Adsorption in Pyrene-Based Metal–Organic Frameworks: The Role of Pore Structure and Topology

Abstract

Pore topology and chemistry play crucial roles in the adsorption characteristics of metal–organic frameworks (MOFs). To deepen our understanding of the interactions between MOFs and CO₂ during this process, we systematically investigate the adsorption properties of a group of pyrene-based MOFs. These MOFs feature Zn(II) as the metal ion and employ a pyrene-based ligand, specifically 1,3,6,8-tetrakis(p-benzoic acid)pyrene (TBAPy). Including different additional ligands leads to frameworks with distinctive structural and chemical features. By comparing these structures, we could isolate the role that pore size, the presence of open-metal sites (OMS), metal–oxygen bridges, and framework charges play in the CO₂ adsorption of these MOFs. Frameworks with constricted pore structures display a phenomenon known as the confinement effect, fostering stronger MOF–CO₂ interactions and higher uptakes at low pressures. In contrast, entropic effects dominate at elevated pressures, and the MOF’s pore volume becomes the driving factor. Through analysis of the CO₂ uptakes of the benchmark materials —some with narrower pores and others with larger pore volumes—it becomes evident that structures with narrower pores and high binding energies excel at low pressures. In contrast, those with larger volumes perform better at elevated pressures. Moreover, this research highlights that open-metal sites and inherent charges within the frameworks of ionic MOFs stand out as CO₂-philic characteristics.

1. Introduction

Pyrene-based metal–organic frameworks (MOFs) have been increasingly used in the fields of luminescence sensing,¹⁻³ photocatalysis,⁴⁻⁷ electrochemistry,^8,9 as well as biomedical applications.¹⁰ Moreover, in recent years, the interest in pyrene-based MOFs for gas adsorption and separation has widely grown.¹¹ Various characteristics make pyrene-MOFs very attractive CO₂-adsorbents: they possess high water stability and retain permanent porosity.^12,13 Some structures may possess relatively large surface areas and pore volumes,¹¹ and the extensive π-aromatic system of the pyrene core can also serve as a preferential binding site for polarizable molecules, such as CO₂.¹⁴⁻¹⁷

Most studies on pyrene-based MOFs for carbon capture applications have concentrated on the synthesis and performance of specific structures.^12,13,16 Only a few publications systematically investigate the features that yield high CO₂ uptakes for CO₂/N₂ separation applications. The separation of these gases is mainly rooted in differences in their physical properties, primarily their kinetic diameter, polarizability, and quadrupole moment¹⁸ (Supplementary Note 1). Building upon the distinctive properties of these gases, previous studies have identified various chemical and topological characteristics of MOFs that enhance the preferential adsorption of CO₂.^14,19,20 Among these, metal–oxygen bridges are frequently encountered in high-performing materials.¹⁴ Given its polarizability and pronounced quadrupole moment, these bridges foster a favorable environment for the adsorption of CO₂, bringing it closer through induced dipole–quadrupole interactions.¹⁹ Another effective characteristic of MOFs for CO₂ capture is the presence of open-metal sites (OMS).¹⁴ Unsaturated metal sites favor the adsorption of CO₂.^21,22 Here, electrostatic interactions between the unsaturated metal centers and the large quadrupole moment of CO₂ enhance gas adsorption. Moreover, ionic MOFs are also a promising class of materials. MOFs with a net charge possess unique charge distributions due to the inherent presence of cations or anions within their structure. This charge localization greatly enhances the electrostatic interactions with CO₂ molecules, which have significant polarizability.²³ MOFs with narrower pores can create a confinement effect for CO₂ molecules, placing them in closer contact with pore surfaces and intensifying interaction potentials. For mixed gas adsorption, separation efficiency is driven by the difference in interaction potential for each gas type with the pore surface.^20,24

In the context of CO₂ capture, it is essential to carry out a systematic and comprehensive understanding of the interactions between CO₂ and the framework and explore the adsorption properties of both new and established MOFs. Our study provides a methodical investigation of the CO₂ adsorption properties of a group of MOFs that are composed of Zn(II) as a metal ion and a pyrene-based ligand: 1,3,6,8-tetrakis(p-benzoic acid)pyrene (TBAPy).

From a geometric point of view, the synthesis of pyrene-based MOFs is nontrivial. The available coordination sites of the metal determine the orientation of the TBAPy ligands. Several TBAPy-MOFs are known to distort upon activation^25,26 due to the rotational freedom of the benzoate groups. The orientation of the ligand can affect the lattice parameters and, therefore, its structure, accessible surface area, and adsorption sites, as well as the performance of the material as an adsorbent. The complexity and potential variability in the framework’s actual structure make in-silico predictions of adsorption behaviors in TBAPy MOFs difficult. In this work, we explore, through a combined experimental and computational study, how we can better understand the impact of these structural changes on different applications of interest.

2. MOFs Studied in This Work

In this work, three different reported structures are characterized before and after the removal of trapped solvent molecules. In particular, Zn₂(TBAPy)(H₂O)₂²⁶ (which we will refer to as Zn₂-(TBAPy) hereafter), Zn_1.5O_0.25(Ade)(TBAPy)_0.5,²⁷ (Zn-(Ade)(TBAPy) in the following, with Ade = adenine), and (Zn₈(TBAPy)₃(6-BA)₄(μ₄-O)(H₂O)₄,²⁸ (Zn-(6BA)(TBAPy) in this study, with 6BA = 6-benzylaminopurine).

The unique pore geometries, chemistries, and structural features of this family of materials present an opportunity to characterize the various CO₂-philic characteristics. Upon activation, these MOFs show different structural arrangements. Such structural changes can significantly impact the properties of these materials.

Zn₂(TBAPy)

Zn₂-(TBAPy) reportedly has a 2D monoclinic structure, with Zn coordinated square-pyramidal in the paddle-wheel motif containing axially coordinated water molecules. Activation of this MOF leads to notable shifts and changes in the intensity of the main diffraction peaks at low angles, suggesting a structural phase transition.²⁶ In exploring this transformation upon desolvation, Stylianou et al. employed classical molecular dynamics (MD) simulations, imposing specific conditions such as removing water molecules and aligning cell parameters with those derived from experimental data. Through a methodical process, the structure was adjusted across different Zn ion coordination environments, ultimately aligning closely with experimental diffraction data (Supplementary Note 2). Our simulation results indicated that, following removing water molecules from the Zn-paddlewheel sites, the TBAPy ligand’s carboxylate groups shift to establish cross-linking bridges between layers. This shift facilitates the formation of 1D Zn–O chains along the c-axis (Figures 1a, b), which, coupled with TBAPy ligands, assembles into a 3D framework.²⁶

Figure 1.

