Articulo 1

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

Physica A 546 (2020) 123995

Contents lists available at ScienceDirect

Physica A
journal homepage: www.elsevier.com/locate/physa

A molecular dynamics simulation of the glass transition


temperature and volumetric thermal expansion coefficient of
thermoset polymer based epoxy nanocomposite reinforced by
CNT: A statistical study

Majid Hadipeykani, Farshid Aghadavoudi, Davood Toghraie
Department of Mechanical Engineering, Khomeinishahr Branch, Islamic Azad University, Khomeinishahr, Iran

article info a b s t r a c t

Article history: A molecular dynamics simulation study is performed to predict the glass transition
Received 3 May 2019 temperature ( Tg ) and the volumetric coefficient of thermal expansion (CTE) of thermoset
Received in revised form 14 August 2019 polymer based nanocomposite reinforced by carbon nanotube (CNT). An atomistic model
Available online 2 January 2020
of cross-linked Diglycidyl ether bisphenol A (DGEBA) epoxy and Diethylenetriamine
Keywords: (DETA) was built as a matrix by employingCondensed-phase optimized molecular
CNT potentials for atomistic simulation (COMPASS27) force field. Different molecular models
Polymer nanocomposites were constructed with various types of CNT embedded in epoxy simulation boxes. Tg was
Glass transition temperature determined based on density variation with temperature. Furthermore, a new method is
Thermal expansion proposed to compute the CTE based on density variation with temperature. The effects
Molecular dynamics of CNT diameter, volume fraction and chirality on Tg and CTE of nanocomposites were
investigated using molecular dynamics simulation. For all cases, studied and CTE were
less than pure epoxy (between 3.77% to 10.05% for Tg and respectively 14.24% to 32.23%
and 23.82% to 41.65% for CTE below and above of Tg ). Increasing the CNT diameter in
nanocomposite increases Tg and CTE (5.0% for Tg and 20.0% for CTE when the diameter of
CNT changed 7.8A0 to 15.6A0 ). On the other hand increasing volume fraction of CNT in
the nanocomposite decreases Tg and CTE (2.7% for Tg and 13.8% for CTE when the volume
fraction of CNT in the nanocomposites changed 3.36% to 5.23%). Chirality studies under
constant weight fraction of nanocomposites show that applying armchair CNT instead
of zigzag CNT, decreases Tg and increases CTE (2.1% for Tg and 5.8% for CTE)
© 2020 Elsevier B.V. All rights reserved.

1. Introduction

Polymer-based nanocomposites, reinforced with carbon nanotubes (CNTs), have been the focus of various studies be-
cause of their wide applications. Nanocomposites are used in modern engineering structures in the extreme temperature
range; therefore, predicting their thermomechanical behavior is very important for designers before construction. Glass
transition temperature (Tg ) is a primary indicator of service temperature or heat resistance of polymeric materials. The
mechanical properties of polymer above Tg are much lower than those below Tg . Coefficient of thermal expansion (CTE)
is one of the other physical properties required by engineers in designing structures at different temperatures. This

∗ Correspondence to: Department of Mechanical Engineering, Islamic Azad University, Khomeinishahr Branch, Khomeinishahr 84175-119, Iran.
E-mail address: Toghraee@iaukhsh.ac.ir (D. Toghraie).

https://doi.org/10.1016/j.physa.2019.123995
0378-4371/© 2020 Elsevier B.V. All rights reserved.
2 M. Hadipeykani, F. Aghadavoudi and D. Toghraie / Physica A 546 (2020) 123995

