CONSPECTUS
The development of adsorbents acting as “molecule traps” for advanced adsorption and separations on specific molecular recognition has garnered increasing attention over the past decade. Molecule trapping can be understood as strong confinement of guest molecules within well-defined local environments inside adsorbents, where adsorption is governed by geometric matching and multiple host−guest interactions rather than nonspecific van der Waals interactions alone. This site-specific trapping strategy enables precise discrimination between molecules with similar physicochemical properties under complex separation conditions. In particular, metal−organic frameworks (MOFs) provide an ideal platform for constructing molecule traps because their modular, designable, and crystalline structures enable the precise spatial organization of binding sites and confined pore environments.
This Account highlights our recent advances in MOF-based molecule-trap adsorbents, from the foundational concept of the single-molecule trap to diverse trapping modes, including aromatic, localized, dynamic, and reactive trapping. These trapping modes enable selective recognition and capture/separation of hydrocarbons, greenhouse gases, and air pollutants through distinct confined interaction environments. Beyond empirical design, molecule trapping is also increasingly being guided by data-driven approaches, including large language model-assisted extraction of structure−function relationships and mechanism-based analysis. Furthermore, advances in scalable synthesis, material shaping, and continuous operation have moved MOF-based molecule-trap adsorbents closer to practical separation processes. These studies suggest that molecule trapping can serve as a useful design strategy for adsorption and separation media, providing a conceptual framework for the development of next-generation porous materials.
1. INTRODUCTION
Adsorption-based technologies are widely used for molecular separation and purification owing to their intrinsic energy efficiency and operational simplicity, supporting applications ranging from chemical manufacturing to environmental remediation. However, as separation targets become increasingly demanding, particularly for trace components, structurally similar molecules, and multicomponent mixtures, the limitations of conventional adsorbent design have become increasingly evident. Under these conditions, separation efficiency is governed not merely by whether molecules can be adsorbed, but by how effectively they can be distinguished. In many cases, the differences between target species are too subtle to be resolved through traditional design variables such as pore size, surface area, or overall adsorption affinity.
Metal−organic frameworks (MOFs) are particularly suited for molecular recognition because their crystalline and modular structures enable precise control over pore geometry and binding environments. Representative studies have demonstrated several approaches to translating this structural control into selective adsorption, spanning applications in hydrocarbon separations, carbon capture, and gas purification. Fe₂(dobdc) preferentially binds olefins over the corresponding paraffins at its open Fe(II) sites, whereas UTSA-16 combines suitable pore cages with terminal water sites for selective CO₂ capture. In NOTT-300, multiple weak supramolecular interactions are cooperatively aligned within the pore to recognize C2 hydrocarbons. SIFSIX-1-Cu and SIFSIX-2-Cu-i further illustrate how binding-site chemistry and pore dimensions can be matched to C₂H₂, affording high C₂H₂ capacity and C₂H₂/C₂H₄ selectivity, respectively. Collectively, these studies show that selective adsorption can be rationally engineered by tailoring local pore structures, where pore chemistry and binding interactions are coordinated at different spatial scales.
In this context, our work has addressed a more specific question: whether a binding environment can be predefined for precise target-molecule recognition with improved sensitivity (gases at low concentration) and selectivity (coexisting gases with little difference in physical/chemical properties). In 2013, we realized this idea in PCN-88 and introduced the concept of a predesigned single-molecule trap (SMT). Opposing Cu(II) sites define a confined cavity that accommodates one CO₂ molecule in a specific configuration, enabling molecular recognition at the level of an individual trapping site. Beyond this initial design, our subsequent studies have explored diverse molecule-trapping modes, including aromatic, localized, dynamic, and reactive trapping (Figure 1). These trapping modes enable selective separation by stabilizing target molecules, regulating their orientation and diffusion, or inducing reversible chemical transformations within confined pores. This Account summarizes our recent advances in MOF-based molecule-trap adsorbents, focusing on the design principles, trapping mechanisms, and representative separation applications of different trapping modes. Finally, we further discuss how molecule trapping serves as a useful strategy for realizing molecular recognition and separation in porous materials under increasingly complex operating conditions.
