Journal of Membrane Science 320 (2008) 248258

Contents lists available at ScienceDirect

Journal of Membrane Science

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

Poly(vinyl alcohol)/cellulose nanocrystal barrier membranes

Shweta A. Paralikar a , John Simonsen b, , John Lombardi c
a
Department of Wood Science & Engineering and Chemical Engineering, Oregon State University, Corvallis, OR 97330, United States
b
Department of Wood Science & Engineering, Oregon State University, Corvallis, OR 97330, United States
c
Ventana Research Corp., 2702 S. 4th Avenue, S.Tucson, AZ 85713, United States

a r t i c l e i n f o a b s t r a c t

Article history: In this study, barrier membranes were prepared from poly(vinyl alcohol) (PVOH) with different amounts
Received 22 February 2008 of cellulose nanocrystals (CNXLs) as ller. Poly(acrylic acid) (PAA) was used as a crosslinking agent to
Received in revised form 4 April 2008 provide water resistance to PVOH. The membranes were heat treated at various temperatures to optimize
Accepted 5 April 2008
the crosslinking density. Heat treatment at 170 C for 45 min resulted in membranes with improved water
Available online 12 April 2008
resistance without polymer degradation. Infrared spectroscopy indicated ester bond formation with heat
treatment. Mechanical tests showed that membranes with 10% CNXLs/10% PAA/80% PVOH were syner-
Keywords:
gistic and had the highest tensile strength, tensile modulus and toughness of all the membranes studied.
Cellulose
Nanocrystal
Polarized optical microscopy showed agglomeration of CNXLs at ller loadings greater than 10%. Differ-
Poly(vinyl alcohol) ential thermogravimetric analysis (DTGA) showed a highly synergistic effect with 10% CNXL/10% PAA/80%
Poly(acrylic acid) PVOH and supported the tensile test results.
Barrier Transport properties were studied, including water vapor transport rate and the transport of
trichloroethylene, a representative industrial toxic material. Water vapor transmission indicated that all
the membranes allowed moisture to pass. However, moisture transport was reduced by the presence of
both CNXLs and PAA crosslinking agent. A standard time lag diffusion test utilizing permeation cups was
used to study the chemical barrier properties. The membranes containing 10% CNXLs or PAA showed
signicantly reduced ux compared to the control. The CNXLs were then modied by surface carboxy-
lation in order to better understand the mechanism of transport reduction. While barrier performance
improvements were minimal, the chemical modication improved the dispersion of the modied CNXLs
which led to improved performance. Of special note was an increase in the initial degradation tempera-
tures of both modied and unmodied systems, with the modied system showing an initial degradation
temperature >100 C higher than the cellulose alone. This may reect more extensive crosslinking in the
modied composite.
2008 Elsevier B.V. All rights reserved.

1. Introduction nanocrystals (CNXLs) (Fig. 1) have been successfully employed as

The eld of polymer nanocomposites is a rapidly expanding
area of research generating new materials with novel properties
[1]. Several new materials have been developed within the last
decade incorporating nano-sized ller material in polymer matri-
ces. Use of nanomaterials has proven to confer various advantages
like improved mechanical, thermal and barrier properties com-
pared to non-lled polymers. These effects are largely due to their
high interfacial area, their aspect ratio, their extent of dispersion
and percolation, which occurs when the ller particles are present
in quantities above the threshold where they start interacting [2].
Nanomaterials, e.g. carbon nanotubes, exfoliated clay, and cellulose
Corresponding author. Tel.: +1 541 737 4217.
E-mail address: john.simonsen@oregonstate.edu (J. Simonsen).
llers in polymer matrices and some systems are being commer-
cialized [3,4].
One application area for these materials is barrier membranes,
where the nano-sized llers impart enhanced mechanical and
barrier properties. Research in this area is evolving rapidly to
enhance the barrier properties and to overcome certain limitations,
e.g. durability, weight, robustness, exibility and packing volume.
Superior barrier membranes nd their use in food and biomedical
packaging where low permeability to oxygen, aromas, oils, or water
are needed [5,6].
While many barrier membranes are designed to prevent the
permeation of hydrophilic substances, such as water, other barrier
membranes are designed to reduce permeability to hydrophobic
substances, such as many toxic chemicals [7]. In this case, the
most effective ller material is likely to be an hydrophilic ller.
Nanocrystalline cellulose thus has the potential to provide

