Report

Synergistic sorbent separation for one-step ethylene purification from a four-component mixture

See allHide authors and affiliations

Science  11 Oct 2019:
Vol. 366, Issue 6462, pp. 241-246
DOI: 10.1126/science.aax8666

Selecting for ethylene

Purification of ethylene from other gases produced during its synthesis, such as acetylene, ethane, and carbon dioxide, is an energy-intensive process. Chen et al. use a mixture of microporous metal-organic framework physisorbents that are selective for one of these four gases. A series of sorbents in a packed-bed geometry produced ethylene pure enough for making polymers.

Science, this issue p. 241

Abstract

Purification of ethylene (C2H4), the largest-volume product of the chemical industry, currently involves energy-intensive processes such as chemisorption (CO2 removal), catalytic hydrogenation (C2H2 conversion), and cryogenic distillation (C2H6 separation). Although advanced physisorbent or membrane separation could lower the energy input, one-step removal of multiple impurities, especially trace impurities, has not been feasible. We introduce a synergistic sorbent separation method for the one-step production of polymer-grade C2H4 from ternary (C2H2/C2H6/C2H4) or quaternary (CO2/C2H2/C2H6/C2H4) gas mixtures with a series of physisorbents in a packed-bed geometry. We synthesized ultraselective microporous metal-organic materials that were readily regenerated, including one that was selective for C2H6 over CO2, C2H2, and C2H4.

Purification of commodities currently consumes 15% of global energy, and commodity demand has been projected to triple by 2050. The production of ethylene (C2H4) and propylene (C3H6) uses 0.3% of global energy production (1) as polymer-grade (>99.9% purity) C2H4 is produced by energy-intensive separation of downstream C2 hydrocarbon gas mixtures during the steam cracking process. Acetylene (C2H2) is removed through catalytic hydrogenation (with a noble-metal catalyst at high temperature and pressure) or solvent extraction (requiring a large volume of solvent and a large plant installation). Removal of C2H6 occurs through cryogenic distillation (2).

The high energy footprint associated with C2H4 production has spurred research into the development of more energy-efficient approaches to purification of these C2 gases. To afford polymer-grade C2H4 in a single step, simultaneous removal of C2H2 and C2H6 from C2H4 would be necessary. Chemical transformation of C2H2 and C2H6 to C2H4, chemisorption, extraction, and membrane-based technologies could in principle address the need, but each approach has drawbacks. The simultaneous separation of C2H2, C2H6, and other trace impurities from C2H4 with a physisorbent could enhance the energy efficiency of C2H4 production but is too challenging for classical physisorbents. The fundamental limitation lies in the respective quadrupole moments and kinetic diameters of the C2 gases; C2H4 (1.5 × 10−26 esu cm2 and 4.1 Å) lies just between C2H2 (7.2 × 10−26 esu cm2 and 3.3 Å) and C2H6 (0.65 × 10−26 esu cm2 and 4.4 Å), which precludes most physisorbents from being highly selective (3).

The task is compounded when CO2 (4.3 × 10−26 esu cm2 and 3.3 Å) is also an impurity in a C2 gas mixture, because a physisorbent would then require strong affinity toward C2H2, C2H6, and CO2 versus C2H4. Metal-organic materials, also known as metal-organic frameworks (MOFs) or porous coordination polymers, have gained attention for gas separations because of their tunability over pore size and pore chemistry (49). Several studies have addressed the separation of C2H2, C2H4, C2H6, and CO2 by physisorbents (e.g., zeolites, activated carbon, porous organic frameworks, and MOFs) (1017), but selectivity that realizes polymer-grade C2H4 production from a quaternary gas mixture (C2H2-C2H4-C2H6-CO2) is still an important goal. Indeed, even for the ternary C2H2-C2H4-C2H6 mixture, only a recent report by Lu and co-workers found discrimination toward C2H4 over C2H2 and C2H6 in the MOF material TJT-100 (18). The task is further compounded for wet gas streams because water vapor can interfere with physisorbent performance through co-adsorption or hydrolytic degradation (19). This would necessitate pretreatment of a gas mixture with a desiccant material.

