Efficient Propylene/Ethylene Separation in Highly Porous Metal–Organic Frameworks

Light olefins are important raw materials in the petrochemical industry for the production of many chemical products. In the past few years, remarkable progress has been made in the synthesis of light olefins (C2–C4) from methanol or syngas. The separation of light olefins by porous materials is, therefore, an intriguing research topic. In this work, single-component ethylene (C2H4) and propylene (C3H6) gas adsorption and binary C3H6/C2H4 (1:9) gas breakthrough experiments have been performed for three highly porous isostructural metal–organic frameworks (MOFs) denoted as Fe2M-L (M = Mn2+, Co2+, or Ni2+), three representative MOFs, namely ZIF-8 (also known as MAF-4), MIL-101(Cr), and HKUST-1, as well as an activated carbon (activated coconut charcoal, SUPELCO©). Single-component gas adsorption studies reveal that Fe2M-L, HKUST-1, and activated carbon show much higher C3H6 adsorption capacities than MIL-101(Cr) and ZIF-8, HKUST-1 and activated carbon have relatively high C3H6/C2H4 adsorption selectivity, and the C2H4 and C3H6 adsorption heats of Fe2Mn-L, MIL-101(Cr), and ZIF-8 are relatively low. Binary gas breakthrough experiments indicate all the adsorbents selectively adsorb C3H6 from C3H6/C2H4 mixture to produce purified C2H4, and 842, 515, 504, 271, and 181 cm3 g−1 C2H4 could be obtained for each breakthrough tests for HKUST-1, activated carbon, Fe2Mn-L, MIL-101(Cr), and ZIF-8, respectively. It is worth noting that C3H6 and C2H4 desorption dynamics of Fe2Mn-L are clearly faster than that of HKUST-1 or activated carbon, suggesting that Fe2M-L are promising adsorbents for C3H6/C2H4 separation with low energy penalty in regeneration.


Introduction
Light olefins, particularly ethylene (C 2 H 4 ) and propylene (C 3 H 6 ), are important raw materials in the petrochemical industry for the manufacture of products such as plastics, solvents, cosmetics, paints, and drugs. Light olefins are traditionally produced from the thermal or catalytic cracking of crude oil. In the past decades, considerable attention has been paid to the production of light olefins (C 2 = -C 4 = ) from other alternative feedstocks, such as coal, natural gas, and biomass, by the Fischer-Tropsch synthesis (FTS) and the methanol-to-olefins (MTO) reaction [1,2]. Some recent breakthrough results showed that coal-and biomass-derived syngas (a mixture of carbon monoxide and hydrogen) could be converted to light olefins in very high selectivities (up to 80%, relative to less valuable saturated hydrocarbons) by advanced catalysts [3,4]. These new innovations would lead to a high demand for efficient separation technology for light olefins in the future. Nowadays, the separation of C 2 H 4 and C 3 H 6 is commonly accomplished by cryogenic distillation, which is a mature but energy-intensive process. Many works have been reported to develop new materials and technologies to efficiently separate C 2 H 4 and C 3 H 6 [5][6][7][8][9]. Adsorptive separation of C 2 H 4 and C 3 H 6 with porous materials is regarded as a promising alternative [10].

Single-Crystal X-ray Diffraction
Single crystals of Fe 2 M-L were picked for single-crystal diffraction experiments. The diffraction data were collected at 100 K with a Rigaku Supernova CCD diffractometer (Tokyo, Japan) equipped with a mirror-monochromatic enhanced Cu-Kα radiation (λ = 1.54184 Å). The dataset was corrected by empirical absorption correction using spherical harmonics, implemented in the SCALE3 ABSPACK scaling algorithm. The structure was solved by direct methods and refined by full-matrix least-squares on F 2 with anisotropic displacement using the SHELXTL software package (Göttingen, Germany) [48]. Nonhydrogen atoms on the frameworks were refined with anisotropic displacement parameters during the final cycles. The hydrogen atoms on the ligands were positioned geometrically and refined by using a riding model. The electron density of the disordered guest molecules in Fe 2 M-L was flattened by using the SQUEEZE routine of PLATON (Utrecht, The Netherlands) [49]. The graphical representations of single-crystal structures of Fe 2 M-L were performed by the Diamond software (Bonn, Germany) [50]. Some strong residual Q peaks were found near the water molecules coordinated with the metal ions in the final refinement cycles, indicating that partially coordinated solvent molecules may be DMF molecules rather than water molecules, which could not be fully modeled. The singlecrystal structure data of Fe 2 M-L have been deposited in the Cambridge Crystallographic Data Centre (CCDC deposition number: 2177907-2177909).

