Fabrication and characterization of a surface-patterned thin filmcomposite membrane

Sajjad H. Maruf, Alan R. Greenberg, John Pellegrino, Yifu Ding n

Membrane Science, Engineering and Technology Center, Department of Mechanical Engineering, 427 UCB, University of Colorado Boulder,1111 Eng. Drive, Boulder, CO 80309-0427, USA

a r t i c l e i n f o

Article history:Received 25 September 2013Received in revised form9 October 2013Accepted 10 October 2013Available online 17 October 2013

Keywords:Surface topographyThin film composite membraneInterfacial polymerizationConcentration polarization

a b s t r a c t

Thin film composite (TFC) membranes are critical components for reverse osmosis (RO) and nanofiltra-tion (NF) processes. Similar to other liquid-based filtration membranes, TFC membranes are susceptibleto concentration polarization and fouling/scaling. Recently, surface topography modification has beenshown as a potential approach for fouling mitigation. However, for TFC membranes, tailoring the surfacetopography remains a challenge. Here, we demonstrate for the first time, successful fabrication of apatterned TFC membrane. A two-step fabrication process was carried out by (1) nanoimprinting apolyethersulfone (PES) support, and (2) forming a thin dense film atop the PES support via interfacialpolymerization (IP) with trimesoyl chloride and 1,3-phenylenediamine solutions. Chemical, topographic,and permeation characterization was performed on the imprinted IP membranes, and their permselec-tivity was compared with that of a flat (non-imprinted) TFC membrane prepared using the same IPprocedure.

& 2013 Elsevier B.V. All rights reserved.

1. Introduction

Thin film composite (TFC) membranes have experiencedremarkable development since the concept of interfacial polymer-ization (IP) was introduced [1]. Cadotte and co-workers developedthe first TFC membrane that has since become the most widelyused membrane for a multitude of reverse osmosis (RO) andnanofiltration (NF) applications [2]. Among the different types ofTFC membranes, the use of a crosslinked aromatic polyamide thinfilm as a barrier layer is a highly-cited chemistry [3–5]. TFCmembranes have become the industry standard over the lastseveral decades because of continuing research and developmentefforts that improved their selectivity and permeability togetherwith excellent mechanical strength and fouling resistance bymodifying both the barrier layer and the porous support [2,6].

A crosslinked aromatic (or semi-aromatic) polyamide barrierlayer in TFC membranes is typically formed over the poroussupports via in-situ IP of a poly-functional amine and an acidchloride at the organic solvent/aqueous solution interface [7,8].Often, ultrafiltration (UF) membranes based on polysulfone (PS),polyethersulfone (PES) and polyvinylidene fluoride (PVDF) are usedas the porous support [9,10]. Like other liquid-filtration membranes,TFC membranes are subject to concentration polarization and

fouling, which reduce membrane performance during filtration[11]. Hence, much research continues to be carried out to addressfouling issues, including various chemical treatments, adsorption ofsurfactants, low-temperature plasma treatments, irradiation meth-ods and addition of hydrophilic particles on the membrane surface[7,12]. By comparison, the use of controlled surface topography tomitigate concentration polarization or fouling in TFC membraneshas been scarce mainly due to the lack of methods to createtargeted topography on the membrane surface. The nature of theIP process, i.e. fast reaction at the organic/water interface, presents amajor obstacle for tailoring the structures and properties of thepolyamide barrier layers [13,14].

On the other hand, a patterned surface has been proveneffective for controlling cellular responses [15,16], which can beutilized as an effective anti-fouling approach. A polydimethylsi-loxane (PDMS) surface containing sharkskin-like micropatterns[17] showed a �86% reduction in the settling density of sporeswhen compared with an otherwise similar, but smooth, PDMSsurface. The extension of surface patterning to membranes hasbeen previously reported in the literature. Micro-molding incombination with phase inversion has been utilized to fabricatesurface-patterned flat membranes [18], as well as hollow fibermembranes [19,20]. Using this method, Won et al. created PVDFmicrofiltration (MF) membranes with �10 μm-scale features [21],which resulted in reduction of microbial fouling [22]. Recently, wedemonstrated the use of nanoimprint lithography (NIL) to impartsub-micron surface patterns directly onto a commercial PES UF

Contents lists available at ScienceDirect

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

Journal of Membrane Science

0376-7388/$ - see front matter & 2013 Elsevier B.V. All rights reserved.http://dx.doi.org/10.1016/j.memsci.2013.10.017

n Corresponding author. Tel.: þ1 303 492 2036; fax: þ1 303 492 3498.E-mail address: yifu.ding@colorado.edu (Y. Ding).

Journal of Membrane Science 452 (2014) 11–19

membrane without sacrificing its permselectivity [23,24]. Thepresence of these patterns significantly reduced deposition ofboth colloidal particles and protein during filtration.

In this study, we report the first fabrication of a functional TFCmembrane with well-controlled surface patterns. The two-stepfabrication process consisted of forming a dense polyamide barrierlayer via IP atop a nanoimprinted UF support membrane. Sys-tematic characterization of the patterned TFC membrane wascarried out, and the results show that this approach can indeedcreate reliable TFC membranes with separation performancecomparable with current commercial TFC RO/NF membranes.The comparison between the patterned and non-patterned TFCmembrane indicates that surface patterns can be an effectiveapproach to mitigate concentration polarization and scaling.

2. Experimental

2.1. Fabrication of patterned and non-patterned TFC membranes

Patterned TFC membranes were fabricated via a two-stepprocess that consisted of (1) nanoimprinting a PES support, and(2) forming a thin dense film atop the PES support via IP process.A commercial PES UF membrane (PW, GE Water and Infrastruc-ture) with a nominal 30 kg/mol molecular mass cutoff (MWCO)was used as the substrate on which the polyamide thin film washand-cast via IP. The nanoimprinting process for the UF membranewas described in detail in our previous work [24]. Briefly, the NILprocess was carried out in an Eitrie 3 (Obducat, Inc.) nanoimprinter,using a silicon mold containing parallel line-and-space gratings (aperiodicity of 834 nm, groove depth of 200 nm, and a line-to-space

ratio of 1:1). The Si mold surface was treated with a Piranhas

solution (3:1 concentrated sulfuric acid to 30% hydrogen peroxidesolution) prior to the imprinting. The NIL process was carried out at120 1C with a pressure of 4 MPa for 180 s, and the mold wasseparated from the membrane samples at 40 1C. The imprinted UFmembranes were cleaned with and stored in de-ionized (DI) waterin the dark until forming the polyamide layer. Non-patterned TFCmembranes that served as a reference were fabricated using thesame IP process on the PES UF membranes.

