Contents lists available at ScienceDirect

Journal of Membrane Science

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

Effect of long-term operation on membrane surface characteristics andperformance in membrane distillation

A.L. McGaughey, R.D. Gustafson, A.E. Childress⁎

University of Southern California, Sonny Astani Department of Civil and Environmental Engineering, 920 Downey Way, Los Angeles, CA 90089-0137, USA

A R T I C L E I N F O

Keywords:Membrane distillationHydrophobicitySurface roughnessSurface morphologyLong-term performance

A B S T R A C T

In this study, significant changes to surface morphology and decreased surface hydrophobicity were observed onboth the feed and distillate sides of membrane distillation membranes after 100 days of operation. Contact anglesdecreased by 56% and 26% on the feed and distillate sides, respectively. Surface roughness also decreased by92% and 57% on the feed and distillate sides, respectively. Moderate morphological changes were also observedafter 20 days of operation. While decreased hydrophobicity and surface roughness on the feed side were asso-ciated with fouling/scaling deposits and not changes to the actual membrane surface, decreased hydrophobicityand surface roughness on the distillate side indicated changes to the actual membrane surface. Often, membranehydrophobicity is assumed to be recoverable if foulants can be removed; however, if membrane hydrophobicitydecreases due to physical changes in the membrane surface, hydrophobicity may not be fully recoverable andmembrane lifetime may be reduced. Despite significant reductions in feed-side hydrophobicity, distillate con-ductivity remained low, indicating that other membrane characteristics, such as distillate-side and internal orpore wall hydrophobicity, may play an important role in maintaining rejection during long-term operation.

1. Introduction and background

Membrane distillation (MD) is a thermally driven water treatmentprocess whereby a warmer feed stream flows on one side of a hydro-phobic, microporous membrane and a cooler distillate stream flows onthe other side. The temperature difference across the membrane createsa vapor pressure difference that causes liquid water to evaporate fromthe feed stream, pass through the membrane pores, and condense intothe distillate stream. MD has the potential to produce high qualitywater while operating at low temperature differences that are achiev-able using waste heat and/or renewable energy sources [1–3]. In directcontact MD (DCMD), the simplest configuration of MD, both liquidstreams directly contact the membrane surfaces. Non-volatile speciesare prevented from passing through the membrane under normal op-eration if membrane hydrophobicity is sufficient to prevent liquid fromwetting the membrane pores [4,5]. Thus, membrane hydrophobicity isa key parameter for assuring rejection of non-volatile species.

MD membranes are typically polymeric, often composed of poly-tetrafluoroethylene (PTFE), polyvinylidene fluoride (PVDF), and/orpolypropylene (PP) [5]. Hydrophobicity is usually highest for PTFEmembranes [4,5]. PTFE membranes are commonly characterized by asurface structure consisting of strands of polymer material called fibrils,which connect at junctions or “nodes” [6]. In general, the surface

morphology of a PTFE membrane can be characterized by surfaceroughness and by the size, shape, and form of the fibrils, nodes, andpores that make up the surface [6,7]. Membrane fibrils also contributeto surface roughness [8]. Surface roughness, in turn, affects surfacehydrophobicity [9] and membrane performance.

Because the vapor pressure driving force in MD does not decreasesignificantly for feed streams with higher total dissolved solids (TDS)concentrations, MD membranes are being considered for a range ofmoderate- to high-salinity applications [3,10–13]. Some of these ap-plications have complex solution chemistries containing organic spe-cies; for this reason, they are often considered high-fouling applica-tions. In these cases, the presence of fouling and scaling can obfuscateobservation of the actual membrane surface; however, characterizationof the membrane surface may not be critical as the feed solution wouldmore likely interact with the foulant/scalant layer than with themembrane itself [14]. On the other hand, in low-fouling applications,characterizing changes in membrane surface morphology with time iscritical for understanding long-term performance. Low-fouling appli-cations may be defined as applications with feed streams that consistprimarily of inorganic contaminants that are not present at high enoughconcentrations to exceed their solubility and form scale; these appli-cations include effluent polishing (removal of targeted contaminantsthat pass through other treatment processes) and re-concentration of

http://dx.doi.org/10.1016/j.memsci.2017.08.040Received 31 May 2017; Received in revised form 15 August 2017; Accepted 16 August 2017

⁎ Corresponding author.E-mail address: amyec@usc.edu (A.E. Childress).

Journal of Membrane Science 543 (2017) 143–150

Available online 18 August 20170376-7388/ © 2017 Elsevier B.V. All rights reserved.

MARK

draw solutions used in forward osmosis systems [10,15,16]. To achievesustainable, long-term performance in these applications, membranehydrophobicity must be maintained; however, there are few long-termstudies in the literature to support or dispute the frequent, implicitassumption that hydrophobicity remains constant in the absence offouling.

Most bench-scale MD studies have durations on the order of minutesto hours and focus on short-term flux or initial membrane performance.Few bench-scale studies with low-fouling feed solutions have beencarried out over long time periods [17–25]. Of these, only a smallsubset [21,23,26] mention possible changes in membrane surfacecharacteristics that occur during operation and suggest a relationshipbetween changes in membrane surface morphology and performance.No studies have systematically evaluated relatively unfouled MDmembrane surfaces after exposure to low-fouling solutions.

In pilot-scale studies, operation over longer time periods is morecommon; however, due to the complexities of extracting membranesfrom modules to perform autopsies, consideration of the relationshipbetween long-term performance and membrane surface characteristicsis even less common than in bench-scale studies, with a few exceptions.In 2011, Guillen-Burrieza et al. [27] studied the performance and ef-ficiency of a pilot-scale solar-powered MD system treating marine saltsolutions of 1 and 35 g/L TDS with commercial PTFE membranes. For aperiod of four months, the system was operated during daylight hoursand shut down overnight. Distillate conductivity increased due tomembrane wetting and salt passage. An integrity test showed that themembrane lacked holes, suggesting to the authors that wetting occurreddue to a decrease in membrane hydrophobicity, although surfacecharacterization was not able to be performed to verify this [27]. In2014, Guillen-Burrieza et al. [28] examined membrane performanceand the efficacy of cleaning procedures for a PTFE-PP compositemembrane used in a pilot-scale solar MD system to treat a 35 g/Lmarine salt solution. The system was operated during the day and shutdown overnight for a total of 17 days. Pore wetting occurred, shown byincreasing distillate conductivity. Various membrane surface char-acteristics, including contact angle and surface morphology, werecompared for the virgin and used membranes. Scaling on the feed sideof the membrane was associated with fibril damage [28]. Cleaningprocedures were applied at the laboratory scale to samples of the usedmembrane. Contact angle recovery varied with cleaning solution butdid not reach virgin membrane values; it was suggested that membranestructural damage due to operation and/or cleaning procedures hadoccurred [28]. To the best of the authors’ knowledge, these are the onlystudies that have examined how membrane surface characteristics areaffected by long-term operation and in turn, affect long-term salt re-jection.

