Structural Characterization of Thin-Film Polyamide Reverse Osmosis Membranes

Structural Characterization of Thin-Film Polyamide Reverse OsmosisMembranesJonathan Albo,† Hideaki Hagiwara,‡ Hiroshi Yanagishita,‡ Kenji Ito,‡ and Toshinori Tsuru†,*†Department of Chemical Engineering, Hiroshima University, 1-4-1 Kagayami-yama, Higashi-Hiroshima 739-8527, Japan‡National Institute of Advanced Industrial Science and Technology (AIST), 1-1-1 Higashi, Tsukuba 305−8565, Japan

*S Supporting Information

ABSTRACT: This study aims to explore the structural characteristics of the inhomogeneous top layer of thin-film compositemembranes when pretreated by different methods: room temperature−oven, ethanol−hexane in a solvent exchange process, andfreeze-drying. An evaluation of the nano-order free-volume pore size of the polyamide samples was carried out bynanopermporometry (NPP) and was quantitatively compared with the free-volume pore estimated from normalized Knudsen-based permeance (NKP) and with positron annihilation characterization (PALS). NPP results denoted a bimodal polyamidemembrane structure described by a dense matrix and highly permeable regions. The application of different condensable vapors(water, hexane, and isopropanol) resulted in a free-volume pore size smaller than dp = 0.6 nm for dense regions, which wasconfirmed after NKP and PALS. In addition, the influence of highly permeable regions on permeance decreased in the followingorder: ethanol−hexane > freeze-drying > room temperature−oven samples, demonstrating an effective membrane structurealteration after different pretreatments.

1. INTRODUCTION

The use of thin-film composite membranes (TFC) for reverseosmosis (RO) processes has been widely extended due to theiradvantageous high flux and rejection provided by the thinaromatic polyamide (PA) separating layer. Most commerciallyavailable RO TFC membranes are formed in situ by theinterfacial polymerization of an aromatic polyamine such as m-phenylenediamine (MPD) with one or more aromatic polyacylhalides (for example, trimesoyl chloride (TMC)). Thesechemical and mechanical resistant aromatic-based membranes1

exhibit excellent performance in many desalination and waterpurification applications and are already in mass production.2,3

However, for an optimized separation performance, a clearunderstanding of PA membrane characteristics at not auniquely macroscopic level (physical and mechanical proper-ties) but at a nanoscale one (local spaces and distribution) isdemanded.The concept that stipulates that it is the local spaces (free

volume) around the permeating molecule that determine thediffusion coefficient is the key to understanding diffusion inpolymer membranes. Polymers are normally divided into twobroad categories, rubbery and glassy, where transport can bedescribed by a solution-diffusion model. In a rubbery polymer,the polymer chains can rotate more freely, making the polymersoft and resulting in higher diffusion coefficients. On the otherhand, in a glassy polymer, steric hindrance along the polymerprohibits the rotation of the polymer segments, resulting in arigid polymer with low diffusion coefficients. However, due toan extraordinarily high and interconnected free volume, somepolymeric membranes, such as PIM (polymer intrinsicmicroporous), can also act as porous materials with poresranging from 0.5 to 1.5 nm.2,4,5 In this case, pore-flow transportoccurs through the so-called free-volume pore spaces of themembrane and separation can be primarily explained by

molecular sieving. Furthermore, membrane structures withsmall (0.42 nm) and also larger (1.2−1.4 nm) free-volumediameter elements have been reported for PTMSP high-free-volume polymer,6 where a crossover from solution-diffusion toKnudsen transport seems to occur. As a rule of thumb, thetransition from solution-diffusion and pore-flow transportmechanism is in a diameter range of 0.5−1 nm2. In this regiona dual-mode transport model is the most appropriatedescription for permeances and selectivities.7,8

Positron annihilation lifetime spectroscopy (PALS) hasdrawn much attention recently in polymeric membraneresearch due to its great ability to explore free-volume holesat the molecular level.9−12 The application of PALS to PA-based-TFC RO membranes has resulted in free-volume holesizes ranging from 0.4 to 0.8 nm in diameter.13−15 Therefore, inPA membranes with subnano holes, the transition betweenpore-flow and solution-diffusion transport seems to occur in aporous−nonporous material.In our previous study, gas permeation results revealed that

the dry PA layer consisted of a dense matrix where chainmobility with temperature enabled permeation by the activateddiffusion of small gases, such as He, and highly permeableregions where larger species, such as N2, could permeateexclusively via the Knudsen mechanism.16 Lately, in thefollowing work, it was found that the transport of anisopropanol (IPA)/water mixture in pervaporation alsoresponded to a two-region structure, where the free volumearising from the wetting of the dense and highly permeablepolymer chains, together with the higher affinity of IPA

Received: October 10, 2013Revised: December 19, 2013Accepted: January 7, 2014Published: January 7, 2014

Article

pubs.acs.org/IECR

© 2014 American Chemical Society 1442 dx.doi.org/10.1021/ie403411w | Ind. Eng. Chem. Res. 2014, 53, 1442−1451

pubs.acs.org/IECR

molecules and PA, defined the separation performance.17 Bothstudies resulted in a comparatively high flux and selectivity anddemonstrated the applicability of the composite PA membranesfor high-temperature separation processes. Additionally, theprevious results suggested that different pretreatments for themembrane could alter gas and vapor permeances. Conse-quently, for further discussion of the effect of membranepretreatments, free-volume pore size and the distribution in thearomatic PA membrane structure would enhance the under-standing of the dominated transport mechanism.Nanopermporometry (NPP) is a methodology that typically