0376-7388/$ see front matter 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.memsci.2008.04.009
 S.A. Paralikar et al. / Journal of Membrane Science 320 (2008) 248258
Fig. 1. TEM image of CNXLs from cotton.
improved mechanical, physical and barrier properties to
hydrophobes in polymer matrices.
Recently, CNXLs have attracted much attention from researchers
for their remarkable reinforcing abilities [810]. Cellulose bers
are easily obtained from biomass and are an abundant, renewable,
biodegradable resource. Nanocrystalline cellulose is the crystalline
portion obtained after acid hydrolysis of cellulose. Typical sizes are
510 nm in diameter and 50200 nm in length [11,12]. Some advan-
tageous properties of CNXLs are their high aspect ratio of around
2050 (length/width), low density of 1.56 g/cm3 , high elastic mod-
ulus and strength reported to be 145 GPa [10,13,14] and estimated
at 7500 MPa, respectively [15]. One example is Orts et al. [9], who
obtained cellulose from various sources and added low concen-
trations to polymer blends to study the reinforcement effect of
cellulose. They found 10.3% cellulose bril content increased the
Youngs modulus by vefold in an extruded starch thermoplastic.
They also observed complex interactions between the components.
Two major disadvantages of using cellulose as a nanoller
are agglomeration due to high hydroxyl content, i.e. obtaining a
good dispersion of cellulose in the polymer matrix, especially in a
hydrophobic matrix. Surface compatibility by physical or chemical
treatment is usually required when composites are prepared with
cellulose and hydrophobic polymers [8].
Zimmermann et al. [10] dispersed cellulose bers in the
hydrophilic polymers poly(vinyl alcohol) and hydroxypropyl cellu-
lose to study their reinforcing effect. Mechanical tests conducted
showed a threefold increase in Youngs modulus and a vefold
increase in tensile strength compared to the base materials. The
elongation at rupture increased 500% with 5 wt% of the brils and
300% at 10 or 20 wt% of the cellulose compared to the pure polymer.
Borges et al. [16] studied the addition of coupling agent 1,4-butyl
diisocyanate, which improved the interaction between cellulose
and hydroxypropyl cellulose. With the coupling agent the yield
stress increased slightly, the rupture stress and Youngs modu-
lus almost doubled, while the % elongation decreased by 10% for
membranes with 10 wt% cellulose bers. Heat treating the same
membrane further increased the rupture stress and yield stress
by 50% while Youngs modulus and % elongation did not improve
signicantly.
Poly(vinyl alcohol) (PVOH) has been extensively studied as
a controlled drug release hydrogel, a membrane material for
249
chemical separations, barrier membrane for food packaging,
pharmaceutical component, manufacturing material for articial
human organs and as a biomaterial [7,1720]. It is resistant to per-
meation of solvents and oils and acts as a good barrier against
oxygen and aromas. Since PVOH is a hydrophilic polymer, dis-
persion of hydrophilic CNXLs into the matrix can be successfully
achieved by blending an aqueous PVOH solution with an aqueous
CNXL dispersion. However, PVOH membranes have poor stability
to moisture unless crosslinked [21,22] because the water molecules
swell the polymer and disrupt its barrier properties. Crosslinking
can be achieved by agents that bind with the hydroxyl groups or by
controlled heat treatment [18,23].
Additionally, crosslinking with monoaldehydes and dialdehydes
has successfully provided water resistance to PVOH, but the pro-
cess is not cost effective and leaves undesirable toxic crosslinking
agents [23]. Heat treatment, a low-cost method to provide insol-
ubility to membranes, provides crystallinity in PVOH matrices at
200 C and 10 min treatment time [24]. Temporary insolubility is
achieved when PVOH bers are heat treated with drawing [22].
Another method for crosslinking is by repeated freeze/thaw cycles
of PVOH gels, which forms a physically crosslinked PVOH with a
three dimensional network. These gels are mechanically strong, but
their long-term stability is an issue [23]. Poly(acrylic acid) (PAA) has
been successfully used as a chemical crosslinking agent for PVOH
[22]. The PVOH/PAA combination has been studied as a hydrogel for
biomedical applications [25] and as a separation membrane mate-
rial [26]. The added PAA creates a highly networked structure and
different mesh sizes can be obtained. By controlling the mesh size
the size of molecules that can pass through the membrane is con-
trolled [27]. A blend with PVOH/PAA ratio of 80/20 showed the best
separation results for a methanolwater pervaporation separation
membrane [28]. For an acetic acidwater pervaporation separating
membrane a ratio of 75/25 (v/v) was found to be optimum with
40 C as the optimum temperature [26]. Kumeta et al. [22] showed
that using partially neutralized PAA in the range of 510 mol% along
with heat treatment enhanced the crosslinking reaction between
PVOH and PAA.
In this work, crosslinking was accomplished by using heat treat-
ment and adding PAA as a crosslinking agent. The carboxyl group
in PAA and the hydroxyl groups of PVOH and CNXLs formed ester
linkages along with presumed hydrogen bonding between PVOH
and CNXLs. This allows the overall nanocomposite to have effec-
tive crosslinking and barrier resistance. The goal is for these barrier
membranes to have the ability to prevent permeation of harmful
chemicals and to be chemically inert, mechanically strong, tough,
exible and water resistant.
In this study, trichloroethylene (TCE) was chosen as a repre-
sentative toxic industrial material (TIM) and used to evaluate the
barrier properties of the PVOH/PAA/CNXL barrier membranes.
In addition, chemical modication of the CNXLs was performed
in order to further explore the mechanism of vapor transport.
2. Materials and methods
2.1. Theory
We evaluated the transport properties of the membranes using
TCE as the permeant in an evaporative ow device: Liquid TCE was
contacted to the feed side of the membrane. Its concentration on
the permeate side of the membrane was kept at essentially zero
by performing the experiment in a hood, where the airow swept
away any permeant from the membrane surface. Thus, the assump-
tions made for the chemical transport experiments were as follows:
(1) Mass transfer occurred in the positive (in this case down) z-
direction only. (2) The temperature and relative humidity of the
250 S.A. Paralikar et al. / Journal of Membrane Science 320 (2008) 248258
system remained constant throughout the experiment. (3) After
a sufcient time interval, a semi-steady state mass transfer was
attained and the ux became constant. (4) The permeant was a
liquid with concentration c0 at the upstream surface of the mem-
brane and a concentration of zero at the downstream surface of the
membrane.
There are generally three different parameters that characterize
the diffusion of a permeant through a polymer membrane in this
type of experiment [29]: (1) The time from initial exposure until
the permeant is rst detected, i.e. the breakthrough time. (2) The
ux (Q), which is obtained from the slope of the line of cumulative
ow through the membrane vs. time. (3) The time lag (tlag ), which
is the x-axis intercept of the same plot.
Under the conditions of this experiment, assuming the diffusion
coefcient (D) is constant and the solubility of the permeant in the
polymer membrane is uniform, Ficks law may be used to derive
the time dependence of Q as
 
