Pore structure characterization of asymmetric membranes:Non-destructive characterization of porosity and tortuosity

Seetha S Manickam a, Jeff Gelb b, Jeffrey R. McCutcheon a,n

a Department of Chemical & Biomolecular Engineering, Center for Environmental Sciences and Engineering, University of Connecticut, Storrs, CT, USAb Carl Zeiss X-ray Microscopy, Inc., Pleasanton, CA, USA

a r t i c l e i n f o

Article history:Received 25 June 2013Received in revised form23 October 2013Accepted 24 November 2013Available online 10 December 2013

Keywords:Pore structureMercury intrusion porosimetryX-ray microscopyTomographyForward osmosisPressure retarded osmosis

a b s t r a c t

Internal concentration polarization (ICP) in osmotic processes is largely influenced by the porousstructure of the support layer of the membrane. Recent publications on osmotic separations havedescribed a ‘structural parameter’, S (function of the thickness, tortuosity, and porosity of the supportlayer), that represents the support layer's contribution to the overall mass transfer resistance duringosmosis. To date, S has only been calculated as a fitted parameter in a model that requires experimentalflux measurements. Such a method is inaccurate since the models fail to account for all of the differentmass transfer phenomena. An alternative is to characterize the thickness, tortuosity, and porosityindependently and thus calculate the actual value of the structural parameter. However, for soft materialslike porous membranes, no standard methods have been established for measuring porosity andtortuosity. In this study, we propose the use of X-ray microscopy (XRM) for determining the structuralparameter of thin film composite (TFC) membrane support layers. The S value could be calculated fromthe XRM images and was compared to the results obtained from more conventional mercury intrusionporosimetry as well as an existing model used with empirical data. Substantial differences between thevalues obtained from the different techniques indicated the need to revise the traditional approaches ofcharacterizing membrane structures.

& 2013 Elsevier B.V. All rights reserved.

1. Introduction

Engineered osmosis (EO) is an emerging technology platformcomprising a number of membrane-based technologies. Theseinclude forward osmosis (FO), pressure-retarded osmosis (PRO)and direct osmotic concentration, which can be used for desalina-tion, power production and dewatering, respectively. These tech-nologies rely on osmotic gradients between a concentrated drawsolution and a relatively dilute feed solution. In EO, water fluxperformance is critical and is dependent on the osmotic pressuregradient over the selective layer of the membrane. The membranesupport layer however, poses a resistance to draw (in FO) andfeed (in PRO) solute mass transport that can dramatically reducethis driving force. This phenomenon is known widely as internalconcentration polarization (ICP) and is largely responsible forpreventing the use of existing commercial reverse osmosis (RO)membranes in EO processes.

Most EO membrane developers have focused on optimizing thesupport layer characteristics in order to reduce the severity of ICP.

The structural parameter, S, has been widely used as a metric toassess the membrane's contribution to ICP. S is defined as

S¼ tτε

ð1Þ

where t is the thickness, τ is the tortuosity, and ε is the porosity. Theseindividual characteristics can be manipulated in order to minimize thevalue of S, which is the goal of manymembrane development teams inindustry and academia. However, when making a newmembrane, theexact value of some of these characteristics, and by association thevalue of S, is unknown. So far, S has only been indirectly calculatedusing models based on experimental flux measurements and anassumption of film theory dictating mass transfer. One such modelis shown below

S¼ DJw

lnBþAπD;b

Bþ JwþAπF ;m

� �ð2Þ

where D is the solute diffusivity, Jw is the water flux, A is the purewater permeability of the membrane, B is the solute permeability ofthe membrane, πD,b is the osmotic pressure of the bulk draw solutionand πF,m is the osmotic pressure of the feed solution at the membraneinterface. It is explicitly clear that the above parameters do not definethe membrane structure and that changes in these values should not

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.2013.11.044

n Correspondence to: 191 Auditorium Road, Unit 3222, Storrs, CT 06269-3222,USA. Tel.: þ1 860 486 4601.

