Self‐Healing Pd(II) Cage Metallogels for Iodine Capture in Water and Heterogeneous Photocatalysis

ABSTRACT

Adaptive multifunctional systems that integrate responsiveness, environmental remediation, and catalysis within a single platform remain exceptionally rare. Herein, two carbazole‐derived metal‐organic cages (MOCs) were constructed through coordination‐driven self‐assembly and subsequently transformed into fibrous metallogels (MOC1G and MOC2G) in mixed DMSO/H₂O media through cooperative π−π stacking and dynamic supramolecular interactions. The resulting metallogels display reversible gel−sol transitions, interconnected nanofibrous architectures, and pronounced viscoelastic behavior. Rheological analyses revealed dominant elastic characteristics with rapid recovery after large mechanical deformation. Remarkably, the metallogels exhibit autonomous self‐healing without external stimuli, restoring their structural integrity within ∼18 min for MOC1G and ∼14 min for MOC2G after mechanical damage. Temperature‐dependent DLS and DSC studies further confirmed reversible thermally driven assembly−disassembly processes. Beyond their adaptive soft‐matter properties, the discrete MOCs demonstrated highly efficient capture of iodine species from aqueous media, achieving uptake capacities up to 3.01 g g^‒1 together with excellent selectivity and fast adsorption kinetics under both static and continuous‐flow conditions. In addition, the cages function as visible‐light‐responsive heterogeneous photocatalysts for aerobic sulfide oxidation, affording sulfoxides under mild conditions. This work establishes the first example of cage‐based metallogels that exhibit adaptive and autonomous self‐healing behavior while integrating environmental remediation and photocatalytic functionality within a single supramolecular platform.

Two carbazole‐derived metal‐organic cages hierarchically self‐assemble into supramolecular metallogels that integrate autonomous self‐healing, reversible stimuli responsiveness, high‐capacity aqueous iodine capture, and visible‐light‐driven photocatalytic sulfide oxidation within a single multifunctional platform.

1. Introduction

Adaptive materials that can recover from mechanical damage while responding to external stimuli are of considerable interest in soft‐matter science [1, 2]. Biological tissues provide a compelling model because they combine self‐repair, environmental responsiveness, and mechanical integrity within dynamic networks [3, 4, 5, 6]. Inspired by these features, synthetic self‐healing gels have been developed using reversible dynamic covalent bonds and noncovalent interactions, allowing damaged networks to reorganize and recover their structure and function [7, 8, 9, 10]. Incorporating thermoresponsiveness into such systems can further regulate network assembly and disassembly, enabling reversible gel−sol transitions and stimulus‐assisted healing [11, 12, 13]. Nevertheless, simultaneously achieving rapid self‐healing, reversible thermal responsiveness, and robust mechanical properties in a simple and modular material remains challenging [14]. Supramolecular gels are particularly attractive in this regard because their networks are sustained by dynamic interactions such as hydrogen bonding, π−π stacking, hydrophobic interactions, van der Waals forces, and metal−ligand coordination [15, 16, 17, 18]. The reversibility of these interactions enables structural adaptation while providing access to stimuli‐responsive behavior and diverse functional properties [17, 19, 20, 21, 22].

Among these systems, metallogels formed through coordination‐driven assembly offer an additional level of molecular control because metal−ligand coordination can define both the primary molecular architecture and the higher‐order network [23, 24]. Metal−organic cages (MOCs) are especially appealing building blocks because they combine well‐defined coordination geometries with enclosed cavities for molecular recognition and host−guest chemistry [25]. Their cavities have been extensively harnessed for the encapsulation of anions [26], neutral molecules [27], radical initiators [28], and pharmaceutical compounds [29], enabling advances in molecular recognition, photophysics, gas adsorption, and catalysis [30, 31, 32, 33, 34]. However, most MOCs are studied as discrete molecular hosts, whereas their use as molecular building blocks for adaptive supramolecular gels remains comparatively underexplored [35, 36, 37]. In particular, Pd₂L₄ cages provide a useful platform for this purpose because their well‐defined cationic cavities can be combined with ligand‐derived functionality and their molecular surfaces can participate in intermolecular interactions. Thus, converting such discrete cages into higher‐order gels offers an opportunity to couple molecular‐scale host–guest properties with macroscopic self‐healing and stimuli‐responsive behavior.

This molecular‐to‐material connection is also attractive for environmental applications [38]. Iodine is widely used across medical, industrial, agricultural, and food‐processing sectors [39], yet radioactive isotopes such as ¹²⁹I and ¹³¹I, produced during nuclear fuel reprocessing, create serious environmental and health hazards. The extremely long half‐life of ¹²⁹I (∼1.6 × 10⁷ years) and the high radiotoxicity of ¹³¹I make their removal from water particularly important [40, 41]. Various porous materials−including MOFs, COFs, HOFs, and porous organic polymers−have been explored for iodine capture [42, 43, 44, 45, 46, 47]. However, many of these systems demonstrate high adsorption capacities only under static conditions, whereas practical water‐treatment processes require efficient iodine removal under continuous flow and at low concentrations [48, 49]. Materials capable of capturing iodine efficiently under such realistic conditions remain limited. The cationic cavities and π‐rich surfaces of MOCs provide a potentially useful molecular environment for stabilizing iodine species through electrostatic, confinement, and π‐interaction/charge‐transfer effects.

The same cage architecture can, in principle, provide a second function through its photophysical properties [50, 51]. Incorporation of heavy metal centers such as Pd(II) promotes intersystem crossing (ISC) and facilitates the generation of reactive oxygen species (ROS) [52, 53, 54], while carbazole‐based chromophores are known to stabilize long‐lived triplet excited states required for ROS formation [55, 56]. These features make metal‐organic cages attractive candidates for photocatalytic aerobic oxidation reactions, such as the conversion of sulfides to sulfoxides an important transformation in organic synthesis and pharmaceutical chemistry [53, 57, 58]. Importantly, the possibility of combining host−guest recognition and photocatalysis within the same molecular cage provides a direct design route toward multifunctional supramolecular materials. Despite advances in cage‐based adsorption, photocatalysis, and adaptive gels as separate fields, systems that integrate cage‐based hierarchical assembly, self‐healing behavior, aqueous iodine capture, and photocatalytic activity within one platform remain rare.

Herein, we report a multifunctional metal−organic cage platform that integrates thermoresponsive metallogel formation, efficient iodine capture from aqueous environments, and heterogeneous photocatalysis within a single architecture. Two carbazole‐derived Pd₂L₄ cages, MOC1 and MOC2, were constructed through coordination‐driven self‐assembly and subsequently organized into fibrous metallogels in a DMSO/H₂O medium. The resulting gels (MOC1G and MOC2G) exhibit reversible gel−sol transitions and autonomous self‐healing, demonstrating how discrete cage units can be translated into adaptive macroscopic materials. The molecular cavities of the corresponding MOCs enable efficient capture of iodine species from aqueous media through π−iodine, and electrostatic interactions, exhibiting rapid adsorption kinetics (0.5 g mg⁻¹ min⁻¹), excellent dynamic flow‐through performance with breakthrough volumes reaching 8.2 L g⁻¹, and near‐quantitative iodine removal from dilute aqueous solutions, reducing residual iodine concentrations to the low ppb range. In addition, the synergistic combination of Pd(II) centers and carbazole chromophores enables efficient ROS‐mediated aerobic oxidation of sulfides under blue light, affording high conversions (≥70%) and excellent selectivity (∼97%) toward sulfoxides under ambient conditions while completely suppressing sulfone formation. Comparison of MOC1 and MOC2 further allows the influence of ligand substitution on assembly, iodine binding, and photocatalytic performance to be examined within a common structural framework. Thus, this work establishes a unified cage‐based strategy that links molecular architecture to hierarchical soft‐matter assembly, environmental remediation, and photocatalysis.

2. Results and Discussion

2.1. Synthesis and Characterization

The carbazole‐derived ligands L1 and L2 were prepared through Ullmann coupling reactions between the corresponding dibromo‐carbazole precursors bearing methyl or phenyl substituents and imidazole, affording the desired ligands in good yields (Schemes S1–S6). The synthesized ligands were comprehensively characterized by ¹H and ¹³C NMR spectroscopy together with high‐resolution mass spectrometry (HRMS) (Figures S1–S14). Subsequent coordination‐driven self‐assembly of ligands L1 and L2 with Pd(NO₃)₂ in DMSO using a 2:1 ligand‐to‐metal stoichiometry at room temperature furnished the corresponding metal−organic cages MOC1 and MOC2, respectively (Scheme 1). The assembly process was completed within 1 h, producing clear brown solutions, and the resulting cages were isolated as white solids upon addition of excess ethyl acetate. The formation of MOC1 and MOC2 was confirmed by NMR spectroscopy, ESI‐MS, and single‐crystal X‐ray diffraction analysis (Figure 1, Figures S15–S29). The formation of the Pd(II)‐directed self‐assembled cages was verified by ¹H NMR spectroscopy, which revealed significant downfield shifts of the imidazole resonances upon metal coordination. In the case of MOC1, the H_a resonance shifted from 8.27 to 8.97 ppm, while the H_c resonance moved from 7.79 to 8.05 ppm (Figure 1a). Likewise, MOC2 displayed pronounced downfield shifts, with the H_a proton moving from 8.26 to 9.60 ppm and the H_c proton from 7.76 to 8.13 ppm (Figures S12–S23), consistent with successful coordination to Pd(II).

