Separation and Purification Technology2.4. Fabrication of membranes The Schematic representation of the membrane fabrication is shown in Scheme 1. The polymers were dissolved in DMF

Contents lists available at ScienceDirect

Separation and Purification Technology

journal homepage: www.elsevier.com/locate/seppur

Study on the effect of crosslinking temperature on microporous polyamidemembrane structure and its nitrogen/cyclohexane separation performance

Yuan Chena, Jinchao Qina, Tong Tonga, Haoli Zhoua,⁎, Xingzhong Caob, Wanqin Jina,⁎

a State Key Laboratory of Materials-Oriented Chemical Engineering, College of Chemical Engineering, Nanjing Tech University, 30 Puzhu Road(s), Nanjing 210009, PRChinab Institute of High Energy Physics, Chinese Academy of Sciences, Beijing 100049, China

A R T I C L E I N F O

Keywords:Microporous polymerMembranePropertiesCrosslinking temperature

A B S T R A C T

Membrane performance is a vital factor for the membrane-based waste VOCs treatment, which is not onlyaffected by the properties of monomers or polymers, but also by crosslinking temperature during the membrane-forming procedure. However, little attention has been paid to the latter. In this work, the effect of the cross-linking temperature on membrane structure and separation performance was studied. The properties of polymersand membranes were first characterized by X-ray photoelectron spectroscopy, positron annihilation lifetimespectra, and so on. Apparent differences in physicochemical properties between polymers and membranes can beobserved. Different crosslinking temperatures were thus chosen to investigate their effects on the membranestructure, and the separation performances of membranes obtained under different crosslinking temperatures forthe molecular sieving of nitrogen over cyclohexane were also evaluated. Results showed that higher temperatureleads to higher crosslink and smaller pore size so that 99% rejection can be obtained when the crosslinkingtemperature over 80 °C was used. While the permeability of nitrogen slightly decreases with increasing tem-perature from 50 to 120 °C.

1. Introduction

VOCs (volatile organic compounds) emitted from many chemicaland pharmaceutical processes not only cause harm to the environmentand human beings, but also lead to the waste of resources [1,2].Therefore, it is necessary to limit their emission. Membrane technologyis regarded as a very effective way for VOCs control because of its highefficiency, easy operation, no need for further regeneration, and en-ergy-efficient property, etc. [3,4].

In the membrane process, membrane performance is a crucial factorthat affects their industrial application, and membrane structure plays avital role in the improvement of membrane performance [5]. Manymonomers have thus been synthesized for the tuning of the membranemicrostructure to enhance membrane performance [6–8]. Besides thedevelopment of monomers, preparation parameters such as crosslinkingtemperature can also have a significant influence on membrane struc-ture [9–11], finally affecting membrane performance. At the same time,the thermal crosslinking method is widely used in the membrane-forming procedure for the preparation of non-porous poly-dimethylsiloxane (PDMS) [3,12] and polyether block amide (PEBA)membrane [13], etc. or microporous polyamide [8] and polyimide

membranes, etc. [14]. Berean et al. [11] studied the effect of thecrosslinking temperature on the gases (CO2, N2, and CH4) permeation ofthe PDMS membrane. Results showed the membrane obtained at atemperature of 75 °C shows the highest gas permeabilities because ofthe relaxed polymer chains, which resulted from the reduction of thecrosslinking density and the increasing fractional free volume withinthe membrane, providing more diffusional channels for these gases.

Besides the non-porous PDMS membranes, microporous polymermembranes can also be significantly affected by crosslinking tempera-tures. As Budd et al. [15] reported that when microporous PIM-1polymer is made into the membrane by thermal crosslinking method,the BET surface area decreases from 860 to ∼600m2/g, indicating thatcrosslinking process affects the membrane structure because the coatingprocedure reportedly does not affect the intrinsic microporosity of apolymer [16], which is also true of the dissolution procedure. There-fore, only the thermal crosslinking process should affect the char-acteristics of a polymer. Furthermore, it is reported that even minorvariations of the membrane structure could lead to a dramatic differ-ence in permeability [17].

In this work, the microporous polyamide was selected to reveal theeffect of crosslinking temperature on membrane structure, because the

https://doi.org/10.1016/j.seppur.2020.117401Received 30 May 2020; Received in revised form 14 July 2020; Accepted 15 July 2020

⁎ Corresponding authors.E-mail addresses: [email protected] (H. Zhou), [email protected] (W. Jin).

Separation and Purification Technology 252 (2020) 117401

Available online 18 July 20201383-5866/ © 2020 Elsevier B.V. All rights reserved.

T

polyamide microporous membrane is also a large class of membranematerials, which can be fabricated into different membranes for gasseparation with high separation performance by the thermal cross-linking method [8,18,19]. For example, high separation performanceswere observed for molecular sieving of nitrogen over cyclohexane bypolyamide membranes [8]. The study on the effect of temperaturewould provide essential data for further optimization of membranestructure to enhance membrane performance.

Here, the properties of the microporous polymer and their corre-sponding membrane were first compared and characterized. Differentthermal crosslinking temperatures were then investigated. Differentcharacterizations such as X-ray diffraction (XRD), fourier transforminfrared spectrometry (FTIR), nuclear magnetic resonance (NMR),scanning electron microscopy (SEM) and so on were performed to re-veal the membrane structural changes. Then polyamide membranesobtained under different crosslinking temperatures were studied for theseparation of a nitrogen/cyclohexane mixture. Finally, the effect ofoperating temperature on membrane performance for the separation ofa nitrogen/cyclohexane mixture was also investigated.

