Breaking the trade-off between selectivity and adsorption capacity for gas separation

Summary

The trade-off between selectivity and adsorption capacity with porous materials is a major roadblock to reducing the energy footprint of gas separation technologies. To address this matter, we report herein a systematic crystal engineering study of C₂H₂ removal from CO₂ in a family of hybrid ultramicroporous materials (HUMs). The HUMs are composed of the same organic linker ligand, 4-(3,5-dimethyl-1H-pyrazol-4-yl)pyridine, pypz, three inorganic pillar ligands, and two metal cations, thereby affording six isostructural pcu topology HUMs. All six HUMs exhibited strong binding sites for C₂H₂ and weaker affinity for CO₂. The tuning of pore size and chemistry enabled by crystal engineering resulted in benchmark C₂H₂/CO₂ separation performance. Fixed-bed dynamic column breakthrough experiments for an equimolar (v/v = 1:1) C₂H₂/CO₂ binary gas mixture revealed that one sorbent, SIFSIX-21-Ni, was the first C₂H₂ selective sorbent that combines exceptional separation selectivity (27.7) with high adsorption capacity (4 mmol·g⁻¹).

Graphical abstract

Highlights

• •Six isostructural hybrid ultramicroporous materials are prepared and characterized
• •Crystal engineering approach enabled fine-tuning of pore size and chemistry
• •Weak CO₂/strong C₂H₂ affinity resulted in high C₂H₂/CO₂ separation selectivities
• •SIFSIX-21-Ni: benchmark selectivity/uptake capacity for C₂H₂/CO₂ separation

The bigger picture

It is generally recognized that porous solids (sorbents) with high selectivity and high adsorption capacity offer potential for energy-efficient gas separations. Unfortunately, there is generally a trade-off between capacity and selectivity, which represents a roadblock to the utility of sorbents in key industrial processes. For example, acetylene (C₂H₂), an important fuel and chemical intermediate, is produced with CO₂ as an impurity, and the similar physicochemical properties of C₂H₂ and CO₂ mean that most sorbents are poorly selective. Hybrid ultramicroporous materials (HUMs) are candidates for gas separations as they exhibit benchmark selectivity for several key gas pairs. Unfortunately, existing HUMs are handicapped by low capacity. We report a new HUM, SIFSIX-21-Ni, that addresses the trade-off between selectivity and capacity that has plagued sorbents, as its high uptake and high selectivity renders it the new benchmark for C₂H₂/CO₂ separation performance.

A new family of six isostructural hybrid ultramicroporous materials that enabled high adsorption capacity as well as high experimental separation selectivity values for C₂H₂/CO₂ separation.

Introduction

Acetylene (C₂H₂) is an important chemical commodity; it is used to manufacture vinyl and acrylate polymers and is a combustion fuel in oxy-acetylene torches.^1,2 The latter application stems from its flammability range, the widest known (2.5%–81%), but can thereby represent an explosion hazard at >2.5% concentrations.³ Whereas the utility of C₂H₂ as an oxy-combustion fuel typically requires >98% purity grade for its use as a chemical feedstock, a higher purity grade is essential.⁴ Bulk C₂H₂ is produced by either oxidative coupling (partial combustion) of methane or downstream thermal cracking of hydrocarbons; CO₂ is a by-product of both processes.⁵ C₂H₂ production generates CO₂ as an impurity, which means that selective separation/purification of high-purity (>99% v/v) C₂H₂ from C₂H₂/CO₂ mixtures is of industrial relevance.⁶ Three technologies are currently used to remove C₂H₂ from C₂H₂/CO₂ mixtures: (1) bulk solvent extraction, resulting in solvent waste such as N,N-dimethylformamide and acetone⁷; (2) partial hydrogenation of C₂H₂ to ethylene, C₂H₄, with costly noble-metal catalysts such as Ag(0)⁸; and (3) cryogenic distillation, an energy-intensive process.⁹ All three processes suffer from high cost and low efficiency. Physisorbents offer potential for energy-efficient gas purification as exemplified by recent reports on physisorbents that exhibit benchmark performance for key separations such as CO₂/N₂,^10,11 CO₂/CH₄,¹² C₂H₂/C₂H₄,^13,14 and C₂H₆/C₂H₄,¹⁵ among others.¹⁶ Nevertheless, commercially viable C₂H₂ capture from CO₂ using physisorbents remains an unmet challenge since traditional physisorbents, such as zeolites, mesoporous silicas, and activated carbons, exhibit poor selectivity for C₂H₂ over CO₂¹⁷ thanks to their similar physicochemical properties (molecular dimensions: C₂H₂ = 3.32 × 3.34 × 5.7 Å³; CO₂ = 3.18 × 3.33 × 5.36 Å³; kinetic diameters for both molecules = 3.3 Å; boiling points: C₂H₂ = 189.3 K, CO₂ = 194.7 K).^18,19 These physicochemical properties also practically rule out the application of molecular sieving.²⁰

