Computational Materials Science 96 (2015) 146–158

Contents lists available at ScienceDirect

Computational Materials Science

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

Evaluation of mechanical characteristics of nano modified epoxy based

polymers using molecular dynamics
B.S. Sindu ⇑, Saptarshi Sasmal
Academy of Scientific and Innovative Research (AcSIR), CSIR Campus, Taramani, Chennai 600113, India
CSIR-Structural Engineering Research Centre, CSIR Campus, Taramani, Chennai 600113, India

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

Article history: Epoxy polymers are finding large scale application in construction industry for repair and retrofitting of
Received 7 February 2014 deficient concrete structures, because of their great compatibility with construction materials and lam-
Received in revised form 24 June 2014 inate composites. There is a scope to enhance the physico-chemical and mechanical properties of com-
Accepted 3 September 2014
mercially available epoxy polymers through nanoscale modification. But, the improvement in the
Available online 4 October 2014
properties of the polymeric matrix mainly depends on the compatibility and extent of interaction of
nanoparticles with the epoxy polymers. With the exponential growth in the field of nanotechnology
Keywords:
and nanoscience, number of potential and compatible nanoparticles is available which can be judiciously
Molecular dynamics
Epoxy polymers
explored to engineer the epoxy polymer to attain the desirable properties. In view of this, Molecular
Carbon nanotube Dynamics (MD) studies have been carried out to investigate the interaction of filament type nano mate-
Nano modification rials such as Carbon Nanotube (CNT) with different types of epoxy polymers. Various types of resins and
Mechanical response hardeners available in market and usually adopted in industrial applications have been considered for the
Affinity index study. Steered Molecular Dynamics simulation has been carried out individually on CNT and epoxy resins.
Pull out simulations CNTs of different chiralities and diameters and epoxy resins are subjected to constant velocity pulling and
their responses are observed. Then, the interaction of each of the epoxy polymer with CNT is investigated.
In order to verify the results obtained from the affinity test using the interaction energy, pull out simu-
lations have also been carried out. It has been found that pull out test results strongly supports the results
obtained from the study using interaction energy in choosing the best and worst performing resins for
nano engineering using CNT. From the study, the suitable resin and compatible epoxy is identified for
CNT modification to ensure better performance of structures towards their repair and retrofitting. A
new parameter has been proposed in the present study to uniquely quantify the extent of interaction
of the polymer with CNT.
Ó 2014 Elsevier B.V. All rights reserved.

1. Introduction ries in determining the reaction rates, vibrational spectrum, ther-

The wide-spread use of nanotechnology paved the way for
development of many experimental and computational techniques
at molecular level for better understanding of nanomaterials and to
find out the source for certain phenomena observed at the macro-
level. Molecular mechanics is a computational technique that lies
between Quantum mechanics and Mesoscale modeling and deals
with the length scale of few angstroms to nanometers. The basic
assumption of molecular mechanics is that the matter consists of
a collection of atoms and from the interaction of atoms, the energy
can be calculated. Molecular mechanics has found its success sto-
⇑ Corresponding author at: CSIR-Structural Engineering Research Centre (CSIR-
SERC), CSIR Campus, Taramani, Chennai 600113, India. Tel.: +91 44 22549239; fax:
+91 44 22541508.
E-mail addresses: sindu@serc.res.in, sindu.lsn@gmail.com (B.S. Sindu).
modynamic and structural properties of the system [1–3].
Molecular Dynamics (MD) and Monte-Carlo (MC) simulations
are the two major simulation types under molecular mechanics.
MD is more widely used than MC owing to its ability to determine
the dynamic properties of materials. The objective of a MD simula-
tion is to compute the particle position as a function of time. It is
assumed that the particles behave classically, thereby, obey New-
ton’s second law of motion. For carrying out MD, the atomic system
should be defined by specific forms of inter-particle interactions,
proper initial and boundary conditions and thermo-dynamic con-
straints. The applications of MD is diverse; it is used in Chemo-
informatics to discover and design new molecules; Proteomics to
study the folding, misfolding and aggregation of proteins and
structural change of a protein upon mutation of one or more amino
acids; Biochemistry to study the drug delivery system; Material

http://dx.doi.org/10.1016/j.commatsci.2014.09.003
0927-0256/Ó 2014 Elsevier B.V. All rights reserved.
 B.S. Sindu, S. Sasmal / Computational Materials Science 96 (2015) 146–158 147

Crack

High Stress Zone High Stress Zone

Crack Propagation Crack Propagation

(a) Plate end FRP delamination (b) Intermediate crack induced FRP delamination

Fig. 1. Schematic diagram of FRP delamination.

Fig. 2. Atomic structure of (a) 7,7 and (b) 13,0 nanotube.

