Separation and Purification Technology 146 (2015) 24–32

Contents lists available at ScienceDirect

Separation and Purification Technology

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

Hollow fiber modules with ceramic-supported PDMS compositemembranes for pervaporation recovery of bio-butanol

http://dx.doi.org/10.1016/j.seppur.2015.03.0291383-5866/� 2015 Elsevier B.V. All rights reserved.

⇑ Corresponding author. Tel.: +86 25 83172266; fax: +86 25 83172292.E-mail address: wqjin@njut.edu.cn (W. Jin).

Danyu Liu, Gongping Liu, Lie Meng, Ziye Dong, Kang Huang, Wanqin Jin ⇑State Key Laboratory of Materials-Oriented Chemical Engineering, College of Chemistry and Chemical Engineering, Nanjing Tech University (former Nanjing University ofTechnology), 5 Xinmofan Road, Nanjing 210009, PR China

a r t i c l e i n f o a b s t r a c t

Article history:Received 1 December 2014Received in revised form 11 March 2015Accepted 12 March 2015Available online 24 March 2015

Keywords:Hollow fiber modulePervaporationCFD simulationCeramic-supported PDMSButanol

The practical application of hollow fiber membranes for pervaporation technology has been receivedgrowing attention in recent years. This work reports the development of hollow fiber modules of cera-mic-supported polydimethylsiloxane (PDMS) composite membranes applied for pervaporation process.Computational fluid dynamics (CFD) technique was used to simulate and optimize the flow field dis-tribution in the modules with different packing density and cross-section layout. The hollow fiber mod-ules with proposed configurations were fabricated in our lab and evaluated by pervaporationmeasurement in model butanol aqueous solution and real fermentation broth. The results suggested thatthe design of packing density and cross-section layout could realize the optimization of module config-uration. The optimized module filled with 7 bundles of hollow fiber membranes at a high packing densityof 560 m2/m3 exhibits a high and stable performance in the real ABE fermentation broth during 120 hcontinuous operation at 40 �C. The average total flux was 1000 g/m2 h and separation factor were 6.4for ethanol, 22.2 for butanol and 28.6 for acetone, respectively. Our results demonstrated that the hollowfiber modules developed in this work could be competitive candidates for the practical application in per-vaporation recovery of bio-butanol.

� 2015 Elsevier B.V. All rights reserved.

1. Introduction

Pervaporation (PV) is a membrane technology that could realizemolecular separation for liquid mixtures in which a feed solution ispassed over a membrane surface and some of the components areable to preferentially pass through the membrane and be concen-trated as vapors in the permeate [1]. With the advantages ofenergy-saving, cost-effective and environmental friendly, perva-poration technology has broad applications in various fields suchas chemical industry, energy, environment, pharmacy and foodengineering [2,3]. Among them, organophilic (usually hydropho-bic) pervaporation membranes have potential applications includ-ing bio-fuels recovery [4], VOCs removal [5], organic/organicmixtures separation [6], gasoline desulfurization [7] and so on.

Most of practical applied pervaporation membranes are com-posite membranes with a thin and dense active layer on a poroussubstrate [2]. Compared with the traditional polymeric substrates,the inorganic substrates exhibit the advantages in the chemical,mechanical and thermal stabilities [8]. Our group designed a newtype of ceramic-supported polymer composite membranes with

the deposition of polymeric active layers on the macroporous cera-mic substrate [9]. This kind of membranes have shown high sep-aration performance both in organophilic [10,11] and hydrophilicpervaporation applications [12]. Excellent results were also foundin other works that pervaporation composite membranes werefabricated using inorganic substrates [13–15]. Meanwhile, hollowfiber substrates have been received increasing attention in devel-opment of high-permeable membranes, because of their low trans-port resistance, high-packing density and self-support structure[16–18]. Zhang et al. proposed several hollow fiber compositemembranes prepared by self-assembly method [19–21]. Recently,Brown et al. reported a novel and scalable interfacial microfluidicprocessing approach for fabricating metal organic frameworks(MOFs) hollow fiber membranes [22]. In our previous work, high-quality ceramic hollow fiber composite membranes weredeveloped based on various kinds of materials, such as PDMS[23], ZIF-8 [24] and graphene oxide [25].

Although with many attractive characteristics, as well as a largenumber of investigations on development of membrane materialsand preparation methods, hollow fiber composite membranes stillhave not been industrially implemented in pervaporation process[2]. One of the major challenges impeding their application is thehigh-performance hollow fiber modules. For pushing them into

Nomenclature

List of symbolsu velocity components of the x directions (m/s)v velocity components of the y directions (m/s)w velocity components of the z directions (m/s)f volume force per phaseU velocity (m/s)J mass flux (g/m2 h)M weight of the permeate (g)A effective area of hollow fiber membrane (m2)t operation time (h)l membrane thickness (lm)Yi weight fractions of component i in permeateXi weight fractions of component i in feedJ0 average flux of hollow fiber membranes (g/m2 h)J1 flux of the module (g/m2 h)Jo constant (g/m2 h)EJ activation energy for permeation (kJ/mol)R gas constant (kJ/mol K)

T operation temperature (K)x axial coordinate (m)Sc Schmidt numberD mass diffusivity (m2/s)dc concentration boundary layer (m)du velocity boundary layer (m)

