Programming Host–Guest Interactions in Covalent Organic Frameworks with Arylazopyrazole Photoswitches

Abstract

Programming host–guest interactions in porous crystalline materials using external stimuli offers a promising strategy for adaptive molecular separations but remains challenging because adsorption is often governed by pore structure rather than tunable molecular interactions. Here, we report a series of photoresponsive covalent organic frameworks (LACOF-1) incorporating arylazopyrazole (AAP) photoswitches that enable light-controlled modulation of CO₂ adsorption through molecular photoisomerization. The framework-embedded AAP units undergo efficient E–Z photoisomerization with an exceptional thermal half-life of ∼76 days while preserving crystallinity and permanent porosity of the framework. Despite exhibiting a lower surface area than the corresponding Z-state, the E-state displays stronger CO₂ adsorption enthalpies, demonstrating that adsorption can be regulated independently of overall pore accessibility. Combined molecular simulations and density functional theory calculations reveal that this behavior arises from enhanced local CO₂ binding within adsorption pockets created by the extended E-AAP configuration. These molecular-level changes translate directly into reversible dynamic CO₂ capture under dry, dilute, and humid conditions, with excellent cycling stability and facile regeneration under ambient conditions. This work establishes molecular photoswitching as an effective strategy for programming adsorption energetics in crystalline porous materials and provides a general design principle for adaptive, stimulus-responsive gas separations.

1. Introduction

The ability to actively regulate host–guest interactions in porous solids represents a longstanding challenge in supramolecular and materials chemistry. − Dynamic control over molecular recognition, diffusion, and adsorption would enable adaptive porous materials with programmable functions that respond reversibly to external stimuli. Among the available stimuli, light is particularly attractive because it provides rapid, remote, and waste-free control with exceptional spatial and temporal precision. , Unlike thermal or pressure–driven processes, photoirradiation selectively alters molecular conformation through localized structural changes without disrupting the overall framework, making it an appealing strategy for programming host–guest interactions and adsorption energetics in crystalline porous materials. −

Covalent organic frameworks (COFs) provide an ideal platform for investigating light-controlled host–guest interactions. Constructed from organic building blocks through reticular chemistry, COFs possess long-range structural order, permanent porosity, and highly tunable pore environments that enable precise placement of functional groups within well-defined channels. − Owing to these characteristics, COFs have attracted considerable attention for gas storage and separation, catalysis, molecular recognition, and optoelectronic applications. − Their modular architectures also provide a unique platform for establishing structure–property relationships between molecular-scale conformational changes and macroscopic adsorption behavior within extended porous networks. ,, This structural versatility has enabled the development of COFs with exceptional gas adsorption performance. For example, Yaghi and co-workers reported the polyamine-functionalized COF-999, which enables efficient direct air capture of CO₂ with regeneration at relatively mild temperatures (∼60 °C). Despite such impressive advances, adsorption–desorption processes in most COF systems remain predominantly thermally driven, motivating the development of alternative low-energy strategies for reversible adsorption control.

Recent efforts have focused on incorporating photoresponsive units into porous frameworks to achieve reversible modulation of pore environments and gas adsorption. In particular, studies of azobenzene-functionalized metal–organic frameworks (MOFs) and COFs have established that reversible trans–cis photoisomerization can modulate pore accessibility and dynamically influence gas adsorption behavior. , In pioneering work, Lyndon and co-workers demonstrated reversible modulation of CO₂ adsorption in an azobenzene-functionalized MOF through trans–cis photoisomerization. More recently, Feng and co-workers reported visible-light-responsive COFs incorporating tetrafluoroazobenzene photoswitches, enabling reversible tuning of pore environments and gas adsorption behavior. Together, these studies established molecular photoswitching as a viable strategy for introducing dynamic functionality into crystalline porous materials. However, most photoresponsive systems reported to date rely on azobenzene derivatives, which suffer from intrinsic limitations, including relatively short thermal half-lives of the Z isomer and limited fatigue resistance. These drawbacks are often exacerbated in rigid porous frameworks, where restricted conformational freedom can reduce switching efficiency and long-term stability. Consequently, achieving efficient, reversible, and durable photoresponsive behavior in crystalline porous materials remains a significant challenge.

