Journal of Molecular Modeling (2018) 24:178

https://doi.org/10.1007/s00894-018-3716-6

ORIGINAL PAPER

Effect of temperature on elastic properties of CNT-polyethylene

nanocomposite and its interface using MD simulations
Akhileshwar Singh 1 & Dinesh Kumar 1

Received: 8 January 2018 / Accepted: 12 June 2018

# Springer-Verlag GmbH Germany, part of Springer Nature 2018

Abstract
This paper investigates the effect of temperature on the elastic modulus of carbon nanotube-polyethylene (CNT-PE) nanocom-
posite and its interface using molecular dynamics (MD) simulations, by utilizing the second-generation polymer consistent force
field (PCFF). Two CNTs—armchair and zigzag—were selected as reinforcing nano-fillers, and amorphous PE was used as the
polymer matrix. For atomistic modelling of the nanocomposite, the commercially available code Materials Studio 8.0 was used
and all other MD simulations were subsequently performed using the open source code Large-Scale Atomic/Molecular
Massively Parallel Simulator (LAMMPS). To obtain the elastic modulus of the nanocomposite, stress-strain curves were drawn
at different temperatures by performing uniaxial deformation tests on the nanocomposite material, whereas the curvatures of the
interfacial interaction energy vs. strain curves were utilized to obtain Young’s modulus of the interface. In addition, the glass
transition temperatures of the polymer matrix and nanocomposites were also evaluated using density-temperature curves. Based
on the results, it is concluded that, irrespective of temperature condition, a nanocomposite reinforced with CNT of larger chirality
(i.e., armchair) yields a higher value of Young’s modulus of the nanocomposite and its interface. It was also found that, at the
phase transition (from a glassy to a rubbery state) temperature (i.e., glass transition temperature), Young’s moduli of the polymer
matrix, nanocomposite, and its interface drop suddenly. The results obtained from MD simulations were verified with results
obtained from continuum-based rule-of-mixtures.

Keywords CNT . Polyethylene . Nanocomposite . Interface . Molecular dynamics . Young’s modulus