We addressed this separation challenge by developing synergistic sorbent separation technology (SSST), which uses the favorable sorption properties of task-specific ultramicroporous physisorbents, each with ultrahigh selectivity for one of the impurities, to enable one-step production of C2H4 from C2 gas mixtures (Fig. 1A). Such an approach was used to address the desulfurization of hydrocarbons, but sulfur impurities with strong polarity are much easier to isolate from hydrocarbons (20). With respect to C2H2 and C2H6 removal from C2H4 streams, TIFSIX-2-Cu-i (TIFSIX = TiF62–, 2 = 4,4′-dipyridylacetylene, i = interpenetrated) offers ultraselective C2H2 capture (21, 22) and, as shown below, Zn-atz-ipa (atz = 3-amino-1,2,4-triazolate; ipa = isophthalate) (23), exhibits strong affinity for C2H6 over C2H4 and the other impurities. In principle, a combination of these two physisorbents could synergistically capture C2H2 and C2H6 in a tandem-packed sorbent bed to produce polymer-grade (>99.9%) C2H4 via physisorption. Further, the addition of SIFSIX-3-Ni (SIFSIX = SiF62–, 3 = pyrazine), an ultraselective CO2 sorbent (19, 24), to the sorbent bed should enable removal of trace levels of CO2 from the corresponding four-component gas mixture, thereby offering a one-step process that offers advantages over present approaches (Fig. 1). The selected physisorbents are stable to humidity, but their sorption performance is reduced in the presence of water vapor (19, 25).

Fig. 1 Synergistic sorbent separation technology (SSST) versus present approaches to purify C2H4.

(A) SSST involves an adsorption bed with three task-specific physisorbents to purify the commodity (red) with specific binding sites for each trace impurity (blue, green, yellow). (B) The present process for producing polymer grade C2H4 involves three energy-intensive steps.

TIFSIX-2-Cu-i and SIFSIX-3-Ni belong to a family of hybrid ultramicroporous materials of general formula M′FSIX-L-M (M = divalent transition metal center; L = dipyridyl organic linker; M′ = Si, Ti, Ge, Zr, Sn). In this family, square lattice networks are formed by transition metal nodes and bipyridyl-type linkers, which are pillared by inorganic linkers to generate primitive cubic topology coordination networks with tunable pore size and chemistry (21, 24). TIFSIX-2-Cu-i and SIFSIX-3-Ni have previously been shown to exhibit benchmark performance for trace C2H2 (21) and trace CO2 sorption, respectively (19, 26), but their pore size and chemistry renders them ill-suited for C2H6-selective sorption. Indeed, we are unaware of any existing sorbents that are even mildly selective toward C2H6 over CO2, C2H2, and C2H4 (table S1).

The ultramicroporous MOF Zn-atz-ipa was selected in this context given its excellent water stability and unusual pore chemistry (23); as detailed below, this pore chemistry makes it suitable for the intended purpose. To evaluate the selected sorbents for C2H4 purification, we first determined their pure gas adsorption properties. Each sorbent was synthesized according to previous reports (19, 22, 23). To verify purity, we collected powder x-ray diffraction (XRD) patterns and sorption data at cryogenic temperatures on as-synthesized materials after activation (figs. S1 and S2). Single-gas isotherms at 273 and 298 K were collected to 1 bar for TIFSIX-2-Cu-i, SIFSIX-3-Ni, and Zn-atz-ipa (figs. S3 to S5). As shown in Fig. 2B, at 298 K and 1 bar, TIFSIX-2-Cu-i exhibited less uptake for C2H6 (2.1 mmol/g) than for C2H4 (2.6 mmol/g), CO2 (4.3 mmol/g), and C2H2 (4.1 mmol/g). C2H2 exhibited the highest uptake from 0 to 0.8 bar for TIFSIX-2-Cu-i. In the case of SIFSIX-3-Ni, CO2 exhibited the highest uptake at 298 K below 0.2 bar (Fig. 2C). For Zn-atz-ipa, all four gases showed similar uptake (1.8 to 2.0 mmol/g) at 1 bar and 298 K (Fig. 2A). However, from 0 to 0.4 bar, we measured higher uptake for C2H6 over CO2, C2H2, and C2H4.