Crystal structure representation of the MOFs along different crystallographic axes. (a, b) Distorted framework of Zn₂-(TBAPy); (c, d) Zn-(Ade)(TBPy). The counterion in the pores were hidden for simplicity; (e, f) OMS version of Zn-(6BA)(TBAPy), without counterions in the pores for clarity. Pore A is defined here as the acid-pore, and it is characterized by the carboxylate-oxygens from the TBAPy ligands that point in the channels. Pore B is the base pore and contains the hydrogen-bonding Watson–Crick face of adenine. Color code: C (gray), O (red), Zn (green), N (blue). The hydrogens (H) and carbons (C) were hidden for clarity.

Zn-(Ade)(TBAPy)

In Zn-(Ade)(TBAPy),²⁷ the metal node unit consists of six tetrahedrally coordinated Zn(II) ions that form octahedral cages with four adenine ligands. In this framework, the deprotonated adenine is coordinated in a tridentate fashion as a bridging ligand. Two chemically distinct channels are formed, which were classified as acid- and base-pore by Anderson et al.²⁷ The acid-pore (pore A in Figure 1c) is characterized by the carboxylate-oxygens from the TBAPy ligands that are pointing in the channels, while the base pore (pore B in Figure 1c) contains the hydrogen-bonding Watson–Crick face of adenine. The unit cell of Zn-(Ade)(TBAPy) contains dimethylammonium (DMA) counter-cations in the pores to balance the net atomic charge of −8 (Figures 1c, d).

Zn-(6BA)(TBAPy)

Zn-(6BA)(TBAPy),²⁸ contains the biorelated 6-benzylaminopurine (6BA) ligand along with TBAPy, both coordinated to Zn(II) ions. This MOF features two distinct types of secondary building units (SBUs): one is a Zn₄O(6-BA)₄(COO)₄ complex, where Zn1 and Zn2 ions connect with two 6BA ligands and a carboxylate oxygen, sharing a μ₄-O atom. This forms a complex SBU that contributes to the framework’s stability. The other SBU involves Zn3 and Zn4 in a distorted tetrahedral geometry, each coordinating with 6-BA, two carboxylate oxygens from TBAPy, and a water molecule, creating a simpler, mononuclear Zn(6-BA)(COO)₂(H₂O) unit. These SBUs are interconnected by 6BA ligands and TBAPy, constructing a 3D MOF. Each benzoate group binds in a monodentate fashion, leaving free carboxylate oxygen extending into the channels. Similarly to Zn-(Ade)(TBAPy), Zn-(6BA)(TBAPy) contains two different types of pores, one with an acidic character containing the free carboxylate-oxygens, while the free sides of the 6BA ligand point in the basic channel. The base pore features alternating narrow channels and cavities. The structure is, therefore, reported to function as a molecular sieve, allowing only small molecules to enter the pores.²⁸ The anionic MOF, with a framework charge of −4 per unit cell, incorporates DMA counterions to maintain charge neutrality and includes coordinated water molecules.

MOFs’ Stability and Morphology

The thermal stability of the structures is evaluated via thermal gravimetric analysis (TGA), which shows a two-step desolvation for Zn₂-(TBAPy) and Zn-(6BA)(TBAPy), indicating the removal of the metal-coordinated water molecules during the second step (Supplementary Note 3). In the case of Zn-(Ade)(TBAPy), solvent loss is observed up to ≈100 °C, followed by a gradual decrease in weight percentage thereafter. Scanning electron microscopy (SEM) images show that this family of materials presents distinct crystal shapes and sizes (Supplementary Note 4).

3. Results and Discussion

3.1. Crystal Structures

To ensure that our simulations closely reproduce real-world behavior, validating our computational models’ accuracy in replicating the crystal structures’ experimental properties is important. We accomplished this by comparing the diffraction patterns and pore volumes from our MOFs’ experimental and computational results. These characteristics are essential in adsorption studies for describing the material,^29,30 and can be computationally calculated from the crystallographic information file (CIF), reducing uncertainties tied to simulation parameters.³¹

The syntheses of the ligand, as well as the as-made materials, are reported elsewhere, and their syntheses are reproduced in this work (see Sections 6.1 and 6.2 for experimental details). Our simulations are based on ideal, defect-free crystals. Matching simulated XRD patterns and pore volumes with experimental results gives us insights into how well our computational representations align with actual experimental structures. These models, representing different states of the activated frameworks, are detailed in the following paragraphs.

Zn₂-(TBAPy)

To obtain a more accurate simulated model of this framework, we distinguish between two computational models: one with coordinated water molecules, which we refer to as simulated CW, and a second model with OMS, despicted as simulated OMS. The optimized structures of the 2D layered network with and without coordinated water molecules are compared to experimental PXRD data (Supplementary Note 5). The experimental PXRD pattern for as-made Zn₂-(TBAPy) matches well with our simulated models, both in the presence (simulated CW) and absence of coordinated water molecules (simulated OMS) ((Figure 2a). Upon activation, we observe a shift in the main diffraction peaks at low angles. Similarly, Stylianou et al.²⁶ showed that ligand geometry distortion, followed by pyrene core rotation and displacement, disrupts metal-paddlewheel motifs. This process leads to the formation of 1D Zn–O chains, which link adjacent layers, transitioning the structure from 2D to 3D. To model accurately this 3D distorted framework, we created a third structure model (which we refer to as simulated distorted) based on the structure reported in the original work.²⁶ Our simulated distorted model reflects the shift in the diffraction peaks observed in experiments of the activated MOF. We have performed a Le Bail fit, showing that the experimental PXRD fits relatively well with our simulated distorted model (Supplementary Note 6, Figure S5a). This agreement between the experimental and simulated patterns supports our understanding that the activation process alters the metal’s coordination environment, leading to significant structural transformations. Consequently, for a more precise comparison with experimental data on CO₂ adsorption, we use the simulated distorted unless stated otherwise. This approach confirms the change in the coordination environment around the metal centers postactivation, as depicted in Figures 2 a and S1.

Figure 2.

Powder X-ray diffraction (PXRD) patterns of the structures, including the simulated patterns from the respective CIFs: (a) Zn₂-(TBAPy). As the diffraction patterns of the simulated CW and simulated OMS models are very similar, only the simulated CW pattern is shown, (b) Zn-(Ade)(TBAPy) and (c) Zn-(6BA)(TBAPy).

