Article pubs.acs.org/Langmuir

All-Nanoparticle Layer-by-Layer Surface Modification of Micro- and Ultrafiltration Membranes Luis Escobar-Ferrand,† Diya Li,‡ Daeyeon Lee,‡ and Christopher J. Durning*,† †

Department of Chemical Engineering, Columbia University, New York, New York 10027, United States Department of Chemical and Biomolecular Engineering, University of Pennsylvania, Philadelphia, Pennsylvania 19104, United States

‡

ABSTRACT: Layer-by-layer (LbL) deposition using primarily inorganic silica nanoparticles is employed for surface modification of polymeric micro- and ultrafiltration (MF/ UF) membranes to produce novel thin film composite (TFC) membranes intended for nanofiltration (NF) and reverse osmosis (RO) applications. A wide variety of porous substrate membranes with different surface characteristics are successfully employed. This report gives detailed results for polycarbonate track etched (PCTE), polyethersulfone (PES), and sulfonated PES (SPEES) MF/UF substrates. Both spherical (cationic/anionic) and eccentric elongated (anionic) silica nanoparticles are deposited using conditions similar to those in prior works for solid substrates (e.g., Lee et al.1). Appropriate selection of the pH for anionic and cationic particle deposition enables construction of nanoparticle-only layers 100−1200 nm in thickness atop the original porous membrane substrates. The surface layer thickness appears to vary linearly with the number of bilayers deposited, i.e., with the number of anionic/cationic deposition cycles. The deposition process is optimized to eliminate drying-induced cracking and improve mechanical durability via thickness control and postdeposition hydrothermal treatment. “Dead-end” permeation tests using dextran standards reveal the hydraulic characteristics and separations capability for the PCTE-based TFC membranes. The results show that nanoparticle-based LbL surface modification of MF and UF rated media can produce TFC membranes with NF capabilities.

1. INTRODUCTION In membrane technology, the term “thin film composite” (TFC) membrane refers to a multilayer film consisting of a porous, nonselective support layer with a very thin selective barrier layer on top. Such structures have become standard for demanding water “filtration” applications including nanofiltration (NF) and reverse osmosis (RO). The main advantage of TFC architectures is the high flux, due to the thinness of the selective top layer, combined with mechanical integrity, due to the porous support layer.2,3 Important examples are the interfacially polymerized polyamide composite membranes for desalination by RO, known for their high rejection and high flux due to the very thin “skin” layer.2 The skin layer and support core are complementary and can be optimized independently.4 For example, Hoek et al.5 added zeolite nanoparticles during the polyamide membrane interfacial polymerization process to tailor the top skin layer for better RO membrane separation performance in a TFC system. As a result, cost-effective TFC membranes have been developed to provide good selectivity and flux with reasonable mechanical, thermal, and chemical stability and even self-cleaning properties.2−4,6,7 Despite the innovation of TFC architectures, the set © 2014 American Chemical Society

of TFC membranes available for the important applications of water nanofiltration and desalination remains limited. For example, at present, mainly the interfacially polymerized polyamides are used for desalination via RO, carrying their known limitations regarding membrane permeability and limited tolerance against chlorine, fouling, and solvents.2,3 In this work we demonstrate modification of surfaces of polymeric microfiltration (MF) and ultrafiltration (UF) membranes with contiguous inorganic nanoparticle thin layers to obtain defectfree TFC membranes. The methodology achieving this employs layer-by-layer deposition (LbL) which could enable fabrication of “engineerable”, highly selective top layers on a wide variety of substrates.8−10 LbL deposition, reported first by Iler more than 40 years ago,11 is a directed self-assembly process, enabling preparation of supported thin films with controlled nanostructure.1,12−15 The LbL process is simple and robust and employs mild, environmentally friendly reagents and conditions. Compared Received: June 12, 2013 Revised: February 21, 2014 Published: February 25, 2014 5545

dx.doi.org/10.1021/la5002489 | Langmuir 2014, 30, 5545−5556

Langmuir

Article

Figure 1. (a) Frontal incidence SEM image of spherical Ludox TM-40 nanoparticles on coated glass substrate. Nanoparticles appear fairly uniform with a particle diameter of 25−30 nm. (b and c) Frontal incidence SEM image of two different anionic silica nanoparticles (Snowtex OUP and Snowtex UP respectively) on a membrane substrate. They are elongated and appear to be short chains of roughly spherical particles 10−15 nm in diameter and 2−10 particles strung together in length.

with some other methods for thin film fabrication, (e.g., vaporphase deposition, surface-initiated polymerization) LbL offers easier preparation and frequently more durable, largely defectfree deposited layers.2,4,7−9,16 In addition, a variety of deposition protocols have been used to generate uniform composite films (e.g., dip/rinse cycles, spin processing, spray processing).17,18 Although minor differences do result from the various protocols, the flexibility with respect to deposition method makes automated, industrial-scale processing feasible. The most common LbL process involves dipping an initially charged substrate (e.g., with positive net surface charge) into a dilute aqueous solution of the complementary (e.g., anionic) polyelectrolyte and allowing it to adsorb and “overcharge” the substrate surface. The negatively charged, coated substrate is rinsed to remove free (unbound) polyanion and then dipped into a solution of a cationic polyelectrolyte, which adsorbs and recreates a positively charged surface. This cycle corresponds to deposition of one “bilayer” of solutes bearing complementary charge. Sequential, alternating adsorptions of anionic and cationic polyelectrolytes allow construction of multilayer polymer thin films of many bilayers.12,14 More recent work has demonstrated that the process works with a variety of supramolecular charged solutes including nanoparticles and nanotubes.1,19,20 These efforts inspire ours, whose aim is to create TFC filtration membranes with NF or even RO capability via LbL surface modification of existing microporous MF and UF membranes, by depositing contiguous thin coatings having nanometer-scale porosity and controlled internal chemistry.

