Composites: Part A 88 (2016) 8697

Contents lists available at ScienceDirect

Composites: Part A
journal homepage: www.elsevier.com/locate/compositesa

Fabrication of carbon-based nanocomposite films by spin-coating

process: An experimental and modeling study of the film thickness
Marialaura Clausi a, M. Gabriella Santonicola b,c, Susanna Laurenzi a,
a
Department of Astronautic Electrical and Energy Engineering, Sapienza University of Rome, Via Salaria 851-881, 00138 Rome, Italy
b
Department of Chemical Materials and Environmental Engineering, Sapienza University of Rome, Via del Castro Laurenziano 7, 00161 Rome, Italy
c
Materials Science and Technology of Polymers, MESA+ Institute for Nanotechnology, University of Twente, P.O. Box 217, 7500 AE Enschede, The Netherlands

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

Article history: Spin-coating is used for the fabrication of nanocomposite thin films, consisting of carbon nanoparticles
Received 6 March 2016 embedded in epoxy matrix, on Mylar substrate. The final thickness of the heat-cured film was measured
Received in revised form 22 May 2016 as a function of the spinning speed and nanoparticle concentration. Multi-walled carbon nanotubes with
Accepted 24 May 2016
carboxyl functionalization (MWCNT-COOH) or exfoliated graphite nanoplatelets (xGnP) were used as fil-
Available online 25 May 2016
lers. Experimental results were in good agreement with the predictions from a model that considered the
rheology and flow behavior of the reinforced resin fluids on a rotating disk. The model was differentiated
Keywords:
for Newtonian and non-Newtonian regime of the spinning polymer fluid. In case of non-Newtonian
A. Multifunctional composites
A. Nanocomposites
behavior of the epoxy resin at high particle concentrations, a semi-empirical approach was used to deter-
B. Viscosity mine the model constants from rheology measurements. Results from this analysis also indicate how rhe-
C. Process modeling ological and wetting properties of the nano-reinforced polymer fluids depend on the aspect ratio of the
graphene nanoplatelets.
2016 Elsevier Ltd. All rights reserved.

1. Introduction applications, a few analytical studies of the process for nanostruc-

Carbon-based nanocomposite films constitute a novel class of
industrially applied coatings, which have received great attention
due to their electrical, mechanical, thermal and optical properties
[14]. Depending on the film composition and preparation meth-
ods, carbon nanocomposite films are used in a wide range of appli-
cations in different technological areas. In particular, such films
find use in microelectronic devices [5,6], as substrates with novel
biosensing or gas-sensing properties [7,8], in protective and optical
coatings [9,10], and as radar absorbing or EMI shielding materials
[11,12].
One of the crucial aspects in the fabrication of homogeneous
nanocomposite films is to guarantee a uniform distribution of the
nanoparticles inside the polymer matrix [13,14]. Besides filler dis-
persion, the deposition process of the films has a major effect on its
overall properties. Spin-coating has become the predominant tech-
nique for the large-scale production of uniform and reproducible
films with thickness of the order of micrometers and nanometers
[15], in particular for microelectronic devices, solar cells and
optical coatings [1619]. In spite of their widespread industrial
Corresponding author.
E-mail address: susanna.laurenzi@uniroma1.it (S. Laurenzi).
tured polymer liquids can be found in the literature [2022]. Spin-
coating of polymer fluids is a complex process to model due to
many mechanisms that are involved at the physical and chemical
level. The final thickness and uniformity of the spin-coated films
are known to depend on several process parameters, including
rotational speed, solution concentration, fluid viscosity and den-
sity, and, when solvents are present, rate of solvent evaporation
[2325]. In the case of nano-reinforced polymer liquids, additional
complexity in the behavior under spinning arises from the molec-
ular interaction between the nanofillers and the polymer fluid. An
accurate analysis of spin-coating for such fluids would allow for a
better control of the process in its various applications.
Following the pioneering work of Emslie et al. [26], the analysis
of spin-coating process can be performed considering the spread-
ing of a thin axisymmetric film of viscous fluid on a rotating disk.
In their work, Emslie et al. made the major assumption that the
fluid exhibited Newtonian behavior, characterized by a linear rela-
tionship between shear stress and shear rate, and neglected the
effect of several forces, such as Coriolis and gravitational forces,
as well as spatial and temporal variations in concentration. The
predictions of the flow pattern and of the final film thickness were
therefore limited to ideal cases not involving suspensions with set-
tling particles or highly viscous fluids. Later, Jenekhe and Schuldt

http://dx.doi.org/10.1016/j.compositesa.2016.05.026
1359-835X/ 2016 Elsevier Ltd. All rights reserved.
 M. Clausi et al. / Composites: Part A 88 (2016) 8697 87

analyzed the surface coating of non-Newtonian fluids during the

process of spin-coating using numerical methods [27]. Numerical
solutions for the film thickness profiles were obtained for
non-Newtonian behavior described by power-law and Carreau
constitutive equations. Results of this work unveiled that the final
thickness and uniformity of films deposited by spin coating are
dictated mainly by the fluid rheological properties, and not by
the initial profile of the film. The effects of film inertia, disk
acceleration, and interfacial shear due to the overlaying gas phase
during spin coating were also investigated by numerical methods
[28].
In this work, we study the fabrication of nanocomposite films
by spin-coating of an epoxy resin reinforced with carbon nanopar-
ticles on polymer substrate, followed by a heat cure cycle. In par-
ticular, this work is focused on Prime 20LV epoxy resin
containing carboxy-functionalized multi-walled carbon nanotubes
(MWCNT-COOH) or exfoliated graphite nanoplatelets (xGnP) at
different loadings in the range 0.55 wt%. The spin-coating process
is modeled considering the flow of a viscous fluid on a planar inter-
face, and differentiating between Newtonian and non-Newtonian
behavior of the nano-reinforced epoxy resin as determined in
steady shear rheological experiments. The main objective is to pre-
dict the thickness of the spin-coated nanocomposite films consid-
ering also the non-linear behavior of nano-reinforced resin fluids at
high particle concentration. A further objective is to investigate the
effect of the geometry of the nanofillers on the rheological proper-
ties and wetting properties of the nanostructured epoxy resins on
Mylar substrate.

