Facilitated transport membranes by incorporating different divalent metal ions as CO₂ carriers

Abstract

Facilitated transport membranes by utilizing π complexation reactions between metal ions and penetrants have been actively explored, however, the different facilitated transport abilities of metal ions remains to be disclosed. In this study, poly(N-vinylimidazole) coated carbon nanotube particles (PVI@CNT) were prepared via precipitation polymerization of N-vinylimidazole monomers on a CNT surface. The PVI@CNT particles were then loaded with four kinds of divalent metal ions, Cu²⁺, Fe²⁺, Ca²⁺ and Mg²⁺, and incorporated into polyimide (PI) to prepare M²⁺–PVI@CNT hybrid membranes. The structure of M²⁺–PVI@CNT particles and PI–M²⁺–PVI@CNT membranes was analyzed by different characterization tools. Taking CO₂/CH₄ as the model separation system, the hybrid membranes containing Cu²⁺ and Fe²⁺ at filler content of 7 wt% showed the maximum increase of CO₂ permeability of 89% and 87% compared with those of a pristine PI membrane. Meanwhile, the selectivity of these membranes shows little increase. However, membranes containing Ca²⁺ and Mg²⁺ show only little enhancement in the separation properties. Such results can be interpreted based on the π complexation mechanism, transition metal ions Cu²⁺ and Fe²⁺ possess a strong CO₂ facilitated transport ability whereas main-group metal ions Ca²⁺ and Mg²⁺ possess a weak facilitated transport ability. Finally, a correlation of the electronegativity of metal ions with their CO₂ facilitated transport abilities was explored.

---

1 Introduction

In recent decades, membrane technology occupies an increasingly important position in many industrial fields due to its excellent separation performance and distinct energy-saving features.¹ Generally, there are two main kinds of membrane separation mechanisms: one is a physical mechanism including Knudsen diffusion, molecular sieving, solution-diffusion, etc., and the other is a chemical mechanism referring specifically to the facilitated transport mechanism.^2,3 Membranes based on the latter mechanism are collectively known as facilitated transport membranes, which are inspired by the facilitated diffusion in cell membranes in which molecules or ions can go across a biological membrane with the aid of specific carrier proteins.^4,5 Active carriers introduced into membranes can achieve the efficient transfer of the specific components as they have a reversible chemical reaction with the carriers, accordingly, these membranes often display excellent separation properties and great potential for industrial applications.^4,6–14 At present, facilitated transport membranes are primarily used in olefin/alkane separation, gasoline desulfurization and carbon capture including decarburization of natural gas and flue gas.^15–24

According to different reversible reactions, facilitated transport membranes can be divided into two categories: the acid–base reaction facilitated transport membrane and π complexation reaction facilitated transport membrane. These two types of membranes have different facilitated transport behavior, acid–base reaction facilitated transport membrane has stronger dependence on water and it is usually operated at the humidified state, while π complexation reaction facilitated transport membrane has good facilitated transport behavior at the dry state.^16,25–29 Besides, the active carriers of π complexation reaction facilitated transport membrane are mainly metals and metal ions, and therefore good antioxidant capabilities and long-term stability can be achieved in the membrane separation process. Currently, many metal ions have been explored as facilitated transport carriers for gas separation, such as Ag⁺, Pd²⁺, Hg²⁺, Cd²⁺, Cu⁺, Au⁺, Co²⁺, Fe²⁺, K⁺, Zn²⁺ and so on. In particular, Ag⁺ is used primarily in olefin/alkane separation and gasoline desulfurization due to its desirable complexation ability with olefin and thiophene.^22,30–32 Wessling et al.³¹ immobilized Ag⁺ into the supported ionic liquid membrane [Ag(olefin)_x]⁺[Tf₂N]⁻. In such kind of membrane, Ag⁺ was not only the component of the ionic liquid membrane, but also the olefin carrier because of the reversible complexation between Ag⁺ and olefin. Kang et al.³⁰ fabricated poly(2-ethyl-2-oxazoline) (POZ) membranes doped with AgBF₄ for propylene/propane separation and acquired the propylene solubility 25 times higher than that in pristine POZ membrane. The mechanism is that olefin molecules can offer their π electrons of occupied 2p orbitals to the empty s orbitals of Ag⁺, thereby generating σ-bonds. In addition, Ag⁺ also donates electrons of the occupied d orbitals back to the empty π*-2p antibonding orbitals to form π-bonding.³³ Liu et al.²³ loaded Ag⁺ on TiO₂ microspheres via electron donor–acceptor coordination bonds and introduced the filler into poly(dimethylsiloxane) (PDMS) matrix to fabricate facilitated transport membranes for gasoline desulfurization. They found that attributing to the specific reversible chemical reaction between Ag⁺ and thiophene, the optimum permeation flux and enrichment factor were 4.14 kg (m² h)⁻¹ and 8.56, respectively, which were much higher than polysiloxane membranes. Furthermore, it has been confirmed that positively polarized copper, K⁺ and Zn²⁺ as the carriers have an obvious effect on facilitating CO₂ transport.^34–40 Kang et al.³⁶ suggested positively polarized copper nanoparticles as a CO₂ carrier to prepare the ionic liquid BMIM⁺BF₄⁻/Cu membrane, and found that the incorporation of Cu nanoparticles improved CO₂ permeability of neat BMIM⁺BF₄⁻ from 17 GPU to 25 GPU, CO₂/CH₄ selectivity from 4.8 to 11, CO₂/N₂ selectivity from 5 to 11. They also employed K⁺ as a CO₂ carrier to prepare PVP/KF membrane.³⁵ It was found that PVP/KF membrane showed a maximum CO₂ permeability of 28 GPU with a CO₂/N₂ selectivity of 4.1. Chung et al.³⁷ loaded Zn²⁺ on POSS particle and then incorporated into polyimide to prepare hybrid membrane. The CO₂/CH₄ selectivity of membrane containing Zn²⁺ was 70% higher than that of membranes without Zn²⁺. They attributed the improvement to the π complexation reaction between C=O bond in CO₂ and Zn²⁺, thus facilitating CO₂ transport through the membrane. Aforementioned facilitated transport mechanism with metal aggregates and metal ions as CO₂ carriers can be collectively referred to as facilitated transport mechanism in terms of π complexation reaction. In comparison with acid–base reaction facilitated transport membranes, the π complexation reaction facilitated transport membranes have been much less exploited. Especially, the different facilitated transport abilities of different kinds of metal or metal ion carriers were rarely elucidated.³⁸ In order to achieve a rational design of membrane materials, it is essential to implement a systematic comparison about the facilitated transport abilities of different metals or metal ions and establish the preliminary guidelines for selecting suitable metal or metal ion carriers.