Both patterned and non-patterned UF membranes were tapedto a glass plate with the skin layer facing upwards, and placed inan aqueous amine monomer solution (Fig. 1). The aqueous aminesolution was prepared by adding 2 g of triethylamine (TEA, 99.5%,Sigma Aldrich), and 4 g of (þ)10-camphor sulfonic acid (CSA,99.0%, Sigma Aldrich), to �80 mL of DI water under vigorousstirring. CSA improves the absorption of the amine solution in thesupport membrane, while TEA accelerates the MPD–TMC reaction[25]. After complete dissolution of the TEA–CSA mixture, DI waterwas added to reach a total solution of 100 mL. Next, 2 g of 1,3-phenylenediamine (MPD, Sigma Aldrich) was added to the TEA–CSA solution. The entire UF membrane was then immersed in theaqueous MPD–TEA–CSA solution for 8 s, and the excess solution onthe membrane surfaces was removed with an air blower. Subse-quently, the amine-soaked UF membrane was immersed in ahexane solution (Fisher scientific) containing 0.1% (w/v) trimesoylchloride (TMC, 99%, Sigma-Aldrich) for 8 s. The resulting mem-brane was withdrawn from the hexane solution, cured at 70 1C for10 min, and washed thoroughly with DI water. Here, the protocolsfor the polyamide thin film formation were based on the formula-tion used by Ghosh et al. with the exception of a much shorter IPexposure time [9,25]. Finally, the as-prepared TFC membranes,

Fig. 1. A schematic representation of the interfacial polymerization process used to fabricate the patterned TFC membranes. The monomers m-phenylenediamine andtrimesoyl chloride react to form a highly cross-linked polyamide layer atop the patterned polyethersulfone UF membrane used as a support.

S.H. Maruf et al. / Journal of Membrane Science 452 (2014) 11–1912

with or without surface patterns, were stored in DI water at 5 1C inthe dark.

2.2. Membrane characterization

Attenuated Total Reflectance-Fourier transform infrared (ATR-FTIR) spectroscopy (Nicolet 6700 FTIR spectrometer, ThermoFisher Scientific, equipped with a diamond ATR crystal) was usedto characterize the polyamide barrier layers. Both the PES UFmembrane support and the corresponding TFC membranes (afterthe IP process) were measured. Three replicate ATR-FTIR spectrawere obtained for each membrane sample with each spectrumaveraged from 128 scans collected from 700 to 2200 cm�1 at1 cm�1 resolution. Membrane samples were extensively rinsedand soaked in DI water for 24 h before they were dried in avacuum oven prior to the ATR-FTIR measurements.

Surface topography and cross-sections of the membranes, beforeand after fabrication, were examined with a field-emission scanningelectron microscope (FESEM, Zeiss, Supra 60) and an atomic forcemicroscopy (AFM, Dimension 3100 AFM, Bruker). Membrane sam-ples were dried in a vacuum oven prior to SEM measurements, andthe membrane cross-sections were prepared using a microtome at�20 1C, and coated with a 4.7 nm gold layer. All AFM measure-ments were performed with the tapping mode under ambientconditions using silicon cantilever probe tips (Veeco, RTESP).

2.3. Filtration experiments

All of the filtration experiments with the TFC membranes wereconducted in a Sterlitech HP4750 high-pressure stainless steelstirred cell (Sterlitech, WA) using a constant-pressure, stirred/unstirred, dead-end (normal flow) filtration configuration. Theschematic of the filtration setup is provided in the supportinginformation (Fig. S1). The cell has an inner diameter of 3.2 cm andan effective membrane area of 8.48 cm2, and uses high-pressurenitrogen to supply the required pressure. The permeation massflow-rate was obtained by weighing samples over timed intervalsusing an automated electronic balance (PI-225DA, Denver Instru-ment). All of the filtration experiments were carried out at roomtemperature (�25 1C).

The entire experimental protocol utilized the following steps.For a given membrane, DI water filtration was carried out at threeoperating pressures, 1.38, 2.07 and 2.76 MPa, for 2 h, following a2.5–3 h membrane compaction at each pressure. Permeate fluxapproached steady state after the compaction period, and compac-tion for the membranes (estimated by the change in membraneresistance relative to the initial, uncompacted state) typicallyranged from 28% to 33%. After the completion of the pressure-stepping, DI water filtration was conducted at 2.76 MPa for 12 h.Subsequently, the pressure was released completely, and the DIwater feed was replaced with a 1000 mg/L aqueous NaCl (Mal-linckrodt, St. Louis, MO) solution. Filtration of the salt solution wascarried out at a pressure of 2.76 MPa over 3 h, and the collectedpermeate was weighed and the conductivity was measured every10 min. The conductivity was measured with an Ultrameter 6 P(Myron L, Carlsbad, CA), and the concentrations were calculatedfrom the calibration curve prepared for the instrument. For eachmembrane sample, NaCl filtration was performed twice for bothstirred and unstirred conditions. After the NaCl filtration thepressure was released, the whole filtration system along withthe membrane sample was rinsed in DI water, and the NaClsolution was replaced by a 1000 mg/L CaCl2 aqueous solution.Filtration of the CaCl2 solution was carried out using the sameprotocol as that for the NaCl solution, at both stirred and unstirredconditions.

After the CaCl2 filtration the pressure was again released, andthe whole filtration system along with the membrane sample wascleaned using DI water. Finally, the solution was replaced by a1000 mg/L CaSO4 (gypsum) solution for a scaling experiment,which was performed using an operating pressure of 2.76 MPaover 24 h under the stirred condition only. After each filtrationexperiment, the membranes sample was collected and rinsed withDI water to remove loosely attached gypsum crystals from themembrane surface and kept in a refrigerator at 5 1C in a sealedcontainer for SEM inspection. SEM images of the scaled mem-branes were taken at different and representative regions acrossthe membrane samples. The SEM samples were prepared bydrying the scaled membranes at room temperature for 24 h andthen resealing them in a Petri dish at 5 1C until the SEM imaging.Prior to SEM imaging, both patterned and non-patterned mem-branes were coated with �4 nm of gold.

