Introduction

Ethylene (C2H4) is a feedstock for the production of plastics, detergents and coatings and its production, now approaching 200 million tons per year, continues to grow1. The energy footprints for purification of C2H4 and propylene (C3H6), also one of the highest volume products of the chemical industry, account for ca. 0.3% of the global energy demand2. The presence of impurities is a consequence of the steam pyrolysis process used to produce C2H4, which in turn results in acetylene (C2H2), carbon dioxide (CO2), ethane (C2H6) and other downstream products including propylene (C3H6), propane (C3H8), hydrogen (H2), C4 and higher light hydrocarbons3,4, which are easily separated by the difference of boiling points. Polymer-grade (>99.95% purity) C2H4 is generated by stepwise removal in downstream purification processes: CO2 is removed using caustic soda; C2H2 is eliminated via catalytic hydrogenation with noble-metal catalysts at high temperature and pressure; cryogenic distillation is typically employed to remove C2H65.

To mitigate the high energy footprint of C2H4 production, researchers have typically focused upon the development of a new generation of physisorbents that exhibit affinity for one gas over others in a gas mixture. The promise of physisorbents lies with their relatively low energy consumption compared to distillation processes thanks to facile regeneration/recycling6. However, physisorbents tend to be unsuitable for ethylene purification because the kinetic diameter of C2H4 (4.1 Å) sits between CO2 (3.3 Å), C2H2 (3.3 Å) and C2H6 (4.4 Å), precluding the possibility of molecular sieving by pore size engineered physisorbents7. Since the quadrupole moment of C2H4 (1.5 × 10−26 esu cm2) also lies between CO2 (4.3 × 10−26 cm2), C2H2 (7.2 × 10−26 esu cm2) and C2H6 (0.65 × 10−26 esu cm2)8, one-step purification of C2H4 by thermodynamics (selective binding) has thus far proven to be elusive. Metal organic materials (MOMs)9, also called metal-organic frameworks (MOFs)10,11 or porous coordination polymers (PCPs)12, have promising applications as C2 and CO2 selective physisorbents for several binary mixtures, including C2H2/C2H4, C2H4/C2H6, C2H6/C2H4, C2H2/CO2 and CO2/C2H213,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34 There are also examples of physisorbents that are effective against ternary C2 and C2-CO2 mixtures such as C2H2/C2H4/C2H6 and C2H2/C2H4/CO235,36,37,38,39,40,41. Certain classes of physisorbents are amenable to systematic fine-tuning of pore chemistry and pore size42,43 and have resulted in “second generation” sorbents with > one order of magnitude improvement in performance18,24,44,45,47. Nevertheless, we are aware of only one report of C2H4 purification from quaternary mixtures using physisorbents as reported by us in 2019 by introducing the synergistic sorbent separation technology (SSST)46 concept, which enables one-step production of high-purity ethylene from a quaternary C2H2-C2H4-C2H6-CO2 mixture by exploiting three benchmark sorbents, Zn-atz-ipa, SIFSIX-3-Ni, and TIFSIX-2-Cu-i. These sorbents were tandem-packed in a single column to sequentially remove C2H6, CO2, and C2H2, respectively (Fig. 1). Unfortunately, the interplay of packing sequence and gas mass transfer in SSST can make industrial-scale processes infeasible and, as illustrated in Fig. 1, a single sorbent that coadsorbs C2H2, C2H6, and CO2 would be desirable for quaternary gas separations such as the purification of C2H4 from a C2H2-C2H4-C2H6-CO2 mixture. Herein we report that the challenge of one-step C2H4 purification from a quaternary mixture (C2H2-C2H4-C2H6-CO2) is achieved by a single physisorbent, the ultramicroporous coordination network Zn-atz-oba.

Fig. 1: Comparison of ethylene purification technologies.
figure 1

State-of-the-art separation technology is compared to synergistic sorbent separation technology (SSST) and “one-sorbent separation technology”, OSST.