Taking CO₂/CH₄ as the model separation system, facilitated transport membranes containing different kinds of divalent metal ions were prepared. First, four kinds of composite particles, i.e., PVI@CNT loaded with Cu²⁺, Fe²⁺, Ca²⁺ and Mg²⁺ respectively were synthesized through precipitation polymerization and dip coating, then introduced into polyimide to prepare hybrid membranes. The morphology and structure of these composite particles and membranes were analyzed by a transmission electron microscopy (TEM), X-ray photoelectron spectroscopy (XPS), Fourier transform infrared spectra (FT-IR), field emission scanning electron microscope (FESEM) and differential scanning calorimeter (DSC). Afterwards, the CO₂ separation properties for CO₂/CH₄ separation system were evaluated and compared. Finally, the correlation between electronegativity of metal ions and their CO₂ facilitated transport abilities was elucidated.

2 Experimental

2.1 Materials and chemicals

Polyimide (PI, Matrimid 5218, purity 99%) was purchased from Huntsman Advanced Materials Americas Inc. Ethylene glycol dimethacrylate (EGDMA, purity 99%) and 3-(methacryloxy) propyltrimethoxysilane (MPS, purity 99%) were purchased from Alfa Aesar China Co., Ltd. Short-hydroxylate multi-walled carbon nanotubes (CNT, purity 95%) were obtained from Nanjing XF NANO Inc. N-Vinylimidazole (purity 99%) was purchased from J&K scientific Inc. Copper chloride (CuCl₂, purity 99.9%), ferric chloride (FeCl₂, purity 99.9%), calcium chloride (CaCl₂, purity 99.9%) and magnesium chloride (MgCl₂, purity 99.9%) were purchased from Aladdin (China). 2,2′-Azoisobutyronitrile (AIBN, purity 95%), N,N-dimethylformamide (DMF, purity 99.8%), acetonitrile (purity 99.5%), ethanol (purity 99.5%) were purchased from Tianjin Guangfu Fine Chemical Engineering Institute. Since all of the reagents were of analytical purity, there was no need for further purification prior to utilization.

2.2 Synthesis of the M²⁺–PVI@CNT composite particles

The M²⁺–PVI@CNT composite particles were synthesized as follows (M²⁺ refers to Cu²⁺, Fe²⁺, Ca²⁺ and Mg²⁺). First, CNT reacted with MPS in anhydrous alcohol to graft C=C bonds on its surface. Subsequently, the precipitation polymerization of N-vinylimidazole monomers was implemented on the modified CNT. More specifically, a certain amount of modified CNT was dispersed homogeneously in the acetonitrile–water solution (3 : 1 vol%) with the assistance of ultrasonic treatment and then it was mixed with 0.4 mL N-vinylimidazole monomer, 0.4 mL EGDMA and 0.016 g AIBN. Afterwards, the mixture was boiled for 80 min in a water bath to trigger the polymerization of N-vinylimidazole monomers on the CNT surface. The PVI@CNT was separated from the mixture by centrifugation and vacuum-dried over 48 h. Finally, M²⁺–PVI@CNT was prepared by magnetic stirring of the PVI@CNT in MCl₂ aqueous solution. The separation and purification process were the same as that of PVI@CNT mentioned above.