3. Results and discussion

3.1. Characterization of surface-patterned TFC membranes

The FTIR spectra of the surfaces of the imprinted PES UFsubstrate with (patterned TFC membrane) and without (supportonly) the IP dense layer are compared in Fig. 2. We note that the IRspectrum of a non-patterned TFC membrane is not included in thefigure since it was identical to that of the patterned TFC mem-brane. For the patterned PES UF membrane, the strong absorptionband at 1760 cm�1 represents CQO stretching. The sharp absorp-tion peaks at 1151, 1244, and 1490 cm�1 were ascribed to thesymmetrical stretching vibration of the SO2 group, C–O–C vibra-tions, and C–S vibration, respectively [26]. All of these character-istic peaks are consistent with the chemical structure of PES.Because the calculated penetration depth of the ATR-FTIR spectro-scopy was about 1–5 mm in the wavelength region of interest, theIR spectrum of the TFC membrane surface shown in Fig. 2 isnecessarily a combination of the polyamide barrier layer and theunderlying PES support. The vibrational signatures associated withthe polyamide layer include the new peaks around 1240, 1290 and1320 cm�1 corresponding to stretching of aromatic amines I, IIand III, respectively, as well as those at�1540 and 1680 cm�1

representing stretching of amides I and II, respectively [26,27].The observed aromatic amines were possibly originated from

both absorbed unreacted MPD monomers in the membrane andunreacted amine groups bonded on the polyamide network.

1000 1500 2000

Wavenumber (cm-1)

Abs

orba

nce

(a.u

.)

patterned TFC membraneafter2hDI water ltration

patternedTFC membrane

1290 cm

1680 cm-1

-1-1

-1-1

1320 cm 1540 cm

1240 cm

1244 cm1760 cm-1

-1-1

-1

1490 cm1151 cm

1290 cm

1680 cm1320 cm 1540 cm

1151 cm

PESsupport

Fig. 2. Comparison of representative ATR-FTIR spectra for a patterned TFCmembrane (thin solid line) and a patterned PES UF support membrane (thick solidline). In addition, a spectrum of the patterned TFC membrane after 2 h of DI waterfiltration (dash-dot line) is presented. Representative spectra of non-patterned TFCmembranes are identical to that of the patterned TFC membrane in the figure (notshown here for clarity).

S.H. Maruf et al. / Journal of Membrane Science 452 (2014) 11–19 13

To leach out the physically absorbed MPD, DI water filtration onthe as-prepared patterned TFC membrane was carried out at2.76 MPa for 2 h. The FTIR spectrum of the membrane after thefiltration showed that the intensity of the amine peaks wasreduced appreciably, but still reasonably strong (Fig. 2). Thisconfirms that both contributions mentioned above were presentin the as-prepared TFC membranes. Nonetheless, FTIR measure-ments confirmed the formation of polyamide barrier layers onboth the patterned and non-patterned UF membrane used as asupport for the IP process described in Section 2.1.

Fig. 3 summarizes the morphological characterization of thepatterned and non-patterned PES UF support membrane and thecorresponding TFC membranes. A detailed characterization of theimprinted UF membranes was recently reported in Refs. [23, 24].Briefly, the non-patterned UF membrane (Fig. 3a) has a smoothsurface with an RMS roughness of less than 10 nm as determinedfrom the AFM surface profile shown in Fig. 3c [24]. After theimprinting process, periodic line-and-space grating patterns(Fig. 3d) with an average pattern height �100–120 nmwere presentin the patterned UF membrane (Fig. 3f). In addition, the imprintingprocess apparently increases the density of the porous PES supportas inferred by a decrease in the MWCO of the membrane from 15.4to 9.20 kg/mol [24]. However, the DI water flux was quite similar forthe non-patterned and patterned UF membranes most likely due tothe increased actual (versus projected) surface area after imprinting.

An IP process was used to form a polyamide layer on both non-patterned and patterned UF membranes. In Fig. 3b the surface ofthe non-patterned TFC membrane appeared very smooth, whichwas confirmed from the AFM measurement which showed an RMSroughness of �14 nm (Fig. 3c). This surface topography is notablydifferent from the much rougher, “ridge-and-valley” structure ofthe typical aromatic crosslinked polyamide films described in theliterature [28]. However, TFC membranes with relatively smoothcrosslinked polyamide barrier layers have also been reported[28,29]. It has been shown that the ridge-and-valley structuredevelops from the growth of the stiff aromatic polyamide chainsperpendicularly to the organic solvent/aqueous phase interface,

and becomes significant only when the overall barrier layer growsabove a certain thickness (over �100 nm) [30]. Here, the thicknessof the polyamide film on the non-patterned TFC membrane wasdetermined as only �40 nm via an AFM scan on the isolatedbarrier layer on a Si wafer (Fig. 3c) using previously describedtechniques [31]. This relatively low value of barrier layer thicknessmight be caused by a combination of air blowing during thesoaking of the MPD solution and the short reaction time (8 s) usedfor the IP process. Indeed, reaction time (t) during IP is a knownfactor affecting the thickness of the polyamide layer as the filmthickness increases as �t0.5 with a constant diffusion constantafter a dense film has developed, and a shorter reaction timewould be expected to yield a thinner polyamide film [9,14].

Fig. 3e shows the top surface morphology of the patterned TFCmembrane. From AFM measurements (Fig. 3f), the patternedsurfaces on the ridge and valley were even smoother than thatof the non-patterned TFC membrane at the same sub-50 nmlength scale. According to Pacheco et al. [28], a denser poroussubstrate used for the IP process produces a smoother polyamidelayer. As noted previously, the imprinted UF membrane wasindeed denser than the non-patterned UF membrane, as deter-mined from both cross-sectional SEM and MWCO measurements[23]. In fact, the starting PW UF membrane (MWCO �15.4 kg/mol[23]), approaches the lower limit of commercial UF membranesand may be even denser than the UF substrates commonly used incommercial TFC fabrication.