E-mail address: jeff@engr.uconn.edu (J.R. McCutcheon).

Journal of Membrane Science 454 (2014) 549–554

influence the support structure. However, investigators still use thismodel as a means of calculating S from osmotic flux measurements[1].

Recently, a method was proposed to standardize FO membranetesting. This investigation found that even when the same condi-tions were used amongst a number of research groups, S valuescould still vary when using this fitted parameter technique [2].One must note that the models used to calculate S are constantlyevolving in the literature in order to distinguish the differentresistances to mass transport in the system and uniquely identifythe resistance offered by the membrane structure itself. Many ofthese studies still rely on assumptions that are likely inaccurate.One such assumption is that external concentration polarization onthe support layer side of the membrane is negligible. Most modelsfail to account for this phenomenon, effectively lumping anyexternal CP into the S parameter calculation. For poorly perform-ing membranes, fluxes are low enough that this assumption is areasonable approximation. However, with the advent of highperformance EO membranes at both the laboratory and commer-cial scale, the resulting high fluxes mean that external CP can nolonger be ignored [3]. Existing models that continue to combineexternal and internal CP will overestimate S values as the fittedparameter of the equation. This results in an unreliable calculationof the structural parameter and an overestimation of themembrane's contribution to mass transfer resistance. If such aparameter could be calculated directly, rather than as a fittedparameter of a model, we would be able to better understandexactly how membrane structure plays a role in osmotic fluxperformance. However, few techniques are available to accuratelycharacterize the structural characteristics of membranes, such asporosity and tortuosity.

A review of the methods used to calculate porosity and porediameter distribution in soft nonwovens has been presented in aprevious publication by the authors [4]. Models that relatetortuosity to porosity and pore architecture, negating the need todirectly measure the tortuosity [5–7], are available. However,these models are empirical and can only be applied to specificisotropic structures. No models are available for asymmetric orcomposite structures, which include many of today's TFC mem-brane supports. Average tortuosity can be measured throughconductivity and diffusivity measurements of a dissolved solutethrough the porous material [8–10], but such efforts are compli-cated, difficult to reproduce, and have limited value in character-izing asymmetric composite structures. At the time of this writing,the only study on pore structure characterization of EO mem-branes is on microscopic characterization [11]. The techniquesused include scanning electron microscopy (SEM), transmissionelectron microscopy (TEM) and confocal laser scanning micro-scopy (CLSM). While this study provides some interesting insightson the structure of the particular membrane studied, the accuracyof using 2D imaging techniques (SEM, TEM) to characterizeasymmetric pore structures is debatable. Also, the two electronmicroscopy techniques were used to image the membrane in thedehydrated state and then comparisons were made to CLSMimages of the wetted membrane. The membrane studied wasmade of cellulose acetate, a hydrophilic polymer, which likelyexhibits swelling when hydrated. In general, techniques should bechosen carefully so that the sample preparation does not signifi-cantly alter or damage the structure being analyzed.

The objective of this study is to evaluate tools for characterizingthe 3D structure of commercially available TFC reverse osmosis(RO) membrane support layers. These membranes were chosensince they possess a composite and anisotropic structure typical ofmany TFC membranes today [12,13]. The membranes tested in thisstudy have also been previously evaluated for their performancesin FO [14]. TFC membranes are now finding broader application in

EO, with Oasys Water™ and Hydration Technology Innovations™both releasing their own commercially available versions in 2012[15,16]. Two characterization techniques have been used as a partof this study – an analytical method, mercury intrusion porosi-metry (MIP) and an imaging technique, X-ray microscopy (XRM).MIP is a widely used tool in the analysis of porous materials [4].XRM is a non-destructive 3D imaging technique that is widelyused in biomedical, geological and archeological applications.Recently, with the advent of improved phase contrast optics ithas been increasingly used to image soft materials [17]. The resultsfrom the two approaches were used to evaluate the membranestructures and calculate the intrinsic structural parameters.These values were then compared to values obtained from theconventional method of using an empirical model. The compar-ison demonstrates the inaccuracy of empirical approaches and theneed for better understanding of mass transport occurring duringosmosis across anisotropic and composite membranes.

