Selective adsorption of CO₂ in TAMOF-1 for the separation of CO₂/CH₄ gas mixtures

Abstract

TAMOF-1 is a robust, highly porous metal–organic framework built from Cu²⁺ centers linked by a L-histidine derivative. Thanks to its high porosity and homochirality, TAMOF-1 has shown interesting molecular recognition properties, being able to resolve racemic mixtures of small organic molecules in gas and liquid phases. Now, we have discovered that TAMOF-1 also offers a competitive performance as solid adsorbent for CO₂ physisorption, offering promising CO₂ adsorption capacity ( > 3.8 mmol g^–1) and CO₂/CH₄ Ideal Adsorbed Solution Theory (IAST) selectivity ( > 40) at ambient conditions. Moreover, the material exhibits favorable adsorption kinetics under dynamic conditions, demonstrating good stability in high-humidity environments and minimal degradation in strongly acidic media. We have identified the key interactions of CO₂ within the TAMOF-1 framework by a combination of structural (neutron diffraction), spectroscopic and theoretical analyses which conclude a dual-site adsorption mechanism with the majority of adsorbed CO₂ molecules occupying the empty voids in the TAMOF-1 channels without strong, directional supramolecular interactions. This very weak dominant binding opens the possibility of a low energy regeneration process for convenient CO₂ purification. These features identify TAMOF-1 as a viable solid-state adsorbent for the realization of affordable biogas upgrading.

Introduction

The presence of CO₂ in gas streams is deleterious to the environment as emissions, but also problematic as an undesirable impurity in industrial feedstocks and in methane-based gas mixtures such as biogas and natural gas^1,2. The former is a green replacement for the latter, where methane is obtained from the anaerobic fermentation of biological residues, as a renewable fuel³. Its exploitation offers a functional solution for the conversion of biowaste into usable renewable energy while reducing greenhouse gases. Biogas upgrading into biomethane to be used as a fuel (high calorific value) which can meet pipeline grade requirements is of great environmental, economic, and technological interest^4,5. However, the separation and capture of the carbon dioxide (typically around 50% in biogas) is an extremely challenging process as (i) the CH₄ and CO₂ molecules have similar kinetic diameters making the separation of mixture difficult by kinetic/size exclusion; (ii) low-pressure operations are preferred, close to 1 bar (low driving force) to avoid high energy-consumption from compression work; and iii) it needs very high efficiency in CO₂ removal (≥ 98%)^6,7.

Gas-liquid absorption in chemical solvents (aqueous amine solution)⁸ is the most mature technology for large-scale CO₂ capture^9,10, including for methane purification¹¹. However, this technology has three major drawbacks: high energy requirement for solvent regeneration, waste management (solid salts and traces of gaseous compounds due to amine degradation), and a large footprint. This calls for the development of alternative solutions based on novel and cutting-edge materials. In particular, the separation of CO₂/CH₄ mixtures directly in the gas phase would be highly desirable. In this context, a promising process for gas separation is physisorption^12,13. Due to the physical nature of the interactions involved, this process typically demands lower energy consumption during the regeneration step when compared to processes ruled by strong chemical interactions such as chemisorption in solid or liquid adsorbents. Different technologies based on physisorption are available, such as pressure swing adsorption (PSA)^14,15, vacuum swing adsorption (VSA)¹⁶, and thermal swing adsorption (TSA)¹⁷. These technologies involve at least two consecutive steps. First, CO₂ is selectively adsorbed by the active component at high pressure and/or low temperature, and then the adsorbent is regenerated by CO₂ desorption at lower pressure/vacuum (PSA/VSA) or/and at higher temperature (TSA).

Several types of porous materials have been investigated as physisorbents for CO₂ removal¹⁸. Benchmark zeolites typically exhibit competitive separation performance, characterized by high adsorption capacity and equilibrium selectivity¹⁹. However, while advantageous for CO₂ removal from dry emission or process streams, this selectivity is often compromised in the presence of water. Water molecules preferentially coordinate with exposed metal sites, outcompeting CO₂²⁰. Furthermore, these materials necessitate high regeneration energy^19,21,22, which consequently increases the operational expenditure (OPEX) in industrial plants striving for high CO₂^15,20. Carbon-based materials offer the advantage of low cost and easy regeneration, although they are limited by a lower adsorption capacity and selectivity. Nowadays, the limitations of the adsorbents are also limiting the commercial use of carbon capture (CC) technologies just to niche applications, where there is no alternative. Suitable materials combining high adsorption capacity at low/moderate working pressures, able to discriminate CO₂ against molecules with similar dimensions (e.g., CH₄ and N₂) with high selectivity, with low-energy regeneration and fast adsorption/desorption kinetics would offer plausible opportunities to realize CC not only for biogas upgrading but also in many additional fields. Especially if CC becomes economically competitive with current CO₂ emission rights.

Metal-organic frameworks (MOFs) are crystalline and porous materials^23,24 prepared by the self-assembly of metal ions or clusters with organic ligands or linkers to form reticular structures^25,26. Due to their exceptional properties²⁷ such as high accessible surface areas (up to 6200 m²g^–1)²⁸, extra high porosities (up to 90%)²⁹, tunability in pore dimensions and morphology³⁰, versatile functionality³¹ as well as thermal and chemical stability³², MOFs are recognized as promising materials for selective gas separations potentially able to overcome intrinsic limitations of common adsorbents^33,34,35. Even though nowadays more than 70000 MOFs have been discovered, just a few of them have shown promising features into adsorption/desorption protocols or as membrane components for CO₂ capture^23,36. Some of the most remarkable candidates include CALF-20, a Zn-based oxalate-bridged framework with high performance in CO₂ capture from flue gas, and an extraordinary robustness under humid conditions³⁷; or Mg₂(dobpdc), a Mg-based tetramine-functionalized framework with extraordinary cyclability and stability during CO₂ capture from flue gas. However, this last material is characterized by chemical CO₂ adsorption and needs a high operating temperature (100 °C)³⁸. In the case of CO₂/CH₄ separations for biogas upgrading, the most relevant example is MUF-16 (Co(Haip)₂, Haip = 5-aminoisophthalic acid)³⁹. This MOF is able to capture carbon dioxide from hydrocarbons with exceptional selectivity although with moderate CO₂ adsorption capacity⁴⁰. Moreover, the low isosteric heat of adsorption allows for an easy regeneration step. Another interesting MOF is Qc-5-Cu-sql, a Cu- quinoline-5-carboxylic acid supramolecular network, also showing excellent CO₂/CH₄ selectivity via the molecular sieving mechanism. However, also in this case moderate CO₂ adsorption capacity is reported⁴¹. Moreover, a thorough explanation of the selectivity of MOFs to adsorb CO₂ is usually missing. Whereas chemisorption is easier to evaluate from structural data, the physisorption mechanism remains difficult to assess. In situ experiments are very rare to discriminate between: (i) size-exclusion principles, (ii) kinetically controlled processes, or (iii) thermodynamically controlled processes⁴².

