The First Sulfate‐Pillared Hybrid Ultramicroporous Material, SOFOUR‐1‐Zn, and Its Acetylene Capture Properties

Abstract

Hybrid ultramicroporous materials, HUMs, are comprised of metal cations linked by combinations of inorganic and organic ligands. Their modular nature makes them amenable to crystal engineering studies, which have thus far afforded four HUM platforms (as classified by the inorganic linkers). HUMs are of practical interest because of their benchmark gas separation performance for several industrial gas mixtures. We report herein design and gram‐scale synthesis of the prototypal sulfate‐linked HUM, the fsc topology coordination network ([Zn(tepb)(SO₄)] _n ), SOFOUR‐1‐Zn, tepb=(tetra(4‐pyridyl)benzene). Alignment of the sulfate anions enables strong binding to C₂H₂ via O⋅⋅⋅HC interactions but weak CO₂ binding, affording a new benchmark for the difference between C₂H₂ and CO₂ heats of sorption at low loading (ΔQ _st=24 kJ mol⁻¹). Dynamic column breakthrough studies afforded fuel‐grade C₂H₂ from trace (1 : 99) or 1 : 1 C₂H₂/CO₂ mixtures, outperforming its SiF₆ ²⁻ analogue, SIFSIX‐22‐Zn.

The prototypal member of a new family of hybrid ultramicroporous material, SOFOUR‐1‐Zn, uses earth‐friendly sulfate anions to selectively capture C₂H₂ from C₂H₂/CO₂ mixtures, enabling recovery of both gases at high purities even from trace mixtures.

Introduction

The amenability of metal–organic materials (MOMs) to design from first principles has afforded families of porous coordination networks (PCNs) with excellent properties for physisorptive separations. ^[1] In this context, hybrid ultramicroporous materials (HUMs) have emerged as an especially attractive class of PCNs. HUMs are typically comprised of an organic linker, an anionic inorganic linker, and a metal node. Their combination of ultramicropores (<7 Å diameter) and pore chemistry (strong electrostatics from the inorganic pillars that line pore walls) can afford highly selective binding sites for gaseous adsorbates. ^[2] The inherent modularity of HUMs is advantageous since it enables first generation HUMs to be systematically developed into sorbent families (platforms) with optimised pore sizes and chemistries. Such a crystal engineering approach offers insight into structure‐function relationships and means that second generation HUMs can offer a degree of control over binding sites and energies that is not readily available in traditional classes of sorbent such as zeolites and porous carbons. ^[3]

Most HUMs follow a simple structural blueprint in which divalent metal cations are 4‐connected at their equatorial positions by four neutral ditopic organic linkers, thereby forming a cationic square lattice (sql) topology coordination network. The axial positions of the metal centres are further linked by inorganic dianion “pillars” to yield a neutral primitive cubic, pcu, topology network. For example, in the archetypal HUM, SIFSIX‐1‐Zn, the metal cation is Zn²⁺, the organic linker is 4,4′‐bipyridine, and the inorganic pillar is SiF₆ ²⁻. ^[4] These building blocks can be substituted to produce new HUMs in a highly modular fashion: Zn²⁺ can be substituted by other M²⁺cations (e.g. Cu²⁺, Ni²⁺, Cd²⁺); 4,4′‐bipyridine can be replaced by longer (e.g. N,N′‐di(4‐pyridyl)‐1,4,5,8‐naphthalene diimide, 15.4 Å) or shorter (e.g. pyrazine, 2.8 Å) organic ligands; distinct platforms are then defined by the type of inorganic pillar used, e.g. MFSIX (e.g. TiF₆ ²⁻, SnF₆ ²⁻), FOXY (e.g. NbOF₅ ²⁻), M′FFIVE (e.g. AlF₅ ²⁻), and DICRO (Cr₂O₇ ²⁻), can replace SiF₆ ²⁻. ^[5] Although most HUMs are constructed in this manner, other topologies can exist: the mmo platform; ^[6] Tripp‐Cu‐MFSIX ([Cu₆(Tripp)₈](MF₆)₃(MF₆)₃]_n ), Tripp=2,4,6‐tris(4‐pyridyl)pyridine) is based upon a tritopic ligand; ^[7] the fsc networks CPM‐131 and fsc‐2‐SIFSIX are sustained by tetratopic ligands. ^[8] The recently reported fsc network ZJU‐280 exploited the tetratopic ligand tetra(4‐pyridyl)benzene (tepb) and was found to exhibit promising C₂H₂/C₂H₄ separation capabilities. ^[9]