2.3 Membrane preparation

Pristine PI membrane and PI-based hybrid membranes were fabricated via the process of solvent casting and evaporation.⁴¹ For pristine membrane, PI powder was dissolved in DMF to obtain the 6 wt% PI solution. Then the PI solution was poured onto a glass mold and kept in the oven at 50 °C for 12 h. Subsequently, the oven temperature was elevated to 80 °C for another 12 h and then 120 °C for 48 h to evaporate DMF solvent completely. With regard to hybrid membranes, a certain amount of M²⁺–PVI@CNT was dispersed in DMF with the assistance of ultrasonic cell disruption machine and then further dispersed by ultrasonic treatment for 1 h to obtain the nanoparticle-DMF suspension. PI powder was dissolved in the aforementioned suspension to form the 6 wt% PI-DMF solution with dispersed M²⁺–PVI@CNT. The subsequent casting and heating treatment processes of hybrid membranes were the same as those of pristine PI membrane. The prepared hybrid membranes were named as PI–M²⁺–PVI@CNT(x) where x meant particle contents (wt%) of these membranes.

2.4 Characterization of nanoparticles and membranes

The structure and morphology of the composite particles were observed by a Tecnai G2 20 S-TWIN transmission electron microscopy (TEM). The cross-sectional structure of these membranes was observed by a Hitachi S-4800 field emission scanning electron microscope (FESEM). The chemical composition of the composite particles was obtained by a Nicolet 6700 Fourier transform infrared spectrometer. The glass transition temperatures (T_g) of these membranes were obtained by a differential scanning calorimeter (DSC). The data was received by a Netzsch DSC 200F3 instrument in the range of 50 °C to 400 °C.

2.5 Gas permeation experiments

Separation performance of these membranes was measured by the equipment described in our previous studies.⁴² Feed gas was the mixture of CO₂/CH₄ (30/70 vol%) and sweep gas was pure N₂. The experiment was conducted at 30 °C and 2 bar. Feed gas (CO₂/CH₄) and sweep gas (N₂) were all at dry state to exclude the interference of water. Dry membrane samples had been kept at 40 °C for 10 days after they were prepared and then used for the gas permeation test. The permeate gas composition was measured by an Agilent 6820 gas chromatography and the data was obtained every 15 min until the equilibrium was achieved. Gas permeability (P_i, Barrer, 1 Barrer = 10⁻¹⁰ cm³ (STP) cm (cm⁻² s⁻¹ cmHg⁻¹)) was gotten by eqn (1) and the final outcome was the average value of the data.where Q_i refers to gas volume flow (cm³ s⁻¹) (STP), l refers to membrane thickness (μm), A refers to effective membrane area (cm²), Δp_i refers to partial pressure difference between the feed side and permeate side of the membrane (cmHg). The CO₂/CH₄ selectivity (α_ij) was calculated by eqn (2).

3 Results and discussion

3.1 Characterization of M²⁺–PVI@CNT composite particles

3.1.1 TEM. The microstructure of four kinds of composite particles is observed by TEM images. As shown in Fig. 1, there is a visible polymer layer with uniform thickness in the outer layer of each CNT. Besides, all particles show interconnected network morphologies without appearance of impurities and agglomeration. In addition, the particles morphologies are very similar to each other, and to that of PVI@CNT particle in our previous studies, which indicates that loading different metal ions has little effect on the morphology of the particles.⁴²

Fig. 1 TEM images of particles (a) Cu²⁺–PVI@CNT, (b) Fe²⁺–PVI@CNT, (c) Ca²⁺–PVI@CNT and (d) Mg²⁺–PVI@CNT.

3.1.2 XPS. The element compositions of these composite particles are obtained by XPS. All the M²⁺–PVI@CNT particles are mainly comprised of C, N, O and metal elements. As shown in Table 1, the metal contents in these particles have only slight difference, indicating that these metal ions are successfully loaded onto the particles. Meanwhile, it can be considered the metal contents in these particles are the same due to the limit of XPS quantitative accuracy to element analysis. Besides, it has been indicated that the coordination is existed between imidazole group and metal ions in our previous studies, so that the metal contents can be stably existed in these composite particles.⁴²

Table 1 Element contents of particles

Particles	C content (%)	O content (%)	N content (%)	Metal content (%)
Cu^2+–PVI@CNT	70.23	25.05	1.23	1.02
Fe^2+–PVI@CNT	70.62	24.93	1.64	1.13
Ca^2+–PVI@CNT	70.13	25.13	1.33	0.82
Mg^2+–PVI@CNT	70.95	25.25	1.18	0.97

---

3.1.3 FTIR. As shown in Fig. 2, the chemical structures of the particles are verified by FTIR with PVI@CNT particles as the comparison. In our previous work, it has been known that the characteristic band at 1571 cm⁻¹ and 1500 cm⁻¹ correspond to C=N stretching vibration, and the band at 916 cm⁻¹ correspond to the imidazole ring stretching, which is consistent with the imidazole ring structure.^42–45 But compared with PVI@CNT, the characteristic bands of these particles loaded with different metal ions have smaller new peaks as well as peak shifts, as the dotted frame showed, which may be attributed to the influence of metal ions on the chemical structure of PVI.

Fig. 2 FTIR spectra of particles.

