Jie Li,Ying Labreche,Naixin Wang,Shulan Ji,Quanfu An

1 Beijing Key Laboratory for Green Catalysis and Separation,College of Environmental and Energy Engineering,Beijing University of Technology,Beijing 100124,China

2 School of Chemical and Biomolecular Engineering,Georgia Institute of Technology,Atlanta,GA 30332,United States

Keywords:Pervaporation Hollow fiber PDMS/ZIF-8 Torlon?PVDF Ultem?Matrimid?

ABSTRACT In order to develop high performance composite membranes for alcohol permselective pervaporation(PV),poly(dimethylsiloxane)/ZIF-8(PDMS/ZIF-8)coated polymeric hollow fiber membranes were studied in this research.First,PDMS was used for the active layer,and Torlon?,PVDF,Ultem?,and Matrimid?with different porosity were used as support layer for fabrication of hollow fiber composite membranes.The performance of the membranes varied with different hollow fiber substrates was investigated.Pure gas permeance of the hollow fiber was tested to investigate the pore size of all fibers.The effect of support layer on the mass transfer in hydrophobic PV composite membrane was investigated.The results show that proper porosity and pore diameter of the support are demanded to minimize the Knudsen effect.Based on the result,ZIF-8 was introduced to prepare more selective separation layer,in order to improve the PV performance.The PDMS/ZIF-8/Torlon?membrane had a separation factor of 8.9 and a total flux of 847 g·m-2·h-1.This hollow fiber PDMS/ZIF-8/Torlon?composite membrane has a great potential in the industrial application.

1.Introduction

In recent years,the production and application of renewable fuels have drawn more and more attention due to the global energy crisis[1,2].Ethanol,as a typical biofuel,has been widely recognized as a green alternative to fossil fuels.For separation of ethanol from dilute biomass fermentation,pervaporation(PV)has been widely studied as a good potential process for recovering bioethanol[3].For PV process,polymeric membranes have played an important role.Membranes for PV are typically fabricated from hydrophobic material,and the most common of which is poly(dimethylsiloxane)(PDMS).However,polymeric membranes are limited by an“upper bound”in performance indicating the trade-off between membrane permeability and selectivity.The incorporation of inorganic particles to form mixed matrix membranes(MMMs)could be an effective way to enhance PV performance.Several types of inorganic materials were used in making MMMs,such as nano-silica[4],zeolites[5]carbon black[6],and MOFs[7].Particularly,zeolite imidazolate frameworks(ZIFs)can be used to form MMMs with high PV performance due to its hydrophobic inner channels[8-10].

With regard to a PV membrane module,the plate-and-frame arrangement is still the dominating module configuration employed in pervaporation.Nevertheless,the development of economical membrane modules for PV is still a major issue.Currently,the dominating arrangement employed in PV is the plate-and-frame module.Compared with the flat membranes,hollow fibers have 20 times greater surface area per unit volume[11].Moreover,hollow fiber has some economical superiorities and advantages such as high-packing density,a selfcontained vacuum channel and a self-contained mechanical support.Lind et al.[12]prepared Free-standing ZIF-71/PDMS nanocomposite membranes for the recovery of ethanol and 1-butanol from water through pervaporation.With the loading of 40 wt%ZIF-71,the selectivity for 1-butanol/water was about 5.7.However,it is challenging and expensive to directly fabricate an asymmetric hollow fiber membrane with a thin defect-free selective layer for high performance,meanwhile mechanical stability can overcome the solvent-induced swelling[13].Therefore,preparation of hollow fiber composite membranes,consisting of a selective layer and a porous support layer,can be an effective alternative for pervaporation separation processes.