Estimation of C 3 H 6 and C 2 H 4 Adsorption Heat
The C 3 H 6 and C 2 H 4 adsorption heats were calculated from the adsorption data recorded at 298 and 273 K. The Toth equation [51] (Equation (1)) was first utilized for fittings the adsorption isotherms (Figures S10-S19), where N stands for the gas uptake, N sat stands for the saturated uptake, P represents pressure, b and t represent two constants.
The following Equation (2) could be obtained from the rearrangement of Equation (1). The C 3 H 6 and C 2 H 4 adsorption heats (Q st ) were estimated by the Clausius-Clapeyron equation [52] (Equation (3)), where C represents a constant, R stands for the universal gas constant, and T stands for temperature. Assuming (lnP) N is the function, and (1/T) is the variable, Q st depending on the gas uptake (N), could be calculated from the slopes data points (Q st /R).

Prediction of IAST C 3 H 6 /C 2 H 4 Selectivity
Ideal adsorbed solution theory [53] (IAST) is a well-accepted way to predict gas adsorption selectivity for gas mixtures by adsorption data of the individual gases. The IAST defines the following equations. In these equations, R stands for the universal gas constant, T stands for the temperature for adsorption experiment, A represents the surface area of adsorbent, x i stands for the ratio of gas i in the gas mixture adsorbed by adsorbent, y i stands for the ratio of gas i in the gas mixture before adsorption, p mix represents the pressure of gas mixture before adsorption, p i 0 represents the pressure of gas i that corresponds to the spreading pressure π of the binary mixture, n i (p) stands for the uptake of gas i at the pressure p.
The adsorption selectivity of gas 1 over gas 2 (S 12 ) can be obtained by Equation (12). For the adsorption of 1:9 C 3 H 6 /C 2 H 4 gas mixture at 1 bar, C 3 H 6 is gas 1, C 2 H 4 is gas 2, p mix is 1 bar, y 1 is 0.1, y 2 is 0.9, and n i (p) can be calculated by fittings adsorption data of the individual gases ( Figures S20-S24). x 1 can then be calculated by solving Equation (11). At last, x 2 can be obtained by applying the x 1 to Equation (5), and S 12 can be calculated from Equation (12). The C 3 H 6 /C 2 H 4 selectivities for a 1:1 C 3 H 6 /C 2 H 4 gas mixture or at other pressures could be calculated similarly.

Gas Mixture Breakthrough
Breakthrough tests for the gas mixture were performed with a 1:9 C 3 H 6 /C 2 H 4 gas mixture. Quartz tubes (6 mm for outer diameter, 3 mm for inner diameter, and 100 mm in length) were packed with the adsorbents. The adsorbents were first activated at 100 • C overnight inside the tubes under a He flow with a flow rate of 10 SCCM. Mass flow controllers (Alicat Scientific, Tucson, AZ, USA) were used to control the gas flow rate. After the adsorbents were cooled down to room temperature, the breakthrough experiments started when the He flow was changed into a flow of the C 3 H 6 /C 2 H 4 gas mixture (flow rate: 2 SCCM). The concentrations of C 3 H 6 and C 2 H 4 gases at outlet were determined by a mass spectrometer (Hiden HPR20). For monitoring the desorption dynamics of the adsorbed C 3 H 6 and C 2 H 4 gases in the adsorbents, the gas flow was changed from the binary C 3 H 6 /C 2 H 4 gas to a He flow of 10 SCCM.

