Bidirectional reactant flux coupling in hollow hierarchical covalent organic framework enabling efficient uranium extraction from seawater Authors: Jiacheng Zhang, Xuewen Cao, Jun Zhang, Xuefeng Tian, Jianfei Du, Rui Huang, Yihui Yuan, Ning Wang Published: 2026 ABSTRACT Photocatalytic uranium extraction can surpass the intrinsic capacity limits of adsorption by converting soluble uranyl into insoluble uranium peroxides via H2O2 generation. However, this process requires simultaneous local enrichment of H2O2 and UO22+ beyond the solubility product constant, which is difficult to achieve in continuously flowing and ultra-dilute seawater. Here we report a hollow hierarchical covalent organic framework (COF) microcavity reactor, HH-COF-(CN/AO)x, that spatially decouples H2O2 generation from uranyl capture to overcome this thermodynamic barrier via a bidirectional reactant flux coupling strategy. A cyano-functionalized inner layer produces and stores H2O2 in a central cavity, whereas an amidoxime-rich outer layer selectively enriches uranyl ions. The convergence of outward H2O2 flux and inward uranyl adsorption establishes a persistent high-concentration interface that supports continuous formation of insoluble uranium peroxide. The optimized HH-COF-(CN/AO)0.35 achieves 25.1 mg g−1 uranium uptake in natural seawater, 3.9 times that of the non-hollow analog, which demonstrates a generalizable strategy for manipulating reactant fluxes in ultra-dilute environments for realizing effective resource extraction from water environment. Introduction As the global transition toward a low-carbon economy accelerates, the demand for nuclear energy as a pivotal clean baseload power source is surging1–4. Currently, uranium (U) is the key resource of nuclear energy, and global uranium consumption is projected to more than double from approximately 69,000 metric tons at present to over 150,000 metric tons by 2040, raising concerns over the long-term sufficiency of terrestrial ores5–8. In contrast, the oceans host an immense reservoir of approximately 4.5 billion tons of uranium, theoretically capable of sustaining nuclear power for millennia9–14. Nevertheless, translating this vast resource into a practical fuel supply remains exceptionally challenging because uranium exists in seawater at an ultra-trace concentration of about 3.3 ppb within a complex chemical environment characterized by abundant competing ions15–19. These characteristics impose severe thermodynamic limitations and prevent conventional adsorption-based approaches from achieving high extraction efficiencies in practical marine environments20–22. Photocatalytic uranium extraction has emerged as a compelling strategy to transcend the thermodynamic barrier of adsorption23–26. By harnessing solar energy to convert adsorbed uranyl ions (UO22+) into insoluble uranium deposits, thereby continuously regenerating adsorption sites and enabling sustained enrichment27–32. A representative pathway involves mineralization of captured UO22+ by photogenerated H2O2 to form uranium peroxide minerals such as studtite ((UO2)O2·nH2O). However, this mineralization pathway is subject to stringent thermodynamic constraints, and studtite nucleation and growth occur only when local concentrations of both UO22+ and H2O2 exceed the solubility-product (Ksp) threshold (Fig. 1a)33–35. In marine environments, where uranium is extremely dilute and H2O2 rapidly diffuses and dissipates, achieving such localized supersaturation is fundamentally difficult. As a result, photocatalytic mineralization often stalls, preventing continuous uranium enrichment.Fig. 1Bidirectional coupling between H2O2 generation and UO22+ capture in HH-COF-(CN/AO)0.35 microcavity reactor.a Schematic illustration showing that studtite forms only when both UO22+ and H2O2 exceed Ksp threshold concentrations. b Schematic mechanism by which HH-COF-(CN/AO)0.35 microcavity reactor enables bidirectional coupling between H2O2 generation and UO22+ capture, with H2O2 produced in the hollow cavity diffusing outward to react with uranyl adsorbed on the amidoxime-functionalized outer surface. c Schematic illustration of the synthesis process for HH-COF-(CN/AO)x. The outer and inner regions denote AO-enriched and CN-enriched domains, respectively, rather than a perfectly uniform geometric bilayer shell. d SEM image of HH-COF-(CN/AO)0.35. e TEM image of HH-COF-(CN/AO)0.35. f PXRD patterns of HH-COF-(CN/AO)0.35. Bidirectional coupling between H2O2 generation and UO22+ capture in HH-COF-(CN/AO)0.35 microcavity reactor. a Schematic illustration showing that studtite forms only when both UO22+ and H2O2 exceed Ksp threshold concentrations. b Schematic mechanism by which HH-COF-(CN/AO)0.35 microcavity reactor enables bidirectional coupling between H2O2 generation and UO22+ capture, with H2O2 produced in the hollow cavity diffusing outward to react with uranyl adsorbed on the amidoxime-functionalized outer surface. c Schematic illustration of the synthesis process for HH-COF-(CN/AO)x. The outer and inner regions denote AO-enriched and CN-enriched domains, respectively, rather than a perfectly uniform geometric bilayer shell. d SEM image of HH-COF-(CN/AO)0.35. e TEM image of HH-COF-(CN/AO)0.35. f PXRD patterns of HH-COF-(CN/AO)0.35. Covalent organic frameworks (COFs), with their crystalline architectures and programmable pore environments, offer a promising platform for integrating adsorption and photocatalysis36–42. However, traditional COF photocatalysts face an inherent trade-off between mass transport and mineralization, because their uniform nanochannels confine H2O2 generation and uranyl binding within the same nanospace43–46. Solid deposits rapidly block the channels, while the open pore geometry fails to retain H2O2 against convective seawater dilution47,48. These limitations prevent the formation of a persistent interfacial reaction microenvironment, in which H2O2 and UO22+ can be continuously enriched to locally exceed the Ksp threshold for studtite nucleation in ultra-dilute conditions. Therefore, a structure that decouples H2O2 generation from uranyl capture while enforcing their continuous interfacial convergence at high concentration is