No long-term studies have yet examined the role of distillate-sidesurface characteristics in MD performance. Studies on composite hy-drophobic-hydrophilic MD membranes have demonstrated good short-term performance and durability (e.g., [29–33]), and improved per-formance has been observed for an MD membrane with hydrophilicfeed-side and hydrophobic distillate-side surfaces [32]. This may in-dicate that distillate-side characteristics or internal material char-acteristics are relevant to the pore wetting process; however, this pos-sibility has not yet been fully explored in the literature. Internal or porewall hydrophobicity of aquaporin nanopores and carbon nanotubes haspreviously been studied theoretically via molecular dynamics simula-tions (e.g., [34,35]). The concept has also been mentioned for in-homogeneous fuel cell materials [36,37]. However, internal hydro-phobicity of porous materials is not yet experimentally accessible [37].Guillen-Burrieza et al. [28] found that membranes with a thin PTFElayer on a less hydrophobic PP support were significantly wetted duringlong-term use despite a feed-side contact angle greater than 90°. Theauthors suggested that this may have been due to active layer com-pression facilitating liquid “bridging” through membrane pores [28].Thus, simply having a hydrophobic layer facing the feed solution may

not be enough to ensure that membrane pores remain dry, especially forlong-term operation.

The overall objective of this study is to evaluate changes in mem-brane surface hydrophobicity, surface morphology, and performanceafter long-term operation. Only saline feed solutions are used in orderto clearly identify the effect of moderate- to high-salinity solutions onlong-term MD membrane performance. Furthermore, the distillate sideof the membrane, which has been largely overlooked in previous stu-dies, is characterized after long-term operation. The results of this studyare used to better understand the impact of long-term operation onmembrane surface characteristics, the potential relevance of distillate-side membrane characteristics to long-term salt rejection, and the im-plications of these phenomena for MD membrane design.

2. Materials and methods

2.1. MD membranes

Commercial flat-sheet expanded polytetrafluoroethylene (PTFE)microfiltration membranes (Clarcor, Franklin, TN) were used for alllong-term testing. The membranes are hydrophobic, single-layer, andsymmetric. The average pore size of the membrane surface was found tobe 0.3±0.17 µm, based on analysis of ten field emission scanningelectron microscope (FESEM) images using ImageJ software (version1.49, National Institutes of Health, Bethesda, MD). Nearly 90% of thepores were less than 0.5 µm in diameter and 61% of the pores haddiameters of 0.2± 0.06 µm.

2.2. Performance characterization

Membrane performance was tested in a bench-scale DCMD systemdesigned for continuous long-term operation. A schematic of the systemis shown in Fig. 1. To eliminate corrosion and rust in the system duringlong-term experiments, as has been reported in other long-term MDstudies (e.g., [28,38]), all system components were non-metallic withthe exception of a titanium heat exchanger, which is highly resistant tocorrosion [39], on the distillate side.

For all performance tests, the feed and distillate solutions werecirculated at 1.5 L/min using gear pumps (Pulsafeeder, Rochester, NY.)The feed and distillate temperature set points were held at 65 and 35 °Cusing proportional integral derivative controllers for the inline heaterand chiller. Feed concentrations of 35 and 200 g/L NaCl were used forthe long- and short-term experiments, respectively. ACS-grade NaCl(VWR, Radnor, PA) and deionized (DI) water were used for all solu-tions. Membranes were housed in a custom-built, clear, acrylic module,providing continuous visual observation of the membrane, with anactive membrane area of 0.014 m2. 1.7-mm mesh spacers (Sterlitech,Kent, WA) were used on either side of the membrane for support and toincrease turbulence in both the feed and distillate channels [40].

Flow rate and conductivity were continuously monitored using di-gital flow meters (Omega, Norwalk, CT) and in-line conductivity probesand transmitters (Cole Parmer, Vernon Hills, IL) on both the feed anddistillate streams, as shown in Fig. 1. Conductivity measurements wereused to characterize salt rejection. Resistance temperature detectors(Omega, Norwalk, CT) were used to measure the temperature at theinlets and outlets of the membrane module on both the feed and dis-tillate sides. Flux through the membrane was monitored by con-tinuously weighing distillate overflow on an analytical balance (Ohaus,Parsippany, NJ). Data from the probes and scale were recorded at 10-min intervals using a data acquisition and control program developed inLabView (National Instruments, Austin, TX).

2.3. Membrane characterization

Contact angle, morphology, chemical composition, and roughnessof the membrane surface were characterized prior to and after both

A.L. McGaughey et al. Journal of Membrane Science 543 (2017) 143–150

144

experiments. Contact angles of the membrane surface were measuredusing the sessile drop method and a goniometer (Model 500, Ramé-Hart, Succasunna, NJ). Prior to use, contact angles were measured on aminimum of three virgin membrane samples; measured values did notvary significantly for the different samples or with the side of thesymmetric membrane. Following use, contact angles were measured onthree samples of approximately 1 × 2 cm cut from the membrane. Forall contact angle measurements, 5-μL droplets of DI water were care-fully deposited onto the membrane surface using an automated pipette.Droplets were deposited on different areas of each sample and reportedvalues represent an average of at least ten measurements.