2. Theoretical thickness of spin-coated films

The spin-coating process can be modeled considering the

behavior of a liquid solution on a rotating disk substrate, and
assuming that the liquid flows on the substrate only in the radial Fig. 1. Schematic of spin-coating process for the fabrication of nano-reinforced thin
composite films. (For interpretation of the references to colour in this figure legend,
direction and is a function of time [29,30]. For viscous fluids
the reader is referred to the web version of this article.)
(Re
1) inertial effects other than centrifugal have a negligible
influence on the rate of thinning of the fluid layer, and centrifugal
forces can be considered constant during spinning [28]. In addition, Z h
qrx2 h3
the effect of interfacial shear on the thinning rate can be consid- Q v r dz 3
0 3g
ered unimportant for moderate spinning times. In our analysis,
we further neglect any effect due to solvent evaporation on the Considering the variation of fluid thickness in time, the equa-
thinning rate of the spinning fluid, given the absence of solvents tion of continuity in radial direction becomes [30]:
in the nano-reinforced epoxy resins. The film thicknesses that are
@h 1 @
measured and compared with the model predictions are those of rQ 0 4
the cured composites films, assuming that no volume shrinkage @t r @r
occurs during the curing and subsequent cooling steps. This Replacing Eqs. (3) into (4), we obtain:
assumption can be considered valid for epoxy matrices.   3
Using cylindrical coordinates (r, h, z) with origin at the center of @h 1 @r 2 h
C 5
rotation and r and h axes rotating with the plane with angular @t r @r
speed x (Fig. 1), the Navier-Stokes equation for a Newtonian fluid
where C = qx
2
(viscosity independent of shear rate) is: 3g
. If we consider the time-dependent solution and that


the film is uniform along the radial direction @h 0 , then integrat-
@2v r @r
g qr x2 1 ing both sides between suitable limits, h h0 for t 0 and h hf
@z2
for t t spin (the selected duration of the spin-coating process), we
where q is fluid density, g is fluid absolute viscosity, v r is the fluid obtain for a Newtonian fluid:
v r where s is the shear
velocity in the direction of r and srz g @@z rz 2 31=2
stress on the z face of a fluid element arising from fluid momentum 1
in the y direction, z is the film thickness direction, and x is the hf h0 4 5 6
4qx2 h20 t spin
angular speed of the spin coater disk. Integrating and using suitable 3g
1
boundary condition of no-slip (v r (r; h; z) = 0 at z = 0) and free sur-