HUMs are particularly amenable to crystal engineering, even compared to most MOFs, ^[3] enabling systematic fine‐tuning of pore size and pore chemistry and optimisation of key properties such as selectivity and working capacity. Indeed, the current top‐performing sorbents for several industrially important gas mixtures are HUMs: SIFSIX‐18‐Ni‐β , NbOFFIVE‐1‐Ni and TIFSIX‐3‐Ni for CO₂/N₂; ^[10] NbOFFIVE‐1‐Ni and TIFSIX‐3‐Ni for CO₂/CH₄; ^[11] SIFSIX‐14‐Cu‐i for C₂H₂/C₂H₄ ^[12] DICRO‐4‐Ni‐i, TIFSIX‐2‐Cu‐i, UTSA‐300, SIFSIX‐21‐Ni and BSF‐3 for C₂H₂/CO₂; ^[13] and CROFOUR‐1‐Ni for Xe/Kr. ^[14] The selection of the inorganic pillar is crucial since it is more than a structural component; pillars form the binding site that enables selective sorption. The use of such anionic pillars to form binding sites sets HUMs apart from other classes of porous coordination networks such as MOFs, which tend to rely upon coordinatively unsaturated metal centres (UMCs) to provide selective adsorbate binding. Importantly, the noncovalent sorbate‐sorbent interactions in HUMs tend to be weaker than UMCs and can lie in a thermodynamic “sweet spot” that allows benchmark selectivity to be combined with low energy desorption. Nevertheless, despite their ability to generate highly selective binding, fluorinated anion pillars can present challenges in terms of synthesis methodology (e.g. use of HF in production), cost and corrosivity (e.g. hydrolytic or thermal decomposition). ^[15] Other pillars, such as Cr₂O₇ ²⁻, CrO₄ ²⁻, and MoO₄ ²⁻ comprise toxic metals, and the boron oxyanions used in BSF‐3 and the recently reported ZNU‐1 are relatively intricate and costly.[ ^6a , ^13e , ¹⁶ ]

There is therefore a need to generate HUMs using earth‐friendly inorganic anions while retaining sorption performance. The sulfate anion (SOFOUR) is cheap, divalent, and metal/fluoride free, making it an ideal candidate for such studies. Herein, we report the first example of a SOFOUR HUM along with its gas sorption properties and separation performance for C₂H₂, acetylene, over CO₂.

We chose to study C₂H₂/CO₂ as it is among the most challenging of gas separations. Since its discovery in Ireland in the 1830s, acetylene has become a very widely produced chemical commodity with industrial utility as a chemical feedstock and fuel. ^[17] Fuel applications require >98 % pure C₂H₂ and use as a feedstock requires even higher purities. However, the predominant processes by which acetylene is produced utilize partial oxidation of alkanes, in which CO₂ is produced as a by‐product and a persistent contaminant.[ ^17b , ^17c , ¹⁸ ] In such processes, the absence of oxygen leads to soot formation, and thus the production of CO and CO₂ necessarily accompanies C₂H₂ production. Although most C₂H₂/CO₂ separation studies focus on 1 : 1 and 2 : 1 C₂H₂/CO₂ mixtures, the yields of the production routes are variable and depend on several factors including feedstock type and purity, oxygen content, temperature, and process considerations.[ ^17b , ^17c ] For instance, high temperature plasma pyrolysis techniques using CH₄ feeds may have C₂H₂ yields as high as 80–90 %, coal‐based processes can show yields between 20 and 80 %, the BASF controlled partial oxidation process has yields of 10–33 %, whereas other methods may have even lower yields, e.g. propane cracking yields only 2 % C₂H₂.[ ^17c , ¹⁸ ] Typically, higher yields are achieved under more demanding conditions and higher temperatures. Therefore, studies considering equimolar mixtures serve as representative examples of C₂H₂/CO₂ separations, but may not address the greater challenge of viably capturing and purifying acetylene from lower‐yield outputs.

