Hydration Topology Manipulates CO₂ Transport in Water-Saturated Micropores of Functionalized Covalent Organic Frameworks

Abstract

The confinement effect of nanochannels (<2 nm) dictates the gas–liquid distribution at solid–liquid interfaces, yet the effect of interfacial chemistry on their kinetic behaviors lacks insightful understanding. Particularly, the interfacial CO₂ transport mechanism is critical for CO₂ capture in hydrated environments. Herein, we employ nonequilibrium molecular dynamics simulations combined with density functional theory calculations to elucidate CO₂ transport mechanisms in fully hydrated one-dimensional micropores of covalent organic frameworks (COFs). Our results demonstrate that pore-wall chemistry regulates transport not directly through adsorption strength alone, but by reshaping the topology and dynamic organization of confined hydration networks. Without functionalization, the pristine COF channels form an ordered hydration structure that anchors CO₂ on adsorption sites and imposes substantial across-layer transport resistance. Functionalization of pore walls disrupts this hydration organization, accelerating CO₂ transport via distinct microscopic regimes and promoting spatial coexistence of CO₂ and water. Notably, fluorine functionalization establishes a mobility-dominated regime that minimizes friction, inverting the preference from H₂O-favored to CO₂-favored transport and achieving a nearly 3-fold increase in axial selectivity. This work reveals that CO₂ transport is controlled by hydration-mediated transport energetics. These findings offer molecular-level perspectives for design of microporous materials by highlighting importance of balancing interfacial affinity and transport resistance within confined hydration environments.

1. Introduction

Molecular transport through nanoconfined channels is central to membrane-based separations for energy and environmental applications, including carbon capture, electrochemical CO₂ conversion, and proton-conducting devices. − Conventional polymeric and inorganic membranes are often limited by disordered pore architectures and poor tunability, restricting their molecular discrimination capabilities at the sub-nanometer scale. − In contrast, two-dimensional covalent organic frameworks (2D COFs) provide atomically defined nanochannels, tunable pore environments, and highly ordered transport pathways, making them attractive model systems for investigating molecular transport mechanisms and designing next-generation separation membranes. −

Among the various 2D COF families, β-ketoenamine-linked COFs have attracted particular attention because of their exceptional chemical and hydrolytic stability arising from irreversible keto–enol tautomerization. , Their well-defined 1D pore channels and readily functionalized pore surfaces offer a unique platform for exploring how molecular transport responds to controlled variations in pore chemistry. , Such characteristics are especially relevant for membrane applications under humid or fully hydrated conditions, where the coupled transport of gas molecules and water critically influences overall device performance. −

Under nanoconfinement, water is an active participant in molecular transport rather than a passive background medium. − Confined environments profoundly reorganize hydrogen-bond networks, induce spatial density heterogeneities, and generate hydration structures that differ substantially from the bulk phase. − Crucially, within these sub-nanometer spaces, guest transport is no longer a simple gas-phase permeation; instead, it evolves into a competitive process controlled by water–gas partitioning at the pore interface. The intensely localized field near the pore walls dictates the spatial partitioning, local concentration, and orientational alignment of both water and CO₂. −

While interfacial water layers under low humidity can conditionally enhance CO₂ adsorption via capillary docking, this beneficial effect vanishes at full saturation. Under such conditions, guest molecules must traverse a dense hydrated environment instead of passing through partially vacant pores. Indeed, water saturation can drastically reduce CO₂ axial velocity by up to 96% as the confined water layers significantly occlude the available transport pathways. Despite extensive investigations of CO₂ adsorption and diffusion in porous materials, most studies have focused on dry or partially hydrated environments. , However, CO₂ transport in fully saturated nanochannels is fundamentally altered. A sophisticated three-phase (COF–H₂O–CO₂) interfacial interaction intrinsically determines gas transport behaviors. The pore-wall chemistry does not merely act as a static binding substrate for CO₂; rather, it primarily attracts or repels water molecules, thereby reshaping the local hydration topology. Although previous studies have heavily emphasized geometric descriptors such as pore size and shape, − how surface chemistry reshapes this hydration landscape to subsequently regulate gas energetics remains largely unexplored. Therefore, a molecular-level investigation linking pore chemistry, confined water organization, and molecular transport energetics is highly desired to understand and predict CO₂ transport under fully hydrated conditions.

To address this critical knowledge gap, in this study, we combine all-atom molecular dynamics (MD) simulations with density functional theory (DFT) calculations to investigate coupled H₂O/CO₂ transport in fully hydrated β-ketoenamine-linked COF micropores (<2 nm) functionalized with −OH, −F, and −SO₃H groups, respectively. Moving beyond traditional static adsorption models, we establish a comprehensive system that links pore-wall chemistry to gas transport kinetics via the spatial organization of confined water, the transient orientation of CO₂, and the resulting free-energy landscape. By integrating these descriptors, we elucidate how specific functional groups modulate the competition between molecular retention and mobility under full hydration. This framework provides molecular-level insights into gas transport in water-saturated nanochannels and offers design principles for 2D materials operating in realistic aqueous environments.

