Highly flexoelectric response for uranium removal and recovery using the organic-inorganic hybrid material

Abstract

Catalysis was an effective method for uranium recovery and environmental remediation. However, the weak flexoelectric response, despite being universal in dielectric materials, greatly limited its appeal for research and catalytic applications. Here, we proposed a strategy to enhance the flexoelectric response by bridging inorganic chains with metal-organic chains within the structure. The resulting hybrid material, Co[C₄H₄N₂]V₂O₆, demonstrated excellent uranyl removal performance, surpassing that of state-of-the-art piezocatalysts. Co[C₄H₄N₂]V₂O₆ showed flexocatalytic uranyl activity across a broad pH range and under high-salinity conditions. Under a dynamic experimental setup, Co[C₄H₄N₂]V₂O₆ showed strong potential for practical flexocatalytic applications. Co[C₄H₄N₂]V₂O₆ could efficiently separate uranyl in contaminated potable water, reducing the uranium concentration (~2.0 ppm) to below the drinking water standard (30 ppb). It could also lower the uranium concentration (~5.6 ppm) in mining wastewater to below the discharge limit (300 ppb). Its intrinsic anisotropic mechanical properties and cantilever-like morphology endowed high deformability, which, together with a large dielectric constant, enhanced its flexoelectric polarization. The revealed flexocatalytic mechanism confirmed that uranyl was converted into insoluble (UO₂)O₂·2H₂O by active species generated through dynamic polarization. This work provided a promising avenue for the design of advanced flexocatalysts and offered an effective strategy for uranium recovery and environmental remediation.

This study reports a highly flexoelectric organic-inorganic hybrid material for uranium removal from wastewater. The work identifies key factors governing flexocatalytic capacity and offers insights for uranium recovery and environmental remediation.

Introduction

The piezoelectric effect attracted growing interest for its potential in mechanical energy conversion and process chemistry¹. The conversion of mechanical energy into chemical energy through piezoelectric materials could drive catalytic reactions, such as water splitting for hydrogen generation and water purification². Piezocatalysts were expected to harness not only environmental mechanical vibrations (such as wind and waves), but also redundant vibrations generated in industrial systems to achieve catalytic activity³. When subjected to external mechanical stress, piezoelectric materials induced polarization, which in turn drove charge carrier transfer between catalysts and reactants and subsequently modulated the redox reactions^4,5. Low carrier concentration, slow migration rates, and insufficient active sites were recognized as major drawbacks in the field of piezocatalysis, significantly restricting the overall efficiency of the catalytic reactions⁶. Several approaches, such as increasing the dielectric constant⁷, controlling the morphology⁸, and introducing defect engineering⁹, were employed to enhance the piezocatalytic efficiency of materials. However, these strategies for the large-scale preparation of piezoelectric materials were relatively complicated and difficult to implement, as they often required precise synthetic conditions and costly processes⁶. Meanwhile, the piezoelectric effect existed only in materials with the non-centrosymmetric structure, limiting the choice of suitable piezocatalysts and their practical applications¹⁰.

Flexoelectric polarization was able to occur in all dielectric materials due to the fourth-order tensor nature of the flexoelectric coefficient^11–13. This polarization, unlike piezoelectric polarization that was confined to only one or two atomic layers, appeared on the material surface and was distributed throughout the bulk over larger length scales due to significant local strain gradients¹⁴. These characteristics significantly broadened the choice of materials available for flexocatalysis. Centrosymmetric MnO₂ nanoflowers and SnO₂ nanorod clusters were shown to efficiently degrade organic dyes through flexocatalysis, while the flexoelectric effect in centrosymmetric SrTiO₃ nanopowder was confirmed to enable hydrogen evolution reactions^3,15,16. Existing studies demonstrated that surface charges generated from flexoelectric polarization, such as electrons (e^-) and holes (h⁺), could participate in reactions with H₂O and dissolved O₂ to produce active oxygen species, including •O₂^- and •OH¹⁷. These active species promoted reactions for water splitting and organic degradation^18,19. Meanwhile, reactive species (such as e^-, •O₂^-, and •OH) generated from photocatalytic and piezocatalytic processes had been confirmed to be capable of altering the chemical speciation of soluble UO₂²⁺, leading to its transformation into insoluble forms^20–22. Therefore, we speculated that flexocatalysis could similarly drive the transformation of UO₂²⁺. In particular, the conversion of UO₂²⁺ into insoluble species was recognized as an efficient method for uranium recovery and the remediation of radioactive contaminants in the nuclear industry^23,24. Nevertheless, approaches for preparing high-performance flexoelectric materials for catalyzing UO₂²⁺remained largely unexplored, along with a clear understanding of the catalytic mechanisms.