2. Materials and methods

2.1. Materials

The membranes used in this study were the BW30 and SW30-XLE thin film composite reverse osmosis membranes from DowWater & Process Solutions™. These membranes were used as-received and characterized in their dry state.

2.2. Methods

2.2.1. Analytical characterizationMercury intrusion porosimetry (MIP) was used to characterize

porosity and tortuosity of these membranes in their dry state.The porosimeter used was a PoreMaster from QuantachromeCorporation. In addition to pore diameter distribution and poros-ity, tortuosity of the pore structure was calculated using a general-ized correlation [18,19].

τ¼ ð2:23�1:13VtotρbÞ 0:924S∑ΔVi

di

� �1þε !

ð4Þ

where τ is the tortuosity factor, Vtot is the total pore volume (cm3/g),ρb is the bulk density of sample (g/cm3), S is the Brunauer–Emmett–Teller (BET) surface area (m2/g), ΔVi is the change in pore volumewithin a pore size interval (cm3), di is the average diameter within apore size interval (cm), and ε is the pore shape exponent. A value ofε¼2.1 was assigned for both membranes in accordance with aprevious study [18].

Triplicate porosimetry experiments were performed foreach membrane to obtain average porosities and tortuosities.The experiment was set up as such that the instrument will onlymeasure pore size below 1 mm (which is the maximum resolutionof the Xradia MicroXRM) in order to enable a fair comparison. Thethicknesses of the two membranes were determined using amicrometer. Ten measurements were taken to obtain an averagethickness.

2.2.2. Imaging characterizationBoth scanning electron microscopy (SEM) and X-ray micro-

scopy (XRM) were used to image the two membranes used in thisstudy. A JEOL 6335F field emission SEM was used to obtain cross-sectional images of the two membranes. The membranes werefractured along orthogonal axes. One cross section represents themembrane in the direction in which the cast polysulfone (PSu)support was introduced into the precipitation bath. The second

S.S. Manickam et al. / Journal of Membrane Science 454 (2014) 549–554550

was the direction orthogonal to that of the previous one. In orderto allow the ease of freeze-fracturing, the nonwoven backing layerwas removed from both membranes and only the PSu supportlayers were imaged.

Two XRM instruments were used in this study for a multi-length scale approach. The Xradia MicroXCT-400 provided resolu-tion to �1 μm, while the Xradia UltraXRM-L200 extended thisresolution to 50 nm. On the MicroXCT, a 40� objective was usedto obtain 4000 projection radiographs at equally-spaced intervalsover a 1801 sample rotation, exposing each radiograph for 10 s.The X-ray power was set at 20 kV and 0.1 mA. In the case of theUltraXRM, 721 projection radiographs were collected over a 1801rotation range using a 64 nm pixel size and Zernike phase contrastimaging mode, exposing each radiograph for 75 s [20]. The X-raypower was set at 40 kV and 30 mA. The reconstructed images fromboth instruments were exported to Avizo™Fire (VisualizationSciences Group) for further image processing and analysis.The images were first filtered to remove background noise andthresholded to binarize the images into pore space and polymermatrix. The porosity was analyzed using ‘volume3d’, a built-inmeasurement tool that is used to compute the density of pixelsabove a certain intensity threshold. Porosity distribution as afunction of thickness of the membrane was determined by makingthis measurement at each slice. Tortuosity was measured using analgorithm described in a previous study [21]. The algorithmquantifies tortuosity by tracking the center of mass of each poreas it goes from one end of the sample surface to the other end. Thetotal length of this path is then divided by the Euclidian distance ofthe entire sample. Two membrane samples were imaged in theMicroXCT in order to obtain average porosities and tortuosities.The cross-sections obtained from XRM were exported to ImageJ(National Institutes of Health) to measure their thicknesses.Ten measurements were taken to obtain an average thickness.