2. Methods

2.1. DFT Calculations

Two complementary levels of DFT calculations were employed, periodic simulations were used to optimize crystalline COF structures, while molecular fragment calculations were carried out to derive accurate atomic charges for force field parameterization. Periodic DFT calculations were performed with the CP2K package to optimize four representative COF unit cells and obtain reliable interlayer distances. The PBE functional with the MOLOPT-TZVP basis set was adopted, using a plane-wave cutoff of 400 Ry and a relative cutoff of 55 Ry. Structural optimizations were performed with the BFGS algorithm until the maximum atomic force converged below 0.001 Ha/Å. Additionally, CO₂–framework interaction energy landscapes were mapped via monolayer DFT calculations using the PBE0 hybrid functional with Grimme’s D3 (BJ) dispersion corrections. Crucially, a rigorous Counterpoise correction protocol was applied to eliminate the basis set superposition error (BSSE), with further details elaborated in Section S2, Supporting Information.

Atomic partial charges for the force field parameterization were derived from molecular fragment calculations using the Gaussian16 package. The minimal repeating fragments of the COFs were first geometrically optimized at the at the PBE/TZVP level of theory, with Grimme’s D3­(BJ) dispersion correction to account for van der Waals interactions. The electrostatic potential was subsequently evaluated, and restrained electrostatic potential (RESP) charges were obtained via Multiwfn package for subsequent molecular dynamics (MD) simulations.

2.2. MD Simulation Protocol and System Construction

For the MD simulations, the original triclinic COF primitive unit cell was transformed into an orthorhombic cell. A 1 × 2 × 6 supercell was then constructed to represent a six-layer AA-stacked COF, resulting in a simulation box with initial dimensions of 39.11 × 45.16 × 80 ų. The system was initially solvated with TIP3P water molecules, followed by rigorous energy minimization using the conjugate gradient algorithm until the maximum force was reduced below 0.1 kcal/(mol·Å). Subsequently, a 500 ps NPT equilibration was conducted at 300 K and 1 bar, employing the Nosé-Hoover thermostat and Parrinello–Rahman barostat to achieve a stable system density. This was followed by a 15 ns NVT equilibration to ensure the structural stability of the COF and an equilibrated water distribution, during which the confined water population reached a stable state, with the final 10 ns trajectory reserved for equilibrium hydration analysis. The resulting equilibrated structures, in which the microporous channels remained completely occupied by confined water molecules without persistent void regions, were defined as the fully hydrated (water-saturated) states and subsequently used as the configurations for hydration topology analyses.

2.3. Nonequilibrium MD (NEMD) for Gas Permeation

To simulate the gas permeation environment, the simulation box was extended by 10 Å along the Z-axis (final dimensions: 39.11 × 45.16 × 90 ų), and 100 CO₂ molecules were randomly inserted into the expanded region, creating a mixed CO₂/H₂O feed zone. Nonequilibrium MD simulations were then performed under a steady-state pressure gradient. Following force-driven NEMD protocols, a constant body force corresponding to an effective pressure of 100 MPa was applied along the permeation direction (+Z) within a 10 Å-thick feed region to drive the permeation of the CO₂/H₂O mixture. It should be noted that this high external pressure is employed as a computational tool to accelerate sampling and allow for the clear observation of relative transport trends within MD time scales, rather than to directly reproduce experimental operating pressures. Previous studies have demonstrated that this force-driven approach effectively maintains a steady chemical potential gradient while preserving numerical stability and reliable relative transport trends.

The system temperature was regulated at 300 K using a Langevin thermostat. The COF framework was parameterized using the PCFF force field, while H₂O and CO₂ were described by TIP3P and EPM2 models, respectively. To preserve the structural integrity during permeation simulations, harmonic positional restraints were applied to all framework atoms using a spring/self potential. The geometry of water molecules was constrained using the SHAKE algorithm. Long-range electrostatic interactions were computed using the Particle-Particle Particle-Mesh (PPPM) method with an accuracy of 1 × 10^–4 and a real-space cutoff of 12 Å. van der Waals interactions were treated via the Lennard-Jones potential with a consistent 12 Å cutoff and tail corrections to account for long-range dispersion. All NEMD production runs employed a 1 fs timestep and were conducted for 10 ns.

To assess the extent to which framework dynamics and external driving conditions may influence the simulated transport behaviors, additional control simulations were performed and discussed in Sections S4, Supporting Information.

3. Results and Discussion

3.1. Hydration Landscapes and Hydrogen-Bonding Networks

To systematically investigate the role of pore surface chemistry in confined mass transport, we selected β-ketoenamine-linked hexagonal 2D COFs as the crystalline platform for molecular simulations. This specific framework architecture is well-suited to investigate humid-state gas separation for the following reasons. The irreversible tautomerism involved in the β-ketoenamine linkage endows the frameworks with exceptional chemical and hydrolytic stability, which is essential to withstand the highly demanding environments of fully filled aquatic states. Furthermore, its aligned 1D nanochannels and easily tailorable aromatic linkers provide an ideal, non-tortuous model system for a controlled comparative study of directional molecular diffusion across different chemical environments.

