Pore Engineering for One-Step Ethylene Purification from a Three-Component Hydrocarbon Mixture.

Ethylene production from C2 hydrocarbon mixtures through one separation step is desirable but challenging because of the similar size and physical properties of acetylene, ethylene, and ethane. Herein, we report three new isostructural porous coordination networks (NPU-1, NPU-2, NPU-3; NPU represents Northwestern Polytechnical University) that are sustained by 9-connected nodes based upon a hexanuclear metal cluster of composition [Mn6(μ3-O)2(CH3COO)3]6+. NPU-1/2/3 exhibit a dual cage structure that was systematically fine-tuned in terms of cage size to realize selective adsorption of C2H2 and C2H6 over C2H4. Dynamic breakthrough experiments demonstrated that NPU-1 produces ethylene in >99.9% purity from a three-component gas mixture (1:1:1 C2H2/C2H4/C2H6). Molecular modeling studies revealed that the dual adsorption preference for C2H2 and C2H6 over C2H4 originates from (a) strong hydrogen-bonding interactions between electronegative carboxylate O atoms and C2H2 molecules in one cage and (b) multiple non-covalent interactions between the organic linkers of the host network and C2H6 molecules in the second cage.

Introduction

Ethylene (C2H4) is one of the most important industrial products, with production levels of >150 million tons in 2016.[1] The energy footprint needed to purify ethylene and propylene (C3H6), another high-volume product, means that they collectively account for ca. 0.3% of global energy.[2] Currently, C2H4 is mainly produced by steam-cracking reaction of carbon-included feedstocks, and high-purity ethylene is afforded by energy-intensive separation processes of downstream C2 hydrocarbon gas mixtures.[3] Acetylene (C2H2) is first removed by catalytic hydrogenation using noble metal catalysts at high temperature and pressure, whereas ethane is later separated from C2H4 by cryogenic distillation. A transition to a more energy-efficient separation technology with a much lower energy footprint and simplified separation process (e.g., simultaneous removal of C2H2 and C2H6 in one step) for ethylene production would be of societal relevance.

Physisorption-based separation processes hold the promise for greatly reducing energy consumption of gas separation, thanks to the low regeneration temperature and fast sorption kinetics typical of physisorbents.[4] Porous coordination networks based upon metal–organic materials (MOMs),[5] also known as metal–organic frameworks (MOFs)[6,7] and porous coordination polymers (PCPs),[8] have emerged as promising C2 light hydrocarbon physisorbents due to their amenability to exquisite control of pore shape and chemistry through reticular chemistry and crystal engineering strategies.[9,10] In this context, recent progress has been made with respect to binary separations such as C2H2/C2H4 and C2H6/C2H4.[11−32,32] Simultaneous removal of both C2H2 and C2H6 in a single step would simplify the purification of C2H4 but remains a challenge for a single physisorbent. This is because the quadrupole moment and kinetic diameter for C2H4 (1.5 × 10–26 esu cm2 and 4.1 Å) sit between those of C2H2 (7.2 × 10–26 esu cm2 and 3.3 Å) and C2H6 (0.65 × 10–26 esu cm2 and 4.4 Å). These physicochemical properties exacerbate the challenge for one-step C2H4 production from C2H2 and C2H6 by either molecular sieving or thermodynamic selectivity. Thus far, we are aware of only three literature reports of C2H4 production from a C2H2–C2H4–C2H6 mixture in a one-step process. In 2018,[33] simultaneous trapping of C2H2 and C2H6 was realized by the MOF TJT-100 thanks to the hierarchy of weak sorbent–sorbate interactions. In 2019,[34] we introduced the concept of synergistic sorbent separation technology (SSST) to enable high-purity ethylene production from the ternary mixture (C2H2–C2H4–C2H6) and quaternary mixture (C2H2–C2H4–C2H6–CO2) by using tandem packing of three MOMs in a fixed-bed sorbent. In SSST, three benchmark sorbents (SIFSIX-3-Ni, TIFSIX-2-Cu-i, and Zn-atz-ipa) were selected for CO2, C2H2, and C2H6 removal, respectively. Most recently,[35] a Th-azole network (Azole-Th-1) enabled selective adsorption of ethane and acetylene over ethylene. The discovery of new physisorbents that enable selective adsorption of C2H2 and C2H6 over C2H4 is timely not just to address the practical utility of such sorbents but also to advance our understanding of C2H2, C2H4, and C2H6 binding sites in porous materials.