2. FUNDAMENTALS OF MOF MOLECULE-TRAP ADSORBENTS
Early studies on MOF adsorbents showed that selective adsorption can arise from the combined effects of pore confinement and framework functionality. In 2008, Li et al. reported a stable guest-free MOF with nitrogen-rich chiral channels, where the confined and chemically functionalized pore environment enabled selective gas adsorption. This study indicated that guest recognition in MOFs can be directed by specific pore environments rather than solely by pore size, providing an early basis for our subsequent works on the development of molecule-trap adsorbents. Building on this foundation, the SMT concept further defined adsorption sites at the level of individual guest molecules. This section discusses how this concept has evolved into several representative trapping modes, including single-molecule, aromatic, localized, dynamic, and reactive trapping. These categories are intended as a working classification rather than mutually exclusive definitions, because many MOF adsorbents combine multiple trapping features in the same pore environment.
2.1. Single-Molecule Trap
The concept of the SMT represents an early realization of site-defined molecule trapping in MOFs, in which a discrete binding environment hosts a single guest molecule with a defined configuration. SMTs rely on preorganized local environments whose size, geometry, and electrostatic field are matched to the target molecule, thereby making adsorption a highly site-specific trapping process. This concept was demonstrated in PCN-88, where pairs of opposing Cu(II) sites defined a confined cavity that matches the size and charge distribution of CO₂ (Figure 2a). The distance between the two metal sites (7.4 Å) is comparable to the molecular length of CO₂, thereby providing effective geometric confinement (Figure 2b). Structural analyses and simulation studies demonstrated that CO₂ is preferentially stabilized near these sites through cooperative electrostatic interactions, whereas CH₄ and N₂ interact only weakly with the pore surface (Figure 2c).
Beyond geometric matching, SMT performance can be further enhanced by cooperative interactions within confined environments. As later reported by Ma and co-workers, pairs of closely spaced open Cu(II) sites (~4.43 Å) in a MOF named by Cu-ATC form a dual-site nanotrap for C₂H₂, enabling simultaneous stabilization of a single molecule via coordination-like interactions with Cu(II) sites and additional dispersive interactions from the surrounding pore environment. As a result, Cu-ATC exhibits a low-pressure uptake of 2.1 mmol g⁻¹ at 10⁻³ bar and a high C₂H₂/CO₂ selectivity of 53.6. These examples show that SMT design depends on constructing a preorganized cavity or dual-site pocket that matches a single guest molecule in size, binding-site distance, and electrostatic distribution, thereby converting adsorption into a one-molecule-per-site recognition process.
2.2. Aromatic-Molecule Trap
Aromatic trapping is characterized by spatially continuous interaction environments, in which adsorption is governed by the cumulative effect of multiple weak interactions within confined geometries, including π−π stacking, C−H···π contacts, and dispersion interactions. An early example is BUT-66, a Zr(IV)-based framework constructed from angular aromatic ligands. The combination of ligand angularity and framework interpenetration in BUT-66 narrows the pore space into one-dimensional channels lined predominantly with hydrophobic phenyl groups. This confined aromatic environment gives rise to steep benzene uptake at low pressure and high volumetric capacity: BUT-66 adsorbs 2.54 mmol g⁻¹ benzene at 0.012 kPa and 25 °C, and reaches 1.75 mmol cm⁻³ at 0.12 kPa and 80 °C, exceeding ZIF-8, MCM-41, MIL101(Cr), PAF-1, and Carboxen 1000 under comparable low-pressure conditions (Figure 3a). Single-crystal analysis of C₆H₆@BUT-66 reveals two well-defined benzene adsorption sites, one inside the window defined by paired 4,4,4′-(benzene-1,3-diyl)dibenzoate (BDB²⁻) ligands and the other at the center of the one-dimensional channel. At both sites, benzene is stabilized by edge-to-face C−H···π contacts with surrounding phenyl rings, with shortest distances of 2.97–3.43 Å.