parameter, like most of the properties of polymers, is influenced by molecular structure of the ingredients. Obviously,
achieving the optimal properties of these materials requires a lot of trial and errors.
Experimental methods for obtaining the properties of polymer-based nanocomposites are time and expense consum-
ing. Therefore, computational approaches are used for modeling and prediction of nanocomposites properties [1,2]. Some
computational methods, such as finite elements, cannot be directly used in the modeling of polymeric nanocomposites
since these methods do not simulate the interatomic forces of polymers and nano reinforcement molecules. Molecular
dynamics (MD) is a time-dependent numerical simulation for modeling the materials, in which the coordination of
the atoms is determined via solving Newton’s second law or the equation of motion. The molecular dynamics method
calculates the atomic forces accurately, so with the help of this method, it is easy to construct and test the new models for
the new materials composition and to estimate their properties. Due to computational limitations, the number of atoms
in this method is much fewer than that of the actual condition of the material, but the results obtained from this method,
based on statistical mechanics, have proven the accuracy of this method in many cases. This method is referred to as the
virtual lab.
Molecular dynamics simulation converts the atomistic level information to microscopic values using statistical me-
chanics. Moreover, in molecular dynamic simulation, integrating Newton’s equations of motion of atoms allows you to
explore the constant-energy surface of a system. Several methods are available for controlling temperature and pressure
in this process. Depending on which state variables are kept constant, different statistical ensembles can be generated. A
variety of properties can then be calculated from the averages or the fluctuations of these quantities over the ensemble [3].
Molecular dynamics simulation and lattice Boltzmann have been applied as alternative approaches to estimate the
physical and mechanical properties of various materials such as fluids, nanofluids, polymers, and nanocomposites [4–12].
Several researchers have used molecular dynamics simulations to predict the thermal and mechanical properties such as
Tg [13,14] and thermal expansion coefficient (CTE) [13] of pure polymers.
Experimental tests show the addition of CNTs can alter important thermomechanical properties of composites, such as
Tg and CTE [15,16]. Based on empirical methods, it can be noticed that in most of the cases SWCNT reinforced composites
showed a decrease of Tg compared to the base resin [17,18]. Ananyo et al. [19] investigated the effect of cross-linking
density on Tg and CTE using molecular dynamics simulation for a pure thermoset polymer. They used EPON862 as epoxy
and DETDA as a hardener and showed with increasing cross-linking density CTE decreases and Tg increases. Herasati
et al. [20] studied the effect on chain morphology and carbon nanotube additives on the Tg and CTE of polyethylene
as a pure polymer employing MD simulation, they reported both the linear and branched semi-crystalline PE have the
same CTE at room temperature and branches significantly change the Tg . They used the specific volume change in the
simulation box for computing the CTE values.They found for the CNT-composites, an addition of only 10% of the long CNT
s can reduce the CTE by 80%. Malagu et al. [21] applied a molecular dynamics simulation to study the effect of CNT on
the Tg of coarse-grained amorphous monodisperse polyethylene as thermoplastic-based nanocomposite, they show Tg of
the polymer matrix increases with the diameter of the CNT while chirality effect is negligible. Allaoui and Bounia [22]
performed a review study on CNT effect on the cure kinetics and Tg of the epoxy composites, they found that various
authors reported CNT effect on thermal properties of nanocomposites while SWNT may lead to a decrease of Tg due to
their bonding tendency, MWNT suggested an increased or unchanged Tg of nanocomposites.
However, it should be noted that all results described above have been on the pure polymer or linear polymer-based
nanocomposites, which were obtained using experimental or MD simulations. However, no previous work, to the best of
the authors’ knowledge, has focused on calculating CTE and Tg of epoxy-based nanocomposite (with epoxy as a complex
cross-linked polymer) using molecular dynamics. Furthermore, the influence of nano inforcement properties on CTE or Tg
has not been thoroughly investigated. The objective of this paper is to investigate the effect of the CNT volume fraction,
size and chirality on Tg and CTE in thermoset base nanocomposite employing atomistic models. Moreover, a new method
for computing CTE was developed based on molecular dynamics results. In this method, the CTE is directly, without the
need for simulation box details, derived from the density values of molecular models at various temperatures using the
line fitting. The simulation results obtained from nanocomposite models suggest that a small change in CNT radius, volume
fraction and chirality cause relatively large changes in the CTE and Tg . Moreover, the CTE values below and above the Tg
were significantly different for all molecular models. The results of this paper for pure epoxy are generally consistent with
experimental and previous modeling.

2. Materials and method

2.1. The molecular structure of resin and hardener

The molecular structure of the DGEBA epoxy resin with the chemical formula C21 H24 O4 and the DETA hardener with
chemical formula C4 H13 N3 is illustrated in Fig. 1. These epoxies have particular importance as a family of thermosetting
polymers with vast applications in the packaging, electronics, and aerospace industries. In a mixture of epoxy resin
monomer and the hardener, the molecules will move and thus, reactive sites will have the chance to get closer together
to establish a bond between the carbon atom of the DGEBA molecule and nitrogen atoms of the DETA molecule. Usually,
the resin and hardener molecules are mixed in a ratio of 1:2. This ratio will determine the number of reactive sites in the
simulation study.
M. Hadipeykani, F. Aghadavoudi and D. Toghraie / Physica A 546 (2020) 123995 3

Fig. 1. Chemical structure of (a) DGEBA and (b) DETA [23].

One of the important parameters in simulating of the polymerization process is reaction cut-off distance. The cut-off
distance is the distance that reactive sites will be able to react only within that distance. This factor plays an important
role in the degree of polymerization and polymer structures. Selecting a short cut-off distance yields a low degree of
polymerization. On the other hand, selecting a large cut-off distance leads to the formation of tense structures that will
be difficult and time-consuming to reach equilibrium. In studies that researchers have done on polymerization reactions,
carbon and nitrogen atoms for creating of covalent bond stay in the field of gravity each other from a distance about
4 times of the equilibrium bond length C-N [24]. In this study, cut off distance was considered 6 Å and polymerization
method in [25] was used in this study.

2.2. Force field and software

In this study, in the process of modeling and simulation, the Materials Studio software and the COMPASS27force
field [26,27] have been used. COMPASS as the ab-into force field is able to predict thermal and mechanical properties of
polymers such as epoxy resins [28], CNT [17], Polymer/CNT nanocomposites [29]. The COMPASS forcefield has extensive
coverage in covalent molecules including most common organic and small inorganic molecules. Moreover, COMPASS
forcefield is capable of being used in the consolidated coverage of organic and inorganic materials models. For these
molecular systems, the COMPASS forcefield has been parameterized to predict different properties for molecules in
condensed phases. The values of these parameters are included in the molecular dynamics software databases such as
Materials Studio. Intermittently, the parameter set of COMPASS forcefield for calculating the atomistic forces is revised
using updating or addition of new functional groups; consequently, various versions of COMPASS are retained and made
available for the purposes of validations and investigation.
The used energy potential function Et in COMPASS, includes the valence energy, cross-coupling energy and non-bonded
energy as belows [23]:
Etotal = Evalence + Ecross + Enon-bond (1)
Evalence = Ebond + Eangle + Etorsion + Eout-of-plane (2)

The terms of energy in Eqs. (1) and (2) are:


∑[
K2 (b − b0 )2 + K3 (b − b0 )3 + K4 (b − b0 )4
]
Ebond = (3a)
b

∑[
H2 (θ − θ0 )2 + H3 (θ − θ0 )3 + H4 (θ − θ0 )4
]
Eangle = (3b)
θ
∑[
V1 1 − cos(∅ − ∅01 ) + V2 1 − cos(2∅ − ∅02 ) + V3 1 − cos(3∅ − ∅03 )
{ } { } { }]
Etorsion = (3c)