Azoheteroarenes have recently emerged as promising alternatives, offering tunable photophysical properties and significantly extended thermal lifetimes. − Among them, arylazopyrazoles (AAPs) are particularly notable for their near-quantitative E–Z photoisomerization and exceptionally long-lived Z isomers, with thermal half-lives ranging from days to years at room temperature. ,, These properties have enabled reversible regulation of host–guest interactions in supramolecular systems and surface-bound assemblies. − In porous materials, AAPs have been explored as guest molecules confined in MOFs, where the framework environment further enhances their thermal stability. Despite these attractive properties, AAP photoswitches have not yet been incorporated into crystalline COF frameworks.

Herein, we report a series of AAP-functionalized COFs (LACOF-1) that enable reversible light-controlled modulation of CO₂ adsorption through molecular photoswitching. By systematically tuning the density of AAP units within a common crystalline framework, we demonstrate that molecular photoswitching can program local host–guest interactions and adsorption energetics independently of overall pore accessibility. Combined experimental characterization and multiscale theoretical calculations reveal the molecular origins of the photoresponsive adsorption behavior. More broadly, this work establishes a molecular design strategy for programming adsorption energetics in crystalline porous materials and provides a general framework for developing adaptive porous materials with externally tunable molecular recognition and gas separation.

2. Results and Discussion

2.1. Characterization and Structural Analysis of LACOF-1

To program host–guest interactions through molecular photoswitching, we designed a series of COFs (LACOF-1) incorporating AAP photoswitches into a two-dimensional imine-linked backbone. The frameworks were constructed via multicomponent polycondensation of 1,3,5-tris­(4-aminophenyl)­benzene (TPB) with varying ratios of AAP-functionalized linker 4,4″-bis­((E)-(1-methyl-1H-pyrazol-3-yl)­diazenyl)-[1,1′:4′,1″-terphenyl]-2′,5′-dicarbaldehyde (MPDTD) and 2,5-dimethoxyterephthalaldehyde (DMTA), affording a family of isoreticular COFs with tunable AAP content (Figure a). Four isoreticular compositions (AAP: OMe = 1:5, 2:4, 3:3, and 4:2) were synthesized to systematically vary the density of AAP photoswitches while preserving the underlying framework topology (Figures S4). This modular design enables systematic tuning of local host–guest interactions while maintaining an identical crystalline framework, providing an ideal platform for establishing structure–property relationships between photoswitch density and adsorption behavior.

1.

Photoresponsive covalent organic frameworks. (a) Schematic illustration of the synthesis and reversible photoisomerization of LACOF-1 with varying arylazopyrazole (AAP)-to-methoxy ratios; E- and Z-LACOF-1 (3:3) are shown as representative examples. (b) Solid-state ¹³C CP-MAS NMR spectrum of LACOF-1 (3:3) with peak assignment. (c) PXRD patterns of E-LACOF-1 with different AAP-to-methoxy ratios (4:2, 3:3, 2:4, and 1:5) compared with simulated and experimental patterns of TPB-DMTP-COF (AA-stacking). (d) N₂ adsorption isotherms measured at 77 K for TPB-DMTP-COF and E-LACOF-1 samples with varying AAP-to-methoxy ratios.

Formation of the imine-linked structures was confirmed by FT-IR and solid-state ¹³C cross-polarization magic-angle spinning (CP/MAS) NMR spectroscopy. The FT-IR spectra of LACOF-1 with varying AAP content all showed the characteristic CN stretching vibrational band at 1592 cm^–1, together with the disappearance of the aldehyde (CO) vibrational band at 1682 cm^–1, suggesting complete polycondensation (Figures S5). Solid-state ¹³C CP/MAS NMR spectra of representative E- and Z-LACOF-1 (3:3) samples show the presence of characteristic carbon resonance of the imine (CN) bond at 157 ppm and the absence of the aldehyde signal at ∼190 ppm (Figures b and S6). The characteristic resonances associated with the AAP units are retained after framework formation, confirming their successful incorporation into the COF backbone. No significant spectral differences were observed between the E- and Z-isomers, likely due to the similar local chemical environments of the two states in the solid phase (Figure S6). Powder X-ray diffraction (PXRD) patterns reveal that all LACOF-1 compositions remain highly crystalline despite the progressive incorporation of bulky AAP photoswitches. The experimental diffraction patterns closely match the simulated AA-stacked structure of the backbone TPB-DMTA-COF (Figure c), demonstrating that systematic variation of AAP density does not alter the overall framework topology. Because the COFs are formed through multicomponent condensation of two aldehyde linkers, different local distributions of the AAP-containing and methoxy-containing linkers may occur even at a fixed overall composition. Although such local distributions cannot be directly resolved by PXRD, they may influence the local pore environment and adsorption-site geometry without altering the average crystalline framework. The preservation of long-range structural order, therefore, enables the influence of AAP content and local pore environments on adsorption behavior to be examined within a common framework topology.