2.2.3. Calculation of structural parameterThe structural parameter was measured in osmotic membranes

using experimental osmotic flux measurements. Details of theosmotic flux tests can be found elsewhere [14].

3. Results and discussion

3.1. Scanning electron microscopy (SEM)

Fig. 1 shows cross-sectional SEM images of the BW30 (a and b)and SW30-XLE (c and d). Fig. 1a and c corresponds to samplesfreeze-fractured perpendicular to the direction in which the castpolysulfone (PSu) support was introduced into the precipitationbath and in b and d, the samples were fractured along thedirection orthogonal to that of a and c. The specific nature ofthe pore structure (e.g. sponge-like vs. finger-like) depends on thesolvent system used [22] which is proprietary to membranemanufacturers. However, the elongation of the macrovoid struc-tures, seen in Fig. 1b and d, was likely caused by the precipitationof the film as it was introduced to the bath at an angle. A skin layerquickly forms, causing a ‘no slip’ condition and shear within thestill liquid but forming a porous support layer. The macrovoids are‘stretched’ in the direction of the moving film as they form.

While these SEM images clearly show the anisotropy in thepore structure throughout the depth of the membrane they alsopoint out the shortcomings of such a 2D imaging technique.A single SEM image cannot provide a complete representation ofthe anisotropic structure. When comparing Fig. 1a and c (similarlyb and d) it can be seen that the macrovoid density is higher forSW30-XLE in the former set of images but slightly higher forBW30 in the latter set. Non-uniformity in multiple dimensions

Fig. 1. FE-SEM images of the cross-sections of (a and b) BW30 and (c and d) SW30-XLE. Samples were prepared for imaging by freeze-fracturing in liquid nitrogen along twodifferent axial directions. (a and c) Samples freeze-fractured perpendicular to the direction in which the cast polysulfone membrane was introduced into the precipitationbath. (b and d) Samples freeze-fractured in the direction orthogonal to that of a and c.

S.S. Manickam et al. / Journal of Membrane Science 454 (2014) 549–554 551

makes 2D imaging less useful and necessitates the use of ananalytical or imaging technique that captures the 3D structure.

3.2. Mercury intrusion porosimetry (MIP)

Fig. 2 shows the pore diameter distributions of the twomembranes as gathered from MIP. The percent contribution toporosity from smaller pores (1–10 μm) was greater for the BW30than SW30-XLE. In other words, BW30 has more smaller poresthan the SW30-XLE. However, if the entry to a pore is smaller thanits bulk size then this technique exhibits a bias towards thesmaller pore sizes. This effect is termed as the ink-bottle effectand causes histograms to artificially shift to the left (towardsmaller pores). It can also be seen that the large pores contributedgreatly to the overall porosity. The effect was more noticeable inthe SW30-XLE membrane which, by SEM images, showed someevidence of having more macrovoids. The structural metricsobtained from MIP are given in Table 1 and these were used tocalculate the structural parameters shown in Table 2. The propa-gated uncertainties based on the individual parameters have notbeen included in the structural parameters since they were foundto be too small (less than 1/100's). The inherent limitations of thistechnique, such as the ink-bottle effect and possible pore structurecompaction can be avoided in a non-destructive 3D characteriza-tion method, such as imaging using XRM.

3.3. Micro X-ray microscopy (Micro-XRM)

Fig. 3 shows the surface renderings of the 3D micron-scale XRMimages of the BW30 (a) and SW30-XLE (b). The polyamide (PA)layers in both membranes were not visible due to their smallthickness (o100 nm) which was below the �1 μm resolution ofthe instrument. The labels on the image indicate the PSu and PETbacking layers. The structure is bi-continuous with the red regionsindicating the polymer phase and the blue regions corresponding

to the open pore structure. These images can be separated intopore structure phase and polymer matrix phase as indicated bythe images on the right. From these images, porosity and tortuos-ity can be analyzed using the Avizo Fire software package.The porosities and tortuosities of the two membranes calculatedfrom these images are shown in Table 1.