Introduction filler, and a few weight-percentage of CNTs reinforcement in a

The discovery of lightweight carbon nanotubes (CNTs) [1] in
1991 stimulated the attention of researchers across the world
in the fields of science and engineering because of their ex-
ceptional physical and chemical properties. For example,
CNTs are found to possess Young’s modulus, thermal and
electrical conductivities in the ranges of 1–5 TPA [2–5],
6000 W/mK [5] and 200,000 S cm−1 [6], respectively.
Combined with these remarkable properties, a very high as-
pect ratio of CNTs, varying from few hundreds to several
thousand on the nanoscale, makes them promising reinforcing
* Dinesh Kumar
dkumar.mech@mnit.ac.in
Akhileshwar Singh
akhileshwarsingh@live.com
1
Department of Mechanical Engineering, Malaviya National Institute
of Technology, Jaipur 302017, India
polymer matrix can significantly improve the mechanical
properties of the resulting nanocomposite. A significant num-
ber of experimental and computational studies have been con-
ducted to investigate the effective elastic properties of CNT-
reinforced polymer nanocomposites (CRPN).
In the last two decades, many experimental works [7–11]
have demonstrated that reinforcement of CNTs in a polymer
matrix invokes the enhancement of mechanical properties of
the nanocomposite. But due to the difficulty and high cost
involved in experimental works at nanoscales, continuum-
and atomistic-based computational approaches have been
used as powerful tools to predict the properties of CNTs and
other nano-fillers as well as their nanocomposites. Atomistic
simulations can accurately capture the real nature of the atom-
ic interactions, but are computationally expensive compared
with continuum-based techniques. Various atomistic-based
techniques used are: ab initio, tight bonding molecular dynam-
ic (TBDM), density functional theory (DFT) and molecular
dynamic (MD). Of these, the first three are based on quantum
178 Page 2 of 11
mechanics, which need a solution of Schrödinger wave equa-
tion for each atom, whereas the latter (MD) requires solving
Newton’s second law of motion and provide an acceptable
level of accuracy with lesser run time than ab initio, TBDM,
and DFT.
There are many research articles [12–20] predicting the
mechanical properties of CRPN using MD simulation. Han
and Elliott [12] investigated the elastic properties of a CRPN
and found longitudinal Young’s modulus of CRPN equal to
94.6 GPa and 138.9 GPa for 12% and 17% volume fractions
of CNT, respectively. Al-ostaz et al. [13] studied the effect of
aligned- and randomly oriented CNT-reinforced polymer ma-
trix and concluded that the clustering of CNTs reduces the
transverse elastic properties as compared to single-CNT rein-
forced nanocomposite. Mandal et al. [15] investigated the ef-
fect of temperature on the elastic properties of CRPN and
reported that, as the temperature increases from 100 K to
400 K, the stiffness of CRPN decreases. Rouhi et al. [17]
investigated the influence of CNT reinforcement on the elastic
properties CNT-reinforced poly(ethylene oxide) nanocompos-
ite and found dense polymer matrix around CNT, and it was
also shown that zigzag-CNT reinforcement provides higher
Young’s modulus to the nanocomposite. The effects of single
adatom (SA) and Stone-Wales (SW) defects in CNT on the
elastic properties of polypropylene matrix nanocomposite was
studied by Lv et al. [18]. Chawla et al. [21] conducted MD
simulations to study the effect of pull-out of CNT from PE
matrix on the mechanical properties of the nanocomposite and
found a decrease in Young’s modulus of nanocomposite with
pull-out displacement. Ozer and Cabuk [20] computed the
structural parameters, and mechanical and electronic proper-
ties of SbXI (X = S, Se, Te) crystals using DFT , and found
that SbXI crystals are mechanically stable, slightly stiff, elas-
tically anisotropic and ductile in nature. Recently, Alian and
Meguid [19] predicted the elastic moduli of the large model
(i.e., up to ∼9 × 105 atoms) of CNT-reinforced representative
volume elements, considering different parameters of CNTs
such as different lengths, curvatures, and bundle sizes to sim-
ulate actual nanocomposites.
Interfacial interactions between the nanofiller and the ma-
trix play a significant role in enhancing the bulk properties of
the resulting nanocomposite, but this area is less explored
because of the experimental limitations in probing material
behavior near interface at the nanoscale. Therefore, atomistic
simulations through MD can be applied to obtain nanoscale
interfacial information to incorporate subsequently into
continuum-level modeling for the theoretical prediction of
the bulk properties of the nanocomposite. There are very
few publications [22–27] on the prediction of the mechanical
behavior of the interface using atomistic-based simulations.
Zheng et al. [22] studied the influence of chirality on the
interfacial interaction energy of CNT-reinforced polymethyl
methacrylate (PMMA) and found that CNTs with larger chiral
J Mol Model (2018) 24:178
angle have higher interfacial interaction energy. Yang et al.
[23] developed a multiscale approach to consider the effect
of CNT size and non-bonded interaction at the interface on the
effective elastic stiffness of CRPN using MD simulations and
a modified multi-inclusion model using continuum
micromechanics. The load transfer behavior of graphene/
CNT-PE hybrid nanocomposite was investigated by Zhang
et al. [24], and it was found that the peak strength of traction
curve for PE matrix is much lower than the interface of PE/
graphene and graphene/graphene. Arash et al. [25] proposed
an approach to predict Young’s modulus of the interfacial
region between CNT-reinforced PMMA nanocomposites
and also presented a three-phase micromechanical model for
predicting the elastic properties of the nanocomposite. Very
recently, Srivastava and Kumar [26] developed a continuum
level model to characterize interphase region and found that
elastic properties of the nanocomposite are enhanced by con-
sidering the effect of interface for CRPN.
Moreover, Young’s modulus and high thermal conductivity
of CNTs can significantly modify the thermo-mechanical
properties of the CNT-polymer nanocomposite. Further, these
properties of CNT, as well as the properties of the resulting
nanocomposite, such as elastic moduli [27] and load transfer
characteristic [28, 29], are greatly influenced by the high-
temperature environments, especially when the temperature
involved is higher than the glass transition temperature (Tg)
of the nanocomposite. The Tg of the nanocomposite is influ-
enced by many parameters, such as the degree of polymeriza-
tion [28] and interfacial interaction between CNT and polymer
matrix [14, 29]. Several studies have also reported the effect of
CNT reinforcement in a polymer matrix on the thermo-
mechanical response and Tg of the CNT-polymer nanocom-
posite. For instance, Qi [30] used the density-temperature
curve to find the Tg of the CNT-polyimide nanocomposite,
and showed that CNT reinforcement reduces the softening
effects of temperature on mechanical properties. Wei et al.
[28] investigated the thermal expansion and diffusion coeffi-
cients of characteristics of CNT-PE composites and found that
CNTs reinforcement to the polymer matrix enhances the Tg,
diffusion, and thermal expansion coefficients of the composite
system. Li and Strachan [31] characterized the thermo-
mechanical response of pure thermoset polymer composed
of epoxy EPON862 and curing agent DETDA, and it was
found that mechanical response and Tg of the depend on the
degree of polymerization.
It is evident from the above literature study that there are
number of experimental and analytical (MD/continuum) re-
search articles on the mechanical characterization of the
CRPN and the interface, as well as few investigations involv-
ing temperature effects on the properties of the nanocompos-
ite. However, there is no comprehensive study on the
temperature-dependent elastic properties of the CNT-
reinforced nanocomposite including its interfacial region.
J Mol Model (2018) 24:178
Fig. 1 Atomistic unit cell of
carbon nanotube (CNT)
Further, there is a scarcity of any investigations on the effect of
Tg on the elastic properties of nanocomposite and its interface.
Therefore, the objective of the present study was to explore
the elastic modulus of a periodic atomistic model of nanocom-
posite (CNT-reinforced polyethylene) and its interface (be-
tween CNT and PE matrix) for different temperatures using
MD simulations. The Tg values of the polymer and the nano-
composite were also evaluated using density-temperature
curves. The atomistic modeling was done using Material
Studio 8.0 [32], and further MD simulations were performed
using Large-Scale Atomic/Molecular Massively Parallel
Simulator (LAMMPS) [33]. The present report is organized
as follows: In Section Atomistic modeling, a general proce-
dure for modeling an atomistic model of the polymer/
nanocomposite is discussed. The force field and its parameters
used in the study (i.e., PCFF) are explained in Force field.
Section Simulation procedure describes the equilibration and
simulation procedures carried out using LAMMPS. The meth-
od to determine elastic modulus of the material system is
provided in Present Study. In the section Validation Study, a
validation study is conducted with a view to validate the meth-
odology followed for MD simulations in the current work.
Finally, the results are presented and discussed in Results
and discussion.
Fig. 2a,b Atomistic unit cells. a
CNT-reinforced polyethylene
(PE) nanocomposite. b Pure PE
matrix
(a)
Page 3 of 11 178
Methods
Atomistic modeling
In the present study, two CNTs—armchair type with chiral vector
(5, 5) and zigzag type with chiral vector (9, 0)—having nearly
same diameters of 6.78 Å and 7.05 Å, respectively, were selected
as reinforcing fillers, and amorphous PE was used as the polymer
matrix. For simplicity and generic representation of polymer ma-
trix, a single chain of PE, with 500 repeating units of –CH2– in
each chain, was chosen as a building block to model the actual
polymer. Further, a periodic atomistic model of CNT-reinforced
polyethylene (i.e., CRPE) was constructed to avoid the end-effect
of CNT. The modeling procedure of all atomistic models (i.e.,
armchair and zigzag CNT-reinforced PE nanocomposites, and
pure bulk PE matrix) was similar and performed using a commer-
cial MD simulation package: Material Studio 8.0. However, for
the sake of brevity, only the modeling details of the armchair-CNT
nanocomposite is discussed in the following paragraph. The poly-
mer consistent force field (PCFF) [34] was used throughout the
analysis, which is considered in section Force field.
Initially, a pristine armchair CNT was placed at the center of a
periodic unit cell (60 Å × 60 Å × 51.65 Å) with its length aligned
along the z-direction, as shown in Fig. 1, and, subsequently, PE
PE chains
CNT
(b)
178 Page 4 of 11 J Mol Model (2018) 24:178