Nitrogen sorption measurements at 77 K further confirmed permanent porosity, with BET surface areas and pore volumes strongly dependent on AAP loading (AAP: OMe = 1:5, 2:4, 3:3, and 4:2; Figure d and Table S1). The progressive decrease in porosity is consistent with partial occupation of pore space by the photoswitching units. Notably, LACOF-1 samples with AAP: OMe ratios of 1:5, 2:4, and 3:3 exhibit relatively comparable BET surface areas and pore volumes, indicating that moderate incorporation of AAP units does not substantially compromise the accessible porosity of the framework. In contrast, a marked reduction in porosity occurs for LACOF-1 (4:2). Compared with LACOF-1 (3:3), which exhibits BET surface areas of 1017 and 1225 m² g^–1 for the E- and Z-states, respectively, LACOF-1 (4:2) displays substantially lower surface areas of only 161 and 182 m² g^–1 for the corresponding states. These results indicate that beyond a critical AAP loading, the increased density of photoswitches occupies a substantial fraction of the pore volume, significantly modifying the local pore environment.

Furthermore, LACOF-1 (1:5) (2:4), and (3:3) exhibit similar pore size distributions centered at approximately 29 Å despite containing different AAP contents, indicating that moderate incorporation of AAP units has little effect on the primary pore dimensions. In contrast, LACOF-1 (4:2) exhibits additional, smaller pores centered at approximately 14 and 17 Å. The emergence of these smaller pore domains suggests that the high density of AAP units partially partitions the original mesoporous channels, creating secondary confined adsorption environments that may strengthen local host–guest interactions. As demonstrated below, this unique pore architecture plays a key role in the light-controlled modulation of adsorption energetics. Thermogravimetric analysis (TGA) further confirms the excellent thermal stability of LACOF-1, with no significant weight loss observed below 350 °C (Figure S7).

2.2. Photoisomerization

Effective light-controlled modulation of adsorption requires that the embedded AAP units retain efficient and reversible photoisomerization after incorporation into the rigid crystalline framework. The AAP-containing precursor 3-((4-bromophenyl)­diazenyl)-1-methyl-1H-pyrazol (BDMP) (Figure a) was used as a molecular model of the framework-embedded photoswitch to probe the intrinsic photoisomerization behavior prior to incorporation into the COF. The wavelength-dependent photoresponse of BDMP was then investigated to identify optimal irradiation conditions.

2.

Photoisomerization behavior of 3-((4-bromophenyl)­diazenyl)-1-methyl-1H-pyrazol (BDMP) and LACOF-1. (a) Reversible photoisomerization of the molecular model compound BDMP. (b) Time-dependent UV–Vis absorption changes at 331 nm during E→Z and Z→E photoisomerization of BDMP under irradiation with 365 nm UV light (red) and 450 nm blue light (black), respectively. (c) Reversible switching behavior of BDMP over 30 irradiation cycles with alternating 365 and 450 nm light (irradiation time: 4 min). (d) Time-dependent UV–Vis absorption changes at 350 nm during E→Z and Z→E photoisomerization of LACOF-1 (4:2) under irradiation with 365 nm UV light (red) and 450 nm blue light (black), respectively. (e) Reversible switching behavior of LACOF-1 (4:2) over 30 irradiation cycles with alternating 365 and 450 nm light. (f) Summary of heats of adsorption for TPB–DMTP-COF and E- and Z-LACOF-1 samples with varying AAP-to-methoxy ratios.