2. EXPERIMENTAL SECTION 2.1. Materials. All materials were used as received. Poly(allylamine hydrochloride) (PAH, Mw = 56 000 Da) and poly(acrylic acid) (PAA, Mw = 100 000 Da, 35 wt % in aqueous solution) were purchased from Sigma Aldrich (St. Louis, MO). Ludox CL cationic spherical silica nanoparticles (30 wt % in aqueous solution; 15 nm average diameter) and Ludox TM-40 anionic spherical silica nanoparticles (40 wt % in aqueous solution; 25 nm average diameter) were purchased from Sigma Aldrich (St. Louis, MO); Snowtex UP (20−21 wt % in basic aqueous solution; 9−15 nm width, 40−100 nm length) elongated anionic silica nanoparticles and Snowtex OUP (15−16 wt % in acidic aqueous solution; 9−15 nm width, 40−100 nm length) elongated anionic silica nanoparticles were obtained through Nissan Chemical America (Houston, TX). Hydrochloric acid (HCl) (5 M) and sodium hydroxide (NaOH) (5 M) solutions to adjust pH were prepared from HCl purchased from Amend Drug and Chemicals Ltd. (Irvington, NJ) and NaOH (ACS grade, ≥ 97%) obtained from Sigma Aldrich (St. Louis, MO). DI water was obtained from Millipore Q-Gard 1 and Progard 2 systems (Billerica, MA). Diced silicon wafers (10 cm diameter) were obtained from Ted Pella Inc. (Redding, CA). A stock polyelectrolyte solution of PAH was prepared with 0.94 g/L of PAH in DI water adjusted to pH 7.5 using 5 M NaOH. A stock solution of PAA was prepared by dissolving 2.06 g/L of PAA in DI water and then adjusting to a pH of 3.5 using 5 M HCl. Stock nanoparticle solutions were prepared at nominally 1.00 g/L for Ludox CL, 0.75 g/L for Ludox TM-40, 1.50 g/L for Snowtex UP, and 1.94 g/ L for Snowtex OUP. pHs were adjusted to 3.0, 3.0, 10.0, and 4.0−6.0 respectively, by 5 M NaOH or 5 M HCl. Concentrations and pH values for stock solutions were chosen to match values reported in the literature, ensuring optimal conditions for building LbL multilayers1,21,22 except for elongated nanoparticles (Snowtex) where we tried different pHs and found the optimal values, as reported in this 5546

dx.doi.org/10.1021/la5002489 | Langmuir 2014, 30, 5545−5556

Langmuir

Article

work. The nanoparticles employed in this study were imaged by scanning electron microscopy (SEM) to independently verify their size and shape after deposition onto cleaned glass substrates treated with a polyelectrolyte precursor layer as described below or onto polycarbonate membrane substrates treated with a precursor (see Figure 1). A variety of commercial porous membranes were used as substrates for this work. Nuclepore track-etch polycarbonate (PCTE) membranes from Whatman (Kent, UK) were used with nominal pore sizes of 0.03, 0.05, 0.08, 0.1, and 0.2 μm. Also, Omega poly(ethersulfone) (PES) membranes with dextran standards 90% molecular weight cutoffs (MWCO) of 100, 300, 500, and 1000 kDa, Supor PES (Supor 100, Supor 100H Thin, Supor 200, Supor 200 WE4, and AA Supor 200) with 100 and 200 nm nominal pore size, nylon AN-15 and AN25 membranes with 100 and 150 nm nominal pore size, respectively, and sulfonated poly(etherethersulfone) (SPEES) membranes with dextran standards MWCO of 10 kDa and 100 kDa were used, all obtained from Pall Corp. (Port Washington, NY). Table 1 summarizes

repeatedly with tap water followed by several rinses with DI water. They were then rinsed in an alternating fashion in double baths of DI water and acetone until bubbles stopped appearing on the glass surface. The clean pieces were finally stored under a 1% v/v HNO3 solution. Before use, the glassware/fixture was taken out from the acid solution and given a careful rinse with DI water followed by air drying in a laminar flow hood. A Microm MS-50 slide stainer (Zeiss, Thornwood, NY) interfaced to a desktop computer was adapted for this work and used as a programmable dipper for performing the LbL coating process on the porous membranes. Cationic and anionic polyelectrolytes and both spherical (cationic/anionic) and elongated (anionic) silica nanoparticles were deposited by this device using solutions and dipping conditions similar to those reported by Lee et al.1 For samples used in morphological characterizations, the porous flat-sheet MF and UF membranes used as substrates were surface modified on the “top” side in the dipper after being adhered to glass slides as follows. Ten bilayers of polyelectrolytes were first deposited onto the glass with the last coat on the glass being the polyanion PAA. To adhere the membrane substrate to the polyelectrolyte-coated glass, it was dipped with 2.5 bilayers of polyelectrolytes, ending on the polycation PAH, and then placed with the appropriate side down on the coated glass under DI water. When dry, the glass slide and membrane substrate adhered strongly. For samples used to determine membrane permeability or selectivity characteristics, the substrates were surface modified after being clamped into special frames exposing only one side to the coating sequence but allowing subsequent release for use of the sample as a membrane in a permeation cell. It is relevant to note that we obtained the most reproducible results depositing nanoparticle layers after depositing a “precursor” polyelectrolyte layer atop the porous substrate, in particular, after the 2.5 polyelectrolyte bilayers ending on polycation, as indicated above, were deposited on the top side of the membrane surface. In particular, the dipping process starts with cationic PAH solution for 10 min followed by two DI rinses for 2 and 1 min. Then follows dipping in the anionic PAA solution for 10 min, with two DI water rinses of 2 and 1 min each. This would complete a first bilayer of polyelectrolyte coating. The precursor coating stops midway during the third layer after a PAH dip and two DI water rinses. Presumably, as a result of this sequence, the sample surface becomes positively charged and ready for a first dip into anionic nanoparticle solution. In what follows we report only the results employing the precursor coating prior to nanoparticle deposition. Figure 2c and 2d shows representative fracture surface SEM micrographs of the polyelectrolyte precursor layer typical for the substrates used. LbL deposition of nanoparticles was carried out immediately following the precursor coating. Dipping commenced in anionic particles (Ludox Cl or Snowtex-UP) for 10 min and proceeded to three rinses with DI water for 2, 1, and 1 min. The sample then proceeded to the next bath containing cationic nanoparticles (LudoxTM 40) for 10 min, again followed by three DI water rinses of 2, 1, and 1 min. This would complete 1 bilayer cycle of nanoparticle deposition. The whole sequence was iterated up to 300 times to achieve the desired number of bilayers. The process could be programmed and controlled by computer. Following nanoparticle deposition the TFC samples were air dried in a laminar flow hood. 2.3. Post-treatment. A post-treatment was applied to samples where indicated. This was done to mechanically stabilize deposited layers as described by Gemici et al.,16 although no systematic investigation of the effects of this post-treatment was carried out by us. After completing the nanoparticle deposition, the TFC membrane samples mounted on glass slides were first air dried for about 30 min in a laminar flow hood and then put into an “autoclave” oven and subject to a wet (near 100% humidity) heating cycle at 121 °C for approximately 1 h.16 After nanoparticle deposition, air drying, and possible autoclaving, samples were stored covered in a laminar flow hood at ambient conditions until analysis or testing. 2.4. Morphological Characterization Methods. Morphological characterization of the substrates and fabricated TFC membranes was conducted using several techniques. Cross-section and frontal