required. Inspired by compartmentalized biochemical systems that localize reactive species in distinct microdomains, we propose a bidirectional reactant flux coupling strategy that spatially separates the H2O2 production and uranyl capture yet forces their mass transport to converge at a common interface49,50. We design a hollow hierarchical COF (HH-COF-(CN/AO)x) microcavity reactor in which a cyano-functionalized (COF-CN) inner layer catalyzes photogenerated H2O2 production, and the hollow cavity serves as the primary reservoir for H2O2 accumulation, while an amidoxime-functionalized (COF-AO) outer layer selectively captures UO22+ from seawater (Fig. 1b). The hollow cavity accumulates H2O2 to concentrations far exceeding those in bulk seawater, driving sustained outward diffusion toward the uranyl-enriched outer surface of the COF microsphere, where outward H2O2 flux converges with the adsorbed UO22+. The diffusion of H2O2 and adsorption of UO22+ enable efficient bidirectional coupling, promoting controlled heterogeneous nucleation and robust growth of studtite on the photocatalyst exterior. As a result, the optimized microcavity reactor HH-COF-(CN/AO)0.35 achieves 25.1 mg g−1 uranium uptake in natural seawater, which is 3.9 times that of the non-hollow COF-(CN/AO)0.35. This bidirectional reactant flux coupling architecture provides a generalizable route toward solar-driven extraction of critical minerals from vast, low-grade oceanic resources. Results Synthesis and characterization of HH-COF-(CN/AO)x microcavity reactor To enable bidirectional coupling between photocatalytic H2O2 generation and UO22+ capture, we constructed a hollow hierarchical covalent organic framework microsphere, denoted HH-COF-(CN/AO)x microcavity reactor, via a sacrificial templating strategy (Fig. 1c). Amino-functionalized silica microspheres (SiO2-NH2) served as hard templates to direct uniform growth of a cyano-functionalized COF (COF-CN) shell, where the surface amino groups initiated interfacial Schiff-base reactions with the COF monomers and thereby promoted heterogeneous nucleation and subsequent shell growth. COF-CN, which was synthesized by using 1,3,5-triformylphloroglucinol (TP) and 4,4′-diamine-[1,1′-biphenyl]-3,3′-dicarbonitrile (DBD), was chosen because it has been demonstrated to drive photocatalytic H2O2 generation in natural seawater without a sacrificial agent48. In addition, its cyano groups can be chemoselectively converted into amidoxime ligands to afford an amidoxime-functionalized COF (COF-AO) that strongly binds UO22+ from seawater51,52. This single COF scaffold therefore enables functional differentiation into an inner COF-CN layer providing photocatalytically active cyano sites and an outer COF-AO layer presenting uranyl-chelating amidoxime groups, while maintaining intimate spatial coupling. The hollow hierarchical microsphere was obtained through controlled COF shell growth followed by spatially biased oximation and template removal. During shell growth, TP underwent Schiff-base condensation with DBD via the surface amino groups on SiO2-NH2 microspheres, yielding SiO2@COF-CN core-shell microspheres. Scanning electron microscopy (SEM) revealed that the pristine SiO2-NH2 microspheres exhibited smooth surfaces and a uniform diameter of approximately 3 μm (Supplementary Fig. 1). After in-situ COF-CN growth, the SiO2@COF-CN particles displayed a compact, continuous shell tightly covering the SiO2 core (Supplementary Fig. 2). Energy-dispersive X-ray spectroscopy (EDS) mapping showed homogeneous C, N, O and Si distributions, confirming complete encapsulation of the SiO2 microsphere templates (Supplementary Fig. 3). Powder X-ray diffraction (PXRD) of SiO2@COF-CN microspheres showed a characteristic reflection at 2θ of 3.49°, assigned to the (100) plane of COF-CN with an eclipsed AA stacking arrangement (Supplementary Fig. 4). The refined lattice parameters of COF-CN are determined to be P6/M group with a = b = 29.83 Å, c = 3.49 Å, α = β = 90°, and γ = 120° excellent agreement factors (Rwp =5.57% and Rp = 4.44%) (Supplementary Table 1). Pawley refinement corroborated the high crystallinity and structural order of the COF-CN layer uniformly coating the SiO2 microsphere surface51. Surface oxidation of SiO2@COF-CN microspheres was conducted for 2, 4, and 8 h to regulate the extent of cyano-to-amidoxime conversion, affording a series of intermediates denoted SiO2@COF-(CN/AO)x. Consistent with a diffusion-limited penetration of hydroxylamine, this transformation proceeded preferentially near the outer surface, generating a spatially differentiated distribution of cyano and amidoxime functionalities. SEM confirmed that all oximated intermediates retained the spherical morphology and ordered framework of the parent material, indicating that the chemical transformation proceeded while preserving the overall COF architecture (Supplementary Fig. 5). After oximation, selective dissolution of the SiO2 templates generated HH-COF-(CN/AO)x microspheres with well-defined hollow interiors and continuous COF shells, as revealed by electron microscopy (Fig. 1d, e and Supplementary Figs. 6, 7). Fourier-transform infrared (FT-IR) spectroscopy was used to monitor and quantify the progressive conversion of C ≡ N into amidoxime functionalities. Normalization to the aromatic C = C band and integration of the C ≡ N stretching region indicated oximation degrees of approximately 35%, 65% and 75% for 2, 4 and 8 h treatments, corresponding to microspheres of HH-COF-(CN/AO)0.35, HH-COF-(CN/AO)0.65 and HH-COF-(CN/AO)0.75, respectively (Supplementary Fig. 8). Consistent with this chemical evolution, the (100) reflection in PXRD gradually shifted in 2θ from 3.49° to 3.55° with increasing amidoxime content, reflecting a slight lattice contraction associated with substitution of cyano groups by amidoxime groups (Fig. 1f and Supplementary Fig. 9). Meanwhile, the characteristic diffraction peaks were well retained after removal of the SiO2 template, indicating that the crystalline framework remained well preserved. Moreover, cross-sectional SEM images of HH-COF-(CN/AO)0.35 microspheres revealed a shell thickness of approximately 80–100 nm (Supplementary Fig. 