Full size image

Results

Synthesis and characterization of Zn-atz-oba

Zn-atz-oba was synthesized following a previously reported protocol (for details see Methods section)47. The bulk phase purity of crystalline samples was confirmed by powder X-ray diffraction, PXRD (Fig. 2c). As shown in Fig. 2a, b, Zn(II) cations are linked by atz− anions to form 2D undulating layers with dinuclear Zn(II) clusters as nodes. These layers are further cross-linked via oba2− ligands to form a pcu topology network. As revealed by thermogravimetric analysis (Supplementary Fig. 1), activated Zn-atz-oba is fully desolvated and stable until ca. 673 K. As calculated by PLATON48, the void space of Zn-atz-oba is 35.9%.

Fig. 2: Pore structure and gas sorption properties of Zn-atz-oba.
figure 2

Views of (a) the pore structure and (b) the Connolly surface of Zn-atz-oba when viewed along the a- and c-axis, respectively. H-atoms are omitted for clarity. (c) PXRD patterns confirm the bulk phase purity of Zn-atz-oba. (d) Adsorption enthalpy profiles (Qst) for Zn-atz-oba. (e) Gas sorption isotherms of Zn-atz-oba at 298 K. (f) Binary 1:1 (v/v) IAST selectivity of Zn-atz-oba at 298 K. color codes for (a) and (b): carbon = gray; nitrogen = blue; oxygen = red; Zn polyhedra in (a) = tyrian purple; Connolly surface in (b) = purple.

Full size image

The pore volume, Langmuir and Brunauer−Emmett−Teller (BET) specific surface area for Zn-atz-oba are 0.287 cm3 g−1, 783.1 m2 g−1, and 710.7 m2 g−1, respectively, as calculated from its cryogenic (77 K) N2 adsorption isotherm (Supplementary Figs. 2–4). Both parameters are in agreement with the crystal structure derived pore volume (0.283 cm3 g−1) and the Langmuir surface area calculated therefrom (755 m2 g−1), respectively. Horvath-Kawazoe model (pore geometry: slit) based pore size distribution analysis revealed the aperture distribution to be ultramicroporous between 0.32 and 0.44 nm (Supplementary Fig. 5).

Adsorption isotherm, selectivity, and enthalpy

The C2H2, C2H4, C2H6, and CO2 adsorption isotherms of activated Zn-atz-oba were recorded at 298 and 273 K (Fig. 2e and Supplementary Fig. 6). Interestingly, C2H2, C2H6, and CO2 exhibited higher adsorption uptakes at 298 K across the entire pressure range tested, 0–100 kPa, especially in the low-pressure region, 0–20 kPa. These uptakes suggest selective adsorption of C2H2, CO2, and C2H6 over C2H4 by Zn-atz-oba. Low-coverage isosteric adsorption enthalpies were determined by virial fit of the isotherm data (see Methods section, Supplementary Note 2: Adsorption enthalpy calculation, and Supplementary Figs. 7–10) and the trends (Fig. 2d) correlate well with their low-pressure (until 20 kPa) saturation uptake capacities as follows: Qst(C2H6) (30.0 kJ mol−1) > Qst(CO2) (29.0 kJ mol−1) > Qst(C2H2) (27.5 kJ mol−1) > Qst(C2H4) (27.0 kJ mol−1). We also note that the adsorption enthalpies of all four gases are below 35 kJ mol−1, an indication that Zn-atz-oba should exhibit a relatively low energy footprint for regeneration27,45.

Adsorption selectivity is also a key indicator of separation performance. The adsorption selectivities of Zn-atz-oba for C2H2/C2H4, C2H6/C2H4, and CO2/C2H4 were calculated using Ideal Adsorbed Solution Theory49 (IAST) after fitting the single-component adsorption isotherms to the Langmuir-Freundlich model (Supplementary Figs. 11–14 and Supplementary Table 1). As shown in Fig. 2f, at 298 K and 100 kPa, the selectivities are 1.43 (C2H2/C2H4), 1.27 (C2H6/C2H4), and 1.33 (CO2/C2H4). C2H6/C2H4 and C2H2/C2H4 selectivities for Zn-atz-oba are comparable to the current benchmark sorbents that enable C2H4 purification from C2H2/C2H4/C2H6 1:1:1 ternary mixtures (1.2 and 1.8 for TJT-10035, 1.46 and 1.09 for Azole-Th-136, 1.32 and 1.4 for NPU-139, 1.4 and 1.07 for UPC-612 and 1.5 and 1.4 for UPC-61341, respectively) (Supplementary Table 3). The selectivity for CO2/C2H4 is also comparable to that of C2H6/C2H4 and C2H2/C2H4. Grand canonical Monte Carlo (GCMC) simulations of binary mixtures containing 1:1 C2H2/C2H4, C2H6/C2H4, and CO2/C2H4 in Zn-atz-oba confirmed that these selectivities are greater than 1 at 298 K and 1 atm (Supplementary Table 5). These selectivities and relatively close adsorption enthalpies suggest that Zn-atz-oba might coadsorb the preferred sorbates C2H2, C2H6, and CO2 when subjected to a mixture feed.