Although the by‐product profiles in each case are different, CO₂ is an important and persistent contaminant, especially in production routes involving partial combustion. The separation of CO₂ from C₂H₂ is challenging due to similar physicochemical properties such as boiling point (194.7 K for CO₂, 188.4 K for C₂H₂) and quadrupole moment (4.3×10⁻²⁶ esu cm² for CO₂, 3.0×10⁻²⁶ esu cm² for C₂H₂). In addition, they have similar molecular dimensions (3.32×3.34×5.7 ų for CO₂, 3.18×3.33×5.36 ų for C₂H₂) and kinetic diameters (3.3 Å for both CO₂ and C₂H₂). ^[19] Further, C₂H₂ is explosive, so it is unsafe to liquefy in cryogenic purification processes such as those used for C₂H₄ and other hydrocarbons. Consequently, energy‐intensive gas–liquid absorption methods are used for the purification of C₂H₂, e.g. with solvents such as N‐methyl pyrrolidone, N,N‐dimethyl formamide, methanol, and acetone or alkaline scrubbing agents for chemical removal of CO₂. There is a large environmental cost and owing to the scale of C₂H₂ production (projected market value of 6.9 billion USD in 2025), even minor improvements to the economics and ecological footprints of these processes could result in major savings. ^[20] In addition, the extremely low flammability limit of acetylene in mixtures (2.5 %) discourages the recirculation of partially separated mixtures and can necessitate the trace removal of acetylene from gas streams for safety reasons. ^[21] In order to address the problem of separations in acetylene‐poor mixtures and feeds below the flammability limit, a number of recent studies have examined the viability of trace C₂H₂/CO₂ separations (1 : 99) by dynamic column breakthrough (DCB) studies.[ ² , ²² ] However, the physisorptive recovery of acetylene of commercial grade from low concentration mixtures of C₂H₂ with CO₂ has not previously been demonstrated.

The thermodynamic “sweet spot” for C₂H₂/CO₂ selective HUMs arises from arrangements of fluorinated inorganic anion pillars. Optimised geometries can enable molecular recognition of C₂H₂ via H‐bonding as seen for HUMs like SIFSIX‐21‐Ni, TIFSIX‐2‐Cu‐i and ZNU‐1.[ ^13b , ^13d , ^13e ] However, when C₂H₂ is preferentially adsorbed, the production of pure C₂H₂ requires desorption of C₂H₂ from the adsorbent and is operationally challenging. Sorbents with “inverse” selectivity, in which CO₂ is adsorbed preferentially over C₂H₂, have been studied as they can produce a pure C₂H₂ effluent stream (eg. SIFSIX‐3‐Ni, Tm(OH‐bdc), Cd‐NP, and CD‐MOF‐2).[ ^13b , ²³ ] Unfortunately, trace C₂H₂ streams present a limitation due to the rapid saturation of the adsorbent bed with CO₂. The optimum adsorbent for capture of C₂H₂ from dilute feeds would therefore exhibit high C₂H₂ uptake at low partial pressures, preferential C₂H₂ binding, and facile regeneration allowing the recovery of high‐purity C₂H₂ during desorption. In this contribution, we address both the challenge of efficient trace C₂H₂/CO₂ separation and the need for a cheap, green alternative inorganic pillar for HUMs.

Results and Discussion

Among the possible divalent anion pillars for possible utility as HUM pillars, sulfate anions stand out as they are exceptionally cheap, non‐toxic, and amenable for use at large scale. Further, the related tetrahedral dianions CrO₄ ²⁻ and MoO₄ ²⁻ form HUMs that exhibit 6‐connected 4⁸.6⁷ mmo topology. ^[6] In order to generate a HUM based on pillared sqls with channels comparable to those in pcu topology HUMs, we selected the tetratopic ligand tepb (tetra(4‐pyridyl)benzene) and Zn²⁺ nodes. Single crystals of the target HUM, [Zn(tepb)(SO₄)] _n , SOFOUR‐1‐Zn, were obtained by layering and studied by single crystal X‐ray diffraction. In addition, the SiF₆ ²⁻ pillared analogue, SIFSIX‐22‐Zn, was prepared using the same methodology to serve as reference point for performance evaluation.

SOFOUR‐1‐Zn crystallised in the orthorhombic space group Cmm2. Its structure is a (4,6)‐connected fsc topology network (Figure 1a, b) in which each octahedral Zn²⁺ moiety serves as a 6‐c node and each tepb ligand serves as a 4‐c node. Each Zn²⁺ moiety is coordinated by two ditopic bridging sulfate anions at the axial positions whereas the equatorial positions are occupied by four tetratopic tepb ligands. Notably, the SO₄ ²⁻ anions serve as pillars between Zn‐tepb 2‐dimensional sqls to afford an fsc topology net (Figure S1). The sulfate anions are disordered over two positions, suggesting they can rotate and modify the pore environment (see Supporting Information for refinement details). ^[24] Figure 1e reveals the outcome of a Cambridge Structural Database (CSD version 5.41 (2019+3 updates)) survey that plots pillaring angle (pillar=SiF₆ ²⁻ or SO₄ ²⁻) vs. M⋅⋅⋅ bridging distance of the pillar. ^[25] The pillaring angle for SOFOUR‐1‐Zn is close to linear, at 173.5(3)°, higher than 98 % of the pillaring angles for SO₄ ²⁻ (Figure 1e). As expected, for SO₄ ²⁻ there is a positive correlation between pillaring angle and M⋅⋅⋅M distance; SOFOUR‐1‐Zn fits this trend, having a relatively large M⋅⋅⋅M distance of 6.5163(2) Å, higher than 90 % of previously reported Zn‐based structures. Further details of the CSD survey and analysis are presented in the Supporting Information.