2. Experimental section

2.1. Materials

The monomer 2,6,14-triaminotriptycene was synthesized followingthe method reported by Chen et al. [20]. Dodecanedioyl dichloride waspurchased from Aladdin Industrial Corporation. The pyridine waspurchased from Sinopharm Chemical Reagent Co., Ltd. N,N′-Di-methylformamide (DMF,> 99.5%) and methyl alcohol were purchasedfrom Shanghai Lingfeng Chemical Reagent Co., Ltd. Deionized waterwas made in house. Cyclohexane was purchased from Shanghai ShenboChemical Co., Ltd. All purified gases (at least 99.95%), such as nitrogen,were supplied by Nanjing Special Gas Co. LTD. The PA substratemembrane with a pore size of 0.1 μm was purchased from ShanghaiXinya Purification Equipment Co., Ltd, China.

2.2. Characterizations

Gel permeation chromatography (GPC) was performed to measurethe molecular weight of polymer using a Waters GPC instrument (LC20,Shimadzu). The polymer was dissolved in the NMP for 2 h. The polymersolution was filtered to remove insoluble polymer, and then the solublepolymer solution was measured. Chemical structures were analyzed byusing a Thermo Nicolet Fourier transform infrared spectroscopy (FTIR)8700 spectrometer in the range of 500–4000 cm−1 and were alsocharacterized using liquid state 1H (500MHz) nuclear magnetic re-sonance (NMR) spectroscopy (Bruker 400 AVANCE III HD spectro-meter). The crystal structures were investigated using an X-ray dif-fractometer (XRD, Bruker, D8 Advance) using Cu Kα radiation in therange of 5–80° at increments of 0.02° at room temperature. The mor-phology and structure were observed by field-emission scanning elec-tron microscopy (FE-SEM, Hitachi- 4800, Japan). The elemental com-positions of the polymers and membranes were characterized by X-rayphotoelectron spectroscopy (XPS, Thermo ESCALAB 250, USA) withmonochromatized Al Kα radiation. The crosslinking of polymer at dif-ferent crosslinking temperature was also analyzed by low field nuclearmagnetic resonance (NMR) spectroscopy by Bruker minispec MQ60(Germany). The operating parameters were: the frequency=60 Hz, theproton 90° pulse width=1.96 us, dead time=0.012ms. Positron an-nihilation lifetime spectra were measured for the prepared polymerpowders and membranes using a conventional fast-slow coincidencesystem at room temperature. A positron source was sandwiched be-tween the membranes that had been stacked to a thickness of∼0.8mm.The γ-rays with energies of 1.27MeV (emitted from a β-decay of 22Na)and 0.511MeV (emitted from positron annihilation in a sample) weremeasured by the start and stop counters, respectively. Each spectrum

had ∼6 million counts and a time resolution of ∼210 ps full width athalf-maximum (fwhm). The lifetime spectrum was analyzed through afinite-term lifetime analysis, which was resolved into four discretecomponents using the LT9.0 program. The lifetimes of each componentare denoted as τ1, τ2, and τ3 for p-Ps, free positron and ortho-posi-tronium (o-Ps) while the related intensities are denoted as I1, I2, and I3,respectively. Lifetime and intensity of longer-lived o-Ps are denoted asτ4 and I4, respectively [21]. The following expression was employed torelate the o-Ps lifetime (τ3 and τ4) to the kinetic diameter (R) of free-volume holes in polymer chains:

= ⎡⎣

−+

+ ⎛⎝

⎞⎠ +

⎤⎦

−τ R

R R πsin πR

R R12

1Δ

12

2Δ

1

(1)

where τ is the o-Ps lifetime (ns), R refers to the volume radius (nm), andΔR is the electron layer thickness with an estimated value of 0.1656 nm.

2.3. Synthesis of polyamides

The typical procedure for synthesizing polymers was carried outfollowing a previously reported method [8]. Dodecanedioyl dichloride(named HT-PA6) was selected for detailed investigation because thelonger monomer molecular chain results in better solubility of a syn-thesized polymer in aprotic solvent (DMF) in our work, and its mole-cular weight of Mn is 323657 g/mol, and Mw is 342089 g/mol, whichresults in a good membrane-forming property. Simultaneously, its re-jection is also good when compared with those of other alkyl acylchlorides (comparison data not shown in this paper).

2.4. Fabrication of membranes

The Schematic representation of the membrane fabrication is shownin Scheme 1. The polymers were dissolved in DMF to fabricate mem-brane solutions with a concentration of 0.04 g/mL, which were stirredfor 8 h at 22 °C, then degassed by vacuum. The membrane solutionswere coated on dry 0.1 μm PA substrate membranes using a coatingknife to obtain composite membranes and on PTFE dishes to obtainfreestanding membranes. The nascent membranes were put into a va-cuum oven immediately and dried at various temperatures (50, 80, 100,and 120 °C), and the drying time varies from 96 h, 72 h, 48, and 24 haccording to the temperature. The thickness of the separation mem-brane was observed by SEM.

2.5. Measurements of gas permeation

Gas permeation properties of mixed gas were tested at differenttemperatures and pressures, and the detailed experimental design andprocedures have been reported previously [8]. The effective membranearea is 12.56 cm2. The compositions of the feed, permeate, and residualstreams were analyzed by gas chromatography (SCION456-GC-SQ). Theoriginal flux (Ji, L·m−2·h−1) was determined using a soap bubbleflowmeter, calculated as Eq. (2), which was repeated at least threetimes with different membranes, and the obtained average deviationwas less than 5%. The gas permeability (Pi, Barrer) was calculated asEq. (3). The gas flux was normalized by pressure to give the permeance

Scheme 1. Schematic representation of the membrane fabrication.