Table 1. Summary of Substrate Characteristics

substrate

a

PCTE (Nuclepore) PES (Omega) PES (Supor-100) PES (Supor100H Thin) PES (AASupor200) PES (Supor200WE4) PES (Supor-200) Nylon (AN-15) Nylon (AN-25) SPEES

rating MF

b

nominal pore size range (nm)

MWCO range (kDa)

30−200

UF

100− 1000

surfacec wettability

surfaced charge

hydrophilic

neutral

hydrophobic

slightly negative neutral

MF

100

MF

100

very hydrophilic hydrophobic

MF

200

hydrophilic

highly positive

MF

200

hydrophilic

neutral

MF

200

neutral

MF

100

very hydrophilic hydrophilic

MF

150

hydrophilic

positive

hydrophilic

negative

UF

10−100

neutral

positive

a

Commercial product name shown in parentheses; information supplied by the manufacturers (Pall Corp. and Whatman). bMF ≡ microfiltration; UF ≡ ultrafiltration. cOn the basis of water contact angle measurements and testing provided by the manufacturers. dOn the basis of testing provided by the manufacturers. the substrate’s properties obtained from the manufacturers. Figure 2 shows representative scanning electron micrographs (SEMs) of two of the substrates employed, indicating the variety of surface architectures employed, from smooth and geometrically regular to very rough and irregular. LbL surface modifications were tried on all of the substrates shown in Table 1; however, more detailed systematic characterizations of the LbL deposition results were carried out only on TFC’s made from PCTE, PES, and SPEES substrates. For “dead-end” permeation tests aimed at determining the fabricated TFC membrane’s hydraulic and selectivity characteristics, we employed dextran standards obtained from Fluka (St. Louis, MO) and Sigma Aldrich (St. Louis, MO), which serve as “calibrated” solutes with dimensions in the range 1−50 nm.23,24 Permeation tests were carried out only on MF-rated (0.2 μm) PCTE-based membranes. 2.2. LbL Deposition Technique. For the LbL coating process, all glassware/fixtures used in any capacity passed through the same cleaning procedure. As-received glassware/fixtures were first washed in surfactant solution (2% w/w solution of Alconox) and then rinsed 5547

dx.doi.org/10.1021/la5002489 | Langmuir 2014, 30, 5545−5556

Langmuir

Article

Figure 2. Representative SEMs of uncoated and polyelectrolyte-only-coated supports. (a) Frontal incidence SEM of uncoated SPEES 100 kDa membrane. (b) Frontal incidence SEM of uncoated PCTE 0.2 μm membrane. (c and d) SEM of fracture surface cross sections of SPEES 100 kDa and PCTE 0.2 μm membranes, respectively, surface modified by LbL with only polyelectrolytes (PAA and PAH). The precursor layer, whose thickness ranges between 15 and 20 nm, is visible as a thin “veil” over the membrane surface that does not completely block the pore structure.

Table 2. Summary of Samples Tested for Dextran Rejection substrate PCTE PCTE PCTE PCTE

0.2 0.2 0.2 0.2

μm μm μm μm

polyelectrolytea precoat (PAH/PAA)

cationic nanoparticles

anionicb nanoparticles

numberc of bilayers

post-treatmentd

2.5 bilayers 2.5 bilayers 2.5 bilayers

spherical 15 nm Ludox CL spherical 15 nm Ludox CL

spherical 25 nm Ludox TM elongated Snowtex-UP

40 40

Y Y Y

Where indicated, 2.5 bilayers of polyelectrolytes ending on cationic PAH were deposited on top of the substrate as a pretreatment to fix the surface charge. bWhere indicated, Snowtex-UP anionic elongated silica nanoparticles whose dimensions are 9−15 nm width and 40−100 nm length were used. cNumber of bilayers of anionic/cationic nanoparticle pairs deposited. dWhere indicated, a hydrothermal post-treatment was applied for 1 h at 121 °C at near 100% humidity. 16 a

incidence scanning electron microscopy (SEM) for relatively low magnifications (less than 40 000×) was conducted with a JEOL JSM5600 LV Microscope (Tokyo, Japan) and for higher magnifications (up to 500 000×) with a Zeiss LEO 1550 high-resolution field emission SEM microscope (Cambridge, UK) equipped with a Schottky Field Emitter (FESEM). Energy-dispersive X-ray spectroscopy (EDX) on fracture cross sections was performed with a prism/ digital spectrometer from Princeton Gamma Tech (PGT) (Princeton, NJ). Atomic force microscopy (AFM) was performed using a TopoMetrix Explorer Microscope (now Veeco Instruments) (Santa Clara, CA) with software version 5.01. Image analysis was done using the Nikon NIS-Elements Advanced Research software (Melville, NY). Samples for SEM and EDX were prepared on aluminum stubs. For frontal incidence SEM (top view) samples were adhered to stubs with double-sided tape or a suitable conductive adhesive (except for polycarbonate membranes). More than one sample could be placed on a stub. For polycarbonate membranes, an Au/Pd coating was placed along the edges of the sample to adhere the membrane to the stub. For cross-section samples, 45° cut aluminum stubs were used. A

rectangular piece of sample about 2 cm long and 0.5−1 cm wide was cut, and using forceps, it was immerse into a wetting fluid for the membrane allowing enough time for the fluid to wick completely into the sample. Isopropanol or Filmex was used as the wetting liquid. Using forceps, the saturated sample was submerged in liquid nitrogen and quickly snapped to break the sample cleanly, producing a freeze fracture surface. The sample was then air dried and adhered to an angled aluminum stub with the fracture edge protruding slightly over stub’s top edge. The SEM instruments generated images which carry a scale bar and sample identification. The main purpose of energy-dispersive X-ray spectroscopy (EDX) is to chemically depth profile samples to determine where nanoparticles reside, i.e., if significant invasion of nanoparticles into the substrate pore structure occurred during the LbL deposition process. Cross-sectional freeze fracture samples, prepared as described previously, are used for the EDX depth profiling. The Princeton Gamma Tech (PGT) instrument, which is attached to the JEOL SEM, first performs a low-resolution area scan of the sample, at approximately 200×, to obtain a record of major, minor, and trace 5548