Fig. 3a–d Schematic

representation of polymer
consistent force field (PCFF)
interaction. a Bond interaction. b
Angle interaction. c Dihedral
interaction. d Non-bonded (a)
interaction (b)

(c) (d)

chains, were packed into the unit cell to form CNT-PE model with
a target density of 0.8 g/cm3, as shown in Fig. 2a. In the process,
the total potential energy of the nanocomposite was minimized
using the steepest descent method [35] with energy convergence
tolerance of 0.001 kcal mol−1. A similar modeling procedure was
followed to prepare the atomic model for the pure PE matrix, as
shown in Fig. 2b.
however, the non-bonded interaction energy is considered be-
cause of van der Waals interaction only, which is defined by
the Lennard-Jones (LJ) 9–6 inter-atomic potential, for a po-
tential cutoff radius (i.e., rcutoff) of 10 Å and having interatom-
ic potential parameters given in Table 1. More detailed infor-
mation and description of the second generation PCFF can be
found in the literature [34].

Force field
Simulation procedure
The second generation PCFF was used in the present study to
describe the covalent interactions of atoms in the CNT and the After building the initial configuration of the unit cell, MD simu-
polymer as well as the non-covalent interaction at CNT–PE lations were performed with LAMMPS utilizing the PCFF force
matrix interface. The schematic representations of bonded and field. The potential energy of the CNT-PE model (i.e., unit cell)
non-bonded interactions of PCFF are shown in Fig. 3. Based was minimized using the conjugate gradient method, before an-
on the PCFF force field, the total potential energy (i.e., Epot) of nealing it to 600 K for 100 ps and subsequently quenching to
an atomistic system includes different components, as given in 100 K in 200 ps by following the NVT ensemble using the
the following equation: Nose-Hoover thermostat. Thereafter, the material system is equil-
ibrated under NPT ensemble for 200 ps at a temperature of 100 K
E pot ðr1 ; r2 …rn Þ ¼ Ebond þ E angle þ Edihedral þ E improper and pressure of 1 atm in all directions, so that the volume and
þ E non−bonded: ð1Þ potential energy of the material become stable and relaxed. The
final equilibrated density of the polymer matrix relaxed around the
value of 0.92 g cm−3, which is lower than the experimental value
where the bonded interaction energy consists of: Ebond, of 0.95 g cm−3 [36]. The difference is because MD simulations
Eangle, Edihedral, Eimproper terms, and the non-bonded interac- generally consider material free from defects, and the veracity of
tion energy (Enon−bonded) in PCFF includes van der Waals the parameterization in the force field. A similar procedure was
(vdW) and electrostatic interactions. In the present work, followed to generate the equilibrated models for the temperatures
(i.e., quenched temperature) of 300 K, 400 K, and 500 K.
Table 1 Non-bonded Lennard-Jones (LJ) 9–6 interatomic potential After the equilibration procedure, MD simulations
parameters were performed at NVT ensemble using the set-up
LJ 9–6 interactiona ε (kcal mol−1) σ (Å) shown in Fig. 4 under uniaxial deformation. While
keeping the lower group of the nanocomposite fixed,
ðC−CÞsp2 0.064 4.010 the upper group is subjected to a constant velocity of
H−H 0.020 2.995 1 × 10−4 Å fs−1, also shown in Fig. 4. In order to get
ðC−CÞsp3 0.054 4.010 the elastic properties of the nanocomposite from the
a
stress-strain curves obtained during uniaxial deforma-
C−C and H–H represent interactions between carbon and hydrogen
tion, the stress (virial) was calculated using the follow-
atoms, respectively, and the subscripts sp2 and sp3 specify the type of
hybridization ing relation [37, 38]:
J Mol Model (2018) 24:178 Page 5 of 11 178