Surface morphology of both sides of the virgin and used membranewas evaluated using an FESEM (JSM-7001, Jeol USA, HuntingtonBeach, CA). Prior to imaging, samples were sputter-coated in gold andpalladium to render the surface conductive. Energy-dispersive X-rayspectroscopy (EDS) was also used for semi-quantitative characterizationof the chemical composition of the membrane surface. Membranesamples were cut adjacent to the membrane sample cut for performancetesting. FESEM-EDS results showed that there were no significant dif-ferences in surface morphology or elemental composition between thefeed and distillate sides of unused membranes. Following performancetesting, FESEM and EDS measurements on both the feed and distillatesides were taken on samples of approximately 1 × 1 cm cut from theused membrane. ImageJ software was used to analyze FESEM imagesfor fibril/node spacing.

Membrane surface roughness was characterized using an atomicforce microscope (AFM) (Innova, Bruker, Billerica, MA). Initial mea-surements were taken on both sides of the unused membrane, withsamples cut adjacent to the membrane sample for performance testing.Measurements on the unused membrane showed that there were nosignificant differences in roughness between different sides of the un-used membrane. Following performance testing, roughness measure-ments were taken on samples of approximately 1 × 1 cm cut from theused membrane. Samples were affixed to mounting disks using double-sided carbon tape and 10 × 10 µm areas of each sample were scannedin tapping mode using an antimony-doped silicon probe (FMV-A,Bruker, Billerica, MA). Images were analyzed using Gwyddion software(version 2.47, Czech Metrology Institute, Brno, Czech Republic) androughness parameters were obtained by averaging horizontal lineprofiles for each image.

3. Results and discussion

3.1. Effects of long-term operation

3.1.1. Membrane performanceWater flux and distillate conductivity versus elapsed time for the

long-term (100-day) experiment with 35 g/L NaCl feed solution areshown in Fig. 2.

Flux decreased fairly linearly with time over the 100 days of con-tinuous operation, with a correlation coefficient of 0.9; overall, fluxdecreased by 16%. The cause for the decrease in flux was not apparentat first; eventually, a brown-colored fouling layer became visible on themembrane surface. It should be noted that low-fouling feed solutionsare not necessarily “clean” and at least some degree of fouling is likelyto occur over long time periods. Although no foulants were added to thefeed solution, trace quantities of organic matter may have been presentin the DI water used to make the NaCl feed solution; also, dust mayhave entered the feed solution via the auxiliary feed tank, which waspartially open to the environment. Flux decline was likely due to gra-dual accumulation of these trace foulants on the surface. The hydro-phobicity of the membrane and elevated operating temperatures usedin MD could result in greater attraction of organics to the membranesurface than in membrane processes using hydrophilic membranes or

Fig. 1. Schematic of bench-scale DCMD system.

Fig. 2. Water flux and distillate conductivity versus time for long-term experiment with35 g/L NaCl feed solution and average feed and distillate temperatures of 63 and 40 °Cmeasured at the module. Flow rates on the feed and distillate sides were maintained at1.5 L/min throughout the experiment. Distillate conductivity is shown as a movingaverage.

A.L. McGaughey et al. Journal of Membrane Science 543 (2017) 143–150

145

operating at ambient temperatures [41–43]. A brief decrease in fluxoccurred at the end of day 40 due to a malfunction in the chiller, whichshut off for 8.5 h overnight (data omitted). When the chiller restarted,there was a slightly elevated driving force, resulting in the subsequentbrief increase in flux shown in Fig. 2. The chiller also briefly mal-functioned on day 68 (data omitted); flux did not significantly increaseafterward. Although the distillate temperature increased by 14 °Cduring both events, a temperature difference of at least 10 °C wasmaintained. Both events were brief and the system rapidly returned tosteady state after each. In previous studies, temporary system shut-downs have been associated with membrane wetting due to vaporcondensation within the membrane pores (e.g., [8,18,28,44]). This wasnot observed in the current study, in which the system was not shut-down and membrane wetting did not occur.

Feed conductivity averaged 89± 3.1 mS/cm and distillate con-ductivity remained below 8 μS/cm for the duration of the experiment.Salt rejection based on conductivity was always greater than 99.8%. Atday 58, distillate conductivity, which was relatively constant for thefirst 50 days of operation, began to slightly increase. At the same time,there was a slight (< 1 L/m2 h) increase in water flux. This indicatesthe possibility of a small volume of liquid feed solution passing throughthe membrane, simultaneously increasing the measured water flux anddistillate conductivity. However, as shown by the low distillate con-ductivity and high rejection maintained throughout the experiment, itis likely that most of the membrane pores remained dry.

3.1.2. Membrane fouling and surface morphologyRepresentative FESEM images and EDS spectra for the membrane

are shown in Fig. 3. A representative image of one side of the virginmembrane surface was selected (Fig. 3a) as the feed and distillate sidesdid not differ significantly. FESEM images of the feed side of the usedmembrane (Fig. 3b) confirmed that the fouling layer fully coveredportions of the membrane. FESEM images of the distillate side of theused membrane (Fig. 3c) showed no external foulants and a similarmorphology as the virgin membrane but with increased fibril and nodewidths. ImageJ software was used to analyze the void space in the

FESEM images of the virgin and distillate side of the used membrane.Void space decreased by an average of 70% after long-term operation.This was supported by visual observation of the images; the virginmembrane has a greater number of fibrils as well as more fibrils withthin diameters (Fig. 3a). In contrast, the used membrane has wider fi-brils and larger, flattened nodes (Fig. 3c). Fibril aggregation was alsoreported by Saffarini et al. [6] when the observed heat-treated PTFEmembranes had shorter fibrils relative to virgin membranes, perhapsdue to aggregation of nodes. Fibril and node changes were attributed totemperature effects [6]. Morphology changes were also observed byBarbe et al. [45] for PP membranes soaked in water for 24–72 h; thesechanges were attributed to water entry into pores resulting in poreexpansion. Interestingly, no changes were observed when using a CaCl2solution [45].