Note that the thickness of the barrier layer on the ridge andvalley of the pattern are as yet unresolved. Thus, the overlay of theTFC layer on the patterned UF membrane shown in Fig. 3f is aschematic representation rather than an actual profile, which isprovided only for enhanced perspective of the fabrication process.The isolation of the patterned barrier layer was unsuccessful dueto film rupture, most likely at the thinnest region on the ridge ofthe patterned support. However, it is important to emphasize thatTFC membranes with periodic surface patterns on the surface wereindeed fabricated successfully, and their filtration characteristicsare discussed in the following section.

Fig. 3. Morphological characterization of the patterned and non-patterned TFC membranes and UF support membranes. Representative top-surface SEM images of (a) thenon-patterned PES UF support membrane, (b) non-patterned TFC membrane, (d) patterned UF membrane support, and (e) patterned TFC membrane. Images (c) and (f) arerepresentative cross-sectional profiles for non-patterned and patterned TFC membranes obtained from AFM scans. Note that the overlay of the TFC layer on the patterned UFmembrane shown in panel (f) is a schematic representation rather than an actual profile, as it is challenging to isolate the barrier layer in this system.

S.H. Maruf et al. / Journal of Membrane Science 452 (2014) 11–1914

3.2. Filtration performance of non-patterned and patterned TFCmembranes

3.2.1. DI water filtrationFirst, we compare the DI water permeate flux of non-patterned

and patterned TFC membranes using the dead-end filtration setupas described in Section 2.2. This arrangement was used instead ofconventional cross-flow filtration because of the limited size of ourcurrent patterned UF membrane supports. As shown in Fig. 4,water flux for both membranes increased linearly with the appliedpressure, which is expected for membranes during DI waterfiltration. Note that at each pressure the membranes were com-pacted until the flux reached a steady-state value. The flux datareported in Fig. 4 are mean values over 2 h of filtration at thesteady-state condition. At higher pressure, slight deviation fromthe linear flux–pressure relationship was observed for bothmembranes, which is attributed to increased membrane compac-tion under higher pressure [32]. Overall, the water permeate fluxfor the patterned and non-patterned TFC membranes was similar,with a slightly higher flux observed for patterned membranes athigher pressures. The patterned TFC membrane may be somewhatmore compaction-resistant, particularly in the relatively highpressure region, possibly because of the densification of the UFmembrane support during the NIL process.

Subsequently, the permeate flux of the non-patterned andpatterned membranes were determined at 2.75 MPa over 8 h offiltration and compared with that of several commercial RO andNF membranes using the same filtration conditions. The permeatefluxes after compaction are summarized in Fig. 5. The fourcommercial TFC RO membranes, XLE-440 (DOW Filmtec), CPA 3(Hydranautics), ACM 2 (Trisep), and TM-700 (Toray), share thesame tri-layer configuration as the TFC membranes fabricated in

this study. Although the exact chemistry of the polyamide layerlikely differs for the different commercial membranes, they are allbased on the MPD–TMC IP process (Fig. 1) [33]. The two NFmembranes, NF 270 (Hydranautics) and ES-10 (Nitto Denko), alsohave similar aromatic polyamide structures, but typically havemuch higher water permeate flux and lower ion rejection (parti-cularly for monovalent ions) than the RO membranes [34]. Furtherinformation regarding the commercial membranes is provided inthe supporting document (Table S1). TFC RO membranes are usedprimarily for water desalination, while NF membranes are used forthe removal of mineral scale, biological matter, colloidal particlesand insoluble organic constituents from water feed streams [35].

As shown in Fig. 5, both patterned and non-patterned TFCmembranes showed higher pure water permeation (PWP) flux ascompared to most of the RO membranes (except XLE-440) andlower flux than the two NF membranes. Since the water permeateflux of a TFC membrane depends on the properties of the barrierlayer (chemistry, crosslinking density, thickness, roughness) aswell as those of the substrate(s) [9,36], further interpretation ofthe differences in Fig. 5 is beyond the scope of this study. However,the data represented in Fig. 5 confirm that the IP procedure thatwas employed for the non-patterned and patterned UF substratesachieved water permeate flux values that are comparable withthose of typical RO and NF membranes that use similar barrierlayer chemistry.

3.2.2. Filtration of NaCl and CaCl2 solutionsFig. 6 presents the permeation results for both patterned and

non-patterned TFC membranes during filtration over 3 h of theNaCl solution. For each membrane, permeate flux (Fig. 6a) and

0.0 0.5 1.0 1.5 2.0 2.5 3.00

10

20

30

40

50

Perm

eate

ux

(Lm

-2h-1

)

Transmembrane pressure (MPa)

Fig. 4. DI water permeate flux as a function of transmembrane pressure for thenon-patterned (solid symbols) and patterned TFC membranes (open symbols).

0

20

40

60

80

100

ES-10

NF 270

NonPatterned TFC

CPA-3

TM-700

ACM-2

Patterned TFC

Perm

eate

flux

(Lm

-2h-1

)

XLE-440

Fig. 5. Comparison of DI water permeate flux values of the non-patterned andpatterned TFC membranes with those of several commercial NF and RO membranes.

20

24

28

32

stirredcondition

Perm

eate

flux

(Lm

-2h-1

)

unstirredcondition

0 40 80 120 160 200 240

90

92

94

96

98

100

unstirredcondition O

bser

ved

reje

ctio

n R

o (%

)

Intri

nsic

reje

ctio

n R

i (%

)

Time (min)

0 40 80 120 160 200 240

Time (min)

stirredcondition

65

70

75

80

85

90

95

100

Fig. 6. Filtration results for the non-patterned and patterned TFC membranes withan aqueous 1 g/L NaCl solution (pH¼7.1, TMP¼2.75 MPa and T¼¼2570.5 1C). (a) Permeate flux and (b) rejection. In (b) the intrinsic rejection is presentedfor non-patterned (filled symbols) and patterned (open symbols) membranes atunstirred (triangle) and stirred (diamond) conditions; the observed rejection isshown for the non-patterned (filled symbols) and patterned (open symbols) TFCmembranes at the unstirred (squares) and stirred (circles) condition.