Flexoelectricity was an electromechanical property that represented the coupling between inhomogeneous strain and electric polarization^12,25. The strain gradient could locally break inversion symmetry, inducing polarization with a preferred direction and thereby enabling a piezoelectric-like response in centrosymmetric materials despite the absence of intrinsic piezoelectric components^26,27. Generally, the charge density generated by flexoelectric polarization was relatively small, greatly reducing its appeal for research and application in the catalytic field²⁶. The flexoelectric polarization of dielectric materials was determined by $P_i={\mu }_{ijkl}\left(\partial u_{kl}\right)/\left(\partial x_j\right)$, where ${\mu }_{ijkl}$ and $\left(\partial u_{kl}\right)/\left(\partial x_j\right)$ were the flexoelectric coefficients and strain gradient components along the $x_j$ direction²⁸. Experiments confirmed that the flexoelectric coefficient was positively correlated with the dielectric constant, with higher dielectric constants corresponding to larger flexoelectric coefficients²⁹. Importantly, enhancing the dielectric constant could reduce the exciton binding energy, thereby facilitating the separation of electrons and holes³⁰. Meanwhile, under external mechanical stress, soft dielectric materials generally produced greater inhomogeneous strain, leading to an enhanced flexoelectric response^31–33. Therefore, designing materials that combine high dielectric constants with excellent flexibility could pave the way for the rational construction of high-performance flexocatalysts. Inorganic materials generally exhibited high dielectric constants and hardness, whereas organic materials exhibited excellent mechanical flexibility but low dielectric constants. Organic inorganic hybrid materials could integrate the high dielectric constants of inorganic components with the mechanical flexibility of organic moieties³⁴. Therefore, we speculated that bridging inorganic chains with metal-organic chains to fabricate hybrids could achieve an excellent flexoelectric response and high flexocatalytic performance. However, this strategy had not yet been reported and remained to be developed.

Herein, we introduced a strategy to fabricate materials with a high flexoelectric response by bridging inorganic chains with metal-organic chains within the material structure. This approach established the hybrid material Co[C₄H₄N₂]V₂O₆ as an excellent flexocatalyst for uranyl removal. Co[C₄H₄N₂]V₂O₆ possessed a bouquet-like morphology assembled from microsheets. The bridging mode endowed these cantilever-like microsheets with facile deformability and a high dielectric constant. These features collectively contributed to the excellent flexocatalytic performance under external mechanical stimuli. Co[C₄H₄N₂]V₂O₆ exhibited catalytic activity across a broad pH range (1.50-12.00) and in high-salinity environments (1.5 M NaCl, 1.5 M CaCl₂, 1.5 M MgCl₂, and 1.5 M NaHCO₃). Meanwhile, Co[C₄H₄N₂]V₂O₆ could efficiently catalyze the conversion of uranyl even under low ultrasonic power of 20 W. Importantly, we evaluated the flexocatalytic ability in various water matrices using a dynamic experimental setup. Co[C₄H₄N₂]V₂O₆ showed outstanding uranyl separation ability in contaminated potable water, contaminated seawater, and uranium mining wastewater. In particular, it could effectively reduce the U levels in contaminated potable water from ~2 ppm to below the acceptable limit (30 ppb) and in mining wastewater from ~5.6 ppm to under the discharge threshold (300 ppb). The revealed flexocatalytic mechanism confirmed that the active radicals (e.g., e^-, •O₂^-, •OH, and H₂O₂) generated from dynamic polarization could participate in reactions with U(VI), leading to the formation of the insoluble form (UO₂)O₂·2H₂O.

Results

Flexoelectric response capability of the centrosymmetric Co[C₄H₄N₂]V₂O₆

Guided by our concept, we successfully synthesized the hybrid material Co[C₄H₄N₂]V₂O₆ by bridging metal-organic chains with vanadium oxide chains via a one-step hydrothermal method (Fig. 1a). The structure of Co[C₄H₄N₂]V₂O₆ was analyzed by Rietveld refinement of the powder X-ray diffraction (PXRD) pattern. Co[C₄H₄N₂]V₂O₆ crystallized in the centrosymmetric space group Pnna and exhibited high phase purity and crystallinity (Fig. 1b). In the structure, each metal-organic [Co(C₄H₄N₂)]_n²ⁿ⁺ chain was cross-linked with three metavanadate [V₂O₆]_n^2n- chains via {-V-O-Co-O-V-} linkages, forming a charge-balanced three-dimensional hybrid framework. Morphological analyses revealed that Co[C₄H₄N₂]V₂O₆ featured a bouquet-like microstructure assembled from microsheets (Fig. 1c, d). These microsheets exhibited preferential growth along the [002] crystallographic direction (Fig. 1d). The (002) interplanar spacing was measured to be 0.396 nm, in good agreement with the refinement result. The preferentially grown microsheets, regarded as cantilever beams³, could be easily flexed and were prone to significant deformation under external mechanical stimuli.

Fig. 1. Synthesis, structure and morphology of Co[C₄H₄N₂]V₂O₆.

a Synthesis and crystal structure of Co[C₄H₄N₂]V₂O₆. b Rietveld refinement plots of PXRD patterns for Co[C₄H₄N₂]V₂O₆, where R_p, R_wp, and S represented the profile factor, weighted profile factor, and goodness-of-fit indicator, respectively. c SEM image of Co[C₄H₄N₂]V₂O₆. d HRTEM image of Co[C₄H₄N₂]V₂O₆.