SCHEME 1.

Illustration of the Pd(II)‐directed coordination‐driven self‐assembly of ligands L1 and L2 into discrete metal‐organic cage architectures (MOC1 and MOC2).

FIGURE 1.

(a) Partial ¹H NMR spectra (700 MHz, DMSO‐d₆ , 298 K) of ligand L1 and the corresponding self‐assembled cage MOC1. Single‐crystal X‒ray diffraction structures of (b) MOC1 and (c) MOC2 (counteranions, hydrogen atoms, and solvent molecules omitted for clarity; Pd, green; C, dark grey; N, blue). Experimental and simulated ESI‐MS isotopic distributions of representative cationic fragments for (d) MOC1 and (e) MOC2.

Detailed signal assignments were established using two‐dimensional ¹H‐¹H COSY and NOESY experiments (Figures S17–S20 and S25–S27). Furthermore, ¹H DOSY NMR measurements confirmed the formation of discrete and monodisperse cage species, as all proton signals exhibited identical diffusion coefficients. The measured diffusion constants of 7.85 × 10^{− 11} m² s^{− 1} for MOC1 and 6.55 × 10^{− 11} m² s^{− 1} for MOC2 correspond to hydrodynamic radii of 14.0 and 16.78 Å, respectively, supporting the formation of well‒defined cage architectures (Figures S21–S28). ESI‒MS further confirmed the successful formation of the cages by revealing characteristic multiply charged species corresponding to the assembled complexes. For MOC1, prominent peaks were observed at m/z = 795.1578 and 509.4470, assignable to [MOC1 + 2NO₃]²⁺ and [MOC1 + 1NO₃]³⁺, respectively (Figure 1d, Figure S22). Similarly, MOC2 displayed signals at m/z = 919.1877 and 592.1317 corresponding to [MOC2 + 2NO₃]²⁺ and [MOC2 + 1NO₃]³⁺, respectively (Figure 1e, Figure S29). In both cases, the experimentally observed isotopic distributions showed excellent agreement with the simulated patterns, confirming the formation of discrete Pd₂L₄ cage architectures.

Conclusive structural confirmation of the assemblies was obtained by single‐crystal X‒ray diffraction analysis. Single crystals suitable for X‒ray measurements were grown by slow vapor diffusion of 1,4‐dioxane and THF into DMSO solutions of MOC1 and MOC2, respectively (Figure 1b,c). MOC1 crystallizes in the monoclinic space group C2/c and features a dinuclear lantern‐type [Pd₂L₄] architecture in which each Pd(II) center adopts a square‐planar coordination geometry defined by four imidazole nitrogen donors from four independent ligands. The asymmetric unit consists of one Pd(II) ion and two ligand fragments, with a Pd···Pd separation of 9.985 Å (Figure 1b, Table S1). In contrast, MOC2 crystallizes in the triclinic space group $P\overline{1}$ and displays a closely related [Pd₂L₄] cage structure with an intermetallic Pd···Pd distance of 9.941 Å (Figure 1c, Table S2).

After confirming the successful formation of the cages, their thermal behavior, porosity, and solid‐state characteristics were systematically investigated. Thermogravimetric analysis (TGA) performed under a nitrogen atmosphere demonstrated that both cages remain thermally stable up to ~210 ^οC, with the initial weight loss corresponding to the release of encapsulated solvent molecules (Figure S53). The porosity of the materials was further evaluated by N₂ adsorption−desorption measurements at 77 K following activation at 110 ^οC under vacuum for 12 h to remove residual solvent molecules completely. Brunauer−Emmett−Teller (BET) analysis revealed surface areas of 42.30 m² g⁻¹ for MOC1 and 74.44 m² g⁻¹ for MOC2, with pore diameters of 1.52 and 1.70 nm, respectively, consistent with the intrinsic cavity dimensions of the cage architectures (Figures S30 and S31, Table S3). Powder X‐ray diffraction (PXRD) analysis indicated that both materials are predominantly amorphous in the solid state, as evidenced by their broad and weak diffraction patterns (Figure S99). Scanning electron microscopy (SEM) imaging revealed distinct morphologies for the two cages, where MOC1 exhibited aggregated block‐like structures, while MOC2 formed interconnected fibrous nanostructures (Figure S32). Furthermore, energy‐dispersive X‐ray spectroscopy (EDS) elemental mapping confirmed the uniform spatial distribution of all constituent elements (e.g., Pd, C, N and O) throughout the materials (Figures S33 and S34).

Having established the formation of discrete and structurally well‐defined Pd₂L₄ cages, we next considered how these molecular architectures could be translated into higher‐order soft materials. Importantly, the cage design incorporates two complementary structural features: the rigid, π‐conjugated carbazole surfaces provide an extended platform for intermolecular aromatic interactions, while the discrete Pd(II)‐directed architecture maintains a defined molecular cavity and dynamic coordination environment. We therefore anticipated that these cages could serve not only as isolated host molecules but also as molecular building blocks for hierarchical supramolecular assembly. In an appropriate solvent environment, cooperative π−π stacking, hydrophobic, electrostatic, and other reversible noncovalent interactions could organize the discrete cages into extended one‐dimensional and three‐dimensional networks. This molecular‐to‐macroscopic hierarchy provides a direct link between the structural precision of the individual cages and the adaptive properties of the resulting soft materials. We therefore investigated whether MOC1 and MOC2 could undergo further organization into self‐supporting metallogels and whether the molecular differences between the two cages would influence their resulting network structures and mechanical properties.

2.2. Gel Synthesis, Characterization, and Self‐Healing

Although discrete metal−organic cages (MOCs) are widely recognized for their defined cavities and host−guest capabilities, their utilization as building blocks for supramolecular metallogels remains comparatively unexplored. Motivated by the strong aggregation tendency of carbazole‐based systems [59, 60], we envisioned that the synthesized MOCs could further organize through hierarchical self‒assembly to generate extended gel networks. In appropriate solvent environments, cooperative noncovalent interactions−primarily π−π stacking and hydrophobic interactions were expected to promote intermolecular organization and ultimately driving their integration into extended, three‐dimensional gel frameworks. Consistent with this hypothesis, addition of an aqueous Pd(NO₃)₂ solution to DMSO solutions of ligands L1 or L2 in a 1:2 metal‐to‐ligand ratio, followed by gentle shaking and standing for 2 h, resulted in the formation of stable light‐brown supramolecular metallogels, MOC1G and MOC2G in a DMSO/H₂O mixed solvent system (3:2, v/v) (Figure 2a). The critical gelation concentration (CGC) was determined to be 2.2% (w/v), and successful gel formation was verified by the vial inversion method for both the MOCGs. Gelation was found to be highly dependent on the solvent composition, while the DMSO/H₂O mixture (3:2) was optimal, other solvents such as DMF, methanol, and acetonitrile failed to produce gels, highlighting the crucial influence of solvent polarity and solvation effects on the self‒assembly process (Table S4). The gelation process is attributed to the hierarchical aggregation of the discrete MOC units through multiple cooperative supramolecular interactions. In particular, the π−conjugated carbazole moieties within the cage framework facilitate strong intermolecular π−π stacking interactions, which are further reinforced by hydrogen‒bonding interactions involving solvent molecules and counter anions, together with anion‐mediated electrostatic interactions and dynamic solvent−ligand exchange processes [36, 37, 61]. The synergistic interplay of these noncovalent interactions drives the organization of discrete MOC1 and MOC2 cages into an interconnected three‐dimensional fibrous network, ultimately producing self‐supporting supramolecular metallogels. We performed control experiments using the pure carbazole‐based ligands in the absence of Pd(II) metal precursors, as well as experiments using the Pd(II) metal acceptor alone under otherwise identical conditions. In both cases, no gel formation was observed, demonstrating that neither the ligand nor the metal precursor individually is capable of generating the observed gel network (Figure S35).

FIGURE 2.

(a) Schematic illustration of supramolecular metallogel formation through Pd(NO₃)₂‒mediated complexation of ligands L1 and L2, together with the reversible thermally induced sol‐gel transition. TEM images of (b) MOC1G and (c) MOC2G showing the fibrous nanostructures of the metallogels. FE‒SEM images of (d) MOC1G and (f) MOC2G highlighting their interconnected gel‐network morphologies. AFM topographic images of (e) MOC1G and (g) MOC2G.