Y. Chen, et al. Separation and Purification Technology 252 (2020) 117401

2

(Pi/l, GPU), which was calculated as Eq. (4).

= ×J VA t

36000·ii

(2)

The gas permeability (Pi) was calculated as

= × × × ×P VA t p T

P l10· ·Δ

273.1576i

i10 0

(3)

The gas permeance (Pi/L) was calculated as

= × × ×P l VA t p T

P/ 10· ·Δ

273.1576i

i6 0

(4)

where Vi refers to the volume (cm3), A is the effective membrane area(cm2), t is the time (s), ΔP represents the transmembrane pressure (cmHg), T is the testing temperature (K), l is the thickness of selective layerof the composite membrane (cm) and P0 is defined as the atmosphericpressure (kPa).

The rejection was calculated as

⎜ ⎟= ⎛⎝

− ⎞⎠

×CC

R 1 100%p

f (5)

where Cp and Cf refer to the permeate VOC concentration and feed VOCconcentration, respectively.

PSI (permeation separation index) was calculated by the followingequations:

= − ∗ PPSI (R 1) i1 (6)

= − ∗ P lPSI (R 1) /i2 (7)

2.6. Molecular dynamics simulations

Molecular dynamics simulations were conducted via the MaterialsStudio software, and the COMPASS force field was used to calculate theinteractions between atoms. The van der Waals and electrostatic in-teractions were taken into account using the “Ewald” method [22]. TheBerendsen method with a decay constant= 0.05 ps [23] and the Nose’sthermostat with Q ratio= 0.001 [24] was used to control the pressureand temperature, respectively.

Three kinds of HT-PA6 polymer chains, completely linear structure(a), partial network structure (b), and complete network structure (c),were established, as shown in Fig. 1. In the amorphous cell, three dif-ferent HT-PA chains were selected to construct three HT-PA structuremodels. The number of repeat units, chain numbers, and initial densityof the HT-PA models was set to 18, 5, and 0.01 g cm−3, respectively.Optimization of the three different models’ geometry was first per-formed using the energy minimization process; then, a 21-step methodwas used to compress, relax, and eventually balance the models[25,26]. Mean-square displacement (MSD) was used to analyze themolecular mobility of the membrane, which can be obtained by theEinstein relationship as

∑= ⟨ + − ⟩ = +=N

r t t r t B D tMSD(t) 1 [ ( ) ( )] 6i

N

i i C1

0 02

(8)

where N is the total number of atoms, +r t t( )i 0 and r t( )i 0 are the po-sitions at time +t t0 and t0, respectively, B is a constant, and D is a self-diffusion coefficient. The GCMC method was used to estimate the ad-sorption of gas in the sorption process. A sorption isotherm can beobtained by performing the simulations over a pressure range of0.01–1.5 atm. At each pressure, 1,000,000 steps of GCMC calculationswere performed using an initial equilibration period of 100,000 steps.Diffusion coefficients were determined by the addition of 10 gas mo-lecules into each independently equilibrated configuration to form anew simulation cell at 300 K. The diffusion simulations were performedfor 600 ps under the NVT conditions.

3. Results and discussion

3.1. Determination of structural alterations in the membrane-formingprocess

The HT-PA6 polyamide polymer was synthesized by solution poly-merization, as shown in Scheme 2. At the end of the polymer chains,carboxyl and amino groups can still exist, which provides the oppor-tunity for them to react during the membrane-forming procedure. Thus,at a suitable temperature, water from the reaction between the carboxyland amino groups can be timely removed, inducing a reaction equili-brium in a positive direction. Finally, two side chains can be connectedby this newly formed bond in the membrane, where a relatively bigpore is divided into two smaller pores, as illustrated in Scheme 3. Such aprocess results in narrower three-dimensional cross-linked networks (asindicated by pink ellipses in Scheme 3), offering remarkable separationperformance of nitrogen/VOC. The formation of new bands may pulltwo polymer chains closer together while avoiding other polymerchains, providing larger aggregate pores. To validate this assumption,polymers and their corresponding membranes were characterized tohighlight their differences.

The chemical structures of a polymer and its corresponding mem-brane were first analyzed by 1H NMR and 13C NMR, as shown in Fig. 2.The first noticeable difference (Fig. 2(a)) that can be seen is char-acteristic peaks at positions 12 and 3.5 ppm, which are ascribed to thehydrogen on the carboxyl and amino groups [8,27]. These hydrogensformed by hydrolysis of the unreacted acyl chloride and unreactedamino groups in the polymer during the preparation disappeared whenthe polymer was made into a membrane. Fig. 2(b) shows the 13C NMRspectra of a polymer and its corresponding membrane. It can be seenthat the carboxyl carbon at 162.2 ppm in the membrane also dis-appeared compared with that in the polymer. It indicates that thesecarboxyl and amino groups in the polymer were reacted during themembrane-forming process, resulting in the absence of characteristicpeaks in the 1H NMR and 13C NMR spectra of the membrane. Thesenewly formed bonds will divide big pores into smaller ones in themembrane, as shown in Scheme 3, leading to a lower density, as shownin Fig. 3(a). High permeability can thus be expected because the lowerdensity generally indicates a higher free volume [28–30]. This resultcan be read as indirect proof supporting the occurrence of interchaincrosslinking during the membrane-forming process. Fig. 3(b) showsBET areas of HT-PA6 polymer and its membrane. Higher BET area ofthe polymer can be observed, which could be a result of interpenetra-tion or blocking of the pores due to incomplete solvent removal [31].

