Introduction

Separations are indispensable in the chemical industry, where thermal-driven processes are predominant ( > 80%). These established processes, e.g., cryogenic distillations, consume a significant amount of energy due to the similar properties of gas components1,2,3. Separations of olefins and paraffins are highly important, over 350 million tons of olefins, i.e., ethylene (C2H4) and propylene (C3H6), are produced mainly by steam cracking of naphtha1,4,5. But the separations of olefins/paraffins are incredibly challenging owing to their close relative volatilities and boiling points raised by the subtle molecular differences in single/double carbon bonds6,7,8. The dehydrogenation reactions of paraffins (i.e., propane and ethane) typically exhibit a conversion yield from 50% to 60%, giving a complex mixture of C2-C4 olefins and paraffins9,10. The prevalent high-pressure cryogenic distillations for producing polymer-grade ( > 99.95%) olefins consume ~7.3 GJ per ton C2H4 and ~12.9 GJ per ton C3H6, which collectively consume 0.3% of the global energy production11,12. In contrast, physical adsorptions utilizing solid adsorbents with different affinities toward guest molecules enable energy-efficient olefin/paraffin separations and purifications8,13,14,15,16.

The molecular-sieving mechanism can achieve complete separation with theoretically infinite selectivity17. The predominant molecular-sieving adsorbents function by differentiating molecule sizes, that is, adsorbing smaller components while excluding larger ones (Fig. 1a)18,19,20,21. Therefore, the adsorbents executing the size-selective mechanism require precise control over pore sizes and apertures22,23,24,25. For instance, given the subtle size difference in kinetic diameter of C2H4 and C2H6 (0.028 nm in kinetic diameter, Supplementary Fig. 1), only three exceptional adsorbents (i.e., UTSA-280, M-gallate (M = Mg, Co), and HOF-FJU-1) have successfully achieved the complete exclusion of C2H6 from C2H4/C2H6 mixtures10,17,26. Although C3H6 and C3H8 have a larger size difference (0.042 nm in kinetic diameter), their structural dissimilarity is even more minor due to the same alkyl end (-CH3 group), making the C3H6 sieving from C3H8 more challenging11,27,28. Therefore, the specific design and tuning of pore sizes are prerequisite for each olefin/paraffin pair. Note that achieving simultaneous molecular sieving of multiple olefin/paraffin pairs, such as C2H4/C2H6 and C3H6/C3H8, is theoretically impossible in rigid adsorbents.

Fig. 1: The structure and pore properties of BFFOUR-Cu-dpds.
figure 1

a Schematic of size-selective and interaction-selective molecular sieving mechanisms. b 1D [Cu(dpds)2(BF4)2]n chains viewed along a axis. c 2D flexible layers interwoven by angular BF4- anions via hydrogen bonds. d The responsive moving of 2D BFFOUR-Cu-dpds layers. e The pores of activated BFFOUR-Cu-dpds viewed along c axis. Color code: Cu dark green, B orange, F light green, C dark grey, H light grey, S yellow, N blue.

Full size image

In addition to the combination of metal ions/clusters and organic ligands in metal-organic frameworks (MOFs), the incorporation of anionic pillars as third components (e.g., SiF62−, GeF62−, SO42−) into hybrid MOF adsorbents allows for facile tuning of pore apertures/functionalities, holding particular promise for the separations of light hydrocarbons29,30,31,32. Generally, the linear organic linkers (e.g., 4,4’-bipyridine, pyrazine, and 1,4-bis(4-pyridyl)benzene) and transit metal ions (e.g., Cu2+, Co2+, and Zn2+) form rigid two-dimensional (2D) layers with coordination bonds, which are further coordinately connected by anion pillars to create three-dimensional (3D) frameworks (Supplementary Fig. 2a)33,34,35. Structural rigidity is typically maintained in this type of adsorbent, while the rotation of anionic pillars may offer a degree of flexibility36,37,38,39. Recently, our group reported a series of rigid-flexible MFSIX-Cu-dps (dps = 4,4’-dipyridyridylsulfide, M = Si, Ge, Ti), the bent dps linker created rigid intralayer cages and flexible interlayer spaces (Supplementary Fig. 2b). The selective expansion of interlayer spaces facilitated efficient C2H2/C2H4 and C2H2/CO2 sieving separations36,40. Therefore, we expect that the judicious selection of anion pillars can decouple 2D interlayers and provide additional dimensions of flexibility (Supplementary Fig. 2c). This intriguing feature may offer an opportunity for intricate C2-C4 olefin/paraffin sieving separations by distinguishing the saturation degree of carbon bonds in a single adsorbent (Fig. 1a).

Herein, we report a molecular sieving adsorbent, BFFOUR-Cu-dpds (BFFOUR = BF4−, dpds = 4,4’-bipyridinedisulfide), for simultaneous sieving of C2-C4 olefins (C2H4, C3H6, and n-C4H8) from their corresponding paraffins (C2H6, C3H8, and n-C4H10). The assembly of adjacent one-dimensional (1D) Cu(dpds)2(BF4)2 chains are interwoven by angular BF4− anions forming flexible interlayer spaces. The adsorption data demonstrates that C2-C4 olefins can expand interlayer spaces, while paraffins are excluded. Additionally, benzene exhibits selective opening of contracted spaces instead of cyclohexane, further confirming its selectivity originating from distinguishing guest-host interactions. Note that high-purity C2H4 (>99.99%) can be directly obtained through a two-column configuration from six-component C2-C4 olefins and paraffins mixtures. The first BFFOUR-Cu-dpds column efficiently separates olefins and paraffins, while high-purity C2H4 can be directly obtained from the second column filled with granular porous carbons. Guest-loaded single-crystal analysis reveals the structure evolution and identifies the adsorption sites. Computational simulations demonstrate that olefins exhibit significantly lower energy barriers for interlayer expansion than paraffins.