Crystal Structure and Porosity
Fe 2 M-L were synthesized from solvothermal reactions of [Fe 2 M(µ 3 -O)(CH 3 COO) 6 ] precursors and H 3 L ligand in DMF at 120 • C. It was found that using the pre-synthesized 6 ] precursors instead of mixtures of Fe/M metal salts as metal ion source is necessary for the formation of final crystalline products. Additionally, the introduction of an excess coordination modulator (acetic acid) is important for the production of Fe 2 M-L crystals in high crystallinity and with a relatively large size (>0.1 mm). It is well-known that coordination modulators affect the nucleation and growth rate, morphology, and crystallinity of MOFs, and only intergrown aggregates with poor crystallinity or even amorphous phases could be obtained without modulators in some cases [17,54]. The presence of two types of metal ions with a Fe to M ratio of 2:1 in Fe 2 M-L was confirmed by ICP-AES measurements for digested samples of the MOFs (Table S2).
Single-crystal X-ray diffraction (SCXRD) experiments and structure analyses revealed that Fe 2 M-L are isostructural and in the R-3c space group (trigonal) ( Table S1). There is one type of [Fe 2 M(µ 3 -O)(−COO) 6 ] (denoted as Fe 2 M hereafter) clusters and one type of L 3− ligands in their structures. Each Fe 2 M cluster is connected with six different but equivalent L 3− ligands, and each L 3− ligand bridges three Fe 2 M clusters ( Figure 1a). The interconnection of the Fe 2 M clusters and L 3− ligands in such a way results in their 3D frameworks, which could be regarded as (3,6)-connected nets from a topological point  Figure S1). The solvent-accessible volumes are~74% of the whole structures, as estimated by Platon [49]. ligands ( Figure 1b). In the octahedral cage, two neighboring Fe2M clusters are all bridged by one L 3− ligand with its two carboxylate groups. The other set of cages is in tetrakaidecahedron shape, each of which is built from 12 Fe2M clusters and 12 L 3− ligands where each L 3− ligand links 3 neighboring Fe2M clusters (Figure 1c). The cavities inside th octahedral and tetrakaidecahedral cages are ~6 and ~9 Å, respectively. Each small cage i surrounded by eight large cages, and each large cage is surrounded by eight small cage and six equivalent large cages by polyhedral face sharing ( Figure 1d). The windows be tween two small cages or between one large cage and one small cage are in a diameter o ~3.8 Å, and the windows between two large cages are in a diameter of ~5.1 Å. Severa isoreticular MOFs to Fe2M-L were previously reported [55][56][57].  The highly open frameworks of Fe 2 M-L can also be regarded as alternate packing of two sets of cages. One set is octahedral, each of which consists of 6 Fe 2 M clusters and 8 L 3− ligands (Figure 1b). In the octahedral cage, two neighboring Fe 2 M clusters are all bridged by one L 3− ligand with its two carboxylate groups. The other set of cages is in a tetrakaidecahedron shape, each of which is built from 12 Fe 2 M clusters and 12 L 3− ligands, where each L 3− ligand links 3 neighboring Fe 2 M clusters (Figure 1c). The cavities inside the octahedral and tetrakaidecahedral cages are~6 and~9 Å, respectively. Each small cage is surrounded by eight large cages, and each large cage is surrounded by eight small cages and six equivalent large cages by polyhedral face sharing (Figure 1d). The windows between two small cages or between one large cage and one small cage are in a diameter of~3.8 Å, and the windows between two large cages are in a diameter of~5.1 Å. Several isoreticular MOFs to Fe 2 M-L were previously reported [55][56][57].