A more cooperative aromatic trapping environment is found in BUT-55, a double-walled Co(II)−dipyrazolate framework. Adjacent BDP²⁻ (H₂BDP = 1,4-di(1H-pyrazol-4-yl)benzene) ligands adopt a criss-cross packing mode, generating one-dimensional rectangular pores and square channels lined by paired aromatic surfaces. This framework differs from the single-walled supramolecular isomer Co(BDP), despite their identical composition, and leads to distinct pore structure, pore chemistry, and benzene adsorption properties. BUT-55 adsorbs 3.28 mmol g⁻¹ benzene at only 7.3 Pa, whereas several benchmark adsorbents require much higher pressures to reach comparable uptake (Figure 3b). Single-crystal analysis of C₆H₆@BUT-55 identifies two benzene adsorption sites. At site I, benzene is confined between neighboring BDP²⁻ ligand pairs and Co(pz)₂ rod building blocks, where multiple CH···π and C−H···N interactions contribute to stabilization, with distances of 2.89–3.58 Å. At site II, benzene lies above paired BDP²⁻ ligands and forms additional C−H···π interactions with distances of 2.78–3.88 Å; these site-II molecules further assemble into a helical chain through guest−guest C−H···π interactions.
2.3. Local-Molecule Trap
Localized trapping is defined by ultralocalized interaction domains in which multiple binding elements are confined within a subnanometer region, generating a concentrated interaction field. Unlike SMTs, which rely on geometrically predefined cavities to host individual molecules, localized trapping emphasizes the spatial convergence of functional groups or metal centers, enabling multiple short-range interactions to converge on the same guest molecule. Accordingly, Cu-ATC is classified as an SMT because its dual Cu sites constitute a predefined pocket for one C₂H₂ molecule, whereas BUT-316 is classified as an LMT because its shorter Cu···Cu separation (~2.5 Å) generates an ultralocalized interaction domain that differentiates the binding geometries of C₂H₂ and CO₂ (Figure 4a). Within its pocket, a single C₂H₂ molecule adopts a side-on configuration and interacts simultaneously with both metal centers. This dual coordination geometry enforces a specific binding mode that is not accessible in more extended environments. In contrast, CO₂ adopts an end-on configuration with weaker interaction overlap, indicating that selectivity arises from differences in interaction geometry within the same local domain rather than solely from pore size exclusion. The resulting local-molecule trap combines high C₂H₂ packing density with moderate adsorption strength, distinguishing it from conventional open-metal-site or large-pore adsorbents.
Localized trapping can also be realized in hydrogen-rich narrow channels, as demonstrated by BUT-167. This Ni(II)-pyrazolecarboxylate framework contains one-dimensional channels of ~3 Å, with pyrazole N−H groups directed toward the pore center. This arrangement creates a confined local environment in which linear guest molecules are surrounded by multiple hydrogen atoms from the framework. Single-crystal structures of gas-loaded BUT-167 show that both N₂O and CO₂ are uniformly located at the center of the channel without crystallographic disorder. N₂O is stabilized by multiple N/O···H contacts with distances of 2.8–3.3 Å, while CO₂ forms analogous O···H contacts of 2.8–3.2 Å (Figure 4b). These contacts are individually weak, but their convergence within a narrow channel generates a localized interaction domain for N₂O and CO₂ binding. The effect of this local binding environment is reflected in the adsorption behavior of BUT-167. At 298 K, N₂O and CO₂ exhibit steep uptake at low pressure, whereas N₂ shows much weaker adsorption over the same pressure range (Figure 4c). BUT-167 reaches a volumetric N₂O uptake of 135.8 cm³ cm⁻³ at 40 kPa and a CO₂ uptake of 101.5 cm³ cm⁻³ at 4 kPa, exceeding representative porous adsorbents in the low-pressure range (Figure 4d).