dx.doi.org/10.1021/la5002489 | Langmuir 2014, 30, 5545−5556

Langmuir

Article

elemental identities. The area surveyed by this first scan is approximately 3.0−3.5 mm2. Following this, we performed additional scans from 500× to 1200× at 512 × 512 pixels for depth profiling, surveying areas as small as 0.8−1.0 mm2. The instrument resolution at the highest magnification is 0.4 μm. The main purpose of AFM scans on TFC membrane surfaces was to inspect the LbL film surface morphology and specifically map the cracking morphology. The instrument was operated in noncontact mode, normally used for soft samples, with a special silicon probe, model LTESP-MT from Bruker AFM Probes Instruments (Camarillo, CA), and adjusting to set an optimal instrument frequency for this probe. The latter was determined by the cantilever maximum amplitude (the resonance value obtained on the spectrum plot of amplitude vs frequency) measured after beam alignment. After setting the optimal frequency, AFM scans for sample topology were performed with a resolution on the order of nanometers. 2.5. “Dead-End” Permeation Tests. The main goal of this work is the preparation of new NF and/or RO capable TFC membranes by the LbL surface modification of MF/UF media. To demonstrate this we carried out permeability and selectivity tests on the TFC membranes based on the MF-rated (0.2 μm) PCTE substrates. Presumably, the results are representative of the performance of TFC membranes with the same surface modification but with other substrates. The samples tested in the permeability/selectivity experiments are summarized in Table 2. These include as controls the uncoated substrates and substrates coated only with a cationic and anionic polyelectrolyte precursor layer as well as TFC membranes with both spherical (cationic/anionic) and elongated (anionic) silica nanoparticles (the majority of the surface layer) deposited after the precursor. The sample mounting frames mentioned previously were designed to allow surface modification on only one side of the membrane and permitted subsequent removal of samples from the frame without damage. The deposited nanoparticle layer thickness varied from about 200 nm for 40 bilayers deposited for the spherical/ spherical particle case to about 600 nm for 40 bilayers deposited for the spherical/elongated particle case. All TFC samples tested for permeability and selectivity were post-treated at 121 °C near 100% humidity for approximately 1 h after nanoparticle deposition to improve the deposited layer’s mechanical durability.16 The selectivity tests designed to evaluate the TFC membrane’s dextran rejection characteristics employ a polydisperse stock solution of dextrans as feed in a stirred cell dead-end permeation experiment. The stock solution contains dextrans with molecular weights ranging from 1.5 to 2000 kDa (see Table 3). The experiments employ the 44.5 mm diameter Amicon 8050 stirred cells supplied by Millipore (Billerica, MA) with 50 mL volume capacity. The experiment operates in a transient dead-end mode, with the feed chamber magnetically stirred to minimize concentration polarization. The uncoated substrates and TFC membranes were first tested in the stir cell with

deionized water at 55 psi and 300 rpm at ambient temperature. The time to collect about 40 mL of permeate was determined and water flux (i.e., the water permeance, J) calculated there from using •

Q J= A •

where Q is the apparent steady volumetric flow rate of water and A is the membrane’s cross-sectional area. From J one determines the pure water permeability K from

K=

conc. (g/L)

1.5 6 10 20 40 70 100 200 500 2000

0.55 0.65 0.65 0.65 0.65 0.6 0.55 0.55 1.10 3

Rga (Å)

Rhb (Å)

C*c (g/L)

44.5 58 69 95 147 270

(2)

⎛ permeate response ⎞ R d = ⎜1 − ⎟ × 100 feed response ⎠ ⎝

23.6 62 80 95 130 200 380

Jl ΔP

where l is the membrane’s total thickness and ΔP is the applied pressure drop. Following this, experiments with mixed dextrans were carried out in the same cell at 5 psi and 220 rpm in the feed side of the cell. The time to collect about 3 mL of permeate was recorded, and the solution flux was calculated there from using eq 1. Permeabilities based on the solution data were then determined using eq 2. Comparison of apparent permeabilities from the flux data on water and the dextrans feed gives an indication if polarization and/or fouling are important factors in the tests on dextran feeds. Finally, the feed and permeate solutions for dextrans were analyzed by high-pressure liquid chromatography (HPLC) for the dextran molecular weight distributions, from which the rejection characteristics were determined. To perform the HPLC analysis, we employed a Thermo Scientific (Waltham, MA) HPLC setup equipped with a RI-150 refractive index detector and using a Tosoh TSK-GEL G4000PWXL column (King of Prussia, PA) for higher molecular weights and a Tosoh TSK-GEL G3000PWXL column for dextrans whose molecular weights were less than 60 kDa. Before running the chromatography, fresh HPLC-grade water (50 μL) was injected to the mobile-phase reservoir repeatedly until a flat baseline was obtained. For all runs the mobile-phase flow rate was fixed at 1.0 mL/min. Dextran standards for HPLC calibration were prepared by measuring 5.0 mg of each dextran fraction into separate 10 mL volumetric flasks and dissolving them in HPLC-grade water. The standards were first filtered using a 0.45 μm Pall Acrodisc syringe filter and then injected (50 μL) into the HPLC at a mobilephase flow rate of 1.0 mL/min. The chromatographs enabled calculation of the retention times from elution peaks for each standard, and these data were added to the historic data on dextrans for these columns for the instrument. A cumulative calibration for dextrans, log(Mp) vs retention time, was therefore available for the instrument and used for evaluation of percentage of rejection. Here Mp means the peak molecular weight for the (nearly) monodisperse standard samples. Feed and filtrate samples from dead-end permeation were analyzed by first filtering them using a 0.45 μm Pall Acrodisc syringe filter and then injecting 50 μL of each sample at a mobile phase flow rate of 1.0 mL/min into the HPLC. A water injection at the beginning and end of each sample run checked for any baseline shift. The response chromatograms (chromatogram deflection from baseline vs elution time) were converted to response vs dextran molecular weight using the column calibration. Selectivity profiles were then plotted as percentage of rejection (Rd) vs dextran molecular weight where

Table 3. Composition of Dextran Stock Solution for Feed in Stirred-Cell Permeation Tests dextrans MW (kDa)

(1)

62 40

(3)

for each molecular weight.

31 24 19

3. RESULTS AND DISCUSSION 3.1. TFC Membrane Morphology. Nanoparticle-coated TFC membranes were first prepared using all of the different porous substrates (shown in Table 1) with between 100 and 300 bilayers of the spherical nanoparticles (Ludox) deposited, resulting in surface layer thicknesses as large as 3 μm atop the porous membrane substrates. Imaging made clear that in all

a

Radius of gyration (Rg).23 bHydrodynamic radius (Rh).23 cOverlap threshold solute concentration (C*) from inverse of intrinsic viscosity [η].24 5549

dx.doi.org/10.1021/la5002489 | Langmuir 2014, 30, 5545−5556

Langmuir

Article

Figure 3. (a and b) Representative cross-section SEMs of Supor 200 WE4 with 200 deposited bilayers of Ludox CL/Ludox TM40 at different magnifications. Note the visually sharp interface between the deposited layer and the substrate as well as the surface cracking. (c) AFM image of the same TFC surface along a crack edge. Note the raised surface along the crack contour; (d) 2D AFM of the same composite showing surface cracking.