As displayed in Figure a, we adopt four β-ketoenamine-linked hexagonal 2D COFs as objects, of which the knots in the framework are cyclic trione and the linkers are phenylene decorated with various functional groups including −F, −OH, and −SO₃H, respectively. Compared to the pristine COF (TpPa), these derivatives allow for the modulation of hydration states under saturated conditions through various pore-wall properties. All the COFs’ pores feature the water accessible surfaces (Figure S1) and are confined in a micropore region (<2 nm), ensuring a high fraction of interfacial water under fully hydrated conditions defined in the Methods section. In this study, the six-layer COF models were constructed (Figure b) and equilibrated in water to obtain the corresponding water-saturated states, which served as the reference configurations for subsequent analyses of confined water organization. These equilibrium hydration landscapes reflect the intrinsic interfacial water organization modulated by the COF pore chemistry prior to CO₂ transport.

1.

(a) Chemical structures of the building blocks (Tp and functionalized Pa-X linkers) and the corresponding top-view models of the AA-stacked TpPa-X COFs (X = H, OH, F, SO₃H). (b) Schematic representation of the NEMD simulation system for CO₂/H₂O permeation. The simulation box (approx. 39 × 45 × 90 ų) comprises a six-layer COF positioned between the feed zone and the permeate zone.

Figure a–d display the 2D in-plane probability density maps of the hydration landscape. For the pristine TpPa, water molecules are arranged into a highly ordered double-layer structure within the ∼1.8 nm pore. The first hydration layer is a continuous hexagonal ring along the pore wall (Figure a), followed by a second, less dense but still ordered ring toward the interior. This hydration character determines the available transport space and induces the surface resistance encountered by passing guest molecules. The spatial modulation of confined water was further quantified by axial density profiles (ρ) and the average number of hydrogen bonds per molecule (NHB) resolved along the transport direction (Figure e–h). Note that the density was calculated using the total volume of each spatial bin, including the volume occupied by the COF framework. Therefore, the reported values represent apparent local densities and are primarily used to compare the relative spatial distribution of confined water among different channels. Outside the COF layers, water exhibits bulk-like behavior (ρ ≈ 1.0 g/cm³; NHB ≈ 3.5). Near the pore entrance, packing effects induce a slight density increase (ρ ≈ 1.1 g/cm³), followed by a sharp decrease inside the channel (ρ ≈ 0.5 g/cm³), where the structural volume of the COF layers significantly reduces the space available for water occupancy. Within the pore interior, the density oscillates periodically, with maxima aligned to the COF layers and minima located in the interlayer channel regions.

2.

(a–d) 2D XY probability density maps of water molecules, averaged along the Z-axis, for (a) TpPa, (b) TpPa–SO₃H, (c) TpPa–OH, and (d) TpPa–F. The color bar indicates water probability density (darker blue = higher occupancy). (e–h) MD results of water density profile (left Y-axis) and average number of hydrogen bonds per water molecule (right Y-axis) along the Z-axis for (e) TpPa, (f) TpPa–SO₃H, (g) TpPa–OH, and (h) TpPa–F. The COF layer plane is parallel to the XY plane. The vertical dashed lines denote the positions of the COF layers.

Interestingly, the hydrogen-bond coordination does not strictly follow the density profile. Instead, NHB reflects the local organization of the hydration network. Regions with reduced water density may still maintain a high hydrogen-bond coordination when water molecules assemble into locally ordered clusters. The interior water adjacent to the framework layers maintains a nearly bulk-like hydrogen-bond coordination (∼3.5), whereas it shows slightly reduced coordination (∼3.4) in interlayer regions along the channel. This correlation indicates that the periodic potential field of the stacked COF layers exerts a template-induced effect on the confined hydration network, endowing it with spatially modulated structural ordering consistent with early reports. ,

In contrast, chemical functionalization disrupts this smooth periodic profile and introduces pronounced heterogeneity in both radial layering and angular water distribution. As shown in Figure b and f, TpPa–SO₃H features the strong polarity and steric bulk of sulfonic groups, fragmenting the continuous annular layer into localized high-density hotspots, particularly near carbonyl sites. Although the interlayer water density increases (ρ ≈ 0.6–0.8 g/cm³), NHB becomes highly heterogeneous (2.8–3.7). The pronounced spatial oscillations indicate the formation of strongly localized hydration domains anchored by the sulfonic acid groups. The −SO₃H functionality does not destabilize the hydration network, but reorganizes water into a highly structured but spatially heterogeneous configuration. For TpPa–OH with the reduced polarity of side groups (Figure c and g), directional hydrogen bonding between −OH groups and water increases the first hydration layer thickness and raises the sector-like density along hydroxyl and carbonyl motifs. Steric constraints elevate interlayer density (∼0.7 g/cm³) while slightly reducing density parallel to the framework planes (∼0.4 g/cm³). Despite this redistribution, the average NHB increases to 3.4–3.6. This combination of localized network disruption and robust interfacial hydrogen-bonding states points toward an interface-stabilized fragmented network topology. Differing from TpPa–SO₃H and TpPa–OH, the fluorinated TpPa–F exhibits hydrophobic exclusion near the pore wall (Figure d), resulting in low-density depletion zones and petal-like water distributions. Crucially, inside the pore channel (Figure h), water molecules exhibit the highest hydrogen-bond coordination (NHB = 3.6–3.8) across all frameworks. This observation indicates that water molecules preferentially aggregate into isolated, highly saturated clusters rather than with the pore surface, thereby establishing a heterogeneous clustered water network topology.