In this context, metal-organic materials (MOMs),²¹ also known as metal-organic frameworks (MOFs)^22,23 or porous coordination polymers (PCPs),²⁴ have emerged as sorbent candidates to serve as C₂H₂ selective physisorbents in C₂H₂/CO₂ separation.²⁵ Unlike traditional classes of sorbents, the modularity of MOMs enables fine-tuning of pore size and pore chemistry using crystal engineering design approaches.²¹ Nevertheless, there are >100,000 MOMs in the CSD MOF subset (2020.3 CSD release).²⁶ To the best of our knowledge, only a few (20) have been experimentally studied for C₂H₂/CO₂ separation under dynamic conditions (e.g., dynamic column breakthrough [DCB] experiments, see Table S1) including recently reported benchmarks set by ATC-Cu,²⁷ FJI-H8-Me,²⁸ and SIFSIX-Cu-TPA.²⁹ For example, ATC-Cu was found to exhibit an ideal adsorbed solution theory (IAST) selectivity of 53.6 for 1:1 C₂H₂/CO₂ separation.²⁷ The need to study sorbent performances under dynamic conditions arises because calculation of separation performance from C₂H₂ and CO₂ single-component isotherms tends to overestimate separation performance. IAST and fixed-bed simulated breakthrough calculations are, therefore, indicative rather than confirmative. Our review of the literature reveals that C₂H₂/CO₂ separation selectivity (α_AC) values for equimolar (v/v) mixtures have been experimentally measured for only 16 sorbents, most of which are classified as MOFs: SIFSIX-Cu-TPA,²⁹ TCuI,³⁰ TCuBr,³⁰ TCuCl,³⁰ JCM-1,³¹ NKMOF-1-Ni,³² FJU-22a,³³ FJU-89a,³⁴ HOF-3,³⁵ FJU-6-TATB,³⁶ SSNU-45,³⁷ JXNU-5,³⁸ FJU-36a,³⁹ FeNiM′MOF,⁴⁰ UTSA-74a,⁴¹ and sql-16-Cu-NO₃-α'.⁴² Whereas these results are promising, they do not fully address the spectrum of performance parameters needed before a material can be considered for commercialization.⁴³ In particular, in addition to selectivity, working capacity and recyclability (including the kinetics and energy of regeneration) are also key performance parameters.^25,44 As revealed in Table S1, UTSA-74a is the only sorbent that offers separation selectivity ≥5 and adsorption capacity ≥3.5 mmolg⁻¹. Unfortunately, UTSA-74a requires a high temperature for sorbent regeneration and activation (473 K under high vacuum) to generate the unsaturated metal centers (UMCs) that drive preferential binding for C₂H₂ over CO₂.

Hybrid ultramicroporous materials (HUMs), sorbents that exhibit <0.7 nm pore diameter and are comprised of both organic and inorganic linker ligands, are an emerging subclass of MOMs.⁴⁵ HUMs are outstanding candidates for physisorptive separation of small gas sorbates as they offer benchmark adsorption selectivities for C₂H₂ over C₂H₄ (S_AE)^14,18 and CO₂ over N₂ (S_CN).^16,19,46 This performance has been attributed to two key features that enhance selective binding: (1) narrow pore sizes and (2) pore surfaces/walls offering strong electrostatics. HUMs are also modular, which enables crystal engineering-driven control of pore size and chemistry to facilitate removal of even trace impurities. Unfortunately, narrow pore size, as seen in the prototypal pyrazine-linked HUMs, tends to limit surface area and uptake capacity. We recently reported that an HUM based on an expanded symmetrical ligand, 3,3′,5,5′-tetramethyl-1H,1′H-4,4′-bipyrazole, pzpz, offered a highly selective CO₂ binding site that was also hydrophobic.¹¹ Six members of a new HUM platform based on an unsymmetrical ligand that is related to pzpz, 4-(3,5-dimethyl-1H-pyrazol-4-yl)pyridine (pypz) are introduced herein. The six HUMs were obtained by inorganic ligand (SiF₆^2-, SIFSIX; TiF₆^2-, TIFSIX; and NbOF₅^2-, NbOFFIVE) and/or metal (Ni²⁺ or Cu²⁺) substitution (Figure 1A) and offer higher gravimetric surface areas (and, therefore, potentially higher working capacity) than the corresponding pyrazine-linked HUMs.⁴⁷ Herein, we report the exceptional adsorptive separation performance for C₂H₂ versus CO₂ of these HUMs as evaluated by single-component gas sorption measurements, dynamic gas breakthrough experiments, in situ IR studies, solid-state NMR experiments, and molecular modeling.

Figure 1.

The family of isostructural HUMs reported herein and their single-crystal X-ray structures, single-component adsorption isotherms, and isosteric heats of adsorption

(A) Schematic illustration of the building blocks and pcu network topology of M′FSIX-pypz-M and NbOFFIVE-pypz-M.

(B) C₂H₂ binding site in SIFSIX-21-Ni viewed across diagonally opposite F atoms of SiF₆^2- pillars.

(C) The ultramicropore in SIFSIX-21-Ni viewed along the crystallographic b axis.

(D) C₂H₂ and CO₂ isotherms of SIFSIX-21-Ni, NbOFFIVE-3-Ni, and TIFSIX-4-Ni at 298 K.

(E) C₂H₂ and CO₂ isotherms of SIFSIX-21-Cu, NbOFFIVE-3-Cu, and TIFSIX-4-Cu at 298 K.

(F) Comparative bar diagram of isosteric heat of adsorptions (C₂H₂ and CO₂). (Color codes in Figures 1A–1C: C, gray; N, blue; Si, yellow; F, light green; Ni, cyan.)