has been used to measure pore sizes of less than 50 nm.18,19

This methodology offers pore size and distribution evaluation,similarly to bubble point, where the permeate gas flow ratethrough a wet porous membrane is measured by increasing thepressure difference across the membrane,20 and biliquidpermporometry, in which a liquid is used to displace anotherliquid from a porous membrane.21 One of the advantages of thepresent NPP methodology is the ability to measure nanosizedpores ranging from 0.6 to 30 nm, while bubble-point andbiliquid permporometry have been used to characterize poresizes in the microfiltration (larger than 100 nm) andultrafiltration (10−100 nm) ranges, respectively. The type ofvapors applied in NPP definitely affected the membranestructural characterization due to different molecule sizes,polarity, and interactions with the membrane surfaces. To date,water, alcohol, cyclohexane, or carbon tetrachlorides have beenmainly applied.22−24 A previous work suggested that the vaporsof water and nonpolar compounds were appropriate formeasuring free-volume pore size distributions smaller than 1nm in microporous ceramic membranes, in contrast to theutilization of vapors with a larger molecular size and someaffinity to the membrane material.19 A clear understanding ofthe effect of vapors applied on NPP and their use for free-volume pore size determination has not been fully achieved forpolymeric materials.In the present study, NPP is used for the first time to

examine the membrane structure and the free-volume pore sizeof PA-based RO membranes under different pretreatmentprocedures. The results are compared quantitatively with free-volume pore sizes estimated from normalized Knudsen-basedpermeance (NKP), which is based on the permeation of gasmolecules and their kinetic diameter,25 and from PALS, whichrelies on the diffusion of a gas into the surface accessible spacesin the membrane structure. Finally, the separation performanceof the PA-based membranes is evaluated for water permeabilityand salt rejection. The results were compared to gaspermeation and discussed in terms of the membrane localspaces. The present study will aid in an understanding of the

transport mechanisms for compounds in the bimodal structureof PA membranes for the development of high-performancemembrane separation processes.

2. EXPERIMENTAL SECTION

2.1. Materials. CPA5, high-rejection RO membranes werekindly provided by Nitto Denko (Osaka, Japan) and adopted inthe present study. The membrane consisted of a TFC with atop-skin aromatic PA layer (∼200 nm), a middle microporouspolysulfone (∼40 μm), and a bottom poly(ethylene tereph-thalate) layer (∼120 μm). The specific chemical composition ofthe PA layers is proprietary information of the supplier.Membranes were delivered in a polyethylene bag containingless than 1% sodium m-bisulfite solution and were kept in arefrigerator at 4 °C before the analysis.All high-purity chemicals of analytical grade used in this

study were purchased from Sigma Aldrich (Tokyo, Japan).2.2. Membrane Pretreatment Methods. Commercial PA

membranes were tested immediately after common membranepretreatment procedures. A detailed description of theprocedures is shown in the Supporting Information (Appendix1).(1) Room temperature−oven (RTO): Membranes were dried

at room temperature and then in an oven at 120 °C.(2) Ethanol−hexane (EH): Membranes were dried in

ethanol−hexane by a solvent exchange process.(3) Freeze-drying (FD): Membranes were in tert-butanol and

dried in freeze-drying equipment.2.3. Membrane Sorption. Sorption experiments were

performed at room temperature (25 ± 2 °C) in a ShimadzuTGA-50 apparatus with a sensitivity of ±0.001 mg. A N2 flowrate of 50 mL/min was introduced in the TG equipment afterbubbling through a humidifier containing pure water, hexane,or IPA. Prior to the measurement, membranes were pretreatedunder RTO procedure to remove the sorbed water from withinthe membrane structure.The sorption uptake, S, in the PA sample was gravimetrically

measured and calculated as follows:

=−

×⎛⎝⎜

⎞⎠⎟S

m mm

100s d

d (1)

where ms and md are the weight of the sorbed and drymembranes, respectively.

2.4. Contact Angle. The membrane wettability testing wascarried from sessile water drops using a goniometer equippedwith a camera device (Kyowa, Tokyo, Japan) at roomtemperature (25 ± 2 °C) in air. Contact angles were measured

Figure 1. Schematic drawing of the NPP experimental setup.

Industrial & Engineering Chemistry Research Article

dx.doi.org/10.1021/ie403411w | Ind. Eng. Chem. Res. 2014, 53, 1442−14511443

by defining a circle around the drop and recording the tangentangle formed at the substrate surface.2.5. Nanopermporometry. Free-volume pore size in the

membrane samples (2.21 cm2) was tested in a NPPexperimental apparatus, as schematically represented in Figure1.He and N2 were used as noncondensable gases, while the

liquids used as condensable vapors were water, hexane, andIPA. The gas flow rate from the cylinder was regulated by amass flow controller (STEC, Kyoto, Japan). The feed was atatmospheric pressure, and the pressure difference across themembrane ranged from 10 to 20 kPa, monitored by a pressure-difference sensor (Sunx, Tokyo, Japan). The temperature of thehumidifier was controlled at 40 °C, while the apparatus was atroom temperature (25 ± 2 °C). Prior to testing, He was fedthrough to remove the vapor inside the system. NPP wasinitiated by measuring the steady permeance of the dry gas as areference initial point. The vapor pressure was then increasedgradually by controlling dry gas, Qd, via MF-1, and wet gas, Qw,via MF-2 mass flow controllers, until gas permeation wasblocked by capillary-condensation or adsorption-inducedswelling.The partial pressure of vapor in the feed stream, p, was

determined based on the flow rate of the dry gas, Qd, and thewet gas, Qw, and the total pressure of the feed stream, pT, whichis the sum of atmospheric pressure, p0, and the pressuredifference, Δp, across the membrane assuming that the liquidwas under saturated vapor pressure, ps, after the mist trap (5 inFigure 1), and that the vapor pressure was sufficiently lowcompared with pT:

=+

pQ

Q Qpw

w ds

(2)

Gas permeance, Pi, was calculated using the following equation:

=Δ

PQ

pAii

(3)

where Qi is the permeate flow rate and A is the membrane area.In a small free-volume pore size (diameter dp), vapor

sorption and condensation at vapor pressure, p, lower than ps,occur, as it is represented in the Kelvin equation:

υ σ θ=⎛⎝⎜

⎞⎠⎟RT

PP r

ln 2cos

s k (4)

where υ, σ, and θ are the molar volume, surface tension, andcontact angle, respectively. This equation permits calculation ofthe Kelvin radius, rk.2.6. Positron Annihilation Lifetime Spectroscopy.