Figure 1.

a) Crystal structure of SOFOUR‐1‐Zn viewed along the crystallographic a‐axis; b) viewed along the crystallographic b‐axis; c) viewed along the crystallographic a‐axis; d) SIFSIX‐22‐Zn viewed along the crystallographic b‐axis; e) a scatter plot of the results for a CSD search of M⋅⋅⋅M distances for SO₄ ²⁻ and SiF₆ ²⁻ pillars versus the angle between one of the coordinating atoms (X=O, F) and the bridged metals (M=Mn, Fe, Co, Ni, Cu, Zn, Cd).

SIFSIX‐22‐Zn, [Zn(tepb)SiF₆]_n , crystallised in the orthorhombic space group Cmma and is isostructural to the recently reported Cu‐based ZJU‐280. ^[9] The pyridyl rings of the tepb ligands in both SOFOUR‐1‐Zn and SIFSIX‐22‐Zn (Figure 1a, b and c, d, respectively) are arranged in a propeller conformation around the Zn²⁺ metal centre. The dihedral angle between opposite rings for SOFOUR‐1‐Zn is 70.72(17)° and 59.10(14)° for adjacent rings whereas for SIFSIX‐22‐Zn, the angles are 59.08(9)°, and 64.65(8)°, respectively. Similarly, the pyridyl rings are also arranged in a propeller conformation around the phenyl ring of the tepb ligand with dihedral angles with the phenyl ring of 49.4(2)° and 51.8(2)° for the two crystallographically distinct pyridyl rings in SOFOUR‐1‐Zn, and 56.02(8)° for the one crystallographically distinct pyridyl ring in SIFSIX‐22‐Zn. Despite the differences in dihedral angles, the ligands from both SOFOUR‐1‐Zn and SIFSIX‐22‐Zn overlay closely with overlap of the atomic ellipsoids drawn at the 50 % probability level for every atom (Figure S2). The interlayer distance is 7.5678(9) Å is SIFSIX‐22‐Zn vs. 6.5163(2) Å in SOFOUR‐1‐Zn.

A CSD survey revealed that 1479 SO₄ ²⁻ bridged coordination polymers have been reported. ^[25] Of these, only 136 were found to have four nitrogen atoms coordinated to a metal centre (Figure S5, S6, Table S2). Permanent porosity was demonstrated experimentally through gas sorption experiments in just five: [Cd(Tppa)(SO₄)(H₂O)], [Cd₂(tpim)₄(SO₄)(H₂O)₂] ⋅ (SO₄)], [Co₂(bpy)₃(SO₄)₂(H₂O)₂](bpy)], [Cd₂L₄(SO₄)(H₂O)] ⋅ (SO₄)] and [Zn₂L₄(SO₄)(H₂O)₂] ⋅ (MeOSO₃)] (Tppa=tris(4‐(pyridyl)phenyl) amine, tpim, L=2,4,5‐tri(4‐pyridyl)imidazole, bpy=4,4′‐bipyridine), each of which is distinct in terms of structure. ^[26] Additionally, the 3,5‐connected net of formula [Cu(tepb)(SO₄)] has been reported without sorption data and is also distinct from the 4,6‐connected fsc net reported herein. ^[27] To our knowledge, SOFOUR‐1‐Zn is not only prototypal for a new HUM platform, it is the first SO₄ ²⁻‐based porous coordination network of any type to be studied for gas separations.

Gram‐scale quantities of SOFOUR‐1‐Zn were synthesised by stirring zinc(II) sulfate and tepb in MeOH at room temperature. The resulting microcrystalline white powder was characterised by powder X‐ray diffraction and thermogravimetric analysis, which indicated phase purity and thermal stability up to ca. 300 °C (Figure S3, S4). An initial solvent loss of 10.5 wt.% was observed below 100 °C, corresponding to loss of MeOH from the as‐synthesized structure. CO₂ sorption at 195 K enabled determination of BET surface areas, 612.1 m² g⁻¹ for SOFOUR‐1‐Zn and 641.0 m² g⁻¹ for SIFSIX‐22‐Zn (Figure 2a). Pore‐size distributions calculated from these isotherms by the Horvath–Kawazoe method indicated maximum pore widths for SIFSIX‐22‐Zn and SOFOUR‐1‐Zn at 4.1 Å and 4.0 Å respectively, validating them as ultramicroporous (Figure 2b).