The supramolecular morphologies of MOC1G and MOC2G, originating from cage‐based hierarchical self‐assembly were comprehensively examined by FE‒SEM, TEM, and AFM analyses. FE‒SEM images show that both metallogels consist of interconnected three‐dimensional fibrous architectures characteristic of supramolecular gel networks (Figure 2d,f). MOC1G forms an entangled network of relatively uniform nanofibers with an average diameter of 127 ± 14 nm, generating an open and continuous framework capable of supporting gelation (Figure 2d, Figure S49a,b). In contrast, MOC2G exhibits comparatively thicker worm‐like fibers with an average diameter of 108 ± 13 nm arranged in a denser and more compact network, likely arising from enhanced π‐stacking interactions and steric effects associated with the phenyl substituents (Figure 2f, Figure S49c,d). TEM analysis further confirmed the presence of elongated and interconnected fibrous assemblies in both metallogels (Figure 2b,c). Consistent with these observations, AFM topographic and height analyses revealed narrower height distributions for MOC1G (103 ± 43 nm), whereas MOC2G displayed significantly thicker fibers and larger height variations (159 ± 35 nm), indicative of a more tightly packed network structure (Figure 2e,g, Figure S50). Three‐dimensional AFM reconstructions clearly visualized the cross‐linked fibrous morphology and hierarchical nanoscale organization of the metallogels. The distinct morphological differences between MOC1G and MOC2G are attributed to the methyl versus phenyl substitution on the carbazole scaffold, which modulates intermolecular packing, fiber bundling, and overall network organization. Collectively, these microscopic studies demonstrate that carbazole‐based discrete MOCs undergo hierarchical self‐assembly into adaptive nanofibrous metallogels with the observed morphological distinctions. In addition, EDS elemental mapping confirmed the uniform distribution of Pd, C, N, and O throughout the metallogel matrices (Figures S45 and S46).

Further evidence for the metallogel properties was obtained from FT‒IR spectroscopy. The stacked FT‐IR spectra of ligands L1 and L2 and their corresponding metallogels show that the characteristic aromatic C─H and C═C vibrational bands remain largely preserved after gel formation. In contrast, noticeable shifts were observed for the imidazole‐associated vibrations upon coordination to Pd(II). The C═N stretching bands shifted from 1653.17 cm⁻¹ (L1) and 1657.55 cm⁻¹ (L2) to 1652.69 cm⁻¹ (MOC1G) and 1655.58 cm⁻¹ (MOC2G), respectively, while the C−N stretching vibrations moved from 1308.46 to 1315.50 cm⁻¹ for MOC1G and from 1309.00 to 1312.57 cm⁻¹ for MOC2G (Figures S44–S51 and S52). These spectral shifts provide strong evidence for coordination of the imidazole nitrogen donors to the Pd centers and confirm the crucial role of metal−ligand interactions in stabilizing the supramolecular metallogel framework [37] We also performed additional spectroscopic investigations, including ESI‐MS, ¹H‐DOSY NMR, and variable‐temperature ¹H NMR spectroscopy on the metallogel material after dissolution (Figures S36‒S43). These additional data provide consistent evidence that the Pd₂L₄ cage units remain structurally intact upon incorporation into the gel network. In particular, the ESI‐MS and DOSY NMR results support the persistence of the discrete cage species, while the variable‐temperature ¹H NMR measurements further confirm the stability of the cage architecture under the investigated conditions. These results support our proposed hierarchical assembly model in which the preformed Pd₂L₄ cages act as discrete molecular nodes/building units that further organize through noncovalent interactions to generate the fibrous metallogel network, rather than undergoing decomposition or extensive structural reorganization during gel formation.

The mechanical behavior and gelation properties of the metallogels were systematically analyzed by oscillatory rheology, including amplitude and frequency sweep measurements, to evaluate their viscoelasticity, structural integrity, and mechanical resilience [21, 61]. Amplitude sweep experiments were performed over a strain range of 0.01%–100% at a fixed angular frequency of 10 rad s⁻¹ to probe the deformation tolerance of the supramolecular networks (Figure 3a, Figure S54). Within the linear viscoelastic region (LVR), both the storage modulus (G′) and loss modulus (G″) remained nearly invariant up to ~1% strain for both metallogels, indicating that the gel structures remain intact under small deformations. Upon increasing the strain, G′ decreased significantly while G″ initially remained relatively stable before declining, reflecting progressive disruption of the gel network. The crossover points, where G″ exceeded G′ and the gels transitioned from a solid‐like to a liquid‐like state, occurred at 31.6% strain for MOC1G and 46.5% strain for MOC2G, indicating distinct mechanical responses of the two gel systems and highlighting the enhanced elasticity of MOC2G. The viscoelastic properties were further investigated through frequency sweep measurements conducted within the LVR at a constant strain well below the critical strain (0.1%) (Figure 3b, Figure S55). In both metallogels, G′ remained consistently higher than G″ across the examined frequency range, confirming the predominately elastic and solid‐like nature of the supramolecular networks. Notably, MOC2G exhibited higher storage modulus values than MOC1G, indicating superior mechanical strength and rigidity. These rheological results demonstrate that both materials behave as viscoelastic supramolecular solids stabilized by reversible noncovalent interactions. The enhanced mechanical performance of MOC2G is likely associated with stronger intermolecular π–π interactions induced by the phenyl‒substituted carbazole units relative to the methyl‐substituted analogue.

FIGURE 3.

(a) Amplitude sweep rheological analysis of MOC2G, showing the strain‐dependent viscoelastic behavior. (b) Frequency‒dependent viscoelastic moduli (storage modulus, G′, and loss modulus, G′′) of MOC2G measured at a constant strain of 0.1 %. (c) Thixotropic behavior of MOC2G under a continuous step‐strain cycle (0.1 % → 50 % → 0.1 % strain). (d) DLS analysis of MOC2G at varying temperatures. (e) Photographs demonstrating the autonomous self‐healing behavior of MOC2G, showing spontaneous restoration of the gel within 14 min after mechanical damage.

Both metallogels also displayed pronounced thermoresponsive behavior characterized by reversible gel‐sol transitions. Visual inspections revealed that mild heating converted the self‐supporting gels into free‐flowing sols, whereas cooling restored the gel state together with its original mechanical integrity (Figure 2a). To further investigate this thermally responsive behavior, differential scanning calorimetry (DSC) and temperature‒dependent dynamic light scattering (DLS) studies were carried out. DSC thermograms of MOC1G and MOC2G exhibited distinct endothermic transitions corresponding to thermally induced gel‒to‒sol transformations (Figure S47) [62, 63]. MOC1G showed a transition in the range of ∼52 °C‒55 °C, while MOC2G exhibited a slightly higher transition temperature (∼54 °C‒60°C), suggesting enhanced thermal stability and stronger supramolecular interactions in the latter system. Temperature‐dependent DLS measurements further provided insight into the reversible nanoscale assembly behavior of the metallogels (Figure 3d, Figure S48). At lower temperatures (20 °C–30 °C), the number‐average size distributions were dominated by large aggregates with hydrodynamic diameters of ~300–500 nm, consistent with the presence of extended fibrous assemblies associated with the gel state. Upon heating to ∼45 °C‒55 °C, these hierarchical aggregates dissociated into smaller species, resulting in hydrodynamic diameters of ~30 nm for MOC1G and ~10 nm for MOC2G, characteristic of the sol state. Cooling the solutions restored the larger aggregate populations, with the hydrodynamic diameters returning to ∼400 nm at 20 °C, thereby confirming the reversible reconstruction of the supramolecular gel network (Figure 3d, Figure S48) [64, 65]. Overall, the DSC, and DLS studies establish that the metallogels undergo a fully reversible, thermally controlled sol‒gel transition driven by dynamic supramolecular interactions.