In the EDS spectrum for the virgin membrane (Fig. 3d), the fluorinepeak represents 73% while the carbon peak represents 25% on anatomic basis, reflecting the structure of the PTFE molecule [46]. In theEDS spectrum for the feed side of the used membrane (Fig. 3e), thefluorine peak represented only 0.4% on an atomic basis; in some scans,it was below detection limits. This indicates that portions of themembrane surface were fully covered by the fouling layer. The greateramount of carbon relative to fluorine and the appearance of oxygenindicated that organic matter was present in the fouling layer. Hy-drogen cannot be identified by EDS; however, the carbon to oxygenratio (55% C to 20% O by weight) is somewhat similar to that of naturalorganic matter (40–60% C to 30–50% O by weight) [47,48]. Althoughnitrogen and sulfur did not appear, these elements can be 0 to less than1% by weight of natural organic matter [48]. A small peak for silica alsoappeared in the feed-side EDS spectra. Silica may have been present invery small concentrations in the DI water used in the feed solution; it isalso possible that silica came from silicone components (e.g., sealant)used in the system although the reported temperature resistance of allsilicone components was well above operating temperatures. Carbonassociated with the silica deposits as silicone may also account for thehigher C:O ratio relative to natural organic matter. EDS results for thedistillate side of the used membrane (Fig. 3f) showed it was free of

Fig. 3. Representative FESEM images of (a) virgin membrane, (b) feed side of used membrane, and (c) distillate side of used membrane and EDS spectra of (d) virgin membrane, (e) feedside of used membrane, and (f) distillate side of used membrane for the long-term experiment with 35 g/L NaCl feed solution and average feed and distillate temperatures of 63 and 40 °Cmeasured at the module. FESEM images were taken at × 4000 magnification; EDS spectra were taken at × 250 magnification to characterize a larger area of the heterogeneous foulingdeposits. EDS spectra were taken in triplicate for each sample and the resulting average atomic percent composition values are shown. Values for the virgin membrane represent anaverage of both sides of the symmetric membrane.

A.L. McGaughey et al. Journal of Membrane Science 543 (2017) 143–150

146

external foulants. Higher magnification EDS scans of the node areabetween fibrils confirmed that these areas were composed only ofmembrane material.

3.1.3. Membrane hydrophobicityContact angles measured on the virgin and used membrane for the

long-term experiment are shown in Table 1.The contact angle of the feed side of the membrane was reduced by

56%, making the surface much more hydrophilic after use; this was notunexpected due to the fouling layer on the surface. Numerous previousstudies have noted that fouling on the feed side of the membrane pro-motes wetting because the more hydrophilic foulant layer provides apath for liquid to enter the membrane pores [21,24,28,42,49–51].Given the presence of the fouling layer, no conclusions regarding thechange in contact angle of the actual membrane surface could be made.The contact angle of the fouled membrane as it was during operationrepresented the “effective hydrophobicity” of the membrane surface, asexperienced by the feed solution. As foulants began to adhere to themembrane surface, the virgin membrane surface no longer interfacedwith the feed solution; instead, the feed solution was in contact with amore heterogeneous and hydrophilic surface composed of PTFE andfoulants. Interestingly, despite the effective hydrophilization of thefeed-side surface of the membrane, salt rejection remained high andconstant. This suggests that another mechanism, or combination ofmechanisms, may have played a significant role in maintaining rejec-tion.

Surprisingly, the contact angle of the distillate side of the membranewas reduced 26% during use. The lack of fouling on the distillate side,as confirmed by FESEM and EDS analyses, means the reduced contactangle is entirely due to physical changes in the membrane itself.Although to the best of the authors’ knowledge reductions in distillate-side contact angles have not been reported in the literature, PP mem-branes soaked in brine solution have been found to be hydrophilizeddue to the formation of hydrophilic groups [50]. Also, heat treatment ofPTFE membranes has been found to result in a lower measured liquidentry pressure (LEP), or pressure at which liquid can penetrate mem-brane pores [6].

Contact angles can be related to LEP according to

=

−

LEPβγ θr

2 cos( )l

p max, (1)

where β is a parameter describing pore geometry, γl is surface tension ofthe wetting liquid, θ is contact angle between the wetting liquid andsurface, and rp max, is maximum pore radius [15]. LEP measurements ofQM022 membranes resulted in an LEP of 450 kPa for pure water at25 °C [15]. The decrease in contact angle on the feed side from 140° to61° corresponds to a decrease in LEP from 450 to 0 kPa, indicating thatthe feed side of the membrane would readily be wetted at ambientconditions. On the distillate side, the decrease in contact angle from140° to 104° corresponds to a decrease in LEP from 450 to 142 kPa.Although this change is significant, a distillate-side LEP of 142 kPawould still likely be high enough to prevent significant wetting giventhat testing in a separate experiment (using stainless-steel pressureprobes) found operating pressures for this system to be consistently lessthan 70 kPa.

Despite possible surface wetting caused by the fouling layer, liquiddid not pass through the membrane. With a wetted surface layer, thefeed solution vapor interface could move inward and may be located atan inner layer of the membrane or within membrane pores [22]. Thissuggests that internal hydrophobicity may be important in maintainingrejection, especially when a fouling layer is present. The membranesused in this study are symmetric and composed of a single layer ofPTFE; however, the internal hydrophobicity may differ from that of themembrane surface. It has been previously noted that the hydro-phobicity of the interior of PVDF membranes may not be equal to thesurface hydrophobicity [52]. If internal hydrophobicity differs fromsurface hydrophobicity, it may contribute uniquely to rejection whensurface hydrophilization occurs during long-term operation. Also, al-though improved performance and rejection have been demonstratedmany times for MD membranes with superhydrophobic surface coatingsor thin surface layers (e.g., [53–55]), their long-term performance isless certain. Gryta [22] also suggested that a substantial hydrophobiclayer may be necessary to maintain rejection based on surface wettingstudies. When the feed side surface is hydrophilized, internal hydro-phobicity and distillate-side surface hydrophobicity together may benecessary to maintain a vapor-filled layer and prevent wetting viabridging of liquid from the feed and distillate streams. However, adetailed study of the role of internal hydrophobicity has not yet beenperformed in the MD literature.