Meanwhile, the deformability was closely correlated with the elastic constants, with smaller values typically leading to larger deformations³⁵. The theoretical elastic constants of Co[C₄H₄N₂]V₂O₆ were calculated and found to be non-negative values (Supplementary Table 1). Moreover, all mechanical stability criteria, including [C₁₁ + C₂₂ + C₃₃ + 2(C₁₂ + C₁₃ + C₂₃)] > 0, (C₁₁ + C₂₂− 2C₁₂) > 0, (C₁₁ + C₃₃ − 2C₁₃) > 0, and (C₂₂ + C₃₃ − 2C₂₃) > 0, were satisfied. These results indicated that Co[C₄H₄N₂]V₂O₆ exhibited sufficient mechanical stability³⁶, making it suitable for flexocatalytic applications. Analysis of the Young’s and shear moduli revealed that the moduli along the [001] direction (E_z = 40.57 GPa, G_z = 18.59 GPa) were lower than those along the [100] and [010] directions (E_x = 73.20 GPa, E_y = 99.16 GPa, G_x = 23.15 GPa, G_y = 45.01 GPa), indicating preferential deformability along the [001] direction (Fig. 2a–c, Supplementary Table 2). Meanwhile, the Poisson’s ratios along the [001] direction (σ_zx = 0.30 and σ_zy = 0.34) approached 0.33 (Fig. 2d), suggesting ductility comparable to that of metallic materials³⁷. Furthermore, 3D surface contours of the Young’s modulus demonstrated a strongly anisotropic mechanical behavior (Fig. 2b). Because C₃₃ was lower than C₁₁ and C₂₂, the Young’s modulus along the [001] direction was reduced (Fig. 2a). The relationship between structure and mechanical behavior was further analyzed to understand the origin of anisotropy. The mechanical property along the [001] direction was dominated by the flexible metal-organic chains [Co(C₄H₄N₂)]_n²ⁿ⁺, resulting in a smaller elastic constant (C₃₃ = 71.86 GPa). In contrast, rigid metavanadate [V₂O₆]_n^2n- chains along the [100] and [010] directions enhanced strain resistance, as reflected in the higher values of C₁₁ (86.82 GPa) and C₂₂ (157.3 GPa). Overall, the combination of intrinsic anisotropic mechanical properties and the extrinsic cantilever-like morphology synergistically endowed the material with excellent deformability.

Fig. 2. Elastic properties of Co[C₄H₄N₂]V₂O₆.

a Planar projections of Young’s modulus (E) on the (001), (010), and (100) planes. b Contour plots of Young’s modulus in 3D space. c Anisotropic Young’s modulus and shear modulus (G) along the X, Y, and Z crystallographic directions. d Poisson’s ratio (σ) along the X, Y, and Z axes.

The flexocatalytic polarization ability was strongly correlated with the dielectric constant, where a larger value corresponded to a higher polarization²⁶. The dielectric constants, including the electronic and ionic dielectric tensors (ɛ_xx, ɛ_yy, ɛ_zz), were calculated (Fig. 3a). The dielectric constants along different crystallographic directions were expressed as ɛ₀ = ɛ_∞+ ɛ_i, where ɛ_∞ and ɛ_i represented the electronic and ionic contributions, respectively^38,39. The total dielectric constant of Co[C₄H₄N₂]V₂O₆ (8.07) was higher than that of the isolated pyrazine molecule (2.95) and previously reported vanadium oxides (4.46 for V₂O₅ and 5.36 for VO₂)^40,41. The enhanced dielectric constant could improve the polarizability of Co[C₄H₄N₂]V₂O₆. Moreover, Co[C₄H₄N₂]V₂O₆ exhibited slightly anisotropic dielectric behavior, with ɛ_xx, ɛ_yy, and ɛ_zz values of 7.67, 8.34, and 8.20, respectively. This anisotropic dielectric response was related to the unique bridging configuration of Co[C₄H₄N₂]V₂O₆. Furthermore, the projected density of states (PDOS) for Co[C₄H₄N₂]V₂O₆ was analyzed. The upper valence bands were primarily composed of Co 3 d states, whereas the lower conduction bands consisted of Co 3 d, V 3 d, O 2p, and N 2p states (Fig. 3b–d). Moreover, each Co atom was octahedrally coordinated with O and N atoms in the Co[C₄H₄N₂]V₂O₆ structure. In the octahedral field, Co²⁺ normally possessed a 3d⁷ electronic configuration $t_{2g}^5{e}_g^2$. However, the orbital energies and electronic occupations of Co 3 d were significantly modified by the octahedral distortions (Fig. 3b, e). This indicated that the octahedral distortions could influence the band bending process and affect the alignment of the VBM and CBM positions relative to the redox potential.

