FORUM ARTICLE

www.acsami.org

Covalent Binding of Single-Walled Carbon Nanotubes to Polyamide

Membranes for Antimicrobial Surface Properties
Alberto Tiraferri, Chad D. Vecitis, and Menachem Elimelech*,

Department of Chemical and Environmental Engineering, Yale University, P.O. Box 208286, New Haven, Connecticut 06520-8286,
United States

School of Engineering and Applied Sciences, Harvard University, Cambridge, Massachusetts 02138, United States
bS Supporting Information
ABSTRACT: We propose an innovative approach to impart
nanomaterial-specic properties to the surface of thin-lm
composite membranes. Specically, biocidal properties were
obtained by covalently binding single-walled carbon nanotubes
(SWNTs) to the membrane surface. The SWNTs were rst
modied by purication and ozonolysis to increase their side-
wall functionalities, maximize cytotoxic properties, and achieve
dispersion in aqueous solution. A tailored reaction protocol was
developed to exploit the inherent moieties of hand-cast poly-
amide membrane surfaces and create covalent amide bonds
with the functionalized SWNTs. The reaction is entirely aqueous-based and entails activation of the carboxylate groups of both the
membrane and the nanomaterials to maximize reaction with ethylenediamine. The presence of SWNTs was veried after sonication
of the membranes, conrming the strength of the bond between the SWNTs and the membrane surface. Characterization of the
SWNT-functionalized surfaces demonstrated the attainment of membranes with novel properties that continued to exhibit high
performance in water separation processes. The presence of surface-bound antimicrobial SWNTs was conrmed by experiments
using E. coli cells that demonstrated an enhanced bacterial cytotoxicity for the SWNT-coated membranes. The SWNT membranes
were observed to achieve up to 60% inactivation of bacteria attached to the membrane within 1 h of contact time. Our results suggest
the potential of covalently bonded SWNTs to delay the onset of membrane biofouling during operation.
KEYWORDS: polyamide membranes, biofouling, single-walled carbon nanotubes, SWNT, SWCNT, thin-lm composite, surface
modication, water purication, amide bonds, nanocomposite, biocidal membrane

INTRODUCTION benchmark material for the TFC membrane active layer, showing
Membrane-based water separation processes utilize semi- unrivaled productivity and selectivity performance and a high
permeable membranes to retain contaminants. The surface of degree of tunability.4 However, one major drawback of poly-
the membranes is in contact with feedstreams that carry dissolved amide is its degradation in the presence of chemical oxidants,4,5
molecules, particulates, and microorganisms that can induce mem- which are normally used to control microbial growth. Thus,
brane fouling.1 Biofouling, the growth of a biolm on membrane alternatives to these degradative oxidants must be considered. To
address this challenge, methods to functionalize the active layer
surfaces, results in a decrease in membrane performance, which, in
have been proposed as solutions to reduce and/or delay poly-
turn, increases the overall energy requirement of the separation
amide membrane biofouling.
process and decreases membrane life.1,2 Thus, biofouling control is
Eorts toward membrane active layer modication have
one of the most pressing challenges faced by the membrane science focused on rendering the surface more hydrophilic, smooth,
community. This challenge can be overcome by the use of well- and less charged.4,6 Examples include the production of novel
designed functional materials that prevent attachment or growth of polyamide-based materials with tailored chemistry and morpho-
microorganisms at the membrane surface, while maintaining high ogy achieved by the addition of monomers or variation of
separation performance during operation. conditions during interfacial polymerization.7,8 However, the
Thin-lm composite (TFC) membranes represent the state- unsurpassed separation properties of polyamide limit the range
of-the-art in dense separation membranes such as those used in of improvements that can be made following this pathway.
reverse osmosis. The active layer of these membranes is in contact Other studies have investigated modications of the thin lm
with the feed solution and separates water from contaminants through
a solution-diusion mechanism.3 Decades of research and develop- Received: May 1, 2011
ment, both in industry and academia, have resulted in the success of Accepted: June 13, 2011
aromatic polyamide TFC membranes. Currently, polyamide is the Published: June 30, 2011

r 2011 American Chemical Society 2869 dx.doi.org/10.1021/am200536p | ACS Appl. Mater. Interfaces 2011, 3, 28692877
ACS Applied Materials & Interfaces FORUM ARTICLE

interface by postfabrication procedures, such as employing sur- phosphate buffered saline (PBS, BioReagent) were used as received
face coating techniques established in polymeric materials re- (Sigma-Aldrich, St. Louis, MO). Sodium chloride (NaCl, crystals, ACS
search. These methods entail attachment of hydrophilic or reagent) from J.T. Baker (Phillipsburg, NJ) was used to adjust ionic
antimicrobial macromolecules via graft polymerization,9,10 free strength of the reacting solutions and for the membrane performance
radical polymerization,11 or coating by deposition.12
15 Results evaluation. The pH of the reacting solutions was adjusted using
obtained using the previously stated, macromolecule-based hydrochloric acid (HCl) or sodium hydroxide (NaOH). Propidium
procedures suggest that progress toward the prevention of iodide (PI) and 40 ,6-diamidino-2-phenylindole (DAPI) were purchased
fouling and biofouling will be limited and call for a paradigm from Invitrogen (Carlsbad, CA). Unless specified, all chemicals were
shift in antifouling membrane design.6 dissolved in deionized (DI) water obtained from a Milli-Q ultrapure
The functionalization of thin-lm membranes with active water purification system (Millipore, Billerica, MA).
nanomaterials may be a novel strategy to tailor membrane antifouling Single-walled carbon nanotubes (SWNTs, lot number SG65
000