These results demonstrate that pore-wall functionalization reorganizes the spatial distribution and hydrogen-bonding connectivity of confined water, producing distinct hydration topologies ranging from continuous interfacial hydration structures (TpPa) to strongly localized hydration domains (TpPa–SO₃H), interface-stabilized reorganized hydrogen-bond networks (TpPa–OH) and heterogeneous clustered water networks (TpPa–F). Such chemically encoded hydration topologies establish the physical foundation for the diverging gas transport behaviors discussed hereafter.

3.2. Nonequilibrium Transport Kinetics of CO₂ and H₂O in the Fully Hydrated Nanochannels

To evaluate the molecular transport performance under operating conditions, we performed nonequilibrium molecular dynamics (NEMD) simulations on fully hydrated TpPa-based COF systems containing both CO₂ and H₂O (Figure b). This setup enables direct observation of CO₂ migration through the pre-organized hydration environments. We first examined the time-dependent competitive adsorption occupancy and molecular residence times (τ) within the nanochannels to track the continuous accumulation of guest species (Figure S3). Across all four frameworks, CO₂ occupancy remains minimal during the first 1 ns and subsequently increases with simulation time, indicating an induction period before CO₂ molecules can penetrate the water-filled pore entrances. After reaching dynamic equilibrium (approximately 4 ns), the pristine TpPa exhibits the highest CO₂ occupancy, fluctuating between 15.0% and 19.4% during the equilibrium stage, whereas TpPa–SO₃H maintains a lower but stable occupancy of 9–11%. In contrast, the TpPa–F and TpPa–OH systems show consistently lower CO₂ occupancies, remaining within 5.3–6.9% and 5.2–6.5%, respectively. The residence-time analysis further reveals that CO₂ remains substantially longer inside the pristine TpPa (τ = 1.23 ns) and TpPa–SO₃H (τ = 1.99 ns) than within TpPa–F (τ = 0.74 ns) and TpPa–OH (τ = 0.69 ns), indicating distinct CO₂ retention behaviors among different hydrated nanochannels.

To assess the relationship between molecular loading and transport behavior, the self-diffusion coefficients (D) were calculated from the linear region of the MSD curves using the Einstein relation. Specifically, we calculated three distinct diffusion coefficients for each guest molecule to resolve the transport anisotropy: the overall 3D diffusion (D _3D ), lateral diffusion within the XY-plane (D _XY ), and axial transport along the Z-channel (D _Z ). This multidimensional metric allows us to precisely distinguish between global migration and the specific microscopic mobility constraints imposed by the confined water network. The comprehensive data on the calculated diffusion coefficients are presented in Figure S2. Here, the separation factor of CO₂ and H₂O is defined and calculated as the ratio of their respective diffusion coefficients:

$$f=D_{C{O}_2}/D_{H_{2}O}$$ 1

Accordingly, f _3D , f _XY , and f _Z represent the overall, lateral, and axial diffusion selectivity, respectively, where a value greater than 1.0 indicates a regime where CO₂ diffuses faster than H₂O. As shown in Figure , the diffusion selectivity demonstrates a remarkable dependence on both the chemical functionalization and the spatial dimensions.

3.

Comparative analysis of CO₂ and H₂O transport kinetics in hydrated TpPa-based COFs. The bar chart presents the calculated diffusion separation factor (f) of CO₂/H₂O, calculated as the ratio of their respective diffusion coefficients.

For both CO₂ and H₂O, the overall diffusion coefficients of pristine TpPa exhibit substantially higher mobility than the −SO₃H modified systems. The overall diffusion coefficient $D_{C{O}_2,3D}$ decreases from 3.00 × 10^–9 m ²/s to 1.66 × 10^–9 m ²/s, and $D_{H_{2}O,3D}$ decreases from 6.32 × 10^–9 m ²/s to 3.72 × 10^–9 m ²/s. In addition, within both individual frameworks, H₂O diffuses approximately twice as fast as CO₂, corresponding to similar separation factors (f _3D = 0.48 for TpPa and 0.45 for TpPa–SO₃H). This higher mobility of H₂O over CO₂ can be attributed to its smaller molecular kinetic diameter (∼2.65 Å vs. ∼3.30 Å) and its ability to propagate cooperatively through the self-sustained hydrogen-bonding networks within the fully hydrated pores. In contrast, the larger, quadrupolar CO₂ molecules experience stronger confinement and interfacial friction during transport. Further decomposition into spatial components reveals that the lateral diffusion coefficients (D _XY ) of CO₂ and H₂O are similar in TpPa and TpPa-SO₃H ( $D_{C{O}_2,\mathrm{XY}}=0.73\times {10}^{-9}{m}^2/s$ vs. 0.63 × 10^–9 m ²/s and $D_{H_{2}O,XY}=2.38\times {10}^{-9}{m}^2/s$ vs. 2.05 × 10^–9 m ²/s). In contrast, the axial diffusion coefficients (D _Z ) in TpPa are approximately twice those in TpPa–SO₃H ( $D_{C{O}_2,\mathrm{Z}}=7.55\times {10}^{-9}{m}^2/s$ vs. 3.72 × 10^–9 m ²/s and $D_{H_{2}O,Z}=14.20\times {10}^{-9}{m}^2/s$ vs. 7.20 × 10^–9 m ²/s). These results indicate that the observed differences in overall mobility originate primarily from transport along the channel direction. In TpPa–SO₃H channels, introduction of strongly polar sulfonic acid groups provides a highly restrictive environment, which suppresses axial migration of both CO₂ and H₂O. Also, this retardation affects the two species to a similar extent. The relative diffusion selectivity remains unchanged (f _Z = 0.53 for both).