Fig. 3. Dielectric constant, electronic structure, and flexoelectric polarization capability of Co[C₄H₄N₂]V₂O₆.

a Dielectric constant along the X, Y, and Z crystallographic directions. b PDOS analysis of Co 3 d orbitals. c PDOS analysis of V 3 d, O 2p, and N 2p orbitals. d Spatial electron density distributions of the valence band maximum (VBM) and conduction band minimum (CBM) plotted at an isosurface value of 1 × 10^-4 e·Å⁻³. e Schematic diagram of the orbital occupation of Co²⁺ 3d⁷ electrons in the octahedral field. f Schematic illustration showing the flexoelectric response of centrosymmetric Co[C₄H₄N₂]V₂O₆ under a nonuniform strain. g Band tilting induced by flexoelectric polarization. Flexoelectric polarization simulated by finite element analysis using the cantilever beam model, showing h strain gradient distribution and i polarization distribution under an applied force of 100 MPa.

The flexoelectric polarization capability was evaluated through finite element simulations. Generally, ultrasonic waves were used as the mechanical excitation source in flexocatalysis, generating pressures of 10⁵-10⁶Pa in the liquid environment^42,43. When the cavitation bubbles reached their critical rupture size and collapsed, they could generate peak pressures of 10⁸-10⁹ Pa, exerting forces on the surrounding liquid environment. The spatial pressure distribution could induce inhomogeneous strain in the Co[C₄H₄N₂]V₂O₆ microsheets, locally disrupting their inversion symmetry and thereby generating internal polarization. This polarization drove e^- and h⁺ to separate and migrate toward opposite directions, resulting in their accumulation on the material surfaces and causing a complete tilting of the energy bands (Fig. 3f, g). To reproduce the mechanical environment in simulations, an external force (100 MPa) was applied to deform the microsheets, which induced a clear strain (Supplementary Fig. 4) and generated substantial strain gradients up to 8 × 10⁵m^-1 and flexoelectric polarization peaking at 0.4 C·m^-2 (Fig. 3h, i). The obtained values surpassed those reported for several inorganic materials^44,45, demonstrating that Co[C₄H₄N₂]V₂O₆ exhibited a robust flexoelectric polarization capability.

Flexocatalytic uranyl mechanism

Uranyl (UO₂²⁺) ions were the common aqueous form of uranium in wastewater, posing significant environmental and health risks due to their high mobility and toxicity. Under ultrasonic excitation, we found that Co[C₄H₄N₂]V₂O₆ could effectively transform soluble UO₂²⁺ into distinct yellow precipitates (Fig. 4a). The final products were further characterized using PXRD, X-ray photoelectron spectroscopy (XPS), Fourier transform infrared spectroscopy (FTIR), and Raman spectroscopy. PXRD pattern confirmed that the yellow products crystallized in the (UO₂)O₂·2H₂O phase, indicating that flexocatalysis effectively altered the uranyl speciation in wastewater (Fig. 4b). XPS spectrum of U 4 f exhibited characteristic peaks at 391.6 eV (U 4f_5/2) and 380.7 eV (U 4f_7/2)⁴⁶, confirming the +6 oxidation state of uranium in the final products (Fig. 4c). FTIR spectrum showed a peak at 609 cm^-1, corresponding to the asymmetric stretching vibration of the U-O bond, while bands at 728 and 908 cm^-1 were assigned to the symmetric and asymmetric stretching vibrations of the U = O bond, respectively (Fig. 4d)¹⁸. The peaks at 1625 and 3420 cm^-1 were attributed to the bending and stretching vibrations of the O-H groups, indicating the presence of adsorbed or crystalline water molecules in (UO₂)O₂·2H₂O⁴⁷. Raman spectrum displayed peaks at 828 and 853 cm^-1, corresponding to the symmetric stretching vibration of the O = U = O bond and the O-O stretching vibration of the peroxide group bridging oxygen ligands in (UO₂)O₂·2H₂O, respectively (Fig. 4e)⁴⁸. These results further demonstrated that, under ultrasonic stimulation, Co[C₄H₄N₂]V₂O₆ generated dynamic flexoelectric polarization, which converted soluble UO₂²⁺ into the insoluble products.

Fig. 4. Flexocatalytic uranyl removal mechanism through Co[C₄H₄N₂]V₂O₆.

a Photograph of the final product after the flexocatalytic uranyl removal process. b PXRD patterns obtained for samples after flexocatalysis by Co[C₄H₄N₂]V₂O₆ in a 100 ppm uranium solution (pH = 4). c XPS spectra of U 4 f, d FTIR spectrum, and e Raman spectrum of the product from flexocatalysis. f Quenching experiments performed during the flexocatalytic process. g Detection of DMPO-•O₂^- and DMPO-•OH radicals by EPR during the flexocatalytic process. h The concentration of H₂O₂ as a function of time in deionized water and uranium solutions during the flexocatalytic process. i Schematic illustration showing the flexocatalytic uranyl removal mechanism.