Simultaneously, PALs also confirms the evolution of pores, as illu-strated in Table 1. It shows two types of pores in the polymer andmembrane: network and aggregate pores [8]. The size of network poresin the membrane is a little smaller than that of polymers because of theformation of new bonds, which results in a high rejection based on theobservation of a molecular-sieving mechanism, and the increasing sizeand number of aggregated pores from the polymer to membraneFig. 1. Three different polymer chains with different structure.

Y. Chen, et al. Separation and Purification Technology 252 (2020) 117401

3

contribute to enhancing the gas permeability [32].XPS analysis was also applied to elucidate the formation of new

chemical bonds in the membrane after analyzing the oxygen content inthe polymer and membrane, as shown in Fig. 4. Two binding energies ofoxygen elements in these polymers and membranes can be detected:one is at 531.4 eV, which is ascribed to the core level O 1 s spectra ofthe amide bond (NeC]O), and the other is at 532.6 eV, which belongsto the remaining unreacted acyl chloride groups (easily hydrolyzed tocarboxyl groups) (OeC]O). Therefore, the ratio of the integrated areaNeC]O/(NeC]O+OeC]O) (i.e., O1/(O1+O2)) shows an in-creasing trend from polymers to membranes, implying that chemicalcrosslinking between the amino and carboxyl groups occurs in themembrane-forming process, and the size of the pores thus varies ac-cordingly from polymers to membranes.

3.2. Effects of the cross-linking temperature

It is confirmed from the characterizations described above thatcross-linking occurs during the membrane-forming process, which canbe affected by temperature. Thus, the effect of temperature on cross-

linking during the membrane-forming process was elucidated.Fig. 5 shows the FTIR spectra of the HT-PA6 polymer and its

membrane fabricated under different crosslinking temperatures. As isshown, the small peaks at 1662, 1602, and 1550 cm−1 correspond tothe stretching vibration of the C]O band in the amide band I, poly-amide aromatic ring, and CeN stretching of amide band II, respectively[33,34]. The characteristic peak at 1415 cm−1 is the stretching of theCeN band of amide III, suggesting the formation of amide groups inboth polymers and membranes. A comparison of polymer and mem-brane shows that the intensity of peak at 3296 cm−1 ascribed to thesymmetric stretching vibration peak of NeH in the primary aminogroup is reduced from the polymer to the membrane. It further supportsour assumption that amino groups in the polymers are reacted duringthe membrane-forming process, leading to their decreasing intensity inthe membranes. Fig. 5 shows similar FTIR spectra for membranes ob-tained at different temperatures. This may be because the synthesizedpolymers have good membrane-forming properties, as shown in Fig. 6,which shows visible defect-free SEM surfaces for all membranes andthat membrane thickness of 1–2 μm can be obtained from 30 to 120 °C.Thus, similar FTIR spectra are observed.

Although FTIR shows similar spectra, comparative analysis of theXRD spectra of the polymer and membrane illustrates differences, asshown in Fig. 7(a). The broad features of the XRD spectra of both thepolymer and membrane suggest their amorphous nature. The peak at2θ≈ 20° is typical d-spacing for densely packed amorphous polymers,which is usually ascribed to an average chain-chain distance [35].

D-spacing is related to gas permeability [36]. Fig. 7(a) shows thatwhen the polymer is fabricated into the membrane, the peak shifts to ahigh value. The d-spacing decreases from 0.445 nm of the polymer to0.442, 0.437, 0.428, 0.427 and 0.427 nm of the membrane obtained at30, 50, 80, 100 and 120 °C, respectively, indicating that temperaturepromotes the reaction and leads to a lower d-spacing value and smallerpore size. Fig. 7(b) shows the simulated XRD spectra obtained by the

NH2

H2N

H2N

O

O

Cl

Cl

DMFsolution

0 oC, 2 h

NH

HN

HNO

O

10

O O10O O

10NH

NH

NH

n

n

Scheme 2. Synthesis procedure of HT-PA6 polyamide by solution polymerization.

Scheme 3. Evolution of pores in thermal crosslinking process.

Fig. 2. 1H NMR spectra (a) and 13C NMR spectra (b) for the HT-PA6 polymer and its membrane obtained at 80 °C.

Y. Chen, et al. Separation and Purification Technology 252 (2020) 117401

4

simulation of polymers with different structures. It shows that the d-spacing of complete network structure (c) is lower than that of partialnetwork structure (b) and linear structure (a), which expresses the sametrend with experimental data. It further explained that crosslinkingtemperature could promote the reaction during the membrane-formingprocedure and enhance the network structure formation.

The phenomenon of enhancement of network structure with theincrease in temperature was further validated by the measurement of T2relaxation time in LF-NMR, as shown in Fig. 8. It shows from the in-version spectra in Fig. 8(a) that three peaks at around 0.22, 5.5, and53ms are observed, which can be assigned to the inter-crosslink chains,dangling ends, and sol chains, respectively [37]. The inter-crosslinkchains can be regarded as the network structure, and the other two canbe assigned to the aggregate structure. It shows that the highest peak isfor intercrosslink chains, indicating the high degrees of the networkstructure in both of the polymer and membranes. With the increase intemperature, peaks of dangling ends and sol chains gradually dis-appeared, suggesting the enhancement of crosslinking by heat. The leftshift of T2 relaxation time, especially for intercrosslink chains, meansthe decreasing mobility of chains [38,39]. Therefore, the more compactof intercrosslink chains (or the decreasing network pore size) with theincreasing temperature can be observed. It will favor the molecularsieving of nitrogen over VOC.