3.2 Characterization of membranes

3.2.1 SEM. As shown in Fig. 3, the morphology of membrane cross section is obtained by SEM. Compared with pristine PI membrane, four kinds of hybrid membranes display a network-like morphology without obvious agglomeration and defect, indicating all the particles have good compatibility with polymer matrix. Besides, there is no significant difference in morphology among these hybrid membranes. The reason may be that the particles morphologies are very similar to each other and they have the similar behavior in the polymer matrix.

Fig. 3 SEM images of cross section of membranes (a) unfilled PI membrane, (b) PI–Cu²⁺–PVI@CNT(5), (c) PI–Fe²⁺–PVI@CNT(5), (d) PI–Ca²⁺–PVI@CNT(5) and (e) PI–Mg²⁺–PVI@CNT(5).

3.2.2 DSC. Glass transition temperatures (T_g) of membranes are obtained by differential scanning calorimetry (DSC). Generally, the polymer chain rigidity can be investigated according to the glass transition temperature (T_g), and the polymer with a lower T_g value implies that it has more flexible polymer chain.⁴⁶ As shown in Fig. 4, the T_g of four kinds of hybrid membranes are slightly higher than that of PI pristine membrane. It is indicated that the incorporation of composite particles can slightly increase the polymer chain rigidity and reduce the chain mobility, which benefits the improvement of gas selectivity.^47–49

Fig. 4 DSC curves of membranes.

3.3 Membrane separation performances

As shown in Fig. 5, with the increased loading of Fe²⁺–PVI@CNT particle, the CO₂ permeability of membrane is significantly enhanced, and the maximum increment is 87% at filler content of 7 wt%. And there are quite similar trends in separation performance between the membrane filled with Cu²⁺–PVI@CNT and the membrane filled with Fe²⁺–PVI@CNT. The increment of CO₂ permeability is up to 89% at the same filler content. But the CO₂ permeability doesn't increase linearly with metal ion loading and it can be explained as followed. At low metal ion loading, the transfer channel formed by modified CNT is insufficient and discontinuous through the membrane, so that the facilitated transport behavior of membrane is severely restricted. When the metal ion loading is up to 7 wt%, the transfer channel is abundant enough to achieve interconnected network in the membrane, which leads to much enhanced CO₂ permeability. Besides, these membranes also show the elevated CO₂/CH₄ selectivity, and the maximum increment is achieved at filler content of 1 wt%, which is up to 31% for PI–Cu²⁺–PVI@CNT membrane and 26% for PI–Cu²⁺–PVI@CNT membrane. However, the selectivity of membranes with low composite particle content are higher than that of membranes with high composite particle content, which is attributed to the interfacial defects between polymer and composite particle when the loading is higher.

Fig. 5 (a) CO₂ permeability and (b) CO₂/CH₄ selectivity of membranes at dry state.

From the TEM images of particles, nanopores in CNT have been blocked by polymer layer after modified with PVI, therefore PVI@CNT and M²⁺–PVI@CNT particles are impermeable. Besides, it can be observed that these nanoparticles are quite small without obvious impurities and agglomeration, so that polymer/particle interfacial area is high enough to achieve efficient utilization of carriers.⁵⁰ Besides, from the result of DSC, it is known that these four composite particles with the same metal contents have a little influence on the polymer chain mobility. As a result, these metal ions must have an important impact on membrane separation properties. In our previous experiment, we have found that filling PVI@CNT particles into PI membrane will decline the CO₂ permeability at the dry state, while PI–Zn²⁺–PVI@CNT membranes show an increase of CO₂ permeability with a maximum increment of 93% at filler content of 7 wt% as well as enhanced CO₂/CH₄ selectivity of 42% at filler content of 1 wt%.⁴² And in this study, composite particles loaded with Cu²⁺ and Fe²⁺ can increase CO₂ permeability and CO₂/CH₄ selectivity of membranes simultaneously. It is revealed that not only Zn²⁺, but also Cu²⁺ and Fe²⁺ can react with CO₂ by π complexation, so that they can be regarded as CO₂ carriers and play a role in facilitating CO₂ transport. It has been demonstrated that transition metal ions can form both normal σ-bond and π-bond with CO₂ and olefin.^38,51,52 And the π-bond is the major contributor in the metal–olefin bonding.^53,54 Fig. 6(a and b) shows the interaction between metal ion and CO₂ that is much similar to the interaction between olefin and metal ion. The empty s orbital of metal ion can receive the π electrons of CO₂ to form σ-bonds and the unoccupied d orbital of metal ion also donates d electrons to the empty π* antibonding orbitals of CO₂ to form π-bonds, thus generating transient presence of metal ions–CO₂ coordination compound. The unstable compound is easy to be resolved to release CO₂, so metal ions can react with CO₂ reversibly, which is the basis of being the CO₂ carrier.^7,16,55 Besides, it has been indicated that the mono bridged complex is formed between imidazole group and metal ions and metal salts have already been dissociated to free metal ions, therefore, these transition metal ions with unoccupied s orbital and unshared d electrons can interact with CO₂.^42,56

Fig. 6 Interaction between metal ion and CO₂ (a) σ-bonding and (b) π-bonding (M = metal ion).