vr  For fluids exhibiting non-Newtonian behavior, the viscosity is a
face ddz 0 at z h we obtain [26]: non-linear function of the shear rate (c_ ), and a more complicated
  equation is required to describe it. A widespread model to describe
qrx2 h2 z z2
v r r; z  2 the viscosity of fluids with shear-thinning behavior is the general
g h 2h2
four-parameter Carreau model [31,32], in which viscosity is related
The radial flow Q is given by: to c_ by:
88 M. Clausi et al. / Composites: Part A 88 (2016) 8697
n1
gc_ g1 g0  g1 1 k2 c_ 2 2
Here, k is the relaxation time constant, n a measure of the devi-
ation from Newtonian behavior (power-law index), and g0 and g1
the upper and lower limiting values of the fluid viscosity, respec-
tively. This equation was successfully used to fit the viscosity data
of nanostructured polymer dispersions containing carbon-based
nanoparticles [33,34]. For non-Newtonian fluids described by the
Carreau model the film thickness becomes uniform after a suffi-
cient spinning time [27].
For a non-Newtonian fluid described by the Carreau model, Eq.
(6) becomes:
2 31=2
6 7
6 1 7 determined by recording videos of the spreading drops and analyz-
hf h0 6
6
7
7 8
4qx2 h20 t spin
4h i 15
n1
3 g1 g0 g1 1k2 c_ 2 2
3. Experimental
3.1. Materials
Multi-walled carbon nanotubes functionalized with carboxyl
groups (MWCNTCOOH) with purity larger than 95% were pur-
chased from MK Nano Division (MK Impex Corp., Canada). Accord-
ing to datasheet, the average outer diameter was 50 nm, and the
nanotube length was in the range of 0.52 lm. Exfoliated graphite
nanoplatelets (xGnP) with different aspect ratio and specific sur-
face area (SSA) were purchased from XG Sciences Inc. (USA) and
used as received. Nanoplatelets of grade C750 (thickness 2 nm,
average diameter <2 lm, SSA 750 m2 g1), grade M5 (thickness
68 nm, average diameter 5 lm, SSA 120150 m2 g1) and grade
H5 (thickness 15 nm, average diameter 5 lm, SSA 50
80 m2 g1) were used in this study. Prime 20LV epoxy resin with
slow hardener (Gurit, UK) was chosen for the polymer matrix.
Mylar sheets (60 mm  60 mm) with thickness 15 lm were used
as film substrate.
3.2. Fabrication of nanocomposite films
Nanocomposites films of different thicknesses were prepared
by spin coating the mixture of carbon nanoparticles and Prime
20LV epoxy using different angular velocity (10004000 rpm), fol-
lowed by thermal curing in oven at 50 C for 16 h. Nano-reinforced
epoxy mixtures with MWCNT or xGnP at 0.5 wt%, 1 wt%, 3 wt%,
and 5 wt% were used. Each polymer dispersion was prepared by
adding a selected grade of nanoparticles at a given weight percent-
age to the epoxy matrix, followed by mechanical mixing for 20 min
and then sonication in ultrasonic bath for 60 min. Next, the curing
agent was gently added to the blend in the ratio of epoxy to curing
agent of 100:26 by weight. A fixed volume of the polymer mixture
was then deposited at the center of a pre-cleaned Mylar substrate
on the spin coater disk. The spin coater instrument was a CaLCTec
FR10 KPA where the Mylar substrate was held by vacuum on a
microporous chuck to prevent wrinkle formation. Prior to spin
coating, Mylar substrates were cleaned by sonication in Milli-Q
water for few minutes, and then dried in air. Organic solvents com-
monly used for surface cleaning, such as ethanol and acetone, were
avoided with Mylar. All nanocomposite films, including the control
sample made of neat epoxy, were prepared in triplicates. At the
end of the spin-coating process, the nanocomposite films on Mylar
were immediately cured. The thickness of the nanocomposite films
after cure was measured using a micrometer screw gauge (accu-
racy 1 lm). Thickness measurements were repeated in quintupli-
7 cate across a sample area of radius 1 cm around the center of the
disk, and values were averaged.
3.3. Contact angle characterization
Contact angles were measured by the sessile drop method using
a DataPhysics OCA15Pro analyzer (DataPhysics Instruments, Ger-
many). Neat Prime 20LV epoxy or nano-reinforced epoxy samples
were used as the testing fluid, which was dispensed through dis-
posable 1-mL syringes. The spreading kinetics of each polymer
drop on Mylar substrate was followed as a function of time until
equilibrium was reached. Values of contact angles were obtained
by Laplacian curve fitting using the module software SCA 20 pro-
vided with the instrument. Contact angles at equilibrium were
ing the frames after equilibrium was reached (typically after 2
3 min). A minimum of five drops (volume 5 ll) in different spots
of the substrate were analyzed for each resin sample. Measure-
ments were carried out immediately after addition of the curing
agent to the nano-reinforced epoxy resin followed by mechanical
mixing for few minutes. Care was taken to prevent the presence
of air bubbles inside the dispensing unit by sonicating the mixtures
for few minutes before loading the syringe. The OCA15Pro instru-
ment was also used to determine the thickness of the nano-
reinforced epoxy layer on Mylar before starting the spin coating
procedure, which is the value of the initial thickness h0 in Eqs.
(6) and (8).
3.4. Rheological characterization
Rheological analyses of the nanocomposite epoxy fluids were
carried out using a rotational stress-controlled MCR 302 rheometer
(Anton Paar, Austria) equipped with a Peltier heating system and
25-mm plate-plate configuration (gap size in the range 0.71 mm
depending on filler concentration). Steady shear measurements
were conducted at 23 C for shear rates in the range of 0.01
100 s1. To assess reproducibility, measurements were repeated
at least in duplicates. Sufficient time was allowed between consec-
utive measurements on the same sample, in order to guarantee
relaxation of stresses induced by the shear rate. Transient rheolog-
ical measurements were also carried out at 23 C and at fixed shear
rates (0.1, 1 and 10 s1) using the parallel plate geometry with a
gap of 0.9 mm and a sampling time of 1 s. Prior to rheological char-
acterization, the nano-reinforced polymer mixtures were sonicated
in ultrasonic bath for 5 min. For the nano-reinforced resins charac-
terized by Newtonian behavior, dynamic viscosities at the shear
rate of 10 s1 were considered in the film thickness analysis by
Eq. (6). For resins exhibiting non-Newtonian behavior, the viscosity
data as a function of the shear rate were fitted using the general
Carreau viscosity model in Eq. (7). The fitting was performed in
the shear rate range 0.01100 s1 using the non-linear least-
squares method (Levenberg-Marquardt algorithm) implemented
in Origin software. To evaluate the validity of the model with
respect to the experimental data, we used the reduced v2 index.
4. Results and discussion
4.1. Wetting behavior of nano-reinforced epoxy resins on Mylar
substrate
A critical aspect of the spin-coating process is the degree of wet-
tability of the surface substrate by the polymer fluid. The strength
of the interfacial adhesion between polymer and substrate indeed
influences the spreading and thinning of the polymer layer during
spin-coating, thus affecting the morphology and uniformity of the
 M. Clausi et al. / Composites: Part A 88 (2016) 8697
deposited polymer films [35]. When polymer fluids containing
nanoparticles are used for spin coating, the situation is further
complicated by the possible formation of aggregates, which alter
the normal flow behavior of the polymer on the spinning substrate
and generates defects in the deposited films. In this work, we ana-
lyzed the wetting behavior of Mylar substrate by the nano-
reinforced epoxy resins by measuring the contact angles between
a sessile drop of the polymer fluid and the Mylar surface as a func-
tion of time. The contact angle values are a useful indicator of the
interfacial adhesion between the resin and substrate, as they are
correlated with the solid/liquid interfacial tension via Youngs
equation [36], and thus provide a measure of the wetting compat-
ibility of the substrate with respect to the resin. Fig. 2 shows the
spreading kinetics of neat Prime 20LV and of nano-reinforced
epoxy samples containing the three different grades of xGnP at fil-
ler loading of 3 wt%. In all cases, the contact angle of each resin
drop on Mylar does not vary significantly after approximately
1 min, with the neat Prime 20LV resin being characterized by the
lowest value of equilibrium contact angle (16.5 2.0 from the
average of five measurements). Such low value is an indication of
excellent compatibility in terms of surface wetting between the
Mylar substrate and the Prime 20LV epoxy system, which is advan-
tageous for the formation of defect-free spin-coated nanocompos-
ite films.
The effect of graphene nanoplatelets with different aspect ratio
on the spreading kinetics of Prime 20LV epoxy is also clearly high-
lighted in Fig. 2. Upon increasing the dimensions (platelet thickness
and size) of the graphene filler, the resin spreading profiles are
shifted consistently upwards to larger contact angle values, indicat-
ing a general decrease in compatibility of the Mylar substrate with
respect to the nano-reinforced epoxy. Such behavior cannot be
explained simply by taking into account a viscosity increase of
the resin upon filler incorporation. For example, the increase of
equilibrium contact angle for Prime 20LV epoxy upon adding 3 wt
% of xGnP-H5 is of the order of 120%, whereas the corresponding
viscosity increase is only about 80% (data in next section).
In Fig. 3 the effect of nanoparticle dimension on the equilibrium
contact angle of the nano-reinforced epoxy systems is summa-
rized. For epoxy systems with the same xGnP concentration, the
value of the equilibrium contact angle on Mylar increases for nano-
Fig. 2. Spreading kinetics of neat epoxy resin and xGnP-reinforced epoxy on Mylar substrate for graphene nanoplatelets grades C750, M5, and H5 (loading 3 wt%). Each
contact angle value is the average of at least three different measurements on Mylar surface. (For interpretation of the references to colour in this figure legend, the reader is
referred to the web version of this article.)
89
platelets with larger thickness (grades C750, M5, and H5 with
thickness of 2 nm, 68 nm, and 15 nm, respectively). Fig. 4 shows
optical images and calculated contact angle values for sessile drops
of Prime 20LV resin containing 5 wt% of the three grades of xGnP,
with the epoxy/xGnP-H5 system showing the largest value of heq
(34.8 1.9).
When looking at the effects of nanofiller loading, the addition of
carbon nanoparticles (MWCNT-COOH or xGnP) to the resin always
causes an increase of its equilibrium contact angle on Mylar,
despite the hydrophobic nature of the fillers that matches that of
the substrate (contact angle of water on Mylar = 84.0 2.8). In
Fig. 3 the equilibrium contact angles for mixtures of Prime 20LV
resin and MWCNT-COOH or xGnP (grades C750, M5 and H5) with
different loadings are reported. For each type of nano-reinforced
resin system, the increase of the equilibrium contact angle as a
function of the nanoparticle concentration up to 5 wt% was not lin-
ear, but it showed a slight decrease at concentrations below 1 wt%
and below 3 wt% for the xGnP and the MWCNT-COOH nanofillers,
respectively. In particular, for the epoxy/MWCNT-COOH mixtures
the value of contact angle at 3 wt% of carbon nanotubes is 16%
lower than the value at 1 wt%. This behavior might be explained
by the presence of a percolation threshold, where the properties
of the nano-reinforced polymer fluids change abruptly. The non-
linearity of the properties of the nano-reinforced resins upon
increasing the filler content has been confirmed in rheological
measurements, the results of which are presented in the next
section.
4.2. Steady shear rheological analysis of nano-reinforced epoxy resins
Polymer fluids that are used in spin-coating processes can exhi-
bit non-Newtonian behavior, meaning that their viscosity depends
on the applied shear rate. In addition, as a consequence of
particle-particle or multi-particle interactions, suspensions of
carbon nanotubes or graphene nanoplatelets are known to have
shear-thinning behavior in polymer fluids [37,38]. Here, we
analyzed the steady shear rheological properties of the nano-
reinforced Prime 20LV epoxy, with MWCNT-COOH or xGnP as fil-
lers, and used the viscosity results for modeling the spin-coating
fabrication process of the nanocomposite films. Fig. 5 shows the
90 M. Clausi et al. / Composites: Part A 88 (2016) 8697