Fig. 8(b) shows the ratio of intercrosslink vs. total crosslink chains atdifferent temperatures for HT-PA6, a decreasing trend after the first

increase can be observed. This may be the fact that both the networkand the aggregate structure will form during the membrane-formingprocess. The rise of network structure may be higher than that of theaggregate structure at first, leading to the increasing ratio. With thefurther increase of the crosslinking temperature such as over 80 °C,fewer dangling ends and sol chains were left for the formation of in-tercrosslink chains as the reaction proceeded. However, the left shift ofT2 relaxation time for intercrosslink chains means that intercrosslinkchains become more compact, and network pore sizes become smaller,which may pull the aggregate pores bigger, resulting in a decrease ofthe ratio. These also support our assumption shown in Scheme 2 thatthe formation of the smaller network pore size and enlarge aggregate

Fig. 3. True densities (a) and BET areas (b) of HT-PA6 polymer and its membrane.

Table 1Free volume properties of the HT-PA6 polymer and its membrane by PALs.

ID τ3(ns) I3(%) D3(Å) τ4(ns) I4(%) D4(Å)

Polymer 1.63 10.2 4.96 1.91 3.20 5.52Membrane 0.87 6.60 2.86 2.28 10.70 6.18

Fig. 4. X-ray photoelectron O1s spectra for HT-PA6 polymer and its membrane.

Fig. 5. FTIR spectra for polymer and HT-PA6 membranes obtained at differenttemperature.

Y. Chen, et al. Separation and Purification Technology 252 (2020) 117401

5

pore size. Therefore, an optimum temperature should be chosen ac-cording to the experiments. These results also indicate that temperatureis an important factor that can be used to tune the pore structure.Different polymers with different active groups may require differentoptimum crosslinking temperatures.

3.3. Nitrogen/VOCs mixed gas separation performance

Membranes formed under different thermal crosslinking tempera-tures may express different separation performances. Thus, their effectson membrane performance have been investigated. Fig. 9 shows theimpacts of thermal temperature on the separation performance of thenitrogen/cyclohexane mixture. It can be seen that increasing the tem-perature enhances the rejection because of the formation of more newbands. Thus, their d-space shown in Fig. 7 or network pore size shownin Fig. 8(a) will decrease, leading to a higher rejection; a rejectionof> 99% can be obtained when the temperature is above 80 °C, whichis higher than the previous report whose rejection about 30% wasachieved by ZIF-8 membrane for the separation of nitrogen/n-hexanemixture [40]. This phenomenon is verified by diffusion simulation, asshown in Fig. 10. The slope or displacement in the MSD diagram(Fig. 10) represents the molecular diffusivity in the different structuralmodels. It can be seen from Fig. 10(a) that the slope of the MSD dia-gram of cyclohexane is almost zero compared to the high slope of ni-trogen, suggesting reduced diffusivity of the cyclohexane moleculebecause of its larger kinetic diameter than the pore size in the mem-brane. A high rejection can thus be obtained. Simultaneously, Fig. 10(b)shows the steepest slope in the c structure model, indicating that thenitrogen molecule diffuses fastest in the network structure. This resultalso means that newly formed bands in the membrane will help knitmore network pores, leading to the formation of more interconnectivityamong pores and accordingly, higher FAV (fractional accessible vo-lume) can be expected as shown in Fig. 11.

Fig. 11 shows the FAV with different chain structures using differentprobe radii obtained by simulation, which can indirectly representmembrane structures obtained under different temperatures. The linearstructure roughly represents a membrane obtained at low temperature,whereas a complete networked structure represents a membrane ob-tained at high temperature. It can be seen that the polymer with linearstructure has the least FAV, and the polymer with complete networkstructure has the highest FAV, suggesting more pores can be connected,leading to the formation of more interpenetration of pores. Therefore,higher FAV can be observed in the membrane with a complete networkstructure, and even its BET surface area is lower [31]. It further in-dicates that a newly formed band would increase the FAV of thepolymer, which is the same as the results presented in Table 1 that thesum of I3 and I4 in the membrane is higher than that of the polymer.

Fig. 6. SEM surface morpgologies (a, 30 °C; b, 50 °C; c, 80 °C; d, 100 °C; e,120 °C) and cross-sectional morphologies (a′, 30 °C; b′, 50 °C; c′, 80 °C; d′,100 °C; e′, 120 °C) of HT-PA6 membranes under different thermal crosslinkingtemperature.

Fig. 7. XRD spectra for the HT-PA6 polymer and its membranes with different crosslinking temperature (a) and simulated XRD spectra using different polymer chainstructure (b).