Eout-of-plane = Kχ χ 2 (3d)
χ

∅Vi , Hi , Ki , ∅i are system dependent parameters that obtained from ab-into quantum mechanics computations [23]. The
terms of b − b0 , θ − θ0 and ∅ − ∅01 shows the length and angle between the two neighbors’ bond from the equilibrium
4 M. Hadipeykani, F. Aghadavoudi and D. Toghraie / Physica A 546 (2020) 123995

Fig. 2. Cross-linked epoxy resin DGEBA/DETA [23].

state θ , ∅ and χ are angles of bond, dihedraltorsion, out of plane respectively. The ECross includes several terms that have
been described elsewhere [30].
The Enon-bond is related to Coulomb and Vander Waals potentials. The latter is expressed by the following equations.
Enon-bond = EVander Waals + ECoulomb (4a)

⎡ ( )9 ( )6 ⎤
∑ r0ij r0ij
EVander Waals = D0 ⎣2 −3 ⎦ (4b)
rij rij
i>j

∑ qi qj
ECoulomb = (4c)
ε rij
i>j

D0 , q and ε are Lenard-Jones potential, atomic electrical load and dielectric constant respectively. In this study, in order to
calculate Van der Waals and Coulomb potentials, the atom based summation with the cut-off distance of 10 Å and Evald
collection method with the accuracy of 10−4 kcal/molwere used respectively.

2.3. The creation process of cross-linking

Cross-linking is the formation of a covalent bond when epoxy molecules react with the hardener agent molecule. The
structure of DGEBA and DETA cross-linked polymer is shown in Fig. 2.
In the process of the reaction, the C-O bond within the epoxide group breaks, and the open end carbon of the epoxy
group reacts with the nitrogen of the amino group in the curing agent molecule. A number of other cross-linking processes
have been proposed by researchers. The pure resin and hardener agent are used in cross-linking simulation [18,31]. In
this research, a dynamic approach is used to simulate the cross-linking process. The process involves a minimization
and an equilibrium cycle and bonding. In the equilibrium step, the ensemble of NVT is used on the molecular model at
the temperature of 600 K with 100 ps simulation time and 1 fs step time. In order to create more space between the
molecules, the initial density of the resin was considered 0.5 g/cm3 . After that NPT ensemble was applied on the model
for 50 ps with 1 fs step time at the atmospheric pressure. Once the system is reached equilibrium, the distances between
the reactive sites were measured pair-wise, and a covalent bond was created between the appropriate sites that were
within the range of the 6 Å to 10 Å distance. For creating cross-linking, the length of the C-N bond was investigated in
four replicates [19]. Eventually, after repeating the cycle, the degree of polymerization was determined to be 50%.

2.4. Simulation details of nanocomposites

In order to reduce the simulation time and number of calculations, one representative molecule was selected from
the cured epoxy model to construct the samples. According to previews studies the selected molecular model has the
same properties as the amount of cross-linked density [9]. In previous studies, the physical and mechanical properties of
M. Hadipeykani, F. Aghadavoudi and D. Toghraie / Physica A 546 (2020) 123995 5

cross-linked epoxy resin (DGEBA + DETA) have been investigated [23,32]. In order to study the nanotube carbon effect
on the Tg and CTE of epoxy nanocomposites, simulation models were made in different volume fractions and weight
fractions, with the same cross-linking density of 50%. Molecular models of CNT (5,0), CNT (10,0), CNT (15,0), CNT (20,0),
CNT (10,10) and CNT(15,15) were made and amorphous cell packing module was used in the Materials Studio software
for the purpose of adding and constructing polymer matrix around the CNT models. At this step, the mass density was
set to 0.5 g/cm3 and the temperature was adjusted to 300 K. In the packing process, the smart algorithm was used for
geometry optimization. Before calculating the parameters such as density for the atomistic model, molecular dynamics
process should be conducted using appropriate ensembles. An ensemble, which is a group of possible microstates of a
system with different microscopic states, specifies a unique macroscopic state. In statistical mechanics, the average of
parameter A is defined in terms of ensemble averages as follows [3]:
∫∫
⟨A⟩ensemble = dpN drN A(PN , rN )ρ (PN , rN ) (5)

In this equation, P is the momentum and r is the position of all N atoms and the integration is calculated based on all
possible variables of r and p where A(PN , rN ) is the observable of interest and ρ (PN , rN ) is the probability density of the
ensemble.
The dynamic and optimization process for each model was performed in three steps as described below:

(a) Geometric optimization and energy minimization by using the Smart algorithm and using the Newton Gaussian
method in order to achieve minimum convergence with a tolerance of 10−4 kcal/mol.
(b) NVT ensemble was used for 100 ps with the time step of 1 fs on each of the models at 600 K for the purpose of
giving sufficient kinetic energy to the system and release the system from internal stresses in the structure of the
matrix. Also, the use of NVT increases the probability of molecular placement to the closet state of the optimal
position. The initial mass density for each model was set to 0.5 g/cm3 .
(c) In order to achieve the final mass density and closet state of local equilibrium, NPT ensemble was used for 100 ps
with the time step of 1 fs at the temperature of 300 K and pressure of 1 atm.
Brendsen and Barostat thermostat were used to control the pressure during the equilibrium process. In all the
simulation boxes, periodic boundary conditions were used to remove surface effects. In order to calculate the
average properties of the molecular structure in both ensembles (NVT, NPT) the frame of the atomic configuration
was stored as the file’s trajectory.