S.H. Maruf et al. / Journal of Membrane Science 452 (2014) 11–19 15

observed salt rejection (Fig. 6b) gradually decreased with filtrationtime while the intrinsic salt rejection remained essentially con-stant. Specifically, the initial flux for each membrane was�29 L m�2 h�1. For the unstirred condition the flux decreasedover time by �21% for the non-patterned membrane and �22%for the patterned one. With stirring, flux reduction was �12% and�8% for the non-patterned and patterned membrane, respectively.Similar to the permeate flux, the observed salt rejection decreased�22% for both non-patterned and patterned membranes in theunstirred condition, and for the stirred condition �13% and �9%for the non-patterned and patterned membranes, respectively.These time-dependent reductions in permeate flux and observedsolute rejection are due to both the increased feed salinity due towater removal, and the concentration polarization at themembrane-solution interface [37].

Initial observed salt rejection for all of the samples was �90%with relative variability (Fig. 6b). Initial permeate was collectedafter 25 min of filtration to allow the system to stabilize. Theobserved salt rejection, Ro, was calculated from the bulk concen-tration of salt in the permeate (Cp) and feed (Cf) solutions,according to Ro (%)¼(1�Cp/Cf). In any RO/NF filtration systemthe observed salt rejection does not represent the true membraneseparation capability due to concentration polarization. Intrinsicsalt rejection, Ri¼(1�Cp/Cm), based on the boundary layer soluteconcentration (Cm) is normally higher than Ro because Cm is higherthan Cf [37,38]. Cm can be determined from a mass balance overthe boundary layer according to

Jv ¼Di

δiln

ðCi; m�Ci; pÞðCi; b�Ci; pÞ

� �¼ ki ln

ðCi; m�Ci; pÞðCi; b�Ci; pÞ

� �ð1Þ

where Jv is the total volumetric flux, Di is the diffusivity of solute iin water, δi is the thickness of the boundary layer, Ci,m is theconcentration in solution at the feed-membrane interface, Ci,p isthe permeate concentration, Ci,b is the bulk concentration, and ki isthe mass-transfer coefficient [39]. The mass-transfer coefficientwas estimated using the osmotic pressure model described bySutzkover et al. [40], which assumes no composition dependencefor water permeance through the membrane. Further informationregarding calculation of the mass-transfer coefficient and bound-ary layer concentration is available in the Supporting information.Accordingly, the Ri for the membrane samples were in the range of98–99% over the period of the filtration time.

Since the UF support membrane had an observed rejection of�7% for NaCl with a permeate flux of �621 L m�2 h�1, the highNaCl rejection (and correspondingly lower flux, 38–24 L m�2 h�1)for both the TFC membrane types verifies the successful formationof dense and continuous polyamide barrier layers. For comparison,MPD/TMC-based TFC RO membranes have been reported to have aNaCl rejection between 65% and 99% [25,41–45], while commercialmembranes can attain over 99.5% rejection [46]. The variability inthese rejection values is caused by the barrier layer properties,operating conditions, and additional membrane modifications[47]. In addition, polyamide-based TFC NF membranes have areported NaCl rejection range between 60% and 80% [48,49]. Thus,the TFC membranes prepared for this study had a NaCl selectivityless than that of commercial RO membranes, and higher thantypical NF membranes. These values are consistent with the lowerDI water flux comparisons presented in Fig. 5 such that thepatterned and non-patterned TFC membranes can be regarded as“tight NF” or “loose RO” membranes.

The practical water (A) and salt permeances (B) [50] for the saltfiltration were calculated using the relationships A¼ Jw/(ΔP�Δπ)and B¼CpJw/(Cm�Cp). Results, which are detailed in the support-ing information (Fig. S2), indicated that the water permeancedecreased significantly while the salt permeance remained rela-tively constant over the filtration period. Concentration build-up at

the membrane barrier layer increases the osmotic pressure, whichin turn decreases the water permeance by reducing the effectiveTMP [40]. For unstirred filtration conditions concentration polar-ization is more severe because the boundary layer, over whichdiffusion returns the solute to the bulk solution, is larger [51]. Backdiffusion is enhanced by advection due to stirring (the boundarylayer moves closer to the membrane surface), which leads to lessconcentration polarization. This explanation is consistent with theeffects shown in Fig. 6.

Although the flux and salt rejections appear quite similar forthe non-patterned and patterned TFC membranes for the unstirredcondition, there are small but important differences when stirringwas applied. Here, the flux and salt rejection for the non-patternedmembranes evidenced a more pronounced decrease over the 3 hfiltration period as compared to their patterned counterparts. Thisbehavior strongly suggests that the presence of the surfacepatterns on the TFC membrane changes the mass transfer in thevicinity of the membrane surface, i.e., enhances back diffusion(transport) to the bulk. The presence of the surface patterns islikely to modify the flow profile and local streamlines of the feedsolution in the proximity of the patterns, producing localizedturbulence and/or large shear stresses [22]. This topography-induced antifouling effect was also observed in our previous studywith patterned PES UF membranes [23,24]. Note that the second-ary flows depend on the Reynolds number (Re) of the tangentialflow over the membrane, and can be much more extensive athigher Re values.

Filtration experiments were also performed with a modeldivalent salt, CaCl2, using the same protocols. Initial salt rejectionfor all of the samples was �97% with somewhat higher variabilitythan that with NaCl. Permeate flux for CaCl2 filtration was lower

12

14

16

18

unstirredcondition

stirredcondition

unstirredcondition

stirredcondition

Perm

eate

flux

(Lm

-2h-1

)

90

92

94

96

98

100

Intri

nsic

reje

ctio

n R

i (%

)

Obs

erve

d re

ject

ion

Ro (

%)

65

70

75

80

85

90

95

100

0 40 80 120 160 200 240

Time (min)

0 40 80 120 160 200 240

Time (min)

Fig. 7. Filtration results for the non-patterned and patterned TFC membranes withan aqueous 1 g/L CaCl2 solution (pH¼7.2, TMP¼2.75 MPa and T¼2570.5 1C).(a) Permeate flux and (b) rejection. In (b) the intrinsic rejection is presented fornon-patterned (filled symbols) and patterned (open symbols) membranes atunstirred (triangle) and stirred (diamond) conditions; the observed rejection isshown for the non-patterned (filled symbols) and patterned (open symbols) TFCmembranes at the unstirred (squares) and stirred (circles) condition.