Pc0 L2
Q = t (1)
L 6D
where P is the permeability of the membrane to the permeant, t:
time, and L: the thickness of the membrane. From Eq. (1) we may
easily derive D from the x-axis intercept (tlag ) of the plot as [30]:
L2
D=
6tlag
Typically the analysis continues as describing P as being com-
posed of two terms, D and the solubility, S. Thus [29]:
P = SD
However, we will see that our assumptions were violated in this
experiment (see Section 3.7).
2.2. Materials
Poly(vinyl alcohol), (99+% hydrolyzed, Mw = 89,00098,000)
was obtained from Sigma-Aldrich Inc. (St Louis MO, USA).
Poly(acrylic acid), (Mw = 2,000) was obtained from Aldrich Chem-
ical Company Inc. (Milwaukee WI, USA). Cellulose nanocrystals
(CNXLs) were prepared in our laboratory from cotton (Whatman
#1 lter paper), which was obtained from the Whatman Company
(Clifton, NJ). Briey, CNXLs were obtained by partial hydrolysis of
ground lter paper with 65% H2 SO4 (v/v) solution at 45 C with
medium stirring for 50 min. The ground paper to acid ratio was
1:10 g/mL. The mixture was centrifuged ve times with DI water
to remove the spent acid. The suspension was then subjected to
ultrasonic irradiation in a Branson Sonier (Danbury, CT) for 15 min
to disperse the CNXLs and break any agglomerates formed. Sonica-
tion appeared to improve the dispersion as there were fewer settled
solids after sonication and settling. The suspension was next ultra-
ltered (Ultrasette tangential ow UF device, pore size = 50 nm,
Pall Corp., Ann Arbor, MI) to remove salts until the conductivity
was <10 S/cm. The dispersed CNXLs were then concentrated in a
Rotavaporizer R110 (Buchi, Flawil, Switzerland) to obtain an aque-
ous dispersion of 1% CNXLs.
2.3. Materials for carboxylation of cellulose nanocrystals
Sodium hypochlorite (NaClO), 10%, was obtained from VWR
international (West Chester, PA, USA). Sodium bromide was
obtained from EM Industries Inc. (Gibbstown, NJ, USA) and 2,2,6,6-
tetramethyl-1-piperidinyloxy radical (TEMPO) was obtained from
SigmaAldrich Inc. (St Louis, MO, USA).
2.4. Surface modication of cellulose nanocrystals
A dispersion of 200 mL of 1% CNXLs was mixed with 0.2 g of
TEMPO and 2 g of NaBr. The oxidation reaction was initiated with
the addition of 10 mL of NaClO solution. Then 10 mL NaClO solu-
tion was subsequently added after 2 and 4 h. The overall reaction
time was 15 h and a pH of 10 was maintained throughout with
the addition of 1N NaOH as necessary. The oxidation reaction
was terminated by the addition of 30 mL of ethanol. The suspen-
sion was ultraltered thrice (Ultrasette tangential ow UF device,
pore size = 50 nm, Pall Corp., Ann Arbor, MI) to remove unreacted
reagents. A calculated amount of 1 M HCl was added to convert the
carboxylates to free acid. The solution was again ultraltered until
a permeate conductivity of <5 S/cm was obtained. A dispersion of
1% carboxylated CNXLs (C.CNXLs) was obtained by concentrating
in a rotavaporizer R110 (Buchi, Flawil, Switzerland). Carboxylated
content (mmols of acid group/g CNXL) of the dispersion was deter-
mined by potentiometric titration using 0.01N NaOH.
2.5. Preparation of the membranes
Stock solutions (5 wt%) of PVOH and PAA were prepared by dis-
solving the powder form in DI water and stirring in an oil bath at
85 C for 30 min.
Blended stock solutions were prepared by blending calculated
(2) volumes of the component solutions and CNXL dispersion. Before
casting, the solutions were sonicated for 25 min to re-disperse any
agglomerates formed. The thickness of the membrane was con-
trolled by pouring a known volume of the blend solution into a
at-bottomed plastic dish. The membranes were formed by air dry-
(3) ing for 48 h. Within the membranes, the CNXL content was varied
from 0 to 20 wt% in increments of 5 wt% and the PAA content was
varied from 0 to 20 wt% in increments of 10 wt% with the remain-
der of the content being PVOH. The obtained membranes were heat
treated in a convection oven at specied times and temperatures.
2.6. Water sorption and dissolution
The cast membranes were weighed (Weight 1) and submerged
in deionized water at room temperature. The membranes were
periodically removed, wiped dry with a tissue to remove any surface
water and weighed (Weight 2). Next, the membranes were dried in
a vacuum oven at 40 C and 20 in. Hg until a constant weight was
obtained (Weight 3). For each soaking cycle, the % water sorption
and % solubility were calculated as follows:
Weight 1 Weight 3
% solubility = 100%
Weight 1
Weight 2 Weight 3
% water sorption = 100%
Weight 3
2.7. Mechanical properties
A universal testing machine (Sintech 1/G, MTS, Cary, NC) was
used for mechanical testing in tensile mode. The heat treated mem-
branes of 2527 m thickness were cut into a dogbone shape
using a cutter. Each test was replicated thrice. The crosshead speed
was maintained at 1 mm/min with an initial span of 20 mm. Four
properties were determined from the obtained load vs. elongation
curves: (1) tensile strength, (2) tensile modulus, (3) elongation at
break and (4) work to failure (toughness). The tensile strength was
the yield load divided by the initial cross-sectional area of the spec-
imen. The modulus was obtained as the slope of the initial linear
portion of the curve. The elongation at break was obtained as % elon-
gation by dividing the extension by the initial span. The value for
 S.A. Paralikar et al. / Journal of Membrane Science 320 (2008) 248258 251

toughness was obtained as the area under the load vs. elongation
curve from zero load to failure.
2.8. Water vapor transmission rate (WVTR)
The prepared heat-treated membranes were glued onto glass
jars half lled with water. The initial weights of the assemblies were
noted. Three replicates of each sample were placed in a controlled
environmental chamber at 30 C and 30% relative humidity. The
weight change was noted as a function of time until a constant ux
was obtained for three consecutive days. The water vapor trans-
mission rate was calculated from the steady state weight loss as
attached to an Eclipse E400 Nikon microscope (Nikon Company,
Melville, NY) using crossed polarizers.
Atomic force microscope (AFM) imaging was performed using a
Dimension 3100 series Scanning Probe Microscope (Veeco Metrol-
ogy Inc., Santa Barbara, CA). The AFM was operated in tapping
mode using a rectangular silicon cantilever NSC 15/AIBS (Mikro-
Mash Wilsonville OR) with a tip radius <10 nm, typical resonance
frequency of 325 kHz and force constant of 46 N/m.
3. Results and discussion
3.1. Crosslinking