For TpPa–F, the hydrophobic fluorinated environment effectively reduces the transport resistance of CO₂, leading to a remarkable increase in axial diffusion ( $D_{C{O}_2,\mathrm{Z}}=15.96\times {10}^{-9}{m}^2/s$ ). As a consequence, the diffusion selectivity is inverted from H₂O-favored to CO₂-favored fashion, with the axial selectivity f _Z reaching 1.49. A less pronounced effect is observed for TpPa–OH, where modification of the local confinement environment facilitates CO₂ migration along the channel direction ( $D_{C{O}_2,\mathrm{Z}}=12.70\times {10}^{-9}{m}^2/s$ ), resulting in an enhanced axial selectivity of 1.30. Nevertheless, this enhancement is not exclusively confined to the axial direction. For H₂O, the lateral diffusion coefficients remain remarkably similar across all frameworks $(D_{{\mathrm{H}}_2\mathrm{O},\mathrm{XY}}=2.05-2.38\times {10}^{-9}{m}^2/s)$ . In contrast, CO₂ exhibits a substantial increase in lateral diffusivity upon fluorine and hydroxyl functionalization, with $D_{C{O}_2,\mathrm{XY}}$ increasing from 0.73 × 10^–9 m²/s in pristine TpPa to 1.36 × 10^–9 and 1.16 × 10^–9 m²/s in TpPa–F and TpPa–OH, respectively. This nearly twofold enhancement suggests that these functional groups facilitate not only axial migration but also local translational motion of CO₂ within the confined pore environment. Therefore, the kinetic selectivity inversion observed in TpPa–F and TpPa–OH arises from a multidimensional enhancement of CO₂ mobility, although the dominant contribution still originates from the dramatic increase in axial transport.

3.3. Microscopic Distribution and Cooperative Transport of CO₂ and H₂O

The microscopic basis of these transport trends was further investigated by 2D CO₂ probability density maps. In pristine TpPa, CO₂ molecules localize predominantly on the carbonyl-rich pore walls (Figure a). These highly concentrated suggests that the framework’s native CO nodes serve as anchoring templates, effectively localizing CO₂ molecules at adsorption sites. Such interfacial confinement forces CO₂ to migrate along the pore walls rather than undergoing unrestricted diffusion through the channel center. In TpPa–SO₃H (Figure b), carbonyl-associated localization persists, though partial penetration toward the pore center is observed. By contrast, TpPa–OH and TpPa–F (Figure c, d) promote a more central distribution of CO₂, with density spreading across the pore cross-section. This transition from wall-associated localization to a more delocalized pore-wide distribution substantially increases the accessible transport volume available to CO₂ and reduces its dependence on specific adsorption sites.

4.

(a–d) 2D XY-plane distribution of CO₂ molecules, averaged along the Z-axis within the COF region, for the (a) TpPa, (b) TpPa–SO₃H, (c) TpPa–OH, and (d) TpPa–F. The color bar denotes CO₂ probability density. (e–h) 1D normalized density profiles of CO₂ (orange) and H₂O (blue) along the transport direction (Z-axis), together with the corresponding co-localization function C­(z) (green dashed line), for the (e) TpPa, (f) TpPa–SO₃H, (g) TpPa–OH, and (h) TpPa–F. The reported overlap value in each panel quantifies the net degree of spatial co-localization between CO₂ and H₂O.

To quantify the spatial relationship between CO₂ and H₂O along the transport axis (z-axis), we introduce a co-localization function. The indicator is defined as

$$C(z)=\frac{{\rho }_{C{O}_2}(z)·{\rho }_{H_{2}O}(z)}{{\rho }_{C{O}_2,\max }·{\rho }_{H_{2}O,\max }}$$ 2

which measures the normalized overlap between the two species. C­(z) approaches unity when both densities simultaneously peak (cohabitation) and approaches zero under strong spatial exclusion. Integration along the channel yields a global overlap metric.

Both H₂O and CO₂ display five density maxima aligned with interlayer regions (Figure e–h), identifying these voids as primary exchange zones. However, their relative intensities differ markedly. TpPa exhibits the lowest overlap (0.570), reflecting strong spatial competition: CO₂ displaces interfacial water and migrates along carbonyl-associated adsorption sites (Figure e). Transport therefore requires frequent rearrangement of the surrounding hydration environment, resulting in stronger competition between CO₂ migration and water occupancy.