However, as shown in Fig. 5, there are only a few changes in the CO₂ permeability when membranes are filled with Mg²⁺–PVI@CNT or Ca²⁺–PVI@CNT. The reason may be that both Ca²⁺ and Mg²⁺ have very stable configuration of extra-nuclear electron and therefore they cannot form π-bond with CO₂. So that Mg²⁺ and Ca²⁺ have weak interaction with CO₂, hence they won't act as effective CO₂ carriers.^7,57 But PI–Ca²⁺–PVI@CNT membranes and PI–Mg²⁺–PVI@CNT membranes show a little increase in the CO₂/CH₄ selectivity with the increased loading of fillers, and the variation trend is different from the other two membranes, which is may be due to the better interfacial morphology.⁵⁸ Although loading different metal ions does not change the morphology of the particles and all these particles show the almost same morphology with each other, there are still some differences existed between these fillers containing different metal ion. So that it is inferred that fillers containing Ca²⁺ and Mg²⁺ may have the better interfacial interaction with membrane matrix at the higher filler content.

The effect of feed gas pressure on CO₂ permeability and CO₂/CH₄ selectivity of hybrid membranes is investigated. As shown in Fig. 7(a), both CO₂ permeability and CO₂/CH₄ selectivity of PI–Fe²⁺–PVI@CNT membrane and PI–Cu²⁺–PVI@CNT membrane show a decreasing trend as the feed pressure increases. The phenomenon may result from the carrier saturation, therefore it also confirms the CO₂ facilitated transport ability of Fe²⁺ and Cu²⁺.¹⁴ However, Fig. 7(b) shows a different change trend in the CO₂/CH₄ selectivity of PI–Fe²⁺–PVI@CNT membrane and PI–Cu²⁺–PVI@CNT membrane. It can be explained as follows: as the feed pressure increases, the polymer chains are compacted, which can reduce the CO₂ permeability and improve the CO₂/CH₄ selectivity.⁵⁹

Fig. 7 Effect of feed gas pressure on CO₂ permeability and CO₂/CH₄ selectivity of hybrid membranes with 7 wt% filler content at dry state.

3.4 Comparison of metal ions for facilitating CO₂ transport

In this study, we have compared the separation performances of different kinds of hybrid membranes to further explain the facilitated transport mechanism of metal ions. As shown in Fig. 5, compared with Ca²⁺–PVI@CNT and Mg²⁺–PVI@CNT, both Cu²⁺–PVI@CNT and Fe²⁺–PVI@CNT in the membranes can make a difference in the enhancement of CO₂ permeability and CO₂/CH₄ selectivity. Besides, their separation performances have the similar trend as the particle content increases.

From the theoretical point of view, it is difficult for main-group metal ions to form strong coordination bond with CO₂ since they have a much stable extra-nuclear electron configuration. But transition metal ions have higher nuclear charge number, smaller atomic radius and the non-full 3d orbital, so that their electron orbits are more complex and they are more susceptible to take place a coordination reaction with CO₂ than main-group metal ions. Besides, the M²⁺–PVI@CNT particles formed by the coordination of imidazole group and metal ions have a certain polarity and it is very well known that the quadrupole moment value of CO₂ is higher than that of CH₄. Therefore CO₂ will be largely affected by these particles, especially, the particles containing transition metal ions with larger polarity.⁶⁰ From the analysis above, it can be indicated that transition metal ions have higher affinity with CO₂ than main-group metal ions. More similar in-depth study can be found in olefin separation and gasoline desulfurization.^52,61,62 As we known, Cu²⁺ and Fe²⁺ are transition metal ions, while Ca²⁺ and Mg²⁺ are main-group metal ions. Combined with the results described above, it can be speculated that transition metal ions could serve as effective CO₂ carriers and then significantly enhance membrane separation properties, while main-group metal ions, more specifically alkali earth metal ions, show no or a little facilitated transport ability.

Electronegativity is the ability that an element attract electrons from other elements, so that a metal ion with higher electronegativity is considered to form its metal-complex ligand more easily, which indicates that electronegativity positively correlates with the π-complexation strength generally. Theoretically, it has been broadly accepted that electronegativity can represent the π complexation strength between a metal ion and ligands. Hur et al.⁶³ studied the competition of Cu²⁺, Pb²⁺, and Cd²⁺ adsorption on the magnetic graphene oxide (GO) surface and found metal capacity was consistent with metal electronegativity: Pb²⁺ > Cu²⁺ > Cd²⁺. Similar conclusions could be found as the adsorbent was goethite instead of GO.⁶⁴ Catherine et al.⁶⁵ pointed out that the adsorption process of metal ions onto activated carbon cloths (ACCs) may be related to their electronegativity and ionic radius. Dissouky et al.⁶⁶ synthesized a new kind of the Schiff base named HPIB and then studied the stepwise stability constants of its metal complexes with Cu²⁺, Ni²⁺, Co²⁺ and Mn²⁺. They discovered that ΔG⁰ and ΔH⁰ of the complex formation process were closely related to the ionic radius, the electronegativity, the ionization enthalpy, and the hydration enthalpy of the metal ions, so that the stabilities of the complexes followed the order Cu²⁺ > Ni²⁺ > Co²⁺ > Mn²⁺. Hanna et al.⁶⁷ used the cellulose acetate (CTA) as the ligand and laid emphasis on the formation of different CTA–metal ion complexes prepared from different metal salts. They observed that the activation energy from 33 kJ mol⁻¹ for CTA–Cu(II) complex decreased to 28.15 kJ mol⁻¹ for CTA–Cr(III) complex. This result can be explained as follows: the electronegativity of Cu²⁺ was higher than that of Cr²⁺, which enforced chemical bonding between the cellulose chain and metal ion. In field of gasoline desulfurization, it has been revealed that electronegativity could represent the intensity of the π complexation between olefins and metals.⁶¹ Consequently, we may use the electronegativity to represent the intensity of the π complexation between CO₂ and metal ions.