Furthermore, the porosity can be studied as a function of depth.Fig. 4 shows the porosity distribution as a function of membranethickness for the BW30 and SW30-XLE. The resolution of the XRMimages used for this analysis was �0.6 mm. Porosities close to thesurface of the membranes can be analyzed using these images,from which it was seen that the BW30 exhibits a higher ‘nearsurface’ porosity than SW30-XLE. In both membranes, a sharpincrease in porosity was seen at the interface between the PSu andPET layers. This is indicative of the macrovoids that exist at theinterface of the PSu and PET layers (see SEM images in Fig. 1). XRMcan also be used to examine the interface between the two layers.

As with MIP, the structural metrics obtained were used tocalculate the structural parameter using the data gathered fromthe XRM. The S values obtained from XRM were found to besmaller than those from MIP. Both methods determine the samethicknesses, but the measured tortuosity and porosity valuesdiffer. In the case of tortuosity, MIP suggests a bigger differencein the tortuosities between BW30 and SW30-XLE than thatsuggested by XRM. It is to be noted that the value reported byMIP accounts for constriction of the pore diameter along with theincrease in effective pore length whereas the algorithm used forXRM image analysis accounts only for the increase in effectivepore length. It should be noted that the MIP experimental protocolwas set to measure pores only down to 1 μm in order to match theresolution of the XRM. Secondly, the porosities calculated by XRMimage analysis were higher than that fromMIP. This was likely dueto two reasons. First, high intrusion pressures compress the softpolymeric structure, lowering the overall pore volume and causinga negative bias in the measurement [4]. Secondly, mercury

Fig. 2. Pore diameter histograms of (a) BW30 and (b) SW30-XLE from mercury intrusion porosimetry. The average porosities are 26.6374.06% and 36.2075.51%respectively (n¼3).

Table 1Porosity, tortuosity and thickness estimates for BW30 and SW30-XLE from analytical porosimetry and XRM imaging techniques. n¼3 and 2 for porosity and tortuositymeasurements obtained from analytical and imaging techniques, respectively. n¼10 for both thickness measurements.

Analytical (MIP) Imaging (MicroXRM)

Porosity, ε (%) Tortuosity, τ Thicknessa, t (mm) Porosity, ε (%) Tortuosity, τ Thickness, t (mm)

BW30 26.6374.06 1.12070.013 148.376.3 34.9171.94 1.21670.046 142.071.9SW30-XLE 36.2075.51 1.63470.006 151.772.5 43.4971.22 1.31570.164 148.773.0

a This measurement was made using a micrometer.

S.S. Manickam et al. / Journal of Membrane Science 454 (2014) 549–554552

intrudes pores in the shape of a capillary [23] and thus preventsthe entire volume in a pore from being detected. XRM is a non-destructive technique that places no external stresses on thesample and thus could be used to obtain more accurate porosityestimates without compaction. A higher resolution XRM, such asthe UltraXRM™, offers resolution down to 50 nm, and may beused to further examine individual features of an asymmetricor heterogeneous structure with high precision. Fig. 5 shows animage of both membranes taken with this instrument, from whichthe BW30 image captures the elongated macrovoids of the PSumidlayer and the SW30-XLE image captures a macrovoid feature.While these images can render the submicron size pores, theyare not necessarily representative of the complete structure.Nevertheless, these images can be used to study the localizedmicrostructure of heterogeneous pore structures and help under-stand transport at this level.

With regard to structural parameter, one other method hasbeen used to measure its value in membranes like these. Sincethese membranes reject salts, osmotic flux tests combined withmass transfer analysis of boundary layer phenomenon can be usedto calculate an effective structural parameter using Eq. (2).

Table 2Estimates of structural parameter, S (mm) from analytical, imaging and experi-mental flux measurements. The S value was obtained as a fitted parameter fromexperimental osmotic flux measurements.