Fig. 2 Structures of physisorbents used herein and their single component gas sorption properties.

(A to C) Adsorption of CO2 (black squares), C2H2 (green stars), C2H4 (red triangles), and C2H6 (blue circles) at 298 K for Zn-atz-ipa (A), TIFSIX-2-Cu-i (B), and SIFSIX-3-Ni (C). (D to F) Structures of Zn-atz-ipa (D), TIFSIX-2-Cu-i (E), and SIFSIX-3-Ni (F). (G) The Connolly surface of Zn-atz-ipa with probe radius of 2.0 Å. (H) Isosteric heat of CO2 (black), C2H2 (green), C2H4 (red), and C2H6 (blue) in SIFSIX-3-Ni, Zn-atz-ipa, and TIFSIX-2-Cu-i. The “strong binding” threshold of 40 kJ/mol is highlighted with a dashed line. (I) Selectivity for adsorbates in SIFSIX-3-Ni (CO2 over C2H2/C2H4/C2H6), Zn-atz-ipa (C2H6 over C2H4/C2H2/CO2), and TIFSIX-2-Cu-i (C2H2 over CO2/C2H4/C2H6) at 298 K and 1 bar for 1:1 gas mixtures from ideal adsorbed solution theory.

To quantify the strength of sorbent-sorbate interactions, we fit 273 and 298 K sorption data by the virial equation (figs. S6 to S11) and calculated the isosteric heat of adsorption (Qst) according to the Clausius-Clapeyron equation. Qst values at low loading of the four gases in TIFSIX-2-Cu-i, SIFSIX-3-Ni, and Zn-atz-ipa are compared in Fig. 2H. Full Qst curves for the four gases in the three ultramicroporous sorbents are given in figs. S12 to S14 and summarized in table S2. Each sorbent exhibited strong selectivity for one gas over the other three according to Qst: CO2@SIFSIX-3-Ni (50.9 kJ/mol), C2H6@Zn-atz-ipa (45.8 kJ/mol), and C2H2@TIFSIX-2-Cu-i (46.3 kJ/mol) (Fig. 2H, dashed line). These results indicate that, at least in principle, CO2, C2H2, and C2H6 in a four-component gas mixture including C2H4 should be selectively captured by SIFSIX-3-Ni, TIFSIX-2-Cu-i, and Zn-atz-ipa, respectively. The pure gas sorption performance of the three selected sorbents therefore met the needed criteria for a one-step SSST process.

The gas adsorption selectivity at 298 K and 1 bar was calculated for pairs of adsorbates in 1:1 gas mixtures using ideal adsorbed solution theory (27, 28). Detailed fitting parameters are provided in figs. S15 to S20 and tables S3 to S5. As shown in Fig. 2I, TIFSIX-2-Cu-i and SIFSIX-3-Ni exhibited high adsorption selectivity for C2H2 and CO2, respectively, over the other three gases. Indeed, new benchmark selectivity values were found under these or similar conditions (13, 29): C2H2/C2H4 (49) and C2H2/C2H6 (98) in TIFSIX-2-Cu-i; CO2/C2H4 (103) and CO2/C2H6 (308) in SIFSIX-3-Ni. High selectivity was also calculated for CO2/C2H2 (6.1 for C2H2/CO2 by TIFSIX-2-Cu-i, 6.9 for CO2/C2H2 by SIFSIX-3-Ni) (22). Zn-atz-ipa exhibited selective C2H6 adsorption with a selectivity of 1.7 for C2H6/C2H2, 2 for C2H6/C2H4, and 5 for C2H6/CO2.Thus, although Zn-atz-ipa is a physisorbent, it exhibited selective adsorption of C2H6 over C2H4, C2H2, and CO2; we consider this result unexpected because C2H6 tends to be weakly adsorbed as a consequence of its low quadrupole moment.