Results and discussion

Synthesis and structural characterization

Single crystals of SIFSIX-21-Ni, TIFSIX-4-Ni, SIFSIX-21-Cu, and TIFSIX-4-Cu, and microcrystalline samples of NbOFFIVE-3-Ni and NbOFFIVE-3-Cu were prepared as detailed in the supplemental information. Single-crystal X-ray diffraction (SCXRD) studies revealed that SIFSIX-21-Ni, TIFSIX-4-Ni, SIFSIX-21-Cu, and TIFSIX-4-Cu are isostructural and crystallize as pcu topology networks in orthorhombic space group Pnna. The crystallographic data and refinement parameters for SIFSIX-21-Ni, TIFSIX-4-Ni, SIFSIX-21-Cu, and TIFSIX-4-Cu are presented in Table S2. The unit cell parameters calculated from powder X-ray diffraction (PXRD) for NbOFFIVE-3-Cu and NbOFFIVE-3-Ni are close to those of the M’FSIX analogs (Figures S3 and S4). Unit cell volumes are as follows: TIFSIX-4-Cu (3,458.8 Å³)> SIFSIX-21-Cu (3,361.1 Å³)> NbOFFIVE-3-Cu (3,338 Å³)> TIFSIX-4-Ni (3,291.7 Å³)> NbOFFIVE-3-Ni (3,236 Å³)> SIFSIX-21-Ni (3,199.2 Å³). Solvent-accessible free volumes were calculated to be ca. 30%. Polycrystalline samples of SIFSIX-21-Ni, TIFSIX-4-Ni, SIFSIX-21-Cu, and TIFSIX-4-Cu used for physicochemical characterization and sorption studies were prepared solvothermally in methanol (see supplemental information for details). Bulk phase purity was established by comparison of experimental and calculated PXRD patterns (Figures S1 and S2). Thermogravimetric analysis and variable temperature PXRD (VT-PXRD) experiments revealed each material retained crystallinity as follows: 613 K (NbOFFIVE-3-Ni) > 573 K (TIFSIX-4-Ni) > 513 K (SIFSIX-21-Ni); 533 K (TIFSIX-4-Cu) > 513 K (NbOFFIVE-3-Cu) > 493 K (SIFSIX-21-Cu) (Figures S5–S10 and S13–S14).

Single-component gas isotherms and binding sites

To evaluate microporosity, N₂ and CO₂ adsorption isotherms were measured at 77 K and 195 K, respectively (Figures S15, S17, S19, S21, S23, and S25). Following the Rouquerol criteria,⁴⁸ Brunauer-Emmett-Teller (BET) surface areas were experimentally determined from the N₂ adsorption isotherms as follows: TIFSIX-4-Ni (931 m²g⁻¹) > SIFSIX-21-Ni (871 m²g⁻¹) > SIFSIX-21-Cu (839 m²g⁻¹) > NbOFFIVE-3-Cu (805 m²g⁻¹) > NbOFFIVE-3-Ni (761 m²g⁻¹) > TIFSIX-4-Cu (747 m²g⁻¹). C₂H₂ and CO₂ single-component gas sorption isotherms were collected at 298 K and 273 K (Figures S16, S18, S20, S22, S24, and S26). All six HUMs were observed to exhibit higher affinity for C₂H₂ than CO₂ with C₂H₂ uptakes >3.5 mmolg⁻¹ and CO₂ uptakes <2.0 mmolg⁻¹ at 298 K and 1 bar (Figures 1D and 1E). SIFSIX-21-Ni was found to exhibit the highest C₂H₂ uptake (~4.05 mmolg⁻¹), followed by NbOFFIVE-3-Cu (3.97 mmolg⁻¹), SIFSIX-21-Cu (3.94 mmolg⁻¹), TIFSIX-4-Ni (3.85 mmolg⁻¹), NbOFFIVE-3-Ni (3.79 mmolg⁻¹), and TIFSIX-4-Cu (3.53 mmolg⁻¹) (Figures 1D and 1E). The stepped isotherms observed for SIFSIX-21-Ni (Figure S16) and SIFSIX-21-Cu (Figure S22) prompted us to conduct in situ PXRD measurements on these two HUMs. These measurements (Figures S11 and S12) reveal no significant change in the PXRD patterns with C₂H₂ loading until 1 bar indicated that only subtle structural changes occurred during C₂H₂ sorption. Isosteric heats of adsorption, Q_st, were determined from virial fits of these experimental single-component gas isotherms. Q_st values for CO₂ at zero loading, Q_st(CO₂), were as follows: 19.8 kJ mol⁻¹ (SIFSIX-21-Ni) < 24.0 kJ mol⁻¹ (SIFSIX-21-Cu) < 25.1 kJ mol⁻¹ (NbOFFIVE-3-Ni) < 25.8 kJ mol⁻¹ (NbOFFIVE-3-Cu) < 27.0 kJ mol⁻¹ (TIFSIX-4-Cu) < 27.5 kJ mol⁻¹ (TIFSIX-4-Ni). Q_st (C₂H₂) values were determined to be as follows: 36.3 kJ mol⁻¹ (SIFSIX-21-Cu) < 36.7 kJ mol⁻¹ (NbOFFIVE-3-Ni) < 37.9 kJ mol⁻¹ (SIFSIX-21-Ni) < 40.6 kJ mol⁻¹ (TIFSIX-4-Cu) < 41.3 kJ mol⁻¹ (TIFSIX-4-Ni) < 41.9 kJ mol⁻¹ (NbOFFIVE-3-Cu) (Figures 1F amd S27–S32; all virial fitting parameters are provided in Figures S33–S44). The differences between Q_st(C₂H₂) and Q_st(CO₂), (ΔQ_st)_AC = [Q_st(C₂H₂) − Q_st(CO₂)], were as follows: 18.1 kJ mol⁻¹ (SIFSIX-21-Ni) > 16.1 kJ mol⁻¹ (NbOFFIVE-3-Cu) > 13.9 kJ mol⁻¹ (TIFSIX-4-Ni) > 13.6 kJ mol⁻¹ (TIFSIX-4-Cu) > 12.3 kJ mol⁻¹ (SIFSIX-21-Cu) > 11.6 kJ mol⁻¹ (NbOFFIVE-3-Ni). These Q_st and ΔQ_st values reveal the relative thermodynamic preferences of the studied sorbents toward the competing sorbates, C₂H₂ and CO₂. Adsorption selectivities were calculated using IAST.⁴⁹ For C₂H₂/CO₂ (v/v: 1:1 and 2:1) at 1 bar and 298 K, IAST selectivities (S_AC) were calculated upon fitting the single-component isotherms to the dual-site Langmuir-Freundlich equation (see supplemental information for details; Figures S51, S52, and Table S3). S_AC (1:1/2:1) values at 1 bar were thereby determined to be 10.0/9.4 (SIFSIX-21-Cu) > 9.5/9.0 (NbOFFIVE-3-Cu) > 8.3/8.1 (TIFSIX-4-Cu) > 7.8/7.8 (SIFSIX-21-Ni) > 7.6/7.4 (TIFSIX-4-Ni) > 6.0/5.9 (NbOFFIVE-3-Ni) (Figures S45–S50). As presented in Table S1, under relevant partial pressures, S_AC for the six HUMs studied in this contribution were found to be comparable to leading C₂H₂-capture sorbents such as SIFSIX-Cu-TPA(5.3),²⁹ MUF-17(6),⁵⁰ ZJU-60a(6.7),⁵¹ FJU-22a(7.1),³³ TCuBr (9.5),³⁰ ZJUT-2a(10),⁵² TIFSIX-2-Cu-i(10),⁵³ FJI-H8-Me(10.4),²⁸ and MIL-100(Fe)(12.5).⁵⁴