Greek lettersb separation factorq flow density (kg/m3)l dynamic viscosity (Pa s)�k second molecular viscosityuJ deviation factor of fluxub deviation factor of separation factorb0 average separation factor of hollow fiber membranesb1 separation factor of the module

D. Liu et al. / Separation and Purification Technology 146 (2015) 24–32 25

the industrial stage, more intensive efforts are needed to study thedesign and application of hollow fiber modules for pervaporationprocess. As for hollow fiber modules applied in water treatmentand gas separation, massive theoretically and experimentalattempts have been made to design the membrane modules[17,26–32]. Flow geometry was optimized to minimize the concen-tration polarization and enhance the mass transfer efficiency of thehollow fiber modules. Moreover, a few works indicated that theperformance of commercial hollow fiber modules for VOCs removalcould be effectively improved with the intensification of masstransfer by optimizing operating conditions [33,34]. However, thereare limited studies on designing hollow fiber modules for perva-poration applications in the open literature. Zhang et al. [20] carriedout a preliminary work on the preparation of pilot-scale hollowfiber membrane modules for hydrophilic pervaporation applica-tion. It was found that the packing density had great influenceson the performance of hollow fiber module. Both total flux andselectivity decreased with the increase of packing density. Untilnow, to the best of our knowledge, it has not yet been reported inthe hydrophobic hollow fiber modules design for pervaporationprocess, which is an essential part of pervaporation technology [3].

Therefore, this paper intended to explore the preparation of hol-low fiber modules loaded with hydrophobic PDMS/ceramic compos-ite membranes developed in our previous work. Computational fluiddynamics (CFD) technique was employed to simulate the flow fielddistribution of the hollow fiber modules with different config-urations. Two critical aspects for the hollow fiber modules, packingdensity and cross-section layout were systematically investigated tooptimize module design. As the typical pervaporation application ofhydrophobic PDMS membranes, the hollow fiber modules wereapplied for bio-butanol recovery from model aqueous solution andreal acetone–butanol–ethanol (ABE) fermentation broth. Keyoperating parameters (i.e., feed temperature, concentration, flowrate and long-term stability) were carried out to study the perva-poration performance of the PDMS/ceramic hollow fiber module.

2. Experimental

2.1. Preparation of hollow fiber modules

The fabrication of ceramic hollow fiber supported PDMS com-posite membranes was described in our previous work [23]. The

inside diameter, outside diameter and length of the hollow fibermembranes are 1.6 mm, 2 mm and 180 mm, respectively. Beforefilled into the module, each hollow fiber membranes were testedin 1 wt% n-butanol/water mixtures at 40 �C to evaluate the originalpervaporation performance. The hollow fiber membranes weresealed in stainless steel modules by epoxy resin. By varying thepacking density and cross-section layout, we prepared hollow fibermodules with different configurations, which are abbreviated asModule 1, Module 2, Module 3, Module 4 and Module 5, respec-tively. As shown in Fig. 1, Module 1, Module 2 and Module 3 con-sist of one bundle with packing density of 140 m2/m3, 280 m2/m3

and 560 m2/m3, respectively. With the same packing density of560 m2/m3, 4 bundles and 7 bundles were designed to prepareModule 4 and Module 5, respectively. The detailed parameters ofthese modules are listed in Table 1.

2.2. CFD simulation

Computational fluid dynamics (CFD) simulations (Ansys 14.0)code FLUENT were performed to simulation the flow situation inthe hollow fiber modules with a three-dimensional model. In orderto simplify the model, there are two assumptions: (1) the pureaquatic system was applied in this simulation; (2) there is no dif-ferential pressure between two sides of membrane. Thus there isno diffusion across the membrane. Based on the above assump-tions, the fluid flow in modules meets the conservation equationscan be written as follows:

Continuity equation:

@u@xþ @v@yþ @w@z¼ 0 ð1Þ

where u, v and w are the velocity components of the x, y and z direc-tions, respectively.

The momentum equation:

@ðquÞ@tþ @ðquuÞ

@xþ @ðqvuÞ

@yþ @ðqwuÞ

@z

¼ qf x �@p@xþ @

@x2l @u

@xþ �kdivU

� �þ @

@yl @v

@xþ @u@y

� �� �

þ @

@zl @w

@xþ @u@z

� �� �ð2Þ

Fig. 1. Schematic of configurations of hollow fiber modules.

Table 1The parameters of the hollow fiber modules.

Parameters Module 1 Module 2 Module 3 Module 4 Module 5

Numbers of hollow fiber 7 14 28 28 28Bundles of hollow fiber 1 1 1 4 7Membrane area of module (cm2) 79 158 316 316 316Packing density of module (m2/m3) 140 280 560 560 560The length of membrane module (mm) 180 180 180 180 180The diameter of membrane module (mm) 20 20 20 20 20

26 D. Liu et al. / Separation and Purification Technology 146 (2015) 24–32

@ðqvÞ@t

þ @ðquvÞ@x

þ @ðqvvÞ@y

þ @ðqwvÞ@z

¼ qf y �@p@yþ @

@y2l @v

@yþ �kdivU

� �þ @

@xl @v

@xþ @u@y

� �� �

þ @

@zl @w

@xþ @u@z

� �� �ð3Þ

@ðqwÞ@t

þ @ðquwÞ@x

þ @ðqvwÞ@y

þ @ðqwwÞ@z

¼ qf z �@p@yþ @

@z2l @w

@zþ �kdivU

� �þ @

@xl @u

@zþ @w@x

� �� �

þ @

@yl

@w@yþ @v@z

� �� �ð4Þ

where q is the flow density; fx, fy and fz are the volume force perphase of the x, y and z directions, respectively; l is dynamic viscos-ity; �k is second molecular viscosity; U is the velocity.