In this contribution, we introduce the new hexanuclear Mn cluster [Mn6(μ3-O)2(CH3COO)3]6+ and its use as a building block for three isostructural 9-connected (9-c) MOMs with varying pore size. The hexanuclear metal cluster can be viewed as a fusion of two well-known trinuclear metal clusters bridged by three acetate anions. The Mn6 cluster reported herein has the potential to sustain 12-c networks, whereas the trinuclear parent typically serves as a 3-c, 6-c, or 9-c node (Scheme ).[36−41] The linking of the Mn6 cluster by rigid dicarboxylate linker ligands (BDC = 1,4-benzenedicarboxylate; NPDC = naphthalene-2,6-dicarboxylate; BPDC = biphenyl-4,4′-dicarboxylate) and the 3-connected pyridyl-based tritopic ligand 2,4,6-tris(4-pyridyl)pyridine (Tripp) afforded three isostructural MOMs (NPU-1, NPU-2, and NPU-3; NPU = Northwestern Polytechnical University) with tunable pore size and pore chemistry. We report herein the C2 sorption and separation properties of these new sorbents.

Scheme 1

M3(μ3-O) Metal Clusters Are Well Studied and Have Previously Been Used to Construct 3-, 6-, or 9-Connected Porous Coordination Networks (Left); Augmented 9-Connected Networks Are Reported Herein from a New Mn6(μ3-O)2 Cluster (Right)

Results and Discussion

Single crystals of NPU-1, NPU-2, and NPU-3 were harvested following solvothermal reaction of manganese acetate tetrahydrate, Tripp, and a dicarboxylic acid ligand (H2BDC, H2NPDC, or H2BPDC) in N,N-dimethylacetamide (DMA) at 373 K. Single-crystal structure analysis revealed that the three materials are isostructural 3D networks based on the aforementioned hexanuclear Mn6 cluster. The novel Mn6 cluster can be regarded as the result of fusion of two M3(μ3-O) clusters through bridging acetate anions. Each manganese cation adopts an octahedral geometry through coordination with two oxygen atoms from two dicarboxylate ligands, two oxygen atoms from acetate anions, one μ3-O atom, and one nitrogen atom from Tripp (Supplementary Figure 1). Overall charge neutrality of these networks requires mixed-valence MnIII and MnII cations, which is supported by X-ray photoelectron spectroscopy (XPS) analysis (Supplementary Figure 2) and suggests electron delocalization. Unsuccessful ion-exchange experiments with charged dye molecules were conducted upon NPU-1 and NPU-3, indicating an absence of extra-framework counterions (Supplementary Figure 3). Such dye exchange experiments have been utilized by others to verify the charge state of host coordination networks.[45] The general formula for NPU-1/2/3 is Mn5IIMnIII(μ3-O)2(CH3COO)3(Tripp)2(L)3 (L = BDC2–/NPDC2–/BPDC2– for NPU-1/2/3, respectively).

M3(μ3-O) clusters are normally connected by three, six, or nine organic linker ligands, but the Mn6 cluster in NPU-1/2/3 is bonded to 12 linker ligands (six dicarboxylate ligands and six Tripp ligands). The potential for rare 12-c nodes exists,[5,42−44] but pairs of Tripp ligands serve as parallel 3-c nodes that connect to the same adjacent Mn6 cluster. Therefore, the Mn6 cluster should be viewed as a 9-c node, and NPU-1/2/3 are classified as 3,9-connected networks with pacs topology[46] (Supplementary Figure 4). A similar “double cross-linking” situation occurred in the metal–organic polyhedron of formula [Cu2(bdc)2]12 (bdc = 1,3-benzenedicarboxylate). In this case, the 24 accessible connection points can serve as 6-c or 24-c nodes to build augmented pcu or rht networks.[47,48] Notably, there are dual cages (A and B) in NPU-1/2/3 (Figure ). Cage A has a trigonal bipyramidal shape, as it is surrounded by five metal clusters, six dicarboxylate ligands, and six Tripp ligands. Whereas the equatorial dimension of Cage A is consistent in NPU-1/2/3 (0.79 nm after subtracting the van der Waals radii of the nearest atoms), the axial dimension increases with the length of the dicarboxylate linker ligands (1.26, 1.77, and 2.24 nm for NPU-1/2/3, respectively). Cage B exhibits a distorted triangular prismatic shape, as it is formed from six adjacent Mn6 clusters, six dicarboxylate ligands, and four parallel Tripp ligands. The axial dimension in cage B varies from 0.38 to 0.66 to 0.91 nm for NPU-1/2/3, respectively. Each cage A is connected to six cage B’s in a face-shared manner, and vice versa. The triangular interconnecting pore windows between these cages, which are surrounded by one Tripp ligand and two carboxylate ligands, exhibit aperture sizes of 0.48 nm for NPU-1, 0.6 nm for NPU-2, and 0.78 nm for NPU-3. As revealed below, the dual cage nature of NPU-1/2/3 is crucial to their selective adsorption of C2H2 and C2H6 over C2H4. The porosity of NPU-1, NPU-2, and NPU-3 as calculated by PLATON[49] was found to be 50.8%, 56.8%, and 62.1%, respectively.