However, it must be noted that the higher bond strength between CO₂ and carriers is not equivalent to the higher facilitated transport ability. In general, the ideal CO₂ carrier should have a moderate CO₂ binding ability, which demands that the carrier should combine with CO₂ quickly and the bond can be also broken readily. In other words, the combination between CO₂ and carriers do not need to be close too much. With electronegativity considered as the bond strength between CO₂ and metal ions, we need to relate the electronegativity of metal ions with their facilitated transport ability, so that we can determine the most suitable metal ions as the CO₂ carrier in terms of the most suitable electronegativity.

The facilitated transport abilities of divalent metal ions involved in this study and previous study are contrasted and the result is as follow: Zn²⁺ > Cu²⁺ ≈ Fe²⁺ > Mg²⁺ > Ca²⁺.⁴² Electronegativity of elements corresponding to these metal ions is shown in Table 2. It can be found that Zn²⁺ with the best facilitated transport ability shows the moderate electronegativity among these metal elements, while electronegativity of Cu²⁺ and Fe²⁺ is close to each other and a little higher than that of Zn²⁺. They also show almost the same ability in facilitating CO₂ transport. However, Mg²⁺ and Ca²⁺ have the relatively lower electronegativity as well as the lower facilitated transport ability. In especial, Ca²⁺ has the lowest electronegativity that corresponds exactly to its poorest separation performance. So it can be revealed that facilitated transport ability of divalent metal ions has tight relevance to the electronegativity and the appropriate electronegativity of metallic carriers may be in the range of 1.5–1.9. Similarly, for olefin separation, metals are suitable to be carriers when their electronegativity is in the scope of 1.6–2.3.⁶¹ Overall, Zn²⁺ may have the most suitable electronegativity to facilitate CO₂ transport and can be as the benchmark for choosing CO₂ carriers: the greater a divalent metal ion differs from Zn²⁺ in electronegativity, the weaker facilitated transport ability it will have. In addition, divalent metal ions with similar electronegativity will also have the similar facilitated transport ability. However, to confirm the universality of this assumption, it needs further relevant experimental studies on more metal ions.

Table 2 Electronegativity of metal elements in experiment

Element	Cu	Fe	Zn	Mg	Ca
Electronegativity	1.90	1.83	1.65	1.31	1.00

---

4 Conclusions

In this study, four different CO₂ facilitated transport hybrid membranes are prepared by introducing different composite particles loaded with Cu²⁺, Fe²⁺, Ca²⁺ and Mg²⁺ respectively into polyimide matrix. Compared with pristine PI membrane, hybrid membranes containing Cu²⁺ and Fe²⁺ perform 89% and 87% increase in CO₂ permeability, respectively. Such increase can be explained based on π complexation mechanism, transition metal ions Cu²⁺ and Fe²⁺ serve as CO₂ facilitated transport carriers, while main-group metal ions Ca²⁺ and Mg²⁺ possess weak facilitated transport ability. The reason is that transition metal ions with more complex electron orbitals are easier to form coordination compounds with CO₂. Furthermore, considering that electronegativity can represent the π complexation intensity between a metal ion and CO₂, a further relation between electronegativity of divalent metal ions and their CO₂ facilitated transport abilities has been studied. The finding is that the higher electronegativity is not equivalent to the higher facilitated transport ability. According to the existing results, Zn²⁺ may have the most suitable electronegativity to facilitate CO₂ transport. The smaller difference in electronegativity between a divalent metal ions and Zn²⁺, the stronger facilitated transport ability it will have. This finding is not only suitable for CO₂ separation system, but also suitable for many other systems where π complexation reaction can be utilized, such as olefin/alkane separation, thiophene/gasoline separation and so forth.

Acknowledgements

The authors gratefully acknowledge the financial support from the National High Technology Research and Development Program of China (2012AA03A611), State Key Laboratory of Organic–Inorganic Composites (oic-201501009), the National Science Fund for Distinguished Young Scholars (No. 21125627), Program of Introducing Talents of Discipline to Universities (B06006), State Key Laboratory of Separation Membranes and Membrane Processes, Tianjin Polytechnic University, (No. M1-201501).