The transformation of UO₂²⁺ into (UO₂)O₂·2H₂O under flexocatalytic conditions remained unclear and was not previously investigated. For Co[C₄H₄N₂]V₂O₆, the external mechanical vibrations could induce flexoelectric polarization. Near the material surface, the generated e⁻ and h⁺ participated in redox reactions with species such as dissolved O₂ and water molecules, leading to the formation of •O₂^- and •OH. The •O₂^- and •OH subsequently promoted the generation of H₂O₂^49,50. The accumulation of H₂O₂ further altered the solution environment, decreasing pH from 6.92 to 5.36 (Supplementary Fig. 6). To confirm the H₂O₂ generation pathway, control experiments under different quenching conditions were performed. In all cases, a significant decrease in H₂O₂ yield was observed, suggesting that e^-, h⁺, •O₂^-, and •OH were all involved in the process (Supplementary Fig. 7). The corresponding reactions were described below:

$$\mathrm{Co}\left[{\mathrm{C}}_4{\mathrm{H}}_4{\mathrm{N}}_2\right]{\mathrm{V}}_2{\mathrm{O}}_6{\to }^{{\mathrm{Dynamic}P}_{\mathrm{flexo}}}{\mathrm{e}}^{-}+{\mathrm{h}}^{+}$$ 1

$${\mathrm{O}}_2+{\mathrm{e}}^{-}\to \cdot {\mathrm{O}}_2^{-}$$ 2

$${\mathrm{H}}_2\mathrm{O}+{\mathrm{h}}^{+}\to \cdot \mathrm{OH}+{\mathrm{H}}^{+}$$ 3

$$\cdot {\mathrm{O}}_2^{-}+{\mathrm{e}}^{-}+2{\mathrm{H}}^{+}\to {\mathrm{H}}_2{\mathrm{O}}_2$$ 4

$$2\cdot \mathrm{OH}\to {\mathrm{H}}_2{\mathrm{O}}_2$$ 5

The above e^- and H₂O₂ also reacted with UO₂²⁺, thereby altering the speciation of uranyl. The e^- could reduce UO₂²⁺ to UO₂²⁰. Subsequently, the resulting UO₂ was transformed into (UO₂)O₂·2H₂O through its reaction with H₂O₂²¹. In addition, H₂O₂ could directly react with UO₂²⁺, leading to the formation of (UO₂)O₂·2H₂O^18,51. Accordingly, the rate of flexocatalytic uranyl conversion could also be inferred to depend on the concentration of dissolved O₂, as it affected the generation of •O₂^- and the subsequent redox reactions. (Supplementary Fig. 8, Supplementary Table 3). The reactions responsible for altering the occurrence state of uranyl were listed below:

$${\mathrm{UO}}_2^{2+}+{\mathrm{e}}^{-}\to {\mathrm{UO}}_2\left(\mathrm{s}\right)$$ 6

$${\mathrm{UO}}_2+{\mathrm{H}}_2{\mathrm{O}}_2+2{\mathrm{H}}_2\mathrm{O}\to \left({\mathrm{UO}}_2\right){\mathrm{O}}_2\cdot 2{\mathrm{H}}_2\mathrm{O}\left(\mathrm{s}\right)+2{\mathrm{H}}^{+}$$ 7

$${\mathrm{UO}}_2^{2+}+{\mathrm{H}}_2{\mathrm{O}}_2+2{\mathrm{H}}_2\mathrm{O}\to \left({\mathrm{UO}}_2\right){\mathrm{O}}_2\cdot 2{\mathrm{H}}_2\mathrm{O}\left(\mathrm{s}\right)+2{\mathrm{H}}^{+}$$ 8

To identify the active species involved, quenching experiments and electron paramagnetic resonance (EPR) measurements were performed. In the quenching experiments, sacrificial agents AgNO₃, (NH₄)₂C₂O₄, p-BQ, and TBA were used to scavenge e^-, h⁺, •O₂^-, and •OH, respectively. The results revealed that e^-, •O₂^-, and •OH were the primary reactive species responsible for uranyl removal during the flexocatalytic process (Fig. 4f, Supplementary Table 4). Meanwhile, EPR spectra showed that the signals of •O₂^- and •OH increased with ultrasonic time, indicating enhanced generation of these reactive species (Fig. 4g). Quantitative analysis indicated that the concentration of H₂O₂ increased with ultrasonic irradiation time (Fig. 4h, Supplementary Fig. 9). Notably, the H₂O₂ concentration (1360.33 μmol·g^-1) in deionized water was significantly higher than that (594.63 μmol·g^-1) in the uranyl solution. This further suggested that H₂O₂ was effectively generated and played a crucial role in the conversion of uranyl speciation. Based on the above analysis, the overall reaction pathway was proposed, as illustrated in Fig. 4i. The solidification of U(VI) under flexocatalytic conditions resulted from reactions with active radicals, including e^-, •O₂^-, •OH, and H₂O₂.