The metal–organic cage‐based gels display remarkable macroscopic self‐healing behavior, highlighting the dynamic and reversible nature of the underlying supramolecular network (Figure 3e, Figure S57). In the undisturbed state, the gels exist as homogeneous, self‒supporting monoliths, however, mechanical cutting generates a visible fracture, disrupting the continuous network and separating the material into two independent pieces. Upon bringing the fractured surfaces back into contact, the damaged interface gradually disappears without requiring any external stimulus, demonstrating spontaneous autonomous healing. Complete recovery of the gel network occurs within ~18 min for MOC1G and ~14 min for MOC2G, after which the gels fully regain their original shape and mechanical integrity (Figure 3e, Figure S57) [21, 66]. The self‒healing performance was further investigated by thixotropic rheological measurements using alternating strain cycles (0.1% → 50% → 0.1%) (Figure 3c, Figure S56). Application of high oscillatory strain (50%) disrupted the gel network, resulting in an immediate transition from gel to sol state, as indicated by the crossover of the storage and loss moduli (G″ > G′). Upon restoring the strain to 0.1%, the storage modulus rapidly recovered and again exceeded the loss modulus, with both G′ and G″ values returning close to their original magnitudes [21, 67]. This rapid and reproducible recovery demonstrates the remarkable ability of the metallogel to reconstruct its supramolecular network following mechanical disruption, resulting in excellent self‒healing efficiency and mechanical resilience. Additionally, we performed multiple self‐healing cycles and quantitatively evaluated the mechanical recovery of the gels. Specifically, frequency and amplitude‐sweep rheological measurements were carried out over three consecutive damage−healing cycles to compare the mechanical properties of the pristine and healed gels (Figures S58‒S61). The corresponding storage modulus (G′) and loss modulus (G″) were monitored before and after each healing cycle, and the quantitative recovery ratios were determined. In both cases, the gels exhibited excellent recovery of G′ and G″, demonstrating that the healing process restores not only the macroscopic integrity of the gel but also its mechanical properties to a high degree (Figures S58‒S61 and Table S5). Although self‐healing gels derived from polymers and low‐molecular‐weight gelators have been extensively investigated, the emergence of autonomous self‐healing behavior in discrete metal‐organic cage‐based architectures remains largely unexplored. Notably, this represents the first example of a metal‐organic cage system exhibiting adaptive autonomous self‐healing behavior, highlighting its promise for reusable and stimuli‒responsive soft‐matter applications.

The hierarchical assembly studies demonstrate that the discrete MOCs retain their molecular identity while gaining new collective properties through reversible intermolecular organization. Importantly, this hierarchical assembly does not eliminate the intrinsic molecular functionality of the cages; rather, it provides a route to integrate their well‐defined cavities and π‐rich surfaces into a macroscopic, recoverable material platform. This feature is particularly relevant for molecular recognition and environmental remediation, where the combination of accessible host‐guest sites and a robust, reusable material architecture is highly desirable. The cationic Pd₂L₄ cages contain electron‐rich carbazole surfaces within and around their confined cavities, providing complementary π‐rich and electrostatic environments for interaction with iodine species. We therefore reasoned that the same molecular characteristics responsible for the host‐guest character of the discrete cages could be exploited independently of the macroscopic gel network to achieve efficient iodine sequestration from water. Thus, having established the hierarchical assembly and adaptive behavior of the cage‐based soft materials, we next examined the intrinsic iodine‐binding capability of the discrete MOCs and evaluated whether their molecular recognition properties could translate into efficient and recyclable iodine‐removal platforms.

2.3. Iodine Adsorption From Aqueous Media

Radioactive iodine released during nuclear fuel reprocessing and energy generation represents a major environmental hazard, especially when iodine species enter industrial wastewater streams or natural water reservoirs [68]. The development of materials capable of rapidly removing iodine from aqueous environments is therefore of considerable importance for both environmental remediation and emergency response applications. Although a variety of porous adsorbents have been reported for iodine capture, most are primarily effective for iodine vapor adsorption or for uptake from nonpolar organic solvents such as hexane and cyclohexane [69]. In contrast, iodine sequestration from water remains particularly challenging because of the low solubility of iodine, the coexistence of multiple iodine species, and the presence of competing ions in aqueous media. To overcome these limitations, we envisioned that cationic metal–organic cages (MOCs) incorporating a high density of electron‒rich carbazole units could provide an effective platform for iodine capture through cooperative host‐guest and electrostatic interactions. The π‐electron‐rich carbazole framework together with the positively charged cage environment was expected to promote strong binding toward various aqueous iodine species, including I₂ and I₃ ⁻, thereby enabling efficient iodine sequestration under realistic aqueous conditions. To evaluate this hypothesis, adsorption studies were carried out using aqueous solutions of I₂ and I₃ ⁻ as separate source phases.

The iodine adsorption performance of the cages was initially evaluated using aqueous I₂ solutions. Upon addition of 2 mg of MOC1 or MOC2 to 3 mL of a 1.2 mm I₂ solution, the characteristic yellow coloration of iodine rapidly disappeared, producing a colorless solution within ~4 min, indicative of fast iodine uptake by both cages (Figures S63–S67). Kinetic analysis of the adsorption data revealed excellent agreement with a pseudo‒second‒order model (R²> 0.99), suggesting that the adsorption process is predominantly governed by chemisorption interactions (Figure S68). The calculated adsorption rate constants were 0.1463 g mg⁻¹ min⁻¹ for MOC1 and 0.2198 g mg⁻¹ min⁻¹ for MOC2, with both systems achieving iodine removal efficiencies exceeding 99% (Figure S68 and Table S6). The equilibrium adsorption behavior was further analyzed using Langmuir and Freundlich isotherm models. The adsorption data showed a significantly better fit with the Freundlich model (R² = 0.93−0.98) than with the Langmuir model (R² = 0.80−0.82), indicating a heterogeneous multilayer adsorption process on the cage surfaces (Figures S71 and S72, Table S7). The maximum iodine uptake capacities were determined to be 2.07 g g⁻¹ for MOC1 and 2.45 g g⁻¹ for MOC2 (Figure 4e, Figure S69), highlighting the enhanced adsorption capability of MOC2. To examine adsorption selectivity under competitive conditions, iodine uptake studies were performed in the presence of excess interfering anions, including F⁻, Cl⁻, Br⁻, HCO₃ ⁻, NO₃ ⁻, and SO₄ ²⁻. Remarkably, even in the presence of 100 equivalents of competing anions, iodine removal efficiencies remained exceptionally high at 94% for MOC1 and 95% for MOC2 (Figure 4d). The corresponding distribution coefficients (K _d) remained on the order of 10⁴ mL g⁻¹ even in the presence of equimolar competing anions, demonstrating the strong binding affinity and excellent selectivity of the cages toward iodine species (Figure S70) [70]. Considering the environmental relevance of radioactive iodine contamination in marine systems during nuclear accidents, the adsorption performance of the cages was also evaluated using iodine‐spiked seawater samples. Under these conditions, MOC1 and MOC2 exhibited uptake capacities of 1.67 and 1.98 g g⁻¹, respectively, reaching equilibrium after 18 and 24 h (Figure 4c). These results demonstrate that the cages maintain high iodine adsorption efficiency even under chemically complex and high salinity conditions, highlighting their potential applicability for practical environmental remediation.

FIGURE 4.

(a) Time‐resolved UV–vis spectra monitoring the progressive removal of I₃ ⁻ following the addition of MOC2. (b) Non‐linear pseudo‐second‐order kinetic fit describing I₃ ⁻ adsorption by MOCs. (c) Time dependent adsorption profile for I₂ capture from seawater by the MOCs. (d) Selective static adsorption of I₂ by the MOCs in the presence of 100 equiv. of competing anions. (e) Comparative adsorption performance of both MOCs toward I₂ and I₃ ⁻ in various aqueous environments. (f) Effect of individual competing anions (1−200 equiv.) on I₃ ⁻ uptake by MOC2.

Beyond molecular iodine, triiodide (I₃ ⁻) constitutes one of the major iodine species encountered in aqueous nuclear waste streams and contaminated groundwater. Upon addition of 2 mg of each cage to 3 mL of an aqueous 0.5 mm I₃ ⁻ solution, rapid iodine removal was observed.