CO₂–H₂O spatial overlap for TpPa–OH (0.840) and TpPa–F (0.892) exhibits near-cooperative coexistence (Figure g, h), while TpPa–SO₃H remains intermediate (0.729) (Figure f). This increasing overlap value reflects a fundamental change in the transport environment. In the pristine TpPa, CO₂ migration is largely confined to framework adsorption sites and therefore competes directly with the surrounding hydration structure. In contrast, functionalization promotes spatial coexistence between CO₂ and water, enabling diffusion through a less restrictive hydrated nanochannel.

3.4. The Mechanism of CO₂ Transport Dynamics

To gain a deeper molecular-level understanding of the observed dual-component transport, we initially evaluated the surface electronic profiles of the four COF systems via Electrostatic Surface Potential (ESP) analysis (Figure a). According to the calculated color scale, the negative electrostatic potential is intensely concentrated over the CO oxygens in TpPa, a local environment further amplified by both −OH and −SO₃H functionalization, but attenuated by the electron-withdrawing −F substituents.

5.

(a) Top-view ESP maps of the structural units for pristine TpPa and functionalized derivatives (TpPa–SO₃H, TpPa–OH, and TpPa–F). The color scale (in a.u.) indicates the surface potential, where red and blue regions represent positive and negative potentials, respectively, highlighting the enhanced localized polarity induced by functional groups. (b) CO₂ binding energies (E_b ) at pore entrance and interior pore wall adsorption site. Representative adsorption configurations of CO₂ pore entrance (c) and interior pore wall (d) adsorption site, with the corresponding distance between O­(CO) and C­(CO₂) annotated as d_C–O .

To quantify how these tailored electrostatic fields direct the CO₂ capture and transport pathways, density functional theory (DFT) binding energies (E_b ) were evaluated across representative structural coordinates from the pore entrance to the channel interior. To visually isolate the thermodynamic driving forces and the resulting localized trapping effects, Figure b highlights the profiles at the initial pore entrance (Entrance, corresponding to Pos 1 in Figure S4) and the deeply confined adsorption framework nodes (Pore wall, corresponding to Pos 3 in Figure S4), while the fully resolved three-site continuous energetic landscapes (including the transition site, Pos 2) are preserved in Figure S4.

As shown in Figure b, the pristine TpPa exhibits the weakest binding affinity (E _b = −0.092 eV) at the pore entrance although a close contact distance (d_C–O = 2.82 Å), reflecting a limited capacity to anchor CO₂ molecules during initial capture. In contrast, functionalization significantly shifts the entry energetics. TpPa–OH, TpPa–F and TpPa–SO₃H provide stronger initial binding at the entrance (−0.130 eV, −0.140 eV and −0.144 eV, respectively). These enhanced surface interactions in all functionalized COFs provide the necessary energetic push for CO₂ to intercalate into the nano-confined hydration layer, ensuring a higher steady-state concentration within the channel and overcoming the initial resistance of the water film.

As the molecule moves toward the interior, the energy landscape further diverges. The pristine TpPa gives the strongest localized binding energy at the carbonyl-rich framework nodes (E _b = −0.212 eV, d_C–O = 2.78 Å), yielding the largest static energy drop (ΔE = 0.120 eV). This pronounced energy drop suggests strong localization of CO₂ at pore-walls adsorption sites. While CO₂ is difficult to capture at the entrance, it becomes strongly trapped once it reaches the pore wall. For TpPa–SO₃H, functionalization enhances the entry affinity, yet it maintains an excessive interior binding of −0.204 eV, resulting in a static energy drop of 0.060 eV. TpPa–F displays a uniquely shallow energy gradient, with E _b evolving minimally from −0.140 to −0.177 eV (a total drop of only 0.037 eV). This nearly flat energetic plateau explains its high microscopic mobility due to minimal guest-host friction. Meanwhile, TpPa–OH presents an intermediate stepped energy profile, where the binding strength remains nearly constant between the entrance and transition sites (−0.130 eV to −0.132 eV), followed by a moderate descent to −0.180 eV within the channel interior, yielding a well-balanced energy difference (ΔE = 0.050 eV). Although the relative potential fluctuations (ΔE) of TpPa–OH and TpPa–SO₃H are closely matched (0.060 eV vs. 0.050 eV), they diverge significantly in their absolute interior energy levels. TpPa–SO₃H possesses a much deeper absolute interior well that approaches the intense trapping baseline of pristine TpPa. These results demonstrate that efficient transport requires a dynamic synergy between molecular capture at the pore entrance and molecular release within the channel interior, rather than maximizing adsorption strength alone.