3.2. Deposited Layer. Given the foregoing qualitative indications of successful deposition on porous substrates, a primary question to be answered was if the LbL process in these systems obeyed the most commonly found linear growth law, i.e., if the thickness of the deposited layer increased linearly with the number of bilayers.1,14,21,22 Therefore, measurements were made of deposited layer thicknesses to determine its variation with the number of bilayers deposited with one bilayer corresponding to the result of one complete deposition cycle of the anionic and cationic solute. Depositions were made onto both silicon wafers with a native oxide coating and a porous membrane substrate (PCTE 0.03 μm) to provide controlled comparisons. Deposited layer thicknesses were calculated from fracture surface cross-section SEMs by NIS software. Experiments were performed for the spherical/spherical and spherical/elongated nanoparticle combinations using the concentrations and pH’s indicated earlier. Note that ellipsometry, the method most commonly employed for determining LbL-deposited layer thicknesses,1,8 could not be used due to the porous nonreflective nature of the substrate surfaces. Unfortunately the method we used has much lower accuracy and greater variability relative to ellipsometry. Figure 5 shows data for these experiments, which indicate, in both cases and for both types of substrates, a monotonic increase of the deposited layer thickness with the number of bilayers. Although exhibiting considerable scatter, the results are most consistent with a linear growth law. It appears that the thickness of deposits increases faster for depositions on silicon than for the case of porous substrates, especially for the

cases the resulting TFC membranes have a very sharp interface between the deposited layer and the substrate (see Figure 3 as representative). This result is surprising, considering that the ratio of pore to particle size lies in the range 2−20 and that the substrates used have a wide range of surface physicochemical characteristics as indicated in Table 1. It is worth noting that a sharp interface was seen regardless of pretreatment with polyelectrolyte. A relevant observation suggests that for all substrates wicking of water into the porous interior did not occur immediately upon contact but was delayed by a time scale on the order or greater than seconds or was completely inhibited for hydrophobic substrates. This may be an important factor in realizing nanoparticle deposition highly localized at the surface. Depth profiling characterization was conducted by EDX elemental analysis on fracture cross sections to check more quantitatively that no significant intrusion of nanoparticles into the substrate pore structure occurred in our systems. Representative results are shown in Figure 4. The technique, based on 1200× SEM images with 512 × 512 pixels, showed the silica confined to the membrane surface with very little getting deposited within the porous interior of substrates to within a resolution of 0.4 μm, which is within a few pore diameters typically. The method indicates a ratio of more than 50 between silica mass density detected in the surfacedeposited layer to that detected in the interior based on compositions measured at locations selected across the TFC membranes thickness. 5550

dx.doi.org/10.1021/la5002489 | Langmuir 2014, 30, 5545−5556

Langmuir

Article

Figure 4. Representative depth profiles by EDX analysis along with cross-sectional SEMs of Supor 100 H Thin with 200 deposited bilayers of Ludox CL/Ludox TM40. (a) Map of silica distribution. Upper left image shows the cross-section SEM at 1200×; upper right shows the corresponding silica distribution (silica represented by yellow) from the local elemental analysis of the image on the upper left with 0.4 μm resolution; lower left shows the sample’s surface average chemical elemental profile. Analysis indicates qualitatively that nanoparticles remain on the membrane surface with minimal intrusion into the porous substructure. (b) Local elemental analysis across a TFC membrane thickness for the same sample shown in a, indicating a silica mass density ratio of at least 50 of the surface to the porous interior, to within a resolution of 0.4 μm. Upper image correspond to a cross-section SEM at 1100×, indicating the positions selected for local elemental analysis across the TFC thickness. The sequence of spectra shown below give the elemental analysis at each position.

3.3. Cracking Phenomena. While the forgoing results clearly indicate LbL deposition can effectively create a surface layer on a porous substrate, it was also clear that the TFCs were typically not intact and that cracks ran through the top coating’s length and breadth (see Figures 3 and 6). The cracking phenomena appeared to be ubiquitous, and efforts were concentrated on determining the origin of this cracking in order to produce crack-free, intact TFC samples required for

spherical/elongated particles case. The slopes of regression lines constrained to the origin of the data in Figure 5 for both cases show acceptable statistics (see the 95% confidence limits) and are much lower than the geometric estimate considering the bilayer thickness to be the sum of the particle diameters with two monolayers of nanoparticles per bilayer (slope = 40 and 35 nm/bilayer for Figures 5a and 5b, respectively). 5551

dx.doi.org/10.1021/la5002489 | Langmuir 2014, 30, 5545−5556

Langmuir

Article

Figure 6. Representative SEM micrograph of drying-induced cracking of LbL-deposited nanoparticle layer on an Omega PES 500 kDa substrate. Note a characteristic length of ∼O(102 μm). Analogous cracking appears in slurries and suspensions, e.g., in dried sand layers,25 exhibiting much larger characteristic length scales ∼O(104 μm).

We examined the morphology of cracks induced mechanically in two different ways. TFC samples were purposely bent through a known radius of curvature, e.g., a “rolling” test schematized in Figure 7, or they were exposed to the stress of folding and flattening the coating sample upon itself. SEM images of samples bent over a fixed radius appear to show separation and widening of the pre-existing cracks and delamination from substrates while the images from folded samples along the fold show a distinctly different crack morphology (small shard-like fragments) compared with the results observed for as-prepared, dried layers in Figures 3 and 6. The evidence taken together, (AFM indicates raised crack edges (Figure 3) and mechanically induced cracks exhibit different failure morphologies of the surface layers than that seen in the as-prepared layers (e.g., Figure 6)), suggests dryinginduced cracking is the main cause of the surface defects shown in Figure 6. If true, by analogy with published results on slurries, the deposited layer thickness is the most important variable to be controlled in preventing such cracking. In fact, analysis of the drying process indicates a thickness threshold hc, below which one avoids the damaging drying stresses generated in a thick over layer.26

Figure 5. Thickness of LbL deposit versus number of bilayers deposited on silicon wafers and on a porous substrate (PCTE 0.03 μm) for (a) spherical/spherical Ludox CL/Ludox TM40 and (b) spherical/elongated Ludox CL/Snowtex UP nanoparticle deposition. Slope values are shown with 95% confidence limits. Lines in a and b are regression fits constrained to the origin.