Notably, MOC2 achieved >84% I₃ ⁻ removal within only ~30 s, whereas MOC1 required ~60 s to attain comparable efficiency (Figure 4a, Figures S74‒S76). Kinetic analyses revealed that the adsorption process follows a pseudo‒second‒order model with high correlation coefficients, indicating a predominantly chemisorption‒controlled mechanism (Figure 4b, Figure S77). Importantly, the adsorption rate exhibited by MOC2 ranks among the fastest reported for aqueous iodine capture systems (Table S12). Maximum uptake experiments performed under excess I₃ ⁻ conditions (300 mg I₂ and 600 mg KI in 20 mL water) revealed exceptionally high adsorption capacities of 2.56 and 2.32 g g⁻¹ for MOC1, and 3.01 and 2.87 g g⁻¹ for MOC2 in the absence and presence of one equivalent of competing anions, respectively (Figure 4e). These uptake values are competitive with state‐of‐the‐art COF, MOF, and POP based iodine sorbents (Table S11). We have also performed I₃ ⁻ adsorption experiments over a range of pH values (pH 4–9) using the cage materials. The adsorption profiles showed no significant variation in I₃ ⁻ uptake or adsorption behavior across this pH range, indicating that the iodine sequestration capability of the cages is relatively insensitive to moderate changes in the solution pH (Figures S82‒S84). Adsorption experiments at higher pH were not feasible because I₃ ⁻ undergoes conversion to IO₃ ⁻ under strongly alkaline aqueous conditions. Remarkably, even in the presence of 200 equivalents of interfering anions (F⁻, Cl⁻, Br⁻, HCO₃ ⁻, NO₃ ⁻, and SO₄ ²⁻), both cages retained outstanding I₃ ⁻ removal efficiencies of ~95% for MOC1 and ~96.5% for MOC2 (Figure 4f, Figure S80). The corresponding distribution coefficients remained on the order of 10⁴ mL g⁻¹, comparable to values obtained in the absence of competing ions, highlighting the exceptional selectivity and binding affinity of the cages toward iodine species in highly competitive aqueous environments (Figure S81) [70]. Additionally, we freeze‐dried the as‐prepared metallogels to preserve their three‐dimensional gel morphology and subsequently evaluated their iodine adsorption performance in aqueous media under conditions identical to those used for the as‐synthesized cage powders (Figures S85–S88). Interestingly, the freeze‐dried metallogels exhibited essentially the same iodine adsorption behavior and adsorption performance as the as‐synthesized cage powders, with no significant difference in the adsorption kinetics or overall uptake (Figures S86 and S87, S89 and S90; Table S10). These results suggest that, although the hierarchical gel morphology provides the material with macroscopic self‐healing and soft‐matter properties, the intrinsic iodine‐binding capability is primarily governed by the molecularly defined cage cavities and their associated host−guest interactions. Overall, these findings establish carbazole‐based metal‒organic cages as highly efficient, selective, and robust adsorbents for the removal of both I₂ and I₃ ⁻ from complex aqueous systems, including seawater, highlighting their strong potential for radioactive iodine remediation in environmental and nuclear applications.

Although MOC1 and MOC2 exhibited excellent iodine uptake under static conditions, practical water‐treatment applications require efficient adsorption under continuous flow conditions. To evaluate their applicability in realistic remediation processes, dynamic flow‒through column experiments were performed, demonstrating that both cages function as highly effective heterogeneous adsorbents for iodine sequestration from aqueous media (Figure 5d). For molecular iodine adsorption, a 1.2 mm aqueous I₂ solution was continuously passed through columns packed with the respective MOCs, leading to immediate decolorization of the effluent solution. Under these dynamic conditions, iodine removal efficiencies of ~98.0% for MOC1 and ~98.87% for MOC2 were achieved (Figure S94). Analysis of the breakthrough curves revealed dynamic adsorption capacities of 1.46 g g⁻¹ for MOC1 and 1.76 g g⁻¹ for MOC2, corresponding to breakthrough volumes of 4.8 and 5.8 L g⁻¹, respectively (Figure 5a). The dynamic adsorption performance was further investigated using aqueous triiodide (I₃ ⁻) solutions. In the absence of competing anions, MOC1 and MOC2 afforded high removal efficiencies of ~97% and ~99%, respectively (Figure S95). Importantly, even in the presence of one equivalent of mixed competing anions, both materials retained excellent adsorption performance, maintaining iodine removal efficiencies of ~96% for MOC1 and ~97.8% for MOC2 (Figure 5d). Breakthrough analysis showed that the adsorption columns could process up to 6.0 L g⁻¹ and 8.2 L g⁻¹ of I₃ ⁻ solution for MOC1 and MOC2, respectively, before saturation, corresponding to dynamic uptake capacities of 1.14 g g⁻¹ for MOC1 and 1.50 g g⁻¹ for MOC2 (Figure 5b, Figure S96). The reusability of the adsorbents was also assessed through repeated adsorption−desorption cycles using methanol as the regeneration solvent. Notably, MOC2 retained high iodine removal efficiency and adsorption capacity over at least three consecutive cycles, demonstrating excellent operational stability and recyclability under continuous flow conditions (Figure 5b). These findings highlight the strong potential of the MOC based systems, particularly MOC2, for practical iodine remediation in dynamic aqueous environments.

FIGURE 5.

(a) Dynamic breakthrough curves for I₂ adsorption by the MOCs under continuous flow conditions. (b) Recyclable breakthrough performance of MOC2 for aqueous I₃ ⁻ removal over three successive adsorption–desorption cycles. (c) Adsorption efficiency of the MOCs for ultratrace I₃ ^‒ removal from different aqueous media. (d) Schematic illustration of the continuous‐flow column adsorption process for various iodine species using the MOCs.

To evaluate the practical applicability of the cages under environmentally relevant low‒concentration conditions, dynamic adsorption studies were performed using 5 ppm I₃ ⁻ feed solutions prepared in distilled water, seawater, and distilled water containing one equivalent of competing anions. The residual iodine concentration in the treated effluent was determined using the leuco crystal violet colorimetric assay (Scheme S7) [49, 71]. Under these conditions, MOC1 and MOC2 effectively reduced iodine concentrations to 24 and 21 ppb, respectively, in distilled water, while residual concentrations of 32 and 47 ppb were obtained in seawater. In the presence of competing anions, the final iodine concentrations remained exceptionally low at 37 ppb for MOC1 and 35 ppb for MOC2 (Figure 5c). These residual iodine levels are close to naturally occurring background concentrations in environmental waters and are significantly below levels considered hazardous to human health [70]. Despite possessing similar cage dimensions and identical cationic charge, MOC2 consistently exhibited superior adsorption performance relative to MOC1. This enhancement is attributed to the phenyl substituent present in ligand L2, which increases the overall π‒electron density of the cage framework and strengthens π‒iodine interactions, thereby facilitating more effective iodine capture under continuous−flow conditions. Collectively, these findings demonstrate that the MOCs, particularly MOC2, function as highly efficient, selective, and recyclable heterogeneous adsorbents for the removal of trace iodine species from complex aqueous environments, highlighting their considerable promise for nuclear wastewater treatment and large‐scale environmental remediation applications.

The adsorption studies demonstrate that both MOC1 and MOC2 exhibits excellent affinity toward molecular iodine (I₂) and triiodide (I₃ ^‒) species in aqueous media. To gain mechanistic insight into the iodine capture process and elucidate the nature of the cage−iodine interactions, the iodine‐loaded materials were systematically analyzed using TGA, SEM−EDS, FT−IR, Raman spectroscopy, XPS, and Density functional theory (DFT) calculations. Thermogravimetric analysis of the iodine‐encapsulated cages revealed pronounced weight losses in the temperature range of 40 ^οC−180 ^οC under nitrogen atmosphere, corresponding to the thermal release of adsorbed iodine species (Figure S98) [72]. The magnitude of the mass loss closely matched the experimentally determined adsorption capacities, confirming substantial iodine incorporation within the cage motifs. Further evidence for strong iodine−cage interactions was obtained from FT−IR spectroscopy. Upon iodine adsorption, distinct shifts in several characteristic vibrational bands were observed for both cages. In particular, the aromatic C═C stretching vibrations shifted from 1516.04 and 1518.23 cm⁻¹ to 1513.19 and 1516.12 cm⁻¹, while the C─H stretching bands moved from 1339.89 and 1325.16 cm⁻¹ to 1333.08 and 1323.66 cm⁻¹ for MOC1 and MOC2, respectively (Figure S97).

These systematic spectral changes indicate direct involvement of the π‒rich carbazole framework in iodine binding through charge‒transfer and π−iodine interactions. Morphological changes associated with iodine uptake were investigated by SEM analysis. Pristine MOC1 and MOC2 exhibited relatively uniform morphologies with broad particle‒size distributions, whereas the iodine‐loaded materials adopted noticeably expanded and fluffy textures, consistent with significant iodine incorporation into the supramolecular assemblies (Figure S100). Correspondingly, EDS spectra of the iodine‐loaded cages displayed intense iodine signals in the 3.5−5.0 keV region, with signal intensities proportional to the iodine loading levels (Figures S101 and S102) [73]. Elemental mapping further confirmed a dense and homogeneous distribution of iodine throughout the cage materials, supporting uniform iodine−cage interactions. Raman spectroscopy provided direct insight into the chemical speciation of the adsorbed iodine species. The iodine‐loaded cages exhibited characteristic Raman bands at ∼106.60 and 107.83 cm⁻¹ corresponding to symmetric stretching vibrations of I₃ ⁻, while bands at ∼135.89 and 138.53 cm⁻¹ were assigned to asymmetric I₃ ⁻ stretching modes for MOC1 and MOC2, respectively (Figure 6c). Additional signals at ∼160.72 and 162.19 cm⁻¹ were attributed to the stretching of I₅ ⁻ species, indicating that the adsorbed iodine undergoes partial conversion into polyiodides within the cage environment. A weak broad feature centered around 215−220 cm⁻¹ further suggested the coexistence of molecular iodine within the MOC1 framework [74, 75]. These observations collectively indicate that the electron‐rich carbazole ligands and cationic cage cavities facilitate stabilization of polyiodide species through charge‐transfer interactions. X‐ray photoelectron spectroscopy (XPS) further clarified the electronic interactions between iodine and the cage framework. The I 3d spectra of iodine‐loaded cages exhibit characteristic doublets at 631.83 and 620.44 eV for I₂@MOC1 and 631.81 and 620.34 eV for I₂@MOC2, corresponding to I 3d_5/2, along with peaks at 630.23 and 618.81 eV for I₂@MOC1 and 630.14 and 618.93 eV for I₂@MOC2 assigned to I 3d_3/2 (Figure 6a, Figure S104). Quantitative deconvolution of the spectra revealed polyiodide compositions of ~64.8% I₃ ⁻ and ~35.2% I₅ ⁻ for I₂@MOC1, and ~63% I₃ ⁻ and ~37% I₅ ⁻ for I₂@MOC2 [76]. Moreover, the aromatic C═C‐associated C 1s signals shifted from 284.92 and 285.37 to 285.42 and 285.56 eV (Figure 6b, Figure S103) [77]. Collectively, these spectroscopic and analytical studies establish that iodine adsorption by MOC1 and MOC2 is dominated by chemisorption involving synergistic π‒donor interactions from the carbazole units together with strong electrostatic attraction within the cationic cage. This cooperative binding environment promotes efficient charge transfer and stabilization of polyiodide species such as I₃ ⁻ and I₅ ⁻, thereby accounting for the exceptional iodine uptake performance of the metal‒organic cages.