To confirm the permanent porosity of Fe 2 M-L, N 2 adsorption isotherms were recorded at 77 K after the as-prepared crystals were activated by guest exchange (methanol as solvent) and then degassing at 80 • C. The three MOFs showed highly similar N 2 adsorption isotherms (Figure 2a), which is consistent with their isostructural structures and close unit cell parameters. The isotherms are type I isotherms typical for microporous materials, showing saturated N 2 adsorption capacities of 837, 854, and 847 cm 3 g −1 at~1 P/P 0 for Fe 2 Mn-L, Fe 2 Co-L, and Fe 2 Ni-L, respectively. The apparent BET/Langmuir surface areas of the three MOFs are estimated to be 3105/3600, 3168/3675, and 3169/3674 m 2 g −1 , respectively. The pore volumes are estimated to be 1.29, 1.32, and 1.31 cm 3 g −1 , respectively, which are almost the same as the predicted pore volumes from their single crystal data (1.30, 1.32, and 1.32 cm 3 g −1 ). The results suggested that the MOF samples were in a pure phase, and their highly open frameworks remained unchanged after the evacuation of guests. The high purity of the batch MOF crystal samples was also confirmed by PXRD measurement results, which showed a good agreement between the PXRD patterns of the MOF samples and the single-crystal structure simulated ones ( Figure S2). orded at 77 K after the as-prepared crystals were activated by guest exchange (methanol 195 as solvent) and then degassing at 80 °C. The three MOFs showed highly similar N2 ad-196 sorption isotherms (Figure 2a), which is consistent with their isostructural structures and 197 close unit cell parameters. The isotherms are type I isotherms typical for microporous ma-198 terials, showing saturated N2 adsorption capacities of 837, 854, and 847 cm 3 g −1 at ~1 P/P0 199 for Fe2Mn-L, Fe2Co-L, and Fe2Ni-L, respectively. The apparent BET/Langmuir surface ar-200 eas of the three MOFs are estimated to be 3105/3600, 3168/3675, and 3169/3674 m 2 g −1 , re-201 spectively. The pore volumes are estimated to be 1.29, 1.32, and 1.31 cm 3 g −1 , respectively, 202 which are almost the same as the predicted pore volumes from their single crystal data 203 (1.30, 1.32, and 1.32 cm 3 g −1 ). The results suggested that the MOF samples were in a pure 204 phase, and their highly open frameworks remained unchanged after the evacuation of 205 guests. The high purity of the batch MOF crystal samples was also confirmed by PXRD 206 measurement results, which showed a good agreement between the PXRD patterns of the 207 MOF samples and the single-crystal structure simulated ones ( Figure S2).

214
The adsorption isotherms of C2H4 and C3H6 were recorded for Fe2M-L at 298 K as 215 well as 273 K. As shown in Figure 2b, the three MOFs showed close gas adsorption capac-216 ities at all pressure ranges. The C2H4 uptakes were 87.5, 94.4, and 94.9 cm 3 g −1 , and the 217 C3H6 uptakes are 291.1, 302.3, and 304.2 cm 3 g −1 at 1 bar for Fe2Mn-L, Fe2Co-L, and Fe2Ni-218 L, respectively. The slightly lower gas uptakes of Fe2Mn-L with respect to those of the 219 other two MOFs are consistent with the results of 77 K N2 adsorption studies, which re-220 vealed that Fe2Mn-L had a slightly lower porosity than Fe2Co-L and Fe2Ni-L. Overall, the 221 results indicate the gas adsorption properties of Fe2M-L are not very dependent on the 222 nature of the M ions. Therefore, only Fe2Mn-L was investigated for further experiments. 223 For comparison, C2H4 and C3H6 adsorption measurements were also carried out for four 224 benchmark adsorbents, namely, HKUST-1, MIL-101(Cr), ZIF-8, and a commercial acti-225 vated carbon. ZIF-8, MIL-101(Cr), and HKUST-1 were all prepared by reported methods 226 [45][46][47], and activated carbon was purchased from Sigma-Aldrich. Before the C2H4 and 227 C3H6 adsorption measurements, PXRD measurements and/or N2 adsorption experiments 228 at 77 K were performed for those adsorbents ( Figure S3-S9). According to the adsorption 229 experiments, their porosities are evaluated and shown in Table 1.