M The rst set of experiments was intended to determine a heat

J=
At treatment procedure to reduce the % solubility and % water sorp-
tion. At 185 C/60 min membrane samples of composition 10%
where J = mass ux (g/m2 day), M = cumulative weight loss of water
PAA/10% CNXL/80% PVOH exhibited a light brown coloration indi-
(g), A = area of the membrane available for mass transfer (m2 ),
cating degradation, so this temperature was discontinued for
t = time (day).
further tests. The samples heat treated at 125 C/60 min completely
disintegrated with water immersion after a few hours, while the
2.9. Fourier transform infrared spectroscopy (FTIR)
samples heat treated at 150 C/60 min and 170 C/60 min did not.
This suggested improvement of the crosslinking density with the
FTIR analysis was performed using a Nexus 470 FTIR spectrom-
increase in heat treatment temperature. The time for heat treat-
eter (Thermo Nicolet, Madison, WI) in an absorbance range of
ment was reduced to 45 min as the membranes assumed a light
4000500 cm1 , to compare the non-heat-treated and the heat-
yellow coloration with 170 C/60 min heat treatment. The increase
treated membranes. Membranes with 10 thickness were cast
in heat treatment temperature from 150 C/45 min to 170 C/45 min
for observation. Since pure PAA is too brittle to be formed into a
clearly decreased the total % solubility in each blend composition of
membrane, KBr pellets were used for this observation. The charts
the membrane (Fig. 2). Thus, the heat treatment protocol of 170 C
obtained were automatic baseline corrected and brought to a com-
for 45 min was applied to all the samples reported in this study.
mon scale for comparison.
The addition of PAA reduced the total % solubility greatly (Fig. 2),
suggesting a decrease in unreacted PVOH molecules, which would
2.10. Thermal analysis
otherwise leach out in the water. The total % solubility of the mem-
branes containing 20% PAA was just 1.3%, which was presumably
A modulated TGA 2950 thermogravimetric analyzer (TGA) (TA
due to the highly crosslinked matrix. However, this membrane was
instruments, Newark, DE) was used to test the thermal degradation
very brittle. The addition of CNXLs in the presence of PAA did not
of the composite membranes. The temperature was ramped from
cause a great difference in the % solubility. Evidently, if CNXLs were
room temperature to 600 C at a heating rate of 20 C/min under
forming any hydrogen bonds with PVOH they had no effect on the
nitrogen. Differential thermogravimetric analysis (DTGA) curves
solubility.
were obtained from the TGA data by numerically differentiating
The presence of crosslinking was investigated using FTIR. A 10%
the later with respect to temperature.
PAA/10% CNXL/80% PVOH membrane showed a broad carbonyl peak
at 1715 cm1 , which upon heat treatment shifted to 1723 cm1
2.11. Chemical vapor transmission rate (CVTR)
(Fig. 3A). Other features, such as the reduction in absorbance of
the OH group (33003400 cm1 ) and the appearance of a weak
This test was conducted according to ASTM standard F 1407-
peak at 1655 cm1 are common to either the PVOH or PAA spec-
99a [31] using 1,1,2 trichloroethylene (TCE) as the permeant. The
membrane was clamped over the permeation cup, immediately
weighed, inverted and placed in a chemical fume hood. Time and
weight of the assembly were noted at regular time intervals. A graph
of cumulative solute transferred through the membrane, Q, was
plotted against time. After a time lag, a linear curve was obtained
representing constant ux. This linear portion was modeled using
a least squares regression. The regression line was extended to the
x-axis intercept to obtain the time lag. The slope of regression line
gave the value of ux (g/m2 h) through the membrane.