References

1. S. L. Liu, L. Shao, M. L. Chua, C. H. Lau, H. Wang and S. Quan, Prog. Polym. Sci., 2013, 38, 1089–1120.
2. P. Pandey and R. S. Chauhan, Prog. Polym. Sci., 2001, 26, 853–893.
3. N. V. Blinova and F. Svec, J. Membr. Sci., 2012, 423, 514–521.
4. J. Zhao, X. T. Zhao, Z. Y. Jiang, Z. Li, X. C. Fan, J. N. Zhu, H. Wu, Y. L. Su, D. Yang, F. S. Pan and J. F. Shi, Prog. Polym. Sci., 2014, 39, 1668–1720.
5. E. M. Kramer and M. J. Bennett, Trends Plant Sci., 2006, 11, 382–386.
6. C. X. Zhang, Z. Wang, Y. Cai, C. Y. Yi, D. X. Yang and S. J. Yuan, Chem. Eng. J., 2013, 225, 744–751.
7. S. Zhao, Z. Wang, Z. H. Qiao, X. Wei, C. X. Zhang, J. X. Wang and S. C. Wang, J. Mater. Chem. A, 2013, 1, 246–249.
8. X. Q. Li, Y. D. Cheng, H. Y. Zhang, S. F. Wang, Z. Y. Jiang, R. L. Guo and H. Wu, ACS Appl. Mater. Interfaces, 2015, 7, 5528–5537.
9. Q. P. Xin, Z. Li, C. D. Li, S. F. Wang, Z. Y. Jiang, H. Wu, Y. Zhang, J. Yang and X. Z. Cao, J. Mater. Chem. A, 2015, 3, 6629–6641.
10. M. Esmaeili, S. S. Madaeni, J. Barzin and N. Yousefimehr, Sep. Purif. Technol., 2011, 81, 371–383.
11. M. Shoji, K. Oyaizu and H. Nishide, Polymer, 2008, 49, 5659–5664.
12. J. H. Ryu, H. Lee, Y. J. Kim, Y. S. Kang and H. S. Kim, Chem.–Eur. J., 2001, 7, 1525–1529.
13. S. Kasahara, E. Kamio, R. Minami and H. Matsuyama, J. Membr. Sci., 2013, 431, 121–130.
14. L. Y. Deng, T. J. Kim and M. B. Hagg, J. Membr. Sci., 2009, 340, 154–163.
15. J. Y. Liao, Z. Wang, C. Y. Gao, S. C. Li, Z. H. Qiao, M. Wang, S. Zhao, X. M. Xie, J. X. Wang and S. C. Wang, Chem. Sci., 2014, 5, 2843–2849.
16. Y. F. Li, S. F. Wang, G. W. He, H. Wu, F. S. Pan and Z. Y. Jiang, Chem. Soc. Rev., 2015, 44, 103–118.
17. L. Ansaloni, Y. N. Zhao, B. T. Jung, K. Ramasubramanian, M. G. Baschetti and W. S. W. Ho, J. Membr. Sci., 2015, 490, 18–28.
18. A. H. Gorji and T. Kaghazchi, Can. J. Chem. Eng., 2008, 86, 1039–1046.
19. L. Xiong, S. Gu, K. O. Jensen and Y. S. S. Yan, ChemSusChem, 2014, 7, 114–116.
20. M. J. Kim, Y. I. Park, K. H. Youm and K. H. Lee, J. Membr. Sci., 2004, 245, 79–86.
21. J. H. Kim, B. R. Min, J. Won and Y. S. Kang, J. Membr. Sci., 2003, 227, 197–206.
22. T. C. Merkel, R. Blanc, I. Ciobanu, B. Firat, A. Suwarlim and J. Zeid, J. Membr. Sci., 2013, 447, 177–189.
23. W. P. Liu, B. Li, R. J. Cao, Z. Y. Jiang, S. N. Yu, G. H. Liu and H. Wu, J. Membr. Sci., 2011, 378, 382–392.
24. A. Hussain and M. B. Hagg, J. Membr. Sci., 2010, 359, 140–148.
25. R. M. Hodge, T. J. Bastow, G. H. Edward, G. P. Simon and A. J. Hill, Macromolecules, 1996, 29, 8137–8143.
26. L. A. El-Azzami and E. A. Grulke, J. Membr. Sci., 2008, 323, 225–234.
27. L. Liu, A. Chakma and X. Feng, J. Membr. Sci., 2008, 310, 66–75.
28. L. A. El-Azzami and E. A. Grulke, J. Polym. Sci., Part B: Polym. Phys., 2007, 45, 2620–2631.
29. Y. Zhao and W. S. W. Ho, Ind. Eng. Chem. Res., 2013, 52, 8774–8782.
30. S. U. Hong, J. H. Jin, J. Won and Y. S. Kang, Adv. Mater., 2000, 12, 968–971.
31. F. Pitsch, F. F. Krull, F. Agel, P. Schulz, P. Wasserscheid, T. Melin and M. Wessling, Adv. Mater., 2012, 24, 4306–4310.
32. S. W. Kang, J. H. Kim, K. Char, J. Won and Y. S. Kang, J. Membr. Sci., 2006, 285, 102–107.
33. J. Won, D. B. Kim, Y. S. Kang, D. K. Choi, H. S. Kim, C. K. Kim and C. K. Kim, J. Membr. Sci., 2005, 260, 37–44.