Upon irradiation at 365 nm, the characteristic n–π* absorption band at 331 nm decreased concomitantly with the appearance of new π–π* absorption bands at 243 and 276 nm. Well-defined isosbestic points were observed throughout the switching process, indicating efficient E→Z conversion, which was completed within 120 s (Figures S8a and b). In contrast, irradiation with 450 nm blue light rapidly restored the E-isomer on a significantly shorter time scale (< 6 s) (Figures S8b and b). Other visible wavelengths, as well as 254 nm UV light, also induced Z→E conversion, albeit with reduced efficiency or slower kinetics (Figure S9 and S10). Quantitative ¹H NMR analysis of BDMP after 365 nm irradiation further showed that the Z-rich photostationary state contains approximately 90% Z isomer (Figure S11), confirming efficient E→Z conversion under the irradiation conditions used in this study.

Beyond rapid switching kinetics, BDMP exhibits the exceptional photochemical stability characteristic of arylazopyrazoles. The Z-isomer possesses a thermal half-life of approximately 76 days at room temperature (Figures S11–S13), consistent with the behavior of AAP photoswitches reported by Calbo et al., while repeated alternating irradiation between 365 and 450 nm produces highly reproducible switching for more than 30 cycles without detectable fatigue (Figure c). The combination of long thermal persistence and excellent fatigue resistance represents a key advantage over conventional azobenzene photoswitches and provides a robust molecular foundation for reversible modulation of adsorption processes.

These results established 365 and 450 nm as optimal wavelengths for reversible switching and were subsequently employed for photoisomerization of the COF samples. Importantly, efficient molecular switching is retained after incorporation into the COF framework. UV–Vis spectra of LACOF-1 (4:2) suspensions, measured using an integrating sphere, showed characteristic spectral changes upon irradiation at 365 and 450 nm, corresponding to reversible E→Z and Z→E switching, respectively (Figures d and S14). Photoisomerization remained remarkably efficient despite the steric constraints imposed by the crystalline lattice, requiring only ∼60 s for E→Z conversion and < 5 s for Z→E conversion. Direct quantitative determination of the E/Z ratio in the COF by UV–vis spectroscopy was not possible because the absorption feature associated with the Z state near 350 nm overlaps with the shoulder of the more intense band centered near 280 nm. Nevertheless, the spectrum obtained after 450 nm irradiation closely matches that of a thermally equilibrated E-state reference obtained by heating at 120 °C for 3 h, indicating near-complete recovery of the E state within the resolution of the UV–vis measurements (Figure S14c). Both photoisomeric states remain unchanged for at least 12 h under dark conditions (Figure S15ab), demonstrating that spontaneous thermal relaxation is negligible on the time scale of the adsorption experiments. Furthermore, repeated alternating irradiation produces highly reproducible photoresponses for more than 30 switching cycles without measurable degradation (Figures e and S15c), confirming that the AAP units retain robust photoactivity after covalent incorporation into the COF. Collectively, these results demonstrate that efficient molecular photoswitching can be maintained within an extended crystalline porous framework while preserving both structural integrity and switching performance, thereby providing the molecular basis for the light-controlled regulation of host–guest interactions explored below.

2.3. Structural Basis of Light-Controlled Gas Adsorption

Having established that the AAP units undergo efficient and reversible photoisomerization within the crystalline framework, we then investigated how photoisomerization influences the pore environment and adsorption behavior of LACOF-1. Nitrogen sorption analysis revealed systematic differences between the E- and Z-states, with Z-LACOF-1 generally exhibiting larger surface areas and pore volumes than the corresponding E-isomers, consistent with changes in local AAP configuration within the pores (Figure S16 and Table S1). Despite these differences in accessible porosity, PXRD patterns of the two photoisometric states remain essentially identical (Figure S17), demonstrating that photoisomerization preserves the long-range crystallinity and framework topology. Molecular dynamics (MD) simulations further support this conclusion, showing that structural equilibration maintains the experimentally observed framework architecture and diffraction features (Figures S18 and S19). Together, these results indicate that photoisomerization perturbs only the local molecular environment while leaving the overall framework intact.