3.1.4. Membrane surface roughnessRMS roughness values obtained from Gwyddion image analysis

software are shown in Table 2 and AFM scans of the virgin and usedmembrane are shown in Fig. 4. A representative image of one side of thevirgin membrane surface was selected as the two sides did not differsignificantly. Membrane fouling on the feed side of the used membraneled to a reduction in roughness of approximately 92% relative to thevirgin membrane (Table 2). This can be observed both in the AFMimages (comparison of Fig. 4a and b) and FESEM images (comparisonof Fig. 3a and b). The fouling deposits effectively smoothed the mem-brane surface on the feed side. A similar effect was observed by Guillen-Burrieza et al. [8]; soaking and intermittently drying PTFE membranesin seawater for 5–20 days resulted in fouling and scaling deposits thatfilled in spaces between fibrils and effectively smoothed the membranesurface. In the current study, it is interesting to note that the morpho-logical changes on the distillate side of the used membrane were alsocharacterized by a significant reduction in roughness – approximately57% relative to the virgin membrane (Table 2). As increased roughnesscorrelates with increased hydrophobicity [9], decreased surfaceroughness is likely to have contributed to the decreased contact anglesseen in Table 1.

For the distillate side of the used membrane, changes in membranesurface hydrophobicity and morphology (Fig. 4c) may have been due totemperature effects or pore wetting, as proposed by previous studies[6,45]. Saffarini et al. [6] observed minor shifts in the surface structureof PTFE membranes at temperatures< 50 °C in air as well as decreasedLEP with increasing temperature. The morphological changes may havealso been due to shear forces exerted on the membrane surfaces by thetransverse flow of liquid over the surface. Computational fluid dy-namics models have found shear stress values of approximately 1–7 Pa,

Table 1Contact angles measured on the virgin and used membrane before and after the long-termexperiment with 35 g/L NaCl feed solution and average feed and distillate temperaturesof 63 and 40 °C measured at the module.

Membrane Side Water contact angle

Virgin Feed 140±4.0°Distillate 140± 3.6°

Used Feed 61±9.1°Distillate 104± 14°

Table 2Roughness values for membrane used in the long-term experiment with 35 g/L NaCl feedsolution and average feed and distillate temperatures of 63 and 40 °C measured at themodule.

Membrane Side RMS roughness (nm)

Virgin Feed 31.0± 5.40Distillate 27.1± 7.58

Used Feed 3.0±0.88Distillate 14.5± 3.11

A.L. McGaughey et al. Journal of Membrane Science 543 (2017) 143–150

147

depending on spacer configuration, at similar cross-flow velocities [56].Morphological changes may also have been due to hydraulic pressureeffects on both sides of the membrane that caused compression overtime even though only low pressures (less than 70 kPa) are required forrecirculation. These forces are likely orders of magnitude lower thanthe Young's modulus of the membrane, as similar polymer membraneshave shown yield stress values of 7.5 MPa [57]. Although shear forcesand hydraulic pressure were likely low in this study, long-term opera-tion may result in permanent deformation over time via creep, espe-cially at elevated temperatures [58]. Polymeric membranes are knownto have relatively low creep resistance [59]. Within the MD literature,creep and membrane compaction have been only briefly mentioned aspotential risks of degassing feed solution, as causes of minor flux de-cline, and as benefits of mechanically strong membrane materials[60–62]. However, creep, especially due to the combined effect of shearforces, hydraulic pressure, and temperature, has not been previouslyreported in the literature to the best of the authors’ knowledge.

3.2. Effect of short-term operation with scaling and confirmation of long-term results

To corroborate the results of the long-term study and to furtherinvestigate membrane resistance to wetting under scaling conditions, ashort-term (20-day) experiment with 200 g/L NaCl feed solution wasperformed, with membrane scaling induced after 15 days. Operatingconditions were otherwise identical to the previous long-term experi-ment. Water flux and distillate conductivity versus elapsed time for theshort-term experiment are shown in Fig. 5.

Flux decreased fairly linearly with time over the first 15 days of

continuous operation, with a correlation coefficient of 0.7; overall, fluxdecreased by less than 5%. This was similar to the long-term experi-ment with 35 g/L NaCl, in which flux declined by 3% over the first 15days.

After day 15, the solenoid valve loop maintaining constant feedconcentration was shut off and the feed solution was allowed to con-centrate in batch mode. Operating conditions, including temperatures,were otherwise held constant. Due to the elevated concentration, avisible layer of crystallized salt (scale) developed on the membranesurface and caused rapid and severe flux decline, reducing and even-tually preventing water flux through the membrane. After replenishingthe feed tank with DI water, flux was partially recovered to approxi-mately 34% of the initial flux and remained constant for the final fivedays of operation. This demonstrated that some portion of the foulingwas reversible, although the irreversible portion was associated withthe majority of the flux decline. The virgin and used membrane surfaceswere characterized by FESEM and EDS, as shown in Fig. 6. A re-presentative image of one side of the virgin membrane surface wasselected as the two sides did not differ significantly.

The feed side of the used membrane was nearly completely coveredby fouling deposits (Fig. 6b), which were composed of carbon, sodium,and chlorine (Fig. 6e). The EDS spectrum also showed oxygen andfluorine, although there was significantly more carbon relative tofluorine, indicating that organic fouling occurred (as in the long-termexperiment). A very small peak for silica was present, likely originatingfrom the same source as the long-term experiment. The appearance ofnickel on the feed side was surprising, especially given the lack of nickelin the long-term experiment; it is possible that this originated fromimpurities in the ACS-reagent-grade NaCl used for feed solutions, whichcontained ~ 2 mg/L heavy metals. The development of the scaling layerat the end of 15 days resulted in a slight increase in distillate con-ductivity (Fig. 5), indicating that some pore wetting occurred. Thescaling layer likely promoted pore wetting by decreasing the effectivehydrophobicity of the surface [8,21,24,28,42,44,50,51]. The increasedsalinity of the feed solution may also have promoted membrane wet-ting. NaCl solutions result in a lower contact angle relative to purewater, but NaCl solutions also have a slightly higher surface tensionthan pure water; the overall effect on LEP has been found to be fairlysmall [6] but may be significant at higher salinities. In this study, de-spite severe scaling of the membrane and elevated feed solution con-centration, the distillate conductivity remained quite low as in the long-term experiment.