2.12. Imaging

The morphology of the membranes was characterized with a

scanning electron microscope (SEM), AmRay 3300FE (AmRay Inc.,
Bedford, MA). The membranes were fractured under liquid nitro-
gen. The fractured surfaces were mounted on aluminum mounts,
PELCO # 16262. The samples were sputter coated with 60/40 wt%
Au/Pd alloy in an EDWARDS S 150B sputter coater.
Polarized optical microscopy (POM) images were obtained with
a Photometrics Cool Snap camera (Roper scientic, Tuscon, AZ) Fig. 2. Total % solubility after 72 h of heat treatment at 150 and 170 C for 45 min.
252 S.A. Paralikar et al. / Journal of Membrane Science 320 (2008) 248258
Fig. 3. FTIR spectra of (A) 10% PAA/10% CNXL/80% PVOH, (B) 100% PVOH membrane
and (C) 100% PAA membrane.
tra (Fig. 3B and C). The carbonyl peak might also originate from
ketone formation resulting from the degradation of PVOH during
heat treatment. However, an FTIR of pure PVOH heat treated for
45 min at 170 C showed that the absorbance of the OH peak at
33003400 cm1 , remained the same, indicating that the carbonyl
peak did not originate from ketone formation by degradation of
PVOH (Fig. 3B).
FTIR spectra of pure PAA were obtained to see if the position of
the carbonyl peak of PAA shifted with heat treatment. No such shift
was observed in the FTIR spectrum of pure PAA (Fig. 3C). Therefore,
it was concluded that the peak at 1723 cm1 (Fig. 3A) was due to
the formation of carboxylic esters formed by the heat treatment.
The total % water sorption decreased dramatically with increas-
ing PAA content and slightly with increasing CNXL content (Fig. 4).
With the heat treatment the polymer chains were interconnected
with ester linkages, the chain mobility was restricted and the
number of hydroxyl groups reduced, reducing water sorption.
Results with the addition of 20% CNXLs showed an increase in
water sorption compared to 10% CNXLs, which was possibly due
to agglomeration of the CNXLs and the consequent formation of a
separate CNXL phase and reduced CNXL content in the matrix.
Fig. 5. Polarized optical microscopy pictures of (A) 5 wt% CNXL/10 wt% PAA, (B) 10 wt% CNXL/10 wt% PAA and (C) 15 wt% CNXL/10 wt% PAA. PVOH makes up the remaining
content of the membranes.
Fig. 4. Total % water sorption, after 72 h of soaking time, for membranes with heat
treatment at 170 C/45 min.
3.2. Agglomeration
Since CNXLs are crystalline and birefringent, they rotate light
and thus appear bright between crossed polarizers (white in the
photos in Fig. 5). The membranes with 5% CNXLs (Fig. 5A) showed
good dispersion, i.e. no white particles except for some large
dirt particles. The 10% CNXLs (Fig. 5B) showing some agglomer-
ation while higher CNXL contents showed obvious agglomeration
(Fig. 5C).
3.3. Mechanical properties
Nanoparticles have been shown to improve both tensile modu-
lus and elasticity in lled nanocomposites [8,9]. With the addition
of 10% PAA and 10% CNXLs to PVOH the ultimate tensile strength
(UTS) showed a 150% improvement compared to pure PVOH (Fig. 6).
The change in UTS with either CNXL or PAA content was not linear. A
large synergistic increase was observed with the 10% CNXL/10% PAA
sample. Increasing PAA content did not show any consistent trend
in UTS. Elongation increased with CNXL content in the absence of
PAA (Fig. 6). Polymer nanocomposites have been reported to show
increases in elongation at low ller loadings [32,33] and the system
studied here behaves similarly.
The addition of PAA increases brittleness in the membranes, as
expected. In fact, the membranes containing 20% PAA showed a 70%
reduction in elongation compared to pure PVOH membrane. The
membranes with 20% PAA were quite brittle and were excluded
 S.A. Paralikar et al. / Journal of Membrane Science 320 (2008) 248258
Fig. 6. Mechanical properties as a function of CNXL and PAA content. Coefcient of variability for the samples ranged from 3 to 10%. The error bars represent 1 standard
deviation of the data.
from further consideration as commercially interesting composi-
tions.
The tensile modulus increased with the addition of CNXLs from
0 to 10%, but then dropped (Fig. 6). We believe that the drop in
modulus with CNXL content was due to agglomeration. Modulus
increased with PAA content from 0 to 10% then leveled off. This
suggested that either there was no additional crosslinking in the
20% PAA sample compared to the 10%, (although the reduction in
water sorption argues against this explanation), or that the 10%
already had sufcient crosslinking to maximize the modulus. Again,
we see a signicant synergy for the 10% CNXL/10% PAA sample.
The energy to break is the area under the stress strain curve,
which is one measure of the toughness of a material. A signi-
cant synergy was observed once again with the 10% CNXL/10% PAA
sample (Fig. 6). With an increase in the PAA content to 20% the brit-
tleness in the membranes increased, the % elongation decreased
and as a result toughness decreased. Again, 10% PAA with 10% CNXLs
showed excellent toughness, an improvement of 2.5 times in UTS
and twice the tensile modulus compared to pure PVOH.
3.4. Electron microscopy
SEM images of fractured membrane surfaces showed a smooth
surface for 100% PVOH (Fig. 7A). A rough texture with small cracks
was observed in the presence of CNXLs and PAA (Fig. 7B). The
fracture initiated crack formation, but well-dispersed CNXLs and
good bonding between the components evidently prevented fur-
ther crack proliferation. The rough texture could be attributed to
the addition of PAA which added to the stiffness and brittleness of
the membranes. Membranes with CNXL contents >10 wt% showed
more cracks which were deeper, wider and longer (Fig. 7C).