Zn-(Ade)(TBAPy)

The original structure of the anionic MOF framework was derived in the cited study²⁷ by single-crystal X-ray diffraction (SCXRD). We use a simulated annealing approach as described in section 6.3.2 to locate the ions in the pores of the ionic framework and relax the atom positions. Our results agree well with the location of the counterions determined using X-ray photoelectron spectroscopy (XPS) experiments in the study of Anderson et al.²⁷ The as-made material agrees well with the simulated PXRD of the optimized framework with charge-balancing ions (Figure 2 b). The activated MOF shows weaker and broader Bragg peaks at the high 2θ (deg) range, which indicates a loss of order on this length scale. We have performed a Le Bail fit on the activated structure showing that the experimental diffraction pattern fits relatively well our simulated structure, and no peaks are missing in the low 2θ (deg) range (Supplementary Note 6, Figure S5b).

Zn-(6BA)(TBAPy)

To explore the crystal structure of Zn-(6BA)(TBAPy) postactivation, we developed two computational models highlighting different features of the activated MOF: one simulating the presence of coordinated water molecules (referred to as simulated CW), and the other showcasing OMS, indicative of a completely activated MOF (labeled as simulated OMS model). The experimental PXRD before activation seems to be in good agreement with the simulated CW model (Figure 2c). Upon activation, OMS are created, and we have performed a Le Bail fit on the activated structure and compared it to the simulated OMS model. While in the high 2θ (deg) range we observe weaker and broader Bragg Peaks, we were also able to get a relatively good fit for this structure (Supplementary Note 6, Figure S5c).

The differences observed in the PXRD patterns could mainly stem from structural distortions and the presence of solvent guest molecules in the framework, which could not be fully captured in the simulations. Nonetheless, the overall trends in unit cell parameters show good agreement with the computationally obtained DFT values (Table S2). The alignment of unit cell parameters and space groups supports our proposed structural models. While some discrepancies remain due to the aforementioned structural distortions and challenges in achieving full activation without structural collapse, these are relatively minor and should not significantly affect the overall analysis.

3.2. Pore Volumes

To further characterize these structures and assess their utility for carbon capture applications, the structural properties of the different MOFs are comprehensively evaluated. N₂ adsorption experiments at 77 K reveal modest pore volumes of 0.28 cm³/g for Zn₂-(TBAPy), and 0.29 cm³/g for Zn-(Ade)(TBAPy). Conversely, Zn-(6BA)(TBAPy) demonstrates almost no N₂ uptake, producing a near-zero pore volume (Supplementary Notes 7 and 8).

For Zn₂-(TBAPy), the slight deviation between computed pore volumes of the distorted structure (0.25 cm³/g) and experimental value (0.28 cm³/g), along with higher pore volumes in the 2D simulated CW and simulated OMS models (0.31 and 0.35 cm³/g, respectively), might suggest the coexistence of the 2D and 3D structures postactivation. This coexistence could stem from incomplete activation, where remaining coordinated water molecules at some metal nodes prevent the ligand rotations necessary for forming the fully distorted 3D framework. We attempted activation at 250 °C, considering the small weight loss observed in the TGA until 425–450 °C. However, while the structure activated at 110 °C maintains still a good crystallinity, the one activated at 250 °C, shows a less crystalline structure, which is evidenced by broader and less intense Bragg peaks in the higher 2θ (deg) range (Supplementary Note 9, Figure S7a). Given the relatively good match of the Le Bail fit and unit cell parameters, we focused our study on the material activated at 110 °C, as it aligns more closely with the simulated distorted PXRD pattern.

The pore volume simulations containing counterions in the acid pore of Zn-(Ade)(TBAPy) (0.27 cm³/g) closely align with the experimental results (0.29 cm³/g), with minor differences that might arise from the static nature of counterions in computational pore volume calculations. Zn-(6BA)(TBAPy)’s data presents a more intricate picture. The negligible experimental N₂ uptake contrasts sharply with the different computational models. Notably, while frameworks without counterions reveal similar pore volumes for both simulated OMS and simulated OMS models, the inclusion of ions results in significant alterations. As detailed in Section 6.3.2, introducing counterions into the framework models induces ligand reorientation and significant shifts in unit cell vectors of the simulated OMS model (∼10% reduction in a and c cell length). The consequence of these changes is a decrease of the pore limiting diameter from 4.62 to1.98 Å, i.e., below the kinetic diameter of N₂, 3.86 Å (Supplementary Note 8). This is important, as the pore-limiting diameter corresponds to the diameter of the smallest molecule that can diffuse through the structure. Hence, this observed decrease makes certain (or all) channels inaccessible for N₂. It is crucial to note that Zeo++, in the computation of the geometrical properties of the framework (as detailed in Section 6.3.1), treats the adsorbate as a spherical probe with a static framework so that actual accessibility may vary based on molecular reorientation or framework atom flexibility. While the experimental results match relatively well the simulated models with coordinated water molecules or OMS, the small discrepancy suggests the possibility of partial activation in the experiments. This partial activation could result in a framework representing an intermediate state, not fully transitioning to the OMS model, with only some of the coordinated water molecules removed. Based on the TGA results, we observe an additional weight loss step at around 250–300 °C, which may be attributed to the formation of open metal sites (OMS) in Zn-(6BA)(TBAPy). Therefore, we attempted to activate the material at 250 °C to reach a fully activated structure. However, the structure becomes more amorphous at 190 °C, with higher-order Bragg peaks significantly weakening (Supplementary Note 9, Figure S7b). We chose to study this material at 190 °C to examine the effect of partially present OMS without causing excessive structural collapse. This may indeed explain the differences observed in pore volume and CO₂ and N₂ isotherms at 25 °C, as OMS have a tremendous effect on CO₂ uptake.