(a) 395000 (a) 45000

40000
345000 Bond Bond
35000
Energy (kcal/mol/Å)

295000 Angle
Energy (kcal/mol/Å)

Angle
30000
245000 Dihedral Dihedral
25000
195000 vdW vdW
20000
Kinec Kinec
145000 15000
Total Total
95000 10000
45000 5000

-5000 0
0 5000 10000 15000 20000 -5 5 15 25 35
-5000
Time Steps Diameter (Å)

(b) 395000
(b) 39300

345000 34300
Bond
Energy (kcal/mol/Å)
Energy (kcal/mol/Å)

295000 Bond 29300 Angle

245000 Angle 24300 Dihedral
Dihedral
195000 19300 vdW
vdW
145000 14300 Kinec
Kinec
9300 Total
95000
Total
45000 4300

-5000 -700
0 5000 10000 15000 20000 -5 5 15 25 35
Time Steps Diameter (Å)

Fig. 3. Energy vs. time for (a) armchair and (b) zigzag. Fig. 4. Energy vs. diameter for (a) armchair and (b) zigzag.

Science to predict the rheological, thermal and mechanical proper-
ties of polymers. Recently, it is being used by the civil engineers to
understand the structure and properties of various materials used
for construction, repair and rehabilitation of structures and to
study the behavior of these materials when some foreign materials
like fibers and nanomaterials are added to it.
Epoxy polymers are widely used in construction industry.
Epoxy polymers have been identified as the best binding agent
for construction materials like wood, metal and concrete. This
excellent property with its strain compatibility with construction
materials has made them to find its application in repair and reha-
bilitation of structures and in developing laminate composites.
148 B.S. Sindu, S. Sasmal / Computational Materials Science 96 (2015) 146–158
Fig. 5. Typical deformation behavior of CNT.
They have been used to develop laminate composites by binding
different layers and make them act as a single entity. They have
also been used as coatings for concrete floors to make them wear-
and chemical-resistant and as primers to improve the adhesion of
paints on metal surfaces and thereby making it corrosion resistant.
Many of the concrete structures are undergoing deterioration due
to aging. In order to strengthen these structures and to increase
its service life, they are bonded with fresh concrete or wrapped
with Fiber Reinforced Polymer (FRP). Epoxy polymers act as the
major binding agent during these applications. However, the
higher strength that is to be achieved due to strengthening is usu-
Fig. 6. Force vs. extension plot for (a) armchair and (b) zigzag.
Fig. 7. Young’s modulus of CNT.
ally not achieved due to debonding of FRP. This is complemented
by horizontal cracking in concrete leading to loss of concrete cover
thereby further reducing the strength of concrete and making it
more brittle. Schematic diagrams showing FRP delamination are
given in Fig. 1.
Also, though epoxy polymers have high bond strength, they fail
easily due to impact, fatigue and vibration loading. Hence, improv-
ing the impact strength, vibrational and other properties of epoxy
polymers will be of great use.
In order to achieve this, researchers are working on nano mod-
ification of epoxy polymers. Gilbert et al. [4] incorporated nano
alumina in epoxy based film adhesives and observed improvement
in the peel and shear strength of the adhesive. Wang et al. [5]
incorporated nano clay in epoxy matrix using a novel method
and found that the incorporation led to the improvement of stor-
age modulus, glass transition temperature, Young’s modulus and
the facture toughness of the matrix. Zamanian et al. [6] incorpo-
rated nano silica in epoxy resin to improve its fracture toughness.
Among different nanomaterials, it has been found that Carbon
Nanotubes (CNTs), owing to its excellent mechanical properties,
Fig. 8. Validation of Young’s modulus of (a) armchair and (b) zigzag nanotube. (See
above-mentioned references for further information.).
 B.S. Sindu, S. Sasmal / Computational Materials Science 96 (2015) 146–158 149

is the best nano modifier for epoxy polymers. Mao [7] incorporated
different percentages of functionalized CNT into epoxy matrix and
found that the nano composite not only had improved compressive
strength, flexural strength and flexural modulus but also had
improved impact strength and vibrational damping. Kordani
et al. [8] added 0.5% of CNT in epoxy matrix and found that the
vibrational characteristics like frequency and damping ratio of
the material improved when compared with the conventional
material.
There are different epoxy polymers as on date. The extent of
enhancing the properties of polymers depends how good CNT has
its affinity with that polymer. Hence, identification of the epoxy
polymer and hardener that has more affinity towards CNT is utmost
important. Yang et al. [9] used MD to study the interaction of CNT
with different polymers and found out that the strength of interac-
tion depends on the specific monomer structure. Al-Haik et al. [10]
observed the effect of chirality of CNT on the adhesion energy of
CNT–polymer composite. Gou et al. [11] determined the interaction
energy and the affinity index of CNT with epoxy resins. Liu et al. [12]
carried out interaction studies of CNT with different polymers and
found out that the mechanical properties of polymers are enhanced
if an appropriate repeat unit arrangement and conformation forms
the first layer around the CNT. Pang and Xu [13] carried out MD
simulations of interactions between CNT and surfactant and
observed that the structure of the surfactant and its concentration
played a significant role in the dispersibility of CNTS. Saha et al.
[14] carried out interaction studies of polymer on CNT and con-
cluded that the interaction is governed by number of repeating
units of polymer. Zaminpayma and Mirabbaszadeh [15] studied
the interactions between CNT and polymers and observed that
the interaction energy between the molecules were influenced by
monomer structure, and morphology of CNT whereas it was not
influenced by temperature. Haghighatpanah and Bolton [16] car-
ried out MD simulations on CNT polyethylene composites and