Results

Structure and porosity analysis

The reaction of dpds ligand and Cu(BF4)2 in methanol/water solution yielded purple cuboid BFFOUR-Cu-dpds (Cu(dpds)2(BF4)2·2H2O) crystals under mild conditions (Supplementary Figs 3, 4). Single-crystal X-ray diffraction (SCXRD) revealed that the as-synthesized BFFOUR-Cu-dpds crystallizes in the monoclinic Ccc2 space group (Supplementary Table 1). The distorted octahedral Cu atom was coordinated by four N atoms from four independent dpds linkers and two F atoms from two trans BF4− anions (Supplementary Fig. 5), which was connected by dpds ligands forming infinite 1D [Cu(dpds)2(BF4)2]n chains (Fig. 1b). Adjacent 1D chains were coupled and stabilized through hydrogen bonds mediated by guest water molecules with an O···H distance of 2.68 Å (Supplementary Fig. 6). The angular BF4− anions further interweaved the assembled flexible 2D layers via hydrogen bonds along a axis (Fig. 1c). Note that the generated 2D layers exhibited non-rigid behaviors and functioned as hinges with a certain degree of flexibility (Fig. 1d). Unlike rigid coordination bonds in MF6-pillared MOFs, the flexible hydrogen bonds created unparallel stackings with a dihedral angle of 64.39° along a axis, thereby imparting additional flexibility (Supplementary Fig. 7). The asymmetric assembly resulted in discrete pore systems with an accessible pore size of 3.9 × 4.0 Å (Supplementary Fig. 8). The single-crystal to single-crystal transition was achieved after the removal of guest water molecules (Supplementary Fig. 4b). The SCXRD analysis disclosed that the activated BFFOUR-Cu-dpds [Cu(dpds)2(BF4)2] retained the same space group but with a slightly reduced pore volume of 2816.3(16) Å3 (Supplementary Fig. 9 and Supplementary Table 1). The enlarged dihedral angle of 76.83° reduced the accessible pore size to 3.3 × 3.8 Å (Fig. 1e and Supplementary Fig. 7b), smaller than the cross-sectional sizes of C2-C4 olefins and paraffins. In the activated BFFOUR-Cu-dpds, the interlayer distance was measured to be 3.55 Å (Supplementary Fig. 6b), longer than that of SiFSIX-Cu-dps (3.35 Å) and ZUL-100 (3.35 Å), indicating increased flexibility40,41.

The phase purity of bulk BFFOUR-Cu-dpds was confirmed by powder X-ray diffraction (PXRD) patterns that matched well with the simulated one (Supplementary Fig. 10). After activation, there was no discernible alteration in the diffraction peaks. Thermogravimetric analysis revealed that BFFOUR-Cu-dpds was thermally stable up to 240 °C (Supplementary Fig. 11). Fourier transform infrared (FT-IR) spectrometer was conducted to confirm the constituents of BFFOUR-Cu-dpds (Supplementary Fig. 12). The B-F stretching vibrations at 1128.23 cm−1 (vs) and 1060.24 cm−1 (vas) were attributed to the BF4- pillars42, while the small sharp peak at 659.53 cm−1 was assigned to the Cu-F bond stretching vibration43. The bands of C-S and S-S bonds were observed at 533.24 cm−1 and 522.61 cm−1, respectively44. The Cu-N bonds that appeared at 499.47 and 439.69 cm−1 confirmed the coordination pattern of Cu2+ ions45. Furthermore, the PXRD characteristic peaks and BET specific surface areas measured at 195 K of BFFOUR-Cu-dpds almost remained intact after soaking in various organic solvents and acidic/basic aqueous solutions (pH = 5–11) for 7 days, indicating its excellent structure and chemical stability (Supplementary Fig. 13). The single crystals could also be retained after immersing in various organic solvents (Supplementary Fig. 14). The porous properties were evaluated by CO2 adsorption isotherm at 195 K, the activated BFFOUR-Cu-dpds exhibited a Brunauer-Emmentt-Teller (BET) specific surface area of 140 m2 g−1 with a pore size centered at 3.8 Å (Supplementary Fig. 15). The experimental total pore volume was calculated to be 0.06 cm3 g−1, which is slightly lower than the value obtained through SCXRD analysis (0.078 cm3 g−1). Meanwhile, the contracted pores resulted in negligible N2 adsorption at 77 K.

Equilibrium and kinetic gas adsorption behaviors

Single-component equilibrium adsorption isotherms of C2-C4 olefins and paraffins were collected at different temperatures (273–323 K). Fig 2a-c illustrated the step-wise adsorption behavior of C2-C4 olefins, which can be attributed to the structural adaptability of flexible MOFs under guest stimuli30,37,39,46,47. At 1.0 bar, the C2H4 uptake reached 31.43 cm3 g−1 and 29.31 cm3 g−1 at 273 K and 298 K, respectively (Fig. 2a and Supplementary Fig. 16a). Similarly, the adsorption capacity for C3H6 was measured to be 33.90 and 27.50 cm3 g−1 at 273 K and 298 K, respectively (Fig. 2b and Supplementary Fig. 17a). Meanwhile, BFFOUR-Cu-dpds could adsorb 21.49 cm3 g−1 n-butene (n-C4H8) at 298 K, which increased to 24.54 cm3 g−1 at 283 K (Fig. 2c and Supplementary Fig. 18a). Note that C2-C4 paraffins, i.e., C2H6, C3H8, and n-butane (n-C4H10), were completely excluded by activated BFFOUR-Cu-dpds even at higher temperatures (Supplementary Fig. 19). Correspondingly, based on the experimental pore volume, the packing density of C2H4, C3H6, and n-C4H8 in BFFOUR-Cu-dpds achieved 471.47 g L−1, 653.92 g L−1, and 680.51 g L−1 at 298 K and 1.0 bar, which was 414.3, 383.1, and 298.8 times higher than the density of gaseous C2H4 (1.138 g L−1), C3H6 (1.707 g L−1), and n-C4H8 (2.276 g L−1) under similar conditions17,48. The time-dependent kinetic adsorption curves at 0.5 bar revealed an abrupt uptake point in <1 min for C2-C4 olefins and quickly reached equilibrium at ~13 min (Fig. 2d). The kinetic adsorption capacity of C2H4 (25.0 cm3 g−1), C3H6 (24.1 cm3 g−1), and n-C4H8 (14.7 cm3 g−1) was in good agreement with their corresponding equilibrium adsorption capacities. Meanwhile, no noticeable adsorption uptakes were observed on C2-C4 paraffins even after a prolonged period of ~70 mins. The kinetic adsorption curves at 1.0 bar showed a similar phenomenon (Supplementary Fig. 20).