2.4. Dynamic-Molecule Trap
Dynamic trapping operates through the coupling between host−guest interactions and molecular transport, where selectivity is determined not only by equilibrium binding strength but also by kinetic accessibility within confined pores. Such behavior does not require large-scale framework breathing and may instead arise from interaction-regulated local confinement or aperture-controlled diffusion. In this mode, subtle differences in molecular size, shape, or polarizability can be amplified through interaction-induced local responses, constrained flexibility, or dynamic confinement. A representative example is the pyrazolate-based framework BUT-53 (Figure 5a). Although the framework is crystallographically rigid, its cavity environment remains dynamically responsive due to a dense array of hydrogen atoms lining the pore surface. Single-crystal analysis shows that SF₆ is stabilized through multiple F···H interactions (2.6–3.0 Å), forming two distinct adsorption sites within the cavity. These interactions generate a local field that both stabilizes the highly polarizable SF₆ molecule and modulates its diffusion pathway. Thus, SF₆ exhibits strong binding and restricted mobility, whereas smaller, less polarizable molecules such as N₂ interact more weakly and show less hindered diffusion.
The second form of dynamic trapping involves aperture-gated transport, as illustrated by Zn-ATZ-IP(OH)-a (Figure 5b). In this framework, strong intraframework O−H···N and O−H···O hydrogen bonds constrain the pore aperture, giving a well-defined window size of 3.8–4.2 Å. This dimension lies between the kinetic diameters of C₃H₆ (~4.0 Å) and C₃H₈ (~4.3 Å), enabling discrimination based on diffusion accessibility. While C₃H₆ can pass through the aperture with relatively low energy barriers, C₃H₈ experiences hindered transport due to steric constraints. Importantly, this selectivity does not require a large-scale framework transformation, but arises from a narrowly defined aperture that permits only limited local motion. Thus, subangstrom differences in molecular size are converted into substantial differences in diffusion rates, producing a sensitive kinetic gating effect.
2.5. Reactive-Molecule Trap
Reactive trapping is distinguished by the formation and stabilization of chemical species within the framework during guest capture. Unlike nonreactive trapping modes, where selectivity is mainly governed by physical confinement and weak host−guest interactions, reactive trapping follows a reaction pathway in which the framework stabilizes reactants, intermediates, or charged species while maintaining sufficient reversibility for regeneration. A representative example is the Zr(IV)-based framework BUT-86, which captures SO₂ through reversible formation of bisulfite under humid conditions (Figure 6a). The trapping process occurs at the interface of Zr₆ clusters and proximal pyrazolate groups, where SO₂ reacts with framework −OH groups to form HSO₃⁻ species. Structural and spectroscopic analyses show that the bisulfite intermediate is stabilized through coordination to the metal node and hydrogen-bonding interactions (Figure 6b). The isosteric heat (Qst) of adsorption for SO₂ is 103.5 kJ mol⁻¹, indicating strong yet reversible sorption.
Reactive trapping can also proceed through hydration-mediated proton transfer, as illustrated by BUT-64(H₂O) (Figure 6c). In this framework, bridging water molecules act as proton donors during NH₃ capture, converting NH₃ into NH₄⁺ species that are stabilized by a hydrogen-bonding network within the pores (Figure 6d). The corresponding Qst value of 101.0 kJ mol⁻¹ reflects strong interactions associated with proton transfer, while the process remains fully reversible. Another distinct form of reactive trapping involves framework redox chemistry, as demonstrated by BUT-45. In this case, O₂ adsorption proceeds via electron transfer at Co₆ clusters, leading to partial oxidation of Co(II) to Co(III) and the formation of superoxo/peroxo species. Spectroscopic evidence supports these intermediates, with in situ Fourier-transform infrared (FT-IR) spectroscopy showing bands at 1166 and 1274 cm⁻¹ assigned to superoxo species in distinct coordination environments. Concomitantly, structural changes occur at the cluster level, including the contraction of Co−N bonds (from 2.00 to 1.92 Å) and a reduction in unit cell volume of over 6%, indicating strong coupling between electronic and structural transformations.