Adsorption Study for C 2 H 4 and C 3 H 6
The adsorption isotherms of C 2 H 4 and C 3 H 6 were recorded for Fe 2 M-L at 298 K as well as 273 K. As shown in Figure 2b, the three MOFs showed close gas adsorption capacities at all pressure ranges. The C 2 H 4 uptakes were 87.5, 94.4, and 94.9 cm 3 g −1 , and the C 3 H 6 uptakes are 291.1, 302.3, and 304.2 cm 3 g −1 at 1 bar for Fe 2 Mn-L, Fe 2 Co-L, and Fe 2 Ni-L, respectively. The slightly lower gas uptakes of Fe 2 Mn-L with respect to those of the other two MOFs are consistent with the results of 77 K N 2 adsorption studies, which revealed that Fe 2 Mn-L had a slightly lower porosity than Fe 2 Co-L and Fe 2 Ni-L. Overall, the results indicate the gas adsorption properties of Fe 2 M-L are not very dependent on the nature of the M ions. Therefore, only Fe 2 Mn-L was investigated for further experiments. For comparison, C 2 H 4 and C 3 H 6 adsorption measurements were also carried out for four benchmark adsorbents, namely, HKUST-1, MIL-101(Cr), ZIF-8, and a commercial activated carbon. ZIF-8, MIL-101(Cr), and HKUST-1 were all prepared by reported methods [45][46][47], and activated carbon was purchased from Sigma-Aldrich. Before the C 2 H 4 and C 3 H 6 adsorption measurements, PXRD measurements and/or N 2 adsorption experiments at 77 K were performed for those adsorbents (Figures S3-S9). According to the adsorption experiments, their porosities are evaluated and shown in Table 1. Table 1. The porosities of HKUST-1, MIL-101(Cr), ZIF-8, activated carbon, and Fe 2 M-L estimated by their N 2 adsorption data.

Adsorbent
Pore Volume (cm 3 g −1 ) S BET (m 2 g −1 ) S Langmuir (m 2 g −1 ) HKUST As shown in Figure 2c, for HKUST-1 and activated carbon, the C 3 H 6 uptakes increase abruptly at the low-pressure range (~140 and 90 cm 3 g −1 at 0.1 bar), and after a gradual increase at the high-pressure range, the uptakes reach 167.0 and 128.9 cm 3 g −1 at 1 bar, respectively. The C 2 H 4 adsorption isotherms of the two adsorbents share a similar profile to the C 3 H 6 adsorption isotherms with lower uptakes, being 136.5 and 98.8 cm 3 g −1 at 1 bar, respectively. In contrast, for MIL-101(Cr) and ZIF-8, the uptakes of C 3 H 6 or C 2 H 4 all gradually increase in the full pressure range. Notably, although MIL-101(Cr) has a larger pore volume (1.47 cm 3 g −1 ) than Fe 2 Mn-L (1.29 cm 3 g −1 ), its C 3 H 6 and C 2 H 4 uptakes at 1 bar (196.6 and 62.1 cm 3 g −1 ) is obviously lower than those of Fe 2 Mn-L (291.1 and 87.5 cm 3 g −1 ). Additionally, the gas uptakes of ZIF-8 are the lowest among all the tested adsorbents, being 80.6 and 26.4 cm 3 g −1 at 1 bar for C 3 H 6 and C 2 H 4 , respectively, although its pore volume is clearly higher than those of HKUST-1 and activated carbon ( Table 1). The C 3 H 6 uptake of Fe 2 Mn-L at 1 bar is higher than those of the other adsorbents, but its C 2 H 4 uptake is lower than that of HKUST-1 and activated carbon. The C 3 H 6 /C 2 H 4 uptake ratios at 1 bar are 3.3, 3.2, 3.1, 1.3, and 1.2 for Fe 2 Mn-L, MIL-101(Cr), ZIF-8, activated carbon, and HKUST-1, respectively. The C 3 H 6 and C 2 H 4 adsorption isotherms of MIL-101(Cr), ZIF-8, and HKUST-1 were also previously reported [20,38,40,[58][59][60][61], and those reported results are basically consistent with those presented in this work. Some slight differences may result from the subtle difference in sample preparation and/or activation.