Figure 1

NPU-1/2/3 are isostructural coordination networks that are based upon assembly of Mn6 clusters, Tripp ligands, and different dicarboxylate linker ligands. The three-dimensional pore structures comprise two distinct cages.

Bulk purity of NPU-1/2/3 was confirmed by powder X-ray diffraction (PXRD, Figure a and Supplementary Figures 6–8). Thermogravimetric analysis (TGA) of as-synthesized and CH2Cl2-exchanged samples revealed that the DMA solvent in the as-synthesized phase can be fully exchanged with CH2Cl2 and that NPU-1/2/3 are thermally stable until 573 K (Supplementary Figures 9–11). N2 sorption experiments were conducted at 77 K to establish permanent microporosity, and NPU-1/2/3 each exhibited reversible type-I adsorption isotherms. By assuming pore filling by liquid N2 at 77 K and 100 kPa, the pore volumes calculated from N2 uptake at 100 kPa are 0.47 for NPU-1, 0.66 for NPU-2, and 0.77 cm3 g–1 for NPU-3. These values match well with the values of 0.53 for NPU-1, 0.68 for NPU-2, and 0.84 cm3 g–1 for NPU-3 calculated from the respective crystal structures. These findings are consistent with the absence of counterions in the pore channels of NPU-1/2/3. Langmuir and Brunauer–Emmett–Teller (BET) surface areas were calculated to be 1557 and 1396, 1844 and 1580, and 2133 and 1834 m2 g–1 for NPU-1/2/3, respectively. Horvath–Kawazoe (pore geometry: cylinder) and NLDFT model (pore geometry: slit)-based pore size distribution analyses were conducted upon the corresponding N2 sorption isotherms (Figure c and Supplementary Figure 12). A gradually increasing trend in pore size distribution was noticed upon increasing the length of the dicarboxylate ligands (peak position: 0.74 nm in NPU-1, 1.05 nm in NPU-2, and 1.22 nm in NPU-3), in agreement with the crystal structures.

Figure 2

PXRD patterns and sorption data for NPU-1, NPU-2, and NPU-3. (a) Experimental and calculated PXRD patterns. (b) N2 sorption isotherms at 77 K. (c) Calculated pore-size distribution from 77 K N2 sorption data. (d) C2 gas sorption isotherms of NPU-1, (e) NPU-2, and (f) NPU-3 at 298 K. (g) IAST selectivity of C2H6/C2H4 (1:1 ratio) and C2H2/C2H4 (1:1 ratio) at 298 K and 100 kPa. (h) C2 gas adsorption enthalpies at low loading. (i) Optical microscope image of NPU-1 crystals.

The microporosity of NPU-1/2/3 prompted us to collect pure gas sorption isotherms at 273 and 298 K for C2H2, C2H4, and C2H6 up to 100 kPa (Figure d–f, Supplementary Figures 13–15). NPU-1 is the sorbent with the smallest pore size and exhibited the largest adsorption capacity for C2H2, C2H4, and C2H6, followed by NPU-2 (intermediate pore size) and NPU-3 (largest pore size). That the gas uptake at ambient temperatures was observed to increase with decreasing pore size is expected for microporous (<2 nm pores) materials.[50] Notably, at 298 K, NPU-1 was found to exhibit higher uptakes of C2H2 (5.1 mmol g–1) and C2H6 (4.5 mmol g–1) than C2H4 (4.2 mmol g–1) at 100 kPa. This trend is maintained at relatively low pressures (0–30 kPa), implying selective adsorption of C2H2 and C2H6 vs C2H4 across the range of loading in NPU-1. Despite lower C2H2 (4.03 and 2.58 mmol g–1, respectively), C2H4 (3.45 and 2.22 mmol g–1), and C2H6 uptakes (4.44 and 3.36 mmol g–1) for NPU-2 and NPU-3 at 298 K and 100 kPa vs NPU-1, a similar phenomenon for C2H2 and C2H6 adsorption selectivity over C2H4 was observed. This trend also occurred in the 273 K C2 sorption isotherms. The C2H2 and C2H6 uptakes for NPU-1 exceed the values of 4.55/3.79 mmol g–1 for TJT-100 and 3.50/4.47 mmol g–1 for Azole-Th-1.