Fig. 2: Adsorption and separation properties of BFFOUR-Cu-dpds.
figure 2

Adsorption isotherms for (a) C2H4 and C2H6, (b) C3H6 and C3H8, and (c) n-C4H8 and n-C4H10 at 298 K. d Kinetic adsorption curves for C2-C4 olefins and paraffins at 0.5 bar and 298 K. Time-dependent in-situ FT-IR spectra for e C2H4 and f C3H6 in BFFOUR-Cu-dpds.

Full size image

The threshold pressure was measured to be 0.103 bar for C2H4, 0.107 bar for C3H6, and 0.157 bar for n-C4H8 at 298 K (Supplementary Fig. 21). As the temperature decreased to 273 K, the threshold pressure correspondingly reduced to 0.03 bar for C2H4, 0.028 bar for C3H6, and 0.039 bar for n-C4H8 (283 K). Intriguingly, the occurrence of similarities in threshold pressures and adsorption capacities of C2-C4 olefins, particularly for C2H4 and C3H6, was rare among reported flexible adsorbents49,50. The adsorption behaviors suggested that the adsorption processes were initiated by opening the pore channels and subsequently fulfilling the voids, rather than relying on molecule-size sieving and adsorption affinities. Therefore, we anticipated that the simultaneous molecular sieving for C2-C4 olefin/paraffin originated from the selective interlayer expansion driven by differences in guest-host interactions. As a proof-of-concept, we also measured the adsorption isotherms for benzene and cyclohexane (Supplementary Fig. 22a). As anticipated, benzene molecules possessing delocalized π-bonds could open the interlayer spaces, whereas the saturated structure of cyclohexane could not enter BFFOUR-Cu-dpds. Additionally, both styrene and ethylbenzene could be adsorbed by BFFOUR-Cu-dpds, indicating that carbon double bonds and benzene rings were capable of opening up interlayer spaces (Supplementary Fig. 22b).

Separation selectivity and adsorption enthalpy

Ideal adsorbed solution theory (IAST) was applied to estimate the separation selectivity for C2-C4 olefins/paraffins (0.5/0.5, v/v). The dual-site Langmuir-Freundlich (DLSF) model was employed to fit the adsorption isotherms with remarkable precision (Supplementary Fig. 23 and Supplementary Table 2). Due to the stepwise adsorption behaviors, the IAST selectivity curves of C2-C4 olefins/paraffins exhibited an increasing trend along the increase of adsorption amounts (Supplementary Fig. 24)51. Specifically, BFFOUR-Cu-dpds showed a high IAST selectivity for C2H4/C2H6 (68.8), C3H6/C3H8 (108.4), and n-C4H8/n-C4H10 (22.9) at 298 K and 1.0 bar, surpassing many leading adsorbents such as Ni-gallate (16.8 for C2H4/C2H6), NOTT-300 (48.7 for C2H4/C2H6), ZnAtzPO4 (12.4 for C2H4/C2H6), Fe2(m-dobdc) (60 for C3H6/C3H8), and Fe2(dobdc) (14.7 for C3H6/C3H8)26,52,53,54,55. To avoid the overestimation of separation performances in molecular-sieving adsorbents by IAST calculations, an intuitive evaluation based on the olefin-to-paraffin uptake ratio was employed23,56,57,58,59. As depicted in Supplementary Fig. 25, the uptake ratio reached 14.88 for C2H4/C2H6, 14.35 for C3H6/C3H8, and 10.53 for n-C4H8/n-C4H10, also outperforming most top-ranking adsorbents, including HIAM-301 (11.47 for C3H6/C3H8), Co-gallate (10.87 for C2H4/C2H6) and JNU-3 (1.21 for C3H6/C3H8), NOTT-300 (5.03 for C2H4/C2H6)11,27,48,52. Although the scarcity of reported data, the uptake ratio for n-C4H8/n-C4H10 also surpassed monolayer AgNO3/SiO2 sorbent (8.33) and Ag+ ion impregnated clay (2.97)60,61. To the best of our knowledge, BFFOUR-Cu-dpds represents the example of simultaneous sieving of C2-C4 olefins and paraffins.

The potential co-adsorption of paraffin molecules in gas-mixtures may arguably compromise the sieving efficiency of olefins/paraffins upon expanding interlayer spaces. Thus, the binary-component adsorption isotherms for C2-C4 olefin/paraffin (0.5/0.5, v/v) were collected (Supplementary Fig. 26). The proximity of uptakes and threshold pressures between olefin/paraffin mixtures and their single-component olefin isotherms indicated the preferential adsorption of olefins over paraffins, even in equimolar gas mixtures. Moreover, the kinetic adsorption curves and capacities for equimolar olefin/paraffin gas-mixtures at 1.0 bar also matched well with these of pure olefin components (Supplementary Fig. 27), e.g., 24.96 cm3 g−1 for C2H4/C2H6 (0.5/0.5, v/v, 1 bar) and 24.21 cm3 g−1 for C2H4 (0.5 bar), further confirming the negligible co-adsorption and solid molecular sieving effect.

After fitting C2-C4 olefins adsorption isotherms by the virial equation (Supplementary Fig. 28 and Supplementary Table 3), the isosteric heat of adsorption (Qst) was calculated to be 34.44, 31.13, and 33.37 kJ mol−1 at near-zero coverage for C2H4, C3H6, and n-C4H8, respectively (Supplementary Fig. 29). These adsorption enthalpy values were notably lower than many leading olefin-selective adsorbents (Supplementary Table 4 and Supplementary Fig. 5), such as PAF-1-SO3Ag (106 kJ mol−1 for C2H4), CuΙ@UiO-66-(COOH)2 (48.5 kJ mol−1 for C2H4), Co-gallate (41 kJ mol−1 for C3H6), and KAUST-7 (57.4 kJ mol−1 for C3H6), suggesting the easy regenerations for BFFOUR-Cu-dpds8,11,62,63. Moreover, the time-dependent in-situ FT-IR spectra was carried out to demonstrate the adsorption and desorption processes. As shown in Fig. 2e, C2H4 molecules were rapidly adsorbed within 3 min, as evidenced by the appearance of stretching vibrations of va,as(CH2) at 2987 ~ 3124 cm−1, v(C = C) at 1608 cm−1, δ(CH2) at 1444 cm−1, σ(CH2) at 948 cm−1, and w(CH2) at 825 cm−1 of adsorbed C2H414,64,65,66. The intensity of these vibrations gradually escalated until reaching equilibrium at 9 min. It should be noted that BFFOUR-Cu-dpds can be facilely regenerated under mild conditions of 10 mL min−1 He flows at 60 °C within 12 min, as evidenced by the disappeared vibrations of adsorbed C2H4. Similarly, the adsorption equilibrium of C3H6 was attained within 9 min in BFFOUR-Cu-dpds and could be easily regenerated under the same conditions within 12 min (Fig. 2f)64,67.