For a better understanding of the C 3 H 6 and C 2 H 4 adsorption behavior and the IAST predicted C 3 H 6 /C 2 H 4 selectivities, the C 3 H 6 and C 2 H 4 adsorption heats (Q st ) were estimated for the adsorbents by Clausius-Clapeyron equation using Toth equation fitting parameters from adsorption results obtained at 273 and 298 K (Figures S10-S19) [51,52].  For a better understanding of the C3H6 and C2H4 adsorption behavior and the IAST predicted C3H6/C2H4 selectivities, the C3H6 and C2H4 adsorption heats (Qst) were estimated for the adsorbents by Clausius-Clapeyron equation using Toth equation fitting parame ters from adsorption results obtained at 273 and 298 K (Figure S10-S19) [51,52]. For C2H adsorption, the Qst values for HKUST-1, activated carbon, Fe2Mn-L, and MIL-101(Cr) de crease as the loadings increase, whereas the Qst values for ZIF-8 increase at low loading and stabilize at higher loadings (Figure 3c The above results revealed HKUST-1 and activated carbon show high affinity to the C2 and C3 olefins, which resulted in their high loading of the olefins even at low-pressure ranges. Relatively, the interactions between the olefins and Fe2Mn-L, MIL-101(Cr), or ZIF 8 are low, although, at low pressures, Fe2Mn-L and MIL-101(Cr) also show quite high Qs values. The interaction between Fe2Mn-L and C3H6 is not strong, but it shows high C3H uptakes, which should be a result of its high porosity and moderate pore size. It is also suggested that the nature of metal ions in the adsorbents does not profoundly affect their The above results revealed HKUST-1 and activated carbon show high affinity to the C2 and C3 olefins, which resulted in their high loading of the olefins even at low-pressure ranges. Relatively, the interactions between the olefins and Fe 2 Mn-L, MIL-101(Cr), or ZIF-8 are low, although, at low pressures, Fe 2 Mn-L and MIL-101(Cr) also show quite high Q st values. The interaction between Fe 2 Mn-L and C 3 H 6 is not strong, but it shows high C 3 H 6 uptakes, which should be a result of its high porosity and moderate pore size. It is also suggested that the nature of metal ions in the adsorbents does not profoundly affect their C 3 H 6 and C 2 H 4 adsorption behaviors. For the adsorptive C 3 H 6 /C 2 H 4 separation, three key parameters need to be considered, namely, C 3 H 6 /C 2 H 4 adsorption selectivity, adsorption capacity, and regeneration energy. Based on the above-mentioned results, among the tested adsorbents, HKUST-1 may outperform the others in high adsorption selectivity, while Fe 2 Mn-L may be advantageous in high adsorption capacity and facile regeneration for C 3 H 6 /C 2 H 4 separation.
The C 3 H 6 /C 2 H 4 separation performances of other types of adsorbents have also been reported. For example, Han et al. reported two covalent organic frameworks (COFs), CR-COF-1 and CR-COF-2, by one-pot Suzuki coupling and Schiff's base reaction [5]. The C 3 H 6 uptakes were 84 and 137 cm 3 g −1 , and the C 2 H 4 uptakes were 38 and 72 cm 3 g −1 at~1 bar and 298 K for CR-COF-1 and CR-COF-2, respectively. The uptakes are obviously lower than those of Fe 2 Mn-L (291.1 and 87.5 cm 3 g −1 ). Zhang et al. prepared a blend membrane by doping 15% (mass) polyethylene glycol (PEG600) into the poly(ether-block-amide) (Pe-bax®2533) polymer matrix [6]. For a 1:1 binary C 3 H 6 /C 2 H 4 gas, the Pebax®2533/PEG600 membrane showed a C 3 H 6 permeability of 273 barrer and a C 3 H 6 /C 2 H 4 selectivity of 4.15 at 293 K and 2 bar, and a C 3 H 6 permeability of 196 barrer and a C 3 H 6 /C 2 H 4 selectivity of 8.90 at 238 K and 2 bar. The permselectivities are comparable to the C 3 H 6 /C 2 H 4 IAST selectivity of Fe 2 Mn-L at 298 K and 1 bar for a 1:1 binary C 3 H 6 /C 2 H 4 gas (7.8).