To better understand the interactions of the four gases with the three sorbents, we conducted grand canonical Monte Carlo (GCMC) simulations. Final optimized results for SIFSIX-3-Ni and TIFSIX-2-Cu-i for C2H2 and CO2 were consistent with earlier reports (21, 30). Binding-site information for all gases is given in Fig. 3 and figs. S21 to S24. In SIFSIX-3-Ni, CO2 binding is driven by interactions with four electronegative F atoms from four independent SiF62– anions. C2H2 was trapped through multiple C-H···F interactions with H···F distances of 3.3 to 4.5 Å between C2H2 and eight SiF62– anions. In contrast, C2H4 and C2H6 exhibited simultaneous interactions with two and six SiF62– anions, respectively. Although there are fewer contacts with anions, shorter distances of 2.51 and 2.62 Å for C2H4 suggest favorable C2H4 binding over C2H6 (2.59 to 2.76 Å).

Fig. 3 Molecular simulation and periodic density functional theory calculations.

(A to L) C2H2 [(A), (E), and (I)], C2H4 [(B), (F), and (J)], C2H6 [(C), (G), and (K)], and CO2 [(D), (H), and (L)] binding sites in SIFSIX-3-Ni (top), TIFSIX-2-Cu-i (middle), and Zn-atz-ipa (bottom). Closest contacts between framework atoms and gas molecules are defined by the distance (in angstroms) between the H atom of hydrocarbons and the closest framework atoms. Adsorbed C2 and CO2 molecules are presented in space-filling display mode (C, gray; H, white; O, red; N, blue; F, cyan; Si, yellow; Ni, lavender; Ti, silver; Cu, gold; Zn, dark gray).

The overall trend of adsorption energy from calculations, CO2 > C2H2 > C2H4 > C2H6, is fully consistent with experimental data. In TIFSIX-2-Cu-i, C2H2, C2H4, and C2H6 are localized so that every molecule could interact with two TiF62– anions through C-H···F interactions. However, C2H2 had shorter contacts (2.46 and 2.50 Å) relative to C2H4 (2.45 and 2.52 Å) and C2H6 (2.62 and 2.90 Å). Moreover, the more acidic C2H2 molecule (pKa = 26, versus C2H4, pKa = 45, and C2H6, pKa = 62) would be expected to form stronger hydrogen bonds. For TIFSIX-2-Cu-i, CO2 molecules interact with two F atoms from one TiF62– anion, with short interaction distances between the C atom of CO2 and the F atoms of the TiF62– anion (2.65 and 3.48 Å). The calculated hierarchy in TIFSIX-2-Cu-i was C2H2 > CO2 > C2H4 > C2H6.

In Zn-atz-ipa, all six H atoms of one C2H6 molecule interacted with the pore surface. This tight-fitting binding site helps to explain the high adsorption energy of 42.2 kJ/mol from GCMC calculations, a value near the experimental value of 45.8 kJ/mol. In contrast, smaller molecules such as C2H4, C2H2, and CO2 only interacted through two or three close contacts, and, although amino groups are considered to improve CO2 binding, the amino group of the atz ligand was not exposed on the pore surface. Thus, CO2 binding was weak and the strength of interactions in Zn-atz-ipa followed the sequence C2H6 > C2H4 > C2H2 > CO2. (See supplementary materials for more details on these simulations.)

We conducted dynamic breakthrough experiments at 298 K on a custom-built apparatus (fig. S25) using an equimolar three-component gas mixture of C2H2/C2H4/C2H6 at a total pressure of 1 bar. All sorbents were preactivated by heating under high vacuum before preparing the breakthrough column. We first conducted control experiments using sorbent beds of TIFSIX-2-Cu-i or Zn-atz-ipa. C2H2 was selectively captured, but C2H4 and C2H6 were not separated by TIFSIX-2-Cu-i (Fig. 4A). For Zn-atz-ipa (Fig. 4B), C2H6 was selectively adsorbed for ~10 min, but C2H2 and C2H4 were not separated. However, SSST with a two-component (tandem) bed containing TIFSIX-2-Cu-i and Zn-atz-ipa cleanly removed C2H2 and C2H6 with C2H4 at >99.9% purity in the effluent stream (Fig. 4C). By increasing the mass ratio of Zn-atz-ipa over TIFSIX-2-Cu-i from 1:1 to 10:1, breakthrough times of C2H2 and C2H6 were optimized for the production of pure C2H4 using SSST (Fig. 4E and figs. S26 to S29), which suggests that the adsorption capacities of the two adsorbents had been fully used in the case of the 10/1 ratio.