2.5. Glass transition temperature (Tg )

The glass transition temperature (Tg ) is the temperature which the state of thermoset polymers changes from hard
and brittle to the soft and rubbery state. Differential scanning colorimetry (DSC), differential thermal analysis (DTA) and
thermal mechanical analysis (TMA) are the most important experimental methods to measure Tg . In order to calculate the
Tg in molecular dynamics simulation, NPT ensemble is applied to the system at various temperatures, and density values
of the system are obtained at each temperature individually. Decreasing the temperature will cause a uniform increase in
density. At the Tg Temperature, a clear gradient change is seen in the density line which interpolates the density values.

2.6. The coefficient of thermal expansion (CTE)

The coefficient of thermal expansion (CTE) is one of the most important thermal properties of nanocomposites defined
as follows:
∂V
( )
1
β= (6)
V ∂T P

where β is the CTE and V, T and P are the volume, temperature and pressure of system, respectively. Based on density
definition the above equation can be represented using the below equation:
∂ρ
( )
−1
β (T) = (7)
ρ ∂T P

In this study, a new method has been applied to determine the CTE, due to the linear relationship between temperature
and density of polymers and polymer nanocomposites, the density can be written as follows.

ρ = aT + ρ0 (8)

where a and ρ0 are constant parameters. By replacement Eq. (6) in the Eq. (7), the β (T) can be obtained as follow.
−a
β (T) = (9)
ρ0 + aT
6 M. Hadipeykani, F. Aghadavoudi and D. Toghraie / Physica A 546 (2020) 123995

Table 1
Details of case studies.
Model Number of molecules in Dimension of simulation Weight fraction Volume fraction
no. structure box (Å) %WCNT %VCNT
1 40DGEBA+20DETA 29.59 × 29.59 × 29.59 0 0
2 50DGEBA+25DETA+CNT(10,0) 34.6 × 34.6 × 34.08 16.41 7.40
3 116DGEBA+58DETA+CNT(15,0) 47.4 × 47.4 × 34.08 11.26 7.40
4 176DGEBA+88DETA+CNT(20,0) 60.1 × 60.1 × 34.08 10.04 7.40
5 30DGEBA+15DETA+CNT(5,0) 26 × 26 × 34.08 15.4 5.23
6 38DGEBA+19DETA+CNT(5,0) 28.85 × 28.85 × 34.08 11.44 4.35
7 50DGEBA+25DETA+CNT(5,0) 32.4 × 32.4 × 34.08 9.75 3.36
8 50DGEBA+25DETA+CNT(10,10) 44.5 × 44.5 × 19.68 16.41 10.61
9 116DGEBA+58DETA+CNT(15,15) 65.5 × 65.5 × 19.68 11.26 9.80

By using the definition of function average in calculus between temperatures T1 and T2 , β (T) as the CTE average can
be calculated by the following equation:
∫ T2
T1
β (T) dT
β (T) = (10)
T2 − T1
By replacing Eq. (8) in the Eq. (9) β (T) can be calculated as:
∫ T2
1 −a0
β (T) = dT (11)
T2 − T1 T1 ρ0 + aT
Base on density variation by using Eq. (7) two points are considered as (T1 , ρ1 ), (T2 , ρ2 ):

ρ1 = ρ0 + aT1
ρ2 = ρ0 + aT2
CTE average is then calculated by two parameters of density and temperature as follows:
ρ1
( )
1
β (T) = ln (12)
T2 − T1 ρ2
As mentioned in the previous section, Tg is the intersection of two lines of density vs. temperature plot. Therefore
according to eq.11, two values for β (T) can be determined by using the following equations,
ρ (Ta )
) (
1
Ta < Tg : β g = ln (13)
Tg − Ta ρ (T )
( g )
1 ρ (Tg )
Tb > Tg : β R = ln (14)
Tb − Tg ρ (Tb )
where β g and β R are volumetric coefficient thermal expansion at the temperatures below of the Tg in the glassy state
and above the Tg in the rubbery state of polymer and nanocomposites respectively. Ta and Tb are arbitrary temperatures
below of the Tg and above the Tg respectively.

3. Case studies

In order to investigate the influencing factors on thermal properties of the nanocomposites nine atomistic models were
constructed. Table 1 shows the details of these cases studies. As can be seen in the table, case 1 has been constructed
for modeling the neat epoxy resin and other cases were used for atomistic modeling of nanocomposites. In case 2, 3 and
4 various CNT including CNT(10,0), CNT(15,0) and CNT(20,0) are embedded in the epoxy resin molecules with constant
volume fraction. Case studies 5, 6 and 7 have the same CNT of (5, 0) as the reinforcement used in constructing molecular
models with different volume fraction. Finally, in cases 8 and 9 CNT (10, 10) and CNT (15, 15) were used respectively.
3D molecular models of cross-linked epoxy (Case 1) and CNT reinforced epoxy (Case 3) are presented as samples of
epoxy and nanocomposite in Fig. 3. In nanocomposites, CNT is located in the center of model and epoxy molecules are
around them, as shown in Fig. 3b.
In all atomistic models, the CNT volume fraction is calculated using Eq. (15):
heq 2
π (RCNT + 2
)
VCNT = (15)
Acell
where RCNT is the radius of CNT, heq is the equilibrium Van der Waals separation distance between epoxy and CNT
molecules and was considered to be 2.8 Å [23], and Acell is the cross section area of simulation box for each model.
M. Hadipeykani, F. Aghadavoudi and D. Toghraie / Physica A 546 (2020) 123995 7

Fig. 3. Molecular model of the cross-linked Epoxy (a) and the CNT reinforced composite (b).

Fig. 4. Density convergence pattern of nanocomposite molecular model [23].