Dynamic Breakthrough of Binary C 3 H 6 /C 2 H 4 Gas
To confirm the separation capacities of the adsorbents for a real binary C 3 H 6 /C 2 H 4 gas, breakthrough experiments were performed for the quartz tubes packed with the adsorbents by using a binary gas of C 3 H 6 /C 2 H 4 (1:9) at room temperature and ambient pressure. The tested adsorbents all show capability to separate the C 3 H 6 /C 2 H 4 gas mixture, which is indicated by the gaps between their C 2 H 4 and C 3 H 6 breakthrough curves (Figure 4a). Among the five adsorbents, HKUST-1 showed the longest breakthrough time for C 3 H 6 (530 min g −1 ), indicating a C 3 H 6 adsorption capacity of 106 cm 3 g −1 . The C 2 H 4 breakthrough time for HKUST-1 was 62 min g −1 , corresponding to a C 2 H 4 adsorption capacity of 112 cm 3 g −1 . Accordingly, about 842 cm 3 g −1 purified C 2 H 4 could be obtained in a breakthrough test of HKUST-1. The C 3 H 6 /C 2 H 4 separation capacity is also high for activated carbon according to its breakthrough curves. It captured 69 cm 3 g −1 C 3 H 6 and 106 cm 3 g −1 C 2 H 4 , and about 515 cm 3 g −1 purified C 2 H 4 could be obtained for a breakthrough run. Fe 2 Mn-L captured about 73 cm 3 g −1 C 3 H 6 before C 3 H 6 started to penetrate the MOF column. Meanwhile, 153 cm 3 g −1 C 2 H 4 was adsorbed, and the productivity of purified C 2 H 4 in each breakthrough run was 504 cm 3 g −1 for Fe 2 Mn-L, slightly less than the productivity of activated carbon. The C 2 H 4 and C 3 H 6 adsorption capacities of ZIF-8 and MIL-101(Cr) were obviously lower, and only about 181 and 271 cm 3 g −1 purified C 2 H 4 could be obtained in each of their breakthrough tests, respectively. uptakes were 84 and 137 cm 3 g −1 , and the C2H4 uptakes were 38 and 72 cm 3 g −1 at ~1 bar 301 and 298 K for CR-COF-1 and CR-COF-2, respectively. The uptakes are obviously lower 302 than those of Fe2Mn-L (291.1 and 87.5 cm 3 g −1 ). Zhang et al. prepared a blend membrane 303 by doping 15% (mass) polyethylene glycol (PEG600) into the poly(ether-block-amide) 304 (Pebax® 2533) polymer matrix [6]. For a 1:1 binary C3H6/C2H4 gas, the Pebax® 305 2533/PEG600 membrane showed a C3H6 permeability of 273 barrer and a C3H6/C2H4 selec-306 tivity of 4.15 at 293 K and 2 bar, and a C3H6 permeability of 196 barrer and a C3H6/C2H4 307 selectivity of 8.90 at 238 K and 2 bar. The permselectivities are comparable to the 308 C3H6/C2H4 IAST selectivity of Fe2Mn-L at 298 K and 1 bar for a 1:1 binary C3H6/C2H4 gas 309 (7.8). 310

311
To confirm the separation capacities of the adsorbents for a real binary C3H6/C2H4 312 gas, breakthrough experiments were performed for the quartz tubes packed with the ad-313 sorbents by using a binary gas of C3H6/C2H4 (1:9) at room temperature and ambient pres-314 sure. The tested adsorbents all show capability to separate the C3H6/C2H4 gas mixture, 315 which is indicated by the gaps between their C2H4 and C3H6 breakthrough curves (Figure 316  4a). Among the five adsorbents, HKUST-1 showed the longest breakthrough time for C3H6 317 (530 min g −1 ), indicating a C3H6 adsorption capacity of 106 cm 3 g −1 . The C2H4 breakthrough 318 time for HKUST-1 was 62 min g −1 , corresponding to a C2H4 adsorption capacity of 112 cm 3 319 g −1 . Accordingly, about 842 cm 3 g −1 purified C2H4 could be obtained in a breakthrough test 320 of HKUST-1. The C3H6/C2H4 separation capacity is also high for activated carbon accord-321 ing to its breakthrough curves. It captured 69 cm 3 g −1 C3H6 and 106 cm 3 g −1 C2H4, and about 322 515 cm 3 g −1 purified C2H4 could be obtained for a breakthrough run. Fe2Mn-L captured 323 about 73 cm 3 g −1 C3H6 before C3H6 started to penetrate the MOF column. Meanwhile, 153 324 cm 3 g −1 C2H4 was adsorbed, and the productivity of purified C2H4 in each breakthrough 325 run was 504 cm 3 g −1 for Fe2Mn-L, slightly less than the productivity of activated carbon. 