TAMOF-1 is the first in a series of homochiral MOFs based on natural amino acid derivatives, by transformation of the α-amino unit into a triazole group^43,44. TAMOF-1 ([Cu(S-TA)₂].xH₂O, S–HTA = (S)-3-(1H-imidazol-5-yl)-2-(4H-1,2,4-triazol-4-yl)-propanoic acid) is easily synthesized at large scale from low-cost raw materials just by reaction of a copper(II) salt with L-histidine derivative (imidazole-5-ylmethyl)-(1,2,4–triazol-4-yl)acetate (L₁) in water. This material has a 3D network, made from 10 Å wide, helicoidal, intercommunicated channels, decorated with multiple dangling functional groups carboxylate, triazole, and imidazole (Fig. 1) exhibiting a BET-specific surface area of 980 ± 50 m² g^–1 and a micropore volume of 0.38 cm³g^–1. TAMOF-1 exhibits exceptional water stability, withstanding both hydration/dehydration cycles without structural degradation or porosity loss⁴⁴. Moreover, activation for gas separation is achieved under mild conditions (353–393 K under a sweep gas flow), preserving the MOF crystallinity—a rare feat for Cu-based MOFs and surpassing the requirements of typical MOFs and zeolites^45,46.

Fig. 1: Crystal structure of TAMOF-1.

Representation of the crystal structure of TAMOF-1, showing the network of open 10 Å-wide channels. Color code: Cu, deep blue; O, red; N, light blue; C, black. Hydrogen atoms omitted for clarity.

In previous reports, we disclosed the performance of TAMOF-1 as a stationary phase for the chromatographic separation of racemic mixtures of organic substances thanks to its porosity and homochirality⁴⁷. The latter was further exploited for the kinetic resolution of chiral substrates by catalytic coupling. Moreover, the characteristic shape and size of its channels and pores, as well as the Cu metal center, make TAMOF-1 a good candidate for separating also a wide variety of volatile organic compounds such as benzene−cyclohexane system and xylene isomers (positional isomers reconnection). The separation of these molecules has been proved in both the liquid and gas phases⁴⁸. Furthermore, TAMOF-1 has been reported to be capable of separating C₂H₂/C₂H₄ and C₂H₂/CO₂ mixtures⁴⁹.

Here, we report that TAMOF-1 is a highly promising physisorbent to resolve CO₂/CH₄ mixtures and, in a broader context, aiming to capture CO₂. In particular, this material has highly promising CO₂ adsorption capacity and CO₂/CH₄ selectivity. Moreover, it allows for a low energy (ambient temperature) regeneration. When compared with other available adsorbents, including other MOFs, the TAMOF-1 performance appears very promising also in terms of operation costs. We anticipate that this material may help to bridge the gap between effective adsorption and affordable regeneration. Noteworthy, the adsorption/desorption performance remains durable over time, even in the presence of water, with only a small reduction observed when exposed to highly concentrated H₂S. By combining specific experimental (neutron diffraction and IR spectroscopy) and theoretical tools (Monte Carlo and molecular dynamics), we have thoroughly elucidated the molecular mechanism of the physisorption and transport properties of gas molecules within the TAMOF-1 network at the origin of its promising performance.

Results and discussion

Gas adsorption

Single gas (CO₂, CH₄, and N₂) adsorption isotherms of TAMOF-1 powder up to 10 bars were measured in the 293–353 K temperature range (Fig. 2a). As a function of temperature, all isotherms maintain their characteristic shapes, with gas uptake increasing as the temperature decreases, typical of a physisorption process. Type I isotherms are obtained for CO₂, without reaching a plateau (saturation) in the investigated pressure range. For example, at 298 K, CO₂ adsorption capacities of 3.8 and 6.5 mmol g^–1 are obtained at 1 and 10 bars, respectively. Linear isotherms are obtained for both CH₄ and N₂, with adsorption capacity following the order CO₂ > > CH₄ > N₂, indicating selective preferential uptake of carbon dioxide.

Fig. 2: Adsorption and desorption properties of TAMOF-1.

a Adsorption/desorption isotherms of CO₂, CH₄ and N₂ in a TAMOF-1 powder sample at 298 K up to 10 bar. b Adsorption isotherms of CO₂ with a temperature range of 293–353 K up to 10 bar. c IAST selectivity of different CO₂/CH₄ gas mixtures with varying ratios at 25 °C up to 1 bar. d CO₂ sorption capacity versus CO₂/CH₄ IAST selectivity for binary 50:50 v/v mixtures at 1 bar and 293 K (IAST selectivity values > 20), adapted from ref. ⁴². e Breakthrough curves of CO₂ (solid line) and CH₄ (dashed line) for binary mixtures with different CO₂/CH₄ ratios: 25:75(blue), 50:50(red), 75:25(green). f CO₂ breakthrough curves in different inlet gas mixtures: 50:50 CO₂/CH₄ (red), 50:50 CO₂/N₂ (cyan), CO₂/CH₄/N₂ = 50/₂5/25 (magenta). Measures in (e and f) were performed with 0.7 g of activated TAMOF-1 powder. TAMOF-1 was activated under vacuum (10⁻¹mbar) in both column top and bottom sides at 393 K for 15 h. g CO₂/CH₄ (50:50) breakthrough curves for activated TAMOF-1 (7.7 g) at 72% relative humidity (RH). TAMOF-1 was activated under vacuum (10^-1 mbar) in both column top and bottom sides at 393 K for 15 h. h CO₂/CH₄ (50:50) breakthrough curves for activated TAMOF-1 (0.7 g) under acid gas conditions (0.9% H₂S, 2.5% H₂O). TAMOF-1 was activated under N₂ flow (170 mil min^–1 g_TAMOF ^–1) at 393 K for 15 h. All breakthrough experiments were performed at 298 K and 1.05 bar. See Supplementary Tables 19–21 for detailed fixed-bed adsorption parameters.

From the different adsorption models available (See SI), the dual-site Langmuir-Freundlich model shows the best fitting for the CO₂ adsorption isotherms (Supplementary Figs. 7–11). See, for instance, the comparison with the single-site Langmuir-Freundlich model in the logq vs. logP plot (Supplementary Fig 12). This indicates that at least two different adsorption sites are responsible for dominant CO₂ uptake. Supplementary Tables 16, 17 show the isotherm parameters and regression coefficients.

The CO₂/CH₄ selectivity calculated by the ideal adsorption solution theory are reported in Fig. 2c and in the Supplementary Table 18. At 293 K and 1 bar, 45.9, 40.9, and 38.30 selectivity values are obtained for respectively 30:70, 50:50, and 70:30 CO₂/CH₄ gas mixtures (in the typical biogas composition range). Interestingly, data indicate that TAMOF-1 is capable of separating CO₂/CH₄ mixtures over a wide temperature range. For example, at 1 bar and 353 K, TAMOF-1 has a CO₂ sorption capacity and 50:50 CO₂/CH₄ ideal selectivity of 1.8 mmol g^–1 and 18, respectively. Another important feature is the complete adsorption/desorption reversibility obtained for all the gases (Fig. 2b) with no hysteresis observed. This behavior aligns with the physical nature of the gas adsorption phenomena within the TAMOF-1 network and underscores the structural rigidity of this material.

The isosteric enthalpy of adsorption, ΔH_ads was estimated via an indirect approach from the CO₂ adsorption isotherms at the different temperatures using the Clausius-Clapeyron equation (for a detailed description of the calculations, see SI). Supplementary Fig. 13 shows the evolution of –ΔH_ads with CO₂ uptake. A high –ΔH_ads is found at low pressure (zero coverage, –ΔH⁰_ads=0), and then it rapidly decreases as CO₂ uptake increases, approaching the bulk-phase sublimation heat of CO₂, 26–27 kJ mol^–1, at an adsorption capacity higher than 2 mmol g^–1 50. This energy range points to a physical gas adsorption behavior and denotes a heterogeneous adsorption process occurring at multiple adsorption sites with different surface energies⁵¹.