Y. Chen, et al. Separation and Purification Technology 252 (2020) 117401

6

Therefore, although the reduction in pore size would help block thetransportation of nitrogen through the membrane, resulting in lowerpermeability [41], the reduction is small from 50 to 120 °C. This isbecause, even though the average pore size in the membrane is lowerthan that of the polymer, the increasing size and number of aggregatepores from polymer to membrane (Table 1) enhances nitrogen perme-ability [32]. Simultaneously, the formation of new bands will increasethe degree of network crosslinking and the interconnectivity of pores inthe membrane, enhancing the FAV and gas permeability [42,43]. Fi-nally, three factors compensate for each other, leading to a gentle re-duction in the nitrogen permeability. Nitrogen permeance is also ex-pressed in Fig. 10 and shows a similar trend to that of permeability. At30 °C, high nitrogen permeability and a low rejection might result from

the existence of invisible pinholes in the membrane due to the cross-linking temperature being too small.

Fig. 9 also shows that the permeability of nitrogen decreases whilethe rejection increases with increasing temperature; therefore, it is hardto determine the optimal temperature. To identify this, a modified PSIconcept was selected, and the results are shown in Fig. 12. Ignoring30 °C because of the defects in the membrane, PSI is shown to decreasewith increasing temperature. According to these results, 50 °C may bethe best temperature; however, its rejection is lower than 99%. Thus,80 °C is a better temperature, combining a high rejection with relativelyhigh permeability. This result also suggests that under the premise of

Fig. 8. T2 inversion spectra (a) and the ratio of intercrosslink vs total crosslink chains at different temperature for HT-PA6.

Fig. 9. Effect of thermal crosslinking temperature on the separation perfor-mance of HT-PA6 membrane (cyclohexane concentration: 34000 ± 1000 ppm;temperature: 25 °C; Feed pressure: 10 kPa).

Fig. 10. Simulation of mobility of nitrogen (a) and nitrogen and cyclohexane (b) in different structure models.

Fig. 11. Simulated fractional accessible volume of polymers with differentstructural models as a function of probe radius.

Y. Chen, et al. Separation and Purification Technology 252 (2020) 117401

7

guaranteeing a high rejection (or selectivity), a much higher cross-linking temperature may ruin the permeability; thus, an optimal tem-perature exists.

At the optimal crosslinking temperature, the obtained membranewas tested for the separation of the nitrogen/cyclohexane mixtureunder different operating temperatures. These results are shown inFig. 13, and a constant rejection and permeability with the increasedoperating temperature can be observed. This may be because the rigid3D structure of triptycene in the polymer chain might interlock withitself, preventing the movement of polymer chains as the operatingtemperature increases [44]. Furthermore, the network structure formedmight also restrict chain movement. Therefore, the membrane perfor-mance remains constant across increasing temperatures.

4. Conclusions

In this work, the crosslinking temperature was confirmed to affectmembrane structure, and apparent physicochemical differences be-tween polymers and membranes were observed, such as the density ofthe membrane is lower than that of the polymer. Then, differentcrosslinking temperatures were investigated to show that temperatureenhances crosslinking, leading to decreasing d-space and network poresize and increasing aggregate pore size. A rejection of> 99% can be

obtained when the temperature is above 80 °C, which is also supportedby molecular diffusivity simulations. Simultaneously, a gentle reductionin nitrogen permeability can be ascribed to the formation of morenetwork pores during the thermal crosslinking procedure, resulting inmore interconnectivity among the pores and an increasing number ofaggregated pores and FAV, which compensate for the decrease in per-meability resulting from the decreasing pore sizes. Because of the op-posite trends of rejection and permeability with crosslinking tempera-ture, it is suggested that an optimal crosslinking temperature exists inthe membrane-forming procedure.

In summary, crosslinking temperature is an important factor thatshould be considered in the membrane-forming procedure. It mayprovide a significant foundation for more accurate tuning of membranestructures with enhanced separation performance.

CRediT authorship contribution statement

Yuan Chen: Methodology, Investigation, Writing - original draft,Writing - review & editing. Jinchao Qin: Methodology, Writing - ori-ginal draft, Investigation. Tong Tong: Methodology, Investigation.Haoli Zhou: Conceptualization, Supervision, Writing - review &editing. Xingzhong Cao: Investigation. Wanqin Jin:Conceptualization, Supervision, Writing - review & editing.

Declaration of Competing Interest

The authors declare that they have no known competing financialinterests or personal relationships that could have appeared to influ-ence the work reported in this paper.

Acknowledgments

This work was supported by the National Key Research andDevelopment Program of China (No. 2017YFC0210901), Six TalentPeaks Project in Jiangsu Province (JNHB-041), the Qing Lan Project,the Fund of State Key Laboratory of Materials-Oriented ChemicalEngineering (ZK201715), the Top-notch Academic Programs Project ofJiangsu Higher Education Institutions (TAPP), and the InnovativeResearch Team Program by the Ministry of Education of China (No.IRT13070). We are also grateful to the High Performance ComputingCenter of Nanjing Tech University for supporting the computationalresources.

References

[1] C. Della Pina, M.A. De Gregorio, L. Clerici, P. Dellavedova, E. Falletta, Polyaniline(PANI): an innovative support for sampling and removal of VOCs in air matrices, J.Hazard. Mater. 344 (2018) 308–315.

[2] W. Wei, Z. Lv, G. Yang, S. Cheng, Y. Li, L. Wang, VOCs emission rate estimate forcomplicated industrial area source using an inverse-dispersion calculation method:A case study on a petroleum refinery in Northern China, Environ. Pollut. 218 (2016)681–688.

[3] Z. Petrusová, K. Machanová, P. Stanovský, P. Izák, Separation of organic com-pounds from gaseous mixtures by vapor permeation, Sep. Purif. Technol. 217(2019) 95–107.