Figure 2.

a) The 195 K CO₂ sorption isotherms of SOFOUR‐1‐Zn and SIFSIX‐22‐Zn; b) Horvath–Kawazoe pore‐size distributions of SOFOUR‐1‐Zn and SIFSIX‐22‐Zn calculated from the 195 K CO₂ sorption isotherms; c) 298 K C₂H₂ and CO₂ isotherms of SOFOUR‐1‐Zn and SIFSIX‐22‐Zn; d) coverage‐dependent isosteric heats of C₂H₂ and CO₂ for SOFOUR‐1‐Zn and SIFSIX‐22‐Zn; e) C₂H₂/CO₂ IAST selectivities of SOFOUR‐1‐Zn and SIFSIX‐22‐Zn at 298 K for various compositions vs. pressure; f) gravimetric kinetics of C₂H₂ and CO₂ sorption on SOFOUR‐1‐Zn and SIFSIX‐22‐Zn plotted as uptake vs. time.

Isotherms measured on SIFSIX‐22‐Zn at 298 K showed type I characteristics with uptakes for CO₂ of 95 cm³ g⁻¹ and 127 cm³ g⁻¹ for C₂H₂. Whereas SOFOUR‐1‐Zn exhibited a lower 1 bar uptake for C₂H₂ (69 cm³ g⁻¹) than CO₂ (81 cm³ g⁻¹), its high C₂H₂ uptake at lower pressures is indicative of strong C₂H₂ binding sites. Specifically, at 298 K and 0.01 bar, SOFOUR‐1‐Zn had uptakes of 1.65 mmol g⁻¹ of C₂H₂ and 0.20 mmol g⁻¹ of CO₂ (Figure 2c). This low pressure C₂H₂ uptake is comparable to top‐performing acetylene sorbents such as MUF‐17 (1.40 mmol g⁻¹), NKMOF‐1‐Ni (1.74 mmol g⁻¹), TIFSIX‐2‐Cu‐i (1.78 mmol g⁻¹), and UTSA‐200a (1.83 mmol g⁻¹).[ ¹² , ^13b , ^22a , ²⁸ ] Experimentally determined isosteric heats of adsorption (Q _st) for SOFOUR‐1‐Zn (33 kJ mol⁻¹ for CO₂, 57 kJ mol⁻¹ for C₂H₂ at low loading, Figure 2d), S7–S10 are consistent with experimental uptakes at low pressures. The difference in Q _st values between C₂H₂ and CO₂ (ΔQ _st), 24 kJ mol⁻¹, is to our knowledge the highest yet reported for a physisorbent (NKMOF‐1‐Ni=19.4 kJ mol⁻¹, CPL‐1‐NH₂ =17.6 kJ mol⁻¹, BSF‐3=17.2 kJ mol⁻¹, sql‐16‐Cu‐NO₃‐α ‘=13.0 kJ mol⁻¹).[ ^13e , ²⁸ , ²⁹ ] Low‐loading Q_st values for SIFSIX‐22‐Zn were determined to be 36.5 kJ mol⁻¹ for C₂H₂ and 25 kJ mol⁻¹ for CO₂, a ΔQ _st of 11.5 kJ mol⁻¹.

Ideal Adsorbed Solution Theory (IAST) calculations using 298 K isotherms indicated that SOFOUR‐1‐Zn and SIFSIX‐22‐Zn display similar C₂H₂/CO₂ selectivities (S_AC) of ca. 6.60 and 6.49 respectively for equimolar mixtures at 1 bar (Figure 2e). The calculated selectivity is nearly constant for SIFSIX‐22‐Zn for compositions of 10 % (7.08), 5 % (7.23), and 1 % C₂H₂ in CO₂ (7.36). Conversely, selectivity values increase for SOFOUR‐1‐Zn (9.55 for 10 %, 10.85 for 5 % and 13.00 for 1 %), indicating potential for trace removal. The differences between SOFOUR‐1‐Zn and SIFSIX‐22‐Zn in terms of their sorption characteristics can be attributed directly to the use of the SO₄ ²⁻ pillar and its effects on interlayer distance and pore electrostatics. Temperature swing cycling experiments conducted gravimetrically under C₂H₂ and CO₂ flow conditions revealed that the adsorption performance of SOFOUR‐1‐Zn and SIFSIX‐22‐Zn was retained in successive cycles. Further, the sorbents were regenerated through a 303 K to 363 K temperature swing (Figure S11). Initial adsorption rates indicated that both SOFOUR‐1‐Zn and SIFSIX‐22‐Zn exhibited faster uptake kinetics for C₂H₂ than for CO₂, favouring SOFOUR‐1‐Zn over SIFSIX‐22‐Zn (Figure S12). Specifically, in the first 2.5 minutes of adsorption, SOFOUR‐1‐Zn adsorbed C₂H₂ equivalent to 20.0 % of its 1 bar saturation uptake, but only 1.2 % of its CO₂ uptake. Kinetics of adsorption in SIFSIX‐22‐Zn are faster but less discriminatory, with 2.5 minute loading equal to 31.5 % and 4.1 % of C₂H₂ and CO₂ saturation uptakes, respectively (Figure 2f).