10). To optimize the monomer loading, the mass ratio of COF monomers (TP and DBD) to SiO2-NH2 microsphere was varied from 0.1:1 to 0.25:1. SEM images of the SiO2@COF-CN microspheres after template removal showed that the 0.1:1 sample fractured, while the 0.15:1 sample exhibited the most complete morphology with a smooth and compact surface. At higher loading ratios, the hollow structures remained intact but became rougher and looser (Supplementary Fig. 11). After amidoximation, samples with similar amidoximation degree (~35%) were tested for photocatalytic uranium extraction in 32 ppm uranium-spiked seawater without sacrificial agent, achieving uranium extraction amounts of 1935.4, 1689.5, and 1105.9 mg g−1 for the 0.15:1, 0.20:1, and 0.25:1 samples, respectively, after 24 h of irradiation (Supplementary Fig. 12). Therefore, 0.15:1 was selected as the optimal monomer loading ratio. Solid-state 13C NMR spectroscopy provided molecular-level evidence for oximation. HH-COF-(CN/AO)0 microsphere exhibited a resonance at approximately 114 ppm attributable to cyano carbons, whereas HH-COF-(CN/AO)0.35 displayed an additional signal at approximately 169 ppm assigned to amidoxime carbons, confirming partial transformation to amidoxime species (Supplementary Fig. 13). Contact-angle measurements showed a monotonic decrease in the water contact angle with increasing amidoxime content, indicating enhanced hydrophilicity and tunable interfacial properties (Supplementary Fig. 14). In addition, the chemical stability of HH-COF-(CN/AO)0.35 was evaluated after immersion in 0.1 M H2O2, 1 M HCl, and 1% HF for 24 h. The PXRD patterns and FT-IR spectra remained essentially unchanged, indicating good stability of the COF (Supplementary Fig. 15). Collectively, these results demonstrate that oximation of SiO2@COF-CN microspheres combined with template removal yields a hollow COF-(CN/AO)x microcavity reactor with graded surface chemical structure and preserved crystallinity. The resulting architecture features an inner COF-CN layer and an outer COF-AO layer that are spatially differentiated yet closely coupled within a hollow microsphere, providing a structurally well-defined platform to facilitate H2O2 generation in proximity to uranyl adsorption sites and thereby underpinning the proposed bidirectional reactant flux coupling strategy. Confined H2O2 accumulation enabling efficient uranyl enrichment By using the fabricated hollow hierarchical framework, we first verified the thermodynamic constraint that governs uranyl peroxide formation under homogeneous conditions. Control experiments were conducted by adding H2O2 at varying concentrations into uranium-spiked natural seawater of different uranium concentrations. At a uranium concentration of 330 ppb, even 10,000 μM H2O2 resulted in negligible uranium removal. When the uranium concentration was increased to 3300 ppb, discernible removal was observed at 500 μM H2O2. Further increasing the uranium concentration to 8000 ppb lowered the apparent onset threshold of H2O2 to about 200 μM (Fig. 2a). These results support the view that uranyl peroxide nucleation is governed by the Ksp constraint and indicate that homogeneous precipitation is strongly disfavored in ultra-dilute marine environments. This finding motivates the dual-confined bidirectional reactant flux coupling strategy, which aims to co-enrich uranyl ions and H2O2 within a localized microenvironment to cross the thermodynamic nucleation threshold.Fig. 2Confined H2O2 accumulation and photocatalytic uranium extraction performance of HH-COF-(CN/AO)x microcavity reactor.a Uranium removal efficiencies at different H2O2 and uranium concentrations in homogeneous systems. b H2O2 concentrations inside the HH-COF-(CN/AO)0.35 microsphere and the corresponding H2O2 enrichment factor after 24 h of visible-light irradiation. The bars represent H2O2 concentrations leaked into water and stored in the microcavity on the left y-axis, while the line represents the H2O2 enrichment factor on the right y-axis. c Confocal fluorescence image of H2O2 accumulation within the cavity of HH-COF-(CN/AO)0.35 microsphere after light irradiation. d Schematic illustration of confined H2O2 retention in HH-COF-(CN/AO)0.35 microsphere. e Uranium adsorption capacity of HH-COF-(CN/AO)0.35 microsphere in natural seawater under dark conditions. f Photocatalytic uranium-extraction kinetics of HH-COF-(CN/AO)0.35 microsphere in uranium-spiked seawater. g Comparison of photocatalytic uranium-extraction capacities of HH-COF-(CN/AO)0.35 microsphere with reported systems in uranium-spiked seawater without extra-added sacrificial agents. Confined H2O2 accumulation and photocatalytic uranium extraction performance of HH-COF-(CN/AO)x microcavity reactor. a Uranium removal efficiencies at different H2O2 and uranium concentrations in homogeneous systems. b H2O2 concentrations inside the HH-COF-(CN/AO)0.35 microsphere and the corresponding H2O2 enrichment factor after 24 h of visible-light irradiation. The bars represent H2O2 concentrations leaked into water and stored in the microcavity on the left y-axis, while the line represents the H2O2 enrichment factor on the right y-axis. c Confocal fluorescence image of H2O2 accumulation within the cavity of HH-COF-(CN/AO)0.35 microsphere after light irradiation. d Schematic illustration of confined H2O2 retention in HH-COF-(CN/AO)0.35 microsphere. e Uranium adsorption capacity of HH-COF-(CN/AO)0.35 microsphere in natural seawater under dark conditions. f Photocatalytic uranium-extraction kinetics of HH-COF-(CN/AO)0.35 microsphere in uranium-spiked seawater. g Comparison of photocatalytic uranium-extraction capacities of HH-COF-(CN/AO)0.35 microsphere with reported systems in uranium-spiked seawater without extra-added sacrificial agents. To probe structure-function coupling in uranium extraction, we performed photocatalytic analysis using an HH-COF-(CN/AO)x microcavity reactor in uranium-spiked seawater under visible-light irradiation (λ > 420 nm, 100 mW cm−2). A clear trade-off was observed between COF-AO-mediated UO22+ adsorption and COF-CN-driven H2O2 photogeneration, whereby increasing the amidoxime content enhanced dark adsorption but attenuated