S.H. Maruf et al. / Journal of Membrane Science 452 (2014) 11–1916

than that for NaCl due to higher osmotic pressure and greaterconcentration polarization from the interaction between thenegatively charged TFC membrane and the divalent Ca2þ salt[52]. As was the case for NaCl (Fig. 6), both permeate flux (Fig. 7a)and observed salt rejections (Fig. 7b) for all of the samplesdecreased from the onset of the experiment while intrinsic saltrejection remained relatively constant. Under unstirred conditions,the flux decreased by �26% and �28% for the non-patterned andthe patterned TFC membranes, respectively. The salt rejectionevidenced similar decreases of �26% for the non-patternedmembranes and �29% for the patterned membranes. With stir-ring, the non-patterned and patterned membranes had fluxdecreases of �13% and �9% and salt rejection declines of 13%and �10%, respectively. Hence, results from filtration of bothmonovalent and divalent salt solutions are consistent and implythat the better performance of the patterned membranes is due toreduced concentration polarization arising from surface-pattern-induced hydrodynamic effects.

3.2.3. Scaling with CaSO4 solutionsScaling experiments with a 1 g/L CaSO4 (gypsum) solution were

performed on both non-patterned and patterned TFC membranesusing dead-end filtration system but only with stirring. As shownin Fig. 8a, the initial permeate flux of both membranes was�25 L m�2 h�1, and subsequently decreased in two distinct stages.After 6–7 h of filtration, the flux decreased �9.7% and�6.7% for thenon-patterned and patterned TFC membranes, respectively, whichis primarily due to the increasing osmotic pressure of the feed andthe associated concentration polarization effect as observed inFigs. 6 and 7. Subsequently, a steeper flux decline was observedfor each membrane type whereby after 24 h the initial values of fluxhad declined by �47% and �40% of for the non-patterned andpatterned membranes, respectively. The second stage of flux declineis attributed to the scaling of gypsum on the membrane surfaces[53,54]. From Fig. 8a, it appears that the onset of scaling on thepatterned TFC membrane (�6 h) was somewhat more rapid thanfor the non-patterned membrane (�7.5 h). This induction timeindicates the point at which CaSO4 reached its solubility limit inthe feed solution such that precipitation on the membrane surface isinitiated. Continued precipitation initiates scaling which leads to amarked decrease in the permeate flux as less surface area isaccessible for the permeate [55,56]. Because of the higher permeateflux of the patterned membrane during the initial stage, over-saturation of the CaSO4 was likely reached sooner.

Fig. 8b and c presents representative SEM images of gypsum onthe non-patterned and patterned TFC membranes after the 24 hfiltration period. Note that crystallization of the CaSO4 duringfiltration can occur via both homogeneous nucleation in the bulk

feed solution and heterogeneous nucleation on the membranesurface. The latter mechanism often produces distinctive plate-likecrystal forms [57]. The morphology of the gypsum crystals (Fig. 8band c) was bulk-like on both membranes with very few needle-like(and no clear plate-like) crystallites in Fig. 8b, which suggests thatthe scaling was dominated by bulk crystallization of the gypsum[58,59]. Statistical analysis of the relatively constant flux valuesobtained over the final hour of the tests indicates that the flux forthe patterned membranes is significantly higher than that for thenon-patterned membranes (one-tailed t-test; po0.05). This differ-ence in permeate flux is consistent with the sparser distribution ofgypsum crystals observed on the surface of the patterned mem-brane (Fig. 8b and c), presumably due to the aforementionedpattern-induced hydrodynamic effects. In addition, we note thatthe crystals formed during filtration with the patterned membraneswere less adherent as judged by the fact that they were more easilyremoved during water rinsing at the completion of the experiment.Overall, the patterned membrane provided a lower cake resistanceper unit mass (surface area) of deposition. Although additional workis clearly required to substantiate the improved fouling character-istics of the patterned membranes, the initial results described hereare quite encouraging.

4. Conclusion

In this study, we demonstrate for the first time fabrication of asubmicron-patterned TFC membrane via interfacial polymeriza-tion on a nanoimprinted UF membrane without adversely affect-ing permselective properties. Thin crosslinked aromatic polyamidebarrier layer films atop a patterned support membrane werecharacterized by FTIR spectroscopy and electrolyte versus waterpermselectivity determined from filtration experiments. The pat-terned TFC membrane exhibited water permeance and salt rejec-tion comparable to that of commercial TFC RO membranes.Compared with their non-patterned counterparts, the patternedTFC membranes demonstrated higher flux and rejection valueswhen convection was present as a result of stirring. Scalingexperiments revealed that gypsum distribution on the surface ofthe patterned membranes was more widely scattered in compar-ison to that on the non-patterned membranes. Overall, theseresults suggest that the surface patterns induced hydrodynamicsecondary flows at the membrane-feed interface which wereeffective in decreasing concentration polarization as well as inreducing scaling effects. Together with our recent work onpatterning UF membranes, this study provides important perspec-tive regarding the promise of surface patterning as an effectivealternative to chemical modification for fouling mitigation forliquid-based separation membranes.

Fig. 8. (a) Permeate flux as function of filtration time for the stirred condition for non-patterned (solid squares) and patterned (open squares) TFC membranes in an aqueous1 g/L CaSO4 solution (pH¼7.4, TMP¼2.75 MPa and T¼2570.5 1C). Representative SEM images of the non-patterned (b) and patterned (c) TFC membrane surface after 24 hof CaSO4 filtration.

S.H. Maruf et al. / Journal of Membrane Science 452 (2014) 11–19 17

Acknowledgment

The authors acknowledge the funding support from the NationalScience Foundation under Grant no. CBET-1031785. The authorsgratefully acknowledge the National Science Foundation (NSF)Industry/University Cooperative Research Center for MembraneScience, Engineering and Technology (MAST) at the University ofColorado Boulder (NSF Award IIP 1034720) for stimulating broad-based and impactful research. Additionally, we acknowledge theOffice of the Vice Chancellor for Research at CU-Boulder for theaward of an Innovative Seed Grant in support of this research.

Appendix A. Supporting information

Supplementary data associated with this article can be found inthe online version at http://dx.doi.org/10.1016/j.memsci.2013.10.017.

References

[1] P.W. Morgan, Condensation Polymers: By Interfacial and Solution Methods,Interscience Publishers, New York, 1965.

[2] R.J. Petersen, Composite reverse osmosis and nanofiltration membranes,J. Membr. Sci. 83 (1993) 81–150.

