecular junction photocatalysts that spatially separate charge carriers and reaction sites¹¹². Integrating light-harvesting units with spillover-functional modules may enable reactions that are difficult or inaccessible under purely thermal conditions, further broadening the functional scope of spillover-enabled SACs.

It should also be noted that intermediate spillover is not universally beneficial. Although it is often discussed as a promotional mechanism, its catalytic consequence depends critically on directionality, rate matching, and the chemical nature of the acceptor site. If spillover is excessively fast, poorly directed, or terminates at a site that cannot efficiently transform the migrated species, it may deplete the local coverage of key intermediates at productive centers, accumulate spectator or poisoning species at receiving sites, or redirect reactive intermediates into undesired side pathways. In this sense, effective spillover should not be judged solely by whether migration occurs, but by whether intermediate transfer remains kinetically synchronized with downstream consumption and selectively coupled to the desired reaction channel. Future catalyst design should therefore aim not only to promote spillover, but also to control its extent, destination, and selectivity under realistic operating conditions.

Practical implementation: toward realistic and scalable spillover-enabled SACs

Despite its conceptual appeal, translating spillover-assisted single-atom catalysis into practical systems remains challenging. Many demonstrations rely on idealized model structures with tightly controlled spatial relationships between SACs and auxiliary sites, whereas scalable synthesis typically yields statistical distributions of single atoms, defects, and interfaces, resulting in heterogeneous spillover behavior across catalyst ensembles. From an application perspective, the key question is not whether spillover can occur at an isolated local site, but whether its cumulative contribution remains significant when averaged over a realistic population of single-atom centers. Variations in coordination environment, support morphology, and local reaction conditions can attenuate spillover contributions, even when the underlying mechanism remains valid at the nanoscale. Industrially relevant conditions further impose constraints, as high reactant fluxes, prolonged operation, and impurities may disrupt migration pathways or favor direct reaction routes.

From an engineering perspective, the biggest challenge is to reproducibly manufacture spillover assisted SACs at scale while preserving their atomic structure and spillover function under practical operating conditions, because structural heterogeneity, atom aggregation, support reconstruction, and long-term instability can easily destroy the intended intermediate transfer pathway. Progress toward application-relevant spillover-enabled SACs will therefore depend on shifting emphasis from ideal structural motifs to ensemble-level performance metrics. Designing catalysts that retain spillover functionality despite structural heterogeneity, aging, and environmental fluctuations may be more impactful than maximizing spillover efficiency at isolated sites. This perspective aligns spillover research with practical deployment considerations and provides a pragmatic pathway toward scalable single-atom catalysis. Beyond traditional catalysis, spillover-enabled surface micro-environments offer opportunities in chemical sensing, environmental monitoring, and medical diagnostics. Coupling spillover-active catalysts with biocatalysis or plasma-assisted processes may further enable hybrid platforms beyond conventional catalytic paradigms.

Overall, intermediate spillover plays a pivotal role in single-atom catalysis by enabling spatial decoupling of elementary steps that cannot be simultaneously optimized on isolated atomic sites. By redistributing key intermediates across distinct yet functionally connected sites, spillover mitigates local poisoning, relaxes intrinsic scaling relations, and sustains catalytic turnover under demanding reaction conditions. In this sense, spillover transforms SACs from isolated reactive centers into dynamically coordinated catalytic networks. Building on this understanding, intermediate spillover reframes single-atom catalysis from the optimization of isolated active sites to the orchestration of spatially and functionally coupled reaction steps. Rather than serving as an auxiliary effect, spillover emerges as a design principle for constructing catalytic networks that remain active, selective, and resilient under realistic conditions. Future advances will depend less on maximizing local reactivity and more on engineering controllable pathways for intermediate redistribution across multiscale catalytic architectures. Viewed in this light, the success of future SACs will be determined not by the intrinsic activity of individual sites, but by the efficiency and controllability of intermediate exchange within integrated catalytic architectures.

Author contributions

T.Z. concepted the manuscript. M.S. drafted the manuscript. T.Z. and J.Y. supervised and revised the manuscript.

Peer review

Peer review information

Communications Chemistry thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

Funding

T.Z. discloses support for the research and publication of this work from Leading Innovative and Entrepreneur Team Introduction Program of Zhejiang [2025R01010], Zhejiang Provincial Natural Science Foundation [ZCLQN26B0301], Ningbo Yongjiang Talent Programme and Fundamental Research Funds for the Provincial Universities of Zhejiang [ZX2025000270]. J.Y. and M.S. declare no relevant funding.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s42004-026-02158-5.