3.3. CO₂ and N₂ Adsorption Isotherms

While examining the distinct experimental isotherms at 25 °C, Zn-(Ade)(TBAPy) records the highest CO₂ uptake of 1.75 mmol/g at 1 bar (Figure 3a). The structure showcases numerous CO₂-favorable characteristics, such as pronounced local charges owing to the framework’s anionic nature, Lewis bases present as free carboxylate oxygens in the ligands, and a narrow pore architecture amplifying interactions with adsorbates. Interestingly, while Zn-(6BA)(TBAPy) outperforms Zn₂-(TBAPy) at lower pressures (0.51 mmol/g for Zn-(6BA)(TBAPy) and 0.37 mmol/g for Zn₂-(TBAPy) at 0.2 bar), this trend reverses at higher pressures with 1.26 mmol/g for Zn-(6BA)(TBAPy) and 1.52 mmol/g for Zn₂-(TBAPy) at 1 bar. In the discussion of the structural properties, we show that Zn₂-(TBAPy) has a larger pore volume than Zn-(6BA)(TBAPy), while — as will be discussed in detail in the next section — Zn-(6BA)(TBAPy) exhibits markedly higher interaction energies with CO₂. The compact pores in Zn-(6BA)(TBAPy) facilitate closer proximity between CO₂ and the framework, enhancing the adsorption energy due to overlapping interactions from the framework’s multiple walls, which is evident from the visualizations of density maps and binding sites provided in Figures 4 and Supplementary Note 10. These enhanced binding energies primarily govern the system’s adsorption capacity at lower pressures. With increasing pressure, however, the adsorption dynamics shift toward being dominated by the framework’s pore volume and the entropy-driven distribution of CO₂ molecules within the available space.³² Consequently, Zn₂-(TBAPy), with its extensive pore volume, allows for a higher saturation of CO₂ molecules and a greater variety of molecular arrangements, leading to its enhanced adsorption at higher pressures once the primary adsorption sites are saturated.³³

Figure 3.

(a) CO₂ and (b) N₂ adsorption isotherms at 25 °C of the Zn-TBAPy structures: Zn₂-(TBAPy) (black), Zn-(Ade)(TBAPy) (red) and Zn-(6BA)(TBAPy) (blue). Experimental data (filled circles), simulated OMS (empty squares), simulated CW (empty triangles), simulated Distorted for Zn₂-(TBAPy) (crosses) or simulated for Zn-(Ade)(TBAPy) (empty circles).

Figure 4.

Crystal structure representation of all MOFs with the CO₂ center of mass distribution derived from classical simulations within each structure. (a, b) Zn₂-(TBAPy); (c, d) Zn-(Ade)(TBPy); (e, f) Zn-(6BA)(TBAPy). Color code: C (gray), O (red), Zn (green), N (blue), CO₂ (yellow). The hydrogens (H) were hidden for clarity.

Marked differences in the CO₂ uptakes can be observed for the different computational models of Zn₂-(TBAPy) and Zn-(6BA)(TBAPy) (Figure 3). The simulated CW and simulated OMS models of Zn₂-(TBAPy) feature wide channels with strong binding sites, consequently rendering the highest computed uptakes. Notably, the OMS model surpasses the structure with coordinated water molecules in uptake, underscoring the influence of open metal centers exposed to the pore surface on CO₂ adsorption. Conversely, the distorted structure model of Zn₂-(TBAPy), with its tighter channels and obscured metal center, predicts lower uptakes. This model aligns most closely with experimental results, confirming Stylianou et al.’s hypothesis that Zn₂-(TBAPy) transitions to a 3D framework upon activation.²¹ For Zn-(Ade)(TBAPy), simulated CO₂ uptakes closely match experimental results. The open-metal centers in the OMS model of Zn-(6BA)(TBAPy) preferentially bind to counterions, and this structure model has a significantly reduced pore volume (Supplementary Note 5). In contrast, the model having coordinated water molecules exhibits a notably higher uptake. The experimentally determined isotherm is positioned between the simulated isotherms for this structure’s CW and OMS models, indicating potentially incomplete activation in experiments (Figure 3a). The computational isotherms also reflect the trend wherein activated Zn-(6BA)(TBAPy)’s uptake surpasses the distorted model of Zn₂-(TBAPy)’s at lower pressures, but inverts at elevated pressures. However, the intersection point of the simulated isotherms occurs at greater pressures, suggesting the coexistence of different phases in the experiments.

From the N₂ adsorption isotherms (Figure 3b), a pronounced low N₂ uptake across all MOFs becomes apparent in simulations and experiments. Charged frameworks often have a much lower affinity for N₂ compared to CO₂.^34,35 This phenomenon is typically attributed to the difference in quadrupole moments and polarizability between CO₂ and N₂ (Supplementary Note 1), with N₂ having a very low affinity for charged framework sites due to its low polarizability and quadruple moment. Accordingly, experiments show a nearly negligible N₂ uptake for both Zn-(Ade)(TBAPy) and Zn-(6BA)(TBAPy), with values below 0.15 mmol/g. In the case of Zn-(6BA)(TBAPy), the experimental outcomes align with the OMS framework model’s simulations, although the CW model predicts a marginally increased uptake of around 0.35 mmol/g at 25 °C and 1 bar. Given the negligible measured pore volume for this framework (i.e., 0.04 cm³/g), the low N₂ uptake can also be attributed to a sieving effect, making it nonporous for N₂. For Zn-(Ade)(TBAPy), both simulations and experiments determine a considerably larger pore volume, and simulations predict a significantly higher N₂ uptake. An increased mobility of counterions at elevated temperatures might obstruct the pathways for N₂. For Zn₂-(TBAPy), the computed isotherm for the distorted 3D model again shows the best agreement with experimental data, confirming previous findings. Experimentally, an uptake of around 0.16 mmol/g is determined, while for the distorted 3D model we obtain approximately 0.37 mmol/g at 1 bar. However, uptakes in the 2D CW and OMS models soar beyond 0.8 mmol/g at 1 bar.

3.4. Density Maps

To gain insights into favored adsorption sites and the degree to which they are filled at low and high pressures, density maps illustrating the center of mass of CO₂ are derived from simulations at 0.2 and 1 bar at 25 °C. For Zn₂-(TBAPy) and Zn-(6BA)(TBAPy) (CW model), the density maps (Figures 4 a, b and e, f) shed light on the noticed variances in CO₂ uptake performance at different pressures. At reduced pressures, both Zn₂-(TBAPy) and Zn-(6BA)(TBAPy) maps reveal that regions with dense adsorbate concentration are primarily centered around binding sites, suggesting that single occupancy sites majorly influence low-pressure adsorption. CO₂ predominantly occupies the smaller elliptical channels running along the c-axis for Zn₂-(TBAPy). Zn-(6BA)(TBAPy)’s superior uptake at these pressures suggests it possesses stronger adsorption sites, as initial uptake often correlates with stronger binding energies. The maps reveal that Zn-(6BA)(TBAPy) has desirable adsorption locations with narrow pores; these charged sites likely enhance interactions with CO₂ via dipole induction. As pressure increases, total available pore volume starts playing a pivotal role. For Zn₂-(TBAPy), the adsorption at 1 bar now extends across both channels, but the adsorption in Zn-(6BA)(TBAPy) is still highly localized. With Zn₂-(TBAPy) offering a more substantial pore volume for CO₂ accommodation, its uptake surpasses Zn-(6BA)(TBAPy) despite the latter’s stronger interaction energies. This reveals that Zn-(6BA)(TBAPy)’s binding sites have deeper potential energy wells compared to Zn₂-(TBAPy), but they are also constricted, leading to swift saturation with increasing pressure. Contrarily, Zn₂-(TBAPy) provides multiple shallow-welled binding sites, enabling CO₂ to distribute across different pore varieties when pressure rises. The density maps illustrating CO₂ center of mass distribution in Zn-(Ade)(TBAPy) (Figure 4b) — the best-performing Zn-TBAPy MOF — highlight that adsorbed CO₂ molecules occupy both channels. At reduced pressures, the proximity to the framework becomes a favored site for adsorption due to the pronounced framework charges. Notably, CO₂ evades areas near the DMA ions within channel A, leading to an X-shaped distribution of the molecules. This could be attributed to physical blockage caused by the counterions or potential electrostatic repulsions from intense charge concentration.