326 The C2H4 and C3H6 adsorption capacities of ZIF-8 and MIL-101(Cr) were obviously lower, 327 and only about 181 and 271 cm 3 g −1 purified C2H4 could be obtained in each of their break-328 through tests, respectively.  The binary gas breakthrough experiment results are in good accordance with the results of adsorption isotherm measurements for the individual gases, IAST selectivity, and adsorption heat calculations. Specifically, HKUST-1 exhibited high Q st and uptakes for both C 3 H 6 and C 2 H 4 , and its C 3 H 6 /C 2 H 4 IAST selectivities were also high relative to those of the other tested adsorbents. A high C 3 H 6 /C 2 H 4 separation performance was indeed observed in the binary gas breakthrough experiment for HKUST-1. Similar results were also observed for activated carbon, except that it exhibited lower adsorption capacities than HKUST-1. For Fe 2 Mn-L, although its C 3 H 6 /C 2 H 4 IAST selectivities were obviously lower than those of HKUST-1 and activated carbon, it still showed a high C 3 H 6 /C 2 H 4 separation performance for a real binary gas. Compared with HKUST-1, Fe 2 Mn-L shows lower productivity of purified C 2 H 4 in a breakthrough run, and the C 2 H 4 productivities of Fe 2 Mn-L and activated carbon are close. However, it should be noted that the Q st values of Fe 2 Mn-L for C 3 H 6 and C 2 H 4 adsorption are much lower than those of HKUST-1 and activated carbon, which would lead to its advantage in regeneration with less energy consumption. This conjecture is further sustained by the comparison of C 3 H 6 and C 2 H 4 desorption dynamics of the adsorbents. After C 3 H 6 and C 2 H 4 were saturatedly adsorbed by the adsorbents during breakthrough experiments, the adsorbents were regenerated by purging a He flow. It was found that the evacuation of adsorbed gases in Fe 2 Mn-L was clearly faster than that of HKUST-1 or activated carbon (Figure 4b).

Conclusions
In summary, three new isostructural MOFs, Fe 2 M-L (M = Mn 2+ , Co 2+ , or Ni 2+ ), have been obtained, which all show large surface areas (BET:~3100 m 2 g −1 ) and high pore volumes (~1.3 cm 3 g −1 ). Adsorption isotherms at room temperature suggest the MOFs uptake~300 cm 3 g −1 C 3 H 6 and~90 cm 3 g −1 C 2 H 4 at 1 bar. The potential of the MOFs in C 3 H 6 /C 2 H 4 separation has been further evaluated by IAST selectivity prediction, adsorption heat calculations, and dynamic binary C 3 H 6 /C 2 H 4 (1:9) gas breakthrough experiments. The predicted IAST C 3 H 6 /C 2 H 4 adsorption selectivity for Fe 2 Mn-L is~8 at 298 K and 1 bar. The C 3 H 6 and C 2 H 4 adsorption heats for Fe 2 Mn-L are estimated to be 28-40 kJ mol −1 and 20-38 kJ mol −1 , respectively. A binary gas breakthrough experiment confirms the capability of Fe 2 Mn-L to selectively adsorb C 3 H 6 over C 2 H 4 , producing 469 cm 3 g −1 purified C 2 H 4 in a breakthrough run. In addition, the C 3 H 6 /C 2 H 4 separation performances of four other benchmark adsorbents, HKUST-1, MIL-101(Cr), ZIF-8, and activated carbon, were also studied for comparison. The results reveal that Fe 2 M-L are promising adsorbents for C 3 H 6 /C 2 H 4 separation with low energy penalty in regeneration, although HKUST-1 shows higher C 3 H 6 /C 2 H 4 adsorption selectivity and productivity of purified C 2 H 4 than Fe 2 M-L and the other tested adsorbents.