Adsorption selectivity is an important metric when evaluating separation performance. Ideal Adsorbed Solution Theory[51] (IAST) was used to calculate the adsorption selectivity of C2H2/C2H4 and C2H6/C2H4 for equimolar gas mixtures at 298 K and 100 kPa, after fitting the single-component 298 K adsorption isotherms to the Langmuir–Freundlich equation (Supplementary Figures 16–18). The adsorption selectivity values for C2H6/C2H4 and C2H2/C2H4 were found to be 1.32 and 1.4 for NPU-1, 1.52 and 1.25 for NPU-2, and 3.21 and 1.32 for NPU-3 (Figure g), making the NPU-1/2/3 platform only the third reported example of C2H2 and C2H6 adsorption selectivity vs C2H4. The C2H6/C2H4 IAST selectivity values for NPU-1/2/3 are comparable to those of the other C2H6-selective sorbents (1.95 for MUF-15,[52] 1.8 for IRMOF-8,[53] 1.76 for NUM-7a,[54] and 1.6 for JNU-2(55)).

To evaluate the interaction strengths of C2H2, C2H4, and C2H6 with NPU-1/2/3, the adsorption enthalpies (Qst) were calculated using the Clausius–Clapeyron equation by fitting the C2H2, C2H4, and C2H6 adsorption isotherms at 273 and 298 K to the virial equation (detailed fitting curves are given in Supplementary Figures 22–24). As expected, NPU-1/2/3 interact somewhat more strongly with C2H2 and C2H6 than with C2H4. Figure h presents C2H2 (27.88 kJ mol–1 for NPU-1, 20.98 kJ mol–1 for NPU-2, and 19.93 kJ mol–1 for NPU-3) and C2H6 (29.10 kJ mol–1 for NPU-1, 19.64 kJ mol–1 for NPU-2, and 18.71 kJ mol–1 for NPU-3) adsorption enthalpies relative to C2H4 (23.95 kJ mol–1 for NPU-1, 18.18 kJ mol–1 for NPU-2, and 17.79 kJ mol–1 for NPU-3) at low loadings. Considering the differences between the low C2 loading enthalpies, particularly those for C2H2 and C2H6, the decreasing Qst trend holds across the full C2 loading range: Qst(C2H2) ≈ Qst(C2H6) > Qst(C2H4). Increased adsorption enthalpy for C2 hydrocarbons with decreasing pore size can be attributed to stronger binding in tighter pores.[11,56] We note that a relatively low adsorption enthalpy value means a relatively low energy footprint for sorbent regeneration.

Molecular simulations were performed to gain insight into the nature of the binding sites for C2H2, C2H4, and C2H6 in NPU-1 (see Supporting Information for full details). The modeling studies revealed that C2H2 localizes in a triangular pocket formed by two BDC linkers and one Tripp linker (Figure a). C2H2 molecules interact with electronegative carboxylate O atoms of the BDC linkers through their CH moieties. C2H2 molecules also interact with the aromatic rings of BDC and Tripp linkers. C2H4 was found to adsorb at the same location as C2H2, although its interaction distances are longer (Figure b), an indication that C2H4 is likely to exhibit weaker interactions with NPU-1. The calculated averaged classical potential energies for both adsorbates about this position (see Supporting Information for details) also support weaker interactions with C2H4. In essence, the D2 symmetry of C2H4 offers a less favorable fit compared to the D∞ symmetry of C2H2.

Figure 3

Molecular modeling determined primary adsorption sites of (a) C2H2, (b) C2H4, and (c) C2H6 in NPU-1. Adsorbed C2 molecules are presented in space-filling mode (C(MOM), gray; C(C2 gas), orange; H, white; O, red; N, blue; Mn, purple).