photocatalytic H2O2 generation. Among all configurations, HH-COF-(CN/AO)0.35 microsphere delivered the highest overall efficiency by combining strong adsorption with sustained H2O2 production, indicating that this composition is the optimal member for spatially separating H2O2 generation from uranyl enrichment (Supplementary Fig. 16). The photocatalytic H2O2 generation performance of the HH-COF-(CN/AO)x microcavity reactor with different oximation degrees under visible light in natural seawater without extra-added sacrificial agents was determined. Increasing the amidoxime content caused a moderate decrease in H2O2 yield, consistent with the progressive replacement of cyano groups that act as the primary catalytic centers (Supplementary Fig. 17). To further elucidate the role of -CN in H2O2 generation, HH-COF-CN and HH-COF-AO were comparatively studied. HH-COF-CN produced about 3.0 mmol g−1 of H2O2 after 24 h of irradiation in natural seawater without a sacrificial agent, whereas HH-COF-AO produced only about 0.35 mmol g−1 under the same conditions (Supplementary Fig. 18). Combined with the UV-vis absorption, Tauc plots, and Mott-Schottky analysis, these results show that HH-COF-CN possesses a suitable band structure for H2O oxidation to H2O2 in natural seawater without sacrificial agents, while HH-COF-AO does not. This confirms that the -CN is mainly responsible for photocatalytic H2O2 generation. Notably, HH-COF-(CN/AO)0.35 maintained an H2O2 generation rate comparable to HH-COF-(CN/AO)0, indicating that partial oximation provides a favorable balance between photocatalytic activity and uranyl-binding functionality. The introduced amidoxime increases surface hydrophilicity and improves water access to reactive sites, thereby helping to offset the partial loss of cyano-derived catalytic centers and to sustain efficient H2O2 production. In addition, HH-COF-(CN/AO)0.35 showed only slight variation in H2O2 yield over the pH range of 4.0-8.2, further demonstrating its good adaptability for photocatalytic H2O2 production under seawater conditions (Supplementary Fig. 19). The ability of the hollow microsphere to function as an H2O2 reservoir was then evaluated. In natural seawater at a solid-liquid ratio of 1:100 (mg: mL), the H2O2 concentration accumulated within the cavity of HH-COF-(CN/AO)0.35 microspheres reached 7414.0 μM after 24 h of visible-light irradiation, 266.3-fold higher than that in the surrounding seawater (Fig. 2b). Even as the solid-liquid ratio was increased to 1:200, 1:500, and 1:1000 (mg: mL), the H2O2 concentration inside HH-COF-(CN/AO)0.35 microspheres remained high at 6932.8 μM, 6667.2 μM, and 6243.7 μM, respectively, while the corresponding enrichment factors increased to 600.7, 997.3, and 1925.99 (Supplementary Fig. 20). This striking concentration gradient indicates that the closed-shell hierarchical architecture effectively retains H2O2 and maintains a persistently oxidative microenvironment. Confocal fluorescence imaging corroborated localized accumulation, showing strong green emission confined within the cavity, accompanied by a faint exterior halo attributed to the diffusion of H2O2 into the surrounding solution (Fig. 2c and Supplementary Fig. 21)53,54. By contrast, no obvious fluorescence was observed in the microspheres before light irradiation under identical conditions, supporting that the detected signal originated predominantly from the H2O2 probe rather than from the intrinsic fluorescence background of the microsphere. This result confirms that the cavity serves as a high-concentration H2O2 pool that can supply an outward H2O2 flux through the shell channels to the exterior interface (Fig. 2d). Complementing H2O2 confinement, the HH-COF-(CN/AO)0.35 microcavity reactor displayed pronounced uranyl enrichment capability from natural seawater. In natural seawater under dark, HH-COF-(CN/AO)0.35 adsorbed 0.2 mg g−1 uranium after 1 day and 1.6 mg g−1 after 7 days, which were greatly enriched compared the initial uranium concentration in natural seawater (Fig. 2e). This substantial accumulation can be attributed to the amidoxime-functionalized outer layer, in which the amidoxime groups coordinate with UO22+ through their O and N donor atoms and the outer-layer configuration shortens the interfacial mass-transfer pathway, thereby jointly promoting rapid uranyl capture and enrichment. This interfacial enrichment establishes a concentrated uranyl reservoir at the solid-liquid interface, ensuring sufficient uranium availability for subsequent H2O2-driven mineralization. The uranium extraction performance of HH-COF-(CN/AO)0.35 in uranium-spiked seawater was evaluated under visible-light irradiation without extra-added sacrificial agents. The effect of pH on uranium extraction was first evaluated at pH 4, 5, 6, 7, and 8.2, where the maximum extraction capacity was achieved at pH 6, reaching 2489.5 mg g−1, while a high extraction capacity of 1972.8 mg g−1 was still retained at pH 8.2 (Supplementary Fig. 22). At pH 8.2, near natural seawater pH, HH-COF-(CN/AO)0.35 reached equilibrium within about 24 h, achieving the capacities of about 167, 614, 1999 and 2644 mg g−1 at initial uranium concentrations of 8, 16, 32 and 64 ppm, respectively, with no obvious pH change observed during the extraction process (Fig. 2f and Supplementary Fig. 23). These capacities place HH-COF-(CN/AO)0.35 among the highest-performing photocatalysts reported under sacrificial-agent-free conditions (Fig. 2g and Supplementary Table 2). This exceptional performance is attributed to a bidirectional reactant flux coupling mechanism, in which H2O2 supplied from the cavity converges with uranyl enriched on the outer surface. This interfacial convergence brings both reactants into the same confined reaction domain and allows the local ion product to exceed the Ksp threshold, thereby supporting continuous mineralization. Notably, after 8 h of light irradiation, the HH-COF-(CN/AO)0.35 microspheres still sustained uranium extraction in the subsequent dark period and gradually ceased within 4 h. This result indicates that the H2O2 accumulated within the microspheres can continue to be released after photogeneration ceases and sustain