At higher magnication, similar results were observed (Fig. 8).
The 100% PVOH appeared to be smooth (Fig. 8A). The 10% CNXL/10%
PAA sample showed raised areas that could be CNXL agglomer-
253
ates (Fig. 8B), but their well-dispersed pattern argues against this.
Higher magnications gave no further elucidation of these fea-
tures on our equipment. Once again, the 20% CNXL/10% PAA sample
showed a much rougher texture (Fig. 8C) indicative of agglomera-
tion and poor fracture resistance.
Supporting the optical microscopy results, CNXLs were found
to be agglomerated in the 20% samples. Within these agglomer-
ates the CNXL surfaces are probably not fully wetted by the matrix.
This results in void spaces and weak interactions between particles
and between particles and the matrix, hence crack propagation is
enhanced in these regions, lowering the mechanical properties of
the membrane.
3.5. Thermal properties
Pure PAA and pure PVOH degraded thermally in three steps at
similar temperatures. Pure CNXL degraded in a complicated pattern
with an initial degradation temperature of 220 C and the maxi-
mum at 273 C (Fig. 9A). For pure PVOH, the maxima occurred at
290, 347 and 454 C. The maxima for pure PAA occurred at 300,
360 and 454 C. When PVOH and PAA were blended together only
two maxima were observed and they shifted to higher tempera-
tures at 374 and 463 C. The presence of CNXLs had little effect
on the degradation temperatures (Fig. 9B). In fact, CNXL in PVOH
without PAA showed little change in its degradation pattern from
pure PVOH. This suggests that it is the crosslinking with PAA that
alters the thermal degradation of these composites. Various blends
of PAA, CNXL and PVOH with 1020% PAA or CNXL were observed
to be similar to the DTGA graph of 10% PAA/10% CNXL and are not
shown.
Overall, the membrane with 10% PAA/10% CNXL/80% PVOH does
not loose its integrity as a composite material and this combi-
nation holds promise for its further development as a barrier
membrane.
254 S.A. Paralikar et al. / Journal of Membrane Science 320 (2008) 248258
Fig. 7. SEM fracture surface images of the cross-sections of membranes of (A) 100%
PVOH, (B) 10% PAA/10% CNXL/80% PVOH, (C) 10% PAA/20% CNXL/70% PVOH and (D)
10% PAA/10% C.CNXL/80% PVOH.
3.6. Transport properties: WVTR
The incorporation of either PAA or CNXLs into the membranes
reduced the WVTR (Fig. 10). This is expected as crosslinking with
PAA would reduce the number of hydroxyl groups in the compos-
ite and thus the hydrophilicity. In addition, CNXLs would provide a
Fig. 8. Higher magnication SEM fracture surface images of (A) 100% PVOH, (B) 10% PAA/10% CNXL/80% PVOH and (C) 10% PAA/20% CNXL/70% PVOH.
physical barrier to moisture ow, creating a tortuous path for the
permeating moisture. Addition of 10% CNXLs reduced the WVTR
more than the addition of 10% PAA, lowering the WVTR to <50%
that of 100% PVOH. The membranes with 10% CNXL showed a lower
WVTR than those with 20% CNXLs. This was contradictory to what
was expected. However, the agglomeration observed at the 20%
level may provide for channels, or domains, in the membrane that
allow for more rapid permeation.
Increasing PAA content from 10 to 20% gave only a slightly
lower WVTR value. We expect that increasing PAA concentration in
the membrane will yield offsetting effects: Increased crosslinking
should reduce WVTR, but if the PAA is not crosslinked, additional
free acid groups should increase WVTR since PAA itself swells upon
contact with water and should increase the free volume (and per-
haps liquid water phase in the membrane) and therefore increase
permeation.
The combination of CNXLs and PAA at the 10% level showed the
lowest WVTR, along with 10% CNXL/20% PAA, which was statisti-
cally equal. We observed that the 10/10 combination did not show
the effects of CNXL agglomeration apparent at higher CNXL con-
centrations, thus the 10% CNXL/10% PAA composition showed the
best performance.
3.7. Transport properties: CVTR
PVOH by itself possesses resistance against diffusion of
hydrophobic TIMs like TCE. The heat treatment reportedly
crosslinks the PVOH and induces crystallization, which improves
resistance to permeation. In addition, the barrier properties were
signicantly enhanced by the addition of CNXLs and crosslinking
agent PAA after heat treatment to promote crosslinking (Fig. 11).
In this initial exploration of these membranes, the long-term
steady state ux was calculated and extrapolated to the time axis
to obtain the time lag. The error in the time lag is complex since it
is the quotient of the (negative) y-intercept of the ux regression
divided by its slope. The time lag error was estimated as the sum
of the coefcients of variability of those two values. However, the
diffusion coefcient calculated from the time lag values gave unre-
alistically low values from Eq. (2), on the order of 1012 cm2 /s. This
value is several orders of magnitude smaller than typical values
of D in polymer membranes. We conclude that the assumptions
on which the equation was based were violated, most likely the
assumption of constant solubility of the permeant in the membrane
and constant D. The long lead time before a steady state is achieved
suggests that the permeant is dissolving in or being absorbed by the
membrane and this retards the time required for a steady state to be
achieved. Both S and D of the permeant in a swollen membrane are
likely to be different than that of the unswollen membrane, violat-
ing the assumptions used in the theoretical analysis of transport in
this system (Section 2.1). Thus, while these values are not appropri-
ate to calculate the diffusion coefcient or solubility, they do allow
 S.A. Paralikar et al. / Journal of Membrane Science 320 (2008) 248258 255