Positron annihilation lifetime measurements were carried outat various positron incident energies, E, ranging from 1.8 to 10keV by utilizing a 22Na-RI-based pulsed-positron beam system(PALS-200A Fuji Imvac, Yokohama, Japan) with a timeresolution of approximately 290 ps full width at half-maximumat the prompt peak of the obtained data. The lifetimes ofpositrons were recorded as the time difference between apulsing trigger and the corresponding detection timing of theannihilation radiation, and the annihilation events wereaccumulated with total counts of 2.0 million. Multiexponentialanalysis was applied to the recorded lifetime data to deduce theaverage lifetime of the long-lived positronium (Ps), τPs, for thefilms. In the analysis, care was taken of the background

component for the lifetime data by using that for a Kapton foil.An average hole radius rPs (nm) was calculated from τPs (ns)based on the following equation:9−12

τπ

π= −

++

+

−⎡⎣⎢⎢

⎛⎝⎜

⎞⎠⎟⎤⎦⎥⎥

rr

rr

0.5 10.166

12

sin2

0.166PsPs

Ps

Ps

Ps

1

(5)

2.7. RO Experiments. LP and salt rejection, R, were testedin a stainless dead-end cell in a single-solute system. Theeffective membrane area was 2.21 cm2. Rejection and waterpermeance in single-solute systems were measured through thepretreated PA-based membranes and calculated according tothe following equations:

= ΔΔ Δ − ΔΠ

Lv

tA p( )P(6)

= −RCC

1 P

f (7)

where Δv is the permeate volume, Δt is the permeation time, Ais the effective membrane area, Δp is the operation pressure,and ΔΠ is the osmotic pressure difference. In eq 7, CP, is thepermeate concentration and Cf is the feed concentration. Inorder to maintain constant ΔΠ, across the membrane over thecourse of the experiment, the feed solution was changedperiodically after the measurements (every 30 min). Theosmotic pressure was calculated from NaCl Cf. Waterpermeances and rejection values were stable after 2 h ofexperimental time.The operation pressure was set at 1.5 MPa, and temperature

was maintained at 25 °C in a water bath. The values wereobtained for a NaCl concentration of 2,000 ppm (wt) indeionized water. Membranes were immersed in the 2,000 ppmNaCl aqueous solution and then set in the dead-end cell priorto use. The electric conductivity was evaluated using an ES-51conductivity meter (Horiba, Kyoto, Japan).

3. RESULTS AND DISCUSSION3.1. Nanopermporometry Results. In our previous work,

gas permeation tests demonstrated that the separationcharacteristics of the TFC RO commercial membranes weremainly attributed to the top aromatic PA layer, which consistedof two different structures.16 First, there was a dense matrix thatenabled the permeation of small gases, such as He. Second,there were highly permeable regions where larger species, suchas N2, could permeate via Knudsen mechanism.16 In order toestimate these different free-volume pore sizes, He and N2 havebeen selected for NPP as noncondensable gases.Figure 2 shows the NPP experimental curve for a CPA5-

RTO sample. He and N2 dimensionless permeances,normalized with the lowest humidity, are plotted as a functionof the relative water humidity in the feed stream (p/ps). A detailof the experimental procedure is shown in the SupportingInformation (Appendix 2).As observed, the permeance of He was decreased at a p/ps <

20%, while N2 remained invariable. After p/ps = 90%, thepermeance of both gases rapidly decreased, which can beexplained as water vapor partial condensation and/or sorptionin the induced-swelling membrane at higher humidity, blockingthe permeation of the noncondensable gases (both He and N2).The two decreasing regions in the He curve may denote abimodal structure described by two different free-volume pore



size regions. The bimodal structure is consistent with ourprevious results on gas separation16 and pervaporation/vaporpermeation.17 N2 (0.36 nm) is able to permeate only throughhighly permeable regions (large free-volume pores), and itspermeation is not reduced at low relative humidities.Alternatively, He (0.26 nm) first showed a decrease inpermeation because vapor blocked local spaces in the densematrix (available for permeation of small gases, such as He),which was followed by a second decrease after p/ps = 90%,where highly permeable regions can be partially plugged. Thisbimodal structure is in agreement with the findings for PTMSPhigh-free-volume polymer, with small and larger free-volumeelement diameters.6

3.2. NPP Using Different Solvents As CondensableVapors. The molecular size of vapors and their interactionwith the membrane may affect the free-volume pore sizemeasurements.19 Figure 3 represents the Kelvin diameter, dk, of

the applied condensable vapors (water, hexane, and IPA) as afunction of p/ps. The curves are defined by the molar volume,surface tension, and contact angles of every vapor at the samep/ps, eq 4. The physicochemical properties of the solvents arepresented in Table 1. A higher sorption rate for IPA wasobtained, probably because the solubility parameter of IPA is

8.8, which is close to the solubility parameter value of thearomatic PA.26 It should be noted that the IPA sorption uptakeobtained in this work, S = 46.4%, is in good agreement withthat value obtained in the literature for TFC PA-basedmembranes, S = 49.5%, where attenuated total reflection−Fourier transform infrared (ATR-FTIR) analysis suggested thepreferential sorption of alcohols in the PA top layer.26 The PAsorption uptake with time and water contact angle view ispresented in the Supporting Information (Appendix 3).Generally speaking, and according to the figure, hexane and

IPA may be candidates for measurements of larger free-volumepores in the membrane, while water can be used to determinesmall spaces. It should be noted that the Kelvin equationcannot be applied for a pore diameter of less than 2 nm.However, the measurement by NPP showed a reasonablecorrelation with the separation performances of the porousmembranes having pore sizes as small as 0.5 nm,19 and,therefore, this may reveal the probable structural characteristicsof the PA membrane, which is the main objective of the presentwork.Figure 4 shows the dimensionless permeance curve for He as

a function of p/ps when using water, hexane, and IPA ascondensable vapors.