uranium extraction until its concentration drops below the effective level required for the reaction (Supplementary Fig. 24). After evaluating the extraction kinetics in uranium-spiked seawater, the deposited product was further characterized. ICP-OES analysis showed that the product mainly contained U, Ca, Mg, and Na, with atomic fractions of 54.96%, 16.45%, 20.53%, and 8.06%, respectively, indicating the incorporation of seawater-derived cations. PXRD analysis suggested the presence of phases related to MgUO4, Na4UO5, and CaU2O7 (Supplementary Fig. 25 and Table 3). HH-COF-(CN/AO)0.35 also exhibited excellent cycling stability, retaining over 80% of its initial extraction capacity after five cycles using 0.5 M HCl as the eluent which removes uranium through weakened uranyl-amidoxime binding and dissolution of uranium peroxide species (Supplementary Fig. 26). After five cycles, XRD, FT-IR, and SEM results confirmed that HH-COF-(CN/AO)0.35 maintained its framework and spherical morphology without significant changes, indicating good structural stability during regeneration (Supplementary Fig. 27). Collectively, these findings demonstrate that the hierarchical hollow microsphere effectively couples the outward flux from confined H2O2 accumulation with the inward flux from interfacial uranyl enrichment, thereby alleviating the kinetic and stability constraints that limit photocatalytic uranium extraction in marine environments. Mechanistic elucidation of photocatalytic uranium extraction To elucidate the photocatalytic extraction mechanism, the characterizations of the HH-COF-(CN/AO)0.35 microcavity reactor and the uranium product were examined. Upon visible-light irradiation in uranyl-spiked ultrapure water, HH-COF-(CN/AO)0.35 gradually changed from reddish brown to pale yellow-white, indicating the enrichment of uranium (Supplementary Fig. 28)55. X-ray photoelectron spectroscopy (XPS) subsequently revealed U 4 f signals characteristic of uranyl species, confirming uranium deposition on the HH-COF-(CN/AO)0.35 (Supplementary Fig. 29)56,57. Electron microscopy provided direct evidence for the morphology and location of the deposited phase. SEM revealed numerous columnar particles, 200–1000 nm in size, densely anchored on the outer surface of HH-COF-(CN/AO)0.35 after irradiation (Fig. 3a). The accumulation of these studtite particles may partially block some pore entrances and cover surface active sites, which may inhibit subsequent H2O2 production to some extent. Importantly, this effect mainly occurs after substantial uranium mineralization and is reversible because the covered micropores and surface sites can be re-exposed after elution. These crystallites were firmly attached to the shell rather than loosely aggregated, indicating that nucleation occurred directly at the microsphere-solution interface. Such well-defined interfacial nucleation has rarely been observed for conventional COF photocatalysts lacking a confined cavity under otherwise similar conditions, which is consistent with the hollow hierarchical structure helping to maintain the elevated local reactant densities required for stable product deposition under dilute conditions23,31. TEM and EDS mapping further confirmed the uranium-rich nature of these deposits (Fig. 3b and Supplementary Figs. 30, 31). High-resolution TEM images revealed clear lattice fringes with an interplanar spacing of 0.31 nm, while the selected-area electron diffraction (SAED) pattern displayed sharp concentric rings, indicating well-crystallized deposits (Fig. 3c)58. The measured ring d-spacings of 0.310, 0.184, and 0.159 nm are consistent with the (2 0 0), (4 0 2), and (2 4 − 2) planes of studtite, respectively. Powder X-ray diffraction (PXRD) identified the deposited phase as studtite, with diffraction peaks matching those of (UO2)(O2)·4H2O (ICSD 98-016-7992) (Fig. 3d)59. Inductively coupled plasma mass spectrometry (ICP-MS) analysis showed that 73.6% of the extracted uranium remained attached to HH-COF-(CN/AO)0.35, whereas 26.4% existed as suspended particles (Fig. 3e). The robustness of the interfacial attachment was further examined by ultrasonication. After 5 min of sonication at 200 W, SEM images showed that the columnar crystals remained largely tightly bound to the HH-COF-(CN/AO)0.35 surface without noticeable detachment, indicating robust anchoring of the uranium peroxide crystals (Fig. 3f).Fig. 3Mechanistic of HH-COF-(CN/AO)0.35 microcavity reactor for uranium extraction.a SEM image of HH-COF-(CN/AO)0.35 after photocatalytic uranium extraction. b TEM image and elemental mapping of uranium distribution on HH-COF-(CN/AO)0.35. c HRTEM image and SAED pattern of surface uranium crystallites. d XRD pattern of uranium-loaded HH-COF-(CN/AO)0.35 and the reference pattern of studtite. e Quantitative distribution of extracted uranium. f SEM images of uranium-loaded HH-COF-(CN/AO)0.35 after 5 min of sonication at 200 W. g Schematic illustration of the photocatalytic uranium extraction mechanism of HH-COF-(CN/AO)x microcavity reactor. Mechanistic of HH-COF-(CN/AO)0.35 microcavity reactor for uranium extraction. a SEM image of HH-COF-(CN/AO)0.35 after photocatalytic uranium extraction. b TEM image and elemental mapping of uranium distribution on HH-COF-(CN/AO)0.35. c HRTEM image and SAED pattern of surface uranium crystallites. d XRD pattern of uranium-loaded HH-COF-(CN/AO)0.35 and the reference pattern of studtite. e Quantitative distribution of extracted uranium. f SEM images of uranium-loaded HH-COF-(CN/AO)0.35 after 5 min of sonication at 200 W. g Schematic illustration of the photocatalytic uranium extraction mechanism of HH-COF-(CN/AO)x microcavity reactor. Taken together with the H2O2 accumulation and uranium enrichment results, these observations support the bidirectional reactant flux coupling mechanism (Fig. 3g)51. Under visible light, the COF-CN inner layer photogenerates H2O2, which is retained and concentrated within the hollow cavity. In parallel, the COF-AO outer layer selectively enriches uranyl species from seawater at the shell exterior. This spatial separation of H2O2 production and uranyl capture, coupled with the forced convergence of their fluxes at the shell-solution