The distillate side of the used membrane also had a small amount ofsodium and chlorine present on the surface, indicating that salt pene-trated to the distillate side of the membrane due to wetting of somemembrane pores; crystallization likely occurred during desiccation. Thelack of oxygen and the similar carbon to fluorine ratio as the virginmembrane indicated that essentially no organic foulant material waspresent on the distillate side of the membrane. As in the long-termexperiment, there was a distinct change in morphology of the distillateside of the used membrane (comparing Fig. 6a and c). The number of

Fig. 4. Representative tapping mode AFM images of (a) virgin membrane, (b) feed side of used membrane, and (c) distillate side of used membrane for the long-term experiment with35 g/L NaCl feed solution and average feed and distillate temperatures of 63 and 40 °C measured at the module.

Fig. 5. Water flux and distillate conductivity versus time for the short-term experimentwith 200 g/L NaCl feed concentration and average feed and distillate temperatures of 63and 40 °C measured at the module. Distillate conductivity is shown as a moving average.

A.L. McGaughey et al. Journal of Membrane Science 543 (2017) 143–150

148

thin membrane fibrils decreased and there was an increase in the ap-parent node or aggregated fibril area, although ImageJ analysis showedno significant decrease in void space between the virgin and usedmembrane.

Contact angles measured on the virgin and used membrane in theshort-term experiment are presented in Table 3.

As before, the contact angle decreased on both sides of the mem-brane. On the feed side of the used membrane the contact angle wasreduced by 65%, making the surface much more hydrophilic; this wasexpected due to the visible layer of salt crystals on the surface. Thecontact angle on the distillate side of the used membrane decreased by19% relative to the virgin membrane. The significant decrease afteronly 20 days of use was unexpected; however, this change was less thanthat observed in the long-term experiment, suggesting that the dis-tillate-side contact angle decreases with increasing duration of use.Morphology changes were also more significant after 100 days of op-eration than after 20 days. The results of the short-term experimentcorroborated the findings of the long-term experiment and supportedthe likelihood that duration of use affects membrane characteristicseven when membranes are exposed only to salt solutions and tracefoulants. The observations made from comparing 100-day to 20-dayoperation further supported the likelihood that feed-side surface hy-drophobicity is not an isolated parameter and other factors, specificallydistillate-side surface hydrophobicity and internal hydrophobicity, mayplay key roles in maintaining salt rejection.

4. Conclusions and implications

In this study, significant changes to MD membrane surface mor-phology, decreased membrane hydrophobicity, and decreased surfaceroughness were observed after long-term use. Distillate conductivityremained low even after significant decrease in the feed-side contactangle; this suggests that maintaining distillate-side hydrophobicity and/or internal hydrophobicity may be more important for long-term per-formance than has been previously suggested. Furthermore, fouling wasobserved even when “clean” feed solutions were used. When fouling ispresent, distillate-side and internal hydrophobicity may be equally ormore important than feed-side hydrophobicity as the distillate surface,internal layers, and pore walls are less affected by foulants that caneffectively hydrophilize the surface. Effects of long-term operation maybe especially significant for membranes with surface coatings or thinactive layers, which may be more vulnerable than internal material anddistillate-side membrane surfaces.

Furthermore, there is significant interest in recovering membranehydrophobicity after fouling has occurred. Typically, membrane hy-drophobicity is assumed to be recoverable if foulants can be removed;however, if membrane hydrophobicity is decreased due to physical orchemical changes in the membrane itself, hydrophobicity may not berecoverable. Permanent reductions in hydrophobicity can result inpermanent reductions in membrane performance and lifetime.

Acknowledgements

The authors would like to acknowledge funding support for thisstudy from the Strategic Environmental Research and DevelopmentProgram (SERDP Project number ER-2237) and from the USEnvironmental Protection Agency Science to Achieve Results (US EPASTAR Grant #83486701). Additionally, the authors would like to ac-knowledge support to this project from two fellowships awarded to R.D.Gustafson: a Viterbi Graduate School Ph.D. Fellowship and a NationalWater Research Institute and Southern California Salinity Coalitionfellowship. The FESEM-EDS and AFM images and spectra were acquiredat the University of Southern California Center for Electron Microscopy

Fig. 6. Representative FESEM images of (a) virgin membrane, (b) feed side of used membrane, and (c) distillate side of used membrane and EDS spectra of (d) virgin membrane, (e) feedside of used membrane, and (f) distillate side of used membrane for the short-term experiment with 200 g/L NaCl feed concentration and average feed and distillate temperatures of 63and 40 °C measured at the module. Distillate conductivity is shown as a moving average. All images and EDS spectra were taken at × 4000 magnification and the resulting atomic percentcomposition values are shown. Values for the virgin membrane represent an average of both sides.

Table 3Contact angles measured on membrane used in the short-term experiment with 200 g/LNaCl feed concentration and average feed and distillate temperatures of 63 and 40 °Cmeasured at the module.

Membrane Side Water contact angle

Virgin Feed 140±3.96°Distillate 140± 3.63°

Used Feed side 49.1± 3.18°Distillate side 114±6.35°

A.L. McGaughey et al. Journal of Membrane Science 543 (2017) 143–150

149

and Microanalysis and Center for Excellence in NanoBioPhysics, re-spectively.

References

[1] F. Suárez, S.W. Tyler, A.E. Childress, A theoretical study of a direct contact mem-brane distillation system coupled to a salt-gradient solar pond for terminal lakesreclamation, Water Res. 44 (2010) 4601–4615.

[2] F. Suárez, J.A. Ruskowitz, S.W. Tyler, A.E. Childress, Renewable water: directcontact membrane distillation coupled with solar ponds, Appl. Energy 158 (2015)532–539.

[3] T.Y. Cath, V.D. Adams, A.E. Childress, Experimental study of desalination usingdirect contact membrane distillation: a new approach to flux enhancement, J.Membr. Sci. 228 (2004) 5–16.

[4] L. Camacho, L. Dumée, J. Zhang, J.-d. Li, M. Duke, J. Gomez, S. Gray, Advances inmembrane distillation for water desalination and purification applications, Water 5(2013) 94.

[5] E. Drioli, A. Ali, F. Macedonio, Membrane distillation: recent developments andperspectives, Desalination 356 (2015) 56–84.