3.5. CO₂ Binding Site Analysis

The minimum energy configurations of a CO₂ molecule across various Zn-TBAPy framework models are also analyzed (Supplementary Note 10). Interaction energies listed in this section are computed based on first-principles simulations. The distinct chemical characteristics within the pores of the Zn-TBAPy MOFs lead the CO₂ molecule to position itself in diverse CO₂-philic sites. These sites include unsaturated metal centers, exposed carboxylate-oxygens, regions of heightened charge in charged-framework-counterion systems, and pronounced aromatic systems offering pi-interaction in the narrow-pore frameworks.

Significant differences are identified in interaction energies across distinct structural models when examining Zn₂-(TBAPy). The 2D simulated CW model features CO₂ centrally placed in the channel surrounded by TBAPy ligands, which leads to moderate interaction energy of −25.7 kJ/mol. In contrast, the 2D simulated OMS model, with its heightened electron density around the metal node, showcases a preferred CO₂ adsorption site proximate to the unsaturated Zn ion (Supplementary Note 10) with an interaction energy of −47.2 kJ/mol. The 3D distorted framework model, characterized by tetrahedral coordination of the metal ions, lacks an exposed metal site. Instead, CO₂ settles at the heart of narrow channel B, where it interacts with the TBAPy ligands’ π-system with the interaction energies in this configuration amounting only to −21.8 kJ/mol. The increased interaction strength in the OMS structure stems predominantly from amplified electrostatic interactions (Supplementary Note 10). For the 2D simulated CW model, van der Waals interactions account for 86% of the total interaction energy, and this figure rises to 99% for the 3D distorted model. However, in the simulated OMS CIF, they represent just about half of the total interaction energy. This indicates that the pore chemistry of the OMS structure is distinctively tailored toward facilitating stronger electrostatic interactions with the adsorbate, differentiating it from the other models.

The minimum energy configuration of CO₂ in Zn-(Ade)(TBAPy) is located in the acidic pore, close to the free carboxylate (Supplementary Note 10). The prominent quadrupole moment of CO₂ fosters a pronounced lateral interaction with the oxygen atom of the carboxylate group on the linker. The interaction energy deduced from quantum mechanical simulations amounts to −33.4 kJ/mol. The heightened local charges within the negatively charged MOF play a pivotal role in achieving this elevated interaction energy through electrostatic interactions. Notably, dispersive forces account for just 65% of the total interaction energy.

In the context of Zn-(6BA)(TBAPy), a distinct evaluation emerges between the simulated CW and OMS structural models in terms of interaction energies with the adsorbate. Within the CW structure, CO₂ positions itself in the cage formed by the Zn-ions, maintaining a close affinity with the coordinated water and the free carboxylate-oxygen of the TBAPy ligand. Conversely, in the OMS model, the molecule is attracted to a compact pocket near the Zn-ion and DMA counterions (Supplementary Note 10). Remarkably, the interaction energies in both scenarios are very similar. Intriguingly, in the CW model, only approximately 50% of the total DFT-interaction energy stems from dispersive interactions. This hints at a substantial influence from electrostatic interactions with the framework’s inherent charges and counterions, an aspect the classical force field might struggle to capture effectively. Conversely, the OMS model, despite potentially offering even stronger electrostatic interactions due to its unsaturated metal site, sees a predominant influence from van der Waals forces. The significant role of dispersive interactions in this model underlines that the weaker dispersive interactions remain instrumental even in the face of potent coordination sites or electrostatic attractions. This likely arises from the CO₂ molecule’s close spatial alignment with the organic linker, attributed to the MOF’s constrained pore dimensions.

The ranking of the single molecule interaction energies for the Zn-TBAPy MOFs among each other depends on the different framework models. Considering the fully activated 2D simulated CW structure of Zn₂-(TBAPy), this MOF has the highest interaction energy, followed by Zn-(6BA)(TBAPy) and Zn-(Ade)(TBAPy). However, the 3D distorted model of Zn₂-(TBAPy) ranks lowest among the studied Zn-TBAPy MOFs by the ranking based on adsorption isotherms presented in the previous section.

4. Comparative Analysis of Pyrene and Other Ligands

When contrasting pyrene-based MOFs with structures containing other comparable ligands, several key insights emerge. Aromatic ligands, like pyrene, may show high uptakes in the low-pressure regime due to strong confinement effects and localized interactions, making it suitable for applications requiring high selectivity at low pressures.^11,36 As shown in this study, Zn₂-(TBAPy) exhibits moderate CO₂ uptake at the investigated pressure range due to its pore structure and strong binding sites. Porphyrin-based MOFs provide an intriguing comparison due to their similar molecular geometries but distinct interaction mechanisms. Like pyrene-based frameworks, porphyrin MOFs offer strong π–π interactions, enhancing CO₂ adsorption. They also often exhibit larger and more flexible pores, facilitating greater CO₂ uptake at higher pressures. An example is PCN-222, which shows high CO₂ uptake due to the presence of OMS and extensive π–π interactions within its large, flexible pores.^37,38 Boyd and co-workers¹⁴ compared the adsorption properties of porphyrin-based Al-PMOF with pyrene-based Al-PyrMOF. Their study showed that the porphyrin-based MOF has higher interaction energies and uptakes at low pressures due to the extended conjugated system. Naphthalene-based MOFs, such as Mg₂(dondc) (H₄dondc = 1,5-dioxido-2,6-naphthalenedicarboxylic acid), provide another comparison. These structures typically have smaller pore sizes and volumes, resulting in lower uptakes at higher pressures compared to pyrene-based MOFs. Pyrene-based MOFs generally provide stronger π–π interactions due to their extended conjugated system. The availability of OMS in naphthalene-based MOFs like Mg₂(dondc) leads to strong CO₂ binding sites within confined small pores.³⁹ Additionally, several naphthalene-based MOFs exhibit selective adsorption of guest molecules through gating mechanisms, enabling reversible opening and closing of pores.⁴⁰

Understanding these differences allows for the rational design of MOFs tailored to specific CO₂ separation needs, leveraging the unique strengths of each ligand type to optimize performance in various applications. This comparison underscores the importance of selecting appropriate ligands and metals to achieve desired adsorption characteristics, paving the way for developing advanced materials for efficient gas separation and carbon capture technologies.