These properties suggest SOFOUR‐1‐Zn as a candidate for trace C₂H₂/CO₂ separations. Indeed, comparison with leading C₂H₂/CO₂‐selective sorbents (S _AC>5) that do not use chemisorptive or UMC binding sites revealed that SOFOUR‐1‐Zn is one of just four MOMs with a C₂H₂ Q _st value in the “sweet spot” between 45 and 60 kJ mol⁻¹ that allows both strong binding and energy‐efficient regeneration. ^[2] Among these, SOFOUR‐1‐Zn is the only adsorbent with a ΔQ _st greater than 20 (Figure 3). Therefore, despite its modest S_AC value, when this is coupled with a kinetic preference for C₂H₂ over CO₂, SOFOUR‐1‐Zn is highly suited to C₂H₂/CO₂ separation at low partial pressures of C₂H₂.

Figure 3.

a) 1 bar C₂H₂ uptake versus low loading ΔQ _st (C₂H₂−CO₂) for leading C₂H₂/CO₂ selective physisorbents; b) low loading ΔQ _st (C₂H₂−CO₂) vs. low loading C₂H₂ Q _st for leading C₂H₂/CO₂ selective physisorbents; SOFOUR‐1‐Zn is the only sorbent with ΔQ _st>20 and C₂H₂ Q _st from 45–60 kJ mol⁻¹.

In order to experimentally validate the separation performance of SOFOUR‐1‐Zn under mixed‐gas conditions, we conducted dynamic column breakthrough (DCB) experiments using a fixed bed and gas mixture compositions between 50 % C₂H₂ and 1 % C₂H₂ at ambient conditions (Figure 4). Identical experimental conditions were used for SIFSIX‐22‐Zn as a reference point. We indeed determined that SOFOUR‐1‐Zn separated an equimolar C₂H₂/CO₂ mixture at a combined flow rate of 1 sccm and ambient conditions with breakthrough times of ca. 22 min g⁻¹ for CO₂ and ca. 83 min g⁻¹ for C₂H₂. The separation factor (α_AC ) was calculated to be 17.5, higher than some leading C₂H₂ selective adsorbents, such as ZJU‐74a (4.3), HOF‐3a (2.0), and NKMOF‐1‐Ni (2.6), but lower than benchmark sorbents such as the TCuX series (33.4–143.1), IPM‐101 (22.5) and sql‐16‐Cu‐NO₃‐α ‘ (78).[ ^22b , ²⁸ , ²⁹ , ³⁰ ] Effluent CO₂ purity was >99.996 % until the elution of C₂H₂. This is equivalent to CO₂ at N4.5 CP Grade specification. Under the same conditions, SIFSIX‐22‐Zn exhibited α_AC of 3.8 and effluent CO₂ purity of >99.99 %, still a strong performance but not as efficient as SOFOUR‐1‐Zn.

Figure 4.

C₂H₂/CO₂ DCB curves for SOFOUR‐1‐Zn with inlet flows of a) 1 : 1 C₂H₂ : CO₂, 1 sccm; b) 1 : 99 C₂H₂ : CO₂, 7 sccm; c) C₂H₂/CO₂ TPD curves for SOFOUR‐1‐Zn after saturation with an inlet flow of c) 1 : 99 C₂H₂ : CO₂, 7 sccm; C₂H₂/CO₂ DCB curves for SIFSIX‐22‐Zn with inlet flows of d) 1 : 1 C₂H₂ : CO₂, 1 sccm; e) 1 : 99 C₂H₂ : CO₂, 7 sccm; f) C₂H₂/CO₂ TPD curves for SIFSIX‐22‐Zn after saturation with an inlet flow of 1 : 99 C₂H₂ : CO₂, 7 sccm.