Flexocatalytic uranyl performance

In aqueous systems, soluble uranyl existed in various forms, which were governed by the pH conditions⁵². Therefore, pH-dependent experiments on the flexocatalytic removal of uranyl using Co[C₄H₄N₂]V₂O₆ were conducted. Co[C₄H₄N₂]V₂O₆ exhibited good acid-base resistance over a wide pH range of 1.50–12.00 (Supplementary Fig. 12). Meanwhile, Co[C₄H₄N₂]V₂O₆ displayed flexocatalytic activity throughout the tested pH range (Fig. 5a, Supplementary Table 7). The catalytic activity peaked at pH = 4.01, where a uranium removal rate of 99.01% was achieved. Notably, Co[C₄H₄N₂]V₂O₆ also exhibited flexocatalytic capability in removing UO₂²⁺ even under conditions (pH > 7), where uranyl predominantly existed as anionic carbonate complexes such as UO₂(CO₃)₃^4- and UO₂(CO₃)₂^2- 53. These results confirmed that Co[C₄H₄N₂]V₂O₆ exhibited flexocatalytic applicability for uranyl removal across a broad pH range. Ultrasonic power was a critical factor in determining the practical applicability of flexocatalysis. Previous studies reported that inorganic materials often required high power (>200 W) to achieve high catalytic activity^3,18,54. The catalytic capability of Co[C₄H₄N₂]V₂O₆ under varying ultrasonic power was systematically investigated. Co[C₄H₄N₂]V₂O₆ exhibited a notable uranium removal ability at 20 W (Supplementary Fig. 14, Supplementary Table 8). Moreover, the uranyl removal efficiency increased progressively with increasing ultrasonic power (Fig. 5b). This improvement resulted from the higher external force generated by stronger ultrasound, which induced more pronounced structural deformation and larger strain gradients, thereby enhancing the catalytic activity¹⁷. Meanwhile, the flexocatalytic activity observed at low ultrasonic power was primarily attributed to the inorganic chains bridging the metal-organic chains of Co[C₄H₄N₂]V₂O₆. This further confirmed that the construction of inorganic-organic materials could enhance the flexoelectronic response.

Fig. 5. Flexocatalytic uranyl removal performance of Co[C₄H₄N₂]V₂O₆.

a Uranyl removal rate through flexocatalysis under different pH conditions. The pseudo-primary reaction kinetic curves of flexocatalysis b with varying catalyst sizes and c under different ultrasonic powers. k represented the reaction rate constant. The R² values at 20, 50, 100, 200, and 300 W were 0.9516, 0.9577, 0.9994, 0.9869, and 0.9941, respectively. For the intervals 50-100 mush, 100-200 mush, and 200−300 mush, the R² values were 0.9923, 0.9919, and 0.9975, respectively. d Reusability performance of the flexocatalyst. e Uranyl removal rate via flexocatalysis under high salinity conditions. f Schematic illustration showing the dynamic experimental setup for flexocatalysis. Dynamic uranyl capture performance in g U-contaminated potable water, h U-contaminated seawater, and i uranium mine wastewater.

The effect of catalyst size on uranyl removal by Co[C₄H₄N₂]V₂O₆ was also investigated. Catalysts of all tested sizes exhibited high uranyl removal rates (≥93.78%) (Supplementary Fig. 15, Supplementary Table 9). Meanwhile, uranium removal efficiency increased with decreasing catalyst size (Fig. 5c). The significant size dependence arose from smaller particles, which generated larger strain gradients at the micro-nano scale under deformation, thereby enhancing flexoelectric polarization and catalytic performance⁵⁵. The influence of catalyst dosage (5, 10, 20, 30, and 50 mg) on uranyl removal from 20 mL of wastewater was further examined. All tested dosages effectively catalyzed uranyl, with the optimal performance observed at a 20 mg dosage, achieving a uranium removal rate of 99.07% (Supplementary Fig. 16, Supplementary Table 10). Higher dosages did not lead to superior performance, likely due to excessive catalyst accumulation, which formed high electric fields at the surface, subsequently affecting radical generation⁵⁶. Moreover, the nuclear wastewater generally contained varying concentrations of uranium. The flexocatalytic ability of Co[C₄H₄N₂]V₂O₆ was evaluated for different uranium concentrations (1.96-95.48 ppm). Co[C₄H₄N₂]V₂O₆ exhibited excellent removal rates, exceeding 97% across all tested uranium concentrations (Supplementary Fig. 17, Supplementary Tables 11–16). The recycling performance of Co[C₄H₄N₂]V₂O₆ was also investigated. Co[C₄H₄N₂]V₂O₆ exhibited excellent cycling stability, maintaining over 97% uranium removal rate throughout five consecutive cycles (Fig. 5d, Supplementary Table 17). A series of experiments demonstrated that Co[C₄H₄N₂]V₂O₆ showed exceptional flexocatalytic performance for uranyl removal, markedly surpassing previously reported piezocatalysts (Supplementary Table 18).