In order to optimize the structure of PV membranes to get high performance,a good understanding of mass transfer mechanisms though hollow fiber composite membranes is essential.The mass transfer across the polymeric active layer is usually described by the solution-diffusion model,which assumes instant desorption at the permeate side of the membrane,where porous layer effects are usually considered to be negligible.Although several studies have clearly demonstrated that the support layer can also influence the mass transfer through the composite membrane and even dominates and reverses the performance of the selective layer[14,15],the mass transfer mechanism through substrate of hydrophobic membranes for pervaporation has been seldom reported.

The focus of this study is to fabricate high performance polymeric hollow fiber membranes for alcohol permselective PV.The influence of the support layer on the mass transfer in hydrophobic PV was investigated.Four kinds of materials with Torlon?,PVDF,Ultem?and Matrimid?were used as the hollow fiber substrates.PDMS was selected as the selective layer since it has been proven to be an excellent selective material for ethanol/water separation.In order to decrease the mass transfer resistance across the porous layer,porosity and pore diameter of the support were studied.Based on the study of mass transfer effect,further improvement in membrane performance was obtained by introducing ZIF-8 to prepare MMMs to improve the PV performance,both the separation factor and the flux increased.

2.Materials and Methods

2.1.Materials

Asymmetric hollow fiber substrate membranes were produced from Torlon?,PVDF,Ultem?,and Matrimid?in our lab[16-20].Different supports with various functional groups or wetting property will have different interactions with PDMS.The physical properties of different membrane materials were shown in Table 1.PDMS of RTV615A(dimethylsiloxane)and RTV615B(a curing agent)were purchased from Momentive Performance Materials,Inc.ZIF-8 crystals(Basolite Z1200,BASF)were purchased from Sigma-Aldrich.Methanol,nheptane and ethanol were obtained from Aldrich Chemical Company,USA.All solvents were used as received.

Table 1 Physical properties of different membrane materials

2.2.Preparation of PDMS and PDMS/ZIF-8 on hollow fiber substrates

Prior to dip-coating,in order to prevent penetration of the coating polymer solution into the pores,the hollow fiber substrate membranes were immersed in deionized water to fill the pores of the supports with water;and dried for 30 min to remove excess water on the support surface[21].RTV615A was dissolved in n-heptane to form a 10 wt%solution and stirred for 1 h at room temperature.Then the RTV615B was added into the polymer mixture(WRTV615A:WRTV615B=10:1),and the resulting solution was stirred for 1 h.The polymer solution was poured into a tube and coated on the surface of the hollow fiber substrate with a dip-coating method(Fig.1a).To produce defect-free membranes,the composite membrane was put at room temperature for 12 h for solvent removal and subsequently in the oven at 80°C for 12 h for crosslinking.The composite membranes were then loaded in modules(Fig.1b)and were tested for ethanol and water separation.

For the preparation of PDMS/ZIF-8 membrane on hollow fiber substrates,ZIF-8 particles were dispersed in n-heptane and ultrasonicated for 1 h.The RTV615A was dissolved in n-heptane to form a 10 wt%solution and stirred for another 1 h at room temperature(WRTV615A:ZIF-8=10:1).Then the RTV615B was added into the polymer mixture(WRTV615A:WRTV615B=10:1),and the resulting solution was stirred for 1 h.Similarly as preparation of the PDMS membranes,the polymer/ZIF-8 solution was poured into a tube and dip-coated on the surface of the hollow fiber substrate and then put into room temperature and an oven for crosslinking.

2.3.Characterization

Pure gas permeation measurements of substrates were tested with pure N2at 35°C and an upstream pressure of 29.4 psi(～0.2 MPa)[22].SEM photographs of membranes were taken using a scanning electron microscope(Hitachi SEM SU8010,Japan).In order to characterize the cross section of membranes,the samples were fractured into liquid nitrogen.All membranes were sputtered with gold in vacuum to make it conductive before observations.The mean pore size of the hollow fiber substrates were measured using a 3H-2000PBL porometer(BeiShiDe Instrument,China).