34. I. S. Chae, M. Kim, Y. S. Kang and S. W. Kang, J. Membr. Sci., 2014, 466, 357–360.
35. J. H. Oh, Y. S. Kang and S. W. Kang, Chem. Commun., 2013, 49, 10181–10183.
36. J. H. Lee, J. Hong, J. H. Kim, Y. S. Kang and S. W. Kang, Chem. Commun., 2012, 48, 5298–5300.
37. F. Y. Li, Y. Li, T. S. Chung and S. Kawi, J. Membr. Sci., 2010, 356, 14–21.
38. Y. Li and T. S. Chung, J. Membr. Sci., 2008, 308, 128–135.
39. G. Nasr, T. Macron, A. Gilles, C. Charmette, J. Sanchez and M. Barboiu, Chem. Commun., 2012, 48, 11546–11548.
40. O. Hosseinkhani, A. Kargari and H. Sanaeepur, J. Membr. Sci., 2014, 469, 151–161.
41. L. Wang, Y. M. Cao, M. Q. Zhou, S. J. Zhou and Q. Yuan, J. Membr. Sci., 2007, 305, 338–346.
42. S. Wang, Y. Liu, M. Zhang, D. Shi, Y. Li, D. Peng, G. He, H. Wu, J. Chen and Z. Jiang, J. Membr. Sci., 2016, 505, 44–52.
43. M. Andersson, O. Hansson, L. Ohrstrom, A. Idstrom and M. Nyden, Colloid Polym. Sci., 2011, 289, 1361–1372.
44. V. Krishnakumar and R. J. Xavier, Spectrochim. Acta, Part A, 2004, 60, 709–714.
45. S. N. Wijenayake, N. P. Panapitiya, C. N. Nguyen, Y. Huang, K. J. Balkus, I. H. Musselman and J. P. Ferraris, Sep. Purif. Technol., 2014, 135, 190–198.
46. B. Yu, H. L. Cong, Z. J. Li, J. G. Tang and X. S. Zhao, J. Appl. Polym. Sci., 2013, 130, 2867–2876.
47. M. A. Aroon, A. F. Ismail, M. M. Montazer-Rahmati and T. Matsuura, Sep. Sci. Technol., 2010, 45, 2287–2297.
48. M. A. Aroon, A. F. Ismail, M. M. Montazer-Rahmati and T. Matsuura, J. Membr. Sci., 2010, 364, 309–317.
49. A. C. Lua and Y. Shen, J. Appl. Polym. Sci., 2013, 128, 4058–4066.
50. S. Kim and E. Marand, Microporous Mesoporous Mater., 2008, 114, 129–136.
51. D. H. Gibson, Chem. Rev., 1996, 96, 2063–2096.
52. J. X. Zhou, Y. C. Zhang and X. W. Guo, Chem. Res. Chin. Univ., 2005, 21, 217–219.
53. H. Y. Huang, J. Padin and R. T. Yang, Ind. Eng. Chem. Res., 1999, 38, 2720–2725.
54. H. Y. Huang, J. Padin and R. T. Yang, J. Phys. Chem. B, 1999, 103, 3206–3212.
55. M. K. Mondal, H. K. Balsora and P. Varshney, Energy, 2012, 46, 431–441.
56. A. Tarushi, X. Totta, C. P. Raptopoulou, V. Psycharis, G. Psomas and D. P. Kessissoglou, Dalton Trans., 2012, 41, 7082–7091.
57. K. Ramasubramanian, Y. Zhao and W. S. Winston Ho, AIChE J., 2013, 59, 1033–1045.
58. T. S. Chung, L. Y. Jiang, Y. Li and S. Kulprathipanja, Prog. Polym. Sci., 2007, 32, 483–507.
59. Y. Liu, S. N. Yu, H. Wu, Y. F. Li, S. F. Wang, Z. Z. Tian and Z. Y. Jiang, J. Membr. Sci., 2014, 469, 198–208.
60. W. Yave, A. Szymczyk, N. Yave and Z. Roslaniec, J. Membr. Sci., 2010, 362, 407–416.
61. R. Faiz and K. Li, Chem. Eng. Sci., 2012, 73, 261–284.
62. Y. Wang, F. H. Yang, R. T. Yang, J. M. Heinzel and A. D. Nickens, Ind. Eng. Chem. Res., 2006, 45, 7649–7655.
63. J. Hur, J. Shin, J. Yoo and Y.-S. Seo, Sci. World J., 2015, 2015, 836287.
64. C. A. Christophi and L. Axe, J. Environ. Eng., 2000, 126, 66–74.
65. C. Faur-Brasquet, Z. Reddad, K. Kadirvelu and P. Le Cloirec, Appl. Surf. Sci., 2002, 196, 356–365.
66. B. Jeragh, D. Al-Wahaib, A. A. El-Sherif and A. El-Dissouky, J. Chem. Eng. Data, 2007, 52, 1609–1614.
67. A. A. Hanna, A. H. Basta, H. El-Saied and I. F. Abadir, Polym. Degrad. Stab., 1999, 63, 293–296.

---