The MD simulations and equilibrium process were performed for each of the case studies at fourteen different
temperature changes between 250 K to 550 K. The density of each molecular model at a specific temperature determined
after the implementation of NPT ensemble in 100 ps. One of the most important conditions for equilibrium in the MD
simulation is the convergence of density during the simulation time; this condition was investigated for each case. A
density convergence pattern of nanocomposite in ref. [23] is shown in Fig. 4. In this study, all samples’ density converged
with similar trend.
The variation of density with temperature for all cases studies have been investigated to find out the Tg and CTE. The
results are presented in the next section.

4. Results and discussion

4.1. Tg and thermal expansion of the pure epoxy resin

The variation of density with temperature for case study.1, pure epoxy resin is shown in Fig. 5. The mass density is
reduced from 1.21 g/cm3 at 250 K to 1.072 g/cm3 at 550 K. As can be seen in the diagram, the density reduced about
(11.4%) from 250 K to 550 K. Using the linear fitting for density values temperature, two lines with different slopes are
plotted as illustrated in Fig. 5, The intersection of two these lines determines the Tg . For pure epoxy resin, the Tg was
obtained about 398 K.
The volumetric coefficient of thermal expansion (CTE) can be calculated using 12, 13 equations. Based on the density
results, CTE for pure epoxy resin was determined about 21.69×10−5 (1/K) below of the Tg and 55.53×10−5 (1/K) above of
Tg . The obtained results for Tg and CTE of pure resin epoxy were compared with the MD and experimental results of the
8 M. Hadipeykani, F. Aghadavoudi and D. Toghraie / Physica A 546 (2020) 123995

Fig. 5. Density as a function of temperature for pure epoxy.

Table 2
Comparison of the obtained results with those of other studies.
Reference Method Material Cross-linking Tg (K) CTE × 10−5 (1/K) CTE × 10−5 (1/K)
% Below of Tg Above of Tg
Current study MD DGEBA+DETA 50 398 21.69 55.53
(33) Exp. DGEBA+DETA 54 408 ± 3.1 21.27 54.21
(24) MD DGEBA+DETDA 54 410–420 – –
(23) MD DGEBA+DETA 54 390 – –

Table 3
The simulated values of the cases 2, 3 and 4.
Model ρ (g/cm3 ) ρ (g/cm3 ) ρ (g/cm3 ) Tg (K) βg (1/K) βR (1/K)
@250K @Tg @550K
2 1.215 1.195 1.111 365 14.7 × 10−5 32.4 × 10−5
3 1.128 1.192 1.112 376 17.1 × 10−5 39.9 × 10−5
4 1.219 1.191 1.110 383 17.7 × 10−5 42.0 × 10−5

nearest references in Table 2. The good agreement was observed between current results and other values which were
reported in related references. As an example, about 2% differences for Tg and 2.5% for CTE were observed between current
study molecular predictions and experimental results [33]. It is noteworthy that all of the materials in Table 2 have similar
base resin but in references [23,24] DREIDING and classical COMPASS were used as force field respectively. But in this
study COMPASS27 have been used which have shown a higher convergence rate in the calculation of polymer density.
COMPASS forcefield has more atomic groups, including sulfate and sulfonate groups, than COMPASS27 [34]. Therefore,
due to the lower number of parameters, the use of COMPASS27 forcefield reduces the computational cost. Consequently,
when COMPASS27 instead of COMPASS is used in this study, the duration of the molecular dynamics calculations by the
software decreases and the convergence rate increases.

4.2. The effect of CNT diameter on Tg and CTE of nanocomposite

In this study, the Tg and CTE for nanocomposites were investigated using molecular simulations. The variation of Tg
and CTE for specified molecular models with a constant volume fraction of CNT and various types of CNT radius were
studied. For this reason, as described before, case studies of 2, 3 and 4 are subjected to MD simulation and equilibrium
process. Employing the curve fitting on density versus temperature diagram and based on Eqs. (13) and (14), CTE was
obtained for all cases. Table 3 shows the obtained values for Tg and CTE.
The variation of Tg with CNT diameter is shown in Fig. 6. It may be observed from Fig. 6 that Tg increases with CNT
diameter. It can be noted that for all nanocomposite cases 2, 3, 4 the Tg values are smaller than Tg of pure epoxy resin.
The weakness of the atomic interactions between CNT and epoxy molecules has caused the rubbery state in nanocom-
posite to occur at lower temperatures than pure epoxy resin. CNT diameter affects the thermal composite properties such
as CTE as can be observed from Fig. 7 in which the CTE increases with the rise in CNT diameter for both cases at the
below and above Tg temperature. Furthermore, for all of the nanocomposite molecular models, the CTE of the glassy state
M. Hadipeykani, F. Aghadavoudi and D. Toghraie / Physica A 546 (2020) 123995 9

Fig. 6. Effect of variation of CNT diameter on glass transition temperature of nanocomposites.

Fig. 7. Effect of variation of CNT diameter on the coefficient of thermal expansion of nanocomposites.

is determined about two times lower than that of the CTE of rubbery state. This is due to the fact that in the glassy state
at temperatures below Tg , the molecules have limited movement comparing to the rubbery state. Therefore, in the glassy
state, the molecules could expand less than those in the rubbery state. In comparison with CTE values of nanocomposites
and epoxy, it can be observed that the CTE significantly reduced in nanocomposites. This reduction can be attributed to
the enhanced stiffness of the CNT/matrix region caused by the increased interaction between the CNT and the matrix
restricting the expansion of polymer chains.