These results show how the application of water and hexaneas condensable vapors resulted in similar tendencies, despitetheir differences in hydrophobicity and polarity. Alternatively,IPA was able to wet the membrane and completely block thegas permeation at p/ps = 20%.The nature of physisorption processes of condensable vapors

is usually divided into the sorption of vapor molecules to thefree-volume pore wall at low vapor pressure and capillarycondensation afterward.24 In the case of water as a condensablevapor, since the molecular size is relatively small compared with

Figure 2. Dimensionless permeance as a function of relative humidity(p/ps) for a CPA5-RTO membrane (vapor: water).

Figure 3. Kelvin diameter as a function of relative humidity (p/ps) forthe three solvents applied.

Table 1. Physicochemical Properties of the Solvents

vapor M (g/mol)aσ(30°C)(mN/m)b

kineticdiameter (nm) θ (deg)c S (%)d

water 18.01 71.2 0.3 52 12.1hexane 86.18 18.4 0.51 0 20.8IPA 60.09 23.1 0.47 0 46.4

aM = molecular weight. bσ = surface tension. cθ = contact angle. dS =sorption uptake.

Figure 4. He dimensionless permeance as a function of relativehumidity (p/ps) with water, hexane, and IPA as condensable vapors inCPA5-RTO.



hexane and IPA, and there is a reduced affinity to PAmembranes, a reduction in dimensionless permeance maydirectly indicate the free-volume pore size distribution of themembrane. Hexane, which is hydrophobic and nonpolar,presented a sorption value that was similar to that of water.The size of hexane molecules after sorption and condensation isthought to be very thin, resulting in a dimensionless permeancecurve that is similar to that of water.19 IPA, however, with akinetic diameter of 0.47 nm and the highest degree of sorptionuptake for a PA membrane, S = 46.4%, is expected to be sorbedinto the wall of the highly permeable region and onto themembrane surface at a low relative pressure, blocking thepermeation of the noncondensable gas.Finally, Figure 5 shows the dimensionless permeance of He

as a function of dk (eq 4) for the three condensable vapors

applied. The curves may correspond to a free-volume pore sizedistribution curve.As shown in the figure, IPA presents a lower dimensionless

permeance than either water or hexane at the same Kelvindiameter. No He permeation at dk > 1.7 nm (corresponding top/ps = 20%) was observed when using IPA, while water andhexane still showed permeation at p/ps = 90%. Consequently,water and hexane as condensable vapors in NPP are able topartially plug membrane permeable regions and reduce the

permeation of He and N2, giving a clear difference between thetwo membrane structures: dense and highly permeable regions.IPA effectively reduces the permeation of the noncondensablegas, and the permeance reduction is more likely related to thesorption of the vapor in the material, instead of to capillarycondensation.In any case, the application of the three condensable vapors

showed an important decrease in dimensionless permeance upto dk = 0.6 nm and then established a stable value. Therefore,permeation of He at dk < 0.6 nm can be assumed to correspondto free-volume pores in the dense regions of the membrane,while dk > 0.6 nm may be associated with highly permeableregions.

3.3. Effect of Membrane Pretreatments on NPP.Membrane pretreatment prior to use can produce shrinkageand swelling of the PA membrane structure.16,17 Thus, it isimportant to understand the effect of pretreatment procedureson the free-volume pore size, since it determines theperformance characteristics.Figure 6 shows the dimensionless permeance of He (a) and

N2 (b) for the membrane after different pretreatments (RTO,EH, and FD). Water is used as condensable vapor, since lowsorption in the PA material is expected (S = 12.1%) and maygive clearer information on the local space distribution of thetwo membrane regions.As shown in Figure 6a, the initial He permeation was reduced

for RTO and FD samples with increasing p/ps of up to 20%.This relative humidity may be enough to block densemembrane permeable spaces, thereby limiting He permeation.However, in the EH sample, He permeance did not vary beforep/ps = 90%. This may indicate that gas permeation throughsamples treated under ethanol−hexane is controlled by thefree-volume pore characteristics of the highly permeableregions. On the other hand, N2 (Figure 6b) remained in astable value before p/ps = 90%, with slight differences betweenmembranes pretreated under different procedures. Besides,RTO showed the smallest dimensionless permeance at the endof the test, in comparisons with FD and EH, which is attributedto the reduced number and size (width and length) of highlypermeable regions of the PA selective layer after drying at hightemperature.Figure 7 graphically represents the transport of He and N2

through the dense and highly permeable membrane regions.

Figure 5. He dimensionless permeance as a function of Kelvindiameter with water, hexane, and IPA as condensable vapors forCPA5-RTO.

Figure 6. Dimensionless permeance of He (a) and N2 (b) in samples after different pretreatments (RTO, EH, and FD) as a function of relativehumidity (p/ps) for water as a condensable vapor.



The transport through these subnano free-volume pores of PAmembranes may be described by size exclusion (molecularsieving), solubility differences (solution-diffusion) and Knudsendiffusivity in the local spaces (free-volume pores) of theporous−nonporous membrane.The relative contribution of dense and highly permeable

regions to overall He permeance, PHe, is presented in Table 2.