[4] C.K. Yeom, S.H. Lee, H.Y. Song, J.M. Lee, Vapor permeations of a series of VOCs/N2mixtures through PDMS membrane, J. Membr. Sci. 198 (2002) 129–143.

[5] J. Zhang, J.A. Schott, S.M. Mahurin, S. Dai, Porous structure design of polymericmembranes for gas separation, Small Methods 1 (2017) 1600051.

[6] H. Sanaeepur, A. Ebadi Amooghin, S. Bandehali, A. Moghadassi, T. Matsuura,B. Van der Bruggen, Polyimides in membrane gas separation: Monomer’s moleculardesign and structural engineering, Prog. Polym. Sci. 91 (2019) 80–125.

[7] Z.-X. Low, P.M. Budd, N.B. McKeown, D.A. Patterson, Gas permeation properties,physical aging, and its mitigation in high free volume glassy polymers, Chem. Rev.118 (2018) 5871–5911.

[8] H. Zhou, F. Tao, Q. Liu, C. Zong, W. Yang, X. Cao, W. Jin, N. Xu, Microporouspolyamide membranes for molecular sieving of nitrogen from volatile organiccompounds, Angew. Chem. Int. Ed. 56 (2017) 5755–5759.

[9] C.H. Lau, P. Li, F. Li, T.-S. Chung, D.R. Paul, Reverse-selective polymeric mem-branes for gas separations, Prog. Polym. Sci. 38 (2013) 740–766.

[10] B. Zuo, M. Inutsuka, D. Kawaguchi, X. Wang, K. Tanaka, Conformational relaxation

Fig. 12. Relationship between cross-linking temperature and PSI for the se-paration of nitrogen/cyclohexane mixture.

Fig. 13. Effect of operating temperature on the separation performance of HT-PA6(80) membranes (cyclohexane concentration: 34000 ± 1000 ppm; Feedpressure:10 kPa).

Y. Chen, et al. Separation and Purification Technology 252 (2020) 117401

8

of poly(styrene-co-butadiene) chains at substrate interface in spin-coated and sol-vent-cast films, Macromolecules 51 (2018) 2180–2186.

[11] K. Berean, J.Z. Ou, M. Nour, K. Latham, C. McSweeney, D. Paull, A. Halim,S. Kentish, C.M. Doherty, A.J. Hill, K. Kalantar-zadeh, The effect of crosslinkingtemperature on the permeability of PDMS membranes: Evidence of extraordinaryCO2 and CH4 gas permeation, Sep. Purif. Technol. 122 (2014) 96–104.

[12] W. Yang, H. Zhou, C. Zong, Y. Li, W. Jin, Study on membrane performance in vaporpermeation of VOC/N2 mixtures via modified constant volume/variable pressuremethod, Sep. Purif. Technol. 200 (2018) 273–283.

[13] Y. Li, J. Shen, K. Guan, G. Liu, H. Zhou, W. Jin, PEBA/ceramic hollow fiber com-posite membrane for high-efficiency recovery of bio-butanol via pervaporation, J.Membr. Sci. 510 (2016) 338–347.

[14] C. Zhang, B. Cao, P. Li, Thermal oxidative crosslinking of phenolphthalein-basedcardo polyimides with enhanced gas permeability and selectivity, J. Membr. Sci.546 (2018) 90–99.

[15] P.M. Budd, E.S. Elabas, B.S. Ghanem, S. Makhseed, N.B. McKeown, K.J. Msayib,C.E. Tattershall, D. Wang, Solution-processed, organophilic membrane derived froma polymer of intrinsic microporosity, Adv. Mater. 16 (2004) 456–459.

[16] O. Ilinitch, V. Fenelonov, A. Lapkin, L. Okkel, V. Terskikh, K. Zamaraev, Intrinsicmicroporosity and gas transport in polyphenylene oxide polymers, Micropor.Mesopor. Mat. 31 (1999) 97–110.

[17] L.M. Robeson, W.F. Burgoyne, M. Langsam, A.C. Savoca, C.F. Tien, High perfor-mance polymers for membrane separation, Polymer 35 (1994) 4970–4978.

[18] S. Bing, J. Wang, H. Xu, Y. Zhao, Y. Zhou, L. Zhang, C. Gao, L.A. Hou, Polyamidethin-film composite membrane modified with persulfate for improvement of perm-selectivity and chlorine-resistance, J. Membr. Sci. 555 (2018) 318–326.

[19] T. Zhu, X. Yang, X. He, Y. Zheng, J. Luo, Aromatic polyamides and copolyamidescontaining fluorene group: Synthesis, thermal stability, and gas transport proper-ties, 30 (2018) 821–83.

[20] C. Zhang, C.-F. Chen, Synthesis and structure of 2,6,14- and 2,7,14-trisubstitutedtriptycene derivatives, J. Org. Chem. 71 (2006) 6626–6629.

[21] W.F. Yong, T.-S. Chung, Mechanically strong and flexible hydrolyzed polymers ofintrinsic microporosity (PIM-1) membranes, J. Polym. Sci. Pol. Phys. 55 (2017)344–354.

[22] P.V.K. Pant, R.H. Boyd, Simulation of diffusion of small-molecule penetrants inpolymers, Macromolecules 25 (1992) 494–495.

[23] H.J.C. Berendsen, J.P.M. Postma, W.F.V. Gunsteren, A. DiNola, J.R. Haak,Molecular dynamics with coupling to an external bath, J. Chem. Phys. 81 (1984)3684–3690.