S (mm) Analytical(MIP)

Imaging(Micro-XRM)

From osmotic fluxmeasurements

BW30 624 489 15,100SW30-XLE 685 402 20,800

Fig. 3. Surface renderings, obtained using Avizo™Fire, of the 3D XRM images of (a) BW30 and (b) SW30-XLE. (a and b) The complete structure of the membrane, where blueregions denote pore space and red regions denote the polymer matrix. These images can be deconvoluted only into pore phase and polymer matrix as shown by the imageson the right. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 4. Porosity distribution as a function of distance through the membrane.The distribution was obtained by using an in-built algorithm used to calculate thefraction of pore space (i.e. number of pore pixels) in each 2-D image constitutingthe 3D volumes shown in Fig. 3a and b. Distance at x¼0 corresponds to the top ofthe polysulfone layer. The resolution of the micro-XRM images used for thisanalysis was �0.6 mm.

S.S. Manickam et al. / Journal of Membrane Science 454 (2014) 549–554 553

This empirical approach has so far been the only means of estimatingS in these materials and is only useable for osmotic membranes,thereby greatly limiting its utility. Structural parameters as measuredby this method are given in Table 2. It is seen that measured S valuesare 1–2 orders of magnitude higher than the values calculated byXRM and MIP, mostly due to a myriad of mass transfer limitations ofthis technique. Poor wetting of a hydrophobic structure, hydrody-namic conditions, and even local mixing (or the absence of it) canimpact these measurements, making them a poor representation ofthe true structural parameter.

4. Conclusions

Internal concentration polarization (ICP) is a major limitationtowards realization of high fluxes and commercialization ofengineered osmosis (EO) processes. The severity of ICP is greatlyinfluenced by the structural parameter of the support layer. In thisstudy, this parameter was measured for two commercial thin filmcomposite reverse osmosis using analytical (mercury intrusionporosimetry (MIP)) and imaging (X-ray microscopy (XRM)) tech-niques. The structural parameter, which could be obtained fromMIP and XRM, was found to differ substantially from that obtainedusing currently used techniques. XRM had the added advantage ofbeing able to measure the porosity as a function of depth. In fact,this type of analysis may be useful as advanced mass transfermodels are developed to predict diffusion in anisotropic struc-tures. No membrane transport model has been developed toincorporate anisotropy into the structural parameter, primarilybecause such intrinsic structural information has until now beenunavailable. The XRM technique can be used to assess the proper-ties of such anisotropic materials and provide insight into thestructure–property relationships in order to better design mem-branes for engineered osmosis.

Acknowledgments

The authors acknowledge financial support from the NationalScience Foundation (CBET-0933553, CBET) and the U.S. Envir-onmental Protection Agency (#R834872). We also wish toacknowledge University of Connecticut Center for Clean EnergyEngineering for use of their porosimeter and the University ofConnecticut Institute of Material Science for use of their FESEM.The authors also wish to thank Dow Water & Process Solutions forproviding the membranes used in this study and Dr. Mike Marsh(VSG) for assistance with the image analysis.

References

[1] M.C.Y. Wong, K. Martinez, G.Z. Ramon, E. Hoek, Impacts of operating condi-tions and solution chemistry on osmotic membrane structure and perfor-mance, Desalination 287 (2012) 340–349.

[2] T.Y. Cath, M. Elimelech, J.R. McCutcheon, R.L. McGinnis, A. Achilli, D. Anastasio,A.R. Brady, A.E. Childress, I.V. Farr, N.T. Hancock, Standard methodology forevaluating membrane performance in osmotically driven membrane pro-cesses, Desalination 312 (2012) 31–38.

[3] J.R. McCutcheon, M. Elimelech, Modeling water flux in forward osmosis:implications for improved membrane design, AIChE J. 53 (7) (2007) 1736–1744.

[4] S. Manickam, S., J.R. McCutcheon, Characterization of polymeric nonwovensusing porosimetry, porometry and X-ray computed tomography, J. Membr. Sci.(2012) 108–115407–408 (2012) 108–115.