Fig. 4 Molecular dynamics (MD) Constant Velocity

simulation set-up for the
nanocomposite under uniaxial
deformation
Upper group of atoms moved
rigidly during simulation

Middle group of atoms are

free to move in any direction
during simulation

Lower group of atoms are

held fixed during simulation

nanocomposite by following the procedure discussed in

∑N mK vKi vKj ∑K f Kj rKi
N
σij ¼ K þ ; ð2Þ Atomistic modeling. Two CNTs—armchair type with chiral
V V vector (5, 5) and zigzag type with chiral vector (9, 0)—having
where V is the volume of the CNT-PE model, N is a diameters of 6.78 Å and 7.05 Å, respectively, were selected as
number of atoms in the model, vKi is the ith component reinforcing fillers to make the nanocomposite. Before
of velocity of Kth atom, rKi is the ith component of the conducting the study, a simulation procedure was performed
position vector of Kth atom, and fKj is jth component of using LAMMPS, described in Simulation procedure, for
force applied on Kth atom. The strain is calculated using equilibration and, subsequently, conducting the uniaxial de-
the following relation [27]: formation, is verified by comparing the result (i.e., elastic
modulus) for CNT-reinforced PE matrix nanocomposite with
ðNo:of stepsÞ  timestep  Velocity
ɛ¼ : ð3Þ that reported by Mahboob and Islam [39]. During the uniaxial
Length of CNT deformation of the nanocomposite, the stresses were calculat-
ed on the atomistic scale using virial stress tensor defined by
Eq. (2). Further, for predicting Young’s modulus of the inter-
Present study facial region, the interfacial interaction energy (i.e., Eint,
a function of applied strain ɛ) between CNT and PE matrix
In the present study, MD simulations were conducted to pre- is calculated after each step of the applied constant velocity
dict the elastic constant of the nanocomposite and its interfa- and a polynomial curve of the second degree is fitted to close-
cial region (between CNT and PE matrix) using second gen- ly interpolate the Eint data points with respect to applied strain.
eration PCFF, as described in the section Force field. Material The Young’s modulus of the interface region at different tem-
Studio 8.0 was used for atomistic modeling of the peratures was then calculated using the obtained expression.

Fig. 5 a Simulation model of

CNT-PE matrix nanocomposite,
reinforced with 9.7% volume
fraction of CNT. b Stress-strain
curve of the nanocomposite

(a) (b)
178 Page 6 of 11 J Mol Model (2018) 24:178

Glassy Glassy
state state

Rubbery Rubbery
state state

(a) (b)
Fig. 6a,b Temperature-density plots. a Bulk PE matrix, b CNT (zigzag and armchair)-reinforced PE nanocomposites

Thereafter, Young’s modulus of the nanocomposite at differ- Results and discussion

ent temperatures, taken in the range of 100 K–500 K, was
obtained. The Tg values of the polymer and nanocomposite
are also evaluated using density-temperature curves. Finally,
for the verification of the results obtained from MD simula-
tions, a comparison of results with the continuum-based rule-
of-mixtures (ROM) is made.
Validation study
To validate the methodology of the present study, a CNT [of
chirality (6,6)] -polyethylene nanocomposite is modeled to
compare its Young’s modulus with that reported by
Mahboob and Islam [39]. The lattice parameters of the unit
cell (30 Å × 30 Å × 51.65 Å) and volume fraction (9.7%)
used were the same as in the reference paper [39]. The slope of
the linear elastic region (taken up to 3% strain) of the obtained
stress-strain curve shown in Fig. 5b was used to predict
Young’s modulus of the nanocomposite. Young’s modulus
obtained at 300 K from the simulation procedure followed in
the present paper was 107 GPa, which is in good agreement
with the value of 103 GPa reported by Mahboob and Islam
[39].
In this section, initially, the Tg values of bulk PE matrix and
CNT-reinforced nanocomposite are evaluated and discussed
in the section Glass transition temperature. Young’s moduli of
PE matrix and CNT, interface, and the resulting nanocompos-
ite are calculated in Elastic moduli of pure PE matrix and
CNT, Elastic modulus of the interfacial region of the nano-
composite and Elastic modulus of the nanocomposite, respec-
tively. In addition, the influence of phase transition (from
glassy to a rubbery state) temperature (i.e., Tg) on Young’s
moduli of the polymer matrix, nanocomposite, and its inter-
face is also discussed in these sections. Finally, in Comparison
of MD results with continuum-based ROM, the obtained
Young’s moduli of PE matrix, CNT, interface and the resulting
nanocomposite from MD simulations are verified using
continuum-based ROM.
Glass transition temperature
MD simulations provide an effective approach for evaluating
the Tg of the bulk polymer matrix and its nanocomposite. To
evaluate Tg, the equilibrated structure of the bulk PE matrix/
nanocomposite at 500 K was slowly cooled down with a rate

Table 3 Initial values of interfacial interaction energy Eint (kcal mol−1)