While the DFT calculations reveal how pore chemistry modifies the static adsorption energetics of CO₂, these local interactions provide the underlying energetic preference for CO₂ approaching and interacting with different pore-wall environments. However, under hydrated conditions, the effective transport landscape is additionally determined by solvent-mediated interactions, molecular rearrangement, and entropic contributions. To capture these dynamic effects, we further analyzed the orientational statistics and free-energy landscapes of CO₂ within the functionalized channels. As displayed in Figure a–d, the 2D orientation–position maps reveal that CO₂ preferentially adopts near-perpendicular configurations (θ ≈ 80–90° relative to the transport axis) in all systems. Upon entering the COF region, this orientational distribution becomes spatially modulated, forming periodic high-probability bands that coincide with interlayer spacing. Such modulation indicates that rotational sampling is coupled to the layered confinement geometry. Functionalization alters the degree of orientational confinement. TpPa–SO₃H affords sharply localized orientational hotspots, reflecting strong positional stabilization, whereas TpPa–F and TpPa–OH allow for broader angular distributions, indicative of enhanced rotational adaptability during axial migration. The pristine TpPa shows intermediate behavior.

6.

(a–d) 2D probability density maps of CO₂ molecular orientation (θ, deg) relative to the Z-axis as a function of axial position. (e–h) Potential of mean force (PMF, kcal/mol) profiles for CO₂ permeation along the transport direction (Z-axis). The shaded gray areas represent the COF region, with Δ _ent and Δ _lay values indicating the free energy barriers for pore entry and interlayer transport, respectively (Δ _lay shows the mean value for six layer). (i–l) Calculated orientational order parameter $(S=⟨\frac{3{\cos }^2\theta -1}{2}⟩)$ ) profiles of CO₂ molecules within the four TpPa-based COFs: (i) pristine TpPa, (j) TpPa–SO₃H, (k) TpPa–F, and (l) TpPa–OH. Schematic illustrations on the right define the physical significance of S: S = 1 corresponds to alignment parallel to the Z-axis, while S = −0.5 indicates a perpendicular orientation (θ = 90°).

To quantify this modulation, the orientational order parameter (S) was evaluated along the transport coordinate (Figure i–l). S is defined as

$$S=⟨\frac{3co{s}^2\theta -1}{2}⟩$$ 3

where θ is the angle between the molecular axis (or dipole vector for water) and the transport direction (Z-axis), averaged over molecules and time within spatial bins along the pore. The interpretation of the orientational order parameter S values is summarized in Table S1. In the bulk regions, CO₂ remains isotropic (S ≈ 0), whereas distinct spatial patterns emerge inside the nanopores. TpPa exhibits alternating positive and negative S values across layer and interlayer regions, indicating periodic reorientation as molecules traverse the stacked framework. TpPa–SO₃H shows pronounced negative S in the layered regions (S ≈ −0.4), consistent with constrained in-plane alignment. TpPa–F maintains predominantly negative S throughout the pore (S ≈ −0.3), suggesting persistent planar configurations. In contrast, TpPa–OH sustains positive S values (S ≈ 0.1) across the channel interior, reflecting partial axial alignment along the transport direction. Notably, TpPa–OH also exhibits reduced spatial oscillation in S, indicating fewer rotational switching events during migration. The orientational statistics indicate that pore chemistry influences not only where CO₂ resides but also how it navigates through the confined environment.

The preservation of COF-dependent hydration characteristics during CO₂ transport was further verified by analyzing water orientational distributions and orientational order parameters (Figure S5). Among all systems, TpPa–SO₃H displays the strongest orientational structure, where strong hydrogen bonding anchors water into layered in-plane configurations. In contrast, TpPa gives a comparatively smooth orientational profile, indicating a less constrained hydration environment. Together with the hydrogen-bond network analysis, these results suggest that pore chemistry determines the organization of confined water, which subsequently influences the local environment experienced by diffusing CO₂ molecules.

To quantify the energetic consequences of these structural differences, the Potential of Mean Force (PMF) was reconstructed based on the density distributions along the transport coordinate z:

$$PMF(z)=-k_{\mathrm{B}}T·\ln \left(\frac{\rho (z)}{{\rho }_0}\right)$$ 4

where ρ(z) is the density profile of the diffusing species along the transport coordinate z, ρ ₀ is the distribution density of molecules in the reference bulk region, k _B is the Boltzmann constant, and T is the temperature (300 K).

The PMF profiles reveal that transport is controlled by two distinct energetic processes: entry into the hydrated pore (Δ _ent ) and migration across successive interlayer regions (Δ _lay ). The water PMF profiles (Figure S5e–h) disclose substantial differences in hydration topology among the functionalized channels. The pristine TpPa presents the flattest water PMF profile with a smooth free energy landscape ((Δ _lay ∼ 0.08 kcal/mol), consistent with its high water diffusivity $(D_{H_{2}O,Z}=14.20\times {10}^{-9}{m}^2/s)$ . In contrast, TpPa–SO₃H confers the largest barrier (Δ _lay ∼ 0.552 kcal/mol), reflecting the formation of a strongly hydrogen-bonding network within the pore, as the sulfonic groups anchor water molecules into structurally constrained, localized water networks. TpPa–OH and TpPa–F display moderate energetic landscapes (Δ _lay ∼ 0.197 and 0.248 kcal/mol), suggesting that while these pores are fully hydrated, the water network is sufficiently fragmented to allow for guest molecule intercalation while offering greater configurational flexibility.