Sahoo et al.⁴² conducted a preliminary comparison of mixed gas phase CO₂ sorption capacity and separation selectivity for various MOFs using IAST (Fig. 2d). While many studies report only single gas sorption capacity (which is typically higher than the mixed gas capacity), the review focused on those that provided mixed gas CO₂ data. Among the MOFs examined, MUF-16 and Qc-5-Cu-sql-β demonstrated both high CO₂ sorption capacity (1.8 and 1.6 mmol g^–1, respectively) and high CO₂/CH₄ separation selectivity (6690 and 3300, respectively). Interestingly, TAMOF-1 (this study) exhibited an even higher mixed CO₂ sorption capacity (~ 3.1 mmol g^–1) with acceptable separation selectivity (> 40). Given the lack of established performance benchmarks for industrial-scale biogas separation, it is premature to definitively assess the potential of these MOFs. We have also to mention that IAST parameters are calculated under idealized equilibrium conditions and do not represent the actual separation performance of adsorbents in dynamic conditions such as those obtained from breakthrough curves. Nevertheless, IAST analysis is usually reported and used to compare the gas separation performance of adsorbents valuable insights into the relative adsorption capabilities of TAMOF-1 with respect to other adsorbents.

Breakthrough measurements

In Fig. 2e, we report methane/carbon dioxide breakthrough curves through a TAMOF-1 powder bed at 1 bar and 298 K at different CO₂/CH₄ ratios. Gas separation parameters obtained from the breakthrough curves are reported in Supplementary Table 19 (See SI for details). Pure methane (≥ 99.9%) elutes first and speedily from the bed. Regular S-shaped curves were instead observed for CO₂. Effective CO₂/CH₄ separation is achieved in all cases as CO₂ is delayed. This confirms slower CO₂ diffusion through the TAMOF-1 bed. By reducing CO₂ molar flow, the related curves shift towards higher breakthrough times (Bt). Bt(CO₂) increases from 16.4 to 18.7 min g^–1 when the CO₂/CH₄ ratio is reduced from 50:50 to 25:75. In contrast, Bt(CH₄) is almost constant, independent of the CH₄ fraction. The CH₄ concentration overshoots its equilibrium value (C/C₀ > 1) before returning to equilibrium (C/C₀ = 1). This roll-up effect indicates preferential adsorption of CO₂, which displaces some of the initially adsorbed CH₄. The precise shape of this overshoot depends on the methane flow rate, concentration, and adsorbent properties²⁰. Adsorption capacities for both carbon dioxide and methane were also measured.

Nitrogen is a gas that can be present in combination with CO₂ and CH₄ in many gas streams, thus we also evaluated CO₂/CH₄/N₂ separations through TAMOF-1 beds. For this purpose, fixed-bed column experiments were performed by varying the CH₄/N₂ ratio at constant CO₂ concentration (50%) (Fig. 2f).

Carbon dioxide breakthrough curves perfectly overlap independently of the CH₄/N₂ ratio. This indicates that CO₂ diffusion through TAMOF-1 is not affected by the type of other gas components (CH₄ or N₂) in the inlet gas mixture. These gases do not affect the adsorption capacity, which depends directly on the CO₂ concentration, in agreement with isotherms measures (Fig. 2a). As a consequence, CO₂ adsorption capacities (Supplementary Table 19) are similar (q_b = 1.7 ± 0.06 mmol g^–1, q_s = 2.5 mmol g^–1, CO₂ = 50%). A similar CO₂/CH₄ selectivity was also observed for a binary CO₂/CH₄ (S_s = 6) or a CO₂/CH₄/N₂ ternary (S_s = 4) inlet gas mixture with the same CO₂ concentration (i.e., 50%). It is important to mention that N₂, when present, exits as the first component and almost immediately from the column with elution time almost coincident with the dead time.

Breakthrough experiments at variable pressure conditions were also collected (Supplementary Table 19). By pressurizing the bed, Bt shifts to higher times with methane remaining always the first gas to elute. Increasing bed pressure from 1.2 to 6 bar at a constant flow rate, Bt(CO₂) increases from 34 to 122 min g^–1, and Bt(CH₄) from < 4.7 to 32 min g^–1. CO₂ adsorption capacities increase from 0.42 to 1.24 mmol g^–1, but CH₄ adsorption also increases from < 0.06 to 0.33 mmol g^–1. This results in a decrease in the CO₂/CH₄ selectivity as a function of increasing pressure, from > 7 to < 4 between 1.2 and 6 bar.

Higher temperatures accelerate the gas elution, and the CO₂ concentration profiles sensibly shift toward lower Bt and lower adsorption capacity (Supplementary Table 19). This is in agreement with the weaker adsorbent-adsorbate interactions at higher temperatures, the faster molecular diffusion and the exothermic character of the adsorption process. A negligible temperature effect was instead observed for CH₄, that speedily diffuses into the bed and almost immediately exits from the bed. Separation effectiveness decreases by increasing the temperature, and at temperatures similar to the activation one (393 K), CO₂ and CH₄ curves nearly overlap. Good separation capability (S_b > 0.9) was found up to 353 K, indicating applicability in a wide temperature range.

Regeneration and stability assessment of TAMOF-1

The regeneration step of a TAMOF-1 bed after saturation appears to be quite simple, avoiding the need for thermal heating. TAMOF-1 can be completely regenerated at a constant temperature just by passing through a sweep gas (N₂). The corresponding CH₄ and CO₂ desorption curves after saturation by an N₂ sweep gas stream were collected for different flow rates, temperatures, and pressures (Fig. 3)

Fig. 3: Gas desorption dynamics of TAMOF-1.

TAMOF-1 desorption curves of CO₂ (full line) and CH₄ (dashed line) in a N₂ sweep gas flow to regenerate the TAMOF-1 after CO₂-CH₄ breakthrough experiments for (a) different concentrations of equimolecular CO₂-CH₄ mixtures, (b) different pressures, (c) different flow rates and (d) different temperatures. In all cases, the processes were performed at constant pressure, temperature, and overall flow rate during both, the breakthrough and desorption steps. TAMOF-1 (0.7 g) was activated under N₂ flow (170 mL min^–1 g_TAMOF^–1) at 393 K for 15 h.

CH₄ desorbs faster than CO₂, as expected. The high CO₂/CH₄ selectivity in desorption allows for the recovery of a high purity CO₂ fraction, which can be reused as a commodity. Moreover, the regeneration times are relatively short, when compared with the Bt times. The regeneration time can be accelerated by increasing the nitrogen flow rate, or reducing the pressure, or increasing the temperature. All these data highlight the potential of TAMOF-1 as an adsorbent for pressure/vacuum/thermal swing adsorption processes.