2.6. LLM-Accelerated Molecule-Trap Discovery
The discovery of molecule-trap adsorbents relied primarily on accumulated literature knowledge, empirical comparisons of framework structures, and iterative experimental screening. However, the descriptors that govern trap-based recognition are often highly localized and condition-dependent, including pore-limiting aperture, dynamic window size, local polarity, and specific host−guest interactions. These features are difficult to capture using simple structural metrics, such as crystallographic pore size or surface area. Therefore, accelerating molecule-trap discovery requires methods that can extract, organize, and interpret structure−function relationships from dispersed experimental reports.
In this context, large language models (LLMs) offer a useful route to transform unstructured literature information into structured design knowledge. One example is the data-driven discovery of MOFs for quinary C5 olefin separation. In this study, LLMs were used to extract key information from the existing sieving literature and construct structure profiles for individual MOFs, including their pore structure, chemical composition, adsorption behavior, and experimental conditions. After prompt optimization and data curation across 1067 publications, a refined data set containing 345 MOFs was obtained. Based on the extracted adsorption and exclusion data, an effective pore size (EPS) range was assigned to each MOF using the molecular dimensions of the adsorbed and excluded guests. Because EPS is derived from experimental results, it reflects the working pore aperture, including the effects of framework flexibility that are absent from static crystallographic pore sizes. The EPS ranges of 345 MOFs were then compared with the dimensions of the five C5 components. This screening identified ZIF-8, which adsorbs 1-pentene, cis-2-pentene, trans-2-pentene, and n-pentane while largely excluding isoprene. Breakthrough experiments consequently afforded 99.99% pure isoprene in a single step. This result demonstrates how LLM-extracted experimental data can identify aperture-controlled molecule traps for a specific separation.
More recently, LLMs have advanced from knowledge extraction to mechanism-oriented reasoning. A representative example is the MOFReasoner framework, in which reasoning pathways were distilled from over 8242 research articles and 500 reviews to construct a domain-specific data set comprising 35800 entries. Through knowledge distillation, the model achieved an accuracy exceeding 90%, significantly outperforming general chemistry tasks (~50%). The resulting model can reconstruct relationships between local interaction environments, molecular geometry, and adsorption behavior, enabling hypothesis generation and mechanism-oriented analysis. Looking forward, this reasoning capability holds great promise for untangling complex cooperative host−guest interactions to guide the predictable, mechanism-driven design of advanced molecule-trap adsorbents.
3. APPLICATIONS OF MOF MOLECULE-TRAP ADSORBENTS
Having outlined the fundamental trapping modes and local interaction mechanisms, we now introduce how these principles are applied to representative separation and purification scenarios (Table 1). The following sections focus on hydrocarbon separation, greenhouse gas capture, and air pollutant control under different molecular systems and operating conditions, followed by scale-up and manufacturing considerations relevant to practical implementation.
3.1. Hydrocarbon Separation and Purification
Hydrocarbon separation and purification represent important application scenarios for MOF molecule-trap adsorbents because many industrial streams contain molecules with similar sizes, shapes, boiling points, and adsorption affinities. These similarities make conventional distillation energy-intensive, thereby creating opportunities for adsorbents that can amplify subtle molecular differences through confined recognition environments. The trapping mode can be selected according to the molecular difference involved: localized interaction environments amplify differences in adsorption affinity, aromatic traps recognize specific molecular structures, and aperture-controlled dynamic traps translate small size differences into transport selectivity. For multicomponent mixtures, adsorbents with complementary recognition windows can be combined sequentially.
For natural-gas upgrading, where C₂H₆ and C₃H₈ are more polarizable than CH₄, localized traps can amplify their differences in adsorption affinity. BUT-315 demonstrates that high selectivity can be achieved without relying on excessively strong adsorption, with selectivities of 54.9 for C₂H₆/CH₄ and 545.8 for C₃H₈/CH₄ at 298 K, together with a high separation potential (ΔQ=26.54 mmol g⁻¹). This performance is maintained under 70% relative humidity (RH), and dynamic breakthrough experiments show a CH₄ productivity of 10.76 mmol g⁻¹ for ternary mixtures. Olefin purification requires the trapping environment to be adjusted according to both the identity and number of impurities. For trace C₂H₂ removal, Cu(OH)INA combines oxygen binding sites with pore confinement to promote dense C₂H₂ packing, reaching 351 mg cm⁻³ at 0.01 bar with a moderate adsorption enthalpy (36 kJ mol⁻¹). When C₂H₂ and C₂H₆ coexist, BUT-321 provides an oxygen-rich pore environment that preferentially retains both impurities, enabling the one-step purification of C₂H₄ from ternary C₂H₂/C₂H₄/C₂H₆ mixtures, with dynamic uptakes of 1.00 and 1.02 mmol g⁻¹ for C₂H₂ and C₂H₆ respectively.