characteristics. Nanomaterial-specic properties can be imparted to 0031), produced by the CoMoCAT process, were obtained from
SouthWest NanoTechnologies, Inc. (Norman, OK). Manufacturer
the membrane by creating nanocomposite structures or coating the
specications of the SWNTs include: tube diameter of 0.8 ( 0.1 nm,
membrane with the nanomaterial. Incorporation of nanoparticles
carbon content >90% by weight, >50% of tubes are (6,5) chirality, and
within the polymeric lm during interfacial condensation is the route
>90% of tubes are semiconducting.
used for thin-lm nanocomposite membranes. For example, silver
Thin-Film Composite Polyamide Membranes. TFC mem-
nanoparticles,16 titanium oxide17 nanoparticles, and silver-exchange
branes were prepared by interfacial polymerization of polyamide onto
zeolites18 have been employed to yield membranes with enhanced commercial polysulfone (PSf) ultrafiltration membranes (PS20, Sepro
antimicrobial properties. One drawback of this type of modication is Membranes, Oceanside, CA), adapting a procedure described in our
that incorporation of nanomaterials during polymerization also previous publication.29 The PSf support was immersed in a 3.4 wt % aqueous
aects the polymer chemistry and, in turn, the thin lm separation MPD solution for 120 s, and an air knife was used to remove the excess
performance in ways that are dicult to predict. Furthermore, the solution from the membrane surface. Next, the MPD-saturated support
majority of the nanomaterial mass is buried in the bulk polymer and membrane was immersed for 60 s in a 0.15 wt % TMC in Isopar G, a
thus, the active nanomaterial surface is rendered useless. Postfabrica- nonpolar organic solvent (Univar, Redmond, WA), to form the ultrathin PA
tion functionalization may be a more eective option for improved layer by interfacial polymerization. Chemicals used for post-treatment of the
control of nanomaterial spatial localization. Recent studies in this polyamide were sodium hypochlorite (NaOCl, available chlorine 10
15%,
direction have proposed strategies to modify the polyamide thin lm Sigma-Aldrich) and sodium bisulfite (NaHSO3, Sigma-Aldrich).
surfaces with TiO2 nanoparticles,19
21 which under UV irradiation SWNT Purification and Functionalization. As-received
were observed to exhibit antimicrobial activity. SWNTs were purified by heating in concentrated hydrochloric acid
Single-walled carbon nanotubes (SWNTs) have been proposed (37%) for 12 h at 70 C. The SWNTs were then rinsed repeatedly with
as antimicrobial agents for a variety of environmental applications.22 DI water until neutral pH was attained. Amorphous carbon was removed
Although the mechanism of toxicity is still not fully understood, from the SWNTs by oxidation at 350 C for 6 h. Treatment of SWNTs by
recent studies suggest a combination of cell membrane perturbation ozone was used for generating sidewall defects to facilitate functionalization,
and oxidative stress as the main cause of bacterial inactivation.23 The reduce the length of the nanotubes, and enhance dispersion in aqueous
physicochemical mechanism of toxicity provided by SWNTs ensures solution.30
32 Approximately 5 mg of purified SWNTs were sonicated for 1
inactivation of a broad range of microorganisms,24 without stimulat- h in 20 mL of DMF. Mats of SWNTs were then prepared by filtering the
ing antibiotic resistance.22 Furthermore, SWNTs can be used as suspension through a 5-m Omnipore PTFE membrane (Millipore) to
stand-alone nanomaterials, as opposed to titanium dioxide nanopar- form a SWNT film. At least 100 mL of ethanol were then filtered through the
ticles, which require UV irradiation to ensure adequate bacterial SWNT-coated filter to remove residual solvent, followed by an extensive
inactivation.19,20 Finally, SWNTs represent nondepleting biocides, as wash with DI water to remove residual ethanol. After air-drying the SWNT-
the related toxicity process does not involve the leaching of ionic or coated filter, it was placed in a UV/O3 generator in ambient laboratory air for
chemical species from the bulk of the nanomaterials, as in the case of 10 h (BioForce Nanosciences, Inc., Ames, IA).
silver nanoparticles.25
28 SWNT Characterization. Raman spectra were acquired utilizing
In this study, we propose a novel approach to impart nano- an excitation wavelength of 532 nm on a WITec CRM 200 Spectro-
material-specic properties to the active surface of thin-lm photometer. Thermogravimetric analysis (TGA) (SETSYS 16/18) was
performed from 200 to 1000 C at a heating rate of 10 C/min. X-ray
composite (TFC) polyamide membranes. In particular, antimicro-
photoelectron spectroscopy (XPS, Thermo Scientific ESCALAB 250) and
bial properties were conferred to polyamide membranes by cova-
transmission electron microscopy (TEM, FEI Titan 80
300) of SWNTs
lently binding cytotoxic single-walled carbon nanotubes to their were conducted at CAMCOR (University of Oregon, Eugene, OR).
surfaces. Our results suggest the potential of these membranes to
SWNT-Polyamide Membrane Surface Reaction. Scheme 1
delay the onset of biofouling in membrane-based separation applica- describes the protocol for covalent binding of SWNTs to the PA surface
tions, as well as the promise of the proposed functionalization to create SWNT-TFC membranes. All reaction chemistry was com-
platform in a variety of other systems requiring reactive surfaces. pleted in an aqueous solution. A 7-cm diameter TFC membrane was
loaded into a custom-made dead-end filtration unit with only the active
MATERIALS AND METHODS (top) surface accessible. The dead-end filtration unit is made of stainless
steel and equipped with a built-in magnetic stir bar. EDC (4 mM) and
Materials and Chemicals. N-(3-Dimethylaminopropyl)-N0 - NHS (10 mM) were dissolved in 125 mL of 0.5 M NaCl solution
ethylcarbodiimide hydrochloride (EDC, 98%), N-hydroxysuccinimide buffered at pH 5 (MES), and the solution was put in contact with the
(NHS, 98%), ethylenediamine (ED, BioXtra), HEPES (>99.5%), MES membrane surface for 1 h (Step 1A-B of Scheme 1). During this step, the
monohydrate (>99.0%, BioXtra), N,N-dimethylformamide (DMF, an- native carboxylate groups of the polyamide surface were converted into
hydrous, 99.8%), 1-methyl-2-pyrrolidinone (NMP, anhydrous, 99.5%), intermediate amine-reactive esters for cross-linking.33,34 The activated
dimethyl sulfoxide (DMSO, 99.5%), 1,3-phenylenediamine (MPD, esters were used promptly for reaction with ethylenediamine (ED) to
>99%), 1,3,5-benzenetricarbonyl trichloride (TMC, 98%), and form amide bonds by contacting the membrane surface with a 125-mL