The presence of humidity is inherent in all raw biogas mixtures, typically reaching saturation levels under water²⁰. From an engineering perspective, and particularly when employing adsorbents under vacuum conditions, partial removal of water from the inlet stream becomes crucial. This necessity arises from the potential for water condensation within the column, leading to obstructions that can diminish vacuum pressure during the desorption phase⁵². However, it is important to acknowledge that a certain degree of water vapor will invariably persist in the biogas mixture entering the adsorbent column. Previous investigations have demonstrated the structural robustness of TAMOF-1 under humid conditions, even in liquid water, with the material exhibiting complete reactivation. Here, we evaluate the competitive adsorption behavior of CO₂ and H₂O, as well as the adsorbent cyclic stability under dynamic adsorption conditions. As depicted in Fig. 2g, even in the presence of high humidity (~ 72% RH), TAMOF-1 retains its CO₂ adsorption capacity (2.3 mmol g⁻¹) with just a 15% reduction compared to the dry inlet mixture. The CO₂/CH₄ selectivity is also maintained within the range of 4–6. Subsequent to the adsorption step, regeneration was conducted under vacuum at a temperature of 80 °C. The separation performance of TAMOF-1 remained consistent over multiple adsorption-desorption cycles, as demonstrated by the overlapping CO₂ and CH₄ breakthrough curves (Supplementary Table 20). This stability indicates negligible degradation of the material after successive cycles and near-complete recovery of working capacity under the operating conditions investigated.

Hydrogen sulfide (H₂S), a common impurity found in raw biogas (~ 0–100 ppm¹), presents a significant challenge to biogas upgrading. H₂S is a highly corrosive acid gas that can compromise the chemical stability of materials used in separation processes. For instance, H₂S can disrupt the coordination bonds between organic ligands and metal centers in the MOFs, leading to structural degradation⁵³. To assess the separation behavior of MOFs in the presence of hydrogen sulfide is therefore crucial for their effective application in biogas upgrading. Analysis of TAMOF-1 in humid (2.5 mol% H₂O) and acidic gas (0.9 mol% H₂S) conditions (Fig. 2h) demonstrates its ability to separate CO₂/CH₄ mixtures, even under these conditions. However, a reduction of adsorption capacities (16%) is observed between the first and the second cycle, while selectivity remains relatively constant (~ 3–5). We observed a reduction in the competitive adsorption between CO₂ and CH₄, with negligible co-adsorption of CH₄. We assign this difference to the very high retention time found for H₂S (no H₂S was detected in the outlet stream not even after the CO₂ equilibrium regime), which competes for adsorption sites. Importantly, this analysis was performed under accelerated conditions, with very high H₂S concentration (9000 ppm), almost two order or magnitude higher than what is usually found in biogases. These data, therefore, demonstrate a relatively high robustness of the material also in strongly acid conditions. However, H₂S removal is recommended prior to biogas upgrading with TAMOF-1 in order to increase material durability.

Structure determination with adsorbed CO₂

We attempted to localize the preferential crystallographic position of the CO₂ molecules when adsorbed in the TAMOF-1 framework by X-ray diffraction analysis. A single crystal was dehydrated, exposed to a CO₂ stream, and then cooled down to 100 K. The analysis of the XRD structural data revealed residual densities in the channels that were compatible with the presence of CO₂ in the crystal matrix. However, the poor selectivity under X-ray diffraction conditions between oxygen and carbon, in addition to the possible presence of residual water, precluded the successful location of CO₂ molecules in the channels. For this reason, we turned to neutron diffraction⁵⁴, where carbon and oxygen atoms can be properly identified thanks to their distinct neutron scattering lengths.

The neutron diffraction structure analysis of an activated TAMOF-1 single crystal confirms the crystal structure obtained by XRD (Fig. 4) with triangular empty channels running along the [111] direction, which is a 3-fold axis due to its cubic symmetry (Supplementary Tables 1–7). After exposure to CO₂, and cooling down to 100 K, these channels are filled with CO₂ molecules exhibiting moderate disorder (Fig. 4), indicating that weak interactions with the TAMOF-1 network are dominant (Supplementary Tables 8–15). Adjacent to this channel, three copper atoms self-assemble in two orientations that are rotated approximately 45° with respect to each other every 13.5 Å along the [111] axis, forming equilateral triangles of approximately 6.7 Å. This arrangement creates a narrower channel where CO₂ molecules align themselves with the oxygen’s lone pairs located between the Lewis sites of two of the copper atoms in the triangle. These CO₂ molecules are the only ones that exhibit a clear, directional interaction with any of the available TAMOF-1 sites. We associate this crystallographic position with the higher isosteric enthalpy of adsorption found at low coverage. The rest of the CO₂ molecules are located along the channels of the structure with apparently random orientations.

Fig. 4: Crystal structure from neutron data showing the occupation of CO₂ molecules into the channels in TAMOF-1.

a Detail of the CO₂ positions within the Cu₃ triangles in the largest void space of TAMOF-1. The distances O1C-Cu1(zxy) and O2C-Cu1 are 2.7726(1) and 2.7068(1) Å, respectively. The view direction is along [111]. Cu1, Cu1(zxy) and Cu1(yzx) are related by a crystallographic three-fold axis, which also relates the CO₂ molecule at C1C to two others, which are not shown. C1C/O1C/O2C and its two congeners have site occupancy factors of 1/3 as a result of disorder about the symmetry axis. Color code: Cu, deep blue; O, red; N, light blue; C, gray; H, white. b View along [111] of the TAMOF-1 open framework (green stick representation) and the positions occupied by the guest CO₂ molecules (red van der Waals’ radii representation) in the TAMOF-1 channels, with the two components shown together (left) and individually (right). View along [100] (c) and view along [111] (d), of parts of the complex void structure that accommodates the guest CO₂ in TAMOF-1. The voids are represented as lighted areas for exterior surfaces and as shadowed areas for interior surfaces of the void boundaries.

The intermolecular interaction between CO₂ molecules located in the channel is consistent with London dispersion forces in agreement instead with the lowest isosteric enthalpy of adsorption approaching the bulk-phase sublimation heat of CO₂. The oxygen atoms in the CO₂ molecules possess lone pairs that are directed toward the more positively charged carbon atoms, which are situated at distances ranging between 2.6 Å and 3.1 Å. Similarly, the carboxylate groups of the TAMOF interact in the same way with the CO₂ molecules, as the oxygen atoms in the carboxylate groups exhibit lone pairs that are directed towards the carbon atoms in the CO₂ molecule, which are positioned at a distance of 3.1 Å.

It is important to note that these neutron diffraction data were obtained at 100 K. Room temperature data collection resulted in a disordered model, indicating that room temperature is enough to break the dominant interactions between CO₂ and TAMOF-1.

Spectroscopic analysis

To further investigate the interactions between CO₂ and the TAMOF-1 framework at room temperature, we collected IR spectroscopy data. The IR spectra of CO₂ when adsorbed on TAMOF-1 (Supplementary Fig. 4) show a band at 2335 cm^–1, which, according with previous reports^55,56, can be ascribed to physisorbed CO₂. The less intense band at 2324 cm^–1 has also been observed before and assigned to a combination band. These results suggest that there are not site-specific strong interactions and CO₂ only physically interacts with the framework.

To confirm the validity of these results, we also investigated CO adsorption for comparison. Supplementary Fig. 4b shows the C-O stretching region of carbon monoxide adsorbed on TAMOF-1. Interaction of the metal-organic framework with adsorbed CO resulted in the formation of an IR absorption band at 2139 cm^–1, which, according to literature, is assigned to (non-localized) physisorbed carbon monoxide⁵⁷, indicating that the metal-organic framework TAMOF-1 does not present accessible coordinatively unsaturated metal centers.