The binding sites for CO₂ and C₂H₂ in SIFSIX-21-Ni were determined by simulated annealing calculations (Figure 2). The binding energies at 0.1 bar obtained from canonical Monte Carlo (CMC) were in agreement with the experimentally derived low loading Q_st(C₂H₂) and Q_st(CO₂) obtained from single-component isotherms (Table S4). Results derived by these two simulation methods revealed that C₂H₂ molecules interact with a pair of diagonally opposite F atoms of the inorganic pillars (SiF₆²⁻) via two (C₂H₂)CH^δ+···F^δ− interactions. There are also CH···C(C₂H₂) interactions. CO₂ molecules form C^δ+···F^δ- and CH···O(CO₂) interactions. To further understand these binding sites, density function theory (DFT) refinements were conducted on the strongest binding sites of SIFSIX-21-Ni as identified by CMC simulations in order to calculate the adsorption enthalpies at low loading (one adsorbate molecule per unit cell). Employing the BEEF-vdW functional,⁵⁵ adsorption enthalpies of −45.3 and −30.1 kJ mol⁻¹ were calculated for C₂H₂ and CO₂, respectively. The adsorption enthalpy difference of 15 kJ mol⁻¹ correlates well with experimental ΔQ_st values (Figure 1F). For the BEEF-vdW optimized structures (see supplemental information), the optimized binding pocket with adsorbed C₂H₂ has two short F^δ-···H^δ+ interactions (2.15 Å each, Figure S53A). For CO₂, F^δ-···C^δ+ interactions of 3.69 Å and CH···O(CO₂) interactions of 2.95 and 3.34 Å (Figure S53B) were determined. For CO₂ adsorption, an alternate binding site with similar adsorption enthalpies (± 1 kJ mol⁻¹) with shorter F^δ-···C^δ+ interactions of 3.10 Å and CH···O(CO₂) interactions of 2.59 and 3.15 Å were observed (Figure S53C). Because the adsorbates maximized their interactions within the binding pocket (Figure S53), these enthalpies are higher than the Q_st(C₂H₂) and Q_st(CO₂) determined from CMC simulations, the latter representing the average binding energy distribution at a certain pressure. Therefore, the loss in translational entropy was found to be higher for C₂H₂ than for CO₂. Experimental Q_st values representing a distribution averaged energy over all possible adsorption configurations of 16.1 kJmol⁻¹ (CO₂) and 40.6 kJmol⁻¹ (C₂H₂) were found to be slightly lower than the adsorption enthalpies computationally predicted from the optimal binding sites. Therefore, one can conclude that the average residence time of C₂H₂ in its most stable binding site compared with that of CO₂ is higher, suggesting that CO₂ will exhibit faster kinetics. When the CO₂ binding site of SIFSIX-21-Ni is compared with that of SIFSIX-3-Ni, unlike the single C^δ+···F^δ- binding interaction in SIFSIX-21-Ni, SIFSIX-3-Ni was found to exhibit four C^δ+···F^δ- binding interactions to electronegative F atoms from four independent SiF₆^2– anions.⁵⁶ Like SIFSIX-18-Ni-β, SIFSIX-21-Ni exhibits C^δ+···F^δ- and CH···O interactions with CO₂, however SIFSIX-18-Ni-β was found to exhibit multiple CH···O interactions thanks to the presence of extra methyl groups in pzpz over pypz.¹¹ Therefore, both SIFSIX-3-Ni and SIFSIX-18-Ni-β exhibit significantly higher CO₂ binding energies compared with SIFSIX-21-Ni.^11,56 Conversely, SIFSIX-21-Ni exhibits CH^δ+···F^δ- interactions with C₂H₂ similar to those observed in SIFSIX-2-Cu-i and TIFSIX-2-Cu-i thanks to their “sweet spots” for C₂H₂ binding (F∙∙∙F distance of ca. 7 Å, see Figure 1B).^13,53 In summary, the binding sites in the pypz HUMs reported herein combine key features that imply strong C₂H₂ affinity versus CO₂: CH^δ+···F^δ+ interaction-driven binding sites for C₂H₂; weak CO₂-sorbent interactions.

Figure 2.