Construction details of hollow fiber membrane module in sim-ulation were shown in Fig. 2. The axial flow was along the x-axisnegative direction, while y and z was the planner coordinate. Thecoordinates-origin was established in the center of the circle atthe top of the modules. Meanwhile, ICEM (Ansys Inc., USA) wasused to simulate the flow field distribution in membrane modulescontaining about 1,500,000 tetrahedral cells. The three-dimen-sional (3D) Navier–Stokes (N–S) equations using Semi-Implicit

Fig. 2. Schematic of pervaporation hollow fiber module.

Method for Pressure-Linked Equations (SIMPLE method) wereemployed in this simulation. No-slip and no-penetration condi-tions were imposed on the walls of membrane module. Standardwall functions were used to specify the wall boundary conditions,which were applied to the membrane surface. Velocity and pres-sure were used as the inlet and outlet boundary conditions, respec-tively. Atmospheric pressure was used as the internal pressurefield. Other parameters in this simulation were fixed according tothe experimental conditions.

2.3. Pervaporation measurement

The pervaporation performance of the hollow fiber moduleswere measured on a home-made apparatus as displayed in Fig. 3.The feed tank with n-butanol/water model solution or acetone–bu-tanol–ethanol (ABE) fermentation broth (described in our previouswork [11]) was maintained at a constant temperature by thewater-bath. The feed flow from the feed tank entered into bottomand left at top of the membrane module. The vapor permeate wascollected by a cold trop immersed in liquid nitrogen. The com-position in feed and permeate side were analyzed by a gas chro-matography (GC-2014, SHIMADZU, Japan) equipped with a flameionization detector (FID). Iso-butanol was selected as an internalstandard substance. Nitrogen was used as carrier gas. The perme-ate sample was diluted by the deionized water and injected intothe gas chromatography.

The pervaporation performance of the hollow fiber moduleswere expressed in terms of the permeation flux J and separationfactor b as below:

J ¼ MAt

ð5Þ

b ¼ Yi=Yj

Xi=Xjð6Þ

where W is the weight of the permeate, A is the effective membranearea of the module, and t is the permeation time interval for thepervaporation; Y and X are the weight fractions of components ior j in the permeate and feed, respectively.

Fig. 3. Schematic of pervaporation set-up equipped with hollow fiber modules.

D. Liu et al. / Separation and Purification Technology 146 (2015) 24–32 27

In order to standardize the performance difference between thehollow fiber membranes and modules, deviation factors of flux uJ

and separation factor ub are defined as follows:

uJ ¼J0 � J1

J0� 100% ð7Þ

ub ¼b0 � b1

b0� 100% ð8Þ

where J0 is the average flux of hollow fiber membranes; J1 is the fluxof the module; b0 is the average separation factor of hollow fibermembranes and b1 is the separation factor of the module.

3. Results and discussion

3.1. Design and fabrication of hollow fiber modules

3.1.1. Effect of packing densityOne of the distinct advantages of hollow fiber membranes is the

high packing density. Thus, different numbers of hollow fibermembranes were put into the module with same volume, so asto prepare hollow fiber modules with various packing density.Fig. 4 shows the effect of packing density on the contours of xvelocities in the middle of Modules 1–3 (x = 0.09 m) predicted by

Fig. 4. Contours of x velocity (m/s) predicted by C

the CFD simulation. Module 1 and Module 2 with low packing den-sity have terrific distributions concerning the flow. The flow dis-tribution in Module 3 with high packing density is extremely lessthan those, especially that inside the bundle of hollow fiber mem-branes. It is indicated that the addition of packing density broughtabout the flow maldistribution. On the contrary, adding the num-bers of the hollow fiber membranes decreases the equivalentdiameter of the module and then increases Reynolds number,which might be beneficial for the separation factor [35].

To further observe the flow distributions of the whole module,we analyzed and calculated the average membrane surface veloci-ties from the bottom to the top of the module, namely, along the xcoordinate. Normalizing axial length between 0 and 1 as the hori-zontal axis, which makes the axial length universal. As shown inFig. 5a, the membrane surface velocity at different positions ofthe x direction generally follows the order of Module 1 > Module2 > Module 3. The relationship between velocity boundary layerand concentration boundary layer can be described as below:

Sc ¼ m=D ð9Þ

dc

du¼ 1ffiffiffiffiffi

Sc3p ð10Þ

where Sc is Schmidt number; m is kinematic viscosity; D is massdiffusivity; dc is concentration boundary layer; du is velocity

FD simulation in Modules 1–3 with hollow.

Fig. 5. (a) The membrane surface velocity at different positions along the modulelength, and (b) the deviation factors of flux and separation factor of Modules 1–3with hollow fiber packing density of 140, 280 and 560 m2/m3, respectively.