5. Conclusion

In this study, we investigated the impact of specific chemical and structural features of TBAPy-based MOFs on CO₂ adsorption. Such a methodical study is essential for tailor-made adsorbents for a given application, including carbon capture. Even though computational high-throughput screenings are frequently employed to pinpoint optimal materials for efficient gas adsorption and separation processes, MOFs’ vast chemical and topological diversity introduces complications, underscoring that a generic approach is not always ideal.

We gain a deeper understanding of the structures and their structural behavior by comparing experimental diffraction data before and after activation, pore volumes, and CO₂ uptakes to our simulated models. Some structures undergo phase changes during activation due to the rotational freedom of the ligand and/or coordinated solvent molecules that are removed in the process. In exploring MOFs for specific applications, it is imperative to conscientiously consider potential structural changes due to various factors. Hence the need for more comprehensive investigations of diverse structures to identify promising adsorbent candidates.

Important CO₂ binding sites identified in the study of Boyd et al.¹⁴ are OMS and bridging oxygens. In this work, we show that the presence of OMS creates adsorption sites with high interaction energies. CO₂ is found to preferentially adsorb close to free carboxylates that form strong binding sites in some of the Zn-TBAPy MOFs. Besides, we investigate charged Zn-TBAPy MOFs with counterions, where Coulomb interactions dominate during adsorption. The MOFs investigated in this study exhibit promising characteristics for CO₂–N₂ separation, primarily due to their significantly lower N₂ uptake compared to CO₂, attributed to the MOFs’ stronger interaction with CO₂. Additionally, the constricted pore architecture of Zn₂-(TBAPy) acts as an effective molecular sieve, efficiently discriminating against larger molecules (e.g., N₂ and potentially CH₄). These findings not only underscore the efficacy of these MOFs in gas separation but also provide valuable insights for future research, guiding the choice of metals and ligands to optimize separation performance. However, for practical applications, especially in CO₂–N₂ separation, understanding the stability of these structures in humid conditions remains crucial, an aspect that warrants further investigation to ascertain their operational viability in real-world scenarios. While focusing on the intrinsic separation capabilities of these MOFs, this study opens avenues for exploring their adaptability to varying operational environments and their comparative performance against other materials documented in the literature.

6. Materials and Methods

6.1. MOFs and Ligand Syntheses

Synthesis of the Pyrene Ligand (TBAPy)

The two-step synthesis of the pyrene ligand used in this study (i.e., 1,3,6,8-tetrakis(p-benzoic acid)pyrene) can be accomplished according to.^25,41

Synthesis of Zn₂-(TBAPy).²⁶

In a 12 mL glass vial, introduce a mixture of Zn(NO₃) ·6 H₂O (0.03 mmol, 9 mg) and TBAPy (0.02 mmol, 10 mg) into 5 mL DMF/dioxane/H₂O (ratio 2/1/1) solvent mixture. Add concentrated HCl (32 wt %) (10 μL). The vial is heated to 120 °C for 72 h. The heating and cooling rates are 2 and 0.2 °C/min, respectively. The solid is recovered by centrifugation and washed with DMF three times and acetone twice. The MOF is activated at 110 °C for 12 h under dynamic vacuum before the adsorption measurements.

Synthesis of Zn-(Ade)(TBAPy).²⁷

In a 12 mL glass vial, introduce a mixture of Zn(NO₃) ·6 H₂O (0.057 mmol, 12 mg), TBAPy (0.015 mmol, 10 mg) and adenine (Ade) (0.059 mmol, 8 mg) into 6 mL DMF/H₂O (ratio 11/1). Add 4 drops of pure nitric acid (HNO₃). The vial is then capped and placed in the oven for 72 h at 120 °C. The heating and cooling rates are 2 and 0.2 °C/min, respectively. The solid is recovered by centrifugation and washed with DMF three times and acetone twice. The MOF is activated at 110 °C for 12 h under dynamic vacuum before the adsorption measurements.

Synthesis of Zn-(6BA)(TBAPy).²⁸

In a 12 mL glass vial, introduce a mixture of Zn(NO₃) ·6 H₂O (0.1 mmol, 29 mg), TBAPy (0.02 mmol, 12 mg) and 6-benzylaminopurine (6BA) (0.02 mmol, 5 mg) into 4.5 mL DMF/H₂O (ratio 8/1). The pH was adjusted to 3 with concentrated HNO₃ (60 μ L). The vial is then capped and placed in the oven for 72 h at 120 °C. The heating and cooling rates are 2 and 0.1 °C/min, respectively. The solid is recovered by centrifugation and washed with DMF three times and acetone twice. The MOF is activated at 190 °C for 12 h under dynamic vacuum before the adsorption measurements.

6.2. Experimental Methods

Powder X-ray Diffraction (PXRD)

PXRD data on all samples were collected on a Bruker D8 Advance diffractometer at ambient temperature using monochromated Cu Kα radiation (λ = 1.5418 Å), with a 2θ step of 0.02° with different 2θ ranges. Simulated PXRD patterns were generated from the corresponding crystal structures using Mercury 3.0.

Thermal Gravimetric Analysis (TGA)

A PerkinElmer Thermogravimetry Analyzer was used to determine the decomposition temperature of the samples. All measurements were performed under airflow up to 700 °C.

Scanning Electron Microscope (SEM)

The morphological characteristics were investigated by SEM (FEI Teneo SEM instrument). For SEM measurements, all samples were deposited on a carbon tape. Conventional TEM images were collected on the FEI Tecnai Spirit instrument at 120 kV acceleration voltage.

Gas Adsorption Measurements

The N₂ adsorption isotherm measurements were performed at 77 K by using BELSORP Mini (BEL Japan, Inc.).