[6] R.B. Saffarini, B. Mansoor, R. Thomas, H.A. Arafat, Effect of temperature-dependentmicrostructure evolution on pore wetting in PTFE membranes under membranedistillation conditions, J. Membr. Sci. 429 (2013) 282–294.

[7] K.J. Kim, A.G. Fane, C.J.D. Fell, T. Suzuki, M.R. Dickson, Quantitative microscopicstudy of surface characteristics of ultrafiltration membranes, J. Membr. Sci. 54(1990) 89–102.

[8] E. Guillen-Burrieza, R. Thomas, B. Mansoor, D. Johnson, N. Hilal, H. Arafat, Effectof dry-out on the fouling of PVDF and PTFE membranes under conditions simu-lating intermittent seawater membrane distillation (SWMD), J. Membr. Sci. 438(2013) 126–139.

[9] D. Quéré, Wetting and roughness, Annu. Rev. Mater. Res. 38 (2008) 71–99.[10] G. Rao, S.R. Hiibel, A. Achilli, A.E. Childress, Factors contributing to flux im-

provement in vacuum-enhanced direct contact membrane distillation, Desalination367 (2015) 197–205.

[11] C.R. Martinetti, A.E. Childress, T.Y. Cath, High recovery of concentrated RO brinesusing forward osmosis and membrane distillation, J. Membr. Sci. 331 (2009) 31–39.

[12] T.Y. Cath, D. Adams, A.E. Childress, Membrane contactor processes for wastewaterreclamation in space: II. Combined direct osmosis, osmotic distillation, and mem-brane distillation for treatment of metabolic wastewater, J. Membr. Sci. 257 (2005)111–119.

[13] T.Y. Cath, S. Gormly, E.G. Beaudry, M.T. Flynn, V.D. Adams, A.E. Childress,Membrane contactor processes for wastewater reclamation in space: Part I. Directosmotic concentration as pretreatment for reverse osmosis, J. Membr. Sci. 257(2005) 85–98.

[14] D.L. Shaffer, M.E. Tousley, M. Elimelech, Influence of polyamide membrane surfacechemistry on gypsum scaling behavior, J. Membr. Sci. 525 (2017) 249–256.

[15] G. Rao, S.R. Hiibel, A.E. Childress, Simplified flux prediction in direct-contactmembrane distillation using a membrane structural parameter, Desalination 351(2014) 151–162.

[16] T.Y. Cath, A.E. Childress, C.R. Martinetti, Combined Membrane-distillation-for-ward-osmosis Systems and Methods of Use, in, Google Patents, 2011.

[17] K. Schneider, W. Hölz, R. Wollbeck, S. Ripperger, Membranes and modules fortransmembrane distillation, J. Membr. Sci. 39 (1988) 25–42.

[18] F.A. Banat, J. Simandl, Theoretical and experimental study in membrane distilla-tion, Desalination 95 (1994) 39–52.

[19] K. Karakulski, M. Gryta, A. Morawski, Membrane processes used for potable waterquality improvement, Desalination 145 (2002) 315–319.

[20] C. Feng, K.C. Khulbe, T. Matsuura, R. Gopal, S. Kaur, S. Ramakrishna, M. Khayet,Production of drinking water from saline water by air-gap membrane distillationusing polyvinylidene fluoride nanofiber membrane, J. Membr. Sci. 311 (2008) 1–6.

[21] M. Gryta, Long-term performance of membrane distillation process, J. Membr. Sci.265 (2005) 153–159.

[22] M. Gryta, Influence of polypropylene membrane surface porosity on the perfor-mance of membrane distillation process, J. Membr. Sci. 287 (2007) 67–78.

[23] M. Gryta, M. Barancewicz, Influence of morphology of PVDF capillary membraneson the performance of direct contact membrane distillation, J. Membr. Sci. 358(2010) 158–167.

[24] M. Gryta, Wettability of polypropylene capillary membranes during the membranedistillation process, Chem. Pap. 66 (2012) 92–98.

[25] M.M.A. Shirazi, A. Kargari, M.J.A. Shirazi, Direct contact membrane distillation forseawater desalination, Desalin. Water Treat. 49 (2012) 368–375.

[26] M. El Fray, M. Gryta, Environmental fracture of polypropylene membranes used inmembrane distillation process, Polimery 53 (2008) 865–870.

[27] E. Guillén-Burrieza, J. Blanco, G. Zaragoza, D.-C. Alarcón, P. Palenzuela, M. Ibarra,W. Gernjak, Experimental analysis of an air gap membrane distillation solar desa-lination pilot system, J. Membr. Sci. 379 (2011) 386–396.

[28] E. Guillen-Burrieza, A. Ruiz-Aguirre, G. Zaragoza, H.A. Arafat, Membrane foulingand cleaning in long term plant-scale membrane distillation operations, J. Membr.Sci. 468 (2014) 360–372.

[29] M.M. Teoh, T.-S. Chung, Y.S. Yeo, Dual-layer PVDF/PTFE composite hollow fiberswith a thin macrovoid-free selective layer for water production via membranedistillation, Chem. Eng. J. 171 (2011) 684–691.

[30] P. Peng, A.G. Fane, X. Li, Desalination by membrane distillation adopting a hy-drophilic membrane, Desalination 173 (2005) 45–54.

[31] F. Edwie, M.M. Teoh, T.-S. Chung, Effects of additives on dual-layer hydro-phobic–hydrophilic PVDF hollow fiber membranes for membrane distillation andcontinuous performance, Chem. Eng. Sci. 68 (2012) 567–578.

[32] M.R. Bilad, F.A. Al Marzooqi, H.A. Arafat, New concept for dual-layer hydrophilic/hydrophobic composite membrane for membrane distillation, J. Membr. Sep.Technol. 4 (2015) 122–133.

[33] M. Khayet, J.I. Mengual, T. Matsuura, Porous hydrophobic/hydrophilic compositemembranes: application in desalination using direct contact membrane distillation,J. Membr. Sci. 252 (2005) 101–113.

[34] C. Han, D. Tang, D. Kim, Molecular dynamics simulation on the effect of pore hy-drophobicity on water transport through aquaporin-mimic nanopores, ColloidsSurf. A: Physicochem. Eng. Asp. 481 (2015) 38–42.