Fig. 3. Contact angles at equilibrium (heq) of carbon-reinforced Prime 20LV on Mylar substrate with different nanofillers. Mean values and standard deviations calculated
from measurements in five different spots of the Mylar surface. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of
this article.)

Fig. 4. Optical images and calculated contact angle values of neat Prime 20LV epoxy resin and epoxy/xGnP systems with different nanoplatelet thickness on Mylar substrate.
(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

shear stress and viscosity profiles of epoxy/MWCNT-COOH disper-
sions with different filler content for measurements conducted at
the temperature of 23 C. As a reference, the neat Prime 20LV
epoxy fluid was investigated, and it showed the typical Newtonian
behavior with linear shear stress profile and constant viscosity pro-
file as a function of the applied shear rate (green lines in Fig. 5). At
filler contents of 0.5 wt%, the epoxy/MWCNT-COOH mixtures exhi-
bit Newtonian behavior in the shear rate range of 0.01100 s1,
with dynamic viscosity values similar to those of the neat resin.
For MWCNT-COOH concentrations of 1 wt% and above, the epoxy
dispersions show a clear transition from Newtonian to shear-
thinning behavior characterized by a distinct decrease of the vis-
cosity values when the shear rate varies from 0.01 s1 to 100 s1.
The viscosity difference between the low and high shear rates
can reach two orders of magnitude (from approximately 100 to
1 Pa s) for nanotube loadings of 3 and 5 wt%. The rheological
behavior of epoxy/MWCNT-COOH can be explained considering
the onset of percolation at loading >1 wt%. At the percolation
threshold, the interaction among MWCNTs is promoted due to a
smaller effective distance between nanotubes, and the stronger
degree of entanglement translates into significantly higher viscos-
ity values for the MWCNT-reinforced epoxy resin. The observed
shear thinning behavior, which becomes stronger as particle load-
ing is increased, is typical of suspensions with particle-particle
interactions. In fact, such interactions are disrupted at higher shear
rates leading to lower measured viscosity values that those at low
shear rates.
The extent of the shear thinning nature of the nano-reinforced
epoxy was quantified by fitting the steady shear rheological data
using the general four-parameter Carreau model in Eq. (7). Table 1
summarizes the fitted parameters for the epoxy Prime 20LV sys-
tem reinforced by MWCNT-COOH (loading 1, 3, and 5 wt%)
obtained from the rheological data reported in Fig. 5, and for the
epoxy resins containing xGnP-M5 (5 wt%) and xGnP-H5 (3 and
5 wt%) obtained from the data in Figs. 7 and 8. In particular, results
in the table show how the shear thinning parameters change as a
function of the nanofiller type and concentration, together with
the goodness of the fit in terms of reduced v2. Referring to the
 M. Clausi et al. / Composites: Part A 88 (2016) 8697 91

Fig. 5. Steady shear rheological measurements (solid lines for shear stresses, lines with symbols for viscosities) of epoxy/MWCNT-COOH mixtures with different
concentrations (0.5, 1, 3, and 5 wt%) of carbon nanotubes at T = 23 C. (For interpretation of the references to colour in this figure legend, the reader is referred to the web
version of this article.)