Gas-loaded single-crystal structures

Single-crystal to single-crystal transformations were attained upon C2H4 and C3H6 loadings (Supplementary Fig. 30). Despite multiple attempts, it was not able to obtain n-C4H8-loaded single-crystals. The SCXRD analysis of gas-loaded BFFOUR-Cu-dpds at 173 K revealed that olefin molecules were accommodated in the interlayer spaces as the preferential binding sites (Fig. 3a, b). Upon C2H4 adsorption, the B-F-Cu dihedral angle decreased from 166.82° to 152.53° to adapt C2H4 molecules, leading to the reduction of unit cell volume from 2816.3 Å3 to 2780.4 Å3 (Fig. 3E and Supplementary Fig. 31a). Due to the larger size of C3H6, the corresponding B-F-Cu dihedral angle was further reduced to 143.9° with a unit cell volume of 2801.4 Å3 (Fig. 3e and Supplementary Fig. 31b). Meanwhile, the pyridine rings underwent rotation to accommodate gas molecules, the dihedral angle between the pyridine plane and metal plane decreased from 68.35° to 67.63° and 66.54°on C2H4- and C3H6-loaded BFFOUR-Cu-dpds, respectively (Supplementary Fig. 32). Similarly, another dihedral angle decreased from 44.86° to 43.11° and 40.52° on BFFOUR-Cu-dpds upon C2H4- and C3H6-loading, demonstrating its structural self-adaptation in response to gas adsorptions.

Fig. 3: The crystal structure of gas-loaded BFFOUR-Cu-dpds.
figure 3

(a) C2H4-loaded and (b) C3H6-loaded BFFOUR-Cu-dpds viewed along c axis. Adsorption sites for (c) C2H4 and (d) C3H6 derived from SCXRD analysis of gas-loaded BFFOUR-Cu-dpds. e Structure change of BFFOUR-Cu-dpds framework after gas loadings. Color code: F green, Cu dark green, C gray, H (in framework) light gray, H (in hydrocarbons) dark purple, N blue, S yellow, blue line: C-H∙∙∙F interaction, green line: C-H∙∙∙π (olefin) binding.

Full size image

The binding conformations of C2H4 and C3H6 within BFFOUR-Cu-dpds framework were probed through SCXRD analysis, revealing multiple interactions between the olefins and interlayer spaces. For instance, each C2H4 molecule was associated with two BF4- anions via forming strong C-H∙∙∙F interactions, in which the negatively charged F atom interacted with weakly acidic H atoms of C2H4 at bonding distances of 2.65 Å, 2.75 Å, and 2.84 Å (Fig. 3c). Moreover, two intermolecular interactions via C-H∙∙∙π (C2H4) interactions with distances of 3.03 Å and 3.09 Å implied the tight contacts between C2H4 and BFFOUR-Cu-dpds. Similarly, each C3H6 molecule formed strong C-H∙∙∙F bindings (1.97 Å, 2.73 Å and 3.01 Å) with F atoms of two BF4- ions, as well as two C-H-π (C3H6) interactions (2.52 Å and 2.63 Å) with BFFOUR-Cu-dpds (Fig. 3d). The binding sites of n-C4H8 were investigated by computational simulations, and a detailed discussion will follow.

Modeling simulation studies

To elucidate interaction-selective adsorption behaviors, molecular simulations were conducted to illustrate the energy changes during corresponding transition states. The diffusion process of gas molecules into the BFFOUR-Cu-dpds framework can be rationally divided into four distinct transit states36. In the initial state (State I), the optimized framework and gas molecules were established as independent energy references. Subsequently, upon adsorption onto the surface of BFFOUR-Cu-dpds (State II), the gas molecules further diffused into pore channels (State III) and were firmly captured by adsorption sites (State IV). In terms of energy, the structural transformation during the diffusion process from State II to State III required additional energy input, representing the highest energy level in host-guest systems. Whereas, State IV was considered as having the minimal conformational energy landscape. Therefore, the energy difference for mitigating gas molecules from the adsorbent surface into the pore channel (State II to State III) was denoted as the energy barrier (ΔE′) for structural deformation and guest accumulation. In other words, it was a quantitative measurement of the permeation ease of the probe molecule into pore channels68,69.

In particular, C2H4, C3H6, and n-C4H8 exhibited comparable ΔE′ values of 0.502 eV, 0.448 eV, and 0.494 eV, respectively (Fig. 4a-c), which was consistent with their close threshold pressures as shown in the adsorption isotherms. In stark contrast, C2H6, C3H8, and n-C4H10 molecules with saturated carbon bonds and large sizes exhibited significantly higher ΔE′ of 2.704 eV, 5.617 eV, and 6.748 eV, respectively (Supplementary Figs 33–35). These findings indicated the considerable energy gap in inducing structural deformation by paraffin molecules to compensate for the more significant transformations of paraffin-loaded frameworks (Fig. 4d).

Fig. 4: The energy landscape of the host-guest conformations.
figure 4

The energy pathway and corresponding energy levels for (a) C2H4, (b) C3H6, and (c) n-C4H8. d Comparison of energy barriers of C2-C4 olefins and paraffins.