Fig. 9. DTGA graphs comparing the thermal degradation of (A) the pure components (PVOH, PAA and CNXLs), (B) the effect of PAA and CNXL and (C) C.CNXL to CNXL-lled
membranes.

Fig. 10. WVTR with varying weight% of either CNXLs or C.CNXLs and varying weight% of PAA. PVOH makes up the remaining content of the membranes. The error bars
represent 1 standard deviation of the data.
256 S.A. Paralikar et al. / Journal of Membrane Science 320 (2008) 248258
Fig. 11. Cumulative vapor transport through selected barrier membranes as a func-
tion of time.
Fig. 12. Time lag values for the PVOH membranes containing the listed components.
for a comparison between membranes, which is sufcient for this
preliminary study.
The time lag, i.e. the time axis intercept, and the steady state ux,
i.e. the slope of the linear regression at long times, show that the
addition of CNXLs reduces the permeation of TCE (Figs. 12 and 13).
The incorporation of 10% PAA was observed to yield a slight decrease
in ux, but an insignicant change in time lag (Figs. 12 and 13). This
points out the critical role that the impermeable nanoparticle ller
plays in the overall performance of the composite. However, adding
CNXLs without PAA does not produce a signicant change in time
Fig. 13. Flux values for the PVOH membranes containing the listed components. The
error bars represent 1 standard deviation of the data.
lag. The 5% CNXL/95% PVOH sample showed no change in time lag,
but some changes in ux, but this was inconsistent with the 10 and
15% values and, while statistically signicant, we concluded it was
not large enough to investigate further. The quality of the dispersion
is again suspected as a source of variability in this case.
Samples containing PAA showed the same synergistic effects in
CVTR at 10% CNXL/10% PAA as was observed in mechanical proper-
ties (Fig. 6), with time lag increasing and ux decreasing. We can
offer no rm conjecture on the mechanism of the synergism in this
case. More research is needed in this area. Again, the reduction in
time lag and increase in ux at the 15% CNXL level is most likely
due to aggregation of the nanoparticles.
3.8. CNXL carboxylation
In order to observe the effect of surface modication of the
CNXLs on mechanical and transport properties, the surface was
carboxylated following the method of Araki et al. [34]. This pro-
cedure results in selective oxidation of the C6 carbons of the
surface anhydroglucose repeat units of cellulose. FTIR conrmed
the presence of carboxyl groups at 1728 cm1 (data not shown). A
heat-treated composite membrane with 10% carboxylated cellulose
(10% C.CNXL/90% PVOH showed only one peak at 1730 cm1 which
conrmed ester bond formation, and thus crosslinking, between
the two components (data not shown).
Atomic force microscopy showed that the carboxylation proce-
dure did not appreciably alter the geometry or size of the CNXLs
(Fig. 14B) when compared to non-carboxylated cellulose (Fig. 14A).
A surface-carboxylated CNXL dispersion was observed to be more
stable against agglomeration than an untreated CNXL dispersion.
This is expected since the C.CNXLs have a higher surface charge den-
sity than the unmodied CNXLs, thus improving colloid stability
[34].
Titration with 0.01N NaOH indicated the presence of 1.4 mmol
of acid groups/g CNXL in the C.CNXLs. Titration of PAA with 0.5N
NaOH indicated the presence of 13.2 mmol of acid groups/g PAA.
PAA and C.CNXLs were added in a calculated amount to maintain
the total mmols of carboxylic acid equal to that of the corresponding
wt% PAA in the comparative sample.
While observation of the CNXLs dispersed in PVOH at a
15% loading using polarized optical microscopy showed obvious
agglomeration (Fig. 5), this agglomeration was not observed in an
equivalent C.CNXL dispersion (data not shown). This suggests that
the C.CNXLs dispersed more readily and were more stable in PVOH
than the unmodied CNXLs, probably due to the increased surface
charge resulting from carboxylation.
SEM images of the fractured surface of a heat-treated 10%
CNXL/10% PAA/80% PVOH membrane showed a rough fracture
(Fig. 7B). In contrast, the heat-treated membrane with 10%
C.CNXL/10% PAA/80% PVOH resulted in a smooth aky fractured
surface (Fig. 7D). This may be explained by both better interfacial
stress transfer between the C.CNXLs and the matrix vs. unmodied
CNXLs and better dispersion of the C.CNXLs in the matrix compared
to unmodied CNXLs.
No signicant differences in strength, modulus, % elongation or
toughness were observed between C.CNXLs and unmodied CNXLs
in equivalent matrices with the exception of % elongation at 10 and
20% PAA (Fig. 6). Here the carboxylated samples had a lower % elon-
gation than their unmodied counterparts. This was presumably
due to the increased crosslinking provided by the C.CNXLs. How-
ever, since the total carboxylation level was maintained constant,
no signicant differences in the other mechanical properties were
observed. Increased crosslinking, i.e. increasing PAA levels, was also
observed to lead to increased brittleness, i.e. lower elongation to
break, in both unmodied and carboxylated CNXLs.
 S.A. Paralikar et al. / Journal of Membrane Science 320 (2008) 248258
Fig. 14. AFM tapping mode images of (A) dried dispersion of CNXLs and (B) dried dispersion of C.CNXLs.
Carboxylating the CNXLs showed a consistent, but small,
decrease in WVTR for the C.CNXL samples (Fig. 10). This is in
accord with our contention that the C.CNXLs provide for additional
crosslinking between the nanoparticle ller and the matrix.
C.CNXLs in PVOH showed insignicant changes in CVTR time
lag compared to unmodied CNXLs (Fig. 12). However, the ux of
the 15% C.CNXL was markedly lower than that of the corresponding
15% CNXL sample (Fig. 13). We attribute this to agglomeration in the
case of the unmodied CNXLs, while the C.CNXLs were observed to
be easier to disperse and gave a more stable dispersion in all cases.
Thermogravimetric analysis showed that the incorporation of
C.CNXLs gave an increase in initial degradation temperature and
peak rate of degradation of approximately 40 C compared to
unmodied CNXLs in the same matrix (Fig. 9C). We conclude
from this that the C.CNXLs are more highly crosslinked to the
PVOH matrix than unmodied CNXLs. However, this large effect
observed in the TGA did not result in correspondingly large
effects in either mechanical properties or transport properties.
When we compare the number of carboxylate groups on the
CNXL surface in the 10% C.CNXL/10% PAA/80% PVOH membrane to
those on the PAA backbone in the same membrane, we see that
CO2 H(C.CNXL)/CO2 H(PAA) = 10.6%. Furthermore, the total number
of carboxyl groups in the membrane was kept constant. So we con-
clude that in terms of mechanical and transport properties, it is
257
the degree of crosslinking (assumed to be proportional to the con-
centration of CO2 H groups) that is important, and not whether the
crosslinking is to a nanocrystal or to the PAA. Since the unmodied
CNXLs are suspected to also form ester groups with the PAA, it is not
surprising that mechanical properties have not changed. It is per-
haps more informative that the transport properties do not change.
One might expect that any change in the interphase would alter
transport properties if surface ow along the percolating nanopar-
ticles governs diffusion through the membrane. However, these
data suggest that the diffusion chemistry of the interphase is not
that sensitive to its composition and that PAACNXL ester bonds
and C.CNXLPVOH ester bonds yield similar interphases and there-
fore similar nal composite properties. An alternative explanation
would be that the transport of the TCE through the membrane
is not related to particle surface ow and occurs primarily in the
matrix, with the CNXLs providing a more tortuous path up to con-
centrations where they begin to agglomerate. Additional research
is required to determine which hypothesis predominates in this
system.
4. Conclusions
Heat treatment improved the crosslinking density within the