2.4.Pervaporation experiments

The pervaporation experiments were carried out using a laboratory made system in Fig.2(a).The detailed equipment setup was described elsewhere[23,24].The experiments were conducted at 60°C with the feed solution of～5 wt%ethanol,and the downstream pressure maintained at 200-300 Pa.To minimize the pressure losses,Helium gas flow was used instead of pulling vacuum as shown in Fig.2(b).Gas flow rate in the bore-side of the fiber was measured with a bubble flow meter.For each set of pervaporation conditions,three modules with the same fabrication were tested.

The effective length of fibers was～37 cm.During the pervaporation process,the feed was recirculated using a plunger pump from the shell side of the module to the feed tank.The permeate vapor was collected on the pore side of the fiber and trapped in liquid nitrogen.The permeate was subsequently analyzed with a refractometer(Leica ARIAS 50),using a pre-calibrated curve of the refractive index(RI)vs.ethanol concentration.The permeate flux(J)was determined according to the following equation:

where W is the mass of the liquid collected in the cold traps,A is the effective area of the membrane,and t is the collection time for the PV.

Fig.2.Experimental apparatus for pervaporation evaluation with different methods:(a)pull vacuum and(b)Helium sweeping.

The selectivity of the composite membranes is expressed as a separation factor α,which is defined as

where YEtOH,YH2O,XEtOHand XH2Oare the mass ratios of ethanol and water in the permeation and feed sides,respectively.

3.Results

3.1.Pore size and pure gas permeance of the hollow fiber

Since the outer surface of the hollow fiber is directly in contact with the PDMS selective layer,its porous structure affects the coating of the PDMS solution.To investigate the pore size of all fibers,the pore size tests and pure gas permeance experiments are performed.The results of N2permeation test could reflect the pore size of the substrates.As shown in Table 2,the pore sizes of the substrates were in an order ofMatrimid?＜Ultem?＜Torlon?＜PVDF?,while the gas permeances for all four substrates follow the same order.It can be concluded that the PVDF has a larger pore size.After coating with 10%PDMS solution,the gas permeance for four composite membranes was in an order of PDMS/Matrimid?＜PDMS/Ultem?＜PDMS/PVDF＜PDMS/Torlon?.After coating PDMS,the permeance of PVDF decreased sharply from 4349 GPU to 10.8 GPU(1GPU=10-6cm3(STP)·(cm2·s·cmHg)-1),mostly as the PDMS could infiltrate into the pores of the PVDF substrate.Besides,in order to investigate the effects of pore size on the gas permeance,Torlon?(named Torlon?-1 and Torlon?-2)with different pore sizes was spun by different drawing processes.The pore sizes and gas permeances for Torlon?were in an order of Torlon?-2＜Torlon?-1 and for composite membrane the gas permeances were in an order of PDMS/Torlon?-2＜PDMS/Torlon?-1,which exactly follows in the same order of the permeances of the substrates.

Table 2 Pore size and pure gas permeance of different membranes

Fig.3.SEM imaging of the different PDMS membranes.(a)PDMS/PVDF composite membrane,(b)PDMS/Torlon?composite membrane,(c)PDMS/Matrimid?composite membrane and(d)PDMS/Ultem?composite membrane.

3.2.Morphologies of hollow fiber composite membranes

In order to compare the composite membranes with different support structures,the cross section photographs of the membranes were taken by SEM.As shown in Fig.3,for the PDMS/Torlon?,PDMS/Matrimid?and PDMS/Ultem? composite membranes,asymmetric hollow fiber composite membranes consist of a thin dense selective layer supported on a porous substrate(Fig.3(b),(c)and(d)).The outer surface was dense and uniform,and a clear PDMS selective layer could be seen.That means the PDMS coated on the top of the substrates.But for the PDMS/PVDF membrane,a nanoporous surface was still clearly observed because of the large pore structure of PVDF(Fig.3(a)).Meanwhile,the wide pores may lead to some of the PDMS not only coating on the surface but also intruding into the pores of the PVDF substrate.