DCB experiments were then conducted using a 1 : 99 C₂H₂/CO₂ inlet stream. SOFOUR‐1‐Zn exhibited a remarkable breakthrough time of ca. 270 min g⁻¹ for C₂H₂ at a total inlet flow rate of 7 sccm. CO₂ broke through the column within 3 min g⁻¹, resulting in calculated uptakes of 21.9 cm³ g⁻¹ of C₂H₂ and 17.4 cm³ g⁻¹ of CO₂, and α_AC of 124.6. No other 1 : 99 α_AC DCB experiments have been reported in the literature. The purity of the effluent CO₂ remained >99.996 % under these conditions. SIFSIX‐22‐Zn exhibited uptakes of 5.6 cm³ g⁻¹ of C₂H₂ and 20.3 cm³ g⁻¹ of CO₂ under the same conditions for α_AC of 27.3. At intermediate inlet gas compositions of 5 % and 10 % C₂H₂, α_AC values of 42.5 and 52.6 were measured, respectively, for SOFOUR‐1‐Zn, while the values for SIFSIX‐22‐Zn were 11.9 and 15.5, respectively (Figure S13, S14).

Temperature‐programmed desorption (TPD) experiments were conducted after saturation in DCB experiments using a He gas stream at 20 sccm and a temperature gradient up to 333 K in order to evaluate the feasibility of recovering purified C₂H₂ by desorption. SOFOUR‐1‐Zn and SIFSIX‐22‐Zn both exhibited rapid desorption of CO₂, followed by peaks associated with desorption of C₂H₂ as the temperature was elevated under both 1 : 1 and 1 : 99 conditions (Figure S15, S16). In the 1 : 1 experiments, much of the C₂H₂ was desorbed with CO₂ at the onset of He flow and C₂H₂ desorption also occurred at elevated temperature. Elution of high purity C₂H₂, >99.5 %, which exceeds instrument grade specifications of purity (>99.0 %), occurred from 15 to 41 min g⁻¹ for SOFOUR‐1‐Zn and from 17 to 30 min g⁻¹ for SIFSIX‐22‐Zn. These values correspond to productivities of 2.07 L kg⁻¹ and 1.95 L kg⁻¹. Such productivity equals the desorptive C₂H₂ recovery performance of ZNU‐1 and the peak C₂H₂ purity is comparable to the benchmark set by TIFSIX‐2‐Cu‐i (99.9 %).[ ^13b , ¹⁶ ] C₂H₂ of purity >98 % was recovered between 15 and 70 min g⁻¹ with a productivity of 3.1 L kg⁻¹ using SOFOUR‐1‐Zn, and between 17 and 46 min g⁻¹ with a productivity of 3.3 L kg⁻¹ using SIFSIX‐22‐Zn. We note that even after adsorptive saturation using a 1 : 99 mixture, C₂H₂ of purity >98 % (fuel grade) was obtained by desorption from SOFOUR‐1‐Zn for the period between 33 and 40 min g⁻¹ and >95 % between 33 and 84 min g⁻¹, corresponding to productivities of 1.01 L kg⁻¹ and 4.66 L kg⁻¹ of >98 % pure and >95 % pure C₂H₂, respectively. To our knowledge, this is the first demonstration of physisorptive recovery of fuel grade C₂H₂ from a dilute (1 : 99) C₂H₂/CO₂ mixture. In contrast, the peak C₂H₂ purity achieved during desorption from SIFSIX‐22‐Zn was 96.7 % (at 30.9 min g⁻¹). We attribute this exceptional performance to the favorable C₂H₂ and CO₂ Q_st values and high C₂H₂ uptake at low partial pressure, both of which are enabled by highly selective C₂H₂ binding sites.

We conducted computational studies to gain insight into SO₄ ²⁻⋅⋅⋅C₂H₂ interactions. The binding sites in SOFOUR‐1‐Zn derived by density functional theory (DFT) calculations for both gases revealed that the interlayer “mezzanine” region which corresponds to the maximum pore diameter plays a key role (Figure 5, S23, S25, S26). C₂H₂ has two hydrogen bonds with SO₄ ²⁻ pillars (H⋅⋅⋅O=2.56 Å and 2.83 Å) while CO₂ has no close C⋅⋅⋅O contacts. (C⋅⋅⋅O=6.12 Å and 6.76 Å). The low loading adsorption enthalpies at 298 K from DFT calculations for C₂H₂ and CO₂ of −53, and −34 kJ mol⁻¹, respectively, are consistent with the Q_st values obtained experimentally. In SIFSIX‐22‐Zn, C₂H₂ has two hydrogen bonds with the framework (H⋅⋅⋅F=3.08 Å and 2.53 Å) while CO₂ has one close C⋅⋅⋅F contact (C⋅⋅⋅F distance=2.99 Å). Lower enthalpies of adsorption were calculated for both gases (Figure S24). Importantly, we found that when SO₄ ²⁻ moieties were afforded freedom to rotate, alternating pairs of SO₄ ²⁻ pillars can synergistically orient an oxygen atom directly into a channel to optimally bind a C₂H₂ molecule, leaving alternating channels without SO₄ ²⁻ oxygen atoms pointing directly into them. That sulfate anions can rotate is experimentally supported by the crystal structure of SOFOUR‐1‐Zn, which exhibited disorder of the sulfate pillars (see Supporting Information for full details). Adsorption of C₂H₂ by the alternating pores with unfavourable electrostatics was calculated to be less exothermic. Such a binding mechanism would explain lower Q _st and reduced C₂H₂ uptake at higher loadings for SOFOUR‐1‐Zn vs. SIFSIX‐22‐Zn as the octahedral SiF₆ ²⁻ pillars preclude an alternating binding site arrangement (Figure S27).