When the molecular differences become smaller, adsorption and diffusion must be regulated more precisely. In C₂H₆/C₂H₄ separation, MOFs such as Ni-cam-pyz achieve C₂H₆ uptakes of up to 68.75 cm³ g⁻¹ at 298 K. For C3 separations, the pore environment can be adjusted toward either size exclusion or guest-responsive transport: ultramicroporous BUT-306 selectively adsorbs C₃H₄ (29.6 cm³ g⁻¹ at 298 K) while excluding C₃H₆, whereas flexible BUT-309 enhances separation through a guest-induced response. For molecules with similar sizes but different chemical structures, aromatic trapping provides another form of recognition. In the benzene/cyclohexane system, Zn-Ade-TCPE selectively confines benzene within its adenine-lined slit pores, achieving adsorption selectivities of 216−1027 and removing over 99% of trace benzene (1000 ppm) to produce cyclohexane with >99.999% purity. A single trapping environment may be insufficient for multicomponent mixtures containing several closely sized molecules. In the quinary C5 separation, adsorbents with different effective pore-size windows are therefore arranged sequentially: ZIF-8 separates isoprene, followed by Co-FA, Co(pz-NH₂)Ni, and Co-gallate for the remaining components (Figure 7a). This combination enables the complete isolation of isoprene, 1-pentene, cis-2-pentene, trans-2-pentene, and n-pentane with high purity (Figure 7b).
3.2. Greenhouse Gas Capture
Greenhouse gas capture presents challenges distinct from those of hydrocarbon separations because many target gases occur at low concentrations, are difficult to selectively recognize, and compete with moisture under practical conditions. Effective adsorbents must therefore combine selective recognition with rapid transport, high regenerability, and structural robustness. An early example is CO₂ capture in PCN-88, where site-defined trapping enables the selective adsorption of CO₂ over N₂ and CH₄ under ambient conditions, demonstrating the feasibility of molecule trapping for greenhouse gas separation.
More demanding challenges arise for non-CO₂ greenhouse gases such as SF₆, which combine a large molecular size, high polarizability, and low concentrations in industrial streams. In BUT-53, dynamic confinement within a hydrogen-rich cavity enables an SF₆ uptake of 2.82 mmol g⁻¹ at 0.1 bar and an SF₆/N₂ selectivity of 2485 at 298 K, while maintaining performance under humid conditions (90% RH). Cooperative trapping provides another route to strengthen SF₆ recognition. In Ni-bpz, adjacent nanotraps allow a single SF₆ molecule to interact simultaneously with multiple binding sites, leading to SF₆/N₂ selectivities of up to 516 for dilute mixtures (1/99). Local electronic tuning offers further control over SF₆ binding. In BUT-125, pyridine functionalization increases the electrostatic potential within the trapping sites, giving an adsorption capacity of 3.57 mmol cm⁻³ at 0.1 bar, a selectivity of 541.8, and a dynamic capture capacity of 3.42 mmol cm⁻³.
Beyond single-component target streams, molecule-trap adsorbents can also address multicomponent greenhouse gas mixtures. The Ni(II)-pyrazolecarboxylate framework BUT-167 enables the simultaneous capture of N₂O and CO₂ from a quaternary mixture (N₂O/N₂/CO₂/O₂), with volumetric capacities of 135.8 cm³ cm⁻³ for N₂O and 101.5 cm³ cm⁻³ for CO₂. The corresponding selectivities reach 135.7 (N₂O/N₂) and 257.6 (CO₂/N₂), while the separation performance is maintained under humid conditions.