Full size image

The host-guest interactions were also illustrated through dispersion-corrected density functional theory (DFT-D) calculations. The interactions between C2H4/C3H6 and frameworks were almost identical with the SCXRD analysis, demonstrating the accuracy of depicting adsorption mechanisms (Supplementary Fig. 36). For the adsorption of n-C4H8, each molecule was trapped and bound to BF4- through three cooperative C-H∙∙∙F bonds with distances ranging from 2.31 Å to 3.14 Å, as well as two cooperative C-H∙∙∙π (n-C4H8) interactions with distances of 2.5 Å and 2.62 Å (Supplementary Fig. 36c). Additionally, charge transfer analysis on the gas-loaded BFFOUR-Cu-dpds structures revealed a significant charge bias between H atoms of olefins and F atoms of BF4- anions (Supplementary Fig. 37)70. The static binding energy was calculated to be 59.43 kJ mol−1, 71.87 kJ mol−1, and 82.52 kJ mol−1 for C2H4, C3H6, and n-C4H8, respectively.

Breakthrough experiments for olefin/paraffin separations

Dynamic breakthrough experiments were conducted on BFFOUR-Cu-dpds columns using binary gas-mixtures of olefin/paraffin (0.5/0.5, v/v) at 298 K to confirm its practical separation performances (Supplementary Fig. 38). As depicted in Fig. 5, BFFOUR-Cu-dpds demonstrated efficient separations of C2-C4 olefin/paraffin binary gas-mixtures within a single adsorption column. For the gas-mixture of C2H4/C2H6 (0.5/0.5, v/v), C2H6 was rapidly eluted from the column at a flow rate of 1.0 mL min−1, while C2H4 exhibited substantial retention in the column for 28 min until saturation (Fig. 5a). Notably, the efficient C2H4/C2H6 separation was also obtained despite slightly decreased retention time under humid conditions (RH = 61.9%). Similarly, both C3H8 and n-C4H10 immediately broke through the column, whereas C3H6 and n-C4H8 were detected at retention times of 32.2 min and 21 min, respectively (Fig. 5b, c). Considering that the IAST selectivity and uptake ratio are determined by equilibrium effect, the dynamic selectivity based on breakthrough curves was calculated to be 9.16, 8.76, and 3.18 for equimolar C2H4/C2H6, C3H6/C3H8, and n-C4H8/n-C4H10, respectively. These values demonstrate comparable performance to top-ranking adsorbents, such as NUS-6(Hf)-Ag (4.4 for C2H4/C2H6)71, ZJU-75a (14.7 for C3H6/C3H8)72, Y-abtc (8.3 for C3H6/C3H8)73 and KAUST-7 (12.0 for C3H6/C3H8)8. The dynamic adsorption capacity for C2H4, C3H6, and n-C4H8 were calculated to be 17.05 cm3 g−1, 19.97 cm3 g−1, and 14.43 cm3 g−1 respectively, which closely matched their static adsorption amounts at 0.5 bar. Furthermore, clean separations of C2H4/C2H6 and C3H6/C3H8 could also be achieved at higher flow rates of 2.0 and 4.0 mL min−1 (Fig. 5a, b). Note that facile adsorbent regeneration was a critical process to obtain high-purity olefins. After reaching the breakthrough point, the column was purged with He sweeping at 5 mL min−1 and 333 K (Supplementary Fig. 40). The productivity of C2H4 and C3H6 with ≥99.5% purity was calculated to be 11.92 L kg−1 and 14.19 L kg−1 in a single adsorption-desorption cycle, which was comparable to the top-ranking adsorbents including UTSA-280 (22.08 L kg−1 99.2% C2H4)17, NOTT-300 (19.66 L kg−1 99.2% C2H4)52, KAUST-7 (10.7 L kg−1 98.3% C3H6)8, and Co-gallate (14.9 L kg−1 98.7% C3H6)11. Meanwhile, the productivity of n-C4H8 was estimated to be 7.4 L kg−1 with ≥90% purity (Supplementary Fig. 40f).

Fig. 5: The olefin/paraffin separation performances.
figure 5

Breakthrough curves for an equimolar mixture of (a) C2H4/C2H6 (0.5/0.5, v/v) and (b) C3H6/C3H8 (0.5/0.5, v/v) at different flow rates on BFFOUR-Cu-dpds. c Breakthrough and desorption curves of n-C4H8/n-C4H10 (0.5/0.5, v/v) at 1.0 mL min−1. d The schematic diagram of one-step C2H4 separation from C2-C4 olefin/paraffin mixture. e Breakthrough curve of an equimolar C2-C4 olefin/paraffin six-component gas-mixture at 298 K. f Breakthrough curve of the desorption stream from BFFOUR-Cu-dpds column in the GBC-900 column at 353 K. g Dynamic olefin uptakes for three consecutive breakthrough cycles. h Five repeated adsorption isotherms for C2H4 o at 298 K and 1.0 bar.

Full size image

In addition, we introduced 20% inert He into olefin/paraffin gas-mixtures (He/olefin/paraffin, 0.2/0.4/0.4, v/v/v) to validate the absence of co-adsorption phenomenon during breakthrough experiments at a flow rate of 2.5 mL min−1. As shown in Supplementary Fig. 41, the paraffins and He almost concurrently outflow from the column, indicating that the negligible paraffin adsorptions in gas-mixtures similar to He74. The dynamic adsorption capacities of C2H4, C3H6, and n-C4H8 were determined to be 15.17 cm3 g−1, 17.55 cm3 g−1, and 13.37 cm3 g−1, respectively, which were in good agreement with their corresponding static adsorption amounts (20.09 cm3 g−1 for C2H4, 22.21 cm3 g−1 for C3H6, and 15.78 cm3 g−1 for n-C4H8) at 0.4 bar.