3.3.Pervaporation performance of the hollow fiber composite membranes

The PV performance of hollow fiber composite membranes for ethanol-water mixture is shown in Table 3.As shown in Table 3,the flux of the hollow fiber composite membranes was in an order of PDMS/Torlon?-1＞PDMS/PVDF＞PDMS/Matrimid?＞PDMS/Ultem?.The optimal PV performance of all these for kinds of membranes was PDMS/Torlon?-1,with a separation factor of 5.1 and a flux of 3258 g·m-2·h-1.

Moreover,in order to further investigate the effect of the pore size for the pervaporation performance,Torlon-2,which has a smaller pore and a lower porosity than Torlon-1 fiber was also investigated.In comparison with PDMS/Torlon?-1,PDMS/Torlon?-2 membranes exhibited lower permeability,the N2permeance for PDMS/Torlon?-2 is 16.7 GPU,while for PDMS/Torlon-1 the N2permeance is 90.7 GPU(Table 2).Moreover,as shown in Table 3,Torlon?-1 exhibited both higher selectivity and higher flux.For example,the PDMS/Torlon?-1 membrane had a flux of 3258 g·m-2·h-1and a separation factor of 5.1.In contrast,these two values were 2541 g·m-2·h-1and 3.2,respectively,when the PDMS/Torlon?-2 membrane was used.

The fibers showed a very low separation factor of 5.1,quite low compared to the values reported for other PDMS composite membranes[25].The reasons for these results could be as follows.Generally,composite pervaporation membranes are formed by a thin,selective,dense top layer and a porous asymmetric support layer.The mass transfer across the selective layer is controlled by the solution-diffusion model,where the effects of porous layer are usually considered to be negligible.This means the porosity and pore size of the support should be as high as possible in order to avoid additional mass transfer resistance.However,this can be challenging from a manufacturing point of view to make defects free membranes.Too large pores result in defects in selective layer and poor separation performance when PDMS casting soluton will intrusd inside the pores of the support(Fig.4(a)).For membranes in Fig.4(b),the mass transfer across the selective layer is controlled by the solution-diffusion model.As when the pores of the substrate membrane were too narrow,mass transfer resistance across the porous layer cannot be negligible.While the support layer(for example contact angle for Torlon?is 87.3°,in Table 1)is more addictive to water than PDMS.Then the mass transfer across the support layer could be controlled by the solution-diffusion model.Therefore,the relatively low separation performance of hollow fiber composite membranes could be due to the resistance of the support layer(Fig.4(b)).And when a support layer is an asymmetric ultrafiltration membrane,the pore diameters varies through the support structure.The pores close to the active layer have the smallest diameters,while those deeper inside the support structure have much larger diameters.If the pores of the support layer are not large enough,the Knudsen diffusion will be exhibited(Fig.4(c)).As the coefficient of the Knudsen diffusion is larger than that of solution-diffusion,the flux of the Torlon?-1 is better than that of Ultem?,Matrimid?and Torlon?-2.Thus,the solution-diffusion coupled with Knudsen-diffusion is beneficial to the separation performance.

Fig.4.Schematic representation of the composite membrane structure and mass transport model.

Table 4 Effects of methods for pervaporation

Table 5 Effects of diameter of hollow fiber

In order to minimize the pressure drop and resistance of the support layer,different methods for pervaporation were tested.As shown in Fig.2(a),for the pervaporation test using a hollow fiber module with pulling vacuum method,one end of the module is sealed and the other is connected to the vacuum side.The feed is cycling on the shell side and the permeate is on the bore side.So there would be a pressure drop on the permeate side[15].In order to minimize the pressure drop,Helium flow was used instead of pulling vacuum.The permeate flux flow conditions inside a hollow fiber membrane were shown in Fig.2(b).The effects of methods for pervaporation were shown in Table 4.The composite membranes showed improved pervaporation performance with the Helium flow.For example,the PDMS/Torlon?membrane tested with Helium sweeping technique had a separation factor of 8.0 and a permeate flux of 499 g·m-2·h-1.In contrast,these two values were 5.1 and 3258 g·m-2·h-1,respectively,when pulling vacuum method was used.