[3] R. Baker, Membrane Technology and Applications, Wiley, 2004.[4] C. Fritzmann, J. Lowenberg, T. Wintgens, T. Melin, State-of-the-art of reverse

osmosis desalination, Desalination 216 (2007) 1–76.[5] K.P. Lee, T.C. Arnot, D. Mattia, A review of reverse osmosis membrane materials

for desalination – development to date and future potential, J. Membr. Sci. 370(2011) 1–22.

[6] N. Hilal, H. Al-Zoubi, N.A. Darwish, A.W. Mohamma, M.Abu Arabi, A compre-hensive review of nanofiltration membranes: treatment, pretreatment, mod-elling, and atomic force microscopy, Desalination 170 (2004) 281–308.

[7] S. Belfer, Y. Purinson, R. Fainshtein, Y. Radchenko, O. Kedem, Surface modifica-tion of commercial composite polyamide reverse osmosis membranes,J. Membr. Sci. 139 (1998) 175–181.

[8] Y. Song, P. Sun, L.L. Henry, B. Sun, Mechanisms of structure and performancecontrolled thin film composite membrane formation via interfacial polymer-ization process, J. Membr. Sci. 251 (2005) 67–79.

[9] A.K. Ghosh, E.M.V. Hoek, Impacts of support membrane structure andchemistry on polyamide-polysulfone interfacial composite membranes,J. Membr. Sci. 336 (2009) 140–148.

[10] P.S. Singh, S.V. Joshi, J.J. Trivedi, C.V. Devmurari, A.P. Rao, P.K. Ghosh, Probingthe structural variations of thin film composite RO membranes obtained bycoating polyamide over polysulfone membranes of different pore dimensions,J. Membr. Sci. 278 (2006) 19–25.

[11] P. Xu, J.E. Drewes, T.U. Kim, C. Bellona, G. Amy, Effect of membrane fouling ontransport of organic contaminants in NF/RO membrane applications, J. Membr.Sci. 279 (2006) 165–175.

[12] D. Rana, T. Matsuura, Surface modifications for antifouling membranes, Chem.Rev. 110 (2010) 2448–2471.

[13] V. Freger, Nanoscale heterogeneity of polyamide membranes formed byinterfacial polymerization, Langmuir 19 (2003) 4791–4797.

[14] V. Freger, Kinetics of film formation by interfacial polycondensation, Langmuir21 (2005) 1884–1894.

[15] C.S. Chen, M. Mrksich, S. Huang, G.M. Whitesides, D.E. Ingber, Geometriccontrol of cell life and death, Science 276 (1997) 1425–1428.

[16] Y. Ding, J. Sun, H.W. Ro, Z. Wang, J. Zhou, N.J. Lin, M.T. Cicerone, C.L. Soles,S. Lin-Gibson, Thermodynamic underpinnings of cell alignment on controlledtopographies, Adv. Mater. 23 (2010) 421–425.

[17] M.L. Carman, T.G. Estes, A.W. Feinberg, J.F. Schumacher, W. Wilkerson, L.H. Wilson, M.E. Callow, J.A. Callow, A.B. Brennan, Engineered antifoulingmicrotopographies-correlating wettability with cell attachment, Biofouling22 (2006) 11–21.

[18] L. Vogelaar, J.N. Barsema, C.J.M. van Rijn, W. Nijdam, M. Wessling, Phaseseparation micromolding – PSμM, Adv. Mater. 15 (2003) 1385–1389.

[19] P.n.Z. Culfaz, E. Rolevink, C. van Rijn, R.G.H. Lammertink, M. Wessling,Microstructured hollow fibers for ultrafiltration, J. Membr. Sci. 347 (2010)32–41.

[20] P.Z. Culfaz, S. Buetehorn, L. Utiu, M. Kueppers, B. Bluemich, T. Melin,M. Wessling, R.G.H. Lammertink, Fouling behavior of microstructured hollowfiber membranes in dead-end filtrations: critical flux determination and NMRimaging of particle deposition, Langmuir 27 (2011) 1643–1652.

[21] Y.J. Won, J. Lee, D.C. Choi, H.R. Chae, I. Kim, C.H. Lee, I.C. Kim, Preparation andapplication of patterned membranes for wastewater treatment, Environ. Sci.Technol. 46 (2012) 11021–11027.

[22] Y.K. Lee, Y.J. Won, J.H. Yoo, K.H. Ahn, C.H. Lee, Flow analysis and fouling on thepatterned membrane surface, J. Membr. Sci. 427 (2013) 320–325.

[23] S.H. Maruf, M. Rickman, L. Wang, J. Mersch, I.v.,A.R. Greenberg, J. Pellegrino,Y. Ding, Influence of sub-micron surface patterns on the deposition of modelproteins during active filtration, J. Membr. Sci. 444 (2013) 420–428.

[24] S.H. Maruf, L. Wang, A.R. Greenberg, J. Pellegrino, Y. Ding, Use of nanoim-printed surface patterns to mitigate colloidal deposition on ultrafiltrationmembranes, J. Membr. Sci. 428 (2013) 598–607.

[25] A.K. Ghosh, B.H. Jeong, X. Huang, E.M.V. Hoek, Impacts of reaction and curingconditions on polyamide composite reverse osmosis membrane properties,J. Membr. Sci. 311 (2008) 34–45.

[26] J. Coates, Interpretation of infrared spectra, a practical approach, in: R.A. Meyers (Ed.), Encyclopedia of Analytical Chemistry, John Wiley and SonsLtd., UK, Chicester, 2000.

[27] R.M. Silverstein, F.X. Webster, Spectrometric Identification of Organic Com-pounds, Wiley, 1998.

[28] F.A. Pacheco, I. Pinnau, M. Reinhard, J.O. Leckie, Characterization of isolatedpolyamide thin films of RO and NF membranes using novel TEM techniques,J. Membr. Sci. 358 (2010) 51–59.

[29] Y. Song, F. Liu, B. Sun, Preparation, characterization, and application of thinfilm composite nanofiltration membranes, J. Appl. Polym. Sci. 95 (2005)1251–1261.

[30] V. Enkelmann, G. Wegner, Mechanism of interfacial polycondensation and thedirect synthesis of stable polyamide membranes, Makromol. Chem. 177 (1976)3177–3189.