If the separation of C2-C4 olefins and paraffins can be achieved in a single adsorption column, it would greatly streamline the process flow for C2H4 production (Fig. 5d). The equimolar six-component gas-mixture of C2-C4 olefins and paraffins (C2H4/C2H6/C3H6/C3H8/n-C4H8/n-C4H10) was introduced into the BFFOUR-Cu-dpds column at 1.0 mL min−1 and 298 K. Surprisingly, three paraffins, i.e., C2H6, C3H8, and n-C4H10, were concurrently eluted from the column at the very beginning, whereas the corresponding olefins, i.e., C2H4, C3H6, and n-C4H8, broke through the column at 14 min (Fig. 5e). Similarly, the clean separations for C2-C4 olefins can also be maintained even at a higher flow rate of 2.0 mL min−1 (Supplementary Fig. 42). These results endow a streamlined process flow for direct C2H4 production from gas-mixtures of C2-C4 olefins and paraffins, if the second adsorption column could capture C3H6 and n-C4H8 during the desorption process of the first BFFOUR-Cu-dpds column (Fig. 5d and Supplementary Fig. 43)7,28,75. Porous carbons exhibited exceptional adsorption properties in the separation of gas molecules with varying carbon atoms with remarkable stability, cost-effectiveness, and high adsorption capacity5,76,77. We utilized the bamboo-derived granular carbon adsorbent (GBC-900), which was previously reported by our group, in the second adsorption column78. The adsorption isotherms showed steep C3H8 and n-C4H10 adsorptions at low pressures, whereas C2H4 was weakly adsorbed (Supplementary Fig. 44). After He sweeping for about 20 min to eliminate weakly absorbed paraffins, the BFFOUR-Cu-dpds column was heated to 353 K. The desorption stream primarily consisted of C2-C4 olefins with trace amounts of paraffins (Supplementary Fig. 45), which was subsequently passed through the GBC-900 column at 1.0 mL min−1 and 353 K. Notably, C2H4 was first detected at the exit at 24.5 min, followed by C3H6 and n-C4H8 at about 38.5 min (Fig. 5f). The high-purity (>99.99%) C2H4 product could be directly collected for a duration of 14 min.

The adsorbents were anticipated to exhibit outstanding stability and reusability. Three consecutive breakthrough experiments were conducted (Supplementary Fig. 46), resulting in nearly overlapped breakthrough curves and closely matched dynamic adsorption capacities (17.05 cm3 g−1 for C2H4, 19.97 cm3 g−1 for C3H6, and 14.43 cm3 g−1 for n-C4H8, Fig. 5g). Moreover, the breakthrough experiments conducted on a six-component gas-mixture of C2-C4 olefins and paraffins showed almost identical curves at flow rates of 1.0 and 2.0 mL min−1 in three successive trials (Supplementary Figs 47, 48). The recorded five repeated single-component olefin adsorption isotherms also demonstrated consistent adsorption capacities without any deterioration (Fig. 5h and Supplementary Fig. 49).

Discussions

The BFFOUR-Cu-dpds adsorbent was successfully designed and synthesized for the simultaneous C2-C4 olefin sieving from their corresponding paraffin. The differences in the strength of guest-host interactions caused the selective expansion of interlayer spaces for the sieving effect rather than molecular sizes. The unique feature made BFFOUR-Cu-dpds the versatile adsorbent for multiple olefin/paraffin pairs sieving. Gas-loaded single-crystal analysis and in-situ FT-IR spectra identified the accommodation patterns and interaction bonds between olefins and adsorbent. Computational simulations revealed that the energy input of structural transformation for olefins was significantly lower than that of paraffins. Furthermore, the direct production of C2H4 from a six-component gas-mixture was achieved in a two-column separation configuration, which could significantly streamline the olefin/paraffin separation processes.

Methods

Materials and reagents

All chemicals were commercially available and used without any further purification, including Cupric tetrafluoroborate hydrate (Cu(BF4)2•xH2O, Aldrich), 4,4’-dipyridyridylsulfide (dpds, 99%, Meryer), and anhydrous methanol (99%, Aladdin). All used gases, including ethylene (C2H4, 99.99%), ethane (C2H6, 99.99%), propylene (C3H6, 99.99%), propane (C3H8, 99.99%), nitrogen (N2, 99.999%), helium (He, 99.999%), C2H4/C2H6 (0.5/0.5, v/v), C3H6/C3H8 (0.5/0.5, v/v), n-C4H8/n-C4H10 (0.5/0.5, v/v), C2H4/C2H6/C3H6/C3H8 (0.25/0.25/0.25/0.25, v/v/v/v) were purchased from Nanchang Jiangzhu Gas Co., Ltd (China).

Preparation of BFFOUR-Cu-dpds

The synthesis of BFFOUR-Cu-dpds powder was as follows: 30 mL methanol solution of dpds (0.0881 g, 0.4 mmol) and an aqueous solution of Cu(BF4)2•xH2O (0.16 g, 0.25 mmol) were mixed at room temperature, then kept undisturbed for 48 h. Subsequently, the obtained purple powders were washed with methanol and dried under a high vacuum at 80 °C for 24 h.

The production of single crystals of BFFOUR-Cu-dpds was prepared by the static diffusion method by using a methanol solution (1.0 mL) of dpds (0.00881 g, 0.04 mmol) on an aqueous solution (1 mL) of Cu(BF4)2·xH2O (0.01 g, 0.07 mmol) at room temperature in a watch glass without any agitation. To control the reaction rate, 2 mL of a 1:1 methanol/H2O solution was layered between the upper and lower solutions. Light purple granular crystals were formed in one week.

Powder and single-crystal X-ray diffraction analysis

The powder diffraction patterns were recorded in a PANalytical Empyrean Series 2 diffractometer with Cu Kα radiation, a step size of 0.0167°, a scanning time of 15 s per step, and 2θ range of 5 to 90° at ambient temperature. The single-crystal X-ray diffraction data were collected at 193(2) K using a Bruker-AXS D8 VENTURE diffractometer equipped with a PHOTON-100/CMOS detector (GaKα, λ = 1.3414 Å). The indexing was performed using APEX2, while data integration and reduction were completed using SaintPlus 6.01. Absorption correction was conducted using the multi-scan method implemented in SADABS. The space group was determined using XPREP implemented in APEX2.1. The structures were solved by direct methods and refined by nonlinear least-squares on F2 using SHELXL-97, OLEX2 v1.1.5, and WinGX v1.70.01 program packages. All non-hydrogen atoms were refined anisotropically. The contribution of disordered solvent molecules was treated using the Squeeze routine implemented in Platon.

Thermogravimetric analysis (TGA)

The thermogravimetric analysis (TGA) data were obtained using a NETZSCH Thermogravimetric Analyzer (STA2500) under an N2 atmosphere with a heating rate of 10 °C min−1 and a temperature range of 25 to 800 °C.