3.3. Air Pollutant Control
Air pollutant control imposes requirements distinct from those of conventional gas separations because pollutant removal is often conducted under humid, chemically aggressive, and low-concentration conditions. Effective adsorbents must therefore combine strong trace-level affinity, high selectivity against water and coexisting species, structural stability, and cyclic durability. For volatile organic compounds (VOCs), particularly aromatic pollutants such as benzene, the key challenge is to achieve sufficient adsorption affinity at trace concentrations without sacrificing regenerability. Doublewalled metal−dipyrazolate frameworks such as BUT-55 show efficient benzene capture at sub-10 Pa pressures. Dynamic breakthrough experiments further confirm effective removal at ppm levels under humid conditions, highlighting the role of π-rich confined environments in the capture of trace aromatic molecules.
For chemically reactive air pollutants, the design challenge shifts from enhancing physisorption affinity to balance strong uptake with reversible regeneration. Strongly interacting acidic gases such as SO₂ require trapping mechanisms beyond conventional physisorption for deep removal. In the Zr(IV)-based framework BUT-86, trace SO₂ is removed under humid conditions to subppb levels, and the captured SO₂ can be converted into value-added products. In this case, reactive trapping enables strong and selective uptake while preserving reversibility. Basic pollutants such as NH₃ present a different challenge, as strong binding often compromises regeneration. In BUT-64(H₂O), NH₃ is captured under humid conditions with a packing density of 0.27 g cm⁻³, while reversible uptake is retained over repeated capture−regeneration cycles.
In practical emission streams, volatile organic compounds and reactive gases may coexist and compete for adsorption sites. The Ni(II)-pyrazolate framework BUT-78 addresses this challenge by enabling the simultaneous capture of benzene and SO₂ under humid conditions, with capacities of ∼0.51 mmol g⁻¹ for 10 ppm benzene and ~1.21 mmol g⁻¹ for 250 ppm of SO₂, while maintaining structural stability and recyclability. Similarly, indoor air pollutants such as formaldehyde (HCHO) require adsorbents that combine high efficiency at low concentration with practical processability. For instance, Cu(II)-based systems achieve a volumetric adsorption amount of 0.87 mg m⁻³ within 3 h, while Zn(II)-based frameworks such as BUT-236 combine pore-size matching and amine functionality to enhance adsorption performance and enable large-scale production (Chinese Patent No. CN119978419B, CN120132811B).
3.4. Scale-Up and Manufacturing
The practical implementation of molecule-trap adsorbents is often constrained by synthesis costs, scalability, shaping, and process compatibility, even after high separation performance has been demonstrated at the laboratory scale. For industrial use, adsorbents must be produced in sufficient quantity, processed into mechanically robust forms, and operated continuously over multiple capture−regeneration cycles. Therefore, the translation of molecule-trap adsorbents requires coordinated progress in scalable synthesis, material shaping, and long-term operation.
A primary requirement is the development of scalable and low-cost synthesis routes. Conventional solvothermal methods often rely on polar aprotic solvents such as N,N-dimethylformamide, require elevated temperatures, and generate substantial solvent waste, thereby increasing both costs and environmental burdens. To address these limitations, aqueous and atom-economical synthesis strategies have been developed. For example, Zn₂(atz)₂ipa can be synthesized in water at 60 °C within 10 min, affording ∼1.2 kg per batch with a yield of 91.3% and a space-time yield of 24 ton m⁻³ day⁻¹. The production cost is reduced to approximately $32 kg⁻¹. The kilogram-scale product shows adsorption behavior consistent with that of the laboratory-scale sample and can be used directly for CH₄/N₂ separation, with stable capacity over repeated regeneration cycles. Similarly, the use of metal oxide precursors avoids the accumulation of anionic byproducts and facilitates solvent recycling. This approach enables scalable MOF synthesis in reactors up to 10 L, reaching space−time yields of ~608 kg m⁻³ day⁻¹ while preserving both crystallinity and adsorption properties.