2870 dx.doi.org/10.1021/am200536p |ACS Appl. Mater. Interfaces 2011, 3, 28692877

ACS Applied Materials & Interfaces FORUM ARTICLE

Scheme 1. Procedure to Covalently Bind SWNTs to the Membrane Surfacea

a
The native carboxylic groups of the polyamide active lm are converted into semistable amine-reactive esters using an EDC/NHS solution at pH 5
(MES buer), followed by reaction with ethylenediamine at pH 7.5 (HEPES buer) and contact with a sonicated suspension of functionalized SWNTs
in the presence of EDC/NHS.

solution of 10 mM ED and 0.15 M NaCl buffered at pH 7.5 (HEPES) for aluminum coating were employed (Tap300A, Bruker Nano Inc., Camarillo,
30 min (Step 2 of Scheme 1). CA). The probe had a spring constant of 40 N/m, resonance frequency of
Approximately 10 mg of ozonized SWNTs were dispersed in 250 mL 300 kHz, tip radius of 8 ( 4 nm, and cantilever length of 125 ( 10 m. Air-
of solution at pH 7 (HEPES and NaOH) by high-power probe sonication dried membranes were scanned in air at 10 randomly selected scan posi-
(40 W, Sonicator 3000, Misonix Inc., Farmingdale, NY) for 3 h. EDC and tions. The surface roughness of each membrane was quantied as the root-
NHS were then dissolved in the sonicated suspension at concentrations of 4 mean-square (rms) roughness, average roughness Ra, maximum roughness
and 10 mM, respectively. The resulting solution was added to the ltration Rmax, and surface area dierence (SAD, determined by dividing the actual
chamber to initiate tethering of SWNTs to the amine groups at the mem- surface area by the planar area).
brane surface (Step 3 of Scheme 1). Constant stirring was provided and a Surface hydrophilicity was evaluated from DI water contact angle
pressure of 50 psi (13.8 bar) was applied to facilitate contact between the measurements using the sessile drop method (VCA Video Contact Angle
SWNTs and the membrane surface. Reaction was carried out at room System, AST Products, Billerica, MA). The system is equipped with
temperature (23 C) for 3 h. Aggregation of the SWNT bundles during this software to determine the left and right contact angles (VCA Optima XE).
reaction time was ruled out, based on dynamic light scattering measurements To account for variations between dierent measurements on the same
of SWNT hydrodynamic diameter carried out for 45 min under the same surface, several desiccator-dried samples were tested in eight random
conditions as those used for the reaction. The measured hydrodynamic locations. The highest and the lowest equilibrium angles were discarded
radius of the SWNTs was 540 ( 38 nm and remained constant during the and the remaining data were averaged. The relative wettability of the
experiment. membranes was evaluated by calculating the solid
liquid interfacial free
After the reaction, the SWNT suspension was discarded and the energy adjusted for roughness by incorporating the surface area dierence,
membrane surface was briey contacted with a solution at pH 11 (NaOH) SAD, determined from AFM measurements35,36
to restore the unreacted carboxylic groups. Finally, the membrane surface  
was rinsed thoroughly with DI water. The strength of the covalent bonds cos

GSL L 1 +
between SWNTs and the surface of the membranes was challenged by SAD
subjecting the SWNT-TFC membranes to two 7-min cycles of bath
sonication (FS60, Fisher Scientic Co., Pittsburgh, PA). where is the average contact angle and L is the pure water surface tension
Membrane Characterization. Pure water permeability, A, and (72.8 mJ/m2 at 25 C).
NaCl permeability, B, of the hand-cast membranes were evaluated in a Surface morphology of membranes was visualized by scanning
dead-end filtration unit before and after reaction with the SWNTs, following electron microscopy, SEM (FEI XL-30 ESEM). The desiccator-dried
procedures described in our previous publication.29 The loaded membrane membranes were sputter-coated with gold and analyzed at a voltage of
(area was 38.5 cm2) was first compacted with DI water at an applied 10 kV. The thin lm of the membranes was analyzed using TGA. To
pressure of 100 psi (27.6 bar) until the permeate flux reached a steady state. prepare the samples for TGA measurements, membranes were placed in
Salt rejection was characterized by keeping the same applied pressure and NMP for 30 min to dissolve the underlying PSf. Once the support layer was
measuring rejection of 20 mM NaCl solution using a calibrated conductivity completely dissolved, the top active layer was collected and dried overnight
meter (Oakton Instruments, Vernon Hills, IL). All experiments were car- at 65 C to volatilize the residual NMP.
ried out at a fixed temperature of 23 C. Evaluation of Antibacterial Activity of SWNTs and of
Membrane characterization was performed on polyamide lms as SWNT Functionalized Membranes. Escherichia coli (E. coli) K12
cast, and on SWNT-TFC membranes before and after the bath sonication MG1655 was used as the model bacteria. The E. coli cells were incubated in
protocol described above. Membrane surface roughness was analyzed using Luria
Bertani (LB) broth with 50 mg/L kanamycin at 37 C and harvested
a Multimode AFM (Veeco Metrology Group, Santa Barbara, CA) in at midexponential growth phase. The bacterial suspension was washed three
tapping mode. Symmetric silicon probes with 30-nm-thick back side times by centrifugation (Sorvall SS-34) at 15 000 rpm for 3 min and

2871 dx.doi.org/10.1021/am200536p |ACS Appl. Mater. Interfaces 2011, 3, 28692877

ACS Applied Materials & Interfaces FORUM ARTICLE

Figure 1. Characteristics of the SWNTs. (A) Oxygen to carbon ratio as obtained by X-ray photoelectron spectroscopy (XPS) analysis for the as-received
(black), puried (red), and ozonized (blue) SWNTs. (B) Dierential thermogravimetric (DTG) plots normalized to the total initial mass of sample for
the as-received (black), puried (red), and ozonized SWNTs in duplicate (solid and dashed blue curves). (C) Representative TEM micrograph of the
ozonized SWNTs showing bundles of single-walled carbon nanotubes. (D) Raman spectra acquired using 532 nm excitation and relative D/G ratio for
the puried (red) and ozonized (blue) SWNTs.