The binding sites in SIFSIX-21-Ni that result in strong C₂H₂ affinity versus CO₂

(A and B) Views of (A) C₂H₂ binding sites and (B) CO₂ binding sites in SIFSIX-21-Ni as determined from molecular simulations (C₂H₂ and CO₂ molecules are shown in space-filling model whereas SIFSIX-21-Ni is presented in ball-and-stick model) (color codes: N, blue; Si, yellow; F, turquoise; Ni, lilac; O, red; H, white; C, gray, except in the C₂H₂ molecule: orange).

Adsorption kinetics

Kinetics is a key factor in determining the efficiency of gas separations.⁴⁴ We, therefore, studied the pure gas adsorption kinetics for C₂H₂ and CO₂ for each of the activated samples (SIFSIX-21-Ni, TIFSIX-4-Ni, NbOFFIVE-3-Ni, SIFSIX-21-Cu, TIFSIX-4-Cu, and NbOFFIVE-3-Cu) by exposure to a constant flow of 10 cm³min⁻¹ of C₂H₂ or CO₂ at 303 K and 1.0 bar (Figures S54 and S55). After 5 cycles of C₂H₂ sorption, the order of uptakes were as follows: SIFSIX-21-Cu (7.8%) < TIFSIX-4-Cu (9.7%) < NbOFFIVE-3-Cu (10%) for the Cu(II) HUMs; NbOFFIVE-3-Ni (4.2%) ~ TIFSIX-4-Ni (4.3%) < SIFSIX-21-Ni (5.9%) for the Ni(II) HUMs (Figures S54 and S55). Meanwhile, the corresponding order after 3 cycles of CO₂ sorption were as follows: NbOFFIVE-3-Cu (0.73%) ~ SIFSIX-21-Cu (0.78%) < TIFSIX-4-Cu (1.5%) and SIFSIX-21-Ni (0.84%) ~ TIFSIX-4-Ni (0.85%) < NbOFFIVE-3-Ni (0.96%) for the Cu(II) and Ni(II) HUMs, respectively. The HUMs with the highest gravimetric uptakes, NbOFFIVE-3-Cu and SIFSIX-21-Ni, reached ca. 99% of their saturation uptakes in 54 and 35 min, respectively. For all HUMs, sorbent regeneration was conducted over five consecutive C₂H₂ adsorption/desorption cycles at 333 K under N₂ flow in <30 min (flow rate: 20 cm³min⁻¹; Figures S54 andS55) for C₂H₂.

Dynamic column breakthrough (DCB) studies

Next, we investigated the C₂H₂/CO₂ separation performances of these HUMs through DCB experiments⁵⁷ with inlet gas mixture ratios 1:1 or 2:1 (v/v) for C₂H₂/CO₂.¹ Binary gas mixtures were passed through a fixed-bed reactor (8 mm diameter) filled with ca. 0.5 g of each HUM with a total gas flow rate of 1 cm³min⁻¹ at 1 bar and 298 K. Pre-activated samples were first heated at 333 K in a 20 cm³min⁻¹ flow of He to remove atmospheric impurities as monitored by gas chromatographic (GC) analysis of the effluent gas stream. The sorbent beds were then cooled to room temperature under continuous He flow and subjected to DCB experiments. Eluted gaseous components were continuously monitored through GC analysis (see supplemental information for details, Figure S56). Figure 3 reveals that CO₂ breakthrough occurred before that of C₂H₂ for each HUM and that SIFSIX-21-Ni was the best-performing sorbent, C₂H₂ breakthrough occurring at 363 and 298 min g⁻¹ for the 1:1 and 2:1 gas mixtures, respectively. In contrast, the corresponding CO₂ breakthrough occurred at 152 and 114 min g⁻¹ for the 1:1 and 2:1 gas mixtures, respectively. The time lags of 211 and 184 min g⁻¹ between C₂H₂ and CO₂ breakthroughs for 1:1 and 2:1 gas mixtures, respectively, imply high C₂H₂ productivities.

Figure 3.

Experimental dynamic column breakthrough curves

(A–F) Binary C₂H₂/CO₂ mixture-based DCB experimental curves at 298 K and 1 bar on the studied family of pypz HUM sorbents (v/v = 1:1, solid line; 2:1, dashed line; C₂H₂, red; CO₂, black) (A), (B), and (C): SIFSIX-21-Ni, TIFSIX-4-Ni, and NbOFFIVE-3-Ni, respectively, and (D), (E), and (F): SIFSIX-21-Cu, TIFSIX-4-Cu, and NbOFFIVE-3-Cu, respectively (see details in the supplemental information).