The calculated values were based on dry gas permeation (p/ps= 0%) results in NPP. The estimation assumes that Hepermeated both membrane regions, while N2 permeated onlyhighly permeable regions via Knudsen diffusion.16 Therefore,the permeation of He through highly permeable regions (largefree-volume pores) can be estimated according to Knudsendiffusion mechanism:

=P PM

MHe,HR NN

He2

2

(8)

where PN2is the N2 permeance and M is the molecular weight

of the gases. The relative contribution of highly permeableregions (HR) on He permeation can be calculated as the ratioof He permeation through highly permeable regions, PHe,HR,and overall He permeation in the membrane, PHe, according tothe following equation:

= ×P

P

M

MHR/% 100N

He

N

He

2 2

(9)

The relative contribution of dense regions (DR) on Hepermeation was calculated as the difference in the overallpermeation. The results may reveal the contribution of each

membrane region to the overall He permeation and give anapproximation of the effect of different pretreatments in themembrane structure.As shown in Table 2, the RTO procedure increased the

relative contribution of dense regions on gas permeance bycomparison with the EH and FD procedures, which can beattributed to the membrane shrinkage that occurred afterdrying at high temperature.16 In the EH procedure, alcohol mayhave swelled the PA chains26 and removed the pluggingcompounds (from membrane preparation) from withinpermeable regions;27 both effects would have resulted in themembrane permeation being controlled by the highlypermeable regions. These results were consistent with Figure6a and confirmed the bimodal membrane structure. In the FDprocedure, the membrane morphology is fixed with aminimized distortion that may have occurred during normaldrying, and thus low membrane shrinkage was expected,28

which resulted in an intermediate influence of highly permeableregions on gas permeation.The averaged free-volume pore size in the samples may be

reduced in the following order: EH > FD > RTO. Generally,membranes with larger average free-volume pore sizes wouldalso have higher permeances, and low permeation discrim-ination to different molecule sizes, resulting in lower separationfactors, as observed with the He/N2 gas selectivity tendency inthe samples (Table 2).

3.4. Normalized Knudsen-Based Permeance. Tofurther explore the PA membrane structure, a normalizedKnudsen-based permeance was then applied. NKP is a simpleand effective method25 to evaluate the average free-volume poresize based on the modified gas-translation model originallyproposed by Xiao and Wei29 and Shelekhin et al.30

NKP is defined as the ratio of the permeance of component i,Pi, to that calculated using He permeance and molecular weight,M, under the Knudsen diffusion mechanism:

=P

PNKP i

MMHe

i

He (10)

NKP can be analyzed using the kinetic diameter of permeatingmolecules, dk, as follows:

=−

−

d d

d dNKP

(1 / )

(1 / )ik, p

3

k,He p3

(11)

For an estimation of dp, first, the NKP for each gas can bemeasured experimentally according to eq 10 and then dp can beobtained by fitting NKP as a function of dk,i, eq 11.According to gas permeation values from our previous

work16 presented in Table 3, Figure 8 shows gas permeancesreported for RTO, EH, and FD membranes, as a function of gaskinetic diameters. The continuous curves show predicted

Figure 7. He and N2 permeation through dense and highly permeablemembrane regions of the membrane.

Table 2. Relative Contribution to Gas Permeance of theMembrane Regions

P (10−8 mol/(m2·s·Pa))a

relative Hepermeation

contribution (%)

CPA5 pretreatment He N2 αHe/N2

b DRc HRd

RTO 4.97 1.28 3.88 31.8 68.2EH 21.3 7.58 2.81 5.7 94.3FD 12.5 3.66 3.41 23.6 77.4

aP = permeance. bα = gas selectivity for He/N2.cDR = dense region.

dHR = highly permeable region.

Table 3. Gas Permeation through the Membranes afterDifferent Pretreatments

P (10−8 mol/(m2·s·Pa))a

pretreatment He H2 CO2 O2 N2 C3H8 SF6

RTO 4.79 5.07 0.89 1.03 1.11 0.89 0.47FD 11.8 13.5 2.75 3.24 3.53 2.78 1.46EH 20.8 29.7 6.38 7.72 7.92 6.47 3.44

aP = permeance.



permeances (Pi = PHe(Mi/MHe)1/2), based on He permeance

under the Knudsen diffusion mechanism, since He is thesmallest molecule with no sorption properties.The experimentally obtained permeances agreed well with

the predicted ones in the EH sample (Figure 8b) and deviatedfor the larger molecules in the cases of RTO and FD.Furthermore, the figure shows the initial decreases in the NKPvalues at lower kinetic diameters for the RTO and FD samples,probably due to the influence of molecular sieving in denseregions. These results may indicate that small gases, such as Heand H2, are able to permeate through dense matrix and highlypermeable regions, compared with larger species, such as N2,that permeate exclusively via Knudsen mechanism throughlarger free-volume pores.16 On the other hand, the constantNKP value for the EH sample may indicate that the transportwas controlled by Knudsen diffusion through highly permeableregions, in accordance with the NPP results (Figure 6 andTable 2). Therefore, the averaged free-volume pore size in thesamples may be reduced in the following order: EH> FD>RTO.Additionally, Figure 9 shows normalized Knudsen-based

permeance as a function of molecular size for the sample afterthe different drying procedures. The dotted line represents theNKP calculations based on eq 11 using dp = 0.6 nm, and it isincluded as a reference. The ratio of experimentally obtainedpermeance to Knudsen-based predicted permeance indicatesthe degree of permeance reduction in dense regions in thefollowing order: RTO > FD > EH. This is inverselyproportional to the relative contribution to gas permeance ofhighly permeable regions: EH > FD > RTO (Table 2).From the experimental curves, a bimodal structure of the PA

membrane was confirmed. The dense regions may consist of astructure with free-volume pore sizes smaller than dp = 0.6 nmaccording to NPP and NKP results, which is in good agreement

with free-volume pore size diameters ranging from 0.4 to 0.8nm based on PALS characterization13−15 for reverse osmosisPA-based thin-film composite membranes.

3.5. Positron Annihilation Lifetime Spectroscopy.Figure 10 shows the variation of the Ps lifetime, τPs, and theaverage diameter, 2rPs, calculated from eq 5, for the membranesafter the different pretreatments (RTO, EH, and FD) as afunction of positron incident energy E.The overall tendency was similar for all samples, that is, τPs

increased from ∼1.8 to ∼2.2 ns with increasing E from 1.8 to10 keV. The obtained τPs is in agreement with that observed fora NF membrane with similar chemistry (1.93 ns at E = 2.0 keV)reported previously.31 By considering the positron meanimplantation depth, the outer layer for each membrane shouldbe associated with an E range between 1.8 and 3 keV, giving anaverage diameter below 0.6 nm. A comparison of rPs among thethree membranes signifies that the hole size in the outer layerwas smaller in the order of EH > FD > RTO, which is

Figure 8. Gas permeance as a function of kinetic diameter. Continuous line shows predicted permeance using Pi = PHe(Mi/MHe)1/2. Points are

experimental data: (a) RTO, (b) EH, and (c) FD.