The influence of these local structural changes becomes evident in the CO₂ adsorption measurements. Single-component CO₂ adsorption isotherms collected at 273 and 292 K (Figure S20a) reveal that incorporation of AAP units substantially enhances CO₂ affinity compared to the parent TPB-DMTP-COF, as reflected by higher heats of adsorption (Figure f and Table S2), which can be attributed to the basic nitrogen-donor sites of the pyrazole units. Notably, E-isomers consistently exhibit higher adsorption enthalpies toward CO₂ than the corresponding Z-isomers despite having lower surface areas (Table S2), indicating that adsorption is controlled primarily by local host–guest interactions rather than accessible surface area. This counterintuitive behavior contrasts with the conventional expectation that larger surface areas lead to higher adsorption capacities and differs from previously reported photoresponsive azobenzene COFs, in which adsorption primarily follows changes in accessible pore volume. These observations establish that molecular photoisomerization regulates adsorption through changes in the local binding environment rather than pore accessibility alone.

The selectivity of this effect is further demonstrated by CH₄ and H₂ adsorption measurements. CH₄ adsorption follows a trend similar to that observed for CO₂, with E-isomers consistently exhibiting higher uptake than the corresponding Z-isomers. However, the magnitude of this effect is considerably smaller than that observed for CO₂. In contrast, H₂ adsorption is largely governed by accessible surface area and pore volume, showing little dependence on AAP incorporation. Consistent with these observations, LACOF-1 (3:3) exhibits CO₂/CH₄ selectivities of 6.32 and 6.92 for the E- and Z-states, respectively, and corresponding CO₂/N₂ selectivities of 91.0 and 34.5 (Figure S21–S22 and Table S3). Collectively, these results demonstrate that AAP photoswitches selectively strengthen CO₂ binding while exerting only a modest influence on more weakly interacting gases such as CH₄ and H₂, thereby enhancing adsorption selectivity through programmed host–guest interactions rather than nonspecific increases in porosity.

To elucidate the molecular origin of the enhanced CO₂ affinity of the E-state, grand-canonical Monte Carlo (GCMC) simulations and density functional theory (DFT) calculations were performed on representative E- and Z-state LACOF-1 models. CO₂ density maps obtained from GCMC simulations (Figure a,b) reveal a higher local CO₂ density around the AAP-functionalized pore regions in the E-state, consistent with the formation of more accessible pyrazole-rich adsorption pockets created by the extended E-AAP configuration. Hotspot maps generated at both 907 Pa and 1.5 bar further show that the preferred adsorption sites remain localized within the AAP-functionalized pore environment in both photoisomertic states across a broad pressure range (Figure S23). Five representative CO₂ binding motifs extracted from the hotspot maps were subsequently truncated into two-layer cluster models for DFT calculations. As shown in Figure c, substantially stronger CO₂ binding was observed for the E-state than for the corresponding Z-state configurations. The representative configurations (Figure c) also provide insight into the possible influence of local linker arrangement. Within a given photoisomeric state, the calculated binding energies span a relatively narrow range, from −48.1 to −51.9 kJ mol^–1 for the E-state and from −25.4 to −26.5 kJ mol^–1 for the Z-state, corresponding to variations of only 3.8 and 1.1 kJ mol^–1, respectively. By comparison, the calculated E/Z differences range from 21.6 to 26.5 kJ mol^–1. This indicates that the effect of local linker arrangement on the representative binding configurations is considerably smaller than that induced by photoisomerization. This trend is consistent with the experimentally determined zero-coverage heats of adsorption for LACOF-1 (3:3), which are 42.6 and 33.5 kJ mol^–1 for the E- and Z-states, respectively, corresponding to an experimental difference of 9.1 kJ mol^–1 (Table S2). The larger E/Z contrast predicted by DFT is expected because the calculations describe site-specific interactions at idealized local binding configurations, whereas the experimental heats of adsorption represent finite-temperature, ensemble-averaged interactions across multiple adsorption sites and local environments. Structural heterogeneity and the distribution of local adsorption environments may further reduce the experimentally observed difference between the two states.

3.

CO₂ adsorption hotspot maps obtained from GCMC simulations at 273 K and 0.27 bar for (a) the Z-state and (b) the E-state of LACOF-1 (3:3). (c) Representative strongest CO₂ binding configurations for the E- and Z-states obtained from DFT calculations.