Nuclear wastewater containing uranium frequently exhibited high salinity. Therefore, the flexocatalytic performance of Co[C₄H₄N₂]V₂O₆ was evaluated in simulated high-salinity solutions to assess its effectiveness under different ionic environments (Fig. 5e, Supplementary Table 19). In the U + 1.5 M NaCl system, the material exhibited excellent performance, reducing uranium concentration from 2.13 ppm to 250.3 ppb (88.25% removal). In the U + 1.5 M NaHCO₃ system, the uranium concentration was reduced from 2.08 ppm to 438.5 ppb, corresponding to a 78.92% removal. Similarly, in the U + 1.5 M MgCl₂ system, the uranium concentration decreased from 2.12 ppm to 486.0 ppb, achieving an 77.08% removal. In the U + 1.5 M CaCl₂ system, Co[C₄H₄N₂]V₂O₆ still maintained significant catalytic activity, lowering the uranium concentration from 2.03 ppm to 714.2 ppb (64.82% removal). These results confirmed that Co[C₄H₄N₂]V₂O₆ retained efficient flexocatalytic activity toward uranyl species under high-salinity conditions. Additionally, in seawater environments, pentavalent vanadium species can be converted to VO₂⁺ during coordination, thus potentially competing with UO₂²⁺ for active sites^52,57,58. Therefore, uranium removal experiments were carried out in the presence of V(V) at various V/U molar ratios (0.94, 2.87, 5.33, 6.84, and 9.36). Co[C₄H₄N₂]V₂O₆ maintained high uranium removal rates (>90%) (Supplementary Fig. 19, Supplementary Table 20), suggesting that the presence of V(V) had no significant impact on the uranyl removal performance.

Due to the excellent catalytic performance, the uranyl removal capability by Co[C₄H₄N₂]V₂O₆ in various water matrices (contaminated potable water, contaminated seawater, and uranium mining wastewater) was further investigated using a dynamic experimental system. In this setup, a circulating water pump was used to maintain ambient room temperature (Fig. 5f). The catalyst served as the stationary phase and was packed into a chromatography column (bed volume = 1.71 mL), which was placed inside an ultrasonic cleaner. Uranium-containing water was then pumped through the column using a peristaltic pump. Results indicated that Co[C₄H₄N₂]V₂O₆ exhibited continuous catalytic capability toward uranyl species, achieving outstanding uranium removal performance across all tested water matrices (Supplementary Data 1–3). When treating 1017 bed volumes of contaminated potable water, Co[C₄H₄N₂]V₂O₆ achieved a uranium removal rate of 98.57-99.26%. The uranium concentration decreased from 2.07 ppm to 15.4-29.7 ppb, which was below the acceptable limit of 30 ppb for potable water defined by the US Environmental Protection Agency (Fig. 5g)⁵⁹. During the treatment of 959 bed volumes of contaminated seawater, Co[C₄H₄N₂]V₂O₆ similarly reduced the uranium concentration (2.16 ppm) to 18.5-29.7 ppb (98.63-99.14% removal) (Fig. 5h). More importantly, during the treatment of 935 bed volumes of uranium mining wastewater, Co[C₄H₄N₂]V₂O₆ could also lower the uranium concentration from 5.64 ppm to 194.5-288.9 ppb (94.88-96.55% removal), meeting the discharge limit (300 ppb) regulated by China’s Ministry of Ecology and Environment (MEE) (Fig. 5i)⁶⁰. These results clearly indicated that the flexocatalyst Co[C₄H₄N₂]V₂O₆ exhibited strong potential for the remediation of U-contaminated potable water, uranium extraction from seawater, and the industrial treatment of uranium mining wastewater. Furthermore, this study was the first to use the dynamic experimental system to evaluate mechanical catalysis performance, providing a reference for its application in other contexts.

To further consolidate extensive parameter-dependent experimental observations into unified, generalizable operational principles, an interpretable data-driven analysis was conducted. This approach followed recent advances that demonstrated data-driven and AI-assisted frameworks could effectively integrate multivariate experimental spaces and identify governing factors beyond empirical trial-and-error optimization^61–63. By considering all experimentally accessible variables in this study, Shapley additive explanations (SHAP) analysis identified reaction time as the most influential parameter, followed by initial uranyl concentration, initial pH, and ultrasonic power (Supplementary Fig. 21a). This indicated that the flexocatalytic rate was predominantly governed by cumulative reaction duration and available energy input, whereas catalyst dosage, particle size, and atmosphere played secondary roles. The SHAP dependence of reaction time showed a monotonic increase with saturation (Supplementary Fig. 21b), reflecting the cumulative generation of reactive species and the progressive conversion of uranyl speciation during flexocatalytic treatment. In contrast, higher initial uranyl concentrations led to lower predicted removal rates. At a fixed reaction time, the amount of reactive species generated was insufficient to fully treat the increased uranyl load, which agreed with the experimentally observed concentration-dependent removal trends. These results not only validated the interpretability of the model but also provided guiding principles for the future optimization of flexocatalytic uranyl removal processes.