Figure 9. NKP as a function of kinetic diameter. Points areexperimental, and the reference curve is calculated based on eq 11using dp = 0.6 nm (dotted line).



consistent with the above argument. Furthermore, the obtainedsize range, i.e., the diameters smaller than dp = 0.6 nm, agreedwell with that from the permeance properties (NPP and NKP).3.6. Effect of Membrane Structure on RO Perform-

ance. In order to correlate the membrane structure with theseparation ability, a CPA5 membrane was evaluated for waterpermeance and salt rejection. The RO performance may offercriteria for evaluating the free-volume pore size distribution ofmembranes pretreated differently. Membrane samples weretreated under RTO, EH, and FD procedures and comparedwith samples without pretreatment (no pretreatment, NP),which were washed only in pure water. Table 4 summarizes the

comparison of water permeance, LP, with the permeance of dryHe in NPP (Table 2). The smaller the He permeance, thesmaller the water permeance through the membrane, whichshows the correlation between the permeation properties of themembrane in RO and gas permeation after different pretreat-ments.Figure 11 presents LP and R for the CPA5 (high-rejection

membrane). The results are compared with a SWC5 seawaterPA membrane (Nitto Denko), which is considered to possess amore rigid structure for high-pressure applications,16,17 and,therefore, the membrane structure is expected to be affectedless by different pretreatments.As observed, SWC5 showed no remarkable differences in

water permeance after the different procedures, which may beattributed to the high degree of chain rigidity, which accounts

for an elevated molecular packing and therefore a smallvariation in the membrane structure after pretreatment.Alternatively, water permeance in the CPA5 sample wasincreased by factors of 1.94 and 1.29 for EH samples comparedwith those permeances in RTO and FD samples, respectively,confirming that different pretreatments effectively producedalterations in desalination performance. Nevertheless, waterpermeance for samples dried under the three procedures wasbelow those permeances obtained in samples that were notdried, NP. This can be attributed to the negative impact of thedrying procedures on membrane−water interactions due to thedehydration of water-swollen hydrogel that fills the membranepores.17 Upon immersion of dried samples in the aqueoussolution, water may be unable (or only slowly able) to accessinterchain hydrogen bonds formed during drying. Besides, aprevious report probed, via PALS and water transport tests indried and hydrated poly(arylene ether sulfone) polymers, thatsorption of water in dried samples may alter the free-volumepore sizes of the polymer. Water molecules may occupy the freevolume of the dried membrane, reducing the cavity-volumesize.32 As a result, water cannot easily permeate polymer chains.As expected, slightly higher rejections were observed from

Figure 11 when using CPA5 (high-rejection RO membrane)compared with SWC5 (seawater). In addition, despite thedifferences in LP, produced by the influence on membraneperformance of the different pretreatments, R remained almostinvariable for each membrane type, with only slight decreases insamples dried at room temperature−oven. Thus, clearly bothdense and highly membrane permeance regions of themembrane (detected by NPP, NKP, and PALS) are effectivefor NaCl rejection. If cracking of the membranes upon dryingwould occur, a remarkable alteration of rejection would beexpected. However, this was not observed. The stable NaClrejections, despite the differences in water permeance, may alsodenote the effective separation of NaCl by molecular sieving inthe pretreated samples. Then, the results may suggest that thestructure of the dried PA layer can be representative of thestructure in the hydrated state.In addition, the rejection in RO is not uniquely described by

the sieving effect, but also for membrane-charged functiongroups33,34 due to the presence of carboxylic groups generatedby the unreacted groups of TMC from membrane synthesis.35

Consequently, the pretreatments applied might not alter thechemically charged structures that are effective for a NaCl size-exclusion mechanism.

Figure 10. Variation of τPs for the membranes with the differentpretreatments (RTO, EH, and FD) as a function of positron E.Average diameter 2rPs obtained from τPs using eq 5 is shown on theright-hand axis. On the upper axis, positron mean implantation depthwas calculated from 40/ρE1.6 with ρ = 1 g/cm3.

Table 4. Comparison of Gas and Water Permeance in ROafter Different Pretreatments

CPA5 pretreatment PHe (10−8 mol/(m2·s·Pa))a LP (L/(m

2·h·bar))b

RTO 4.97 0.82 ± 0.15EH 21.3 1.59 ± 0.12FD 12.5 1.23 ± 0.09NP 4.37 ± 0.31

aP = permeance. bLP = water permeance.

Figure 11. CPA5 and SCW5 membrane performance in sampleswithout pretreatment (no pretreatment, NP), and after RTO, FD, andEH procedures (25 °C, 1.5 MPa, and 2,000 ppm).



In short, the water permeance results from RO testsconfirmed the average free-volume pore size tendency observedfrom NPP, NPK, and PALS (EH > FD > RTO) anddemonstrated that different membrane pretreatments mayaffect the characteristics of dense matrix and highly permeableregions in the bimodal PA structure.