Table 2 Elastic modulus of PE matrix at different temperatures
of nanocomposite at different temperatures
Temperature (K) Density (g cm−3) Young’s modulus (GPa)
CNT type (chirality) Temperature (K)
100 0.92 2.05
100 300 400 500
300 0.85 1.20
400 0.77 0.89 Zigzag (9,0) −353.95 −323.49 −259.30 −230.10
500 0.72 0.72 Armchair (5, 5) −366.77 −335.60 −269.15 −244.92
J Mol Model (2018) 24:178
Interfacial region
Fig. 7 An atomistic model of the nanocomposite featuring interfacial
region
of 0.5 K ps−1 from 500 K to 50 K at NPT ensemble. The
variation of density is monitored over the wide range of tem-
perature values, and is plotted in the form of a density-
temperature plot in Fig. 6a,b along with estimated error bars,
obtained from the density-temperature data using segmented
regression (R2 > 0.99). The Tg values of the bulk PE matrix
and the nanocomposite are estimated at ~295 K and ~320 K,
respectively, obtained at the intersection points of linear fits to
the initial (i.e., glassy-state) and final (i.e., rubbery-state) re-
gions of density-temperature data points, as shown in Fig.
6a,b. It is also important to mention that the Tg of the nano-
composite is found to be same for both armchair- and zigzag-
CNT reinforcements. As reported in several research studies
[28, 40–43], the Tg of long-chained PE falls in the range of
250 K–350 K, which depends on the molecular weights and
degrees of polymerization in the PE matrix. Wei et al. [28] also
found that the CNT reinforcement in PE matrix enhances the
Tg of the nanocomposite in comparison to the bulk polymer
(a)
Fig. 8 Variations in interfacial interaction energy at different temperatures with respect to applied strain. a Armchair-CNT nanocomposite, b zigzag-
CNT nanocomposite
Page 7 of 11 178
matrix. Hence, the results obtained are consistent with the
literature. The influence of Tg, evaluated in this section, on
the Young’s moduli of the polymer matrix, nanocomposite,
and its interface is discussed in detail in the following sections.
Elastic moduli of pure PE matrix and CNT
To evaluate the elastic modulus of the PE matrix, the PE matrix
was equilibrated by following the procedure discussed in
Simulation procedure at different temperatures, and relaxed to
obtain mass density in the range of 0.92–0.72 g cm−3. The ob-
tained values of elastic modulus and density of the PE matrix at
different temperatures are given in Table 2. The values of elastic
modulus of PE matrix evaluated at 100 K and 300 K (i.e.,
2.05 GPa and 1.20 GPa, respectively) are found to be in good
correspondence with the value of 2 GPa reported by Awasthi et
al. [44] at 100 K and the value of 1.22 GPa by Al-ostaz et al. [13]
at 300 K. An expected observation made from the Table 2 is that
the highest and lowest values of Young’s modulus for PE matrix
are found at the highest and lowest temperatures, respectively.
Alternatively, it can be stated that an increase in temperature
would result in softening of the matrix material. Further, a sig-
nificant drop in elastic modulus of the PE matrix was noted when
the temperature changes from 100 K to 300 K and, this indicates
the phase transition of PE matrix from a glassy state to rubbery
state, as shown in Fig. 6a.
The elastic properties of the two CNTs—armchair type
with chiral vector (5, 5) and zigzag type with chiral vector
(9, 0)—having nearly the same diameters at 6.78 Å and 7.05
Å, respectively, were also determined using MD simulations
at a temperature of 300 K. The values of Young’s modulus of
the armchair- and zigzag-CNTs were found to be nearly same
(i.e., 1056.95 GPa) calculated at 300 K.
(b)
178 Page 8 of 11
Fig. 9 Radial distribution function (RDF) of PE atoms of the polymer
matrix around the carbon atoms of the embedded CNT at different
temperatures
Elastic modulus of the interfacial region
of the nanocomposite
The interfacial interaction energy between the CNT and PE
matrix is governed by the non-bonded interaction energy (i.e.,
here EvdW interaction energy only). The interfacial interaction
energy is calculated as the difference between the total poten-
tial energy of nanocomposite (ENC) and the sum of potential
energies of the CNT (ECNT) and polymer matrix (Em), as giv-
en in the following relation:
E int ¼ E NC −E m −E CNT : ð4Þ
The initial values of interfacial interaction energy between
CNT and PE matrix before the uniaxial deformation of the
nanocomposite are represented in Table 3, at different temper-
atures. The negative value of interaction energy shows the
existence of attractive interactions between CNT and PE ma-
trix (Fig. 7). It can be seen from the Table 3 that, irrespective
of temperature, armchair-CNT nanocomposites have more in-
teraction energy than zigzag-CNT nanocomposites; moreover,
an increase in temperature results in a decrease of Eint for both
the nanocomposites. The above finding of higher interaction
Table 4 Elastic modulus (in GPa)
of nanocomposite and its CNT (chirality) Interface
interface at different temperatures
100 K
Zigzag (9, 0) 25.02
Armchair (5, 5) 28.02
J Mol Model (2018) 24:178
energy for an armchair-CNT nanocomposites than zigzag-
CNT nanocomposites is found to be in good concurrence with
results reported by Zheng et al. [22].
Thereafter, the interfacial interaction energy Eint corre-
sponding to the applied strain ε (up to 5%) is calculated after
each step of applied constant velocity at different tempera-
tures. Based on the variation of Eint with respect to applied
strain ɛ, a polynomial curve of second-degree is fitted to close-
ly interpolate the data points (Eint, ɛ) at various temperatures.
The corresponding Eint–ɛ curves are shown in Fig. 8a,b for
armchair- and zigzag-CNT nanocomposites, respectively. The
associated error estimates, for the fitted curves, are obtained
from the Eint–ɛ data using regression analysis, where the
values of the coefficient of determination (R2) for different
curves are as follows: R2 > 0.99 for 100 K and 300 K, R2 >
0.97 for 400 K, and R2 > 0.96 for 500 K.
Thereafter, the interface Young’s modulus is calculated
using the following relation:


1 ∂2 E int ðεÞ
Y int ¼ ; ð5Þ
V int ∂ε2
where, Eint(ε) represents the interfacial interaction energy as a
function of strain, and Vint is the volume of the interface region
(between CNT and PE matrix), which is obtained using the
following equation.
h i
V int ¼ πð rCNT þ t int Þ2 −πr2CNT L; ð6Þ
where L is the length of the CNT-PE model along the axis of
the CNT, rCNT is the radius of the CNT, and tint represents the
thickness of the interfacial region. The thickness of the inter-
facial region (also called as vdW separation distance) is ob-
tained from the radial distribution function (RDF) curve. The
variation of RDF for PE atoms around the CNT in the periodic
unit cell is plotted using VMD [44] in Fig. 9. It can be ob-
served from the Fig. 9 that the RDF of PE atoms is zero until
the radial distance of ~2.7 Å from CNT. This value indicates
the approximate thickness of the interface region between
CNT and PE matrix.
Finally, the values of Young’s modulus calculated at differ-
ent temperatures for the interfacial region of the nanocompos-
ite are given in Table 4. It can be noted from Table 4 that the
highest value of Young’s modulus of the interface is obtained
as 28.02 GPa at a temperature of 100 K for armchair-CNT
nanocomposite, whereas the lowest value of 10.11 GPa is
Nanocomposite
300 K 400 K 500 K 100 K 300 K 400 K 500 K
21.94 15.20 10.11 32.12 29.59 21.01 17.33
25.57 17.16 11.75 34.94 32.95 23.96 21.00
J Mol Model (2018) 24:178
(a)
Fig. 10 Stress-strain curves for a armchair, and b zigzag nanocomposite at different temperatures
evaluated for zigzag-CNT nanocomposite at a temperature of
500 K. It is also important to highlight from Table 4 that,
because of the transition of the state of nanocomposite from
glassy to rubbery around Tg (i.e., 320 K), a significant drop in
Young’s modulus of the interface is also observed between
300 K and 400 K.
Elastic modulus of the nanocomposite
Two types of nanocomposites made by reinforcing armchair-
CNT [with chiral vector (5, 5)] and zigzag-CNT [with chiral
vector (9, 0)] into the pure PE polymer matrix were consid-
ered. The slope of the linear elastic region (taken up to 3%
strain) of the obtained stress-strain curve for a particular nano-
composite at different temperatures taken in the range 100 K–
500 K was used to predict Young’s modulus of the nanocom-
posite. The stress-strain curves drawn at different temperatures
are shown in Fig. 10a,b for armchair- and zigzag-CNT nano-
composites, respectively, and the corresponding evaluated
values of elastic modulus are given in Table 4. It can be seen
from the Table 4 that Young’s modulus of the armchair-CNT
Interface PE matrix
Solid CNT
Fig. 11 Schematic of three-phase continuum-based model of CNT-
reinforced PE matrix featuring interfacial region
Page 9 of 11 178
(b)
nanocomposite shows higher value than zigzag-CNT nano-
composite at all temperatures, and this observation found
good in concurrence with the result presented by Zuberi and
Esat [45] that the larger chirality angle of CNT reinforcement
yields greater Young’s modulus of the nanocomposite.
Further, it can be noted from Fig. 10a,b that the stress-strain
data is grouped into two distinct categories, one with 100 K
and 300 K, and the other with 400 K and 500 K. The Tg (i.e., ~
320 K) of the nanocomposite, as marked in Fig. 6b, lies be-
tween these two distinct categories.
Comparison of MD results with continuum-based
ROM
In this section, an indirect validation of the above findings of
the elastic properties of the CNT, polymer, interface, and
nanocomposite obtained from the MD simulations done
through continuum-based ROM is presented. ROM is based
on the theory of continuum mechanics and assumes that there
is perfectly bonding between the three-phases of the nano-
composite (i.e., CNT, polymer matrix and its interface), as
shown in Fig. 11. By considering the compatibility equations
for strains and equilibrium equations for stresses, Young’s
modulus of a three-phase nanocomposite is expressed mathe-
matically as [26]:
Y NC ¼ V CNT
f Y CNT þ V int
f Y int þ V f Y m ;
m
ð7Þ
where, YNC, YCNT, Yint and Ym represent the elastic moduli of
the nanocomposite, CNT, interface and the polymer matrix,
respectively, and V CNT
f , V int m
f and V f are the volume fractions
of CNT, interfacial region and the polymer matrix in the nano-
composite, respectively. Further, the volume fractions of CNT,
interface and polymer must satisfy the following relationship:
178 Page 10 of 11 J Mol Model (2018) 24:178

Table 5 Verification of molecular dynamics (MD) simulation results through rule-of-mixtures (ROM)

CNT (chirality) Volume fraction Young’s modulus (GPa)