Mechanism of gas adsorption by GCMC simulations

GCMC simulations were conducted upon Zn-atz-oba at 273 and 298 K and afforded C2H4 and CO2 uptakes that are in good agreement with the corresponding experimental measurements at low pressure (≤0.2 atm), but are slightly higher than experimental values at higher pressures (Fig. 3). Moreover, it can be observed that the simulated uptakes for C2H2 and C2H6 significantly overestimate their corresponding experimental values for most of the state points considered. Since polarization contributes to less than 10% of the total energy for simulations of all four gases in Zn-atz-oba (see Supplementary Fig. 21), the overestimation of the theoretical uptakes compared with experiment might be attributable to the partial charges (Supplementary Data 1) and/or the repulsion/dispersion parameters (Supplementary Data 2) that were used for the MOF atoms (for crystallographic distances between MOF atoms see Supplementary Data 3; further details of the modeling study are provided in Supplementary Note 3: Modeling Study). Future work could exploit multiple MOF force field parameters for the simulations and investigate their effect upon the reliability of gas uptakes in this material. Although the simulated uptakes in Zn-atz-oba are inconsistent with experimental data, the simulations still properly predicted that this material exhibits the lowest uptake for C2H4 within the considered pressure range at 273 and 298 K. In addition, according to the simulations, Zn-atz-oba was expected to display higher uptake for C2H6 within the low-pressure region (<0.1 atm) compared to the other three gases. This is also consistent with the finding that this MOF exhibited the highest initial Qst value toward C2H6 according to both experiment and simulation (Fig. 2d and Supplementary Table 4).

Fig. 3: Molecular modeling guided determination of primary adsorption sites and adsorption isotherms in Zn-atz-oba.
figure 3

The primary adsorption sites of (a) C2H2, (b) C2H4, (c) C2H6, and (d) CO2 in Zn-atz-oba. Adsorbed gas molecules are presented in space-filling mode. color codes: C (Zn-atz-oba and CO2): gray; C (C2 gases): orange; H: white; O: red; N: blue; Zn: silver. Experimental (solid lines with circles) and simulated (squares) adsorption isotherms for C2H2 (green), C2H4 (red), C2H6 (blue), and CO2 (black) at (e) 273 K and (f) 298 K and pressures up to 1 atm in Zn-atz-oba.

Full size image

Molecular simulations revealed that the most favorable binding site for all adsorbates lies within the confined region enclosed by four atz linkers and three oba linkers (Fig. 3). In this region, C2H2 interacts with the –NH2 group of atz with N···H–C distances of 2.34 Å (Fig. 3a). C2H4, C2H6, and CO2 exhibit different orientations than C2H2, which allow them to make multiple contacts with the surrounding atoms of the framework. Most sorbent–sorbate interaction distances are longer for C2H4 versus C2H6 and CO2, indicating weaker interactions between C2H4 and the pores of Zn-atz-oba (Fig. 3b–d). The GCMC-calculated initial Qst values were determined and exhibit the following trend: C2H6 > CO2 > C2H2 > C2H4 (see Supplementary Table 4). The greater Qst value for C2H6 versus CO2 in Zn-atz-oba is supported by shorter interactions between C2H6 and Zn-atz-oba, perhaps because of the larger molecular dimensions of C2H6 and repulsive interactions between the negatively charged O atoms of CO2 and the surrounding electronegative N atoms of the atz linkers at the binding site. Further, the close N···H–C interactions exhibited by C2H2 are shorter than those of C2H4. Overall, the modeling studies support the experimental finding that C2H4 forms weaker interactions with Zn-atz-oba than the other three gases.