Fig. 4 Experimental column breakthrough results.

(A and B) Experimental column breakthrough curves for C2H2/C2H4/C2H6 separation (1:1:1 mixture) on TIFSIX-2-Cu-i and Zn-atz-ipa at 298 K and 1 bar. Breakthrough experiments were conducted in a column (inside diameter, 8 mm) at a flow rate of 2.1 ml/min. (C) Experimental column breakthrough curves for C2H2/C2H4/C2H6 separation (1:1:1 mixture) on a tandem-packed column of TIFSIX-2-Cu-i (~250 mg) and Zn-atz-ipa (~600 mg) at 298 K and 1 bar. The x axis is displayed as minutes per gram of TIFSIX-2-Cu-i + Zn-atz-ipa. The orange dashed line highlights the cutoff time for C2H4 with purity >99.9% in this and other plots. (D) Experimental column breakthrough curves for C2H2/C2H4/C2H6 separation (1:49.5:49.5 mixture) on a tandem-packed column of TIFSIX-2-Cu-i (~120 mg) and Zn-atz-ipa (~1200 mg). (E) SSST sorption beds. From left to right: 1:1 to 1:10 TIFSIX-2-Cu-i + Zn-atz-ipa; 1:1.25:10 TIFSIX-2-Cu-i + Zn-atz-ipa + SIFSIX-3-Ni; physical mixture of TIFSIX-2-Cu-i + Zn-atz-ipa after breakthrough experiments. (F) Experimental column breakthrough curves for CO2/C2H2/C2H4/C2H6 separation (1:1:1:1 mixture) on a tandem-packed column of TIFSIX-2-Cu-i (~120 mg), SIFSIX-3-Ni (~150 mg), and Zn-atz-ipa (~1200 mg) at 298 K and 1 bar (packing order: SIFSIX-3-Ni@Zn-atz-ipa@TIFSIX-2-Cu-i). (G) The effect of packing order of the SSST sorbents on ethylene purity. (H) Temperature-programmed desorption curves recorded on the column in (F) at 60°C under He flow of 20 cm3/min.

The four-component equimolar mixture of C2H2/C2H4/C2H6/CO2 was studied using SIFSIX-3-Ni in a three-component sorbent bed. A ratio of 1:1.25:10 (TIFSIX-2-Cu-i, 120 mg; Zn-atz-ipa, 1.2 g; SIFSIX-3-Ni, 150 mg) was adopted on the basis of the single-component sorption data. Breakthrough results revealed that CO2, C2H6, and C2H2 were captured (Fig. 4F), producing polymer-grade C2H4 as effluent (working capacity 0.14 mmol/g). The breakthrough sequence follows the order C2H4/C2H6/CO2/C2H2 at 20.3, 24.6, 24.9, and 28.3 min, respectively. Regeneration of the SSST column under He flow (20 ml/min, 1 hour, 60°C) revealed unchanged performance after nine cycles (fig. S30). Tests on the individual sorbents showed facile regeneration with no capacity loss after 10 cycles (figs. S49 to S51). The low energy footprint of the SSST columns was validated by temperature-programmed desorption experiments (Fig. 4H and figs. S52B to S59B).

In industrial C2 hydrocarbon gas streams, C2H2 typically makes up only ~1% of the total flow. To examine the performance of SSST with more industrially relevant and challenging gas mixtures, we also tested C2H2/C2H4/C2H6 (1:49.5:49.5) and C2H2/C2H4/C2H6/CO2 (1:33:33:33) gas mixtures. Polymer-grade C2H4 with working capacities of 0.32 and 0.10 mmol/g was harvested from 1:49.5:49.5 and 1:33:33:33 gas mixtures, respectively (Fig. 4D and fig. S31). The uptake of C2H6 revealed by its pure gas isotherm with Zn-atz-ipa at 0.495 versus 0.33 bar contributed to the higher working capacity of the 1:49.5:49.5 gas mixture.