28 D. Liu et al. / Separation and Purification Technology 146 (2015) 24–32

boundary layer. According to Eqs. (9), (10), the value of concentra-tion boundary layer decreases with the reduction velocity boundarylayer. Thus, phenomenon of concentration polarization mentionedin the text will be improved. This result suggests that the increaseof the packing density accelerated the concentration polarizationin the module, leading to a decrease in the membrane surface veloc-ity. Compared with Module 3, the changes in the membrane surfacevelocity along x coordinate in Module 1 and Module 2 are muchsmaller, indicating that the membrane surface velocities in themodules with low packing density were more uniform. In the caseof membrane module with high packing density, the membranesurface velocity is significantly reduced in the middle of membranemodule. These results are in accordance with the velocity contoursin Fig. 4. The highest membrane surface velocity could be observedat x = 0.18 m, where closes to the inlet of membrane modules.

Based on the CFD simulation results, hollow fiber modules withdifferent packing density were prepared and tested in 1 wt% n-bu-tanol/water mixtures at 40 �C. The aim of simulation is to optimizethe module design for the improvement of performance in butanol/water separation. Thus, the deviation factor is defined to fairlyevaluate the pervaporation performance difference between thehollow fiber membranes and modules. Generally, a lower deviationfactor indicates a better membrane module.

The effect of packing density on the deviation factors of flux andseparation factor of the hollow fiber modules (Modules 1–3) wereexperimentally investigated, and shown in Fig. 5b. The flux of theModule 1 with packing density of 140 m2/m3 is decreased only

2.5% compared with the average flux of the hollow fiber mem-branes. However, the initial flux is significantly decreased 22.8%in Module 3 with high packing density of 560 m2/m3. It can beexpected that the deviation factor of flux (uJ) would be increasedwith the numbers of hollow fibers in the module, because the poorflow distribution in the high packing density hinders the masstransport on the membrane surface [27,29], as indicated inFigs. 4 and 5a. Meanwhile, it is found that the deviation factorsof separation factor (ub) all modules are relatively large (17.1–27.1%). It might be due to the reduction of Reynolds numbers fromthe single hollow fiber membranes to that in the modules.However, the lowest ub of Module 2 indicated that the spacing-likepacking of hollow fibers would improve the regional turbulence[28], which can be observed in Fig. 4.

3.1.2. Effect of cross-section layoutBesides of the packing density, the cross-section layout is also a

very important parameter for designing membrane module [29].As for the hollow fiber membrane modules, packing the hollowfibers with different bundle numbers is considered as a convenientapproach to vary the cross-section layout. Thus, by filling the samenumbers of hollow fiber membranes into the module to control thesame packing density, we designed different cross-section layoutsto optimize the module configuration. Considering the industrialapplication, Module 3 with the highest packing density (560 m2/m3) was chosen to be optimized by designing the cross-sectionlayout.

The simulated contours of x velocities in the middle (x = 0.09 m)of the modules with different cross-section layouts (Modules 3–5)are also shown in Fig. 6. It can be clearly seen that the cross-sectionlayout played a significant role in determining the velocity dis-tributions in the modules. The increased bundles of hollow fibermembranes distinctly improved the bulk flow patterns and as aresult the velocity boundary layer of the membrane surfacebecame thinner [36]. This illustrates that the cross-section layoutscould strongly influence the performance of the modules. A favor-able flow distribution could be obtained by a reasonable design ofthe cross-section layout, even the module with high packingdensity.

Meanwhile, the average membrane surface velocities along themodule length were calculated in order to study the flow dis-tributions of the whole module. The result is given in Fig. 7a.With the increase of hollow fiber bundles in the module, the mem-brane surface velocities in the entire module are graduallyimproved. They could reach up to 0.052–0.066 m/s as high as thatin the lower packing density (Module 2), as filling the hollow fibermembranes with 7 bundles (Module 5). The division of 1 bundleinto 4 or even 7 bundles made more membrane areas of hollowfibers adequately contact with the feed, contributing to the signifi-cant enhancement of flow distribution on most of the membranesurface. It is also found in the velocity contours of Fig. 6, thereare more and more high-velocity areas on the membrane surfaceof the hollow fibers with the increase of bundles. Furthermore, itcould be noted that cross-section layout with several bundles aresimilar to several units of hollow fibers, which provide a facileassembly and sealing of hollow fibers in the practical application.

The pervaporation performance of the hollow fiber module withdifferent cross-section layouts was also measured in the 1 wt% n-butanol/water at 40 �C. The results are shown in Fig. 7b that givesthe effect of hollow fiber bundles on the deviation factors of fluxand separation factor of the Modules 3–5 with the same packingdensity of 560 m2/m3. It was interesting to find that uJ and ub

are remarkably and simultaneously reduced with increasing thebundles of hollow fiber membranes in the module. As predictedby CFD simulation, the modules with increased hollow fiber bun-dles showed better flow distribution because of the enhancement

Fig. 6. Contours of x velocity (m/s) predicted by CFD simulation in Modules 3–5 with hollow fiber bundles of 1, 4 and 7, respectively.

Fig. 7. (a) The membrane surface velocity at different positions along the modulelength, and (b) the deviation factors of flux and separation factor of Modules 3–5with hollow fiber bundles of 1, 4 and 7, respectively.