resuspended in an isotonic saline solution (0.9% NaCl) to remove residual sonicated for 7 min. This procedure was observed to resuspend bacterial cells
macromolecules and other growth medium constituents. from surfaces without aecting their viability,38 which was conrmed in our
To evaluate SWNT cytotoxicity, we used a SWNT-coated lter laboratory (data now shown). The PBS suspensions of cells were spread on
assay.23,37 Briey, ozonized carbon nanotubes were dispersed in DMSO at solid LB agar growth plates with kanamycin. In addition, the sonicated
a concentration of 0.1 to 0.2 mg/mL using a 15-min probe sonication. coupons were also rinsed with PBS solution and gently tapped on LB agar
SWNT-coated lters were then prepared following the protocol de- plates in 5 dierent locations to account for live cells that were not resus-
scribed earlier in this paper. To facilitate direct contact between the pended during sonication. The colonies formed after 12 to 16 h incubation at
SWNT mat and the bacteria, E. coli cells (1  107 cells/mL) were gently 37 C were counted. The number of colonies counted from the SWNT-
deposited by vacuum ltration onto the SWNT-coated lters. After a 1-h TFC surfaces was normalized by the coupon area and by the number of live
incubation, the cells were stained with PI for 15 min, and then counter- cells counted from PA control membranes. The procedure was repeated for
stained with DAPI for 5 min in the dark. Fluorescence images were then 3 separately cast and modied membranes. In dierent experiments, bacteria
taken under an epiuorescence microscope (Olympus BX40) with a U lter were not resuspended from PBS-rinsed coupons. Instead, cells were xed
(excitation 364 nm/emission 440 nm) for detecting cells stained with both with 3% glutaraldehyde, sputter-coated with gold, and then viewed under
PI and DAPI, and with an IB lter (excitation 464 nm/emission 604 nm) for an SEM.
detecting cells stained with PI. Ten representative images were taken at 10X
magnication at various locations. Dead cells and the total number of cells
were determined by direct cell counting on the SWNT-coated lter. The RESULTS AND DISCUSSION
percentage of dead cells (or loss of viability) was determined from the ratio
of the number of cells stained with PI divided by the total number of cells. SWNT Characteristics. Short, highly functionalized, and
To evaluate inactivation of bacteria by the SWNT functionalized debundled SWNTs exhibit enhanced antibacterial activity, due to
membranes, a plate counting method was adopted. Control membranes and their enhanced dispersivity in aqueous solution and the larger
sonicated SWNT-TFC membranes were attached to a glass plate using contact area with the bacterial cells.39
41 In this study, ozonolysis
laboratory tape, with only the active surface accessible to solution. E. coli cells was employed to shorten and functionalize the nanotubes before
(1  107 cells/mL) were pipetted onto the surfaces using approximately they were reacted with the membranes. Previous studies have shown
0.5 mL of suspension per 1 cm2 of membrane. The cell-covered surfaces that ozone can preferentially attack the defect sites at the nanotubes
were then incubated for 1 h at 23 C under gentle stirring. After this time, the walls,42 thereby creating carboxylic functional groups,31,32,43 redu-
excess bacterial suspension was discarded and two 1-in. coupons were cing particle size,31,43 enhancing dispersion in water,43 and max-
punched for each sample. The samples were rinsed extensively with PBS and imizing the O/C ratio.32,43
each coupon was placed in a 50-mL Falcon tube containing 10 mL of PBS XPS data (Figure 1A) demonstrated an increase in the O/C
solution. The Falcon tubes were then immersed in water and gently bath ratio for the ozonized sample compared to the as-received or puried
2872 dx.doi.org/10.1021/am200536p |ACS Appl. Mater. Interfaces 2011, 3, 28692877
ACS Applied Materials & Interfaces FORUM ARTICLE