⎡ GMϕ R3 ⎤1/2 ⎡ 2γ ⎤3/2 r ⎥ ⎢ hc = 0.64⎢ ⎥ ⎢⎣ 2γ ⎥⎦ ⎣ −PmR ⎦

membrane filtration applications. AFM microscopy studies on the cracks determined that in most of the cases the crack contours are elevated relative to the rest of the sample (see Figure 3). It was noticed from frontal incidence SEMs a pattern to the cracking with a characteristic length, reminiscent of findings on drying-induced cracking of particle beds in the colloid literature.25 Figure 6 shows representative results for our systems. The crack’s geometric pattern resembles closely the ones observed in colloidal systems except for a change in scale. The drying-induced cracking mechanism anticipates raised edges along crack contours which are produced due to lateral compressive stresses generated during the drying process that causes the cracking.25−27 To further support the assertion that the cracking exhibiting patterns such as in Figure 6 was due to drying and not to mechanical failure during handling of samples after coating, several experiments were designed.

(4)

Here, G is the shear modulus of the particles, M the coordination number of the “packed bed” of particles, ϕr the particle volume fraction at close packing, R the particle radius, γ the water−air interfacial tension, and Pm the maximum capillary pressure. From eq 4, the maximum theoretical thickness for crack-free films of our systems corresponds to a nanoparticle film of about 100 bilayers. Testing the threshold limit predicted by eq 4 was carefully done by examining the surface in one system of a series of deposited films of varying thickness. Indeed crack-free TFC membranes for several combinations of substrates/nanoparticles are observed for thin enough depositions as illustrated in Figure 8. Sample post-treatment may also play a role in fabrication of defect-free composite membranes. The thermal post-treatment 5552

dx.doi.org/10.1021/la5002489 | Langmuir 2014, 30, 5545−5556

Langmuir

Article

Figure 7. (a) Schematic of “rolling experiment” showing how flexible strips were cut from a TFC membrane and bent over a cylinder of known radius of 2.94 mm. (b) Control fracture surface cross-section SEM of PCTE 0.2 μm coated with 40 bilayers of Ludox CL/Ludox TM40. The same TFC membrane sample after (c) mechanical bending by rolling and (d) mechanical bending by folding.

trolyte precursor layer followed by 40 bilayers of Ludox CL/ Ludox TM (cationic spherical/anionic spherical) or Ludox CL/ Snowtex UP (cationic spherical/anionic elongated) silica nanoparticle combinations (see Table 2). Table 4 and Figure 9 show results. As expected from the manufacturer’s specification of (nominal) pore size (0.2 μm), there is no significant dextran rejection over the molecular weight range studied (1.5−2000 kDa) by the MF-rated substrate alone (see Figure 9). Indeed, based on our SEM surface imaging, the MF-rated PCTE 0.2 μm membrane has a mean pore size ∼200 nm, significantly larger than even the largest dextran in solution (see Table 3), consistent with the fact that all dextrans tested pass through the PCTE membrane unhindered. Also, for these MF-rated samples there is no significant effect on dextran rejection of depositing only the thin polyelectrolyte layer on the substrate’s surface (note that the weak 8−10% shift of the entire plot upward in Figure 9 for the PE-coated samples relative to the

seems to enhance adhesion of the top layer and perhaps promote sintering of the nanoparticles by chemical hydrolysis of SiO2 bonds. This may enhance the film durability and likely somewhat compacts the top layers. 3.4. Hydraulic and Selectivity Characteristics. We studied the hydraulic and selectivity characteristics of TFC membranes from MF-rated (0.2 μm) PCTE substrates (Table 2) to determine the viability of the LbL surface modification described here for imparting at least NF capability. As a control, the water and dextran feed solution fluxes, the corresponding permeabilities, and the dextran rejection of the MF (PCTE 0.2 μm) rated substrates without surface modification were determined. As a second control, these characteristics were then evaluated for substrates modified with only the very thin polyelectrolyte “precursor” layer (2.5 bilayers of PAA and PAH atop the membrane surface ending on the cationic PAH). Finally, the fluxes, permeabilities, and rejection were determined for TFC membranes, modified with a polyelec5553

dx.doi.org/10.1021/la5002489 | Langmuir 2014, 30, 5545−5556

Langmuir

Article

Figure 9. Dextran rejection Rd vs dextran molecular weight at 25 °C for PCTE 0.2 μm uncoated substrate, after substrate modification with 2.5 bilayers of the polyelectrolytes PAA and PAH, and after modification with polyelectrolytes plus 40 bilayers of Ludox CL/ Ludox TM (cationic spherical/anionic spherical) silica nanoparticles or with 40 bilayers of Ludox CL/Snowtex UP (cationic spherical/ anionic elongated) silica nanoparticles. Error bar shows the range of variation in several repeat experiments.

Figure 8. SEMs demonstrating crack-free TFC membrane fabrication: (a and b) SEM of fracture surface cross-section and frontal incidence SEM of SPEES 10 kDa with 40 deposited bilayers of Ludox CL/ Snowtex UP, respectively; (c and d) fracture surface cross-section and frontal incidence SEM of PCTE 0.03 μm with 60 deposited bilayers of Ludox CL/Snowtex OUP, respectively.

commercial nanofilters and tight ultrafilters.28−30 The values of J for pure water for the TFC samples are also in the range of values for commercial NF operations.2,31,32 The solution permeabilities recorded for the TFC samples in Table 4 lie significantly below the values for pure water. Correcting ΔP in eq 2 for the feed solution osmotic pressure Π to find an effective pressure drop (ΔPeff)

uncoated samples is within the experimental repeatability). Although rejection is not affected, the permeability is reduced significantly by the PE precursor (see Table 4), indicating that a fraction of the pores is blocked. Significantly, the PCTE-based TFC membranes modified with the polyelectrolyte precursor layer and 40 bilayers of Ludox CL/Ludox TM (cationic spherical/anionic spherical) or Ludox CL/Snowtex UP (cationic spherical/anionic elongated) silica nanoparticle combinations show typical “S”-shaped selectivity plots exhibiting cutoff characteristics from the HPLC analysis (see Figure 9). The effect of adding nanoparticles is dramatic. Very similar results were found for the TFCs made on this substrate, with 40 bilayers of spherical particles (anionic and cationic) and 40 bilayers including anionic elongated nanoparticles. The data indicate an average rejection from repeated experiments of 60% for the smallest dextran in the stock solution (1500 Da) and 90% or higher for dextrans with molecular weight 20 kDa or larger, that is, the modified membranes exhibit an “R90” of about 20 kDa. These rejection characteristics are clearly governed by the nanoparticle layer and imply a characteristic pore size in the range of

ΔP eff = ΔP − Π with Π estimated by the ideal solution law

(5) (6)

Π = nRT

where n is the total (molar) density of dextran solute in the feed solution (see Table 3), cannot account for the differences between the solution and pure water permeabilities. Consequently, the data indicate that concentration/polarization and/or dextran fouling play a role in the rejection experiment results.