D. Liu et al. / Separation and Purification Technology 146 (2015) 24–32 29

of surface velocity and alleviation of concentration polarization,leading to the higher flux and separation factor. Although withthe double packing density, the 4-bundles module exhibited nearlythe same pervaporation performance as 1-bundle module. As thenumber of bundles further increases to 7, the pervaporation perfor-mance of the module could be much higher. It was confirmed thatthe cross-section layout optimizing is a powerful tool to designhighly packed hollow fiber modules with high-performance.

The packing density of tubular pervaporation membrane mod-ules is generally limited to 100 m2/m3. Wang et al. [37] found thatthe flux of zeolite hollow fiber module with 1-bundle and 360 m2/m3 packing density dropped by ca. 60% compared with the singlehollow fiber membrane. In another reported hollow fiber modulesfilled with polyelectrolyte pervaporation membranes, the flux wasreduced by 5 times with the packing density increased from100 m2/m3 to 500 m2/m3 [20]. In contrast, our 7-bundles hollowfiber module with 560 m2/m3 packing density only had only13.3% and 11.1% loss in flux and separation factor of the single hol-low fiber membrane, respectively. It is the first time to realize thedesign of cross-section layout to obtain high-performance hollowfiber modules with high packing density (>500 m2/m3). The valuecan be expected to be further increased by the use of hollow fiberswith smaller dimensions.

3.2. Application of hollow fiber modules

Butanol, a good solvent and important chemical feedstock, hasbeen widely used in the chemical, plastic, cosmetic, paint indus-tries, etc. It is also a new kind of advanced biofuel, which has theadvantages of being less volatile and flammable, having a higherenergy content, having water insensitivity, and being less haz-ardous to handle compared with bio-ethanol [38]. Therefore, theoptimized module (Module 5: 7-bundles, 560 m2/m3 packing den-sity) filled with the hollow fiber PDMS/ceramic composite mem-branes, was applied for pervaporation recovery of n-butanol fromthe model aqueous solution as well as the real ABE fermentationbroth. Various operating parameters including feed temperature,concentration and flow rate were carried out to study the perva-poration performance of the module. Moreover, by normalizingthe permeate flux with respect to the driving force will be usefulto further understand the permeate-specific intrinsic membraneproperties of the hollow fiber modules. These intrinsic membraneproperties are permeance and selectivity calculated in supportinginformation. Additionally, the long-term stability of the modulein real ABE fermentation broth was investigated as well.

3.2.1. Effect of feed flow rateThe effect of feed flow rate on the pervaporation performance of

the hollow fiber module is shown in Fig. 8. With increasing the feedflow rate into the module, all the fluxes (n-butanol, water andtotal) firstly increase a lot and then almost kept stable. So as theseparation factor, butanol content in permeate of the module.Compared with the high feed flow rate, the low feed flow rate iseasier to generate polarization of the temperature and concentra-tion. Besides, increasing the feed flow rate could decrease the

Fig. 8. Effect of the feed flow rate on pervaporation performance of Module 5: (a)flux and (b) separation factor and butanol content in permeate (feed conditions:1 wt% n-butanol/water mixtures, 40 �C).

Fig. 9. (a) Effect of the feed temperature on pervaporation performance and (b)Arrhenius plots of partial fluxes of Module 5 (feed conditions: 1 wt% n-butanol/water mixtures, feed flow rate of 124 L/h).

30 D. Liu et al. / Separation and Purification Technology 146 (2015) 24–32

boundary layer thickness and bring about improvement of masstransfer coefficient [39]. Hence, higher flux and separation factorwere obtained at larger feed flow rate. However, these effects areno more significant as further raising the feed flow rate (120–220 L/h). The permeance and selectivity of the hollow fiber moduleunder different feed flow rate are shown in Fig. S1. In contrast tothe higher water flux, the permeance of butanol is much higherthan that of water by normalizing the flux with respect to the driv-ing force. Thus the hollow fiber module is selective permeation forbutanol rather than water. It is also found similar trends in the per-meance and selectivity by increasing the feed flow rate. This isbecause the driving force of the pervaporation process is nearlyindependent on the flow rate. Since higher feed flow rate consumesmore energy, 124 L/h feed flow rate is appropriate for the moduleoperation.

3.2.2. Effect of feed temperatureFig. 9 displays the effect of feed temperature on the pervapora-

tion performance of the hollow fiber module. As expected, themodule shows a progressively improvement in fluxes whereas adecline in the separation factor. The elevated temperature pro-motes the molecular transport in the module in two ways [1]:(1) increasing the driving force of the feeds and (2) enlarging thefree volumes of the hollow fiber membrane. As shown in Fig. 9a,because of the smaller molecular kinetic diameter, water has amuch higher flux increasing rate than the n-butanol, resulting inthe decrease of separation factor. In summary, the total or partial

flux varied with the temperature can be described by theArrhenius expression [40]:

J ¼ Joexp�EJ

RT

� �ð11Þ

where Jo is constant; EJ is the activation energy for permeation; R isthe gas constant; T is the operation temperature in Kelvin. The lin-ear Arrhenius plots for water and n-butanol permeation fluxes areshown in Fig. 9b. The calculated average activity energies of waterand n-butanol permeation through the module are 37.5 kJ/moland 24.7 kJ/mol, respectively. This indicates that the flux of waterwas more sensitive to the temperature variation than that of n-bu-tanol, leading to the decrease of separation factor with feed tem-perature. In addition, the permeance and selectivity werecalculated with the aim of distinguishing the effects of driving forcefrom the intrinsic membrane properties altered by feed tempera-ture. As shown in Fig. S2, the permeance of butanol decreases alot, while the water permeance almost keeps constant, leading tothe reduced selectivity. It is possible due to that the preferentialsorption of the membrane surface was restrained by increasingthe feed temperature.