Table 1
Carreau equation parameters for shear-thinning carbon-reinforced Prime 20LV determined from non-linear least-squares fitting of the viscosity data (shear rate range
0.01100 s1) using Eq. (7). Reduced v2 is used as index of goodness of fit.

Nano-reinforced fluid g0 (Pa s) g1 (Pa s) k (s) n Red. v2

Epoxy/MWCNT-COOH 1 wt% 3.60 0.47 57.6 0.683 0.0007

Epoxy/MWCNT-COOH 3 wt% 213.90 0.08 138.9 0.359 1.3205
Epoxy/MWCNT-COOH 5 wt% 591.44 0.80 705.9 0.305 0.1523
Epoxy/xGnP-M5 5 wt% 50.84 3.05 37.5 0.277 2.3009
Epoxy/xGnP-H5 3 wt% 2.80 0.86 29.8 0.635 0.0016
Epoxy/xGnP-H5 5 wt% 142.26 1.49 838.4 0.435 0.1141

epoxy/MWCNT-COOH samples, the value of the power-law index
n, which measures the deviation from Newtonian behavior,
decreases from 0.683 to 0.305 upon increasing the nanotube con-
centration from 1 wt% to 5 wt%. As for the relaxation time constant
k, we observe an increases from 57.6 s for MWCNT-COOH loading
of 1 wt% to 705.9 s for loadings of 5 wt%, indicating that the transi-
tion from Newtonian to non-Newtonian behavior (onset of power-
law region at c_ c  1=) occurs at lower shear rate values as the nan-
otube loading increases.
The rheological data for Prime 20LV epoxy reinforced with gra-
phene nanoplatelets of grades C750, M5, and H5 are shown in
Figs. 57, respectively. Nano-reinforced epoxy samples containing
xGnP-C750 particles show the typical Newtonian behavior, with
constant viscosity profiles as a function of the shear rate, at all
nanofiller loadings up to 5 wt% (Fig. 6). The steady shear rheologi-
cal data of epoxy with xGnP-M5 obtained at 23 C are reported in
Fig. 7. For this system, mixtures with xGnP-M5 loading values from
0.5 wt% to 3 wt% show Newtonian behavior for the entire range of
shear rate, whereas at higher concentrations the viscosity trend
starts to deviate. In particular, strong shear thinning behavior
(n = 0.277) was found for xGnP-M5 loading of 5 wt% in the shear
rate range under consideration. A similar rheological behavior
was observed for the epoxy samples containing xGnP-H5 nanopla-
telets (Fig. 8), except that the transition from Newtonian to shear
thinning behavior occurred at a lower xGnP-H5 loading (below
3 wt%).
Fig. 9 shows the absolute viscosities of the graphene-
reinforced epoxy samples, measured at the shear rate of 10 s1
and at 23 C, as a function of the nanofiller content. All three
grades of graphene nanoplatelets are reported: xGnP-C750
(thickness 2 nm, SSA = 750 m2 g1), xGnP-M5 (thickness
7 nm, SSA = 120150 m2 g1), and xGnP-H5 (thickness
15 nm, SSA = 5080 m2 g1). As expected, the viscosity of the
nano-reinforced samples becomes larger when the nanofiller
content increases. The increase trend is approximately linear in
the range up to 5 wt% for the epoxy/xGnP-C750 samples, which
exhibit Newtonian behavior in such loading range, whereas it
deviates significantly for the nanoplatelets with larger thickness
and lower values of SSA. In particular, after approximately 3 wt%
of nanoplatelets loading, absolute viscosities of epoxy/xGnP
fluids increase markedly with the increase of xGnP thickness.
This behavior is related to the degree of nanoplatelet dispersion
inside the epoxy matrix, which is controlled by interactions at
both interfaces nanoplatelet/nanoplatelet and nanoplatelet/
epoxy. Such interactions are determined by several factors,
including the aspect ratio (average thickness and diameter) of
the nanoplatelets and their specific surface area. Experimentally,
we observed that the xGnP-C750 platelets were those exhibiting
92 M. Clausi et al. / Composites: Part A 88 (2016) 8697

Fig. 6. Steady shear rheological measurements (solid lines for shear stresses, lines with symbols for viscosities) of epoxy/xGnP-C750 mixtures with different concentrations
(0.5, 1, 3, and 5 wt%) of nanoplatelets at T = 23 C. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 7. Steady shear rheological measurements (solid lines for shear stresses, lines with symbols for viscosities) of epoxy/xGnP-M5 mixtures with different concentrations
(0.5, 1, 3, and 5 wt%) of nanoplatelets at T = 23 C. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

the higher degree of dispersion in Prime 20LV epoxy. In addition,
from the rheological data analysis of the epoxy/xGnP samples, it
is evident that the lower the specific surface area of the nanopla-
telets, the lower is the value of critical wt% above which the
nano-reinforced Prime 20LV fluid assumes non-Newtonian
behavior. This threshold loading value was identified at 4 wt%
for the epoxy/xGnP-M5 system (particle aspect ratio = 714), and
at 2 wt% for the epoxy/xGnP-H5 (particle aspect ratio = 333).
In order to assess any effect of particle alignment on the viscos-
ity data, particularly for dispersions with non-Newtonian behavior,
consecutive rheological measurements in the shear rate range
0.1100 s1 were carried out on the nanocomposite epoxy fluids
with different loadings of xGnP-M5 and xGnP-H5 (see Figs. S1
and S2 in the Supporting Information). Results show that devia-
tions of viscosity data in time, considering consecutive
measurements with 5 min interval, are in general not large for
 M. Clausi et al. / Composites: Part A 88 (2016) 8697 93

Fig. 8. Steady shear rheological measurements (solid lines for shear stresses, lines with symbols for viscosities) of epoxy/xGnP-H5 mixtures with different concentrations
(0.5, 1, 3, and 5 wt%) of nanoplatelets at T = 23 C. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 9. Absolute viscosities of epoxy/xGnP mixtures from rheological measurements at frequency of 10 s1 and T = 23 C for the three grades of nanoplatelets (C750, M5, and
H5) with specific surface area as reported by producer. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this
article.)