The molecular origin of the stronger E-state binding is evident from the local adsorption geometry. In the E-state, the extended AAP configuration exposes the pyrazole nitrogen atoms, creating favorable CO₂ binding pockets through electrostatic interactions (Figure S25). In contrast, the bent Z-AAP configuration partially shields these nitrogen atoms and produces a less favorable local binding geometry, resulting in weaker CO₂ binding. Together, the computational and experimental results establish that molecular photoisomerization programs local host–guest interactions by controlling the accessibility and geometry of the pyrazole-containing binding sites, thereby tuning adsorption energetics through changes in the local binding environment rather than through variations in surface area or pore volume.

2.4. Dynamic Gas Separation under Continuous-Flow Conditions

Having established that molecular photoisomerization programs the local CO₂ binding environment without disrupting the framework, we next asked whether these molecular-level changes translate into dynamic gas separation under continuous-flow conditions. Multicomponent breakthrough analysis provides a stringent test of whether light-controlled host–guest interactions translate directly into dynamic CO₂ capture under continuous-flow conditions. Prior to each measurement, the packed COF bed was irradiated with either 450 nm blue light or 365 nm UV light to generate the E- or Z-dominant state, respectively. Breakthrough experiments were then performed by alternately switching between E and Z states over 10 consecutive cycles using the same sample, enabling direct evaluation of reversible light-modulated adsorption.

Dynamic breakthrough measurements were performed for LACOF-1 (2:4), (3:3), and (4:2) under 15% and 5% CO₂ feed conditions (Figures S26–S30), and the corresponding uptake capacities are summarized in Figure and Table S4. The dynamic CO₂ uptake of LACOF-1 falls within the range reported for selected COFs under their respective adsorption conditions, while additionally providing reversible light-modulated adsorption, humid tolerance, and mild regeneration (Table S4). ,, Remarkably, LACOF-1 (4:2) exhibits the highest dynamic CO₂ uptake among the series despite possessing the lowest BET surface area (161 and 182 m² g^–1 for the E- and Z-states, respectively). This observation indicates that surface area alone does not determine breakthrough performance and suggests that local adsorption energetics also contribute substantially to CO₂ uptake. The higher density of AAP units creates more confined, pyrazole-rich adsorption environments that strengthen local CO₂ binding and compensate for the reduction in surface area and pore volume. Consequently, LACOF-1 (4:2) outperforms the more porous LACOF-1 (3:3) and (2:4) samples under breakthrough conditions. These results further support the conclusion that local adsorption energetics play a key role in governing CO₂ capture in this photoresponsive framework.

4.

Dynamic CO₂ breakthrough performance of LACOF-1. (a) Average CO₂ uptake of E- and Z-LACOF-1 (2:4), (3:3), and (4:2) under a dry 15% CO₂/N₂ feed at 22 °C. (b,c) CO₂ uptake of E- and Z-LACOF-1 (4:2) over 10 breakthrough cycles under dry feeds containing (b) 15% and (c) 5% CO₂ in N₂ at 22 °C. (d) CO₂ uptake of E- and Z-LACOF-1 (4:2) over 10 breakthrough cycles under a humid 15% CO₂/N₂ feed at 15% RH and 22 °C.

Under a mixed-gas feed containing 15% CO₂ in N₂ at 22 °C, distinct breakthrough profiles were observed for the two photoinduced states, with the E-state consistently exhibiting delayed CO₂ breakthrough and higher adsorption capacity than the corresponding Z-state. For LACOF-1 (4:2), CO₂ uptake reached ∼1.35 mmol g^–1 for the E-state, compared with ∼1.00 mmol g^–1 for the Z-state under dry conditions (Figure a). The enhanced uptake of the E-state is consistent with the stronger local host–guest interactions revealed by gas adsorption measurements and multiscale simulations. The differences in breakthrough time and uptake remained reproducible over repeated switching cycles, confirming that the adsorption behavior can be reversibly modulated by light without measurable loss of performance (Figure b). In addition, the dynamic adsorption process exhibited rapid kinetics, reaching half-capacity within ∼3 min under these conditions (Figure S26). Collectively, these results demonstrate that molecular photoisomerization can translate molecular-level differences in host–guest interactions into reversible dynamic gas separation under continuous-flow conditions.