In contrast to C2H2 and C2H4, C2H6 is too bulky to be adsorbed in the triangular channels, and the modeling studies suggested that the most favorable binding site for C2H6 is between two adjacent Tripp linkers (Figure c). Its larger molecular dimensions and the presence of multiple H atoms allow C2H6 to engage in multiple weak interactions with surrounding Tripp linkers. C2H6 can also interact with the phenyl ring of a nearby BDC linker. This binding site is quite favorable, as the potential energy for C2H6 at this site is comparable to that for C2H2 at its energy minimum position on the basis of simulated annealing calculations (see Supplementary Table 3). Overall, molecular simulations support the following trend in sorbent–sorbate interactions for NPU-1: C2H2 ≈ C2H6 > C2H4. This trend is consistent with the experimental findings.

To evaluate separation performance, dynamic column breakthrough experiments were performed for NPU-1 and NPU-2. Two separate columns with tightly packed powder samples of 2.9 g of NPU-1 and 2.5 g of NPU-2 were prepared by pre-activating at 353 K under He flow (flow rate of 20 cm3 min–1). In a typical breakthrough experiment at 298 K, a 1:1:1 mixture of C2H2/C2H4/C2H6 at a total gas pressure of 100 kPa was passed through the packed adsorbent column, and the outlet gas signal was detected by gas chromatography. As shown in Figure a, NPU-1 effectively captured C2H2 and C2H6 from the 1:1:1 gas mixture and afforded polymer-grade C2H4 in the effluent stream. Ethylene breakthrough occurred first at ca. 93 min, followed by C2H2 and C2H6 at ca. 100 min. Before the breakthrough of C2H2 and C2H6 from the column outlet, C2H4 with polymer-grade purity (>99.9%) was harvested, establishing the ability of NPU-1 to produce polymer-grade ethylene from 1:1:1 C2H2/C2H4/C2H6 in a single separation step. Furthermore, dynamic breakthrough experiments with less NPU-1 (1.3 g) exhibited similar separation performance but with slightly reduced working capacity (Supplementary Figure 31). For the NPU-2 column, the breakthrough curve follows the same time sequence (C2H4 first, C2H2 second, and C2H6 last) as NPU-1, but ethylene in high purity was not afforded because of the very close breakthrough times of C2H2 and C2H4. This result was presumably a consequence of the lower adsorption enthalpy of all three gases for NPU-2 vs NPU-1 and the minimal differences between C2H2 and C2H4 isotherms in NPU-2 at lower pressures. NPU-3, with even lower adsorption capacity and enthalpy, was not subjected to dynamic breakthrough experiments. To confirm the ease of regeneration of NPU-1, we conducted temperature-programmed desorption experiments after saturation of the separation column. As seen in Figure c, the NPU-1 column can be fully activated after heating at 313 K for 40 min under He flow (10 cm3 min–1). Upon increasing the activation temperature to 333 K, the regeneration time was reduced to 25 min. To test the recycling ability of NPU-1, 10 cycles of ternary gas mixture breakthrough and single-gas ethane adsorption experiments were conducted. No performance loss was detected. In addition, the PXRD pattern of NPU-1 collected after exposure to ambient air over a month verified its moisture stability.

Figure 4

Dynamic breakthrough data and stability test. Experimental breakthrough curves at 298 K for C2H2/C2H4/C2H6 separation (1:1:1 mixture; total gas pressure of 100 kPa; total gas flow of 2.1 cm3 min–1) based on (a) NPU-1 column with 2.9 g sample and (b) NPU-2 column with 2.5 g sample. (c) Temperature-programmed desorption curves for NPU-1-packed column activated at 313 and 333 K under He flow of 20 cm3 min–1. (d) Dynamic breakthrough data of NPU-1 in the 1st, 5th, and 10th cycles for ternary gas separation (C, outlet gas concentration; C0, inlet gas concentration). (e) PXRD patterns of NPU-1 under different conditions. (f) Ten-cycle comparison of C2H6 adsorption capacity at 298 K and 100 kPa for NPU-1 (n0, C2H6 uptake for first cycle; n, C2H6 uptake for the specific cycle).

Conclusion

In summary, we report a new hexanuclear metal cluster, [Mn6(μ3-O)2(CH3COO)3]6+, that can serve as a node in three porous coordination networks (NPU-1/2/3) with the same pacs topology and high connectivity. Pore size and pore chemistry in the dual cages were controlled by linker ligand substitution, affording a pore environment suitable for effective separation of C2H2 and C2H6 from C2H4. NPU-1 thereby enabled production of polymer-grade ethylene from a 1:1:1 C2H2/C2H4/C2H6 gas mixture at ambient conditions in one step. Sorbent–sorbate interactions, as delineated by molecular simulations, revealed that the dual cage nature of NPU-1 is the reason for the observed performance.