For gas adsorption measurements at room temperature, 80–100 mg of sample were placed in Micromeritics adsorption cells and activated under vacuum (0.02 mbar) following a heating ramp up to the required activation temperature and remained at the same conditions for 12 h using an activation station Micromeritics VacPrep 061. After activation, the samples were allowed to cool down naturally to room temperature and backfilled with argon. Next, the evacuated cells containing degassed samples were transferred to a balance and weighed to determine the mass of sample after activation. The adsorption cells were then transferred to the analysis ports of the instrument Micromeritics 3Flex, where CO₂ (99.998% gas purity) and N₂ (99.999% gas purity) adsorption isotherms measurements at 293 K were performed using an isothermal water bath.

6.3. Computational Methods

Classical Monte Carlo (MC) combined with first-principles Density Functional Theory (DFT) simulations were performed to generate accurate framework models and characterize the adsorption processes on an atomistic level. In this section, details on the applied methods are outlined.

All DFT calculations were executed using the Quickstep code of the CP2K package (version 9.1)⁴² unless stated differently. This code is an efficient DFT implementation for large and complex structures by exploiting the mixed Gaussian and plane waves (GPW) method based on pseudopotentials. The orbital transformation (OT) technique is employed to optimize the wave function. The double-ζ DZVP-MOLOPT-SR contracted basis sets and GTH pseudopotentials represented the electronic wave function. The plane waves are mapped on a 4-level multigrid with a cutoff of 600 Ry, a relative cutoff of 50 Ry and a progression factor of 3. The exchange-correlation energy is approximated using the Perdew–Burke–Ernzerhof (PBE) functional,⁴³ combined with the DFT-D3(BJ) model⁴⁴ to correct for dispersive many-body interactions (”PBE-D3 functional”. The integration grid and DFT functional settings were adapted from Ongari.⁴⁵ In this work, the authors tested and optimized the DFT settings for robust convergence in high-throughput calculations of covalent organic frameworks (COFs) and MOFs. In addition, the functional choice in this study was supported by a benchmark study of Nazarian et al.,⁴⁶ who assessed DFT predictions of MOF framework structures and properties of a diverse test set. They found that, on average DFT calculations using dispersion corrected functional (PBE-D2,⁴⁷ PBE-D3⁴⁴ and vdW-DF2⁴⁸ outperformed other functionals commonly used for MOF structure and property predictions including the GGA functionals PBE⁴³ and PW91^43,49 and the meta-GGA functional M06L.⁵⁰ PBE-D3 was selected in this study. The PBE functional is commonly used and tested in our group, and unlike PBE-D2, the dispersion coefficients used in PBE-D3 are geometry dependent. They are adjusted based on the coordination number of the atoms.

The RASPA molecular simulation software for adsorption and diffusion in nanoporous materials⁵¹ was used to perform MC simulations. The optimized framework geometries were kept rigid in all classical simulations. As stated above, the benzoate groups of the ligands possess various degrees of rotational freedom in the different frameworks and might rotate upon adsorbate loading, making narrow pores accessible. Therefore, no blocking spheres were used,⁵² keeping all cavities accessible. To describe the energy surface, solely dispersive interactions represented by the Lennard-Jones potential (LJ) and Coulomb interactions were considered. Periodic boundary conditions were employed with a cutoff radius of 12 Å, including tail corrections to remedy the truncation. The Ewald summation technique⁵³ was used to model the charges in periodic cells. The TraPPE FF⁵⁴ was utilized to model gas–gas interactions for CO₂ and N₂. The dispersion interactions of the framework and the gases were modeled with the Universal FF (UFF)⁵⁵ LJ parameters for both the gas and MOF atoms as suggested by Ongari et al.⁵⁶

To ensure the reproducibility and direct comparability of computed data, unified workflows were used for several parts of this study. The Automated Interactive Infrastructure and Database for Computational Science (AiiDA)⁵⁷ was employed to orchestrate the different steps, managing the interaction of different codes and providing automation and similarity of the calculations. The workflows are published and maintained as the ”AiiDA-LSMO” plugin on GitHub.⁵⁸

6.3.1. Structures Optimization

First, a preliminary cell optimization step was performed for a maximum of 200 cycles to optimize the frameworks, keeping the unit cell angles fixed. This step ensures that the atom positions were relaxed and an adequate minimum was found based on the experimental cell parameters before optimizing the unit cell vectors. As a result, the simulated cell was kept closer to experiments. Subsequently, a final cell optimization was performed without constraints on the unit cell’s angles. In the first step, the Broyden-Fletcher-Goldfarb-Shanno (BFGS) minimizer is used, and the limited-memory BFGS minimizer is employed for the final cell optimization. The threshold for the pressure was set to 100 bar. To provide a proper representation of the electron density by the employed OT technique, supercells were built with every cell length being >12 Å. Structure optimizations of these frameworks were then performed in two steps based on the high-throughput optimization protocol developed by Ongari et al.⁴⁵ Convergence criteria and the number of cycles were adjusted for this study. The goal was to increase the accuracy of the structure optimization calculations and find a proper minimum of the atom positions before optimizing the structure’s unit cell. In this study, the system was considered converged, when the maximum forces on the atoms dropped below 0.45 mHA a₀^–1 and the geometry change between the current and the last optimizer iteration was lower than 3 ma₀, according to the default settings implemented in CP2K.⁵⁹ Because of the large quadrupole moment of CO₂, it is important to have an accurate model of atomic point charges to evaluate the electrostatic interactions in classical simulations. The Density Derived Electrostatic and Chemical (DDEC) approach is exploited to assign net atomic charges (NACs) to the framework atoms of the optimized structures. In this work, the DDEC6 method as implemented in the Chargemol software⁶⁰ was used to perform atomic population analysis based on the electron density obtained from a DFT calculation using the settings as detailed above.

Ab initio molecular dynamics (MD) simulations were performed in the constant temperature and pressure ensemble using a flexible cell (NPT_F) at 300 K and 1 bar. The CSVR thermostat⁶¹ was used, in combination with the barostat developed by Martyna et al.⁶²

Geometric properties of the materials (pore volumes and pore diameters) were evaluated using the software Zeo++.⁶³ Pore volumes were accessed using the probe-occupiable pore volume model as implemented in the software. This technique was developed by Ongari et al.,⁶⁴ providing a computational pore volume definition directly related to experimental pore volumes obtained from nitrogen isotherms.

6.3.2. Counterions Simulations

A model of the dimethylammonium cations (DMA) cations in the pores of Zn-(Ade)(TBAPy) and Zn-(6BA)(TBAPy) was built using the Maestro software.⁶⁵ The atomic coordinates were then optimized using DFT calculations implemented in the C.01 Revision of the Gaussian 16 program.⁶⁶ The molecule was described using the PBE exchange and correlation functional and 6-311G(d,p) triple-ζ Pople basis sets^67,68 with polarization and diffuse function. The D3 version of Grimme’s dispersion with Becke-Johnson damping was used to correct for van-der-Waals interactions.⁴⁴ Point charges were then assigned to the molecule atoms. As electrostatic forces dominate the interactions of charged molecules, it is critical to model the charges accurately. Two different methods were tested:

• The first protocol was designed similar to the methods used to assign point charges to the framework atoms (see Supplementary Note 7.3.1: A Mulliken population analysis⁶⁹ was performed in CP2K and the DDEC method was then used to calculate NACs.