A similar energetic behavior is captured in the CO₂ PMF profiles (Figure e–h), with transport being dictated by two distinct energetic barriers. Specifically, the pristine TpPa gives a moderate entry barrier (Δ _ent ∼ 0.528 kcal/mol) but the highest periodic internal barriers (Δ _lay ∼ 0.927 kcal/mol). This behavior is consistent with the strong localization of CO₂ near the carbonyl-rich framework sites identified by both the density maps and DFT calculations, requiring molecules to repeatedly overcome localized energetic constraints during migration.

Pore functionalization via −SO₃H groups introduces an additional transport penalty. Although the sulfonic acid groups enhance the initial adsorption affinity, they simultaneously generate the highest entry barrier (Δ _ent ∼ 0.603 kcal/mol) and maintain substantial interlayer resistance (Δ _lay ∼ 0.412 kcal/mol). Together with the deep adsorption traps (E _b = −0.204 eV), these features promote molecular retention but significantly suppress mobility. In contrast, TpPa–F and TpPa–OH both substantially alleviate the interlayer migration barrier (Δ _lay ∼ 0.259 kcal/mol), offering a low-friction pathway inside the pore. This indicates that the framework-induced localization is weakened to facilitate transport through the hydrated interlayer regions. Meanwhile, they achieve the enhanced mobility via entirely distinct microscopic pathways.

As visually conceptualized in the multilevel cascade of Scheme , the PMF trends mirror the diffusion behavior obtained from molecular dynamics simulations. Specifically, the pronounced entry and migration barriers encountered in TpPa–SO₃H correspond to its lowest CO₂ diffusion coefficient, whereas the reduced barriers in TpPa–F and TpPa–OH benefit their enhanced CO₂ mobility. To decode the underlying structural origins, we integrate these PMF profiles with insights from DFT calculations, spatial density distributions, and orientational statistics to establish a transport mechanism. Pore functionalization constructs the unique hydration topology within the microporous channels, which subsequently regulates the spatial distribution, orientational freedom, and free-energy landscape for CO₂ molecules. As a result, different functional groups promote distinct transport strategies. Pristine TpPa favors molecular accumulation but suffers from strong localization during migration. TpPa–SO₃H further amplifies this retention-dominated behavior through deep adsorption traps and strong confined hydration structures. TpPa–F represents a mobility-dominated regime, where reduced across layer transport resistance enables the fastest CO₂ diffusion along the low-friction pathway between the clustered water and pore wall. Meanwhile, TpPa–OH offers an intermediate regime that combines facile pore access with low migration barriers in a reorganized hydration network. Therefore, CO₂ transport originates not from adsorption strength alone, but from the balance between molecular retention and molecular mobility established by the hydration topology of the microporous channels.

1. Schematic Illustration of the CO₂ Transport Mechanism Within the Hydrated Microporous COFs.

Although direct visualization of molecular-scale hydration topology within confined COF channels remains challenging, the proposed mechanism provides experimentally accessible signatures for validation. The differences in confined water organization induced by functional groups could be investigated using water-sensitive spectroscopic techniques, such as quasi-elastic neutron scattering , and solid-state NMR, which can probe water mobility and interfacial hydrogen-bonding environments. Meanwhile, humidity-controlled CO₂ permeation experiments on functionalized COFs could evaluate the predicted transport regulation under hydrated conditions by comparing CO₂ permeability and CO₂/H₂O selectivity. These measurements would provide experimental insights into the relationship between pore chemistry, hydration structure, and molecular transport proposed in this work.

4. Conclusion

In summary, we provide a comprehensive molecular-level understanding of CO₂ transport mechanisms within fully hydrated β-ketoenamine-linked microporous COFs. Our findings demonstrate that pore wall chemistry dictates gas transport kinetics not via single-component adsorption affinity alone, but intrinsically through a sophisticated three-phase (COF–H₂O–CO₂) interfacial interaction that reshapes the confined hydration topology. Without modification, the pristine COF induces continuous and spatially organized hydration structure in the 1D micropores, forming a highly periodic transport resistance. The introduction of polar −SO₃H strengthens water anchoring, producing the highest entry and interlayer penalties and thereby suppressing CO₂ diffusion. While incorporating −OH groups with weakened hydrophilicity, the functionalized COF–OH reorganizes the confined hydration structure for intercalated CO₂ shuttling and minimizes the entry barrier to improve CO₂ loading. For the hydrophobic functionalization, the microporous channels of COF–F facilitate a low friction mobility of CO₂ molecules. The continuous water network is disrupted into water clusters at the liquid-solid interface and the kinetic preference is reversed from H₂O-favored to CO₂-favored transport, yielding a nearly 3-fold increase along the axial direction. Therefore, instead of maximizing static binding strength between COF and CO₂, the three-phase interaction at the microporous interface dominates the free-energy landscape along the entire transport coordinate. Our work provides crucial molecular-level guidelines for designing high-performance microporous materials in CO₂ transport and conversion under realistic aqueous conditions.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacsau.6c01003.

• Calculation methods for diffusion, adsorption energy, and pore surface analysis; control simulations validating the effects of framework flexibility and driving forces on confined transport; additional analyses of molecular diffusion, adsorption configurations, free energy profiles, and transport behaviors under different conditions; as well as an interpretation of the water orientational order parameter (PDF)

The authors declare no competing financial interest.