[31] S.H. Maruf, D.U. Ahn, A.R. Greenberg, Y. Ding, Glass transition behaviors ofinterfacially polymerized polyamide barrier layers on thin film compositemembranes via nano-thermal analysis, Polymer 52 (2011) 2643–2649.

[32] R. Rautenbach, T. Linn, L. Eilers, Treatment of severely contaminated wastewater by a combination of RO, high-pressure RO and NF-potential and limitsof the process, J. Membr. Sci. 174 (2000) 231–241.

[33] T.U. Kim,Transport of Organic Micropollutants Through Nanofiltration (NF)and Reverse Osmosis (RO) Membranes: Mechanisms, Modeling, and Applica-tions, ProQuest, University of Colorado at Boulder, 2006.

[34] V. Freger, Swelling and morphology of the skin layer of polyamide compositemembranes: an atomic force microscopy study, Environ. Sci. Technol 38(2004) 3168–3175.

[35] Y. Lin, Effects of physicochemical properties of nanofiltration membranes on therejection of small organic DBP precursors, J. Environ. Eng. 139 (2013) 127–136.

[36] L.F. Greenlee, D.F. Lawler, B.D. Freeman, B. Marrot, P. Moulin, Reverse osmosisdesalination: water sources, technology, and today's challenges, Water Res. 43(2009) 2317–2348.

[37] S.S. Sablani, M.F.A. Goosen, R. Al-Belushi, M. Wilf, Concentration polarizationin ultrafiltration and reverse osmosis: a critical review, Desalination 141(2001) 269–289.

[38] E.M.V. Hoek, M. Elimelech, Cake-enhanced concentration polarization: a newfouling mechanism for salt-rejecting membranes, Environ. Sci. Technol. 37(2003) 5581–5588.

[39] Z.V.P. Murthy, S.K. Gupta, Estimation of mass transfer coefficient using acombined nonlinear membrane transport and film theory model, Desalination109 (1997) 39–49.

[40] I. Sutzkover, D. Hasson, R. Semiat, Simple technique for measuring theconcentration polarization level in a reverse osmosis system, Desalination131 (2000) 117–127.

[41] M. Hirose, H. Ito, Y. Kamiyama, Effect of skin layer surface structures on theflux behaviour of RO membranes, J. Membr. Sci. 121 (1996) 209–215.

[42] C. Kong, M. Kanezashi, T. Yamomoto, T. Shintani, T. Tsuru, Controlled synthesisof high performance polyamide membrane with thin dense layer for waterdesalination, J. Membr. Sci. 362 (2010) 76–80.

[43] A. Prakash Rao, S.V. Joshi, J.J. Trivedi, C.V. Devmurari, V.J. Shah, Structure-performance correlation of polyamide thin film composite membranes: effectof coating conditions on film formation, J. Membr. Sci. 211 (2003) 13–24.

[44] J. Jegal, S.G. Min, K.H. Lee, Factors affecting the interfacial polymerization ofpolyamide active layers for the formation of polyamide composite mem-branes, J. Appl. Polym. Sci. 86 (2002) 2781–2787.

[45] X. Lu, X. Bian, L. Shi, Preparation and characterization of NF compositemembrane, J. Membr. Sci. 210 (2002) 3–11.

[46] T.U. Kim, J.E. Drewes, R.S. Summers, G.L. Amy, Solute transport model for traceorganic neutral and charged compounds through nanofiltration and reverseosmosis membranes, Water Res. 41 (2007) 3977–3988.

[47] W. Xie, G.M. Geise, B.D. Freeman, H.-S. Lee, G. Byun, J.E. McGrath, Polyamideinterfacial composite membranes prepared from m-phenylene diamine,trimesoyl chloride and a new disulfonated diamine, J. Membr. Sci. 403-404(2012) 152–161.

[48] G. Hagmeyer, R. Gimbel, Modelling the rejection of nanofiltration membranesusing zeta potential measurements, Sep. Purif. Technol. 15 (1999) 19–30.

[49] J.M. Skluzacek, M.I. Tejedor, M.A. Anderson, NaCl rejection by an inorganicnanofiltration membrane in relation to its central pore potential, J. Membr. Sci.289 (2007) 32–39.

[50] J.G. Wijmans, The role of permeant molar volume in the solution-diffusionmodel transport equations, J. Membr. Sci. 237 (2004) 39–50.

[51] W.R. Bowen, F. Jenner, Theoretical descriptions of membrane filtration ofcolloids and fine particles: An assessment and review, Adv. Colloid InterfaceSci. 56 (1995) 141–200.

[52] S. Bason, V. Freger, Phenomenological analysis of transport of mono anddivalent ions in nanofiltration, J. Membr. Sci. 360 (2010) 389–396.

[53] D. Hasson, A. Drak, R. Semiat, Inception of CaSO4 scaling on RO membranes atvarious water recovery levels, Desalination 139 (2001) 73–81.

S.H. Maruf et al. / Journal of Membrane Science 452 (2014) 11–1918

[54] S. Lee, J. Kim, C.H. Lee, Analysis of CaSO4 scale formation mechanism in variousnanofiltration modules, J. Membr. Sci. 163 (1999) 63–74.

[55] J. Gilron, D. Hasson, Calcium sulphate fouling of reverse osmosis membranes:flux decline mechanism, Chem. Eng. Sci. 42 (1987) 2351–2360.

[56] A.G. Pervov, Scale formation prognosis and cleaning procedure schedules inreverse osmosis systems operation, Desalination 83 (1991) 77–118.

[57] A. Antony, J.H. Low, S. Gray, A.E. Childress, P. Le-Clech, G. Leslie, Scaleformation and control in high pressure membrane water treatment systems:a review, J. Membr. Sci. 383 (2001) 1–16.

[58] Z. Amjad, Applications of antiscalants to control calcium sulfate scaling inreverse osmosis systems, Desalination 54 (1985) 263–276.

[59] W.-Y. Shih, A. Rahardianto, R.-W. Lee, Y. Cohen, Morphometric characterizationof calcium sulfate dihydrate (gypsum) scale on reverse osmosis membranes, J.Membr. Sci. 252 (2005) 253–263.

S.H. Maruf et al. / Journal of Membrane Science 452 (2014) 11–19 19