SWNTs. This observation indicates a higher amount of oxidized Visual inspection of the membranes before and after reaction
carbon to elemental carbon and a higher density of oxygen-rich with SWNTs yields insight into the extent of nanomaterial
functional groups. It is likely that the short-lived ozonide intermedi- attachment. Figure S3 in the Supporting Information shows
ates formed during ozonation are converted into carboxylic acids in digital images of active layer surfaces and oven-dried thin lms for
the presence of UV irradiation and atmospheric moisture.30,32 the control and SWNT-TFC membranes. Surface functionaliza-
Raman analysis of the puried and ozonized SWNTs is tion was apparent through the darkening of the membrane upon
presented in Figure 1D. The diameter-dependent frequency of reaction, suggesting a signicant presence of tightly anchored
the radial breathing mode (RBM) appeared at a wavenumber of SWNTs even after membrane sonication. Ultrasonication re-
276 cm
1 for both samples. The tube diameter, d, was estimated sulted in the removal of the loosely bound nanomaterials and the
from the RBM37,44 (d = 248/), yielding a value of approximately larger SWNT bundles and aggregates. The anchored SWNTs
0.9 nm, consistent with the manufacturer specications. The that were able to endure sonication suggest an irreversible
ratio between the SWNT disorder-induced D-band (1320 cm
1) functionalization of the membrane surface.
and the tangential mode G-band (1585 cm
1) is an empirical The amount of SWNTs on a sonicated SWNT-TFC mem-
measure of the sample purity and number of defects.44 This ratio brane was estimated by gravimetric analysis to be approximately
was higher for the ozonized sample, indicating that the functio- 7 wt % of the total thin lm, equivalent to a total SWNT loading of
nalization process increased the concentration of defects and of 0.03 mg/cm2 on the functionalized membrane (see Figure S4 in the
sp3-hybridized carbon atoms on the SWNT sidewalls.30,45 The Supporting Information). TGA measurements performed on the
increase in the D/G ratio measured by Raman was similar to the thin lms collected after PSf dissolution showed a shift in the material
increase in the O/C ratio obtained by XPS, suggesting that the thermal degradation peak to a lower temperature (see Figure S4
increased defect density observed by Raman was correlated to in the Supporting Information). A lower thermal stability suggests
the oxidation of the nanotube sidewalls.32 that the presence of SWNTs may decrease the intermolecular
The ndings from the XPS and Raman analyses were con- bonding and the aromaticity of polyamide during sample pre-
rmed by thermogravimetric analysis. DTG plots (Figure 1B) show paration,53,54 and possibly accelerate the diusion of oxygen and
sharpening and shifting of the mass loss peak to a lower temperature volatile thermo-oxidative products between the gas phase and the
for the ozonized SWNTs (blue) with respect to the puried SWNTs bulk polymer.55
(red). This result is representative of the production of sp3-oxy Figure 2 presents representative SEM micrographs and AFM
functional groups that reduce the oxidative stability of the SWNTs,46 images of control membranes and sonicated SWNT-TFC mem-
and of the presence of a greater number of sidewall defects that causes branes. Roughness data from AFM analysis is also reported. The
the nanotubes to combust at a lower threshold temperature.47 representative topographic image (Figure 2A) and SEM surface
Representative TEM micrographs of ozonized SWNTs micrograph (Figure 2C) of a control membrane show a uniform
(Figure 1C) conrmed that nanotubes were single-walled and ridge-and-valley morphology, which is typical of polyamide thin
that their sidewall structure remained intact after treatment. lms formed by interfacial condensation.4 The membrane sur-
Negligible metal impurities and amorphous carbon were visible. face roughness (rms = 42.2 ( 12.9 nm) was at the lower end of
TGA data (see Figure S1 in the Supporting Information) the range observed for commercial reverse osmosis polyamide
indicated that the purication and functionalization procedures membranes.15 The lm thickness can be estimated from the
decreased the concentration of residual metal impurities from roughness data (300 nm), a value characteristic of layers cast
4.3% for the as-received to 2.2% and 2.1% for the puried and under similar conditions.56
ozonized SWNTs, respectively. Signicant dierences were observed via AFM for the func-
Antibacterial Properties of SWNTs. E. coli K12 was used to tionalized SWNT-TFC membranes as compared to the unmo-
evaluate the SWNT cytotoxicity. Bacteria in direct contact with died membranes. The ridge-and-valley morphology appeared
SWNT-coated filters were incubated for 1 h at 37 C. The more leveled and at times completely attened. Furthermore,
bacteria-coated SWNT filters were then stained and viability loss bulkier features were detected that correlated well with the size of
was determined by fluorescence microscopy. The percent loss of the larger SWNT aggregates measured by DLS. Figure 2B is a
E. coli viability for the purified and ozonized SWNTs was >95% representative image displaying some of these characteristics.
and significantly higher than that of the purified-only carbon These changes in membrane morphology are attributed to the
nanotubes (80%). The results are in agreement with previous presence of SWNTs at the surface. No change in surface rough-
studies on the antibacterial properties of SWNTs.23,37,40,41 ness was observed after exposing the membranes to only those
SWNT antimicrobial activity requires direct contact with the chemical compounds used during reaction with SWNTs (see
bacterial cells, and the cytotoxic effects involve a combination of Figure S5 in the Supporting Information).
cell membrane perturbation and oxidative stress.23,40,48 The The SEM micrograph in Figure 2D shows that the ridge-and-
observed cytotoxic properties of SWNTs correlated with the valley features of the SWNT-TFC membrane surface are overlain
extent of ozonolysis treatment, as shown in Figure S2 of the by other features of comparable or larger size. These features are
Supporting Information. assumed to be bundles or aggregates of SWNTs. SWNTs were
Membrane Surface Characteristics. The hand-cast thin-film observed to be scattered and interspersed on the polyamide and
polyamide membranes possess native surface carboxylate groups may be the reason for the apparent smoother morphology
resulting from incomplete reaction and hydrolysis of the TMC acyl imaged by AFM. Large SWNT aggregates were also observed
halides during interfacial polymerization.49 These reactive moieties on the surface, comparable to the larger size features on the
were utilized as SWNT binding sites, following the reaction membrane surface observed by AFM. More SEM micrographs of
procedure delineated earlier in this paper. During reaction, covalent the sonicated SWNT-TFC membranes are presented in the
cross-linking between the carboxylic groups at the membrane Supporting Information (Figure S6). Figure S6 in the Supporting
surface and at the SWNT walls occurred via amide bond formation Information also shows SEM surface micrographs of SWNT-
with ethylenediamine (Scheme 1).45,50
52 TFC membranes before sonication. Very large SWNT
2873 dx.doi.org/10.1021/am200536p |ACS Appl. Mater. Interfaces 2011, 3, 28692877
ACS Applied Materials & Interfaces FORUM ARTICLE

Figure 2. Imaging by (A, B) AFM and (C, D) scanning electron microscopy of the membrane surface (A, C) before and (B, D) after functionalization.
AFM shows the typical ridge-and-valley morphology of polyamide for the control membrane (A), and the smoothening of these ridge-and-valley
structures and appearance of larger features for the functionalized membrane (B). These structures are observable in the representative SEM
micrographs: the SWNT-reacted membrane (D) shows the presence of bundled SWNTs overlaying the polyamide surface, some of which are
highlighted in blue. Roughness parameters are reported at the bottom of the gure.