• In the CM5 model⁷⁰ partial atomic charges are obtained by mapping from those calculated by Hirshfeld population analysis⁷¹ of density functional electronic charge distributions. The predictions performed with this model are reported to be more accurate on average than results from the Mulliken scheme.⁷⁰

Interaction energies between DMA ions and Zn-(Ade)(TBAPy) and Zn-(6BA)(TBAPy), respectively, calculated with both schemes, differ by less than 1%. Therefore, DDEC charges were used to model DMA point charges consistent with the calculated framework charges. The LJ parameters of the molecule were adapted from the work of Anderson et al.²⁷ Herein, the authors applied the nonbonded interaction parameters of DMA to study nucleobase pairing and photodimerization in Zn-(Ade)(TBAPy).

The interactions between framework atoms and cations are very high, i.e. exceeding −60 000 kJ/mol for Zn-(Ade)(TBAPy) and approximately −90 000 kJ/mol for Zn-(6BA)(TBAPy) respectively, and dominated by coulomb forces. Due to the high interaction energies, it can be assumed that the mobility of the ions in the frameworks is very low. Therefore, the ions were kept fixed in all classical simulations and treated as part of the framework.

To generate structure models of the anionic MOFs with counterions in their pores, we initially optimized the frameworks with a net negative charge, calculating point charges as detailed in section 6.3.1. Subsequently, we employed simulated annealing, outlined in the following paragraph, to position the counterions within the anionic MOFs’ structures, maintaining the framework’s rigidity throughout the process. A subsequent optimization was performed on the ions-MOF system, resulting in a neutrally charged system with designated atomic point charges. The Zn-(Ade)(TBAPy) framework has a net charge of −8,²⁷ necessitating the placement of 8 counterions per unit cell. As can be seen in the framework model illustrated in Figure 1, it was found that the DMA ions locate in the acidic pores close to the free carboxylates where negative local charges prevail. For Zn-(6BA)(TBAPy), which has a −4 charge per unit cell,²⁸ we inserted 4 counterions into the unit cells of both structural models following the same methodology. Notably, the counterions located at different positions for the two framework models: For the structure with coordinated water molecules, the ions are situated close to the free carboxylate-oxygens, whereas the ions reside in the vicinity of the coordinatively unsaturated Zn-ions for the framework model with OMS. Furthermore, upon removal of the coordinated solvent, a compression of the lengths a and c and a decrease of the monoclinic angle β of the crystallographic unit cell could be observed. The changes in the unit cell parameters can be explained by the positioning of the counterions close to the metal centers which might lead to a tighter packing, and the removal of the coordinated water molecules and thus the interactions they formerly participated in. Without these interactions, the framework might be inclined to undergo structural adjustments to minimize void space or optimize interactions between remaining components.

6.3.3. Binding Site Workflow

The minimum energy configuration of a single CO₂ adsorbate molecule in the different frameworks was determined using the simulated annealing technique.⁷² To this end, Monte Carlo simulations in the canonical ensemble (NVT ensemble) were performed at decreasing temperatures to move the molecule to the global minimum configuration. At each temperature step, 10⁵ MC cycles were used to determine the molecule position, which then served as the starting configuration for the subsequent temperature step. The molecule position was then fine-tuned in a minimization step.

Interaction energies derived from DFT calculations were obtained by first relaxing the configuration obtained from simulated annealing in a geometry optimization step (fixed unit cell vectors). The framework was kept rigid to ensure direct comparability with results obtained from classical simulations. Consequently, the only iterables were the positions of CO₂ atoms. To find the global minimum of the CO₂ configuration with these restrictions, a tight convergence threshold was used (maximum forces of 0.01 mHA a₀^–1 and geometry change below 0.05 ma₀). The system energies were then corrected for the basis set superposition error (BSSE) using the counterpoise method proposed by Boys and Bernardi.⁷³ The interaction energy obtained from classical simulations (E_FF) was then compared to the energy obtained from first-principles simulations (E_DFT) in the same configuration. Comparing these energies is a benchmark on how accurately the FF parameters model the interactions of a single CO₂ molecule with the framework atoms.

The interaction energy obtained from classical simulations (E_FF) was then compared to the energy obtained from first-principles simulations (E_DFT) in the same configuration. The comparison of these energies serves as a benchmark on how accurately the FF parameters model the interactions of a single CO₂ molecule with the framework atoms.

6.3.4. Isotherm Workflow

The adsorption isotherms were simulated in the Grand Canonical ensemble (GCMC ensemble). Here, 10⁴ cycles were used for equilibration and 10⁵ cycles for production. Simulations at subsequent pressure points were performed starting from the restart file of the previous pressure step, thus reducing the number of cycles necessary for initialization.

Data Availability Statement

All data generated during this study are included in this article and respective Supporting Information. The characterization and simulation data that support the findings of this study are also available on Zenodo (DOI: 10.5281/zenodo.10684869).⁷⁴

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsami.4c05527.

• Physical properties of CO₂ and N₂ (Table S1), Zn₂-(TBAPy) structural changes (Figure S1), thermal gravimetric analysis (TGA) (Figure S2), scanning electron-microscopy (SEM) images (Figure S3), PXRD showcasing the distortion of Zn₂-(TBAPy) (Figure S4), Le Bail fits of the activated structures (Figure S5) with the corresponding unit cell parameters obtained (Table S2), N₂ isotherms at 77 K (Figure S6), pore volume analysis (Table S2), PXRD of the activated materials (Figure S7), CO₂ binding sites (Figure S8 and Table S4) (PDF)

Author Contributions

^§ M.J.P. and N.P.D. contributed equally to this work.

Author Contributions

N.P.D., F.P.U., and C.P.I. conceived and designed the project. N.P.D. performed the experiments, and N.P.D., F.P.U., and C.P.I. analyzed the data. J.E. and N.P.D. measured CO₂ and N₂ isotherms of all MOFs. M.J.P. performed the simulations, and M.J.P. and A.O.G analyzed the computational data. N.P.D., M.J.P., and B.S. wrote the manuscript with the input of all authors.

The authors declare no competing financial interest.