[24] U.H. Hansmann, Y. Okamoto, F. Eisenmenger, Molecular dynamics, langevin andhydrid monte carlo simulations in a multicanonical ensemble, Chem. Phys. Lett.259 (1996) 321–330.

[25] A. Gonciaruk, K. Althumayri, W.J. Harrison, P.M. Budd, F.R. Siperstein, PIM-1/graphene composite: A combined experimental and molecular simulation study,Micropor. Mesopor. Mat. 209 (2015) 126–134.

[26] A. Del Regno, A. Gonciaruk, L. Leay, M. Carta, M. Croad, R. Malpass-Evans,N.B. McKeown, F.R. Siperstein, Polymers of intrinsic microporosity containingTröger base for CO2 capture, Ind. Eng. Chem. Res. 52 (2013) 16939–16950.

[27] J.W. Jeon, D.-G. Kim, E.-H. Sohn, Y. Yoo, Y.S. Kim, B.G. Kim, J.-C. Lee, Highlycarboxylate-functionalized polymers of intrinsic microporosity for CO2-selectivepolymer membranes, Macromolecules 50 (2017) 8019–8027.

[28] D.F. Sanders, Z.P. Smith, R. Guo, L.M. Robeson, J.E. McGrath, D.R. Paul,

B.D. Freeman, Energy-efficient polymeric gas separation membranes for a sustain-able future: a review, Polymer 54 (2013) 4729–4761.

[29] G. Dlubek, D. Bamford, A. Rodriguez-Gonzalez, S. Bornemann, J. Stejny, B. Schade,M.A. Alam, M. Arnold, Free volume, glass transition, and degree of branching inmetallocene-based propylene/α-olefin copolymers: Positron lifetime, density, anddifferential scanning calorimetric studies, J. Polym. Sci. Pol. Phys. 40 (2002)434–453.

[30] Y.L. Xue, J. Huang, C.H. Lau, B. Cao, P. Li, Tailoring the molecular structure ofcrosslinked polymers for pervaporation desalination, Nat. Commun. 11 (2020)1461.

[31] K.S. Walton, R.Q. Snurr, Applicability of the BET method for determining surfaceareas of microporous metal–organic frameworks, J. Am. Chem. Soc. 129 (2007)8552–8556.

[32] S.H. Kim, S.-Y. Kwak, T. Suzuki, Positron annihilation spectroscopic evidence todemonstrate the flux-enhancement mechanism in morphology-controlled thin-film-composite (TFC) membrane, Environ. Sci. Technol. 39 (2005) 1764–1770.

[33] M. Namvar-Mahboub, M. Pakizeh, S. Davari, Preparation and characterization ofUZM-5/polyamide thin film nanocomposite membrane for dewaxing solvent re-covery, J. Membr. Sci. 459 (2014) 22–32.

[34] T. Wang, L. Dai, Q. Zhang, A. Li, S. Zhang, Effects of acyl chloride monomerfunctionality on the properties of polyamide reverse osmosis (RO) membrane, J.Membr. Sci. 440 (2013) 48–57.

[35] N. Ritter, M. Antonietti, A. Thomas, I. Senkovska, S. Kaskel, J. Weber,Binaphthalene-based, soluble polyimides: The limits of intrinsic microporosity,Macromolecules 42 (2009) 8017–8020.

[36] S. Luo, J.R. Wiegand, P. Gao, C.M. Doherty, A.J. Hill, R. Guo, Molecular origins offast and selective gas transport in pentiptycene-containing polyimide membranesand their physical aging behavior, J. Membr. Sci. 518 (2016) 100–109.

[37] U. Heuert, M. Knörgen, H. Menge, G. Scheler, H. Schneider, New aspects oftransversal1H-NMR relaxation in natural rubber vulcanizates, Polym. Bull. 37(1996) 489–496.

[38] J. Wu, W. Lin, Z. Wang, S. Chen, Y. Chang, Investigation of the hydration of non-fouling material poly(sulfobetaine methacrylate) by low-field nuclear magneticresonance, Langmuir 28 (2012) 7436–7441.

[39] Y. Liu, Z. Cai, L. Sheng, M. Ma, Q. Xu, Y. Jin, Structure-property of crosslinkedchitosan/silica composite films modified by genipin and glutaraldehyde under al-kaline conditions, Carbohyd. Polym. 215 (2019) 348–357.

[40] J. Li, J. Zhong, W. Huang, R. Xu, Q. Zhang, H. Shao, X. Gu, Study on the devel-opment of ZIF-8 membranes for gasoline vapor recovery, Ind. Eng. Chem. Res. 53(2014) 3662–3668.

[41] J. Weber, N. Du, M.D. Guiver, Influence of intermolecular interactions on the ob-servable porosity in intrinsically microporous polymers, Macromolecules 44 (2011)1763–1767.

[42] M.F. Jimenez-Solomon, Q. Song, K.E. Jelfs, M. Munoz-Ibanez, A.G. Livingston,Polymer nanofilms with enhanced microporosity by interfacial polymerization,Nature Mater. 15 (2016) 760–767.

[43] D.L. Gin, R.D. Noble, Designing the next generation of chemical separation mem-branes, Science 332 (2011) 674–676.

[44] Y.J. Cho, H.B. Park, Gas separation properties of triptycene-based polyimidemembranes, Modern Applications in Membrane Science and Technology, AmericanChemical Society, 2011, pp. 107–128.

Y. Chen, et al. Separation and Purification Technology 252 (2020) 117401

9