3.2.3. Effect of feed concentrationThe effect of feed concentration on the pervaporation perfor-

mance of the hollow fiber module is shown in Fig. 10. Withincreasing the n-butanol concentration in feed, the butanol andtotal flux increase significantly with linear relations, while the

Fig. 10. Effect of the feed concentration on pervaporation performance of Module5: (a) flux and (b) separation factor and butanol content in permeate (feedconditions: feed flow rate of 124 L/h, 40 �C).

D. Liu et al. / Separation and Purification Technology 146 (2015) 24–32 31

water flux has a very slow addition. As a result, there is 45.5%improvement of butanol content in permeate and ca. 21% reduc-tion in separation factor with the butanol feed concentrationranges from 0.5 wt% to 5.3 wt%. The pervaporation membranewould be more swelled by the component whose polarity is moreclose to the membrane material [41]. The polarity of n-butanol ismore similar to that of cross-linked PDMS than water [4]. As aresult, more n-butanol molecules dissolved in the polymer chainswith the n-butanol increased, which led to the larger swelling ofthe membrane. Both of the n-butanol and water permeatedthrough the membrane easily. Due to the water molecules aresmaller than n-butanol, the diffusion rate of water is larger thanthat of n-butanol, resulting in the decreased separation factor. Asshown in Fig. S3, with increasing the feed concentration the buta-nol permeance increases at first and then keeps stable anddecreases at last, while the water permeance has a few increases.Thus, membrane selectivity decreased with raising the butanolcontents in feed. The complex variation of butanol permeancemay be attributed to the competitive adsorption and diffusioneffects resulted from increased butanol molecules in themembrane.

Fig. 11. Long-term stability of hollow fiber module: (a) flux and (b) separationfactor (feed conditions: feed flow rate of 124 L/h at 40 �C, ABE concentrations:0.6 wt% acetone, 1.2 wt% n-butanol and 0.2 wt% ethanol).

3.2.4. Long-term stability in ABE fermentation brothIn the process of practical industrial application, a membrane

module with excellent performance is not only relied on high per-meation flux and separation factor, but also having a good long-term stability in the real separation system. Bio-butanol is

produced from the biomass ABE fermentation process [38]. Thus,the performance and stability of the hollow fiber module weretested in the pervaporation of real ABE fermentation broth. Thebroth is more complex than the model butanol/water mixturesbecause it contains not only the variation of density, pH and viscos-ity, but also the inorganic salt, glucose, live and dead microbialcells and several other metabolic compounds [11].

Fig. 11 presents the long-term stability of the hollow fibermembrane module in ABE fermentation broth (0.6 wt% acetone,1.2 wt% n-butanol and 0.2 wt% ethanol) at 40 �C. It was found thatall the fluxes and separation factors kept stable over a long periodof 120 h continuous operation. A few deviations of performancewere due to the fluctuation of ABE feed concentration. The moduleexhibited a high performance in the ABE fermentation broth, withaverage total flux of 1000 g/m2 h and separation factor of 28.6 foracetone, 22.2 for butanol, 6.4 for ethanol, respectively.

The separation performance of various pervaporation mem-branes in ABE fermentation broth is listed in Table S1. It could befound that the separation performance of our hollow fiber moduleowns a high flux and good separation factor than that of the poly-meric membranes, and even the mixed matrix membranes. It ismainly due to the low transport resistance of the ceramic hollowfiber substrate and the thin and defect-free PDMS separation layer.Moreover, the optimal-designed hollow fiber module with a highpacking density of 560 m2/m3 almost performs the original perva-poration performance of the single PDMS/ceramic hollow fibermembrane. It is indicated that the hollow fiber module with cera-mic-supported PDMS composite membranes would be a promisingcandidate for the practical application in bio-butanol production.

32 D. Liu et al. / Separation and Purification Technology 146 (2015) 24–32

4. Conclusions

Hollow fiber modules of ceramic-supported PDMS compositemembranes were successfully prepared in this work. The flow fielddistribution in the modules analyzed by CFD simulation revealedthat the module design could be effectively optimized by control-ling the packing density and cross-section layout. A high-perfor-mance hollow fiber module was obtained by filling 7 bundles ofhollow fiber membranes with a high packing density of 560 m2/m3. Both feed temperature and concentration played importantroles in determining the pervaporation performance of the module.The hollow fiber module performed a high and stable performancein the real ABE fermentation broth during 120 h continuous opera-tion: with average total flux of 1000 g/m2 h and separation factorof 6.4 for ethanol, 22.2 for butanol and 28.6 for acetone, respec-tively. Our work demonstrated that the hollow fiber module filledwith ceramic-supported PDMS composite membranes shows anattractive prospect in the pervaporation application of practicalbio-butanol production.