GC data revealed that, for the 1:1 experiments, C₂H₂ levels in the effluent CO₂ gas stream were 623, 910, 1,751, 2,368, 2,368, and, 2,718 ppm for SIFSIX-21-Ni, NbOFFIVE-3-Cu, TIFSIX-4-Cu, SIFSIX-21-Cu, TIFSIX-4-Ni, and NbOFFIVE-3-Ni, respectively. Unlike the low uptakes of CO₂ registered in single-component isotherms (Figures 1D and 1E), 1:1 and 2:1 C₂H₂/CO₂ DCB experiments (Figure 3) revealed higher adsorbed CO₂ amounts indicated by coadsorption at lower partial pressures in dynamic experiments. Until C₂H₂ breakthrough occurred, CO₂ purity values in the effluent streams were found to be as follows: >99.9% for SIFSIX-21-Ni and NbOFFIVE-3-Cu; >99.7% for TIFSIX-4-Cu, SIFSIX-21-Cu, TIFSIX-4-Ni, and NbOFFIVE-3-Ni, i.e., higher than the commercial specification for CO₂ (N2.0, 99%). C₂H₂ uptakes calculated from the breakthrough curves for SIFSIX-21-Ni, NbOFFIVE-3-Cu, TIFSIX-4-Cu, SIFSIX-21-Cu, TIFSIX-4-Ni, and NbOFFIVE-3-Ni were observed to be 3.22, 3.08, 2.8, 3.07, 3.02, and 2.88 mmolg⁻¹, respectively (see supplemental information for details)^58,59; these values are consistent with the respective isotherm-based uptakes at 0.5 bar. The DCB experiments enabled calculation of the separation selectivities (α_AC) for 1:1 and 2:1 C₂H₂:CO₂ mixtures: SIFSIX-21-Ni (27.7/10.0) > NbOFFIVE-3-Cu (16.9/7.9) > NbOFFIVE-3-Ni (15.0/6.5) > TIFSIX-4-Cu (5.4/4.1) > SIFSIX-21-Cu (4.6/3.1) > TIFSIX-4-Ni (4.4/3.1). SIFSIX-21-Ni exceeds the separation selectivities reported for UTSA-74a (20.1), JXNU-5 (9.9), JCM-1 (4.4), FJU-89 (3), SNNU-45 (2.9), NKMOF-1-Ni (2.6), FJU-6-TATB (2.3), FJU-36a (2.1), HOF-3 (2), SIFSIX-Cu-TPA (1.97), FJU-22a (1.9), and FeNi-M’MOF (1.7) (Table S1). After full saturation in the equimolar DCB experiments, temperature programmed desorption measurements were conducted at 333 K (Figure S57). The desorption profiles revealed complete adsorbent regeneration in <120 min under a He flow of 20 cm³ min⁻¹ for SIFSIX-21-Ni and NbOFFIVE-3-Cu.

A mapping of α_AC versus equilibrium single-component C₂H₂ uptakes at 1 bar indicates that SIFSIX-21-Ni, NbOFFIVE-3-Cu, NbOFFIVE-3-Ni, TIFSIX-4-Cu, and UTSA-74a have potential to address the trade-off between adsorption capacity and separation selectivity for C₂H₂/CO₂ separation as they are the only sorbents that exhibit α_AC ≥ 5 and adsorption capacity ≥ 3.5 mmolg⁻¹ (see Figure 4A for a comparison of performance parameters). The UMCs in UTSA-74a require high temperature for activation (473 K under high vacuum (Figure 4A). Conversely, the HUM sorbents require heating to only 333 K for sorbent regeneration. Overall, SIFSIX-21-Ni outperforms the other HUMs reported herein like UTSA-74a and other known C₂H₂ selective sorbents thanks to its high separation selectivity (27.7), C₂H₂ uptake (4.0 mmolg⁻¹), and low regeneration temperature. Interestingly, the high C₂H₂/CO₂ separation selectivities for M’FSIX-pypz-M/NbOFFIVE-pypz-M can be attributed as much to weak CO₂ binding as to strong C₂H₂ affinity as reflected in (ΔQ_st)_AC values. SIFSIX-21-Ni exhibits a relatively low Q_st(CO₂) of 19.8 kJ mol⁻¹ at low loading versus other C₂H₂ selective physisorbents (Table S1), resulting in a high (ΔQ_st)_AC of 18.1 kJ mol⁻¹. Indeed, (ΔQ_st)_AC at low loading for SIFSIX-21-Ni is close to that of NKMOF-1-Ni (19.4) kJ mol⁻¹ (Figure S59), which exhibits an exceptionally high Q_st(C₂H₂) of 60.3 kJ mol⁻¹. However, as Figure S58 reveals, the Q_st(C₂H₂) for NKMOF-1-Ni rapidly declines to 46.0 kJ mol⁻¹ at half loading, at which point (ΔQ_st)_AC is reduced to 9.5 kJ mol⁻¹ (Figure 4B). (ΔQ_st)_AC values at zero coverage, although widely used, tend to overestimate separation performance at relevant partial pressures. In our experience, for most C₂H₂ selective sorbents, particularly those with high surface areas, Q_st(C₂H₂) at half and full coverage decline from their zero loading values because of weak multilayer adsorption. Therefore, comparing (ΔQ_st)_AC at low coverage does not always translate well to a prediction of relative binding affinity and is why we consider (ΔQ_st)_AC at half loading to be a suitable metric to estimate relative binding for an equimolar mixture.⁴² Figure 4B reveals that NbOFFIVE-3-Cu and SIFSIX-21-Ni set new benchmarks of (ΔQ_st)_AC values at half loading, 15.2 and 14.9 kJmol⁻¹, respectively. This result is consistent with their DCB separation performances especially higher α_AC relative to those of TIFSIX-4-Ni, SIFSIX-21-Cu, TIFSIX-4-Cu, and NbOFFIVE-3-Ni. Overall, (ΔQ_st)_AC values of at least 7.5 kJmol⁻¹ for all HUMs reported herein suggests thermodynamic preference for C₂H₂ over CO₂ and correlates well with their high α_AC values.

Figure 4.

Comparison of separation selectivity versus uptake capacity and ΔQ_st

(A) Comparison of C₂H₂/CO₂ separation selectivity α_AC (v/v = 1:1) and gravimetric C₂H₂ uptake at 1 bar in benchmark C₂H₂/CO₂ separating adsorbents; regeneration/activation temperatures (range 298 to 473 K shown on the right side).

(B) Comparison of (ΔQ_st)_AC for the best-performing adsorbents at half loadings.