FIGURE 6.

(a) High‐resolution XPS spectrum of the I 3d region for iodine‐loaded I₂@MOC2. (b) Comparative C 1s XPS spectra of pristine and iodine‐adsorbed MOC2. (c) Raman spectra of the MOCs before and after iodine adsorption. (d) EDS elemental mapping of iodine distribution in I₂@MOC2. DFT‒optimized structures illustrating different interaction sites and the corresponding binding energies for (e) MOC2@I₂ and (f) MOC2@I₃ ^‒.

Methanol proved to be an efficient regeneration solvent, as demonstrated by the gradual release of iodine monitored using UV–vis spectroscopy (Figures S92‒S67). Time dependent increases in the absorbance at 360 and 445 nm confirmed the controlled recovery of adsorbed iodine species, validating the recyclability of the adsorbent system. Importantly, ¹H NMR spectra of the regenerated cages remained identical to those of the pristine materials, confirming that the cage structures remain intact after repeated adsorption−desorption cycles and highlighting their excellent structural robustness for long term iodine remediation applications (Figures S105 and S106).

We made multiple attempts to obtain suitable single crystals of the iodine‐loaded MOC complexes under various crystallization conditions, including different solvents, solvent combinations, iodine concentrations, and crystallization protocols. However, despite these extensive efforts, we were unable to obtain single crystals of sufficient quality for single‐crystal X‐ray diffraction analysis. Therefore, to obtain deeper molecular‐level insight into the iodine adsorption process, DFT calculations were carried out for the cages MOC1 and MOC2 interacting with I₂ and I₃ ⁻ species [78]. Electrostatic potential (ESP) surface analysis revealed that the cationic nature of the cages, arising from the Pd centers together with positively polarized regions between neighboring aromatic ligands, generates a highly favorable electrostatic environment for iodine binding (Figure S107) [79]. These positively charged regions strongly promote strong interactions with iodine species, particularly with the anionic I₃ ⁻ species. Consistent with this interpretation, the calculated binding energies indicate substantially stronger interactions for I₃ ⁻ than for neutral I₂, highlighting the dominant contribution of electrostatic attraction between the polyiodide anions and the cationic cage structure. The theoretical calculations further demonstrate that MOC2 exhibits significantly higher binding affinity toward both iodine species relative to MOC1. For MOC2, the calculated binding energies were −174.63, −158.67, and −153.06 kcal mol⁻¹ for different I₂ binding configurations and −362.43, −310.53, and −310.13 kcal mol⁻¹ for I₃ ⁻ binding configurations, whereas comparatively weaker interactions were obtained for MOC1, with binding energies ranging from −154.54 to −131.80 kcal mol⁻¹ for I₂ and from −345.76 to −292.82 kcal mol⁻¹ for I₃ ⁻ (Figure 6e,f, Figures S108‒S111). These results are fully consistent with the experimentally observed superior iodine uptake performance of MOC2. For both cage systems, iodine encapsulation within the internal cavity was calculated to be energetically more favorable than adsorption at external surface sites. In particular, cavity‒confined iodine species exhibited significantly stronger binding energies than externally associated species, emphasizing the important role of confinement effects and host−guest stabilization within the cage cavities (Figure 6e,f, Figures S108–S111) [80]. Together, the DFT results support the experimental findings and confirm that the exceptional iodine capture performance arises from the synergistic interplay of electrostatic interactions, confinement effects, and charge‐transfer interactions within the π‒electron‐rich cage framework.

The iodine adsorption studies further establish that the functional performance of these cages originates from their molecularly engineered internal and external environments. In particular, the π‐electron‐rich carbazole framework and cationic Pd₂L₄ architecture provide a cooperative platform in which electrostatic attraction, confinement, and π−iodine/charge‐transfer interactions collectively promote guest stabilization. Notably, these same structural and electronic features are also expected to influence the excited‐state behavior of the cages. The extended π‐conjugated carbazole chromophores can act as light‐responsive units, while the heavy Pd(II) centers may facilitate intersystem crossing and the generation of reactive excited states capable of activating molecular oxygen. Thus, the electronic characteristics that enable efficient iodine recognition also provide a molecular basis for a second function−photosensitized catalytic reactivity. The photocatalytic activity therefore emerges from the same ligand‐engineered cage platform, with the carbazole chromophores and Pd(II) centers collectively controlling both host‐guest interactions and excited‐state processes. On this basis, we next investigated the photophysical properties of MOC1 and MOC2 and explored their potential as heterogeneous photocatalysts for visible‐light‐driven aerobic sulfide oxidation.

2.4. Photocatalytic Activity

The preceding studies establish that the multifunctionality of the present system originates from a common molecular design rather than from three independent properties. The Pd₂L₄ architecture provides a well‐defined and cationic molecular framework, while the carbazole ligands simultaneously contribute π‐rich surfaces for host−guest recognition and extended conjugation for light harvesting. The resulting combination of structural confinement, electronic polarization, and Pd(II)‐mediated excited‐state processes provides a molecular basis for both iodine sequestration and photocatalytic activity [53]. In this context, the iodine‐binding studies demonstrate the ability of the cages to engage strongly with electron‐deficient iodine species through electrostatic, confinement, and π−iodine/charge‐transfer interactions, whereas the same π‐conjugated framework is expected to support long‐lived excited states and reactive oxygen species generation [59, 60]. We therefore investigated whether rational modification of the carbazole substituent could simultaneously tune the cage assembly, electronic structure, and photocatalytic response. In particular, comparison of MOC1 and MOC2 provides an opportunity to correlate ligand substitution with the hierarchical material properties, iodine‐binding behavior, and photochemical performance of the resulting cage systems.

The photophysical properties of MOC1 and MOC2 were examined by UV−Vis absorption spectroscopy in DMSO. Both cages display strong absorption bands within the UV region (254−380 nm), indicating their suitability for activation under blue‐light irradiation (Figure 7a). Optical band gaps estimated from Tauc plot analysis were determined to be 3.34 eV for MOC1 and 3.29 eV for MOC2 (Figure 7a, inset). The slightly narrower band gap observed for MOC2 originates from the enhanced π‐conjugation imparted by the phenyl substituent in ligand L2, which promotes greater electronic delocalization relative to the methyl‐substituted analogue L1. To gain further insight into their electronic structures, Mott−Schottky analyses were conducted using an Ag/AgCl reference electrode at multiple frequencies. From the linear extrapolation of the plots, conduction band energies of −0.94 eV for MOC1 and −0.81 eV for MOC2 were obtained (Figure 7b). Using the relation E_g = E_CB   − E_VB , the corresponding valence band positions were calculated to be 2.40 and 2.48 eV for MOC1 and MOC2, respectively, confirming that both systems possess energetically favorable band structures for photocatalytic redox transformations (Figure 7c). Photoluminescence measurements revealed strong blue emission in the range of ∼360–420 nm for both cages, indicative of efficient radiative recombination of photoexcited charge carriers (Figure S114).

FIGURE 7.

Photophysical, electrochemical, and mechanistic investigations of the MOCs. (a) UV−vis absorption spectra of the MOCs (inset: corresponding Tauc plots). (b) Mott–Schottky plots illustrating the flat‐band potentials. (c) Band energy diagrams illustrating the valence band (VB) and conduction band (CB) positions of MOC1 and MOC2. (d) Transient photocurrent responses recorded under intermittent light irradiation. (e) Excited‐state lifetimes obtained from time‐correlated single‐photon counting (TCSPC) measurements for MOC1 and MOC2. (f) Light on‐off control experiments highlighting the photo‐dependency of the catalytic reaction. (g) EPR spectra confirming the generation of singlet oxygen through oxidation of TEMP to TEMPO in the presence of the MOC2 under light irradiation. (h) Stern−Volmer quenching analysis of pristine MOCs with thioanisole. (i) Kinetic profiles for the selective aerobic oxidation of thioanisole under optimized conditions.