Discussion

To summarize, we developed a novel strategy to enhance the flexoelectric response of centrosymmetric materials by bridging inorganic chains with metal-organic chains. Co[C₄H₄N₂]V₂O₆ was shown to be an excellent flexocatalyst for uranyl separation. The bridging approach endowed the cantilever-like microsheets of Co[C₄H₄N₂]V₂O₆ with facile deformability and a high dielectric constant, both of which contributed to its excellent flexocatalytic polarization capability. Under ultrasonic stimulation, dynamic polarization enabled Co[C₄H₄N₂]V₂O₆ to achieve outstanding uranyl-separation performance compared with previously reported piezocatalysts. The advantages of Co[C₄H₄N₂]V₂O₆ for flexocatalytic uranyl separation included (a) straightforward synthesis, (b) high catalytic activity under challenging conditions (broad pH range and high-salinity environment), (c) low ultrasonic power requirement, (d) excellent reusability, and (e) superior uranyl separation ability in various water matrices under a dynamic experimental system. We also elucidated the pathway by which soluble uranyl was transformed into insoluble (UO₂)O₂·2H₂O. The active species (e.g., e⁻, •O₂⁻, •OH, and H₂O₂), generated by dynamic polarization, played a key role in driving the solidification of U(VI). This work not only introduced an approach to fabricate high-performance flexocatalysts and established the structure-property relationship governing flexocatalytic performance, but also provided a new paradigm for the remediation of U-contaminated potable water, uranium extraction from seawater, and the industrial treatment of uranium mining wastewater. Moreover, the demonstrated effectiveness and universality of flexocatalysis highlighted its potential for mechanocatalytic applications across diverse fields.

Methods

Materials and flexocatalyst preparation

The detailed information on chemicals and reagents was provided in the Supplementary Note 1. The bridging hybrid material Co[C₄H₄N₂]V₂O₆ was synthesized through a modified solvothermal method⁶⁴, which yielded bouquet-like crystals with an efficiency of approximately 85%. The specific synthesis procedure was as follows. 0.18 g of V₂O₅ was added to 20 mL of deionized water and stirred to form an orange-yellow mixture. 0.09 g of LiOH·H₂O was added, and the mixture was stirred for 30 min to yield a light brown transparent solution. 0.42 g of CoSO₄·7H₂O and 0.12 g of pyrazine were sequentially added and homogenized, followed by pH adjustment to 6.70 using dilute H₂SO₄. The orange-yellow mixture was transferred into a 40 mL Teflon-lined autoclave and subjected to hydrothermal treatment at 135 °C for 24  h. After cooling to room temperature, the product was vacuum-filtered, washed several times with deionized water and ethanol, and then dried overnight in a vacuum oven at 60 °C.

Experimental procedures and calculation methods

Catalyst characterizations were provided in the Supplementary Note 2. During the flexocatalytic experiments, ultrasonic waves were used as the mechanical excitation source. The temperature was maintained at room temperature by a circulating water system, and all experiments were conducted in the dark. Detailed flexocatalytic procedures were described in the Supplementary Note 3. Periodic density functional theory (DFT) calculations were performed using the Vienna Ab initio Simulation Package (VASP 6.4) to investigate the electronic structure of the catalyst⁶⁵. The Cauchy stress-strain method was employed to calculate the mechanical properties⁶⁶, and the Voigt-Reuss-Hill (VRH) approximation was used to determine the bulk modulus (B), Young’s modulus (E), and shear modulus (G)⁶⁷. Electron-phonon coupling calculations were applied to obtain the dielectric constants⁶⁸. Finite element method (FEM) simulations were performed in COMSOL Multiphysics 6.3 to analyze the flexoelectric response¹⁵. Detailed simulation methods were described in in the Supplementary Note 4. An interpretable data-driven analysis was conducted to identify the key operational factors governing flexocatalytic uranyl removal. A gradient boosting regression model was used to capture nonlinear relationships between experimental parameters and removal performance, with model interpretability achieved via SHAP to quantify parameter contributions at both the global and local levels. Detailed methods were provided in Supplementary Note 5.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

C.M. conceived the concept of this work, designed the experiments, and carried out theoretical calculations. D.F. performed most of the experiments and wrote the manuscript. C.L. and Z.C. conducted characterization measurements. X.W. (the sixth author), B.C., and B.H. helped with data interpretations. Z.S. analyzed the experimental data using a data-driven and AI-assisted framework. Z.Z., Y.L., and X.W. (corresponding author) supervised the project and discussed the results. All authors commented on the manuscript.

Peer review

Peer review information

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

Data availability

All data supporting the findings of this study are available within the article and Supplementary Information/Source data. Source data are provided with this paper.

Code availability

The code used for the machine learning analysis is available from the corresponding author upon request.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-72085-5.