Computational analysis and adsorption mechanism

We calculated the single-component adsorption isotherms for CO₂, CH₄, and N₂ molecules in TAMOF-1 in the 263–333 K, and 1–10⁷Pa range. The good agreement between simulation and experiments validates our models (Fig. 5). Excess adsorption, following the protocol of Myers and Monson⁵⁸, allows comparison with the experiment. Some energy parameters related to carbon capture were also calculated from the single-component isotherms of CO₂ and N₂. We estimate that TAMOF-1 exhibits low parasitic energy, a volumetric working capacity of 23.46 kg m^–3, and a final CO₂ molar purity of the mixture of 0.904, which are excellent features for carbon capture. The calculated adsorption isotherms were fitted to a dual-site (for CO₂) and single site (for CH₄ and N₂) Langmuir-Freundlich model using the RUPTURA software⁵⁹ (solid lines in Fig. 5). Subsequently, we proceeded to compute the binary isotherms and determine the adsorption selectivity for CO₂:CH₄ and 14 CO₂: 84 N₂ mixtures (representing the composition of dry post-combustion flue gas). The resulting adsorption selectivities are depicted in Fig. 5d. In all cases, the adsorption of CO₂ surpasses that of CH₄ or N₂, primarily due to the stronger interactions between CO₂ molecules and the adsorption surfaces.

Fig. 5: Computational modeling of gas-TAMOF-1 thermodynamic properties.

a–c Calculated excess adsorption isotherm (empty circles) and experimental isotherms (solid points, see legend for more details) at 263.15, 303.15, 313, 323, and 333 K. Errors are estimated from the standard deviation in the adsorption processes. d, Selectivity ${s}_{j}^{i}={q}^{i}/\,{q}^{j},$ where i is CO₂, and j is CH₄ or N₂.

Simulation shows that at low values of pressure (1–500 Pa) the molecules of carbon dioxide are located in the narrow channels formed by the triad of Cu atoms. There is a dominant interaction between CO₂ and two Cu atoms (as shown in Fig. 4 and discussed in the SI, neutron diffraction analysis section), resulting in high values for the heat of adsorption, 50 kJ mol^–1, in agreement with experimental results (${Q}_{{{{\rm{st}}}}}\, \sim -\left\langle {U}_{{hg}}\right\rangle+{RT}$ where $\left\langle {U}_{{hg}}\right\rangle$ is the mean interaction energy between the CO₂ and the TAMOF-1 at very low pressure). However, as the CO₂ pressure increases, the CO₂-CO₂ interactions become strong enough to displace some of the adsorbed CO₂ molecules from these energetically favorable binding sites. This can be seen from radial distribution functions between CO₂-CO₂ and CO₂-Cu atoms, displayed in Fig. 6.

Fig. 6: Computational modeling of CO₂-TAMOF-1 interactions.

a Histogram for the CO₂-TAMOF-1 interaction energy at several values of pressure. b Radial distribution function for (intermolecular) pairs, and atoms.

Figure 6a shows that the interaction energy between the CO₂ molecule and the adsorption surface remains stable at around – 45 kJ mol^–1 up to 3 kPa and progressively decreases with pressure to – 23 kJ mol^–1. Although this trend is qualitatively equivalent for CH₄ and N₂, the range of energies and pressures differs, and the effect on adsorption is neglected. For example, for CH₄, the interaction energy varies from – 25 kJ mol^–1 (low pressure) to – 12 kJ mol^–1 (high pressure).

To conclude, the adsorption selectivity of CO₂ over CH₄ or N₂ at intermediate pressures (Fig. 5d) is mainly explained by two causes: the electrostatic interactions between CO₂ and the square planar Cu sites, and the confinement of CO₂ inside the TAMOF-1 pocket-cages. The narrow pores formed by the triad of Cu atoms contribute to decrease this selectivity at high pressures.

The self-diffusivity coefficients (D) of CO₂ and CH₄ were calculated using MD simulations at various temperatures and adsorption loadings (Fig. 7). CO₂ consistently exhibits lower diffusivity than CH₄. At low loading (1 to 4 CO₂ molecules per unit cell), the molecules of CO₂ are confined in adsorption pockets formed by the triad Cu atoms, resulting in low diffusivity (i.e., D_S^CO2 ~ 10^–10 m s^–1). As pressure increases, the molecules of CO₂ diffuse from the pocket adsorption site to the main channel. This leads to a maximum diffusivity at ~ 3 mol kg^–1 of adsorption loading for all temperatures (e.g., at 333 K, the self-diffusivity is approximately D_S^CO2 ~ 10^–9 m·s^–1). Beyond this loading (three molecules per Cu triad), the molecules of CO₂ are mainly adsorbed in the (chiral) channel, preventing transport and causing a progressively decreasing of diffusivity. In contrast to CO₂, for CH₄ molecules, self-diffusivity reaches its peak at infinite dilution for all temperatures (see Fig. 7b). The diffusion of CH₄ decreases with the increase on adsorption. As an example, for 333 K and for 1 molecule per unit cell, the self-diffusivity coefficient is D_S^CH4 ~ 6 × 10^–9 m s^–1.

Fig. 7: Gas diffusion properties modeling.

Self-diffusivity coefficient (D) for CO₂ (a) and CH₄ (b) at various temperatures and adsorption loadings. Solid lines are fitted to the data as guiding references to assist the reader (u.c. = unit cell).

Breakthrough curves for CO₂ and CH₄ were computationally predicted to validate the experiments in a column bed length of 5.8 cm and a gas pressure of 1.05–1.3 bar at 298 K (Fig. 8). The fitted Langmuir-Freundlich parameters were used to predict this adsorption dynamic behavior. Helium was used as a carrier gas to ensure high gas velocity. Three distinct gas compositions were investigated: 95% He and 5% of a CO₂/CH₄ mixture with ratios of 1.5:3.5, 2.5:2.5, and 3.5:1.5 (v/v). Initially, the column was fully saturated with helium. The void fraction of the adsorbent bed was 0.3. Simulation data are in agreement with the experimental results (Fig. 8). The simulated elution gas behavior remains qualitatively and quantitatively consistent with experimental data across the range of gas mixture compositions investigated (3.75%, 2.5%, and 1.25% v/v CH₄). Although a direct comparison with data reported in Fig. 2 is not accurate, the rapid elution of CH₄ observed in this case may be attributed to its higher diffusivity at higher gas velocity (0.015 m s⁻¹) and to its low gas partial pressure (Fig. 7b).

Fig. 8: Experimental and computational breakthrough analysis.

Calculated (lines) and experimental (square) breakthrough curves for CO₂ and CH₄ at 1 bar and 298.15 K for three composition ratios: 25 CO₂: 75 CH₄ (a), 50 CO₂: 50 CH₄ (b), and 75 CO₂: 25 CH₄ (c). Experiments performed with helium as the carrier gas (95% v/v). Experiments performed using 7.7 g of activated TAMOF-1. TAMOF-1 was activated under vacuum (10⁻¹mbar) in both column top and bottom sides at 393 K for 15 h.