V CNT
f V int
f MD simulation ROM

CNT Interface Nanocomposite Nanocomposite

Zigzag (9, 0) 3.00 3.29 1056.95 21.94 29.59 33.49

Armchair (5, 5) 2.89 3.20 1056.95 25.57 32.95 32.44

evaluated. The Tg values of the polymer and the nanocompos-

V CNT
f þ V int
f þ V f ¼ 1:
m
ð8Þ
Based on the basic micromechanics assumption (i.e., con-
tinuum reinforcement) of ROM, a hollow CNT has to be
modeled first as a solid CNT, and then the volume fraction,
V CNT
f of CNT is computed. The equivalent radius of solid
CNT (rsCNT) is computed by considering the cross-sectional
area of the hollow CNT equal to that of the solid CNT, as
reported by Rafiee and Madhavi [27]. The equations for radius
and volume fraction of solid CNT are derived as:
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
rsCNT ¼ 2rCNT t CNT ; ð9Þ
πr2sCNT
V CNT ¼ ; ð10Þ
f
Acs
where tCNT and rCNT are the wall-thickness and radius of hol-
low CNT, respectively, and Acs represents the cross-sectional
area (50.2 Å × 50.2 Å) of the CNT-PE model (i.e., cross-
sectional area of the final CNT-PE model after NPT
equilibration).
It can be observed from Table 5 that the results obtained
from the continuum ROM for Young’s modulus of the nano-
composite in the axial direction, evaluated using Eq. (7), are
reasonably close to the MD simulations results obtained at
300 K. Therefore, the values of Young’s modulus of CNT,
polymer, interface, and nanocomposite obtained from the
MD simulations were validated through the continuum-
based ROM.
Conclusions
MD simulations were conducted in this study to predict the
elastic properties of a periodic atomistic model of CNT-
reinforced PE nanocomposite and its interfacial region (be-
tween CNT and PE matrix) using second generation PCFF.
The two CNTs—armchair type with chiral vector (5, 5) and
zigzag type with chiral vector (9, 0)—were selected as rein-
forcing fillers for the nanocomposite. Further, the influence of
temperature, taken in the range of 100 K– 500 K, on the elastic
modulus of the interfacial region and the nanocomposites was
ite were also evaluated using density-temperature curves.
Finally, the results obtained from MD simulations were veri-
fied with ones obtained from continuum-based ROM.
Based on the above results and discussion, the following
important conclusions and findings are highlighted:
& Young’s modulus of nanocomposite and its interfacial re-
gion is sensitive to the temperature variation, i.e., an in-
crease in temperature results in a decrease of Young’s
modulus of the nanocomposite and its interfacial region.
& To utilize the full potential of CNT reinforcement for
yielding a high value of Young’s modulus of the resulting
nanocomposite and its interface at all temperatures, CNTs
of larger chirality (i.e., armchair-type CNT) are to be used.
& CNT reinforcement in the PE matrix enhances the Tg of
the material, and in the temperature range containing the
Tg of the nanocomposite, Young’s modulus of nanocom-
posite and its interface drop suddenly.
References
1. Iijima S (1991) Helical microtubules of graphitic carbon. Nature
354:56–58
2. Lourie O, Wagner HD (1998) Evaluation of Young’s Modulus of
carbon nanotubes by micro-Raman spectroscopy. J Mater Res 13:
2418–2422
3. Salvetat J-P, Bonard J-M, Thomson NH (1999) Mechanical prop-
erties of carbon nanotubes. Appl Phys A Mater Sci Process 69:255–
260
4. Yakobson BI, Avouris P (2001) Mechanical properties of carbon
nanotubes. In: Dresselhaus MS, Dresselhaus G, Avouris Ph (eds)
Carbon nanotubes. Springer, Berlin, pp 287–327
5. Allaoui A, Bai S, Cheng HM, Bai JB (2002) Mechanical and elec-
trical properties of a MWNT/epoxy composite. Compos Sci
Technol 62:1993–1998
6. Ebbesen TW, Lezec HJ, Hiura H et al (1996) Electrical conductivity
of individual carbon nanotubes. Nature 382:54–56
7. Qian D, Dickey EC, Andrews R, Rantell T (2000) Load transfer and
deformation mechanisms in carbon nanotube-polystyrene compos-
ites. Appl Phys Lett 76:2868–2870
8. Thostenson ET, Chou T-W (2002) Aligned multi-walled carbon
nanotube-reinforced composites: processing and mechanical char-
acterization. J Phys D Appl Phys 35:L77
J Mol Model (2018) 24:178
9. Cadek M, Coleman JN, Barron V et al (2002) Morphological and
mechanical properties of carbon-nanotube-reinforced semicrystal-
line and amorphous polymer composites. Appl Phys Lett 81:5123–
5125
10. Gijny FH, Wichmann MHG, Fiedler B, Schulte K (2005) Influence
of different carbon nanotubes on the mechanical properties of ep-
oxy matrix composites—a comparative study. Compos Sci Technol
65:2300–2313
11. Moniruzzaman M, Winey KI (2006) Polymer nanocomposites con-
taining carbon nanotubes. Macromolecules 39:5194–5205
12. Han Y, Elliott J (2007) Molecular dynamics simulations of the
elastic properties of polymer/carbon nanotube composites.
Comput Mater Sci 39:315–323
13. Al-Ostaz A, Pal G, Mantena PR, Cheng A (2008) Molecular dy-
namics simulation of SWCNT-polymer nanocomposite and its con-
stituents. J Mater Sci 43:164–173