Acknowledgments

This work was supported by the National Natural ScienceFoundation of China (Nos. 21406107, 2149580015, 21476107),Innovative Research Team Program by the Ministry of Educationof China (No. IRT13070) and the Project of Priority AcademicProgram Development of Jiangsu Higher Education Institutions(PAPD).

Appendix A. Supplementary material

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.seppur.2015.03.029.

References

[1] X. Feng, R.Y.M. Huang, Liquid separation by membrane pervaporation: areview, Ind. Eng. Chem. Res. 36 (1997) 1048–1066.

[2] A. Jonquieres, R. Clement, P. Lochon, J. Neel, M. Dresch, B. Chretien, Industrialstate-of-the-art of pervaporation and vapour permeation in the westerncountries, J. Membr. Sci. 206 (2002) 87–117.

[3] B. Van der Bruggen, P. Luis, Pervaporation as a tool in chemical engineering: anew era?, Curr Opin. Chem. Eng. 4 (2014) 47–53.

[4] G. Liu, W. Wei, W. Jin, Pervaporation membranes for biobutanol production,ACS Sustain. Chem. Eng. 2 (2013) 546–560.

[5] G. Busca, S. Berardinelli, C. Resini, L. Arrighi, Technologies for the removal ofphenol from fluid streams: a short review of recent developments, J. Hazard.Mater. 160 (2008) 265–288.

[6] P. Luis, J. Degrève, B.V. der Bruggen, Separation of methanol–n-butyl acetatemixtures by pervaporation: potential of 10 commercial membranes, J. Membr.Sci. 429 (2013) 1–12.

[7] L.G. Lin, Y.Z. Zhang, Y. Kong, Recent advances in sulfur removal from gasolineby pervaporation, Fuel 88 (2009) 1799–1809.

[8] M. Pera-Titus, Porous inorganic membranes for CO2 capture: present andprospects, Chem. Rev. 114 (2013) 1413–1492.

[9] G.P. Liu, W. Wei, W.Q. Jin, N.P. Xu, Polymer/ceramic composite membranes andtheir application in pervaporation process, Chin. J. Chem. Eng. 20 (2012) 62–70.

[10] F.J. Xiangli, W. Wei, Y.W. Chen, W.Q. Jin, N.P. Xu, Optimization of preparationconditions for polydimethylsiloxane (PDMS)/ceramic compositepervaporation membranes using response surface methodology, J. Membr.Sci. 311 (2008) 23–33.

[11] G.P. Liu, W. Wei, H. Wu, X.L. Dong, M. Jiang, W.Q. Jin, Pervaporationperformance of PDMS/ceramic composite membrane in acetone butanolethanol (ABE) fermentation-PV coupled process, J. Membr. Sci. 373 (2011)121–129.

[12] Y.X. Zhu, S.S. Xia, G.P. Liu, W.Q. Jin, Preparation of ceramic-supportedpoly(vinyl alcohol)-chitosan composite membranes and their applications inpervaporation dehydration of organic/water mixtures, J. Membr. Sci. 349(2010) 341–348.

[13] T.A. Peters, C.H.S. Poeth, N.E. Benes, H.C.W.M. Buijs, F.F. Vercauteren, J.T.F.Keurentjes, Ceramic-supported thin PVA pervaporation membranes

combining high flux and high selectivity; contradicting the flux-selectivityparadigm, J. Membr. Sci. 276 (2006) 42–50.

[14] X.-L. Liu, Y.-S. Li, G.-Q. Zhu, Y.-J. Ban, L.-Y. Xu, W.-S. Yang, An organophilicpervaporation membrane derived from metal–organic frameworknanoparticles for efficient recovery of bio-alcohols, Angew. Chem. Int. Ed. 50(2011) 10636–10639.

[15] S. Liu, W.K. Teo, X. Tan, K. Li, Preparation of PDMSvi–Al2O3 composite hollowfibre membranes for VOC recovery from waste gas streams, Sep. Purif. Technol.46 (2005) 110–117.

[16] Z.B. Wang, Q.Q. Ge, J. Shao, Y.S. Yan, High performance zeolite LTApervaporation membranes on ceramic hollow fibers by dipcoating-wipingseed deposition, J. Am. Chem. Soc. 131 (2009) 6910–6911.

[17] X. Feng, J. Ivory, Development of hollow fiber membrane systems for nitrogengeneration from combustion exhaust gas: Part I. Effects of moduleconfigurations, J. Membr. Sci. 176 (2000) 197–207.

[18] S. Darvishmanesh, F. Tasselli, J.C. Jansen, E. Tocci, F. Bazzarelli, P. Bernardo, P.Luis, J. Degrève, E. Drioli, B. Van der Bruggen, Preparation of solvent stablepolyphenylsulfone hollow fiber nanofiltration membranes, J. Membr. Sci. 384(2011) 89–96.

[19] G.J. Zhang, X. Song, S.L. Ji, N.X. Wang, Z.Z. Liu, Self-assembly of inner skinhollow fiber polyelectrolyte multilayer membranes by a dynamic negativepressure layer-by-layer technique, J. Membr. Sci. 325 (2008) 109–116.