Relative to MOC1, MOC2 exhibited a slight red shift in emission, reflecting differences in the electronic environment induced by the phenyl substituent. Time‐correlated single‐photon counting (TCSPC) experiments further demonstrated that MOC2 possesses a longer excited‐state lifetime (8.31 ns) than MOC1 (7.52 ns) (Figure 7e), suggesting more effective suppression of electron−hole recombination and improved photosensitization efficiency. To directly probe charge‐transfer dynamics, photoelectrochemical studies were performed.

Electrochemical impedance spectroscopy (EIS) measurements showed that MOC2 possesses a substantially smaller semicircle in the Nyquist plot, corresponding to a lower charge‐transfer resistance (Rct = 1.66 kΩ) compared with MOC1 (Rct = 3.72 kΩ), indicating more efficient interfacial electron transport (Figure S112). Consistent with this result, linear sweep voltammetry (LSV) under illumination revealed higher photocurrent densities for MOC2 (2.23 mA cm⁻²) than for MOC1 (1.86 mA cm⁻²) over the potential range of −1.6 to −0.1 V vs. Ag/AgCl (Figure S113). Transient photocurrent response experiments further supported these observations, with MOC2 exhibiting a markedly enhanced photocurrent response of 0.8 µA cm⁻² at 0 V vs. Ag/AgCl relative to MOC1 (Figure 7d). Collectively, these results demonstrate that phenyl substitution in MOC2 significantly improves π‐conjugation, narrows the optical band gap, prolongs excited‐state lifetimes, suppresses charge recombination, and enhances charge‐transfer efficiency, thereby leading to superior photocatalytic performance compared with MOC1 [81].

The combined photophysical and photoelectrochemical studies establish MOC1 and MOC2 as promising visible‐light‐responsive heterogeneous photocatalysts. Motivated by these results, we next explored their catalytic performance in the aerobic oxidation of sulfides to sulfoxides using molecular oxygen as the terminal oxidant. Thioanisole (methyl phenyl sulfide, MPS) was selected as a representative model substrate to optimize the photocatalytic reaction conditions. A detailed optimization study involving catalyst loading, irradiation wavelength, light source, solvent, and reaction time was conducted to identify the most efficient catalytic conditions (Figure S117). Among the two cages, MOC2 consistently exhibited superior photocatalytic activity, affording methyl phenyl sulfoxide (MPSO, 1) in >99% yield under blue LED irradiation (λ_max = 390 nm) using methanol as the reaction medium (Figure 7i). The catalytic efficiency was found to strongly depend on the irradiation wavelength, with maximum activity observed at 390 nm, while alternative solvents led to significantly lower product yields. Under the optimized conditions (0.5 mol% MOC2, methanol, O₂ purging for 2 min, blue LED irradiation at 390 nm, room temperature, 6 h), the oxidation proceeded with near‐quantitative conversion and exceptional selectivity (~99%) toward the corresponding sulfoxide (Figure S117). Comparative studies demonstrated that MOC2 outperformed MOC1 in both catalytic efficiency and product selectivity (Figure 7i).

Control experiments confirmed the critical roles of light, oxygen, and catalyst in the reaction process the yield decreased substantially (∼64%) in the absence of oxygen, no product formation was observed in the dark, and catalyst‐free reactions produced only ∼19% yield. The conversion of MPS and formation of MPSO were monitored by ¹H NMR spectroscopy (Figure 7i, Figure S121). Time‐dependent spectral analysis showed a gradual decrease in the characteristic resonances of MPS at δ 7.10, 7.24, and 2.43 ppm accompanied by the appearance of new signals corresponding to MPSO at δ 7.68, 7.58, and 2.67 ppm (Figure S121). Importantly, MOC2 enabled complete conversion of MPS to MPSO within 6 h in methanol without detectable formation of the overoxidized sulfone byproduct (<1%), highlighting the remarkable chemoselectivity of the catalytic system toward partial oxidation (Figure 7i, Figure S121). Kinetic investigations revealed that the oxidation reaction follows apparent zero‐order kinetics, with MOC2 exhibiting a rate constant of 4.505 ×10^‒4 mol L^‒1 s^‒1 and a corresponding half‐life (t_1/2) of 178.6 min (Figure S122) [82]. Furthermore, the turnover frequency (TOF) of MOC2 (29.42 h⁻¹) exceeds that of MOC1 (27.93 h⁻¹) and surpasses the values reported for most previously described photocatalysts (Tables S1–S15) [82, 83]. To demonstrate the versatility and substrate tolerance of the photocatalytic system, a diverse range of thioanisole derivatives containing both electron‐donating and electron‐withdrawing substituents was examined under the optimized reaction conditions (Table 1). In all cases, MOC2 efficiently catalyzed the aerobic oxidation of the sulfides to their corresponding sulfoxides, generally affording yields ≥70% together with excellent selectivity of ~97%. Importantly, overoxidation to the undesired sulfone byproducts remained below 3% for all substrates investigated, highlighting the exceptional chemoselectivity of the catalytic system. Although MOC1 also exhibited high selectivity, its catalytic activity was consistently lower than that of MOC2, emphasizing the beneficial role of the extended π−conjugation and improved charge‐transfer properties present in MOC2. As anticipated, substrates bearing electron‐donating substituents displayed enhanced reactivity and higher turnover frequencies (TOFs), with p‐Me and p‐OMe derivatives of thioanisoles exhibiting TOF values of 27.81 and 29.13 h^‒1, respectively (Table 1, Figures S125–S128). In contrast, substrates containing electron‐withdrawing groups showed relatively slower reaction rates because of reduced electron density at the sulfur center. Nevertheless, even strongly deactivating substituents such as aldehyde and nitro groups underwent efficient oxidation, providing high product yields (typically ≥70% for MOC2) and excellent selectivity, with TOFs of 20.15 h^‒1 for p‐CHO and 23.19 h^‒1 for p‐NO₂ derivatives (Table 1). Halogen‐substituted aromatic thioethers were also well tolerated, particularly para‐halogenated substrates, for which MOC2 consistently achieved product yields exceeding 80% while maintaining selectivity above 97% (Table 1, Figures S129 and S130). Furthermore, replacing the methyl substituent in thioanisole with an ethyl group resulted in almost quantitative conversion with MOC2 (99%) and high conversion with MOC1 (86%), while preserving outstanding selectivity (99%) in both cases (Table 1, Figures S123 and S124). The broad substrate scope, high catalytic activity, and exceptional selectivity demonstrated by MOC2 highlight its effectiveness as a heterogeneous photocatalyst for the selective aerobic oxidation of sulfur‐containing compounds. These findings further suggest its potential applicability in environmental detoxification and remediation processes involving sulfur‐containing toxic agents. The recyclability and operational durability of the catalysts were also investigated to assess their practical applicability. The photocatalysts could be readily recovered and reused over five consecutive catalytic cycles with only a slight decrease in activity. Even after the fifth run, MOC1 and MOC2 retained high catalytic efficiencies, providing product yields of 84% and 87%, respectively (Figure 8a). We have evaluated the photocatalytic efficiency of the freeze‐dried MOCGs under the same reaction conditions used for the corresponding as‐synthesized MOC powders. Interestingly, the MOCGs exhibited no significant difference in catalytic activity compared with the corresponding MOC powders, indicating that the hierarchical gel morphology does not adversely affect the intrinsic photocatalytic performance of the cage units (Figures S131‒133).

TABLE 1.

Substrate scope for the heterogeneous photocatalytic oxidation of various thioethers to sulfoxides using the MOC catalysts with high conversion and selectivity ^a .

Entry	Reactant	Product	Con.(%) b	Con.(%) c	Sel.(%) d	TOF (h−1) e
1			96	>99	>99	29.42
2			86	99	>99	29.57
3			83	93	98	27.81
4			85	98	97	29.13
5			74	86	98	25.78
6			72	81	97	25.59
7			69	80	99	24.71
8			58	70	99	20.15
9			71	76	98	23.19

^a Reaction condition: Catalyst MOC (0.5 mol%); 1mL of MeOH; 390 nm Light irradiation for 6h; 0₂ purge for 2 min; Room Temperature.

^b Conversion By MOC1.

^c Conversion by MOC2.

^d Selectivity.

^e TOF using the catalyst MOC2. Conversion and selectivity was measured by the ¹H NMR spectroscopy.

FIGURE 8.