boundary, establishes a confined interfacial microenvironment where both reactants reach elevated local concentrations. Consequently, the local ion product of uranyl and H2O2 exceeds the Ksp threshold, triggering uranium-peroxide nucleation directly on the exterior surface. Continuous H2O2 supply from the cavity then sustains growth into columnar studtite crystallites, which remain stably anchored under marine flow, likely aided by the rough, heterogeneous exterior surface that promotes mechanical interlocking and interfacial retention. Scavenging experiments further indicate that, besides H2O2, other reactive species such as ·OH, e−, and O2− also participate in the photocatalytic process. However, given that these species mainly serve as intermediates for H2O2 generation and that a substantial amount of H2O2 was detected under visible-light irradiation, photogenerated H2O2 is considered the dominant oxidant directly responsible for uranium peroxide formation (Supplementary Fig. 32). Thus, HH-COF-(CN/AO)x maintains persistent interfacial mineralization by coordinating the outward H2O2 flux with the inward uranyl flux, overcoming the thermodynamic and kinetic barriers that otherwise suppress uranium peroxide formation in ultra-dilute seawater. Uranium extraction from natural seawater To assess the practical applicability of the HH-COF-(CN/AO)x microcavity reactor for uranium extraction from natural seawater, we carried out a continuous-flow test using natural seawater collected from Wenchang Mulan Bay, China. Natural seawater was circulated across a membrane-supported HH-COF-(CN/AO)0.35 under visible-light irradiation (λ > 420 nm, 100 mW cm−2) (Fig. 4a). A non-hollow COF-(CN/AO)0.35 synthesized without the SiO2 template was tested under identical conditions. Both materials exhibited comparable amidoximation degrees of around 35%, whereas the non-hollow control adopted a willow-leaf-like morphology without an internal cavity (Supplementary Figs. 33, 34). Over seven days, HH-COF-(CN/AO)0.35 gradually changed from red to nearly white, indicating progressive surface mineralization (Supplementary Fig. 35). HH-COF-(CN/AO)0.35 reached a uranium uptake of 25.1 mg g−1 within 7 days, corresponding to 3.9 times that of the non-hollow COF-(CN/AO)0.35 (Fig. 4b). This indicates that the uranium extraction performance is mainly attributed to the hollow hierarchical architecture rather than compositional differences. HH-COF-(CN/AO)0.35 exhibited high selectivity toward UO22+ in natural seawater containing abundant competing ions, delivering a U/V ratio of 4.8 (Fig. 4c). The high U/V ratio is attributed to the synergistic effect of selective binding of uranium by amidoxime groups and conversion of bound uranium into insoluble peroxide species driven by photogenerated H2O2, resulting in enhanced uranium fixation on the microspheres.Fig. 4Photocatalytic uranium extraction from natural seawater by HH-COF-(CN/AO)0.35 microcavity reactor.a Diagram of the uranium extraction device. b Uranium extraction capacities of HH-COF-(CN/AO)0.35 and non-hollow COF-(CN/AO)0.35 in natural seawater under visible-light irradiation without external additives. c Extraction selectivity of HH-COF-(CN/AO)0.35 in natural seawater. d Comparison of the photocatalytic uranium extraction performance of different photocatalysts in natural seawater without external additives. Photocatalytic uranium extraction from natural seawater by HH-COF-(CN/AO)0.35 microcavity reactor. a Diagram of the uranium extraction device. b Uranium extraction capacities of HH-COF-(CN/AO)0.35 and non-hollow COF-(CN/AO)0.35 in natural seawater under visible-light irradiation without external additives. c Extraction selectivity of HH-COF-(CN/AO)0.35 in natural seawater. d Comparison of the photocatalytic uranium extraction performance of different photocatalysts in natural seawater without external additives. This uranium uptake capacity ranks among the highest reported for photocatalytic systems tested directly in natural seawater without extra additives, such as sacrificial donors or antimicrobial substances (Fig. 4d and Supplementary Table 4). Unlike static laboratory conditions, natural seawater imposes continual hydrodynamic dilution, H2O2 loss, and rapid mass exchange, which tend to suppress interfacial nucleation. The exceptional performance of the HH-COF-(CN/AO)0.35 microcavity reactor is therefore consistent with its ability to preserve a localized interfacial microenvironment against constant seawater flushing. The hollow cavity concentrates H2O2 and can supply it to the outer surface, helping to replenish interfacial H2O2 even under convective flow. Meanwhile, the amidoxime-rich outer layer helps to sustain a high local uranyl concentration at the interface, ensuring a persistent uranium supply under large-volume throughflow. Together, these bidirectional reactant flux coupled establish a reactive interfacial layer at the microsphere surface, where local H2O2 and uranyl concentrations are maintained at levels favorable for uranyl peroxide nucleation despite the dynamic ocean environment. In addition, to assess the economic potential of HH-COF-(CN/AO)0.35 for uranium extraction from seawater, a preliminary cost analysis was conducted. The fabrication cost was estimated to be 40,509 USD/t, corresponding to an extraction cost of ~367 USD/kg-U. Compared with previously reported materials, HH-COF-(CN/AO)0.35 exhibits a competitive cost together with markedly faster extraction kinetics, indicating its potential for cost-effective uranium extraction (Supplementary Table 5–7). Overall, these results demonstrate that HH-COF-(CN/AO)0.35 effectively resists hydrodynamic dilution, maintains H2O2-uranyl bidirectional reactant flux coupling in moving natural seawater, and enables efficient solar-driven uranium extraction from ultra-dilute oceanic resources. Discussion In summary, we demonstrate a bidirectional reactant flux coupling strategy enabled by a hollow hierarchical COF microcavity reactor that helps to mitigate the thermodynamic and kinetic limitations of photocatalytic uranium extraction under flowing natural seawater conditions. By spatially organizing H2O2 generation and uranyl