[35] S. Sriraman, I.G. Kevrekidis, G. Hummer, Coarse nonlinear dynamics and metast-ability of filling-emptying transitions: water in carbon nanotubes, Phys. Rev. Lett.95 (2005) 130603.

[36] Y.-I. Chou, Z.-Y. Siao, Y.-F. Chen, L.-Y. Sung, W.-M. Yang, C.-C. Wang, Water per-meation analysis on gas diffusion layers of proton exchange membrane fuel cells forTeflon-coating annotation, J. Power Sources 195 (2010) 536–540.

[37] W. Song, H. Yu, L. Hao, Z. Miao, B. Yi, Z. Shao, A new hydrophobic thin filmcatalyst layer for PEMFC, Solid State Ion. 181 (2010) 453–458.

[38] M. Gryta, Effect of iron oxides scaling on the MD process performance, Desalination216 (2007) 88–102.

[39] F. Mansfeld, G. Liu, H. Xiao, C.H. Tsai, B.J. Little, The corrosion behavior of copperalloys, stainless steels and titanium in seawater, Corros. Sci. 36 (1994) 2063–2095.

[40] R.D. Gustafson, J.R. Murphy, A. Achilli, A stepwise model of direct contact mem-brane distillation for application to large-scale systems: experimental results andmodel predictions, Desalination 378 (2016) 14–27.

[41] G. Naidu, S. Jeong, S. Vigneswaran, Interaction of humic substances on fouling inmembrane distillation for seawater desalination, Chem. Eng. J. 262 (2015)946–957.

[42] L.D. Tijing, Y.C. Woo, J.-S. Choi, S. Lee, S.-H. Kim, H.K. Shon, Fouling and itscontrol in membrane distillation—a review, J. Membr. Sci. 475 (2015) 215–244.

[43] M. Gryta, Fouling in direct contact membrane distillation process, J. Membr. Sci.325 (2008) 383–394.

[44] J. Gilron, Y. Ladizansky, E. Korin, Silica fouling in direct contact membrane dis-tillation, Ind. Eng. Chem. Res. 52 (2013) 10521–10529.

[45] A. Barbe, P. Hogan, R. Johnson, Surface morphology changes during initial usage ofhydrophobic, microporous polypropylene membranes, J. Membr. Sci. 172 (2000)149–156.

[46] National Institute of Standards and Technology Chemisty WebBook, in: NIST (Ed.),U.S. Secretary of Commerce, 2016.

[47] N. Hilal, A.F. Ismail, T. Matsuura, D. Oatley-Radcliffe, Membrane Characterization,Elsevier, Amsterdam, Netherlands, 2017.

[48] M. Sillanpää, Chapter 1 - general introduction, in: Natural Organic Matter in Water,Butterworth-Heinemann, 2015, pp. 1–15.

[49] G. Naidu, S. Jeong, S.-J. Kim, I.S. Kim, S. Vigneswaran, Organic fouling behavior indirect contact membrane distillation, Desalination 347 (2014) 230–239.

[50] M. Gryta, J. Grzechulska-Damszel, A. Markowska, K. Karakulski, The influence ofpolypropylene degradation on the membrane wettability during membrane dis-tillation, J. Membr. Sci. 326 (2009) 493–502.

[51] D.M. Warsinger, J. Swaminathan, E. Guillen-Burrieza, H.A. Arafat, J.H. Lienhard V,Scaling and fouling in membrane distillation for desalination applications: a review,Desalination 356 (2015) 294–313.

[52] E. Guillen-Burrieza, M.O. Mavukkandy, M.R. Bilad, H.A. Arafat, Understandingwetting phenomena in membrane distillation and how operational parameters canaffect it, J. Membr. Sci. 515 (2016) 163–174.

[53] Y. Liao, R. Wang, A.G. Fane, Engineering superhydrophobic surface on poly(viny-lidene fluoride) nanofiber membranes for direct contact membrane distillation, J.Membr. Sci. 440 (2013) 77–87.

[54] K.-J. Lu, J. Zuo, T.-S. Chung, Tri-bore PVDF hollow fibers with a super-hydrophobiccoating for membrane distillation, J. Membr. Sci. 514 (2016) 165–175.

[55] J. Zhang, Z. Song, B. Li, Q. Wang, S. Wang, Fabrication and characterization ofsuperhydrophobic poly (vinylidene fluoride) membrane for direct contact mem-brane distillation, Desalination 324 (2013) 1–9.

[56] M. Shakaib, S.M.F. Hasani, I. Ahmed, R.M. Yunus, A CFD study on the effect ofspacer orientation on temperature polarization in membrane distillation modules,Desalination 284 (2012) 332–340.

[57] X. Li, X. Yu, C. Cheng, L. Deng, M. Wang, X. Wang, Electrospun superhydrophobicorganic/inorganic composite nanofibrous membranes for membrane distillation,ACS Appl. Mater. Interfaces 7 (2015) 21919–21930.

[58] A. Sadeghi Alavijeh, R.M.H. Khorasany, A. Habisch, G.G. Wang, E. Kjeang, Creepproperties of catalyst coated membranes for polymer electrolyte fuel cells, J. PowerSources 285 (2015) 16–28.

[59] E. Chabanon, D. Mangin, C. Charcosset, Membranes and crystallization processes:state of the art and prospects, J. Membr. Sci. 509 (2016) 57–67.

[60] H. Maab, A. Al Saadi, L. Francis, S. Livazovic, N. Ghafour, G.L. Amy, S.P. Nunes,Polyazole hollow fiber membranes for direct contact membrane distillation, Ind.Eng. Chem. Res. 52 (2013) 10425–10429.

[61] B. Li, K.K. Sirkar, Novel membrane and device for direct contact membrane dis-tillation-based desalination process, Ind. Eng. Chem. Res. 43 (2004) 5300–5309.

[62] A.E. Jansen, J.W. Assink, J.H. Hanemaaijer, J. van Medevoort, E. van Sonsbeek,Development and pilot testing of full-scale membrane distillation modules for de-ployment of waste heat, Desalination 323 (2013) 55–65.

A.L. McGaughey et al. Journal of Membrane Science 543 (2017) 143–150

150