(a) Reusability of the MOCs in the photocatalytic reaction. (b) Proposed mechanistic pathway for the MOC catalyzed photocatalytic oxidation of sulfides to sulfoxides under visible‐light irradiation.

We have performed a comprehensive set of experiments to evaluate the solubility, heterogeneity, and possible leaching of MOC1 and MOC2 under the catalytic conditions. First, the solubility of both cages was assessed in MeOH, H₂O, acetonitrile, CHCl₃, DCM, and DMSO using ¹H NMR spectroscopy (Figures S134 and S135). Well‐resolved cage resonances were observed only in DMSO‐d ₆, confirming appreciable solubility in DMSO, whereas no characteristic cage signals were detected in the other solvents, indicating that MOC1 and MOC2 are essentially insoluble in these media. These results are consistent with the heterogeneous nature of the photocatalytic reaction performed in methanol. We further conducted a hot‐filtration experiment to directly evaluate whether catalysis arises from dissolved molecular species (Figures S136 and S137). During the reaction, the reaction mixture was filtered at the reaction temperature to remove the suspended solid catalyst, and the resulting filtrate was subsequently maintained under identical reaction conditions in the absence of the solid catalyst. No significant increase in product formation was observed after hot filtration, indicating that the catalytically active species is predominantly associated with the heterogeneous cage material rather than a dissolved molecular species.

To further assess possible catalyst leaching, the filtered reaction mixtures were analyzed by ¹H NMR spectroscopy and ICP‐OES. No characteristic cage resonances were observed in the ¹H NMR spectra of the filtrates in DMSO‐d ₆, indicating that significant dissolution of the cage complexes did not occur (Figures S138 and S139). ICP‐OES analysis further showed that only ~3% of the cage material leached into the reaction solution for both MOC1 and MOC2 (Table S13). We also examined the structural integrity of the catalysts before and after the catalytic reaction using ¹H NMR and XPS spectroscopy. No appreciable changes were observed in either the ¹H NMR or XPS profiles, indicating that the cage catalysts remain structurally intact under the reaction conditions (Figures S141‒S144). Mechanistic investigations, together with previous reports, indicate that the photocatalytic aerobic oxidation of thioanisole proceeds predominantly through reactive oxygen species (ROS), particularly singlet oxygen (¹O₂), generated by the photoexcited metal−organic cages [84]. Radical trapping experiments using 2,2,6,6‐tetramethylpiperidin‐1‐oxyl (TEMPO) significantly suppressed the reaction efficiency, reducing the conversion to ~15%, thereby confirming the involvement of radical‐mediated intermediates during the catalytic process. Generation of ¹O₂ proceeds via a photosensitized pathway, wherein energy is transferred from the triplet excited state of the MOC to ground‐state molecular oxygen (³O₂₎. Consistent with this mechanism, electron spin resonance (ESR) spectroscopy performed under light irradiation in the presence of TEMP exhibited the characteristic three‐line signal corresponding to TEMPO formation, confirming the generation of singlet oxygen in methanol in the presence of the MOC catalyst. In contrast, no detectable ESR signal was observed in the absence of light irradiation (Figure 7g). We performed additional control experiments using a series of selective radical and charge‐carrier scavengers under otherwise identical photocatalytic conditions. Specifically, NaN₃ was employed as a singlet‐oxygen (¹O₂) scavenger, isopropanol (IPA) as a hydroxyl‐radical (^•OH) scavenger, AgNO₃ as an electron scavenger, KI as a hole scavenger, 1,4‐benzoquinone (BQ) as a superoxide‐radical (O₂ ^•−) scavenger, and 1,1‐diphenylethylene (DPE) as a sulfur‐radical trapping reagent. Interestingly, a pronounced suppression of product formation was observed only in the presence of NaN₃, whereas the addition of IPA, AgNO₃, KI, BQ, or DPE did not cause any significant change in the product yield (Figures S118 and S119). These results strongly support ¹O₂ as the dominant reactive oxygen species responsible for the photocatalytic oxidation, while providing no significant evidence for the involvement of ^•OH, O₂ ^•−, photogenerated electrons/holes, or sulfur‐radical intermediates under the investigated conditions. In addition, intermittent light on/off experiments revealed that the reaction immediately ceased when the light source was removed, ruling out the possibility of a self‐sustaining radical chain process operating independently of continuous photoexcitation (Figure 7f). Collectively, these findings demonstrate that the catalytic oxidation process is governed by photoinduced ROS generation through a dominant energy‐transfer pathway involving singlet oxygen.

To further understand the interaction between the excited‐state cages and the sulfide substrate, steady‐state photoluminescence quenching experiments were performed using methyl phenyl sulfide (MPS) as a quencher for both MOC1 and MOC2. Gradual addition of MPS resulted in significant quenching of the emission intensities of both cages (Figure S120). Stern−Volmer analysis using the relation I₀/I = 1 + K _sv[Q] afforded quenching constants (K _sv) of 350.8 m ⁻¹ for MOC1 and 530.6 M⁻¹ for MOC2. The larger quenching constant observed for MOC2 indicates more efficient excited‐state interaction and energy transfer between the photoexcited cage and the sulfide substrate (Figure 7h), consistent with the enhanceds photocatalytic activity exhibited by MOC2 during sulfide oxidation [81].

Based on these mechanistic studies and related literature precedents, a plausible photocatalytic pathway for sulfide oxidation is proposed (Figure 8b) [82, 83, 85, 86]. Upon visible‐light irradiation, the MOC is initially excited to its singlet excited state (MOC*), followed by rapid intersystem crossing to the long‐lived triplet excited state (MOC^T). This triplet excited species subsequently transfers energy to ground‐state molecular oxygen (³O₂), generating highly reactive singlet oxygen (¹O₂) while regenerating the ground‐state catalyst. The generated ¹O₂ then reacts with methyl phenyl sulfide to form a persulfoxide intermediate, which subsequently reacts with another sulfide molecule to furnish the corresponding sulfoxide product, thereby completing the catalytic cycle. The predominance of this singlet oxygen‐mediated energy‐transfer pathway accounts for both the high catalytic efficiency and the remarkable selectivity observed during the photocatalytic oxidation process. To the best of our knowledge, this study represents the first example of a self‐assembled coordination cage functioning as a heterogeneous photocatalyst under visible‐light irradiation. Importantly, the present catalytic system operates efficiently under mild visible‐light irradiation without the need for additional oxidants such as H₂O₂ [87], thereby minimizing energy consumption and avoiding the use of stoichiometric oxidizing reagents compared to previously reported systems [88]. Despite these mild reaction conditions, the MOC catalysts exhibit excellent photocatalytic performance, highlighting the effectiveness of rational ligand engineering. In particular, the phenyl‐functionalized MOC2 demonstrates that modulation of ligand π−conjugation provides an effective strategy for tuning the electronic structure, narrowing the band gap, and enhancing the photocatalytic activity of metal–organic cage systems.

3. Conclusion

Collectively, these results reveal a common molecular design principle underlying the multifunctionality of the present cage platform. The Pd(II)‐directed assembly defines a robust and cationic molecular cavity, while the π‐conjugated carbazole ligands simultaneously govern hierarchical intermolecular organization, iodine recognition, and photoinduced electronic processes. Consequently, the same structural features enable the discrete cages to undergo higher‐order assembly into adaptive self‐healing metallogels, selectively stabilize iodine species through cooperative confinement and π−iodine/electrostatic interactions, and generate reactive oxygen species for selective aerobic photocatalysis. The systematic differences between MOC1 and MOC2 further demonstrate that ligand‐level modulation of π‐conjugation can tune both supramolecular organization and functional performance. Thus, rather than representing three independent functionalities, the present study establishes a unified cage‐based design strategy in which molecular architecture, hierarchical assembly, host–guest recognition, and excited‐state reactivity are interconnected within a single supramolecular platform. This approach provides a general framework for developing multifunctional, adaptive soft materials that integrate structural responsiveness, environmental remediation, and sustainable catalysis.

[CCDC 2523369 and 2523370 for the cages MOC1 and MOC2, respectively contains the supplementary crystallographic data for this paper. These data can be obtained free of charge from The Cambridge Crystallographic Data Centre via www.ccdc.cam.ac.uk/data_request/cif.]

Author Contributions

Monotosh Dalapati: investigation, validation, methodology, conceptualization, formal analysis, writing – original draft, writing – review and editing. Raghunath Singha: methodology, data curation, formal analysis, investigation. Debashree Manna: formal analysis, investigation. Dipak Samanta: conceptualization, data curation, visualization, methodology, investigation, supervision, project administration, funding acquisition, validation, formal analysis, writing – original draft, writing – review and editing.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File: advs78042‐sup‐0001‐SuppMat.docx.

Data Availability Statement

The data that supports the findings of this study are available in the Supporting Information of this article.