epoxy/xGnP-H5 at loadings of 3 and 5 wt% (Fig. S1). For example,
viscosity difference is about 10% at shear rate of 1 s1, and becomes
smaller (<5%) at shear rates larger than 10 s1. Further, it is inter-
esting to note that no significant difference can be seen in consec-
utive viscosity measurements of epoxy loaded with the thinner
xGnP-M5 nanoplatelets (Fig. S2) as compared to those with thicker
xGnP-H5 (Fig. S1). To investigate the presence of an unsteady
regime of viscosity in the initial period of spinning due to nanopar-
ticle alignment, transient rheological measurements were con-
ducted on the epoxy/xGnP-M5 and epoxy/xGnP-H5 dispersions
with non-Newtonian behavior at various shear rates (Figs. S3S5
in the Supporting Information). For all dispersions, the transient
viscosity regime becomes significantly reduced at higher shear
rates. In particular, at the shear rate of 1 s1 the initial transient
regime is limited to 5 s, whereas no unsteady regime of viscosity
is measured above 1 s for shear rates of 10 s1.
4.3. Thickness analysis of spin-coated nanocomposite films
For the analysis of the film thickness in relation to Newtonian or
non-Newtonian behavior of the nano-reinforced fluids, composite
samples made of epoxy/MWCNT-COOH or epoxy/xGnP were
94 M. Clausi et al. / Composites: Part A 88 (2016) 8697

Table 2 variation from 1000 to 4000 rpm is more significant than in ranges
Fitted values of shear rate c_ for carbon-reinforced Prime 20LV with non-Newtonian with higher speeds. For all type of samples and considering the
behavior determined from non-linear least-squares fitting of experimental film
thicknesses using Eq. (8) (spinning speed range 10004000 rpm).
same spin-coating time and angular speed, the experimental val-
ues of film thicknesses increase with higher loadings of carbon
Nano-reinforced fluid c_ fit (s1) Standard error Adj. R2 nanoparticles as a consequence of the larger viscosity of the
Epoxy/MWCNT-COOH 1 wt% 9.04 1.458 0.996 nano-reinforced epoxy.
Epoxy/MWCNT-COOH 3 wt% 5.68 0.218 0.996 In the case of films made from nano-reinforced fluids with New-
Epoxy/MWCNT-COOH 5 wt% 2.80 0.133 0.995
tonian behavior, the experimental thicknesses were measured and
Epoxy/xGnP-M5 5 wt% 4.20 0.883 0.990
Epoxy/xGnP-H5 3 wt% 3.57 0.716 0.997 directly compared with those calculated using Eq. (6). In this equa-
Epoxy/xGnP-H5 5 wt% 2.92 0.224 0.997 tion, the shear-independent viscosity values from rheological mea-
surements at shear rate of 10 s1 were used. For films made from
carbon-reinforced epoxy with non-Newtonian behavior, we used
fabricated by spin-coating with angular speeds in the range of the Carreau equation parameters previously determined by fitting
10004000 rpm, immediately followed by a heat cure cycle. This the viscosity data as a function of the shear rate (Table 1), and then
range of spinning speed was chosen because the film thickness fitted the measured film thicknesses using Eq. (8) for spinning

Fig. 10. Calculated thickness of spin-coated nanocomposite films from epoxy/MWCNT-COOH mixtures as a function of spinning speed x, and comparison with measured
experimental values (symbols) for spin-coating time of 10 s and viscosity data at T = 23 C. Each experimental thickness is the average of at least five measurements (error
bars represent standard deviations). Theoretical thicknesses calculated using Eq. (6) for Newtonian dispersions (MWCNT-COOH 0.5 wt%) and Eq. (8) for non-Newtonian
dispersions (MWCNT-COOH 1, 3, 5 wt%) with viscosity parameters in Table 1. (For interpretation of the references to colour in this figure legend, the reader is referred to the
web version of this article.)

Fig. 11. Calculated thickness of spin-coated nanocomposite films from Newtonian epoxy/xGnP-C750 mixtures as a function of spinning speed x using Eq. (6), and
comparison with measured experimental values (symbols) for spin-coating time of 10 s and viscosity data at T = 23 C. Each experimental thickness is the average of at least
five measurements (error bars represent standard deviations). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of
this article.)
 M. Clausi et al. / Composites: Part A 88 (2016) 8697 95

Fig. 12. Calculated thickness of spin-coated nanocomposite films from (a) epoxy/xGnP-M5 and (b) epoxy/xGnP-H5 mixtures as a function of spinning speed x, and
comparison with measured experimental values (symbols) for spin-coating time of 10 s and viscosity data at T = 23 C. Each experimental thickness is the average of at least
five measurements (error bars represent standard deviations). Theoretical thicknesses calculated using Eq. (6) for Newtonian dispersions (xGnP-M5 0.5, 1, 3 wt%; xGnP-H5
0.5 and 1 wt%) and Eq. (8) for non-Newtonian dispersions (xGnP-M5 5 wt%; xGnP-H5 3 and 5 wt%) with viscosity parameters in Table 1. (For interpretation of the references
to colour in this figure legend, the reader is referred to the web version of this article.)