4.3. The effect of CNT volume fraction on Tg and CTE of nanocomposites

The volume fraction of reinforcement influences the density versus temperatures of the Tg that can be determined for
all cases. Furthermore, CTE values are determined by using Eqs. (13) and (14). Fig. 8 shows the variation of the volume
fraction of CNT on the Tg temperatures. As can be seen in the diagram, low volume fraction cases have higher Tg . This is
due to the fact that increasing the interactions and decreasing the covalent bond may cause a reduction in the Tg . Similar
to the previous results, Tg of all nanocomposite molecular models is less than that for pure epoxy.
The effect of CNT volume fraction on the CTE has been investigated in Fig. 9. It can be observed from Figure that
increasing the CNT volume fraction may cause a reduction in the coefficient of thermal expansion. Moreover, the CTEs
for the rubbery state, where the temperature of nanocomposites at above the Tg , are 2 to 2.5 times higher than those for
the glassy state.
10 M. Hadipeykani, F. Aghadavoudi and D. Toghraie / Physica A 546 (2020) 123995

Fig. 8. Effect of variation of CNT volume fraction on the glass transition temperature of nanocomposites.

Fig. 9. Effect of variation of CNT volume fraction on the coefficient of thermal expansion of nanocomposite.

4.4. The effect of chirality on Tg and CTE of nanocomposites

CNT chirality is an impressive parameter that significantly influences the thermomechanical properties of nanocompos-
ites. In order to study the effect of chirality on glass transition temperature (Tg ) and the coefficient of thermal expansion
(CTE), the cases of 8 and 9 in Table 1 have been considered. Armchair nanotubes as reinforcement were used in these
models. By performing a linear fit on the graph of density with temperature on these models, the Tg was obtained for
each of them. Due to the repetition of diagram shapes, those were not presented here. By using Eqs. (12) and (13), CTE
below and above the Tg was obtained for each model. For the purpose of investigating the chirality effect, these results
should be compared with those of cases 2 and 3. In Fig. 10, the Tg bar graph is presented into zigzag and armchair CNTs.
In this diagram, the Tg of 2 & 8 cases, and 3 & 9 cases have respectively come along with each other.
In this graph, the size effect of carbon nanotube and chirality effect on Tg are observable simultaneously. Diameters of
CNT (10, 10), CNT (15, 15) are respectively 13.56 Å and 19.68 Å. It is shown in Fig. 10 that with increasing in armchair-CNT
diameter, Tg increases from 358 K to 368 K. It can be concluded that with increasing in CNT-armchair diameter in the
constant weight fraction of carbon nanotube as reinforcement in polymer nanocomposites, the glass transition increases.
But in any cases, these values are less than Tg of pure polymer. Fig. 10 shows that by using CNT-armchair instead of
CNT-zigzag in nanocomposites, glass transition temperature decreases. In Fig. 11, the coefficient of the thermal expansion
bar graph is presented in different CNT types, armchair and zigzag.
In this graph, the size effect of carbon nanotube and chirality effect on the coefficient of thermal expansion are
observable simultaneously. The CTEs are shown below and above Tg for each case. In Fig. 11, it is shown that with an
increase in armchair-CNT diameter, the coefficient of thermal expansion increases. It can be concluded that if the weight
M. Hadipeykani, F. Aghadavoudi and D. Toghraie / Physica A 546 (2020) 123995 11

Fig. 10. Effect of chirality on the glass transition temperature of nanocomposites.

Fig. 11. Effect of chirality on the coefficient of thermal expansion of nanocomposites.

fraction of carbon nanotube remains constant in polymer nanocomposites, by increasing the CNT-armchair diameter,
the coefficient of thermal expansion will increase. Fig. 11 shows that by using CNT-armchair instead of CNT-zigzag in
nanocomposites, the coefficient of thermal expansion has been increased.

5. Conclusion

Through atomistic modeling base on molecular dynamics simulation and using compass 27 force field, the effects of
CNT diameter, CNT volume fraction and chirality for CNT reinforced polymeric nanocomposite have been studied on Tg
and CTE. The epoxy molecules as a matrix have been modeled by considering the cross-link bonds. Also, an equation
was proposed for determining the CTE based on the density and temperature of polymer and nanocomposites. From the
molecular simulation results, it was observed that Tg and CTE of nanocomposites are less than those of pure epoxy. The
conclusions regarding size, volume fraction and chirality effects lead to several summaries as below:

• MD simulations applying the COMPASS27 force field presents Tg and CTE predictions in good agreement with the
previous studies. In this study, this subject was investigated for pure epoxy.
• Size effect: The Tg and CTE of the nanocomposites increase with a rise in the diameter of the nanotube while the
volume fraction of CNT was constant. The results show an increase of about 5% in Tg and 20% in CTE while the
diameter of CNT in nanocomposites increases from 7.8 Å to 15.6 Å, and the volume fraction of the samples is equal
to 7.4%.
12 M. Hadipeykani, F. Aghadavoudi and D. Toghraie / Physica A 546 (2020) 123995

• Volume fraction effect: the Tg and CTE of the nanocomposites decrease with a reduction in the volume fraction of
CNT in nanocomposites while the diameter and type of CNT are constant. The Tg and CTE can respectively reduce
by 2.7% and 13.8% while the volume fraction decreases from 5.23% to 3.36%.
• Chirality effect: by using armchair CNT instead of zigzag CNT in nanocomposites, the Tg of nanocomposites decreases
and the CTE increases, while the weight fraction of CNT is constant. We obtain 2.1% reducing in Tg and 5.8% increasing
in CTE when the weight fraction of CNT is 16.41% and CNT (10, 0) relocates with CNT (10, 10). In other cases, by
using CNT (15, 15) instead of CNT (15, 0) while the weight fraction of CNT is 11.26%, the drop in Tg and the rise in
CTE are respectively 2.2% and 5.4%.