Fig.5.Effects of gas flow rate on pervaporation performance.(Pervaporation conditions:feed temperature 60°C,permeate pressure 100 Pa,EtOH content in feed solution 5 wt%.)

Table 6 Comparison of the pervaporation performance in different membranes

For the hollow fiber module,in order to minimize operational costs,the fiber characteristics need to closely match system demands.In this regard,the specific surface area of the fiber cross-section should be maximized,i.e.,the fiber should have both a small OD and a large ID.The effects of diameter of hollow fiber on the pervaporation performance of composite membranes were evaluated in this study.As shown in Table 5,the membrane with an OD of 423 and an ID of 199 substrate showed better performance.

In this study,Helium acts as the sweeping gas.The effects of flow rate on the PV performance were carried out.As shown in Fig.5,the separation factor decreased with the flux increased and then decreased while the gas flow rate increased from 90 ml·min-1to 180 ml·min-1.So low flow rate was chosen to keep a better separation factor.

Preparing a mixed-matrix membrane could be an effective way to improve the separation performance[8].Therefore,ZIF-8 was introduced to the PDMS polymer based on the investigation of PDMS/Torlon?membranes.As shown in Table 6,the hollow fiber PDMS-ZIF-8/Torlon?membrane shows higher ethanol-water separation performance compared with PDMS/Torlon?membrane.The membrane had a separation factor of 8.9 and a total flux of 847 g·m-2·h-1,while these two values were 5.1 and 3258 g·m-2·h-1for PDMS/Torlon?membrane.This is the first report on the preparation of ZIF-8 nanohybrid hollow fiber membranes for the ethanol permselective pervaporation.

Table 7 compares the PV performance of the ethanol-water system against the separation factors and fluxes of various membrane materials reported in the literature[9,25-32].Compared with the state-of-the-art polymeric and hybrid membranes listed in Table 4,the hollow fiber PDMS-ZIF-8/Torlon?membrane shows high ethanol-water separation factor and a sustainable flux.The membrane had a separation factor of 8.9 and a total flux of 847 g·m-2·h-1at a feed temperature of 60°C,which proved the technical feasibility of the nanohybrid hollow fiber membrane for pervaporation.

4.Conclusions

In this study,PDMS and PDMS-ZIF-8 membranes were successfully prepared on four different kinds of hollow fiber substrates.The membranes were applied for the pervaporation of ethanol purification.We noted that the porous hollow fiber support layer had high impact onmass transfer in composite pervaporation membranes.The Knudsen effect was discussed with different membrane structures.The membrane manufacturing demands are for proper porosity and pore diameter of the support to minimize the Knudsen effect of the substrate.To minimize the pressure losses,Helium gas flow was used to improve the PV performance.The separation performance was further optimized by studying the effects of different pervaporation methods,gas sweeping flow rate and substrate diameter.The hollow fiber PDMS-ZIF-8/Torlon?membrane shows high ethanol-water separation factor and a sustainable flux.In particular,the PDMS-ZIF-8/Torlon? membrane had a separation factor of 8.9 and a total flux of 847 g·m-2·h-1at a feed temperature of 60 °C.Considering the benefits of hollow fiber,the PDMS-ZIF-8/Torlon? composite membrane may be a promising candidate for pervaporation.

Table 7 Comparison of the pervaporation performance in different membranes

Acknowledgments

The authors thank William J.Koros for guidance in analyses of diffusion,Ryan P.Lively for guidance in analyses of fibers and thank Wulin Qiu for the pervaporation test of hollow fiber membranes.