capture, this architecture directs outward H2O2 flux to converge with inward uranyl enrichment at the microspheres-seawater interface. This interfacial convergence establishes a locally enriched reactive layer with elevated concentrations of both reactants, sustaining heterogeneous nucleation of uranium peroxides while reducing the pore blockage and reactant dilution that can occur in conventional single-channel systems. Accordingly, the optimized microcavity reactor HH-COF-(CN/AO)0.35 achieves a uranium uptake of 25.1 mg g−1 in natural seawater, 3.9 times that of its non-hollow analog. These findings advance the development of solar-driven uranium extraction from seawater and highlight seawater potential as a supplementary reservoir for nuclear fuel. Beyond uranium harvesting, this work illustrates how spatial control of reactant transport and accumulation in an ultra-dilute water environment can be used to alleviate mass-transfer constraints in low-concentration resource extraction. Methods Materials and reagents Amino-functionalized silicon dioxide (SiO2-NH2) microspheres were purchased from Zhichuan Intelligent Technology (Suzhou) Co., Ltd., tetrahydrofuran (THF, 97%), acetone (AC, 97%), ethanol (EtOH, 98%), hydroxylamine hydrochloride (NH2OH·HCl, 99%), triethylamine (Et3N, 99%), hydrofluoric acid (HF, 0.5%), nitric acid (HNO3, 99%), potassium hydrogen phthalate (C8H5KO4, 99%) and UO2(NO3)2·6H2O (99%) were obtained from Aladdin. 2,4,6-triformylphloroglucinol (TP) and 4,4’-diamino-[1,1’-biphenyl]−3,3’-dicarbonitrile (DBD, 97%) were supplied by Jilin Chinese Academy of Sciences-Yanshen Technology Co., Ltd. All chemicals were used as received without further purification, and ultrapure water was used in all experiments. Synthesis of SiO2@COF-CN microsphere SiO2@COF-CN microsphere was synthesized via a solvothermal procedure adapted from previous reports. The mass ratio of COF monomers (TP and DBD) to SiO2-NH2 microspheres was optimized to 0.15:1. Accordingly, TP (56 mg) and DBD (94 mg) were each dissolved in THF (3.6 mL) to afford monomer solutions. The mixed monomer solution was introduced via a syringe pump at a rate of 0.5 mL h−1 into a dispersion containing SiO2-NH2 (1 g), 6 M acetic acid (1 mL) and THF (60 mL). The mixture was stirred continuously at 1000 rpm to promote the Schiff base condensation on the surface of SiO2-NH2 microsphere. The solid was collected by filtration and transferred into a Teflon-lined autoclave containing a mixed solvent of mesitylene (4 mL), 1,4-dioxane (4 mL), and 6 M acetic acid (2 mL). The suspension was ultrasonicated and purged with N2 for 20 min, followed by heating at 120 °C for 72 h. The resulting precipitate was washed three times with acetone and dried in air at 60 °C for 12 h to afford SiO2@COF-CN microsphere. Samples with other mass ratios of COF monomers to SiO2-NH2 microsphere (0.1:1, 0.20:1, and 0.25:1) were also prepared following the same procedure for comparison. Synthesis of HH-COF-(CN/AO)x microcavity reactors Safety statement Hydrofluoric acid (HF) is highly toxic and corrosive and can cause severe burns and systemic toxicity. All procedures involving HF were performed by trained personnel in a certified chemical fume hood with appropriate personal protective equipment, including safety goggles/face shield, lab coat, and HF-compatible gloves. Calcium gluconate gel was kept readily available, and HF-containing waste was collected and disposed of according to institutional safety regulations. NH2OH·HCl (100 mg) was dissolved in EtOH (50 mL), and the pH of the solution was adjusted to 7.0 using Et3N. SiO2@COF-CN microspheres (600 mg) were then dispersed in the above solution and reacted at 60 °C under stirring at 500 rpm. By varying the reaction time (2, 4, and 8 h), the conversion of the -CN groups was controlled to obtain SiO2@COF-(CN/AO)x microsphere composites with different degrees of oximation. After the reaction, the solid was collected and etched with 30 mL of 0.5% HF solution for 2 min to remove the SiO2 template. The resulting hollow HH-COF-(CN/AO)x microspheres were dialyzed against deionized water for 24 h using dialysis bags with an MWCO of 1000 kDa to remove residual fluoride ions and other impurities. Finally, the dialyzed samples were lyophilized to afford HH-COF-(CN/AO)x microcavity reactors with tunable oximation degrees. Characterizations Functional groups were analyzed by Fourier-transform infrared spectroscopy (FT-IR, LR 64912 C, PerkinElmer, USA). The morphology and elemental distribution were examined using field-emission scanning electron microscopy (FE-SEM, S-4800, Hitachi, Japan) equipped with an energy-dispersive X-ray spectroscopy (EDS) detector. The microstructure, elemental mapping, and lattice fringes at the atomic scale were characterized by transmission electron microscopy (TEM, Talos F200X, FEI, USA) equipped with an EDS detector. Confocal fluorescence imaging was performed using a confocal laser scanning microscope (LEICA STELLARIS 5, Leica Microsystems, Germany) to visualize the distribution of H2O2 inside the microspheres. X-ray photoelectron spectroscopy (XPS) was conducted on a Thermo VG RSCAKAB 250X spectrometer (Thermo Fisher Scientific, USA) with an Al Kα radiation source. X-ray diffraction (XRD) patterns were collected on an X-ray diffractometer (XRD-6000, Shimadzu, Japan) using Cu Kα radiation. The extraction capacities for uranium and competing ions were quantified by inductively coupled plasma mass spectrometry (ICP-MS, 7900, Agilent, USA) and ultraviolet-visible (UV-vis) absorption spectroscopy (UV-1600, Mapada, China). Verification of uranyl peroxide formation under homogeneous conditions Natural seawater was filtered through a 0.22 μm membrane and spiked with uranyl nitrate hexahydrate (UO2(NO3)2·6H2O, analytical grade) to prepare solutions containing 330, 3300, and 8000 ppb of uranium. These concentrations were selected to be significantly higher than the natural uranium level in seawater (~3.3 ppb) to enable controlled comparative evaluation. For each experiment, 50 mL aliquots of the uranium-spiked seawater were transferred into sealed glass vials. Hydrogen peroxide (30 wt%) was then added to achieve final H2O2 