Breakthrough experiments of Zn-atz-oba

The gas separation performance was determined by dynamic column breakthrough experiments performed with Zn-atz-oba. In a typical breakthrough test at 298 K and ambient pressure, C2H4/C2H6, C2H2/C2H4/C2H6, and C2H2/C2H4/C2H6/CO2 mixtures (equimolar mixtures, total gas pressure = 100 kPa) were passed through a packed column and the effluent streams were monitored by gas chromatography. In accordance with the pure gas sorption results and the molecular simulations derived understanding of the binding sites, the Zn-atz-oba fixed-bed column eluted C2H4 with ultra-high purity in a single step from all three gas mixtures. As shown in Fig. 4a–c, C2H4 breaks through first at ca. 236, 177, and 130 min, respectively, and the impurities (C2H6, C2H2, and CO2) break through 14, 12, and 10 min later, respectively. Before breakthrough of C2H6, C2H4 of polymer grade purity (>99.95%) was collected at the outlet, revealing that Zn-atz-oba achieved one-step purification of C2H4 by the coadsorption of C2H2, C2H6, and CO2 from C2H2/C2H4/C2H6/CO2 (>99.95% C2H4 productivity, 0.106 mmol g−1). To explore the recycling performance of Zn-atz-oba, ten cycles of four-component breakthrough experiments were conducted. There was no loss of C2H4 retention time (Fig. 4d). In industrial C2 hydrocarbon gas streams, C2H2 only constitutes ~1% of the total flow50. We also tested a C2H2/C2H4/C2H6/CO2 (1/33/33/33) mixture under dynamic breakthrough using 7.4 g Zn-atz-oba packed in a fixed-bed. As shown in Fig. S25, polymer-grade ethylene was harvested from 1/33/33/33 gas mixture, followed by C2H6, CO2, and C2H2. As a typical contaminant in downstream feedstocks, water vapor is present in the industrially produced C2-CO2 stream51, so breakthrough experiments in the presence of water vapor were conducted. As revealed by Fig. 4e, fractional pressure reduction of C2H2, C2H4, C2H6, and CO2 in the presence of water vapor reduced the adsorption capacity of Zn-atz-oba. This is reflected in a reduced C2H4 breakthrough time, from c.a. 170 min to c.a. 150 min. 4-component C2H2/C2H4/C2H6/CO2 (1:1:1:1) breakthrough experiments in the presence of water vapor (relative humidity 52%) also resulted in a reduced retention time but a consistent gas outflow sequence. Nevertheless, C2H4 with effluent purity >99.95% was collected at the column outlet. We attribute the shorter retention time under 52% humidity to the presence of water vapor in the breakthrough pipeline and competing sorption of water with the other four gases. After ambient air exposure for 30 days or water immersion for 10 days, PXRD patterns (Fig. 4f) and N2 isotherms at 77 K (including pore distribution, Supplementary Figs. 26 and 27) of regenerated Zn-atz-oba indicated that it is stable to both humid air and liquid water. Temperature-programmed desorption of the fully adsorbed column was conducted after achieving full saturation and revealed that the Zn-atz-oba bed could be regenerated within 50 min at 333 K under He flow (30 cm3 min−1). Even under ambient conditions (298 K and 100 kPa), regeneration was achieved within 140 min under He flow of 20 cm3 min−1 (Supplementary Fig. 23), confirming that there is indeed a relatively low energy footprint for Zn-atz-oba regeneration.

Fig. 4: Dynamic breakthrough experiments and stability tests.
figure 4

(a)–(c) Experimental breakthrough curves at 298 K for C2H4/C2H6 (1:1) (a), C2H2/C2H4/C2H6 (1:1:1) (b), and C2H2/C2H4/C2H6/CO2 (1:1:1:1) (c) separations (equimolar mixtures; total gas pressure 100 kPa; total gas flow 1.4, 2.1 and 2.8 cm3 min−1, respectively) based on Zn-atz-oba (6.3 g) packed column (C: Outlet gas concentration, C0: Inlet gas concentration). (d) Dynamic breakthrough data obtained with Zn-atz-oba fixed-bed in 1st, 5th, and 10th cycle when subjected to four-component C2H2/C2H4/C2H6/CO2 (1:1:1:1) mixture. e Four-component (1:1:1:1) breakthrough experiment in the presence of water vapor at 288 K and 100 kPa based on Zn-atz-oba (7.0 g) packed column (total gas flow of 2.8 cm3 min−1). (f) PXRD patterns of Zn-atz-oba after being subjected to air exposure and water immersion stability test conditions.