Structure stability Tests

To evaluate the solvent stability, 100 mg BFFOUR-Cu-dpds were placed separately in 20 mL vials with 15 mL of different organic and pH solvents for 7 days. The resulting solid was filtered, activated at 80 °C for 12 h, and characterized by PXRD and adsorption analysis. The used acid was sulfuric acid solution and sodium hydroxide solution with various concentrations.

Gas adsorption measurements

A Micromeritics Three-Flex adsorption apparatus (Micromeritics Instruments, USA) was used to measure the adsorption isotherms of single-component gases (C2H4, C2H6, C3H6, C3H8) at 273 K, 298 K, and 323 K, while n-C4H8 and n-C4H10 were measured at 283 K, 298 K, and 313 K. To eliminate residual guest solvents in the framework, the fresh powder samples were subjected to evacuation under a high vacuum at 353 K for 24 h before each adsorption measurement. Liquid nitrogen and a dry ice-acetone bath were used for adsorption isotherms at 77 K or 195 K to calculate the Brunauer-Emmett-Teller (BET) specific surface area. The degas procedure was repeated on the same sample between measurements for 24 h.

Kinetic adsorption measurements

Intelligent Gravimetric Analyzer (IGA-100, HIDEN) was used to measure the time-dependent adsorption profiles of olefins and paraffins. About 80 mg of BFFOUR-Cu-dpds were subjected to evacuation under a high vacuum at 353 K for 24 h before each adsorption measurement. After being cooled to a specific temperature, a single-component gas or olefin/paraffin gas mixture was introduced into the chamber, and the mass of the sample loaded with gas-molecules was continuously recorded for 80 min.

FT- and In-situ infrared (IR) spectroscopic measurements

The IR spectra data were collected using a Bruker Tensor 27 FTIR spectrometer with a liquid N2-cooled mercury cadmium telluride (MCT-A) detector. A Specac Ltd. vacuum cell (product number P/N 5850c) was placed in the sample compartment, and the BFFOUR-Cu-dpds powder (~30 mg) was placed at the focal point of the beam. As for in-situ IR spectra, the cell was connected to gas lines for C2H4 and C3H6 and a vacuum line for evacuation. The sample was first activated by heating up to 80 °C under vacuum and then cooled to 25 °C for recording the reference spectrum. The spectra data were recorded during gas adsorption saturation, and the reference spectrum was retaken after loaded gases were fully evacuated. After adsorption saturation, the olefin gas was switched to 10 mL min−1 He for desorption at 60 °C.

Transient breakthrough experiments

The breakthrough experiments were carried out in a homemade setup. BFFOUR-Cu-dpds (1.17 g) was packed into a Φ 6 × 200 mm stainless steel column in the sample holder. The adsorption bed was purged with a carrier gas (He ≥ 99.999%) at 353 K for 12 h and then cooled to room temperature. Then, binary C2H4/C2H6 (0.5/0.5, v/v), C3H6/C3H8 (0.5/0.5, v/v), and n-C4H8/n-C4H10 (0.5/0.5, v/v) gas-mixture at a flow rate of 1 mL min−1 was separately introduced into the column at 298 K and 1 bar. Gas flows were regulated using a mass flow meter, and the outlet gas from the column was continuously monitored using gas chromatography (A91 Plus PANNA) with a flame ionization detector (FID). For the breakthrough experiments with water vapor, the gas mixture passed through a water vapor saturator under ambient conditions. The humidity of the mixture was measured by a relative humidity meter (UT333, UNI-T, China) at the inlet of the column. Following each breakthrough experiment, the sample was desorbed in-situ in the column with a He sweeping of 5 mL min−1 at 333 K for 2 h and attained high-purity olefin gas (99.5%). To illustrate the co-adsorption problem, the required gases He/C2H4/C2H6 (0.2/0.4/0.4, v/v/v), He/C3H6/C3H8 (0.2/0.4/0.4, v/v/v), and He/n-C4H8/n-C4H10 (0.2/0.4/0.4, v/v/v) were mixed independently in the pipeline and introduced to the column at a total flow rate of 2.5 mL min−1.

To further confirm its separation potential between olefin and paraffin and acquire higher purity ethylene (>99.99%), the equimolar C2-C4 olefin/paraffin six-component mixture was mixed in the pipeline and introduced to two columns. The first column was still primarily filled with 1.17 g BFFOUR-Cu-dpds, and then purged the trapped olefins with 1 mL min−1 He at 353 K in BFFOUR-Cu-dpds after first column adsorption saturation. The desorption stream went through the second column filled with granular porous carbons GBC-900 at 353 K. The outlet gas from the column was monitored using gas chromatography (A91 Plus PANNA) with an FID coupled with a thermal conductivity detector (TCD). Based on the mass balance, the gas adsorption capacities can be determined as follows:

$${Q}_{i}=\frac{{C}_{i}V}{22.4\times m}\times {\int }_{0}^{t}\left(1-\frac{F}{{F}_{0}}\right){dt}$$
(1)

Here Qi is the dynamic adsorption capacity of gas i (mmol g−1), Ci is the feed gas concentration, V is the volumetric feed flow rate (cm3 min−1), t is the adsorption time (min), F0 and F are the inlet and outlet gas molar flow rates, and m is the mass of the adsorbent (g).

Dynamic separation selectivity (α) obtained by calculating the integral area of the desorption curves of alkanes and olefins during the blowing process. The calculation formula is as follows:

$${S}_{i}={\int }_{t1}^{t2}\left(1-\frac{F}{{F}_{0}}\right){dt}$$
(2)
$$\alpha=\frac{{S}_{1}/{S}_{2}}{{y}_{1}/{y}_{2}}$$
(3)

Here Si is the integral area of the desorption curve of gas i, t1 and t2 represent the start and end times of desorption (min), F0 and F are the inlet and outlet gas molar flow rates, y1 and y2 represent the mole fractions of 1 and 2.