Spectroscopic studies

In situ infrared spectroscopy

To further probe the binding sites of C₂H₂ and CO₂ in these HUMs, we carried out in situ infrared (IR) spectroscopy measurements of C₂H₂ and CO₂ adsorption in the two best-performing adsorbents, NbOFFIVE-3-Cu and SIFSIX-21-Ni. The difference spectra for NbOFFIVE-3-Cu (Figure 5) and SIFSIX-21-Ni (Figure S60) demonstrate characteristic stretching bands for adsorbed C₂H₂ and CO₂ molecules, i.e., ν_as(C₂H₂) at 3,309–3,209 cm⁻¹ and ν_as(CO₂) at 2,338 cm⁻¹, whereas the perturbations of vibrational bands in the HUMs are characterized by the decrease in ν(CH)_phenyl peak intensity and its derivative feature in the shorter region, 1,700–1,000 cm⁻¹. The perturbed bands v(CH)_phenyl, ν(CC)_phenyl, δ(CH)_{ip, phenyl}, and ν(CN)_phenyl (see Figures 5, S60, and S61) indicate that C₂H₂ and CO₂ interact with the linker ligand’s phenyl rings,⁶⁰ as predicted computationally (Figure 2). Analysis of these data revealed that concomitant loading of C₂H₂ perturbs the vibrational bands of the HUMs more than CO₂, which is also indicative of stronger sorbent-sorbate interactions for C₂H₂. Particularly, two distinct ν_as(C₂H₂) bands are present at 3,209 and 3,309 cm⁻¹ in both NbOFFIVE-3-Cu and SIFSIX-21-Ni. These bands suggest two types of C₂H₂ binding sites with different binding strengths. This is in contrast to adsorbed CO₂, which displays a single peak at 2,338 cm⁻¹ in the corresponding IR spectrum, indicating only a single type of adsorbed CO₂ molecule. The 3,209 cm⁻¹ IR band decays slightly slower than the one at 3,309 cm⁻¹ (Figure S63), also implying stronger C₂H₂ binding inside the HUMs. Higher relative intensity suggests that the stronger binding sites are more populated with C₂H₂. C₂H₂ is well-known to be relatively acidic (pK_a = 25) in the context of hydrocarbons and, thus, tends to form hydrogen bonding interactions, as observed in several MOMs.^56,61 Similar to the well-studied OH or N-H stretch vibrations,⁶² the ν_as(C₂H₂) band undergoes a downward shift with respect to the gas phase value (at 3,287 cm⁻¹) upon forming intermolecular hydrogen bonds. The 3,209 cm⁻¹ band represents a red-shift of 78 cm⁻¹ and is consistent with the C₂H₂ binding site identified by simulation in Figure 2A. Overall, these in situ IR studies support the high adsorption selectivities for C₂H₂ over CO₂.

Figure 5.

In situ infrared spectra and ²H static NMR spectra

(A) Difference IR spectra showing the adsorbed CO₂ (orange) and C₂H₂ (pink) upon loading at 298 K and 1 bar adsorbate pressure into NbOFFIVE-3-Cu and subsequent evacuation of the gas phase within 3 s, respectively. Each is referenced to the spectrum of activated HUMs. Top inset shows the ν_as(CO₂) band and the bottom inset shows the decay of ν_as(C₂H₂) bands under vacuum. Notation and acronym: ν, stretching; δ, deformation; ip, in plane.

(B) Experimental ²H static NMR spectra of C₂D₂ adsorbed in NbOFFIVE-3-Cu (at a loading level of 0.4 C₂D₂ per Cu) as a function of temperature (blue lines) and simulated ²H spectrum of static C₂D₂ (red line).

Solid-state NMR spectroscopy

Solid-state NMR spectroscopy is a powerful technique for investigating the behavior of gaseous molecules adsorbed by porous materials.63, 64, 65 To better understand the adsorptive properties of the HUMs studied herein, ¹³C and ²H static solid-state NMR experiments were conducted to directly monitor the behavior of ¹³CO₂ and C₂D₂ molecules adsorbed by NbOFFIVE-3-Cu. Figure S64 illustrates the ¹³C NMR spectra of NbOFFIVE-3-Cu loaded with ¹³C-labeled CO₂ at various temperatures. At 373 K, the spectrum contains a relatively sharp, symmetric peak at 125 ppm superimposed on a very broad profile. Comparing the spectrum of the HUM loaded with CO₂ with the spectrum of the empty HUM (Figure S64) indicates that the broad resonance in the spectrum of CO₂-loaded HUM likely originates from the linkers in the framework. The position of the sharp signal indicates that this signal is from adsorbed CO₂. The sharpness of the resonance suggests that CO₂ molecules are rather mobile, implying weak interaction of CO₂ with the framework. Considering that ¹³C enrichment of ¹³CO₂ is 99% and that the framework carbon atoms are at natural abundance (1.1%), the relative intensity of the two signals suggests that a relatively small amount of CO₂ was adsorbed by the HUM, which is consistent with the poor S/N ratio of the spectrum. Previous studies showed that for MOMs featuring good affinity toward CO₂, the ¹³C spectra of adsorbed CO₂ usually exhibited a significantly higher intensity under similar adsorption conditions.66, 67, 68 Overall, the ¹³C NMR results indicate that NbOFFIVE-3-Cu does not adsorb CO₂ well. It is worth noting that the isotropic chemical shift of adsorbed CO₂ (125 ppm) is the same as that of CO₂ adsorbed in the diamagnetic MOMs,66, 67, 68 suggesting the lack of a significant paramagnetic interaction between CO₂ and Cu(II). This is likely due to the distances between the carbon of CO₂ and nearby paramagnetic metal ions being rather long, i.e., 8.03, 7.16, 6.03, 6.31 Å (Figure S65), and that CO₂ is highly mobile. Similar situations have been observed in other Cu(II)-MOMs.^69,70 Upon lowering the temperature, the sharp characteristic peak of CO₂ gradually became broader and merged with the broad framework peak at 253 K. We attribute this to reduced mobility of CO₂ with decreasing temperature and chemical shift anisotropy, which is largely averaged by molecular motions at higher temperatures, becoming dominant and broadening the signal.