4. CONCLUSIONIn this work, polyamide membranes were pretreated and theirfree-volume pore sizes were estimated by nanopermporometry(NPP), normalized Knudsen-based permeance (NKP), andpositron annihilation lifetime spectroscopy (PALS). The mainconclusions of the work are as follows.(1) In NPP, the permeance of He was decreased at p/ps =

0−20%, while N2 remained invariable. At p/ps = 90%, thepermeance of both gases rapidly decreased. The two decreasingregions in the dimensionless permeance curve denoted abimodal membrane structure, which consisted of dense andhighly permeable regions. The application of water as acondensable vapor, with a comparatively small molecular sizeand a reduced affinity to polyamide materials, indicated thefree-volume pore size distribution of the membrane. Thereduction in He permeance for the three applied condensablevapors (water, hexane, and isopropanol) at dp = 0.6 nm mayindicate the size of the local space in dense regions, which wasconfirmed after NKP tests.(2) The application of different pretreatment procedures

influenced the membrane structure. In particular, the influenceof highly permeable regions on permeance was found todecrease in the following order: ethanol−hexane > freeze-drying > room temperature−oven samples. The comparativelyreduced average free-volume pore size in oven pretreatedsamples was attributed to membrane shrinkage during high-temperature drying, while in ethanol−hexane treated polyamidechains, swelling and plugging compounds might have beenremoved, thereby decreasing the membrane resistance to gaspermeation.(3) In PALS, dependence of the Ps lifetime on positron

incident energy clearly showed that the holes in the outer layerof the present membranes were smaller in the order ofethanol−hexane > freeze-drying > room temperature−ovensamples and that the estimated hole sizes were below 0.6 nm indiameter, which is in good agreement with the expected localspace size in the dense regions.(4) Water permeance in RO for the CPA5 membrane was

increased by factors of 1.94 and 1.29 for samples treated inethanol−hexane compared with those treated at roomtemperature−oven and freeze-drying, respectively. This dem-onstrated the correlation between the permeation ability of themembranes in RO and gas permeation systems. On the otherhand, salt rejection remained almost invariable after thedifferent drying procedures despite the differences inmembrane structure, probably due to membrane shrinkage,which produce an effective separation of NaCl by molecularsieving in the pretreated samples, but also due to the chemicallycharged structures that were effective for a NaCl exclusionmechanism, that were not altered after different pretreatments.In summary, this work proposed for the first time the

application of the NPP technique for the structure character-ization of polyamide membranes. The obtained results wereconsistent with NKP and PALS data, especially in the size rangebelow 0.6 nm, and showed a bimodal free-volume pore sizedistribution in the membrane. The results aid in the

understanding of compound transport mechanisms in theinhomogeneous structure of aromatic polyamide membranesfor the development of a new generation of membranes withimproved separation performance.

■ ASSOCIATED CONTENT*S Supporting InformationText describing membrane pretreatment methods, nano-permporometry measurements, and physicochemical propertiesof the solvents and figures showing the time course for Hepermeance, water, hexane, and IPA sorption uptake in the ROpolyamide membrane, and the side view of a water drop on theCPA5 PA membrane after RTO drying. This material isavailable free of charge via the Internet at http://pubs.acs.org.

■ AUTHOR INFORMATIONCorresponding Author*Tel.: +81 824 24 7714. Fax: +81 824 22 7191. E-mail: [email protected] authors declare no competing financial interest.

■ ACKNOWLEDGMENTSWe gratefully acknowledge the financial support from the JapanSociety for the Promotion of Science, under the PostdoctoralFellowship for Foreign Researchers FY2012.

■ NOMENCLATUREA membrane areaC concentrationdk Kelvin diameterdp free-volume pore sizeE positron incident energyL water permeancem massM molecular weightp pressureP permeanceQ flow raterk Kelvin radiusrPs hole radiusR salt rejectionS sorption uptaket timev volumeGreek Lettersθ contact angleΠ osmotic pressureσ surface tensionτPs positronium lifetimeυ molar volume

■ REFERENCES(1) de Abajo, J.; de la Campa, J. G.; Lozano, A. E.; Alvarez, J. C.Thermally stable polymers: Novel aromatic polyamides. Adv. Mater.1995, 7, 148−151.(2) Baker, R. W. Membrane Technology and Applications, 2nd ed.;John Wiley & Sons: Chichester, U.K., 2004.(3) Greenlee, L. F.; Lawler, D. F.; Freeman, B. D.; Marrot, B.;Moulin, P. Reverse osmosis desalination: Water sources, technology,and today’s challenges. Water Res. 2009, 43, 2317−2348.(4) Pinnau, I.; Toy, L. G. Transport of organic vapor through poly[1-(trimethylsilyl)-1-propyne]. J. Membr. Sci. 1996, 116, 199−209.



http://pubs.acs.org

mailto:[email protected]

mailto:[email protected]