To evaluate the performance of LACOF-1 at lower CO₂ concentration, breakthrough experiments were further conducted using a 5% CO₂/N₂ feed at 22 °C. Under these conditions, the E- and Z-states of LACOF-1 (4:2) exhibited CO₂ uptake capacities of ∼0.52 and ∼0.36 mmol g^–1, respectively (Figures S27, c and Table S4). The E-state maintained a higher CO₂ uptake than the corresponding Z-state at both 15% and 5% CO₂. Because adsorption at lower partial pressure is increasingly influenced by the highest-affinity binding sites, this result further supports the conclusion that light-modulated local host–guest interactions contribute to the enhanced CO₂ capture performance of the E-state.

The persistence of light-modulated adsorption under humid conditions further demonstrates that the AAP-dependent difference between the E- and Z-states is retained in the presence of water, an important consideration for postcombustion CO₂ capture. Under humid conditions (15% CO₂, 15% RH, 22 °C), LACOF-1 (4:2) exhibited CO₂ uptake capacities of ∼0.96 and ∼0.48 mmol g^–1 for the E- and Z-states, respectively (Figures d and S28). Although competitive adsorption by water reduced the absolute CO₂ uptake relative to dry conditions, the E/Z uptake difference and stable cycling performance were maintained over 10 consecutive adsorption–desorption cycles. Furthermore, the adsorbed CO₂ could be readily removed by N₂ purging at ambient temperature, demonstrating facile regeneration under mild operating conditions.

Postbreakthrough characterization further confirmed that the framework remained structurally intact after repeated operation. PXRD patterns, FT-IR spectra, and gas sorption isotherms remained essentially unchanged from those of the as-synthesized material (Figure S31), indicating preservation of crystallinity, chemical structure, and permanent porosity after repeated adsorption–desorption cycling. Collectively, these results demonstrate that reversible molecular photoisomerization can be translated into dynamic gas separation without compromising framework integrity. The combination of rapid photoresponsive switching, excellent cycling stability, facile regeneration, and tolerance to humid conditions highlights the promise of this molecular design strategy for reversible, light-modulated CO₂ capture under mild operating conditions.

3. Conclusion

In summary, we have developed a series of photoresponsive covalent organic frameworks that enable reversible, light-modulated CO₂ adsorption through molecular photoisomerization. Whereas previously reported photoresponsive porous materials have often relied on changes in pore accessibility to modulate adsorption, LACOF-1 demonstrates that local host–guest interactions can be tuned independently of overall pore accessibility through arylazopyrazole (AAP) photoswitches within a crystalline imine-linked framework. The embedded AAP units undergo efficient and reversible E–Z photoisomerization with excellent thermal stability while preserving the crystallinity and permanent porosity of the framework. This molecular switching translates directly into distinct adsorption behavior, with the E-state consistently exhibiting stronger CO₂ affinity and higher dynamic CO₂ uptake than the corresponding Z-state under dry, dilute, and humid conditions. Multicomponent breakthrough experiments further demonstrate reversible light-modulated adsorption, excellent cycling stability, and facile regeneration by ambient-temperature N₂ purging. Combined experimental and computational studies reveal that the enhanced CO₂ affinity of the E-state originates from stronger local host–guest interactions associated with the extended E-AAP configuration rather than from larger surface area or pore volume. More broadly, this work establishes molecular photoswitching as an effective strategy for programming adsorption energetics in crystalline porous materials and provides a design principle for developing adaptive porous adsorbents with externally tunable gas adsorption behavior.