In contrast to the ¹³C spectra of CO₂ loaded in NbOFFIVE-3-Cu, the ²H static NMR spectra of C₂D₂ loaded in this HUM exhibited much stronger signals (Figure 5B), suggesting that NbOFFIVE-3-Cu has a higher affinity for C₂D₂ versus CO₂ under the same loading conditions. The ²H static spectrum of adsorbed C₂D₂ at room temperature (293 K) exhibits a narrow pattern, suggesting a lack of significant paramagnetic interaction between the deuterons of C₂D₂ and Cu(II). This is consistent with the modeling study herein that indicates that two deuterium atoms in C₂D₂ are distant from nearby metal centers (5.1–6.4 Å). Inspection of the spectrum also reveals a well-defined line-shape with characteristic horns, shoulders, and “feet.” However, such a pattern cannot be simulated by a single site. Instead, analytical simulation using WSolids software⁷¹ revealed two components: a narrower component with a quadrupolar coupling constant (C_Q) of 43 kHz and a non-zero asymmetry parameter (η_Q) of 0.60 and a broader component with a C_Q of 65 kHz and η_Q of 0.0 (Figure S66B). The overall breadths of both patterns are markedly smaller than that of a static C₂D₂ (Figure 5B), which has a C_Q of 198 kHz and η_Q = 0.⁷² This observation indicates that the first-order quadrupolar interaction is averaged by molecular motions experienced by C₂D₂. Seeing two separate patterns suggests that there are two types of C₂D₂ molecules in the unit cell and that they undergo different motions. In an attempt to identify the types of motions, dynamic simulations were performed by using the Express software package.⁷³ Based on the simulation results, we propose that the narrower pattern results from the C₂D₂ molecules that simultaneously undergo two motions: (1) localized wobbling motion modeled by a C₃ rotation and (2) a delocalized hopping about a C₂ axis (Figure S66D). A small portion (around 10%) of C₂D₂ only wobbles at its absorption site (Figure S66C), yielding a broader line. The spectra between 293 and 353 K look similar, implying that the motions are in the fast exchange regime. In the temperature range 273–173 K, the pattern for each spectrum is gradually broadened and loses its characteristic discontinuities. The overall breadth of the pattern at 173 K is much broader than at 293 K, inferring that C₂D₂ adsorbed by the HUM becomes much less mobile. However, it is difficult to analyze the spectra at low temperature. When loading was increased, the ²H spectrum comprised a sharp peak in the middle and a broad component at the bottom (Figure S67). The sharp peak is likely due to the C₂D₂ molecules inside the pores undergoing fast exchange with the C₂D₂ outside the HUM. They do not interact with the framework strongly. The broad component is attributed to C₂D₂ molecules inside the pores that have strong interactions with the framework. The preference of C₂H₂ versus CO₂ indicated by solid-state NMR spectra further supports our results from single-component sorption, DCB experiments, molecular modeling, and in situ FTIR spectroscopic studies.

Accelerated stability tests⁷⁴ (313 K and 75% RH for 14 days) revealed that NbOFFIVE-3-Ni, SIFSIX-21-Cu, TIFSIX-4-Cu, and NbOFFIVE-3-Cu exhibit excellent hydrolytic stability (Figures S68C–S68F). Accelerated stability tests also revealed that both SIFSIX-21-Ni and TIFSIX-4-Ni (Figures S68A and S68B) underwent a phase change after exposure to humidity. Such phase changes have been observed in other HUMs and were attributed to inorganic pillar ligands being replaced by aqua ligands to afford corresponding sql networks.⁷⁵ SIFSIX-21-Ni was regenerated by heating the humidity-exposed sample at 358 K for 10 h in MeOH (Figure S69), whereas TIFSIX-4-Ni was regenerated by heating at 393 K (Figure S70).

Conclusions

In summary, six new HUMs, SIFSIX-21-Ni, TIFSIX-4-Ni, NbOFFIVE-3-Ni, SIFSIX-21-Cu, TIFSIX-4-Cu, and NbOFFIVE-3-Cu, were studied with respect to their ability to separate C₂H₂ from CO₂. Single-component sorption isotherms and equimolar C₂H₂/CO₂ binary gas mixture DCB experiments revealed that four of these HUMs, SIFSIX-21-Ni, NbOFFIVE-3-Ni, TIFSIX-4-Cu, and NbOFFIVE-3-Cu, break the trade-off between high adsorption capacities ≥3.5 mmol·g⁻¹ and high separation selectivities ≥5. SIFSIX-21-Ni outperformed all sorbents thanks to its benchmark separation selectivity (27.7) and high adsorption capacity (4 mmol·g⁻¹). The key to the performance of SIFSIX-21-Ni is pore size and chemistry that enables relatively high surface area and strong C₂H₂ binding sites. This study once again highlights the ability of ultramicroporous sorbents to offer hitherto unattainable selectivity values for key binary gas separations.²⁵

Experimental procedures

Resource availability

Lead contact

Further information and requests for resources should be directed to and will be fulfilled by the lead contact, Michael J. Zaworotko (xtal@ul.ie).

Materials availability

All materials generated in this study are available from the lead contact on request.

Data and code availability

The accession number for the crystal structures reported in this paper is Cambridge Crystallographic Data Centre, CCDC: 2052024, 2052025, 2052046, and 2052047.

Published: August 10, 2021