(5) Yong, W. F.; Li, F. Y.; Xiao, Y. C.; Li, P.; Pramoda, K. P.; Tong, Y.W.; Chung, T. S. Molecular engineering of PIM-1/Matrimid blendmembranes for gas separation. J. Membr. Sci. 2012, 407−408, 47−57.(6) Hill, A. J.; Freeman, B. D.; Jaffe, M.; Merkel, T. C.; Pinnau, I.Tailoring nanospace. J. Mol. Struct. 2005, 739, 173−178.(7) Koros, W. J.; Chan, A. H.; Paul, D. R. Sorption and transport ofvarious gases in polycarbonate. J. Membr. Sci. 1977, 2, 165−190.(8) Ulbricht, M. Advanced functional polymer membranes. Polymer2006, 47, 2217−2262.(9) Jean, Y. C.; Hung, W. S.; Lo, C. H.; Chen, H.; Liu, G.; Chakka,L.; Cheng, M. L.; Nanda, D.; Tung, K. L.; Huang, S. H.; Lee, K. R.; Lai,J. Y.; Sun, Y. M.; Hu, C. C.; Yu, C. C. Applications of positronannihilation spectroscopy to polymeric membranes. Desalination 2008,234, 89−98.(10) Ito, K.; Oka, T.; Kobayashi, Y.; Suzuki, R.; Ohdaira, T. Variable-energy positron annihilation study of subnanopores in SiOCH-basedPECVD films. Radiat. Phys. Chem. 2007, 76, 213−216.(11) Ito, K.; Chen, Z.; Zhou, W.; Oshima, N.; Yanagishita, H.;Suzuki, R.; Kobayashi, Y. Subnanoscopic holes in compositemembranes for desalination elucidated by energy-tunable positronannihilation. Kobunshi Robnunshu 2012, 69, 443−447.(12) Ito, K.; Ujihira, Y.; Yamashita, T.; Horie, K. Free-volume changeduring the volume phase transition of polyacrylamide gel studied bypositron annihilation: Solvent-composition dependence. J. Polym. Sci.,Part B: Polym. Phys. 2000, 38, 922−933.(13) Shintani, T.; Shimazu, A.; Yahagi, S.; Matsuyama, H.Characterization of methyl-substituted polyamides used for reverseosmosis membranes by positron annihilation lifetime spectroscopy andMD simulation. J. Appl. Polym. Sci. 2009, 113, 1757−1762.(14) Shimazu, A.; Goto, H.; Shintani, T.; Hirose, M.; Suzuki, R.;Kobayashi, Y. Vacancy profile in reverse osmosis membranes studiedby positron annihilation lifetime measurements and moleculardynamics simulations. J. Phys. Conf. Ser. 2013, 443, 012050.(15) Chen, Z.; Ito, K.; Yanagishita, H.; Oshima, N.; Suzuki, R.;Kobayashi, Y. Correlation study between free-volume holes andmolecular separations of composite membranes for reverse osmosisprocesses by means of variable-energy positron annihilationtechniques. J. Phys. Chem. C 2011, 115:37, 18055−18060.(16) Albo, J.; Wang, J.; Tsuru, T. Gas transport properties ofinterfacially polymerized polyamide composite membranes underdifferent pre-treatments and temperatures. J. Membr. Sci. 2014, 449,109−118.(17) Albo, J.; Wang, J.; Tsuru, T. Application of interfaciallypolymerized polyamide composite membranes to isopropanoldehydration: Effect of membrane pre-treatment and temperature. J.Membr. Sci. 2014, 453, 384−393.(18) Tsuru, T.; Takata, Y.; Kondo, H.; Hirano, F.; Yoshioka, T.;Asaeda, M. Characterization of sol/gel derived membranes and zeolitemembranes by nanopermporometry. Sep. Purif. Technol. 2003, 32, 23−27.(19) Tsuru, T.; Hino, T.; Yoshioka, T.; Asaeda, M. Permporometrycharacterization of microporous ceramic membranes. J. Membr. Sci.2002, 186, 257−265.(20) Mulder, M. Basic Principle of Membrane Technology, 2nd ed.;Kluwer Academic: Dordrecht, The Netherlands, 1996.(21) Kim, K. J.; Fane, A. G.; Ben Aim, R.; Liu, M. G.; Jonsson, G.;Tessaro, I. C.; Broek, A. P.; Bargeman, D. A comparative study oftechniques used for porous membrane characterization: Porecharacterization. J. Membr. Sci. 1994, 87, 35−46.(22) Katz, M. G.; Baruch, G. New insights into the structure ofmicroporous membranes obtained using a new pore size evaluationmethod. J. Membr. Sci. 1986, 58, 199−211.(23) Cuperus, F. P.; Bargeman, D.; Smolders, C. A. Permporometry.Characterization of anisotropic UF-membranes: Top layer thicknessand pore structure. J. Membr. Sci. 1991, 61, 73−83.(24) Cuperus, F. P.; Bargeman, D.; Smolders, C. A. Permporometry.The determination of the size distribution of active pores in UFmembranes. J. Membr. Sci. 1992, 71, 57−67.

(25) Lee, H. R.; Kanezashi, M.; Shimomura, Y.; Yoshioka, T.; Tsuru,T. Evaluation and fabrication of pore-size-tuned silica membranes withtetraethoxydimethyl disiloxane for gas separation. AIChE J. 2011, 57(10), 2755−2765.(26) Agrawal, V. K.; Singh, P. S. Alcohol-induced change of ‘‘reverseosmosis’’ polyamide membrane surface. J. Appl. Polym. Sci. 2012, 124,290−299.(27) Solomon, M. F. J.; Bhole, Y.; Livingston, A. G. High fluxmembranes for organic solvent nanofiltration (OSN)Interfacialpolymerization with solvent activation. J. Membr. Sci. 2012, 423−424,371−382.(28) Kuo, C.-Y.; Su, S.-L.; Tsai, H.-A.; Su, Y.-S.; Wang, D.-M.; Lai, J.-Y. Formation and evolution of a bicontinuous structure of PMMAmembrane during wet immersion process. J. Membr. Sci. 2008, 315,187−194.(29) Xiao, J.; Wei, J. Diffusion mechanism of hydrocarbons inzeolites-I. Theory. Chem. Eng. Sci. 1992, 47, 1123−1141.(30) Shelekhin, A. B.; Dixon, A. G.; Ma, Y. H. Theory of gas diffusionand permeation in inorganic molecular-sieve membranes. AIChE J.1995, 41, 58−67.(31) Tung, K. L.; Jean, Y. C.; Nanda, D.; Lee, K. R.; Hung, W. S.; Lo,C. H.; Lai, J. Y. Characterization of multilayer nanofiltrationmembranes using positron annihilation spectroscopy. J. Membr. Sci.2009, 343, 147−156.(32) Xie, W.; Ju, H.; Geise, G. M.; Freeman, B. D.; Mardel, J. I.; Hill,A. J.; McGrath, J. E. Effect of free volume on water and salt transportproperties in directly copolymerized disulfonated poly(arylene ethersulfone) random copolymers. Macromolecules 2011, 44, 4428−4438.(33) Tsuru, T.; Urairi, M.; Nakao, S.; Kimura, S. Reverse osmosis ofsingle and mixed electrolytes with charged membranes: Experimentand analysis. J. Chem. Eng. Jpn. 1994, 24, 518−524.(34) Tsuru, T.; Nakao, S.; Kimura, S. Ion separation by bipolarmembranes in reverse osmosis. J. Membr. Sci. 1995, 108:3, 269−278.(35) Lee, K. P.; Arnot, T. C.; Mattia, D. A review of reverse osmosismembrane materials for desalinationDevelopment to date andfuture potential. J. Membr. Sci. 2011, 370, 1−22.



Structural Characterization of Thin-Film Polyamide Reverse Osmosis Membranes

Documents