Calculation of isosteric heat of adsorption (Q st)

The adsorption data for C2H4 and C3H6 at temperatures of 273 K, 298 K, and 323 K were used to calculate the experimental adsorption enthalpy (Qst) and evaluate the binding strength between the adsorbent and adsorbate. For n-C4H8, the data was collected at temperatures of 283 K, 298 K, and 313 K. The viral equation was used to fit the adsorption curves:

$${{{{\mathrm{ln}}}}}P={{{{\mathrm{ln}}}}}N+\frac{1}{T}\mathop{\sum }\limits_{i=0}^{m}{a}_{i}{N}^{i}+\mathop{\sum }\limits_{i=0}^{n}{b}^{i}{N}^{i}$$
(4)
$${Q}_{{st}}=-R\mathop{\sum }\limits_{i=0}^{m}{a}_{i}{N}^{i}$$
(5)

Here N is gas uptake (mg g−1), P is pressure (mmHg), a and b are virial coefficients, m and n are the coefficients required to describe the isotherm adequately.

Ideal adsorbed solution theory (IAST) selectivity calculations

To evaluate the separation selectivity of BFFOUR-Cu-dpds for binary gas-mixtures, the dual-site Langmuir-Freundlich (DSLF) model was used to fit the single-component adsorption isotherms of C2H4, C2H6, C3H6, C3H8, n-C4H8, and n-C4H10. The adsorption isotherms and gas selectivity for C2H4/C2H6 (0.5/0.5, v/v), C3H6/C3H8 (0.5/0.5, v/v), and n-C4H8/n-C4H10 (0.5/0.5, v/v) at 298 K were calculated using the ideal adsorbed solution theory (IAST).

Dual-site Langmuir-Freundlich (DSLF) model is listed below:

$${{{{{\rm{N}}}}}}={{{\mbox{N}}}}_{1}^{\max }\times \frac{{{{\mbox{b}}}}_{1}{P}^{{c}_{1}}}{1+{{{\mbox{b}}}}_{1}{{{\mbox{P}}}}^{{c}_{1}}}+{{{\mbox{N}}}}_{2}^{\max }\times \frac{{{{\mbox{b}}}}_{2}{P}^{{c}_{2}}}{1+{{{\mbox{b}}}}_{2}{P}^{{c}_{2}}}$$
(6)

Here P (kPa) is the pressure of the bulk gas at equilibrium with the adsorbed phase, N (mol kg−1) is the adsorbed amount per mass of adsorbent, N1max and N2 max (mmol g−1) are the saturated capacities of site 1 and site 2, b1 and b2 (kPa−1) are the affinity coefficients of site 1 and site 2, and c1 and c2 represent the deviations from an ideal homogeneous surface.

The adsorption selectivity for the gas-mixtures of C2H4/C2H6 (0.5/0.5, v/v), C3H6/C3H8 (0.5/0.5, v/v), and n-C4H8/n-C4H10 (0.5/0.5, v/v) are defined by

$${S}_{{ads}}=\frac{{q}_{1}/{q}_{2}}{{p}_{1}/{p}_{2}}$$
(7)

In the above equation, q1 and q2 are the absolute component loadings of the adsorbed phase in the mixture with partial pressures p1 and p2. Adsorption isotherms and gas selectivity were calculated by IAST for mixed C2H4/C2H6 (0.5/0.5, v/v), C3H6/C3H8 (0.5/0.5, v/v), and n-C4H8/n-C4H10 (0.5/0.5, v/v).

Density functional theory (DFT) calculations

The first-principles DFT calculations were carried out using the CASTEP module in Materials Studio 202379. The van der Waals interactions were accounted for by incorporating dispersive forces into the conventional DFT80. The calculations were performed under the generalized gradient approximation (GGA) with Perdew-Burke-Ernzerhof (PBE) exchange-correlation functional. A cutoff energy of 600 eV and a 2 × 2 × 3 k-point mesh were used to achieve a total energy convergence of within 0.01 meV/atom. The crystal structures of the synthesized materials were optimized from the reported crystal structures. The pristine structure and an isolated gas molecule were optimized and relaxed as reference structures to calculate the binding energy. The gas molecules of C2H4, C3H6, and n-C4H8 were introduced to different locations of the channel pore, and their structures were fully relaxed. According to the output files from the CASTEP modules, the “NB dispersion corrected est. 0 K energy* (Ecor-0.5TS)” was selected as the energy at 0 K. The static binding energy (at T = 0 K) was calculated using the following equation:

$$\Delta {{\mbox{E}}}={{\mbox{E}}}({{\mbox{gas}}})+{{\mbox{E}}}\left({{\mbox{adsorbent}}}\right)-{{\mbox{E}}}({{\mbox{adsorbent}}}+{{\mbox{gas}}})$$
(8)

The energy barrier calculation method

The energy barrier calculations were carried out using the DMol3 module in Materials Studio 2023. The periodic slab models with periodic boundary conditions were used to represent the pore surface of BFFOUR-Cu-dpds. The unit cells were optimized until the interaction force was reduced to below 0.002 Ha/Å, achieving SCF convergence of 10−6. A Global orbital cutoff of 5.1 Å was employed. The unit conversion was performed using the equation: 1 Ha = 27.212 eV. The diffusion behavior of guest molecules was investigated by calculating transition state energies through the climbing nudged elastic band (cNEB) method81,82. First, the surface model and host structure were optimized by employing experimentally obtained single crystal structures as initial geometries, followed by full structural relaxation. The isolated guest molecules (C2H4, C2H6, C3H6, C3H8, n-C4H8, and n-C4H10) were placed in a supercell and relaxed to serve as references. Subsequently, the guest molecules were introduced onto both the host surface and various locations within the pores of BFFOUR-Cu-dpds. Full structural relaxation was performed for each configuration. The optimized configurations with the lowest energy after optimization were selected for subsequent analysis and calculation. The transition state search calculations were conducted to identify transition states associated with guest transport between the two energy minimum configurations.

The optimized host-guest structures were defined as state I, and its system energy was set as the reference. The state II denoted the initial state, characterized by introducing guests onto the surface of optimized host-guest structures. State III was defined as the transition state, and state IV represented the final state where guests were pulled into the pore channel of optimized host-guest structures. Notably, E(Initial State) = E(Final State). The energy barrier was determined using the following:

$$\Delta {{{\mbox{E}}}}^{\prime }={{\mbox{E}}}\left({{\mbox{Transition State}}}\right)-{{\mbox{E}}}({{\mbox{Initial State}}})$$
(9)

Where E(Transition State) is the transition energy, E(Initial State) is the energy of the optimized host-guest structure where guests were introduced onto the host surface.