aggregates were apparent, which detached during bath sonica- of the surface water interaction is also presented, which incorpo-
tion. The low surface-to-volume ratio of these large aggregates rates SAD values to account for surface roughness. Polyamide
minimizes the density of binding sites between the mem- control membranes had a contact angle of 70, consistent with
brane and SWNTs, which facilitates detachment during previous studies.35 The presence of SWNTs on the membrane
ultrasonication. surface did not signicantly aect surface wettability, as the average
The higher average roughness of the modied membranes is equilibrium contact angle was within experimental error of the value
attributed to the signicant increase in the value of Rmax, which is measured for the control polyamide surface. Consequently, the
related to the large SWNT aggregates. However, a comparable calculated interfacial free energy of water was comparable for poly-
value of surface area dierence (SAD) to the polyamide control was amide and for SWNT-TFC surfaces (
95 mJ/m2). The carbon
observed, due to smoothing of the ridge-and-valley features dis- nanotubes become relatively hydrophilic as a result of the functio-
cussed above. From AFM and SEM data analysis, we can conclude nalization by the ozonolysis treatment. This nding was conrmed
that the SWNT coverage of the membrane was heterogeneous by using the same highly functionalized SWNTs to functionalize
on a microscopic scale. These results conrm the complexity and more hydrophobic polyamide membranes. The same equilibrium
diculty of dispersing and debundling narrow-diameter single- contact angle as that presented in Figure 3 was observed after
walled carbon nanotubes in aqueous solutions without the use of modication of the more hydrophobic membranes (see Figure S7 in
surfactants. Future studies should focus on developing methods to the Supporting Information).
achieve more homogeneous functionalization through optimization Membrane Transport Properties. The pure water perme-
of nanomaterial dispersion. ability coefficient, A, and salt permeability coefficient, B, were
Figure 3 presents contact angle data of DI water in contact measured for five separately cast and modified membranes,
with the surface of the control, SWNT-TFC, and sonicated before and after reaction with SWNTs. A small increase in A
SWNT-TFC membranes. The calculated interfacial free energy was observed for the SWNT-TFC membranes compared to the
2874 dx.doi.org/10.1021/am200536p |ACS Appl. Mater. Interfaces 2011, 3, 28692877
ACS Applied Materials & Interfaces FORUM ARTICLE

control membranes. A small loss of NaCl rejection was also of the concentration and contact time with ethylenediamine
measured, translating into a small increase in B (see Figure S8 in would minimize the impact on membrane transport properties.
the Supporting Information). Overall, the membrane transport Antimicrobial Properties of Membrane Active Layer. The
parameters did not change significantly, indicating that postfab- objective of postfabrication functionalization is to confer biocidal
rication surface functionalization is a nondestructive route that functionality to the membrane surface through covalent tethering of
minimizes performance loss of the polyamide active layer. The SWNTs. The antibacterial activity of the SWNT-TFC membranes
relatively small increase in A and B is attributed to the contact of was assessed by a plate count assay using E. coli as model bacteria.
the membrane with ethylenediamine during reaction, as the same Cells were incubated with the membranes for 1 h in 0.9% NaCl
behavior was observed for membranes that were subjected to the solution, and then resuspended in PBS solution using bath sonica-
same reaction treatment, but in the absence of SWNTs (see tion. The suspension was spread on LB agar plates and colonies were
Figure S8 in the Supporting Information). Further optimization counted after overnight incubation. Three functionalized membranes
and three control membranes were evaluated. For each membrane
tested, a new membrane was cast, and two different 1-in. diameter
coupons from each cast membrane were tested independently.
Figure 4B presents the average surface density and standard
deviation of culturable cells on the SWNT-TFC membranes,
normalized by the number of culturable cells on each respective PA
control membrane. The number of culturable bacteria on the
SWNT-TFC membranes was signicantly lower (44%) than that
on the PA control membranes. Furthermore, the loss of viability of
the replicate membrane samples was quite similar, yielding a
negligible standard deviation. This result suggests that although
membrane functionalization is heterogeneous on the microscale, a
statistically comparable amount of cytotoxic SWNTs are available
on the surface of the larger size (1-in. diameter) membrane samples
utilized for the enumeration of culturable cells.
To validate the culturable bacterial enumeration method by
Figure 3. Surface tension and contact angle with deionized water for resuspension, the results were compared to experiments where
(black) control, (blue) reacted, and (red) reacted and sonicated mem- the test membranes were pressed onto agar plates for direct
branes. Bars and standard deviations represent the average of six inde-
pendently cast and functionalized membranes. The average membrane

transfer of bacteria. The number of resuspended culturable cells
water interfacial free energy for control polyamide membranes and for was always at least 2 orders of magnitude greater than the number
membranes that were modied and sonicated was calculated from of live cells counted after contacting the sonicated test coupons
contact angle and roughness data. The computed value is reported on LB agar plates. Live cells still present on the sonicated test
below the plot of contact angle for each respective bar. coupons were enumerated, and comparable results to those

Figure 4. Bacterial inactivation properties of the membrane surface. (B) Colony-forming units enumerated from E. coli bacteria resuspended from functionalized
SWNT-TFC membranes (red) normalized with those of control membranes (shaded). The tests involved incubation of bacterial suspension in contact with the
membrane surface for 1 h in 0.9% NaCl at room temperature (23 C). Bars with standard deviation represent the average of three separately cast and reacted
membranes. The gure presents SEM micrographs displaying E. coli at the surface of a (A) control membrane and (C) a SWNT-functionalized membrane at the
end of the cytotoxicity test (some cells with lost integrity are highlighted in orange). (D) Magnied view of the surface of a SWNT-TFC membrane with E. coli
cells: the magnied view is a representative image and does not correspond to the enlargement of micrograph C.