4. CONCLUSIONS LbL deposition of contiguous nanoparticle layers atop MF- and UF-rated porous supports appears to work for a variety of substrates regardless of surface structure and chemistry. Indeed,

Table 4. Fluxes (Permeances) and Permeabilities of Water and Feed Solution for the Samples Tested for Dextran Rejection samplea PCTE 0.2 μm − control PCTE 0.2 μm + polyelectrolytesc PCTE 0.2 μm + polyeletrolytesc + 40 bl SPH/ SPHd PCTE 0.2 μm+ polyeletrolytesc + 40 bl SPH/ ELe

water fluxb [cm/s] × 103

water permeabilityb [cm3s/g] × 1012

solution fluxb [cm/s] × 103

solution permeabilityb [cm3s/g] × 1012

1010 ± 57f 18 ± 7 2.8 ± 0.5

270 ± 15 4.6 ± 1.8 0.75 ± 0.15

130 ± 35 0.27 ± 0.29 0.032 ± 0.09

360 ± 105 0.72 ± 0.89 0.096 ± 0.036

3.3 ± 0.5

0.91 ± 0.22

0.17 ± 0.05

0.52 ± 0.11

All TFC membranes were hydrothermally post-treated for 1 h at 121 °C at near 100% humidity.16 bThe fluxes and permeabilities for water and dextran feed solution, detailed in Table 3, were determined from eqs 1 and 2. cTwo and one-half bilayers of polyelectrolytes ending on cationic PAH were deposited on top of the substrate as a pretreatment to fix the surface charge. dLudox CL cationic spherical silica nanoparticles 15 nm diameter and Ludox TM anionic spherical silica nanoparticles 25 nm diameter were used. eLudox CL cationic spherical silica nanoparticles 15 nm diameter and Snowtex−UP anionic elongated silica nanoparticles whose dimensions are 9−15 nm width and 40−100 nm length were used. fUncertainties

a

•

were calculated based on eqs 1 and 2 and the uncertainties in Q determined by repeated experiments and using a propagation of errors. 5554

dx.doi.org/10.1021/la5002489 | Langmuir 2014, 30, 5545−5556

Langmuir

■

ACKNOWLEDGMENTS The authors thank Mr. Mike Steves and Ms. Pauline Adejo from Pall Corp. Microscopy Lab, where most of the imaging was conducted, for suggestions and help with microscopy techniques. We also acknowledge the Materials Microscopy Lab of SUNY Stony Brook University where the high-resolution FESEM was performed. The authors also acknowledge Dr. Elliot Campbell for his help with equipment and use of his laboratory. Assistance with sample preparation by Mr. Amrut Biswal, Ms. Lydia Ngai, and Mr. Ji Seung Kim is gratefully acknowledged. The authors thank Pall Corp.’s Dr. Tom Gsell and Dr. Amarnauth Singh for ideas and discussions. Financial support for this research was provided by a grant from Pall Corp. D.L. is supported by a NSF CAREER award (DMR1055594).

the silica nanoparticle surface layers showed a very sharp interface between the top layer and the substrate in all cases we examined. One explanation for this is that wicking of the dipping solution into the pore structure is inhibited, or at least delayed, by the surface texture.33 Alternatively, it may be that internal deposition of nanoparticles is intrinsically self-limiting for the range of pore to nanoparticle sizes used here, as indicated in several recent simulation studies performed for nanoparticles and polyelectrolytes.34−36 The deposited layer thickness increases with respect to the number of bilayers for porous substrates very much like that on smooth contiguous solids, although at a lower rate, and appears to follow a linear growth law. Defect-free TFC membranes were consistently achieved for these surface layers of low-aspect ratio, rigid particles, despite their intrinsic tendency to crack upon drying due to compressive stresses that develop, by considering the thickness threshold below which damaging drying-induced stresses are avoided. Crack-free surface layers are possible with thin enough layers. Hydrothermal post-treatment by autoclaving likely stabilizes these layers.16 Permeation and rejection tests demonstrate the successful preparation of filtration-quality TFC membranes suitable for water purification applications in the NF range by the LbL surface modification described here of existing MF- and UFrated membranes with a surface layer comprised primarily of low aspect ratio inorganic nanoparticles. In particular, dextran rejection by surface-modified MF-rated PCTE supports is comparable with at least some NF-rated membranes for NOM removal,28−30 whereas the bare substrates or TFC’s with surface modification by only a few bilayers of polyelectrolytes show no selectivity over the range of dextran molecular weights from 1.5 to 2000 kDa. Meanwhile, measured water permeabilities for the nanoparticle-modified membranes are significantly higher than most commercial NF membranes,2,31,32 indicating that these TFCs have good potential for improved performance relative to existing materials for at least some NF applications. This work shows the feasibility of the preparation of defectfree thin film composite (TFC) membranes through LbL surface modification of polymeric porous MF/UF membranes using nanoparticles. Such membranes hold promise for a variety of NF/RO applications and may provide new options for key water purification applications.37 Supporting this we can also point to a report38 published just after submission of this work, brought to the attention of the authors by a reviewer, of TFC membranes fabricated by essentially the same technique exhibiting permeance J acceptable for NF applications (although no selectivity data were reported). Further improvements regarding fabrication of TFC’s using the LbL deposition technique are certainly possible. The choice of the assembly method, e.g., dip coating vs spray coating, could reduce the time scale for sample preparation. The robustness of the method developed suggests that it could be applicable to many substrate materials and a variety of nanoparticles with different nanoarchitectures.