speeds in the range from 1000 to 4000 rpm. Characteristic shear
rate values (c_ fit ) were so determined in order to relate the experi-
mental film thicknesses to the shear-thinning behavior of the
nano-reinforced epoxy/carbon fluids. The characteristic values of
shear rate fitted with the non-linear least-squares Levenberg-
Marquardt algorithm are reported in Table 2. In general, the value
of c_ fit decreases when the loading of carbon nanoparticle increases,
and it is in the range of 2.809.04 s1 for nanocomposite films from
non-Newtonian MWCNT-COOH-reinforced epoxy, and in a lower
range (2.924.20 s1) for films made from epoxy/xGnP mixtures.
These results indicate that for a theoretical thickness prediction
of films spin-coated from shear-thinning nanocomposite fluids, it
is possible to consider the viscosity values corresponding to shear
rates towards the end of the power-law region, which generally
occurs around 10 s1.
An important factor that could influence the results of our anal-
ysis is the role of shear-induced nanoparticle alignment, which
could translates into unsteady viscosity values in the initial stage
of spinning. In this regard, transient rheological measurements at
various shear rates showed that for non-Newtonian nanocompos-
ite fluids containing nanoplatelets grades xGnP-M5 and xGnP-H5,
the unsteady viscosity regime due to nanoplatelets alignment
under shear forces progressively reduces with increasing the shear
rate. In particular, the transient viscosity regime is limited to few
seconds at shear rates of 1 s1 and approximately 1 s at shear rates
of 10 s1 (Figs. S3S5). Considering that spin-coating is a process
characterized by high shear rates, we therefore neglected any
influence of transient viscosity regimes on the predicted film thick-
ness. In addition, in the previous section, we have discussed results
of consecutive measurements of steady-shear rheological data for
96 M. Clausi et al. / Composites: Part A 88 (2016) 8697
epoxy systems containing xGnP-H5 (Fig. S1) and xGnP-M5
(Fig. S2), showing that viscosities difference at shear rates above
10 s1 are less than 5% for epoxy/xGnP-H5, and less than 3% for
epoxy/xGnP-M5.
Fig. 10 shows the measured and predicted thicknesses of films
spin-coated from epoxy/MWCNT-COOH mixtures as a function of
the spinning speed x. In the case of nanocomposite fluids with
Newtonian behavior (MWCNT-COOH loading of 0.5 wt%), the film
thickness values predicted using Eq. (6) are in good agreement
with the experimental values, with an average error of 4% for
spinning speeds in the range 10004000 rpm. The predicted and
measured thicknesses of films spin-coated from the Newtonian
epoxy/xGnP-C750 fluids are reported in Fig. 11. For these samples,
the average errors for each nanoplatelet loading (considering
spinning speeds from 1000 to 4000 rpm) are in the range
5.711.7%. Average errors below 10% were obtained for calculated
thicknesses of nanocomposite films from Newtonian
epoxy/xGnP-M5 (filler loadings 0.53 wt% in Fig. 12a) and
Newtonian epoxy/xGnP-H5 (filler loadings 0.51 wt% in Fig. 12b).
5. Conclusions
Nano-carbon composite films were fabricated by spin-coating
mixtures of thermosetting Prime 20LV and carbon nanofillers,
MWCNT-COOH or exfoliated GnP, at different loadings and spin-
ning speeds, followed by a heat cure cycle. The experimental film
thicknesses were compared to values predicted using analytical
equations containing shear-independent viscosities for the Newto-
nian nano-reinforced epoxy or Carreau-modeled viscosities for the
shear-thinning epoxy/nanofiller mixtures as determined in steady
shear rheological investigations. Rheological measurements were
useful to determine the particle loading at which the epoxy/-
nanofiller systems transition from Newtonian (shear-independent
viscosities) to non-Newtonian (viscosities dependent on the
applied shear rate) behavior due to the onset of percolation. For
the Newtonian epoxy/nanofiller fluids, analytical film thicknesses
were a good approximation of experimental values (error below
10%). On the other hand, in the case of non-Newtonian mixtures,
our fitting procedure was useful to determine the range of shear
rates corresponding to viscosities that best fitted the experimental
values. We found that these characteristic shear rate values for all
the non-Newtonian samples that we investigated are in a well-
defined range (39 s1) towards the end of the power-law region.
The rheological properties of the nano-reinforced epoxy fluids
and their wetting behavior on Mylar substrates were used to
investigate the role of the geometrical features of the carbon
nanoparticles (platelet size and thickness) on the final composite
films. Results showed that the surface spreading of the graphene-
reinforced epoxy decreases upon increasing the nanoplatelet thick-
ness, thus indicating a higher compatibility with the underlying
substrate when using particles characterized by larger values of
specific surface area. Indeed, the epoxy/graphene samples with
xGnP-C750 platelets were the most compatible fluids with respect
to Mylar. Rheological investigations showed that the transition
from Newtonian to non-Newtonian behavior of the nano-
reinforced epoxy is also dependent on the particle specific surface
area. In particular, the threshold loading value was lower for the
epoxy/xGnP-H5 system, characterized by thicker nanoplatelets
(aspect ratio = 333), than for the epoxy/xGnP-M5 system with
thinner nanoplatelets (aspect ratio = 714).
Acknowledgements
This research was funded by Sapienza University of Rome
(Research Project Grant C26A14337Z). M.G.S. acknowledges
financial support from the Italian Ministry of Education, University
and Research (MIUR) through the Rita Levi Montalcini
Programme.
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.compositesa.
2016.05.026.
References