2875 dx.doi.org/10.1021/am200536p |ACS Appl. Mater. Interfaces 2011, 3, 28692877

ACS Applied Materials & Interfaces FORUM ARTICLE

observed from the resuspension assay were obtained. Specically, contact angle of DI water with surface of hand-cast membrane
46 ( 15% of live cells were counted on the SWNT-TFC test on hand-cast support before and after reaction with SWNTs
coupons relative to the PA control coupons. The strong agree- (Figure S7); transport parameters of membranes before and after
ment between these two experiments validates the sonication reaction with (Figure S8A) and without (Figure S8B) SWNTs;
and resuspension assay for enumerating culturable bacteria surface SEM micrographs of control membranes and sonicated
attached to the membrane surface. SWNT-TFC membranes after bacterial cytotoxicity static test
Figure 4 also presents representative SEM images of E. coli in (Figure S9). This material is available free of charge via the
contact with control (Figure 4A) and sonicated (Figure 4C,D) Internet at http://pubs.acs.org.
SWNT-TFC membrane surfaces after 1 h of incubation in 0.9%
NaCl. Viable and dead bacteria showed distinct dierences in cell
AUTHOR INFORMATION
morphology, as shown in past studies of bacteria in contact with
SWNTs.40 Most of the cells on the PA control membranes Corresponding Author
appeared healthy and their membranes were intact. In contrast, a *Tel.: +1 (203) 432-2789. Fax: +1 (203) 432-4387. E-mail:
large number of bacterial cells on the SWNT-TFC sample had [email protected].
lost their membrane integrity. Specically, the impacted cells
appeared either attened/dehydrated or displayed a compro-
mised cell membrane. The various bacterial morphologies are ACKNOWLEDGMENT
observable in the higher magnication SEM found in Figure 4D. We acknowledge the NWRI-AMTA Fellowship for Mem-
Additional images of the cells on membrane surfaces are pre- brane Technology, awarded to A.T., and the Water-CAMPWS, a
sented in Figure S9 in the Supporting Information. A higher and Science and Technology Center of Advanced Materials for the
more homogeneous distribution of SWNTs on the surface would Purication of Water with Systems under the National Science
likely enhance the SWNT-TFC membrane surface cytotoxicity. Foundation Grant CTS-0120978. We also acknowledge the
CAMCOR facilities and technician for XPS and TEM analysis.
CONCLUDING REMARKS The CAMCOR TEM Facility is supported by grants from the
W. M. Keck Foundation, the M. J. Murdock Charitable Trust, the
In this study, we proposed a methodology for functionaliza- Oregon Nanoscience and Microtechnologies Institute, the Air
tion of surfaces containing native moieties that can be exploited Force Research Laboratory (under Agreement FA8650-05-1-
to tether nanomaterials with targeted functionalities. The possi- 5041), and the University of Oregon.
bilities for functionalization are virtually limitless, as the variety of
nanoparticle interfacial properties and methods to create surface
binding sites are vast. This range of platforms represents a REFERENCES
remarkable opportunity to target and increase performance of (1) Ridgway, H. F.; Safarik, J. Biofouling and Biocorrosion in Industrial
systems requiring specic surface functionalities while maintain- Water Systems 1991, 81111.
ing intrinsic properties of the bulk material. In particular, addition (2) Herzberg, M.; Elimelech, M. J. Membr. Sci. 2007, 295, 1120.
of antimicrobial nanomaterials to a membrane surface to increase (3) Paul, D. R. J. Membr. Sci. 2004, 241, 371386.
the availability of their active/reactive surface is a crucial factor in (4) Petersen, R. J. J. Membr. Sci. 1993, 83, 81150.
optimizing performance of membrane-based separation systems (5) Glater, J.; Hong, S. K.; Elimelech, M. Desalination 1994,
that are prone to biofouling. 95, 325345.
(6) Rana, D.; Matsuura, T. Chem. Rev. 2010, 110, 24482471.
To this purpose, we demonstrated functionalization of dense
(7) Liu, L. F.; Yu, S. C.; Zhou, Y.; Gao, C. J. J. Membr. Sci. 2006,
hand-cast membranes with single-walled carbon nanotubes that 281, 8894.
were selected for their well documented, nondepleting antimi- (8) Liu, M. H.; Yu, S. C.; Tao, J.; Gao, C. J. J. Membr. Sci. 2008,
crobial activity. The biocidal properties of the SWNTs were 325, 947956.
successfully conferred to the membrane surface upon covalent (9) Kang, G. D.; Liu, M.; Lin, B.; Cao, Y. M.; Yuan, Q. Polymer 2007,
surface modication. Analysis of the morphological character- 48, 11651170.
istics of the functionalized surfaces suggests that an improved (10) Wei, X. Y.; Wang, Z.; Zhang, Z.; Wang, J. X.; Wang, S. C.
aqueous SWNT dispersion will result in a greater and more J. Membr. Sci. 2010, 351, 222233.
homogeneous surface coverage. (11) Niu, Q. J. Reverse Osmosis Membrane with Branched Poly-
(Alkylene Oxide) Modied Antifouling Surface. U.S.
(12) Wilson, D. R. Permeable membranes having high ux-density and
ASSOCIATED CONTENT low fouling-propensity. U.S. 4756835
(13) Yu, S. C.; Lu, Z. H.; Chen, Z. H.; Liu, X. S.; Liu, M. H.; Gao, C. J.
bS Supporting Information. Thermogravimetric analysis J. Membr. Sci. 2011, 371, 293306.
(TGA) of puried and ozonized SWNTs (Figure S1); E. coli (14) Tang, C. Y. Y.; Kwon, Y. N.; Leckie, J. O. J. Membr. Sci. 2007,
inactivation rate by SWNTs as a function of ozonolysis time for 287, 146156.
both puried SWNTs (Figure S2A) and unpuried SWNTs (15) Tang, C. Y. Y.; Kwon, Y. N.; Leckie, J. O. Desalination 2009,
(Figure S2B); digital images of membrane surfaces (Figure S3A) 242, 168182.
and dry thin lm (Figure S3B); estimation of SWNT loading on (16) Lee, S. Y.; Kim, H. J.; Patel, R.; Im, S. J.; Kim, J. H.; Min, B. R.
Polym. Adv. Technol. 2007, 18, 562568.
sonicated SWNT-TFC membranes and TGA measurements on
(17) Lee, H. S.; Im, S. J.; Kim, J. H.; Kim, H. J.; Kim, J. P.; Min, B. R.
the thin lm of polyamide membranes and SWNT-TFC mem- Desalination 2008, 219, 4856.
branes (Figure S4); surface roughness and representative AFM (18) Lind, M. L.; Jeong, B. H.; Subramani, A.; Huang, X. F.; Hoek,
images of membranes reacted without SWNTs (Figure S5); E. M. V. J. Mater. Res. 2009, 24, 16241631.
surface SEM micrographs of control membranes and SWNT- (19) Kim, S. H.; Kwak, S. Y.; Sohn, B. H.; Park, T. H. J. Membr. Sci.
TFC membranes before and after sonication (Figure S6); 2003, 211, 157165.