In conclusion, TAMOF-1, a stable, robust, homochiral metal-organic framework (BET specific surface area = 980 ± 50 m² g^–1), has shown selective adsorption of CO₂ from CO₂/CH₄ gas streams, offering a promising performance. Although comparing breakthrough data across studies is often challenging due to inconsistencies in experimental details, the performance parameters exhibited by TAMOF-1 appear competitive for CO₂/CH₄ separations when compared with state-of-the-art (Fig. 2d and Supplementary Table 22). As demonstrated by structural and theoretical data, CO₂ uptake is dominated by a variety of weak interactions. At low coverage, the preferred binding sites are the square planar Cu²⁺ centers. Once these positions are full, the TAMOF-1 open channels accommodate additional CO₂ molecules with weaker interactions with the dangling functional groups (carboxylate, triazole, imidazole) in the framework. Adsorption isotherms confirm the CO₂ preferential uptake with respect to CH₄ or N₂ in a large pressure range, resulting in good separation parameters for CO₂/CH₄ gas mixtures. CO₂/CH₄ mixtures can be effectively separated with high selectivity in a wide CO₂/CH₄ concentration (1–75%), pressure (1–6 bar), and temperature range (293–353 K). Elution behavior of CO₂ through the TAMOF-1 bed depends just on the CO₂ molar flow and it is not affected by the presence of the other gases investigated (CH₄ and/or N₂).

TAMOF-1 retains separation performance, even under high humid conditions (~ 72% RH) showing no degradation and high cycling stability. Separation capability is observed also in extremely acidic conditions (9000 ppm H₂S) although a small performance degradation is observed. These data demonstrate TAMOF-1 robustness, also in harsh acidic environments although H₂S removal is preferred to increase material durability.

These properties are crucial for industrial-scale CO₂ capture applications. The weak CO₂–TAMOF-1 interactions facilitate facile, low-energy regeneration of the adsorbent bed. These attributes, coupled with high selectivity, render TAMOF-1’s performance in biogas upgrading superior to other reported materials⁴². Consequently, TAMOF-1 appears well-suited for integration into pressure/vacuum/thermal swing adsorption processes. A convenient regeneration protocol, adaptable to specific time and energy constraints through operational parameter tuning, is readily achievable. Current efforts are focused on scaling and pelletization of TAMOF-1, advancing the development of a CO₂ capture and purification technology based on this promising material.

Methods

Materials

All reagents were of commercial grade and used without further purification: L-histidine (≥ 98%, Iris Biotech GmbH), thionyl chloride (SOCl₂, ≥ 99%, Sigma Aldrich), hydrazine monohydrate (NH₂NH₂·H₂O, reagent grade, 98%, Sigma Aldrich), sodium carbonate anhydrous (Na₂CO₃, ACS reagent, ≥ 99.5%, Sigma Aldrich). All solvents were of commercial grade and used without further purification: HPLC-grade ethanol, isopropanol, hexanes, acetonitrile and tert-butyl methyl ether (VWR, Chem-Lab and Sigma Aldrich), N,N-dimethylformamide (peptide grade, ≥ 99.9%, Iris Biotech GMBH) and diethyl ether (≥ 99%, VWR).

Synthesis

(E)-N’-((E)-(dimethyl-amino)methylene)-N,N-dimethylformohydrazonamide (2)

This reagent was prepared following a two-step synthesis:

Part I: In a 250 mL 3-neck round bottom flask (RBF), 150 mL DMF was added and cooled down in an ice-bath. Afterwards, 28.6 mL SOCl₂ (0.39 mol) was slowly added under stirring with a syringe (dropwise). The reaction mixture was stirred for 24 h. After this time, the RBF was introduced in an ice bath, and once the mixture was cold, 5 mL NH₂NH₂·H₂O (0.1 mol) dissolved in 15 mL DMF was added very slowly, under stirring, with a syringe (dropwise and carefully, CAUTION: very exothermic reaction!!). The reaction mixture was stirred at room temperature for 48 h. The precipitate was then filtered and washed with DMF (20 × 3 mL) and Et₂O (10 × 3 mL) to yield 20.5 g (0.095 mol, 95% yield) of (CH₃)₂N(CH) = N-N = (CH)N(CH₃)₂·2HCl (1).

Part II: 20.5 g of 1 (0.095 mol) was weighted and dissolved in 100 mL H₂O. In a different beaker, 10.6 g of Na₂CO₃ (0.1 mol) was dissolved in 100 mL H₂O. Both water solutions were then mixed (the Na₂CO₃ on top of the azine solution), and the desired product (CH₃)₂N(CH) = N-N = (CH)N(CH₃)₂ (2) was purified by continuous extraction in 750 mL Et₂O, at 50 °C during 2 days. The organic layer was concentrated in vacuo, and 12.95 g of 2 (0.09 mol, 96% yield) was obtained as an orange solid which was used without any further purification.

Synthesis of (S)-3-(1H-imidazol-5-yl)-2-(4H-1,2,4-triazol-4-yl)propanoic acid

2 (6.3 g, 0.044 mol) and L-histidine (3.1 g, 0.02 mol) were mixed together in 150 mL EtOH. The mixture was refluxed with stirring for 48 h. The final solution was filtered off from unreacted L-histidine, and the solvent was removed under reduced pressure to afford an orange gel that was washed with EtOH to yield a white precipitate of S-HTA, that was collected by filtration, washed with EtOH (15 × 3 mL), Et₂O (15 × 3 mL) and dried in air. Yield: 3.5 g (0.017 mol, 85%).

Synthesis of [Cu(H₂O)₂(C₈H₈N₅O₂)₂].6H₂O (TAMOF-1)

Cu^II (CH₃COO)₂·H₂O (0.48 g, 2.4 mmol) was dissolved in 50 mL of H₂O, and the resulting solution was slowly added to a 100 mL aqueous solution of (S)-3-(1H-imidazol-5-yl)-2-(4H-1,2,4-triazol-4-yl)propanoic acid (1.0 g, 4.8 mmol). The solution was stirred for 30 min, allowing for the appearance of a precipitate. This blue polycrystalline powder (TAMOF-1) was filtered with a Nylon membrane filter (pore size: 0.45 μm, Filter-Lab), washed with H₂O (50 × 3 mL), and dried in air. Yield: 1.1 g (70%). The structure and purity of the samples was confirmed by elemental analysis and powder X-ray diffraction (Supplementary Fig. 1). Anal. Calcd for C₁₆H₃₂CuN₁₀O₁₂: C 30.99, H 5.2, N 22.59; found: C 30.89, H 4.93, N 22.40. IR (ATR): ν = 3289, 1615, 1532, 1495, 1395, 1355, 1263, 1242, 1211, 1180, 1116, 1087, 1015, 979, 891, 845, 752, 701, 660, 647, 549, 511, 472 cm⁻¹. TAMOF-1 thermal stability up to ≈ 473 K was confirmed by thermogravimetric analysis (TGA) performed using a TGA/SDTA851 Mettler in an air atmosphere (Supplementary Fig. 2). Single crystals of ≈ 1 mm³ dimensions, suitable for neutron diffraction experiments, were obtained by slow diffusion of the aqueous solution of Cu(OAc)₂ and the aqueous solution of (S)-3-(1Himidazol-5-yl)-2-(4H-1,2,4-triazol-4-yl)propanoic acid in an H-Cell with the two compartments separated by a glass frit.