Finally, the proposed approach and equation for predicting the thermal properties of epoxy-based nanocomposites are
generally appropriate to any thermoset polymeric matrix molecular models. The quantitative predictions of constructed
atomistic models can be used to optimize the properties of CNT polymeric nanocomposites as a class of new materials.
The extension of this paper for MDS according to previous works [35–44] affords engineers a good option for nanoscale
and statistical simulation
Abbreviations

1. MD Molecular dynamics
2. CNT Carbon nanotube
3. DGEBA Diglycidyl ether bisphenol A
4. DETA Diethylenetriamine
5. TgGlass transition temperature
6. CTE Coefficient of thermal expansion
7. COMPASS Condensed-phase optimized molecular potentials for atomistic simulation studies

References

[1] M. Quaresimin, M. Salviato, M. Zappalorto, Strategies for the assessment of nanocomposite mechanical properties, Composites B 43 (5) (2012)
2290–2297.
[2] E.T. Thostenson, C. Li, T.W. Chou, Nanocomposites in context, Compos. Sci. Technol. 65 (3–4) (2005) 491–516.
[3] D. Chandler, Introduction to Modern Statistical Mechanics, by David Chandler, vol. 288, Oxford University Press, 1987, ISBN-10: 0195042778,
ISBN-13: 9780195042771.
[4] P. Alipour, D. Toghraie, A. Karimipour, M. Hajian, Modeling different structures in perturbed poiseuille flow in a nanochannel by using of
molecular dynamics simulation: Study the equilibrium, Physica A 515 (2019) 13–30.
[5] F. Aghadavoudi, H. Golestanian, K.A. Zarasvand, Elastic behaviour of hybrid cross-linked epoxy-based nanocomposite reinforced with GNP and
CNT: experimental and multiscalemodelling, Polym. Bull. (2018) 1–20.
[6] J.L. Tack, D.M. Ford, Thermodynamic and mechanical properties of epoxy resin DGEBF crosslinked with DETDA by molecular dynamics, J. Mol.
Graph. Model. 26 (8) (2008) 1269–1275.
[7] P. Alipour, D. Toghraie, A. Karimipour, M. Hajian, Molecular dynamics simulation of fluid flow passing through a nanochannel: Effects of
geometric shape of roughnesses, J. Molecular Liquids 275 (2019) 192–203.
[8] A. Moshfegh, H.H. Afrouzi, M. Ahmadian, D. Toghraie, A. Javadzadegan, Statistical analysis of pulsating non-newtonian flow in a corrugated
channel using Lattice-Boltzmann method, Physica A 535 (2019) 122486.
[9] A. Afshari, M. Akbari, D. Toghraie, M.E. Yazdi, Experimental investigation of rheological behavior of the hybrid nanofluid of
MWCNT-alumina/water (80%)-ethylene-glycol (20%), J. Therm. Anal. Calorim. 132 (2) (2018) 1001–1015.
[10] B. Karbasifar, M. Akbari, D. Toghraie, Mixed convection of Water-Aluminum oxide nanofluid in an inclined lid-driven cavity containing a hot
elliptical centric cylinder, Int. J. Heat Mass Transf. 116 (2018) 1237–1249.
[11] B. Ruhani, P. Barnoon, D. Toghraie, Statistical investigation for developing a new model for rheological behavior of Silica-ethylene glycol/Water
hybrid newtonian nanofluid using experimental data, Physica A 525 (2019) 616–627.
[12] A. Javadzadegan, M. Joshaghani, A. Moshfegh, O.A. Akbari, H.H. Afrouzi, D. Toghraie, Accurate meso-scale simulation of mixed convective heat
transfer in a porous media for a vented square with hot elliptic obstacle: An LBM approach, Physica A 537 (2020) 122439.
[13] H.B. Fan, M.M. Yuen, Material properties of the cross-linked epoxy resin compound predicted by molecular dynamics simulation, Polymer 48
(7) (2007) 2174–2178.
[14] B. Arab, A. Shokuhfar, S. Ebrahimi-Nejad, Glass transition temperature of cross-linked epoxy polymers: a molecular dynamics study, in:
Proceedings of the International Conference Nanomaterials: Applications and Properties (No. 1, (1) 01NDLCN11-01NDLCN11), Sumy State
University Publishing, 2012.
[15] F.H. Gojny, M.H. Wichmann, B. Fiedler, W. Bauhofer, K. Schulte, Influence of nano-modification on the mechanical and electrical properties of
conventional fibre-reinforced composites, Composites A 36 (11) (2005) 1525–1535.
[16] L. Ci, J. Bai, The reinforcement role of carbon nanotubes in epoxy composites with different matrix stiffness, Compos. Sci. Technol. 66 (3–4)
(2006) 599–603.
[17] A. Hernández-Pérez, F. Avilés, A. May-Pat, A. Valadez-González, P.J. Herrera-Franco, P. Bartolo-Pérez, Effective properties of multiwalled carbon
nanotube/epoxy composites using two different tubes, Compos. Sci. Technol. 68 (6) (2008) 1422–1431.
[18] J. Shen, W. Huang, L. Wu, Y. Hu, M. Ye, Thermo-physical properties of epoxy nanocomposites reinforced with amino-functionalized multi-walled
carbon nanotubes, Composites A 38 (5) (2007) 1331–1336.
[19] A. Bandyopadhyay, P.K. Valavala, T.C. Clancy, K.E. Wise, G.M. Odegard, Molecular modeling of crosslinked epoxy polymers: The effect of crosslink
density on thermomechanical properties, Polymer 52 (11) (2011) 2445–2452.
[20] S. Herasati, H.H. Ruan, L.C. Zhang, Effect of chain morphology and carbon-nanotube additives on the glass transition temperature of polyethylene,
J. Nano Res. 23 (2013) 16–23.
[21] M. Malagù, A. Lyulin, E. Benvenuti, A. Simone, A molecular-dynamics study of size and chirality effects on glass transition temperature and
ordering in carbon nanotube-polymer composites, Macromol. Theory Simul. 25 (6) (2016) 571–581.
M. Hadipeykani, F. Aghadavoudi and D. Toghraie / Physica A 546 (2020) 123995 13