■

Article

■

REFERENCES

(1) Lee, D.; Rubner, M. F.; Cohen, R. E. All-Nanoparticle Thin-Film Coatings. Nano Lett. 2006, 6, 2305−2312. (2) Petersen, R. J. Composite reverse osmosis and nanofiltration membranes. J. Membr. Sci. 1993, 83, 81−150. (3) Lau, W. J.; Ismail, A. F.; Misdan, N.; Kassim, M. A. A recent progress in thin film composite membrane: A review. Desalination 2012, 287, 190−199. (4) Schmidt, G.; Malwitz, M. Properties of polymer−nanoparticle composites. Curr. Opin. Colloid Interface Sci. 2003, 8, 103−108. (5) Jeong, B.; Hoek, E. M. V.; Yan, Y.; Subramani, A.; Huanga, X.; Hurwitz, G.; Ghosha, A. K.; Jawor, A. Interfacial polymerization of thin film nanocomposites: A new concept for reverse osmosis membranes. J. Membr. Sci. 2007, 294, 1−7. (6) Ramakrishnan, S.; McDonald, C. J.; Prud’homme, R. K.; Carbeck, J. D. Latex composite membranes: structure and properties of the discriminating layer. J. Membr. Sci. 2004, 231, 57−70. (7) Lee, D.; Gemici, Z.; Rubner, M. F.; Cohen, R. E. Multilayers of Oppositely Charged SiO2 Nanoparticles: Effect of Surface Charge on Multilayer Assembly. Langmuir 2007, 23, 8833−8837. (8) Hong, S. U.; Malaisamy, R.; Bruening, M. L. Optimization of flux and selectivity in Cl−/SO42− separations with multilayer polyelectrolyte membranes. J. Membr. Sci. 2006, 283, 366−372. (9) Ouyang, L.; Malaisamy, R.; Bruening, M. L. Multilayer polyelectrolyte films as nanofiltration membranes for separating monovalent and divalent cations. J. Membr. Sci. 2008, 310, 76−84. (10) Ouyang, L.; Dotzauer, D. M.; Hogg, S. R.; Macanás, J.; Lahitte, J-F; Bruening, M. L. Catalytic hollow fiber membranes prepared using layer-by-layer adsorption of polyelectrolytes and metal nanoparticles. Catal. Today 2010, 156, 100−106. (11) Iler, R. K. Multilayers of Colloidal Particles. J. Colloid Interface Sci. 1966, 21, 569−594. (12) Decher, G. Fuzzy Nanoassemblies: Toward Layered Polymeric Multicomposites. Science 1997, 277, 1232−1237. (13) Chen, W.; McCarthy, T. J. Layer-by-Layer Deposition: A Tool for Polymer Surface Modification. Macromolecules 1997, 30, 78−86. (14) Decher, G.; Schlenoff, J. B. Multilayer Thin Films; Wiley-VCH: New York, 2003. (15) Crespilho, F. N.; Zucolotto, V.; Oliveira, O. N., Jr.; Nart, F. C. Electrochemistry of Layer-by-Layer Films: a review. Int. J. Electrochem. Sci. 2006, 1, 194−214. (16) Gemici, Z.; Shimomura, H.; Cohen, R. E.; Rubner, M. F. Hydrothermal Treatment of Nanoparticle Thin Films for Enhanced Mechanical Durability. Langmuir 2008, 24, 2168−2177. (17) Kolasinska, M.; Krastev, R.; Gutberlet, T.; Warszynski, P. Layerby-Layer Deposition of Polyelectrolytes. Dipping versus Spraying. Langmuir 2009, 25, 1224−1232. (18) Izquierdo, A.; Ono, S. S; Voegel, J. C.; Schaaf, P.; Decher, G. Dipping versus Spraying: Exploring the Deposition Conditions for Speeding Up Layer-by-Layer Assembly. Langmuir 2005, 21, 7558− 7567.

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest. 5555

dx.doi.org/10.1021/la5002489 | Langmuir 2014, 30, 5545−5556

Langmuir

Article

(19) Podsiadlo, P.; Michel, M.; Critchley, K.; Srivastava, S.; Qin, M.; Lee, J. W.; Verploegen, E.; Hart, A. J.; Qi, Y.; Kotov, N. A. Diffusional self-organization in exponential layer-by-layer with micro- and nano scale periodicity. Angew. Chem., Int. Ed. 2009, 48, 7073−7077. (20) Podsiadlo, P.; Kaushik, A. K.; Arruda, E. M.; Waas, A. M.; Shim, B. S.; Xu, J.; Nandivada, H.; Pumplin, B. G.; Lahann, J.; Ramamoorthy, A.; Kotov, N. A. Ultrastrong and stiff layered polymer nanocomposites. Science 2007, 318, 80−83. (21) Lee, D.; Omolade, D.; Cohen, R. E.; Rubner, M. F. pHDependent Structure and Properties of TiO2/SiO2 Nanoparticle Multilayer Thin Films. Chem. Mater. 2007, 19, 1427−1433. (22) Bravo, J.; Zhai, J.; Wu, Z.; Cohen, R. E.; Rubner, M. F. Transparent Superhydrophobic Films Based on Silica Nanoparticles. Langmuir 2007, 23, 7293−7298. (23) de Belder, A. N. Dextran; Amersham Biosciences Collection: Uppsala, Sweden, 2003. (24) Granath, K. A. Solution Properties of Branched Dextrans. J. Colloid Sci. 1958, 13, 308−328. (25) Shorlin, K.; de Bruyn, J. Development and geometry of isotropic and directional shrinkage-crack patterns. Phys. Rev. E 2000, 61, 6950− 6957. (26) Singh, K.; Tirumkudulu, M. S. Cracking in Drying Colloidal Films. Phys. Rev. Lett. 2007, 98, 1−4. (27) Xu, P.; Mujumdar, A. S.; Yu, B. Drying-Induced Cracks in Thin Film Fabricated from Colloidal Dispersions. Drying Technol. 2009, 27, 636−652. (28) 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. (29) Thorsen, T.; Fløgstad, H. Nanofiltration in Drinking Water Treatment; Techneau: Nieuwegein, The Netherlands, 2006. (30) Ulbricht, M. Advanced functional polymer membranes. Polymer 2006, 47, 2217−2262. (31) Riley, R. L. Reverse Osmosis in Membrane Separation Systems; Noyes Data Corp.: Park Ridge, NJ, USA, 1991; Vol. 2. (32) Schaefer, A. I. Natural Organic Removal Using Membranes; Technomic Publication: Lancaster, PA, USA, 2001. (33) Quere, D. Wetting and Roughness. Annu. Rev. Mater. Res. 2008, 38, 71−99. (34) Jeon, J.; Panchagnula, V.; Pan, J.; Dobrynin, A. V. Molecular Dynamics Simulations of Multilayer Films of Polyelectrolytes and nanoparticles. Langmuir 2006, 22, 4629−4637. (35) Carrillo, J-M. Y; Dobrynin, A. V. Layer-by-Layer Assembly of Charged Nanoparticles on Porous Substrates: Molecular Dynamics Simulations. ACS Nano 2011, 5, 3010−3019. (36) Carrillo, J-M. Y; Dobrynin, A. V. Layer-by-Layer Assembly of Polyelectrolyte Chains and Nanoparticles on Nanoporous Substrates: Molecular Dynamics Simulations. Langmuir 2011, 28, 1531−1538. (37) Durning, C. J.; Escobar-Ferrand, L. International Patent WO 2012/151197 A1, 2012. (38) Kim, J.; Sotto, A.; Chang, J.; Nam, D.; Boromand, A.; Van der Bruggen, B. Embedding TiO2 Nanoparticles versus Surface Coating by Layer-by-Layer deposition on Nanoporous Polymeric Films. Microporous Mesoporous Mater. 2013, 173, 121−128.

5556

dx.doi.org/10.1021/la5002489 | Langmuir 2014, 30, 5545−5556