Single crystal neutron diffraction

Single-crystal neutron diffraction data was collected at 100 K on the four-circle diffractometer D19 at the Institut Laue-Langevin (ILL), Grenoble. The instrument was equipped with an Oxford Cryostream 700 Plus low-temperature device and was operated using a neutron wavelength of 1.4556(1) Å, provided by a flat Cu(220) monochromator (at 2θ_M = 69.91° take-off angle). A single crystal was selected from a batch of crystals of TAMOF-1 using an optical microscope. The selected specimen, with dimensions of approximately 1 × 0.8 × 0.6 mm³, was found to be suitable for single crystal neutron diffraction measurements using D19. The crystal was glued to a vanadium pin of 0.5 mm diameter and mounted on a standard goniometer head. Three short ω-scans were collected at room temperature (RT) to check the sample quality and obtain a preliminary orientation matrix. This yielded a cubic unit cell with a unique axis length of 20.19 (6) Å, further confirming that the TAMOF-1 structure remained intact. The cubic cell length determined by X-ray diffraction at T = 100 K was 20.120 (4) Å.

Two heating cycles were performed in order to activate (dehydrate) the crystal. The first cycle consisted of heating to T = 400 K (0.1 K min^–1) in a stream of nitrogen gas, after which the crystal was kept at 400 K for slightly less than 30 min. The temperature of the sample, still in a stream of nitrogen gas, was lowered at a rate of 1 K min^–1 to room temperature. Once at room temperature, the crystal was placed in an atmosphere of CO₂ for 2.5 h, after which it was placed in the nitrogen gas stream on the diffractometer; and the temperature was lowered to T = 100 K for diffraction measurements (three omega scans of 80° each, with a step size of 0.07° and counting time of 25 sec/step). Structure analysis revealed that there was still water in the cavities of the MOF, so the activation cycle was repeated.

For the second dehydration cycle, the sample temperature was first raised to 400 K (ramp rate 0.1 K min^–1). The crystal was kept at T = 400 K for 70 min, during which time a color change from turquoise to black was observed, but without loss of crystallinity. The nitrogen stream was then used to lower the sample temperature to 100 K. Once this temperature had been reached, the crystal was exposed to a current of CO₂ for five minutes, while the N₂ stream was simultaneously cut off using a purpose-built, one-use valve that permitted one of the two gases to impinge on the sample while blocking the other. At the end of the CO₂ exposure, the flow was switched back to the N₂ stream at T = 100 K.

Full data collection ensued, consisting of three omega scans, with an exposure time of approximately 45 s per image (5 × 106 monitor counts). The resulting data revealed that CO₂ alone was present in the cavities of the TAMOF-1 sample, vide infra.

A regeneration step was then performed to evacuate the CO₂ from the sample. This consisted of heating the sample once again to T = 400 K for 70 min, and then lowering the temperature to 100 K for data collection, which again consisted of 3 omega scans with a count time of approximately 45 s per image (5 × 106 monitor counts). Data analysis confirmed that all of the CO₂ had been evacuated from the sample.

For full data collection, the measurement strategy consisted of a series of five ω-scans with steps of 0.07° at different χ and φ positions. The acquisition time was ca. 45 s per step, which corresponds to 500000 counts on the instrument monitor. The data collection was performed using the instrument control program NOMAD from ILL. We collected additional reciprocal space where no reflections were expected. We were able to do this thanks to the wide-angle solid-state detector at the D19 line in the Institut Laue-Langevin (ILL). We should have been able to detect any crystalline phase transition (by breaking of symmetry or appearance of domains) under the CO₂ atmosphere. Indexing and unit-cell determination were done using the programs PFIND and DIRAX⁶¹. The integration of the raw data and refinement of the UB-matrix, including the offsets, were done using the programs RETREAT and RAFD19, respectively⁶²; the Lorentz correction was applied to the intensities. Structure refinement was conducted using the program ShelxL2019/3⁶⁰. Neutron scattering lengths for structure analysis were taken from the Neutron Data Booklet, ILL Grenoble (2003) ISBN 0-9704134-7-4.

When the data collection was completed, the resulting structure model confirmed the cubic symmetry of the sample, in agreement with the preliminary X-ray diffraction data. The analysis also confirmed that the MOF was activated, with empty channels potentially capable of hosting guest molecules. We measured continuous sections of reciprocal space -- i.e., not just the expected positions for Bragg peaks -- using the wide angle 2D detector at the D19 beamline at the Institut Laue-Langevin (ILL). We are confident that we would have detected any crystalline phase transition (by breaking of symmetry, or appearance of domains) under the CO₂ atmosphere. No deviation from the cubic symmetry was detected (Supplementary Tables 1 and 8). The average I/σ(I) for all merged reflections is 20.2. Any reflection violating the systematic absence conditions for the space group (P4₃32) was indistinguishable from the background. CCDC 2356946 and 2356947 contain the supplementary crystallographic data for this paper. These data can be obtained free of charge from The Cambridge Crystallographic Data Center via www.ccdc.cam.ac.uk/data_request/cif. Crystal structure representations were generated with OLEX2⁶³, Diamond⁶⁴, and Mercury 4.0⁶⁵ visualization programs.

Infrared spectroscopy

Thin self-supported wafers of the TAMOF-1 were prepared and activated (outgassed) in a dynamic vacuum (residual pressure smaller than 10^–4 mbar) at 423 K for 4 h inside a 2000-A-C AABSPEC cell that allowed us to perform: (i) in situ sample activation, (ii) gas dosage and (iii) variable-temperature IR spectroscopy of adsorbed molecules while simultaneously recording temperature and equilibrium pressure. A K-thermocouple (in contact with the sample wafer) connected to a digital thermometer (TEMP, XS Instruments) and a capacitance pressure gauge (MKS, Baratron) were used for that purpose; the precision of the measurements was about ± 2 K and ± 2 × 10^–2 mbar for temperature and pressure, respectively.

After thermal activation of the sample wafer, the cell was dosed with the probe molecule and closed, and transmission FTIR spectra were recorded while simultaneously registering the spectrum. Transmission FTIR spectra were collected, at 3 cm^–1 resolution on a Bruker Vertex 80 v spectrometer equipped with an MCT cryo-detector (Supplementary Fig. 4).

Gas isotherms

Adsorption isotherms of carbon dioxide, methane, and nitrogen up to 10 bar at different temperatures were carried out with iSorb HP1 200 (Anton Paar). Prior to analysis, the sample was degassed and dehydrated (activated) for 18 hours at 423 K, 10^–6 Torr.

Breakthrough measurements

For the gas breakthrough experiments, different tubes loaded with different TAMOF-1 amounts were used as separation module. Noteworthy, the experiments gave consistent per-mass results. Two different columns were used: Column 1: 0.7 g TAMOF-1 (dehydrated), bed length: 1 cm; Column 2: 7.7 g TAMOF-1 (dehydrated), bed length: 18.3 cm. Scaling the system (about 18 times) did not affect TAMOF-1 performance. In both, TAMOF-1 was packed within the module with glass wool support at the top and the bottom of the column. TAMOF-1 powder has the following properties: random particle size between 0.2 and 10 μm⁴⁴, bulk density = 0.689 g cm⁻³ (hydrated TAMOF-1) and 0.530 (dehydrated TAMOF-1), and porosity = 0.598.