14. Khare KS, Khabaz F, Khare R (2014) Effect of carbon nanotube
functionalization on mechanical and thermal properties of cross-
linked epoxy-carbon nanotube nanocomposites: role of strengthen-
ing the interfacial interactions. ACS Appl Mater Interfaces 6:6098–
6110
15. Mandal A, Singh SP, Prasad R (2016) Dynamic mechanical char-
acterization of CNT–PP nanocomposites. J Mol Model 22:1–7
16. Sharma S, Chandra R, Kumar P, Kumar N (2016) Molecular dy-
namics simulation of functionalized SWCNT–polymer composites.
J Compos Mater 21998316628973
17. Rouhi S, Alizadeh Y, Ansari R (2016) On the elastic properties of
single-walled carbon nanotubes/poly (ethylene oxide) nanocom-
posites using molecular dynamics simulations. J Mol Model 22:41
18. Lv Q, Wang Z, Chen S et al (2017) Effects of single adatom and
Stone-Wales defects on the elastic properties of carbon nanotube/
polypropylene composites: a molecular simulation study. Int J
Mech Sci 131–132:527–534
19. Alian AR, Meguid SA (2018) Large-scale atomistic simulations of
CNT-reinforced thermoplastic polymers. Compos Struct 191:221–
230
20. Ozer T, Cabuk S (2018) First-principles study of the structural,
elastic and electronic properties of SbXI (X= S, se, Te) crystals. J
Mol Model 24:66
21. Chawla R, Sharma S (2017) Molecular dynamics simulation of
carbon nanotube pull-out from polyethylene matrix. Compos Sci
Technol 144:169–177
22. Zheng Q, Xue Q, Yan K et al (2008) Influence of chirality on the
interfacial bonding characteristics of carbon nanotube polymer
composites. J Appl Phys 103:44302
23. Yang S, Yu S, Kyoung W et al (2012) Multiscale modeling of size-
dependent elastic properties of carbon nanotube/polymer nanocom-
posites with interfacial imperfections. Polymer 53:623–633
24. Zhang Y, Zhuang X, Muthu J et al (2014) Composites: part B load
transfer of graphene / carbon nanotube / polyethylene hybrid nano-
composite by molecular dynamics simulation. Compos Part B 63:
27–33
25. Arash B, Wang Q, Varadan VK (2014) Mechanical properties of
carbon nanotube/polymer composites. Sci Rep 4:6479
Page 11 of 11 178
26. Srivastava A, Kumar D (2017) A continuum model to study inter-
phase effects on elastic properties of CNT/GS-nanocomposite.
Mater Res Express 4:25036
27. Rafiee R, Mahdavi M (2016) Characterizing nanotube-polymer in-
teraction using molecular dynamics simulation. Comput Mater Sci
112:356–363
28. Wei C, Srivastava D, Cho K (2002) Thermal expansion and diffu-
sion coefficients of carbon nanotube-polymer composites. Nano
Lett 2:647–650
29. Velasco-Santos C, Martínez-Hernández AL, Fisher FT et al (2003)
Improvement of thermal and mechanical properties of carbon nano-
tube composites through chemical functionalization. Chem Mater
15:4470–4475
30. Qi D, Hinkley J, He G (2005) Molecular dynamics simulation of
thermal and mechanical properties of polyimide–carbon-nanotube
composites. Model Simul Mater Sci Eng 13:493–507
31. Li C, Strachan A (2011) Molecular dynamics predictions of thermal
and mechanical properties of thermoset polymer EPON862/
DETDA. Polymer 52:2920–2928
32. Material Studio 8.0. Accelrys Inc, San Diego
33. Plimpton S (1995) Fast parallel algorithms for short-range molecu-
lar dynamics. J Comput Phys 117:1–19
34. Sun H, Mumby SJ, Maple JR, Hagler AT (1994) An ab initio
CFF93 all-atom force field for polycarbonates. J Am Chem Soc
116:2978–2987
35. Fletcher R, Powell MJD (1963) A rapidly convergent descent meth-
od for minimization. Comput J 6:163–168
36. Peacock A (2000) Handbook of polyethylene: structures: proper-
ties, and applications. CRC, Boca Raton
37. Nayebi P, Zaminpayma E (2016) A molecular dynamic simulation
study of mechanical properties of graphene–polythiophene com-
posite with Reax force field. Phys Lett A 380:628–633
38. Frankland SJV, Harik VM, Odegard GM et al (2003) The stress-
strain behavior of polymer-nanotube composites from molecular
dynamics simulation. Compos Sci Technol 63:1655–1661
39. Mahboob M, Zahabul Islam M (2013) Molecular dynamics simu-
lations of defective CNT-polyethylene composite systems. Comput
Mater Sci 79:223–229
40. Hendra PJ, Jobic HP, Holland-Moritz K (1975) Low temperature
crystallization in polyethylene and the value of Tg J Polym Sci Part
C Polym Lett 13:365–368
41 Lam R, Geil PH (1981) Amorphous linear polyethylene: annealing
effects J Macromol Sci Part B Phys 20:37–58
42 Koyama A, Yamamoto T, Fukao K, Miyamoto Y (2001) Molecular
dynamics studies on local ordering in amorphous polyethylene. J
Chem Phys 115:560–566
43. Brown D, Clarke JHR, Okuda M, Yamazaki T (1994) The prepa-
ration of polymer melt samples for computer simulation studies. J
Chem Phys 100:6011–6018
44. Awasthi AP (1989) An atomistic study of the mechanical behavior
of carbon nanotubes and nanocomposite interfaces. J Chem Inf
Model 53:160
45. Zuberi MJS, Esat V (2015) Investigating the mechanical properties
of single walled carbon nanotube reinforced epoxy composite
through finite element modelling. Compos Part B 71:1–9