After scalable synthesis, powder adsorbents must be converted into practical forms suitable for industrial separation units. Using raw MOF powders directly in packed columns is challenging because they can cause pressure drops, dust formation, and poor mechanical stability, making direct use in packed columns difficult. Shaping is therefore essential, but it must be achieved without blocking pores or disrupting the local trapping environments responsible for selectivity. Pelletization offers a practical solution. For example, Cu(OH)INA can be processed into millimeter-scale pellets using hydroxyethyl cellulose as a binder, which forms interparticle hydrogen-bonded bridges that enhance mechanical strength while preserving the crystal structure. In fixed-bed breakthrough experiments, these pellets separate C₂H₂/C₂H₄ mixtures (1:99, v/v), producing ethylene with purity exceeding 99.99% and a productivity of 12.0 mol kg⁻¹. The retention of separation performance after shaping demonstrates that molecule-trap adsorbents can be successfully integrated into continuous-flow separation systems.
Long-term operation further requires that trapping performance be maintained over repeated capture−regeneration cycles. Strong adsorption or reactive trapping may lead to incomplete regeneration, irreversible binding, or gradual performance decay, thereby increasing operational costs. Consequently, regeneration and material reuse strategies are critical for practical deployment. In BUT-321, a full life-cycle approach integrates large-scale aqueous synthesis with 98% yield, pelletization, and recovery of spent materials (Figures 8a and 8b). Deactivated pellets are dissolved and recrystallized using the postsynthesis supernatant, achieving a reconstruction yield of up to 85%. The regenerated material recovers its crystallinity, surface area, and adsorption properties, showing nearly identical performance to the original sample (Figure 8c). This closed-loop strategy enables the repeated use of adsorbents while reducing raw material consumption and waste generation.
4. CONCLUSION AND OUTLOOK
In this Account, we summarize our studies over the past decade on MOF-type molecule-trap adsorbents, ranging from the single-molecule traps to aromatic, localized, dynamic, and reactive trapping modes. These studies show that adsorption selectivity can be achieved not only by pore size or overall adsorption affinity, but also by site-defined host−guest interactions, confined molecular orientation, regulated transport, and reversible in-pore transformation. This perspective establishes molecule trapping as a useful design concept for adsorbents operating under demanding separation conditions.
Nevertheless, several challenges must be addressed before molecule trapping can become a more predictive and broadly applicable design strategy. Current designs remain largely case-specific, and more general rules are needed to connect local structural features with separation performance in new target mixtures. Mechanistic understanding also requires further validation under working conditions, where competitive adsorption, moisture, concentration gradients, and flow may alter the dominant binding sites or transport pathways. In addition, shaping, binder addition, regeneration, and long-term cycling may affect the accessibility or chemical identity of the sites responsible for selectivity. Future studies may benefit from integrating data-guided discovery, operando characterization, and process-oriented material design to develop shaped, regenerable, and scalable adsorbents with preserved trapping functions.
Although industrial translation remains at an early stage, recent advances in scalable synthesis, shaping, and cyclic operation suggest that molecule-trap adsorbents may be increasingly relevant to practical adsorption processes. As demands continue to expand in petrochemical purification, carbon management, environmental protection, and energy-gas storage, molecule-trap adsorbent design should not only emphasize molecular recognition and selectivity, but also consider working capacity under practical operating conditions. Our recent methane-storage work on metal-dipyrazolate frameworks further extends molecule-trap adsorbent design from separation selectivity to storage-oriented working capacity through guest-induced framework transformation and pressure-window matching. By linking site-defined interactions with macroscopic separation performance and working capacity, this strategy provides a promising foundation for next-generation adsorption-based separations and storage systems.
期刊：ACCOUNTS OF CHEMICAL RESEARCH
DOI：https://doi.org/10.1021/acs.accounts.6c00409 作者：Guang-Rui Si, Lin-Hua Xie, Xin Zhang, Tao He, Xiang-Jing Kong, Jian-Rong Li* 单位：北京工业大学材料低碳循环利用国家重点实验室、化学工程系