[22] A. Allaoui, N.E. El Bounia, How carbon nanotubes affect the cure kinetics and glass transition temperature of their epoxy composites?–a review,
Express Polymer Lett. 3 (9) (2009) 588–594.
[23] F. Aghadavoudi, H. Golestanian, Y. TadiBeni, Investigating the effects of resin crosslinking ratio on mechanical properties of epoxy-based
nanocomposites using molecular dynamics, Polym. Compos. 38 (2017) E433–E442.
[24] C. Li, A. Strachan, Molecular simulations of crosslinking process of thermosetting polymers, Polymer 51 (25) (2010) 6058–6070.
[25] A. Shokuhfar, B. Arab, The effect of cross linking density on the mechanical properties and structure of the epoxy polymers: molecular dynamics
simulation, J. Mol. Model. 19 (9) (2013) 3719–3731.
[26] H. Sun, P. Ren, J.R. Fried, The COMPASS force field: parameterization and validation for phosphazenes, Comput. Theor. Polym. Sci. 8 (1–2)
(1998) 229–246.
[27] M.J. McQuaid, H. Sun, D. Rigby, Development and validation of COMPASS force field parameters for molecules with aliphatic azide chains, J.
Comput. Chem. 25 (1) (2004) 61–71.
[28] W.H. Duan, Q. Wang, K.M. Liew, X.Q. He, Molecular mechanics modeling of carbon nanotube fracture, Carbon 45 (9) (2007) 1769–1776.
[29] B. Arash, Q. Wang, V.K. Varadan, Mechanical properties of carbon nanotube/polymer composites, Sci. Rep. 4 (6479) (2014).
[30] M. Mahboob, M.Z. Islam, Molecular dynamics simulations of defective CNT-polyethylene composite systems, Comput. Mater. Sci. 79 (2013)
223–229.
[31] S. Yang, J. Qu, Computing thermomechanical properties of crosslinked epoxy by molecular dynamic simulations, Polymer 53 (21) (2012)
4806–4817.
[32] E.N. Brown, S.R. White, N.R. Sottos, Fatigue crack propagation in microcapsule-toughened epoxy, J. Mater. Sci. 41 (19) (2006) 6266–6273.
[33] X. Fernández-Francos, D. Santiago, F. Ferrando, X. Ramis, J.M. Salla, A. Serra, M. Sangermano, Network structure and thermomechanical properties
of hybrid DGEBA networks cured with 1-methylimidazole and hyperbranched poly (ethyleneimine) s, J. Polymer Sci. Part B: Polymer Phys. 50
(21) (2012) 1489–1503.
[34] Materials Studio Help, Accelrys Inc., CA, 2011.
[35] D.T. Semiromi, A. Azimian, Nanoscale Poiseuille flow and effects of modified Lennard–Jones potential function, Heat Mass Transf. 46 (7) (2010)
791–801.
[36] H. Noorian, D. Toghraie, A.R. Azimian, Molecular dynamics simulation of Poiseuille flow in a rough nano channel with checker surface
roughnesses geometry, Heat Mass Transf. 50 (1) (2014) 105–113.
[37] D. Toghraie, A.R. Azimian, Molecular dynamics simulation of nonodroplets with the modified Lennard-Jones potential function, Heat Mass
Transf. 47 (5) (2011) 579–588.
[38] M.M.A. Pourshirazi, D. Toghraie, A.R. Azimian, Study effect of deformation nanochannel wall roughness on the water-copper nano-fluids
Poiseuille flow behavior, J. Simul. Anal. Novel Technol. Mech. Eng. 8 (1) (2015) 29–40.
[39] M. Rezaei, A.R. Azimian, D. Toghraie, Molecular dynamic simulation of copper and platinum nanoparticles poiseuille flow in a nanochannels,
Physica A 84 (2016) 152–161.
[40] M. Rezaei, A.R. Azimian, D. Toghraie, The surface charge density effect on the electro-osmotic flow in a nanochannel: a molecular dynamics
study, Heat Mass Transf. 51 (5) (2015) 661–670.
[41] Masoumeh Tohidi, Davood Toghraie, The effect of geometrical parameters, roughness and the number of nanoparticles on the self-diffusion
coefficient in Couette flow in a nanochannel by using of molecular dynamics simulation, Physica B 518 (2017) 20–32.
[42] H. Noorian, D. Toghraie, A.R. Azimian, The effects of surface roughness geometry of flow undergoing poiseuille flow by molecular dynamics
simulation, Heat Mass Transfer 50 (2014) 95–104.
[43] D. Toghraie, A.R. Azimian, Molecular dynamics simulation of annular flow boiling with the modified Lennard-Jones potential function, Heat
Mass Transf. 48 (1) (2012) 141–152.
[44] D. Toghraie, A.R. Azimian, Molecular dynamics simulation of liquid–vapor phase equilibrium by using the modified Lennard-Jones potential
function, Heat Mass Transf. 46 (3) (2010) 287–294.

You might also like