Prior to breakthrough measurements, TAMOF-1 was activated in situ for water removal at 393 K for a 15 h vacuum (10⁻¹mbar) in both column top and bottom sides or a nitrogen flow (170 ml min^–1 g^–1_(TAMOF)). Breakthrough data were collected by using the experimental set-up shown in Supplementary Figs. 5, 6. Gas cylinders of CH₄ (Linde, 99.998%), CO₂ (Air Liquide, 99.998%), CO₂/CH₄ mixture with composition of 50.3/49.7% (Messer) and N₂ (Praxair, 99.999%) were used. The flows/compositions of the inlet gas stream were controlled upstream of the separation module by a set of calibrated mass flow controllers (Bronkhorst EL-FLOW). A manometer and a backpressure controller (Bronkhorst, EL-PRESS) were placed downstream of the separation module. A manometer was also used upstream of the bed to monitor the actual pressure at the bed inlet under no pressure control conditions. Unless otherwise stated, all pressure values are reported in units of absolute bar (bar). The separation module was heated by a linear power silicone heating wire (Ø 3 mm FOR-FLEX NORMAL, Electricfor) rolled around the column, the temperature was measured with a K-type thermocouple (Thermocoax) placed in between the module and the silicon heating wire, and in contact with both. It was controlled with a temperature controller EZ-Zone (Watlow). The outlet stream of the column was on-line analyzed by micro gas chromatograph (MGC, Agilent MicroGC 490) equipped with Molsieve MS5A, using Ar as carrier gas (99.999 % purity), and Poraplot U column, using He as carrier gas (99.999 % purity), along with thermal conductivity detectors (TCD).

In the experiments performed with a partial flow of N₂ in the inlet gas mixture lower than 25 Nml min^–1, a nitrogen flow of 30 Nml min^–1 was mixed with the gas coming out from the separation module in order to meet the flow requirements of the GC. In this case, a non-return valve with negligible pressure drop was used to avoid reverse flow. The transit time or death time of the gas in the set-up was also evaluated and subtracted at each operation condition.

Regeneration assessment of TAMOF-1 (Fig. 3) was conducted flowing N₂ sweep gas within the column (0.7 g). Several regeneration conditions were evaluated as: a) different concentrations of equimolecular CO₂–CH₄ mixtures, b) different temperatures, c) different flow rates, and d) different pressures. In all cases, the composition of the gas exiting the column was analyzed using the GC.

Gas adsorption experiments under humid conditions were conducted using a custom-built glass bubbler containing a known volume of water. A CO₂/CH₄ (50:50) gas mixture at a flow rate of 30 ml min⁻¹ was passed through the bubbler and subsequently fed into the adsorption column. Temperature control was achieved using a chiller with ethylene glycol as the cooling fluid. A K-type thermocouple (Thermocoax) placed within the bubbler provided accurate temperature measurements. To prevent condensation within the analyzer, a nitrogen flow of 120 Nml min⁻¹ was used to dilute the gas exiting the separation column. Effluent gas composition (m/z) was continuously monitored via a quadrupole mass spectrometer (Pfeiffer OmniStar GSD 301 C1). Following CO₂ saturation of the TAMOF-1 column, regeneration was performed under vacuum (10⁻² mbar) at 80 °C during 15 h. This adsorption–regeneration cycle was repeated 5 times to assess the stability of the material.

For the evaluation of the stability of TAMOF-1 under dynamic acid conditions, H₂S was generated in situ by carbonation conversion of aqueous sodium sulfide⁶⁶: NaHS_(aq) + CO_2(aq) + H₂0 $\leftrightarrows {\mbox{H}}_2{\mbox{S}}({\mbox{g}})\,+{\mbox{NaHCO}}_3({\mbox{aq}})$. Experiments were conducted at 28 °C (room temperature) and 1 bar. The H₂S concentration evaluated was 1% (v/v) (0.27 ml min⁻¹), corresponding to a water concentration of 2.3% (v/v) (0.725 ml min⁻¹). A total flow rate of 29 ml min⁻¹ was maintained, consisting of 16 ml min⁻¹ of a CO₂/CH₄ (50:50 v/v) gas mixture and the balance N₂. Prior to the initial cycle, the MOF was activated. Between subsequent cycles, the MOF was purged with 250 ml min⁻¹ of N₂ for at least 15 h.

Computational methods

Monte Carlo (MC) and Molecular Dynamics (MD) simulations were used to investigate the sorption and transport properties of carbon dioxide, methane, and nitrogen molecules both individually and as mixtures within the TAMOF-1 material. Our theoretical analysis focused on discerning the mechanisms that explain the separation process of mixtures of interest from a molecular perspective. In addition to sorption and transport properties, we also calculated the breakthrough curves. To conduct these simulations, we performed simulations at various temperatures, from 263.15 K to 333.15 K. We used a cubic supercell with dimensions of 2 × 2 × 2 for TAMOF-1. The initial relaxation of the atomic coordinates was achieved using an extended semi-empirical tight-binding model known as GFN2-xTB⁶⁷, which was performed using DFTB + code (v. 22.2)⁶⁸.

Our CO₂ model comprises three Lennard-Jones sites with charges centered on each atom. The carbon-oxygen bonds remain fixed, while the Lennard-Jones parameters and the molecular structure of CO₂ were derived from research conducted by García-Sánchez et al.⁶⁹. For the N₂ molecule, we adopted the model proposed by Potoff and Siepmann⁷⁰, which incorporates three charged sites and two Lennard-Jones sites centered on each N atom. Regarding the CH₄ molecule, we used an uncharged united atom model⁷¹, considering the CH₄ group as a single interaction center with its own effective potential. We assumed the TAMOF-1 framework to be rigid during both the MC and MD simulations. To accurately capture the interactions between the adsorbates and framework, we employed van der Waals interactions. Furthermore, Coulombic interactions were considered for the CO₂ and N₂ molecules. The Ewald method was used in these cases. The Lennard-Jones parameters were derived from the well-established DREIDING potential⁷², except for the Cu atoms, for which we adopted parameters from the UFF potential^73,74. The mixed Lennard-Jones parameters were calculated using Lorentz-Berthelot mixing rules.

To perform the MC and MD simulations, we used the reliable RASPA code⁷⁵ and followed the established protocols. This protocol has been used successfully for other Cu-paddlewheel MOF (Cu-BTC)⁷⁶. The adsorption isotherms were determined through MC simulations using a Grand Canonical ensemble (μVT). The pressure values were converted to fugacity using the Peng-Robinson equation of state. In our simulations, we conducted a minimum of 10⁶ MC cycles. Within each MC cycle, the following trial MC moves were randomly selected for a given molecule: translation, rotation, reinsertion at random positions, and swapping (insertion or deletion). Binary isotherms and selectivities were calculated using IAST using the RUPTURA code⁵⁹.

For the MD simulations, we used a time step of 0.5 fs to carry out 10⁸ cycles of simulation. During the MD simulation, the radial distribution functions and atomic mean squared displacements were calculated. To study the transport properties of CO₂ and CH₄, we equilibrated the systems using a canonical ensemble (NVT). Subsequently, using the microcanonical ensemble (NVE), we measured the calculated self-diffusion coefficients for CO₂ and CH₄ from the measured mean-squared displacement at several concentrations and temperatures. Finally, we evaluated the TAMOF-1 features for carbon capture from the calculated sorption properties, following the work of Huck et al.⁷⁷ and Ongari et al. by calculating energy parameters⁷⁸.