[1] Yu X, Rajamani R, Stelson K, Cui T. Fabrication of carbon nanotube based
transparent conductive thin films using layer-by-layer technology. Surf Coat
Technol 2008;202(10):20027.
[2] Huang H, Liu C, Wu Y, Fan S. Aligned carbon nanotube composite films for
thermal management. Adv Mater 2005;17(13):16526.
[3] Peng H. Aligned carbon nanotube/polymer composite films with robust
flexibility, high transparency, and excellent conductivity. J Am Chem Soc
2008;130(1):423.
[4] Tang W, Santare MH, Advani SG. Melt processing and mechanical property
characterization of multi-walled carbon nanotube/high density polyethylene
(MWNT/HDPE) composite films. Carbon 2003;41(14):277985.
[5] Li J, Hu L, Wang L, Zhou Y, Grner G, Marks TJ. Organic light-emitting diodes
having carbon nanotube anodes. Nano Lett 2006;6(11):24727.
[6] Eda G, Chhowalla M. Graphene-based composite thin films for electronics.
Nano Lett 2009;9(2):8148.
[7] Santonicola MG, Coscia MG, Botti S, Laurenzi S. Graphene/DNA nanostructured
films for bioinspired sensing of UV radiation effects. Proceedings of the
International Astronautical Congress, IAC 2014;9:63137.
[8] Philip B, Abraham JK, Chandrasekhar A, Varadan VK. Carbon nanotube/PMMA
composite thin films for gas-sensing applications. Smart Mater Struct 2003;12
(6):935.
[9] Zhang L, Ni Q-Q, Shiga A, Natsuki T, Fu Y. Preparation of
polybenzimidazole/functionalized carbon nanotube nanocomposite films for
use as protective coatings. Polym Eng Sci 2011;51(8):152532.
[10] Smith JG, Connell JW, Watson KA, Danehy PM. Optical and thermo-optical
properties of space durable polymer/carbon nanotube films: experimental
results and empirical equations. Polymer 2005;46(7):227684.
[11] Li Y, Chen C, Zhang S, Ni Y, Huang J. Electrical conductivity and
electromagnetic interference shielding characteristics of multiwalled carbon
nanotube filled polyacrylate composite films. Appl Surf Sci 2008;254
(18):576671.
[12] Yang Y, Gupta MC, Dudley KL. Towards cost-efficient EMI shielding materials
using carbon nanostructure-based nanocomposites. Nanotechnology 2007;18
(34):345701.
[13] Laurenzi S, Sirilli M, Pinna M, Santonicola MG. DNA-assisted dispersion of
multi-walled CNTs in epoxy polymer matrix. Mater Res Soc Symp Proc
2014;1700:1038.
[14] Safadi B, Andrews R, Grulke E. Multiwalled carbon nanotube polymer
composites: synthesis and characterization of thin films. J Appl Polym Sci
2002;84(14):26609.
[15] Norrman K, Ghanbari-Siahkali A, Larsen NB. 6 Studies of spin-coated polymer
films. Annu Rep Sect C (Phys Chem) 2005;101(0):174201.
[16] Chang J-F, Sun B, Breiby DW, Nielsen MM, Slling TI, Giles M, et al. Enhanced
mobility of poly (3-hexylthiophene) transistors by spin-coating from high-
boiling-point solvents. Chem Mater 2004;16(23):47726.
[17] Chang C-C, Pai C-L, Chen W-C, Jenekhe SA. Spin coating of conjugated polymers
for electronic and optoelectronic applications. Thin Solid Films 2005;479
(1):25460.
[18] Komikado T, Inoue A, Masuda K, Ando T, Umegaki S. Multi-layered mirrors
fabricated by spin-coating organic polymers. Thin Solid Films 2007;515
(7):388792.
[19] Krebs FC. Fabrication and processing of polymer solar cells: a review of
printing and coating techniques. Sol Energy Mater Sol Cells 2009;93
(4):394412.
[20] Singh A, Mukherjee M. Discrete film thickness in polyacrylamide-CdS
nanocomposite ultrathin films. Phys Rev E: Stat, Nonlinear, Soft Matter Phys
2004;70(5 1). 051608-16.
[21] Yu Y-Y, Chen W-C. Transparent organicinorganic hybrid thin films prepared
from acrylic polymer and aqueous monodispersed colloidal silica. Mater Chem
Phys 2003;82(2):38895.
[22] Haas DE, Quijada JN, Picone SJ, Birnie III DP, Effect of solvent evaporation rate
on skin formation during spin coating of complex solutions. In: Proc. SPIE, vol.
3943, Sol-Gel Optics V;2000. p. 280284.
[23] Flack WW, Soong DS, Bell AT, Hess DW. A mathematical model for spin coating
of polymer resists. J Appl Phys 1984;56(4):1199206.
[24] Yonkoski RK, Soane DS. Model for spin coating in microelectronic applications.
J Appl Phys 1992;72(2):72540.
[25] Ohara T, Matsumoto Y, Ohashi H. The film formation dynamics in spin coating.
Phys Fluids A 1989;1(12):194959.
 M. Clausi et al. / Composites: Part A 88 (2016) 8697
[26] Emslie AG, Bonner FT, Peck LG. Flow of a viscous liquid on a rotating disk. J
Appl Phys 1958;29(5):85862.
[27] Jenekhe SA, Schuldt SB. Coating flow of non-Newtonian fluids on a flat rotating
disk. Ind Eng Chem Fundam 1984;23(4):4326.
[28] Rehg TJ, Higgins BG. The effects of inertia and interfacial shear on film flow on
a rotating disk. Phys Fluids 1988;31(6):1360.
[29] Higgins BG. Film flow on a rotating disk. Phys Fluids 1986;29(11):3522.
[30] Kistler SF, Schweizer PM. Liquid film coating. Chapman Hall; 1997.
[31] Carreau PJ. Rheological equations from molecular network theories. Trans Soc
Rheol 1972;16(1):99127.
[32] Shenoy A. Rheology of filled polymer systems. Springer; 1999.
[33] Battisti A, Skordos AA, Partridge IK. Monitoring dispersion of carbon
nanotubes in a thermosetting polyester resin. Compos Sci Technol 2009;69
(10):151620.
97
[34] El Sawi I, Olivier PA, Demont P, Bougherara H. Investigation of the effect of
double-walled carbon nanotubes on the curing reaction kinetics and shear
flow of an epoxy resin. J Appl Polym Sci 2012;126(1):35866.
[35] Ren X, Kanezashi M, Nagasawa H, Tsuru T. Plasma treatment of hydrophobic
sub-layers to prepare uniform multi-layered films and high-performance gas
separation membranes. Appl Surf Sci 2015;349:4159.
[36] Yuan Y, Lee TR. Contact angle and wetting properties. Surface science
techniques. Springer; 2013. p. 334.
[37] Fan Z, Advani SG. Rheology of multiwall carbon nanotube suspensions. J Rheol
2007;51(4):585604.
[38] Li Y, Zhu J, Wei S, Ryu J, Sun L, Guo Z. Poly (propylene)/graphene nanoplatelet
nanocomposites: melt rheological behavior and thermal, electrical, and
electronic properties. Macromol Chem Phys 2011;212(18):19519.