[5] S. Mauran, L. Rigaud, O. Coudevylle, Application of the Carman–Kozeny corre-lation to a high porosity and anisotropic consolidated medium: the com-pressed expanded natural graphite, Transp. Porous Media 43 (2001) 355–376.

[6] L. Pisani, Simple expression for the tortuosity of porous media, Transp. PorousMedia 88 (2011) 193–203.

[7] L. Shen, Z. Chen, Critical review of the impact of tortuosity on diffusion,Chemical Engineering Science (2007) 3748–375562 (2007) 3748–3755.

[8] C. Labbez, P. Fievet, A. Szymczyk, C. Simon, A. Vidonne, A. Foissy, J. Pagetti,Hydraulic resistance measurements combined with electrical or diffusionalresistance measurements for determination of pore size in MF membranes,Desalination (2001) 291–299141 (2001) 291–299.

[9] W.C. Maskell, Tortuosity factor in non-homogeneous membranes, Ber. Bun-senges. Phys. Chem. 97 (1993) 680–683.

[10] F.L. Tye, Tortuosity, J. Power Sources 9 (1983) 89–100.[11] Y.-N. Wang, J. Wei, Q. She, F. Pacheco, C.Y. Tang, Microscopic characterization of

FO/PRO membranes – a comparative study of CLSM, TEM and SEM, Environ.Sci. Technol. 46 (2012) 9995–10003.

[12] T.Y. Cath, A.E. Childress, M. Elimelech, Forward osmosis: principles, applica-tions, and recent developments, J. Membr. Sci. 281 (2006) 70–87.

[13] S. Zhao, L. Zou, C.Y. Tang, D. Mulcahy, Recent developments in forwardosmosis: opportunities and challenges, J. Membr. Sci. 396 (2012) 1–21.

[14] J.T. Arena, B. McCloskey, B.D. Freeman, J.R. McCutcheon, Surface modificationof thin film composite membrane support layers with polydopamine:enabling use of reverse osmosis membranes in pressure retarded osmosis,J. Membr. Sci. (2011) 55–62375 (2011) 55–62.

[15] Hydration Technology Innovations, cited, Available from: http://www.htiwater.com/, 2013.

[16] McGinnis, R., G. McGurgan, Forward Osmosis Membranes, WO Patent WO/2011/028,541, 2011.

[17] S.H. Lau, K.S.C. Wilson, G. Fernando, C. Hauyee, T. Andrei, F. Michael,Y. Wenbing, Non invasive, multiscale 3D X-ray characterization of porousfunctional composites and membranes, with resolution from mm to sub50 nm, J. Phys.: Conf. Ser. 152 (2009) 012059.

[18] S.C. Carniglia, Construction of the tortuosity factor from porosimetry, J. Catal.102 (1986) 401–418.

[19] Micromeritics, AutoPoreIV Manual, 2013.[20] A. Tkachuk, F. Duewer, H. Cui, M. Feser, S. Wang, W. Yun, X-ray computed

tomography in zernike phase contrast mode at 8 keV with 50-nm resolution usingCu rotating anode X-ray source, Z. Kristal. – Cryst. Mater. 222 (2007) 650–655.

[21] D. Gostovic, J.R. Smith, D.P. Kundinger, K.S. Jones, E.D. Wachsman, Three-dimensional reconstruction of porous LSCF cathodes, Electrochem. Solid-StateLett. (2007) B214–B21710 (2007) B214–B217.

[22] M. Mulder, Basic Principles of Membrane Technology, Second edition, KluwerAcademic Publishers, Norwell, MA, USA, 1996.

[23] H. Giesche, Mercury porosimetry: a general (practical) overview, Part. Part.Syst. Charact. 23 (2006) 9–19.

Fig. 5. 3D representations of the polysulfone matrix of (a) BW30 and (b) SW30-XLE from nano-scale XRM (Xradia UltraXRM). (a) The presence of long macrovoids in thematrix and (b) the surface rendering of a single macrovoid showing the presence of pores along the macrovoid wall.

S.S. Manickam et al. / Journal of Membrane Science 454 (2014) 549–554554