The effect of packing order on SSST performance was assessed with six parallel SSST columns and breakthrough experiments with a 1:1:1:1 gas mixture at 298 K and 1 bar. The results (table S6) revealed the importance of packing order (Fig. 4G and figs. S55A to S59A). SIFSIX-3-Ni@Zn-atz-ipa@TIFSIX-2-Cu-i afforded the highest working capacity (0.14 mmol/g), whereas two other combinations, both with Zn-atz-ipa as the final sorbent (SIFSIX-3-Ni@TIFSIX-2-Cu-i@Zn-atz-ipa and TIFSIX-2-Cu-i@SIFSIX-3-Ni@Zn-atz-ipa), failed (Fig. 4H). The effects of different selectivity values, kinetics, and co-adsorption (22) were likely the cause of this observation. Particle size and amount of sorbent had little effect, with smaller particle size and larger sample amounts resulting in slightly improved C2H4 purification (figs. S33 to S47). A column with looser packing, however, offered much-reduced performance (fig. S48). The use of a physical mixture also failed. When we used 120 mg of TIFSIX-2-Cu-i and 1200 mg of Zn-atz-ipa on an equimolar gas mixture of C2H2/C2H4/C2H6, C2H2 was not effectively removed before C2H4 breakthrough (fig. S32). Further, C2H6 concentration was not reduced to the required specification of <0.1%.

We were able to take advantage of the variations in pore geometry and pore chemistry of three ultramicroporous sorbents to address one-step C2H4 purification using SSST. The choice of task-specific ultraselective sorbents in tandem-packed sorbent beds of the type used here is unlikely to be limited to the three sorbents or target gas we investigated. Sorbents with higher selectivity, higher uptake capacity, or both could likely be substituted to optimize overall performance. The strong performance of SSST with respect to the purification of C2 gas mixtures and the availability of an ever-increasing number of ultraselective physisorbents suggests that the scope of SSST is likely to be broad enough to address the high energy footprint of other industrial commodity purifications.

Supplementary Materials

science.sciencemag.org/content/366/6462/241/suppl/DC1

Materials and Methods

Figs. S1 to S59

Tables S1 to S6

References (3144)

References and Notes

Acknowledgments: We thank J.-W. Cao and S. Sanda for synthesis of some samples and particle size analysis, respectively; T. Curtin (UL) for use of her dynamic breakthrough equipment; and the Analytical and Testing Center of Northwestern Polytechnical University for access to the PXRD testing facility. Funding: Supported by Science Foundation Ireland awards 13/RP/B2549 and 16/IA/4624 (M.J.Z.); National Natural Science Foundation of China grant 21805227 and Fundamental Research Funds for the Central Universities grant 3102017jc01001 (K.-J.C.); NSF grant DMR-1607989, including support from the Major Research Instrumentation Program (award CHE-1531590) (T.P., K.A.F., and B.S.); and ACS Petroleum Research Fund grant 56673-ND6 (B.S.). Computational resources were made available by XSEDE grant TG-DMR090028 and by Research Computing at the University of South Florida. Author contributions: M.J.Z. and K.-J.C. designed the experiments. K.-J.C., D.G.M., J.K., T.P., B.S., and M.J.Z. co-wrote the paper. K.-J.C. and A.K. synthesized compounds. K.-J.C. performed the gas adsorption experiments and data analysis. D.G.M. and S.M. conducted dynamic breakthrough experiments. S.M. conducted sorption cycling and temperature programmed desorption experiments. T.P., K.A.F., and B.S. conducted molecular simulation. experiments. All authors discussed the results and commented on the manuscript. Competing interests: K.-J.C., D.G.M., S.M., A.K., and M.J.Z. are inventors on patent application EP19197407.0 submitted by University of Limerick that covers the use of ultramicroporous sorbents for one-step purification of gas mixtures. Data and materials availability: All data are available in the manuscript and supplementary materials.
View Abstract

Stay Connected to Science

Navigate This Article