concentrations ranging from 0 to 10,000 μmol L−1. The mixtures were magnetically stirred at room temperature (25 ± 1 °C) for 6 h to ensure homogeneity and subsequently left undisturbed for 7 days to reach equilibrium. After the reaction period, the suspensions were filtered through 0.22 μm polyethersulfone (PES) syringe filters to remove any precipitates. The filtrates were immediately acidified with 2% HNO3 to prevent reprecipitation and then stored at 4 °C before analysis. The residual uranium concentration in the filtrates was quantified by inductively coupled plasma mass spectrometry (ICP-MS), and the uranium removal efficiency (%) was calculated according to:1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Removal\,rate(\%)=\frac{C_{0}-C_{t}}{C_{0}}\times 100\%$$\end{document}Removalrate(%)=C0−CtC0×100%where C0 and Ct (ppm) are the initial and equilibrium uranium concentrations, respectively. Evaluation of photocatalytic uranium extraction performance Natural seawater was filtered through 0.22 μm membranes to remove suspended particulates and then spiked with a 1000 ppm UO2(NO3)2·6H2O stock solution to obtain initial uranium concentrations of 8, 16, 32, and 64 ppm. The solution pH was then adjusted to 8.2 using 0.1 M NaOH. In a typical experiment, 5 mg of HH-COF-(CN/AO)xwas dispersed in 500 mL of uranium-spiked seawater by ultrasonication for 1 min, followed by magnetic stirring. The suspension was irradiated at room temperature under visible-light irradiation (λ > 420 nm, 100 mW cm−2). No external sacrificial agents were added. Dark adsorption experiments were carried out in parallel under identical conditions but without light irradiation. At predetermined time intervals, 1 mL aliquots were withdrawn, filtered through 0.22 μm polyethersulfone syringe filters, and subjected to ICP-MS analysis to determine the residual uranium concentration. The uranium extraction capacity qt (mg g−1) at time t was calculated according to:2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$q_{t}=\frac{(C_{0}-C_{t})\times V}{m}$$\end{document}qt=(C0−Ct)×Vmwhere C0 and Ct (ppm) are the initial and residual uranium concentrations in solution, respectively, V(L) is the volume of the solution, and m (g) is the mass of the photocatalyst. Determination of H2O2 production The concentration of H2O2 in the reaction solution was determined using an iodometric titration method adapted from a reported procedure. In a typical measurement, 1 mL of the bulk reaction solution was mixed with 1 mL of 0.4 M KI and 1 mL of 0.1 M C8H5KO4 and allowed to stand for 1 h. Under acidic conditions, H2O2 oxidizes I− to triiodide (I3−), which exhibits a characteristic absorption peak at 350 nm. The absorbance at 350 nm was recorded by UV-vis spectroscopy and used to quantify I3−, from which the H2O2 concentration in solution was calculated. To evaluate the H2O2 concentration confined within the cavities of HH-COF-(CN/AO)x microsphere, the materials were separated from the reaction solution by filtration and gently ground in an ice bath. The resulting solid was then dispersed in deionized water, diluted to a fixed volume, and filtered to remove insoluble residues. The filtrate was analyzed using the same iodometric titration procedure described above, enabling the determination of the H2O2 produced and retained within the cavities of HH-COF-(CN/AO)x. The amount of H₂O₂ retained within the microsphere cavities was then converted to concentration by dividing the measured H2O2 amount by the total cavity volume of the microspheres. The cavity volume for 1 mg of COF microspheres was determined to be 4.19 μL. Therefore, the confined H2O2 concentration was calculated according to:3\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$C(H_{2}O_{2})=\frac{{{{\rm{n}}}}(H_{2}O_{2})}{V}$$\end{document}C(H2O2)=n(H2O2)Vwhere C(H2O2) is the amount of H2O2 quantified from the filtrate after release from the microsphere cavities, and V is the total cavity volume of the microspheres used in the experiment. Fluorescence imaging of H2O2 accumulation within the cavity of microspheres H2O2 accumulated within the cavity of microspheres was visualized using ROSGreenTM H2O2 Probe, which becomes fluorescent upon reaction with H2O2 and exhibits characteristic Ex/Em = 490/514 nm. Confocal imaging was performed with 488 nm excitation and 520–530 nm emission collection, a spectral window chosen to minimize interference from the intrinsic fluorescence background of the microspheres under 488 nm excitation. Microspheres with or without prior light irradiation were incubated with ROSGreenTM H2O2 Probe for 10 min and then imaged under identical acquisition settings, with optimized exposure time to minimize background interference from the intrinsic fluorescence of the microspheres. Uranium extraction from natural seawater One hundred liters of natural seawater were collected from Wenchang Mulan Bay, Hainan, China. The seawater was transported to the laboratory and immediately filtered through 0.22 μm polyethersulfone membranes to remove suspended particulates. The extraction test was conducted using a closed-loop circulation system consisting of a reservoir, a peristaltic pump, and a flat-sheet polysulfone membrane with a pore size of 0.8 μm, which served as the support for the photocatalysts. Ten milligrams of photocatalysts were uniformly dispersed across the membrane surface before each experiment. The filtered seawater was circulated through the system at a constant flow rate of 2 L h−1. The membrane plane was irradiated under visible-light (λ > 420 nm, 100 mW cm−2). The temperature of the circulating solution was maintained at 25 °C ± 2 °C throughout the experiment to minimize thermal effects. Two materials were tested under identical conditions. One was HH-COF-(CN/AO)0.35, and the other was the non-hollow COF-(CN/AO)0.35, which exhibits no cavity. During operation, aliquots of seawater were collected at predetermined time intervals, then acidified and analyzed by ICP-MS to monitor uranium extraction capacity. Supplementary information Supplementary Information Transparent Peer Review file Supplementary Information Transparent Peer Review file Source data Source Data Source Data