Figure 5.

The binding sites of a) C₂H₂ and b) CO₂ in SOFOUR‐1‐Zn and c) C₂H₂ and d) CO₂ in SIFSIX‐22‐Zn obtained by DFT calculations. (Colour codes: N, blue; Si, gold; S, yellow; F, turquoise; Zn, lavender; O, red; H, white; C, grey and C (C₂H₂), orange. The distances are in Angstrom (Å). C₂H₂ and CO₂ molecules are shown in space‐filling mode.

Temperature swing cycling experiments revealed retention of breakthrough performance over three consecutive adsorption–desorption cycles (Figure S17), as well as good retention of the initial rates of sorption from gravimetric experiments. However, whereas both SOFOUR‐1‐Zn and SIFSIX‐22‐Zn were found to be stable to multiple regeneration cycles and storage under ambient conditions for at least 4 months (Figure S21), water vapor sorption isotherms conducted on both HUMs revealed that SIFSIX‐22‐Zn displayed a dramatic negative deviation in uptake at 80 % R.H., corresponding with a phase transformation reminiscent of other well‐studied HUMs (Fig S18). ^[31] In contrast, SOFOUR‐1‐Zn exhibited reversible Type I water sorption, with no discernible phase change occurring, although very minor peak broadening was observed by PXRD.

The stabilizing effect of the SOFOUR pillar is corroborated by PXRD studies on samples exposed to 75 % R.H. and 313 K in line with previously reported accelerated moisture stability tests. ^[10c] We observed that SIFSIX‐22‐Zn underwent a phase change within 12 hours of exposure, whereas the onset of the phase change for SOFOUR‐1‐Zn occurred after 96 hours (Figure S19, S20). Therefore, the pillaring strategy outlined here resulted in improved stability and performance. In terms of the cost of manufacture, tepb can be prepared through a simple one‐step synthesis (SI) although it is not widely commercially available. Mechanochemical synthesis of SOFOUR‐1‐Zn was attempted, but the material thus obtained exhibited limited porosity (Figure S22).

Conclusion

To conclude, we report the gram‐scale room temperature synthesis of the prototypal sulfate‐pillared HUM, SOFOUR‐1‐Zn. The use of sulfate pillars makes SOFOUR‐1‐Zn greener, cheaper, more stable, and more effective in the separation of C₂H₂ from CO₂ than previously reported materials and its SIFSIX analogue, SIFSIX‐22‐Zn. SOFOUR‐1‐Zn was found to exhibit benchmark performance in for trace separation of C₂H₂ from CO₂ and is the first sorbent that yields fuel grade C₂H₂ (>98 % purity) from a 1 : 99 C₂H₂/CO₂ stream on desorption. DFT calculations provided insight into C₂H₂ binding in SOFOUR‐1‐Zn, revealing that it is enabled by SO₄ ²⁻⋅⋅⋅C₂H₂ H‐bonding. This work reiterates that ultramicroporous physisorbents are highly effective for trace gas separations and demonstrates that they can be prepared using cheap and ubiquitous building blocks. Further research will focus on constructing SO₄ ²⁻‐pillared HUMs with commercially available linkers and improving their stability to humid conditions.

Conflict of interest

The authors declare no conflict of interest.

Supporting information

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supporting Information

Supporting Information

Supporting Information

Supporting Information

Supporting Information

D. Sensharma, D. J. O'Hearn, A. Koochaki, A. A. Bezrukov, N. Kumar, B. H. Wilson, M. Vandichel, M. J. Zaworotko, Angew. Chem. Int. Ed. 2022, 61, e202116145; Angew. Chem. 2022, 134, e202116145.

Data Availability Statement

The data that support the findings of this study are available in the supplementary material of this article.