membranes made of varying contents of CNXL, PAA and PVOH. Sol-
258 S.A. Paralikar et al. / Journal of Membrane Science 320 (2008) 248258
ubility results indicated that higher temperature resulted in higher
crosslink densities within the matrix. The FTIR spectra indicated
that the heat treatment resulted in the formation of ester linkages
between PAA and PVOH. Imaging showed that CNXLs were well
dispersed in blend membranes of PVOH and PAA at 10% content by
weight and lower, but agglomerated at 20%. The presence of CNXLs
and crosslinking almost doubles the strength, stiffness and tough-
ness, while the elongation is reduced by 20%. The DTGA result
supports the synergistic effect of 10% CNXLs and 10% PAA in a PVOH
matrix.
The CVTR experiments showed that increasing PAA or CNXL con-
tent increased the lag time and decreased the ux compared to a
pure PVOH membrane. A synergistic composition was observed at
10% CNXL/10% PAA. Substituting C.CNXLs for CNXLs showed little
effect on transport properties. The FTIR, mechanical, transport and
thermal properties suggest that crosslinking is occurring between
the PAA and the PVOH and probably between the C.CNXLs and
PVOH. Crosslinking between CNXLs and PAA is expected, but the
data to date are not conclusive. Moisture transport was reduced by
both PAA and CNXLs, both modied and unmodied. The C.CNXLs
were observed to disperse more easily in PVOH aqueous solutions
and remain more stable in the resulting membranes than unmod-
ied CNXLs. This effect also is employed to explain differences in
mechanical and transport properties for the higher ller contents.
The DTGA results support the synergistic effect of 10% CNXLs
and 10% PAA in a PVOH matrix. The DTGA of C.CNXLs showed a
large shift of +40 C which suggested that C.CNXLs have a greater
interaction within the matrix than unmodied CNXLs.
The synergistic 10% CNXL/10% PAA combination performed well
and shows potential for the development of improved barrier mem-
branes against toxic chemicals.
Acknowledgement
We gratefully acknowledge the support from the National
Research Initiative of the USDA Cooperative State Research, Edu-
cation and Extension Service, grant number 2003-35103-13711.
References
[1] G. Vaia, R. Krishnamoorti, Polymer nanocomposites: introduction, in: R.
Krishnamoorti, R.A. Vaia (Eds.), Polymer Nanocomposites: Synthesis, Char-
acterization and Modeling. ACS Symposium Series 804, American Chemical
Society, Washington, DC, 2001.
[2] Y. Brechet, J. Cavaille, E. Chabert, L. Chazeau, R. Dendievel, L. Flandin, C. Gau-
thier, Polymer based nanocomposites: effect of llerller and llermatrix [29] E.L. Cussler, Diffusion, Mass Transfer in Fluid Systems, Cambridge University
interactions, Adv. Eng. Mater. 3 (2001), 571-207.
[3] G. Cao, Nanostructures & Nanomaterials: Synthesis, Properties & Applications,
Imperial College Press, London, UK, 2004.
[4] W. Gacitua, A. Ballerini, J. Zhang, Polymer nanocomposites: synthetic and nat-
ural llersa review, Maderas, Ciencia y Technologia 7 (2005) 159178.
[5] F. Fakhouri, P. Tanada-Palmu, C. Grosso, Characterization of composite biolms
of wheat gluten and cellulose acetate phthalate, Braz. J. Chem. Eng. 21 (2004)
261264.
[6] B. Flaconneche, J. Martin, M. Klopffer, Transport properties of gases in polymers:
experimental methods, Oil Gas Sci. Technol. 56 (2001) 245259.
[7] J. Lange, Y. Wyser, Recent innovations in barrier technologies for plastic
packaginga review, Packag. Technol. Sci. 16 (2003) 149158.
[8] A. Bledzki, J. Gassan, Composites reinforced with cellulose-based bers, Prog.
Polym. Sci. 24 (1999) 221274.
[9] J. Orts, J. Shey, S. Imann, G. Glenn, M. Guttman, J. Revol, Application of cel-
lulose microbrils in polymer nanocomposites, J. Polym. Environ. 13 (2005)
301306.
[10] T. Zimmermann, E. Pohler, T. Geiger, Cellulose brils for polymer reinforcement,
Adv. Eng. Mater. 6 (2004) 754761.
[11] S. Noorani, J. Simonsen, S. Atre, Polysulfone nanocomposites, in: K. Oksman, M.
Sain (Eds.), Cellulose Nanocomposites: Processing, Characterization and Prop-
erties, ACS Symposium Series no. 938, American Chemical Society, Washington,
DC, 2006.
[12] M. Roman, W.T. Winter, Cellulose nanocrystals for thermoplastic reinforce-
ment: effect of ller surface chemistry on composite properties, abstracts of
papers, in: Proceedings of the 229th ACS National Meeting, San Diego, CA,
United States, March 1317, 2005, CELL-214.
[13] F. Vilaseca, T. Peijs, Cellulose-based nanocomposites, in: Proceedings of the 8th
International Conference on Woodber-Plastic Composites (and Other Natural
Fibers), Madison, WI, United States, May 2325, 2005, pp. 283295.
[14] A. Sturcova, G. Davies, S. Eichhorn, Elastic modulus and stress transfer proper-
ties of tunicate cellulose whiskers, Biomacromolecule 6 (2005) 10551061.
[15] R. Marks, Cell Wall Mechanics of Tracheids, Yale University Press, 1967.
[16] J. Borges, M. Godinho, M. Belgacem, A. Martins, New bio-composites based
on short bre reinforced hydroxypropylcellulose lms, Comp. Interf. 8 (2001)
233241.
[17] S. Bourke, M. Al-Khalili, T. Briggs, B. Michniak, J. Kohn, L. Poole-Warren, A
photo-crosslinked poly(vinyl alcohol) hydrogel growth factor release vehicle
for wound healing applications, AAPS Pharma Sci. 5 (2003) 33.
[18] C. Finch, Polyvinyl Alcohol Developments, John Wiley & Sons, New York, 1992.
[19] R. Schmedlen, K. Masters, J. West, Photocrosslinkable polyvinyl alcohol hydro-
gels that can be modied with cell adhesion peptides for the use in tissue
engineering, Biomaterials 23 (2002) 43254332.
[20] W. Wan, G. Campbell, Z. Zhang, A. Hui, D. Boughner, Optimizing the tensile
properties of poly(vinyl alcohol) hydrogel for the construction of a bioprosthetic
heart valve stent, J. Biomed. Mater. Res. 63 (2002) 854861.
[21] K.F. Arndt, A. Richter, S. Ludwig, J. Zimmermann, J. Kressker, D. Kuckling,
H.J. Alder, Poly(vinyl alcohol)/poly(acrylic acid) hydrogels: FT-IR spectroscopic
characterization of crosslinking reaction and work at transition point, Acta
Polym. 50 (1999) 383390.
[22] K. Kumeta, I. Nagashima, S. Matsui, K. Mizoguchi, Crosslinking reaction of
poly(vinyl alcohol) with poly(acrylic acid) by heat treatment: effect of neu-
tralization of PAA, J. Appl. Polym. Sci. 90 (2003) 24202427.
[23] C. Hassan, N. Peppas, Structure and applications of poly(vinyl alcohol) hydrogels
produced by conventional crosslinking or by freezing/thawing methods, Adv.
Polym. Sci. 153 (2000) 3765.
[24] N. McCrum, B. Read, G. Williams, Anelastic and Dielectric Effects in Polymeric
Solids, Wiley & Sons, New York, NY, 1967.
[25] O. Sanli, G. Asman, Release of diclofenac through gluteraldehyde crosslinked
poly(vinyl alcohol)/poly(acrylic acid) alloy membranes, J. Appl. Polym. Sci. 91
(2003) 7277.
[26] G. Asman, O. Sanli, Characteristics of permeation and separation for acetic
acidwater mixtures through poly(vinyl alcohol) membranes modied with
poly(acrylic acid), Sep. Sci. Technol. 38 (2003) 19631980.
[27] N. Peppas, S. Wright, Solute diffusion in poly(vinyl alcohol)/poly(acrylic acid)
interpenetrating networks, Macromolecules 29 (1996) 87988804.
[28] J. Rhim, H. Kim, K. Lee, Pervaporation separation of binary organic-
aqueous liquid mixtures using crosslinked poly(vinyl alcohol) membranes. IV.
Methanolwater mixture, J. Appl. Polym. Sci. 61 (1996) 17671771.
Press, 1997.
[30] J. Crank, G.S. Park, Diffusion in Polymers, Academic Press Inc., 1968.
[31] ASTM, F 1407-99a Standard Method of Resistance of Chemical Protective Cloth-
ing Materials to Liquid Permeation, American Society of Testing and Materials,
West Conshohocken, PA, 1999.
[32] C.L. Wu, M.Q. Zhang, M.Z. Rong, K. Friedrich, Tensile performance improvement
of low nanoparticles lled-polypropylene composites, Compos. Sci. Technol. 62
(2002) 13271340.
[33] Y. Choi, Cellulose Nanocrystal-lled Carboxymethyl Cellulose Composites, M.S.
Thesis, Oregon State Univ. Dept. Wood Sci. & Eng., 2005.
[34] J. Araki, W. Masahisa, S. Kuga, Steric stabilization of a cellulose microcrystal
suspension by poly(ethylene glycol) grafting, Langmuir 17 (2001) 2127.