Full size image

Discussion

If one compares the performances of previously reported physisorbents in the context of ethylene purification from binary mixtures15,16,17,18,19,20,22,23,24,25,26,27,28,29,30,32,46 (C2H2/C2H4, C2H6/C2H4, and C2H4/CO2) and ternary mixtures35,36,39,40,41 (C2H2/C2H4/C2H6, C2H2/C2H4/CO2, and C2H6/C2H4/CO2), it is evident that Zn-atz-oba represents a benchmark in terms of its performance parameters (Fig. 5). Specifically, Zn-atz-oba coadsorbs three gases, C2H2, C2H6, and CO2, to produce polymer-grade (>99.95%) C2H4 in just one step. We attribute this performance to the unusual pore environment in Zn-atz-oba, which enables roughly equal affinity towards C2H2, C2H6, and CO2 over C2H4. Future studies will focus upon crystal engineering of second-generation variants to further improve the purification performances of other gas mixtures of industrial relevance.

Fig. 5: Physisorbents studied for C2H4 purification.
figure 5

Comparison of physisorbents that can purify C2H4 from binary (sky blue)15,16,17,18,19,20, 22,23,24,25,26,27,28,29,30, 32, 46, ternary (olive green)35, 36, 39,40,41 and quaternary mixtures (lilac, this work) under ambient conditions.

Full size image

Methods

General

All reagents were obtained from vendors and used as received without further purification. Powder X-ray diffraction (PXRD) data were collected with a Rigaku-Miniflex-600 diffractometer at a scanning rate of 5° min−1 using Cu Kα radiation. Thermogravimetric analysis (TGA) data were obtained using Mettler TG DSC 3+ Thermogravimetric Analyzer. In TGA measurements, the sample was heated from 298 K to 1073 K under Ar atmosphere at a heating rate of 10 K min−1. The morphology was examined using scanning electron microscopy (FESEM, Verios G4, FEI).

Synthesis of Zn-atz-oba

A mixture of Zn(NO3)2·6H2O (10 mmol, 2.98 g), H2oba (5 mmol, 1.29 g), Hatz (10 mmol, 0.84 g), DMF (40 mL), MeOH (40 mL), and H2O (20 mL) was sealed in a 250 ml borosilicate bottle and sonicated for five minutes. The reaction mixture was then solvothermally treated at 403 K for 3 days before naturally cooling it to room temperature. The as-synthesized sample Zn-atz-oba was washed with fresh DMF, MeOH, and H2O, and exchanged with MeOH three times daily for three days and then activated at 353 K in vacuum for 12 h.

Single-component gas sorption experiments

A Micromeritics 3Flex was used for recording all gas sorption isotherms, N2 (77 K), C2H2, C2H4, C2H6, and CO2 (each, 273 and 298 K). For N2 adsorption isotherms, the temperature was controlled at 77 K using a Dewar containing 4 L liquid N2. Precise control of 273 and 298 K temperatures was implemented by a dc-2006 from Ningbo Scientz Biotechnology, which contained a cyclic control system of ethylene glycol and water mixture (v/v = 1:1). Zn-atz-oba was degassed at 298 K under high vacuum for 4 h to regenerate in between consecutive isotherm measurements.

Dynamic gas breakthrough experiments

Breakthrough curves were recorded by an in-house custom-built rig (Supplementary Fig. 22). Equimolar C2H6/C2H4 (1:1), C2H2/C2H4/C2H6 (1:1:1) and C2H2/C2H4/C2H6/CO2 (1:1:1:1) gas mixtures (total gas pressure and flow: 100 kPa and 1.4, 2.1, and 2.8 cm3 min−1, respectively) were subjected through the Zn-atz-oba packed column (6.3 g) at 298 K, and the outlet gas concentrations and composition were monitored by a gas chromatography analyzer (Carrier gas: He, TCD-Thermal Conductivity Detector, detection limit 0.1 ppm). During gas breakthrough cycling tests, Zn-atz-oba packed in the column was regenerated under He flow of 30 cm3 min−1 at 333 K for 2 h, after each breakthrough experiment.