2876 dx.doi.org/10.1021/am200536p |ACS Appl. Mater. Interfaces 2011, 3, 28692877

ACS Applied Materials & Interfaces FORUM ARTICLE

(20) Kwak, S. Y.; Kim, S. H.; Kim, S. S. Environ. Sci. Technol. 2001, (56) Mi, B. X.; Coronell, O.; Marinas, B. J.; Watanabe, F.; Cahill,
35, 23882394. D. G.; Petrov, I. J. Membr. Sci. 2006, 282, 7181.
(21) Mo, J.; Son, S. H.; Jegal, J.; Kim, J.; Lee, Y. H. J. Appl. Polym. Sci.
2007, 105, 12671274.
(22) Mauter, M. S.; Elimelech, M. Environ. Sci. Technol. 2008,
42, 58435859.
(23) Vecitis, C. D.; Zodrow, K. R.; Kang, S.; Elimelech, M. ACS Nano
2010, 4, 54715479.
(24) Kang, S.; Mauter, M. S.; Elimelech, M. Environ. Sci. Technol.
2009, 43, 26482653.
(25) Choi, O. Y.; Yu, C. P.; Fernandez, G. E.; Hu, Z. Q. Water Res.
2010, 44, 60956103.
(26) Fabrega, J.; Fawcett, S. R.; Renshaw, J. C.; Lead, J. R. Environ.
Sci. Technol. 2009, 43, 72857290.
(27) Hwang, E. T.; Lee, J. H.; Chae, Y. J.; Kim, Y. S.; Kim, B. C.; Sang,
B. I.; Gu, M. B. Small 2008, 4, 746750.
(28) Sotiriou, G. A.; Pratsinis, S. E. Environ. Sci. Technol. 2010,
44, 56495654.
(29) Tiraferri, A.; Yip, N. Y.; Phillip, W. A.; Schiman, J. D.;
Elimelech, M. J. Membr. Sci. 2011, 367, 340352.
(30) Cai, L. T.; Bahr, J. L.; Yao, Y. X.; Tour, J. M. Chem. Mater. 2002,
14, 42354241.
(31) Chen, Z. Y.; Ziegler, K. J.; Shaver, J.; Hauge, R. H.; Smalley,
R. E. J. Phys. Chem. B 2006, 110, 1162411627.
(32) Simmons, J. M.; Nichols, B. M.; Baker, S. E.; Marcus, M. S.;
Castellini, O. M.; Lee, C. S.; Hamers, R. J.; Eriksson, M. A. J. Phys. Chem.
B 2006, 110, 71137118.
(33) Grabarek, Z.; Gergely, J. Anal. Biochem. 1990, 185, 131135.
(34) Staros, J. V.; Wright, R. W.; Swingle, D. M. Anal. Biochem. 1986,
156, 220222.
(35) Ghosh, A. K.; Jeong, B. H.; Huang, X. F.; Hoek, E. M. V.
J. Membr. Sci. 2008, 311, 3445.
(36) Wenzel, R. N. J. Phys. Colloid Chem. 1949, 53, 14661467.
(37) Schiman, J. D.; Elimelech, M. ACS Appl. Mater. Interfaces
2011, 3, 462468.
(38) Obrien, R. D.; Lindow, S. E. Phytopathology 1989, 79, 619627.
(39) Upadhyayula, V. K. K.; Gadhamshetty, V. Biotechnol. Adv. 2010.
(40) Kang, S.; Herzberg, M.; Rodrigues, D. F.; Elimelech, M.
Langmuir 2008, 24, 64096413.
(41) Kang, S.; Pinault, M.; Pfeerle, L. D.; Elimelech, M. Langmuir
2007, 23, 86708673.
(42) Akdim, B.; Kar, T.; Duan, X. F.; Pachter, R. Chem. Phys. Lett.
2007, 445, 281287.
(43) Li, M. H.; Boggs, M.; Beebe, T. P.; Huang, C. P. Carbon 2008,
46, 466475.
(44) Dresselhaus, M. S.; Dresselhaus, G.; Saito, R.; Jorio, A. Phys.
Rep.
Rev. Sect. Phys. Lett. 2005, 409, 4799.
(45) Banerjee, S.; Wong, S. S. Nano Lett. 2004, 4, 14451450.
(46) Arepalli, S.; Nikolaev, P.; Gorelik, O.; Hadjiev, V. G.; Bradlev,
H. A.; Holmes, W.; Files, B.; Yowell, L. Carbon 2004, 42, 17831791.
(47) Hemraj-Benny, T.; Bandosz, T. J.; Wong, S. S. J. Colloid
Interface Sci. 2008, 317, 375382.
(48) Obraztsova, E. A.; Lukashev, E. P.; Zarubina, A. P.;
Parkhomenko, I. M.; Yaminsky, I. V. Moscow Univ. Phys. Bull. 2009,
64, 320323.
(49) Wamser, C. C.; Gilbert, M. I. Langmuir 1992, 8, 16081614.
(50) Dhar, S.; Liu, Z.; Thomale, J.; Dai, H. J.; Lippard, S. J. J. Am.
Chem. Soc. 2008, 130, 1146711476.
(51) Hirsch, A.; Vostrowsky, O. Func. Mol. Nanostruct. 2005,
245, 193237.
(52) Ma, H. Y.; Zhang, L. P.; Pan, Y.; Zhang, K. Y.; Zhang, Y. Z.
Electroanalysis 2008, 20, 12201226.
(53) Ghaemy, M.; Mighani, H.; Behmadi, H. J. Appl. Polym. Sci.
2008, 109, 23882394.
(54) Marani, D.; Di Vona, M. L.; Traversa, E.; Licoccia, S.; Beurroies,
I.; Llewellyn, P. L.; Knauth, P. J. Phys. Chem. B 2006, 110, 1581715823.
(55) Zhang, X. G.; Loo, L. S. Polymer 2009, 50, 26432654.

2877 dx.doi.org/10.1021/am200536p |ACS Appl. Mater. Interfaces 2011, 3, 28692877