[20] G.J. Zhang, N.X. Wang, X. Song, S.L. Ji, Z.Z. Liu, Preparation of pilot-scale innerskin hollow fiber pervaporation membrane module: effects of dynamicassembly conditions, J. Membr. Sci. 338 (2009) 43–50.

[21] J.X. Guo, G.J. Zhang, W. Wu, S.L. Ji, Z.P. Qin, Z.Z. Liu, Dynamically formed innerskin hollow fiber polydimethylsiloxane/polysulfone composite membrane foralcohol permselective pervaporation, Chem. Eng. J. 158 (2010) 558–565.

[22] A.J. Brown, N.A. Brunelli, K. Eum, F. Rashidi, J.R. Johnson, W.J. Koros, C.W. Jones,S. Nair, Interfacial microfluidic processing of metal–organic framework hollowfiber membranes, Science 345 (2014) 72–75.

[23] Z.Y. Dong, G.P. Liu, S.N. Liu, Z.K. Liu, W.Q. Jin, High performance ceramic hollowfiber supported PDMS composite pervaporation membrane for bio-butanolrecovery, J. Membr. Sci. 450 (2014) 38–47.

[24] K. Huang, Z.Y. Dong, Q.Q. Li, W.Q. Jin, Growth of a ZIF-8 membrane on theinner-surface of a ceramic hollow fiber via cycling precursors, Chem. Commun.49 (2013) 10326–10328.

[25] K. Huang, G.P. Liu, Y.Y. Lou, Z.Y. Dong, J. Shen, W.Q. Jin, A graphene oxidemembrane with highly selective molecular separation of aqueous organicsolution, Angew. Chem.-Int. Ed. 53 (2014) 6929–6932.

[26] M.M. Teoh, S. Bonyadi, T.-S. Chung, Investigation of different hollow fibermodule designs for flux enhancement in the membrane distillation process, J.Membr. Sci. 311 (2008) 371–379.

[27] D. Xiao, W. Li, S. Chou, R. Wang, C.Y. Tang, A modeling investigation onoptimizing the design of forward osmosis hollow fiber modules, J. Membr. Sci.392–393 (2012) 76–87.

[28] X. Yang, R. Wang, A.G. Fane, Novel designs for improving the performance ofhollow fiber membrane distillation modules, J. Membr. Sci. 384 (2011) 52–62.

[29] X. Yang, R. Wang, A.G. Fane, C.Y. Tang, I.G. Wenten, Membrane module designand dynamic shear-induced techniques to enhance liquid separation byhollow fiber modules: a review, Desalin. Water Treat. 51 (2013) 3604–3627.

[30] J. Ivory, X. Feng, G. Kovacik, Development of hollow fiber membrane systemsfor nitrogen generation from combustion exhaust gas: Part II: Full-scalemodule test and membrane stability, J. Membr. Sci. 202 (2002) 195–204.

[31] S. Kim, S.H. Han, Y.M. Lee, Thermally rearranged (TR) polybenzoxazole hollowfiber membranes for CO2 capture, J. Membr. Sci. 403–404 (2012) 169–178.

[32] H.Y. Hwang, S.Y. Nam, H.C. Koh, S.Y. Ha, G. Barbieri, E. Drioli, The effect ofoperating conditions on the performance of hollow fiber membrane modulesfor CO2/N2 separation, J. Ind. Eng. Chem. 18 (2012) 205–211.

[33] S.V. Suggala, P.K. Bhattacharya, Real coded genetic algorithm for optimizationof pervaporation process parameters for removal of volatile organics fromwater, Ind. Eng. Chem. Res. 42 (2003) 3118–3128.

[34] A.A. Keller, B.G. Bierwagen, Hydrophobic hollow fiber membranes for treatingMTBE-contaminated water, Environ. Sci. Technol. 35 (2001) 1875–1879.

[35] S. Schnabel, P. Moulin, Q.T. Nguyen, D. Roizard, P. Aptel, Removal of volatileorganic components (VOCs) from water by pervaporation: separationimprovement by Dean vortices, J. Membr. Sci. 142 (1998) 129–141.

[36] A.G. Fane, S. Chang, E. Chardon, Submerged hollow fibre membrane module –design options and operational considerations, Desalination 146 (2002) 231–236.

[37] X.R. Wang, Y.Y. Chen, C. Zhang, X.H. Gu, N.P. Xu, Preparation andcharacterization of high-flux T-type zeolite membranes supported on YSZhollow fibers, J. Membr. Sci. 455 (2014) 294–304.

[38] N. Qureshi, T.C. Ezeji, Butanol, ‘a superior biofuel’ production from agriculturalresidues (renewable biomass): recent progress in technology, Biofuels,Bioprod. Biorefin. 2 (2008) 319–330.

[39] H. Yu, X. Yang, R. Wang, A.G. Fane, Numerical simulation of heat and masstransfer in direct membrane distillation in a hollow fiber module with laminarflow, J. Membr. Sci. 384 (2011) 107–116.

[40] F. Xianshe, R.Y.M. Huang, Estimation of activation energy for permeation inpervaporation processes, J. Membr. Sci. 118 (1996) 127–131.

[41] P. Shao, R.Y.M. Huang, Polymeric membrane pervaporation, J. Membr. Sci. 287(2007) 162–179.