4. Experimental Section

4.1. Synthesis of LACOF-1

LACOF-1 was synthesized with modifications to a reported procedure. The molar ratio of 2,5-dimethoxyterephthalaldehyde (DMTA) and the AAP-functionalized linker 4,4″-bis­((E)-(1-methyl-1H-pyrazol-3-yl)­diazenyl)-[1,1′:4′,1″-terphenyl]-2′,5′-dicarbaldehyde ((E,E)-MPDTD) was varied to afford a series of COFs with different AAP contents. As a representative example, LACOF-1 (3:3) (Figure a) was synthesized by condensing 1,3,5-tris­(4-aminophenyl)­benzene (TPB, 0.085 mmol, 30.0 mg), DMTA (0.064 mmol), and (E, E)-MPDTD (0.064 mmol) in n-BuOH (1 mL) with aqueous acetic acid (6 or 9 M) in a Pyrex tube (7 mL). The reaction mixture was sonicated for 30 s, flash frozen in liquid N₂, and degassed by three freeze–pump–thaw cycles. The tube was then sealed after being flushed with N₂ and heated at 120 °C for 3 days. The resulting precipitate was collected by centrifugation, washed with THF (×6), and further purified by Soxhlet extraction with THF for 24 h. The product was then collected and dried under vacuum at 80 °C overnight to afford the corresponding COF. All polymerization and purification steps were carried out in the absence of light.

4.2. Photoisomerization Study of LACOF-1

LACOF-1 samples were dispersed in 2-ethoxyethanol and irradiated with 450 or 365 nm light at a distance of 1–2 cm from the sample. Photoisomerization kinetics and thermal stability were monitored by time-resolved UV–vis spectroscopy, while solid-state irradiation experiments were performed for 2 h under temperature-controlled conditions. Additional experimental details are provided in the Supporting Information.

4.3. Gas Sorption and Breakthrough Measurements

Single-component gas sorption isotherms were measured on a Micromeritics ASAP 2020 Plus after activation of the samples under dynamic vacuum at 120 °C for 12 h. Dynamic breakthrough experiments were performed using a Micromeritics Breakthrough Analyzer equipped with a mass spectrometer. Approximately 100 mg of LACOF-1 was packed into a quartz column and irradiated in situ with 450 nm blue light or 365 nm UV light for 2 h prior to measurement to generate the E- or Z-dominant state. Breakthrough experiments were conducted under dry and humid mixed-gas feeds containing 15% or 5% CO₂. CO₂ adsorption capacities were determined by subtracting the corresponding blank-column contribution from the measured breakthrough capacity and normalizing the resulting value by the mass of the COF sample. Additional experimental details are provided in the Supporting Information.

4.4. Computational Methods

Molecular dynamics (MD), grand canonical Monte Carlo (GCMC), and density functional theory (DFT) calculations were performed to investigate framework stability and the origin of the photoresponsive adsorption behavior. Representative E- and Z-state LACOF-1 models were constructed from AA-stacked TPB–DMTP-COF structures with varying AAP-to-methoxy ratios. MD simulations were used to evaluate structural stability and generate representative framework configurations. GCMC simulations were employed to identify preferred CO₂ adsorption regions and adsorption hotspot distributions, and DFT calculations were performed to quantify local CO₂ binding interactions. Full computational details are provided in the Supporting Information.

The data that support the findings of this study are available from the corresponding author upon reasonable request.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.chemmater.6c02293.

• The authors have cited additional references within the Supporting Information. − Experimental procedures; synthesis and characterization of molecular precursors and LACOF-1; NMR, mass spectrometry, FT-IR, PXRD, TGA, and UV–vis data; photoisomerization kinetics and stability; gas sorption, pore-size, heat-of-adsorption, and selectivity analyses; breakthrough adsorption and blank-correction data; molecular dynamics, GCMC, and DFT methods and results (PDF)

X.Y. synthesized the materials, performed material characterization, gas sorption measurements, and breakthrough adsorption experiments, analyzed the experimental data, and wrote the original draft of the manuscript. M.D. performed computational studies, including molecular dynamics simulations, grand canonical Monte Carlo simulations, and density functional theory calculations, and contributed to the interpretation of the computational results. T.K. conducted the solid-state NMR measurements and data analysis. A.R. performed gas adsorption isotherm measurements and assisted with data collection. H.C. supervised the computational studies, secured funding for the computational work, and contributed to the study design and interpretation of the simulation results. L.Q. contributed to project supervision and manuscript revision. Y.L. conceived and supervised the project, secured funding, guided data analysis and interpretation, and revised the manuscript. All authors discussed the results, contributed to the preparation of the manuscript, and approved the final version of the manuscript.

The authors declare no competing financial interest.