Fig. 9. Atomic structure of epoxy resins (a) DGEBA, (b) DGEBF and (c) TGDDM.
150 B.S. Sindu, S. Sasmal / Computational Materials Science 96 (2015) 146–158

2
d ri
mi 2
¼ Fi; i ¼ 1; . . . ; N ð1Þ
dt
where mi is the mass of the particle ‘i’ and Fi is the force acting on
the ‘i’th particle which can be obtained from,
X
N
Fi ¼ f ij ð2Þ
j–i

where fij is the force acting on atom i from all other atoms j in the
system. This equation can also be obtained from Lagrange equations
of motion. The Lagrangian is defined as,
Lðx1 ; y1 ; z1 ; . . . ; x_ N ; y_ N ; z_ N Þ ¼ Kðx_ 1 ; y_ 1 ; z_ 1 . . . x_ N ; y_ N ; z_ N Þ
 Uðx1 ; y1 ; z1 . . . xN ; yN ; zN Þ ð3Þ
where K is the kinetic energy given by,
X1  2
dr i
K¼ mi ð4Þ
2 dt
and U is the total potential energy of the collection of N particles
given by,
X
N1
U¼ U inter-atomic ð5Þ
i¼1

Here, U(rij) is the interatomic potential (or force-field function)

between atoms i and j.
Fig. 10. Simulated structure of epoxy resins (a) DGEBA, (b) DGEBF and (c) TGDDM. The Lagrange equations of motion are,
 
d @L @L
 ¼ 0; i ¼ 1; 2; . . . ; N ð6Þ
found out that short CNTs does not increase the Young’s modulus of dt @ x_ i @xi
the system whereas long CNTs do.
Substituting Eqs. (4) and (5) in Eq. (6) and simplifying we obtain
In this study, MD is used to identify the epoxy resin and the
the following scalar forms of the equations of motion,
hardener that has more affinity towards CNT thereby leading to
an enhanced epoxy compound that can be used with the construc- 2
d xi @U
tion materials. To get a clear picture on the mechanical properties mi 2
¼ ¼ F xi
dt @xi
of CNTs and epoxy compounds, Steered Molecular Dynamics (SMD) 2
d yi @U
is carried out. CNTs of various chiralities and diameters are sub- mi 2
¼ ¼ F yi ð7Þ
jected to uniaxial pulling and its response to the applied force is dt @xi
2
studied. Similarly, epoxy resins are also subjected to uniaxial pull- d zi @U
mi ¼ ¼ F zi
ing and its force–displacement relationship is obtained. To study dt
2 @xi
the interactions between epoxy polymers and CNT, MD simula-
tions are carried out with NVT ensemble. Each of the epoxy poly- The above set of 3N equations are solved using an appropriate
mers is placed near CNT and their interaction is observed. A new numerical integration procedure.
parameter called affinity index has been used to quantify the inter-
action of CNT with epoxy polymer. From this study, the extent of 2.1. Steered Molecular Dynamics
nano modification of epoxy resin and hardener can be ascertained.
Thus, this study provides the avenue to the scientists and engineers Steered Molecular Dynamics (SMD) is a type of simulation in
to tap the alluring scope of nanotechnology and nano engineering which atoms (single or group) are subjected to an external force
to develop an improved epoxy based polymers for effective usage along the desired degrees of freedom and the response of the sys-
in structural repair and retrofitting. tem to this external force is studied. The atoms that are being sub-
jected to force are called steered atoms. There is another group of

2. Molecular dynamics: Process 40000

35000
Consider a simulation box of size L  L  L containing N atoms.
_ 30000
The initial position and velocity of these atoms are x(0) and xð0Þ
respectively. The energy between atoms is defined by a specific 25000
Force (pN)

form of inter-particle interaction (interatomic potential). Inter- 20000 DGEBA

atomic potential is analogous to strain energy in classical mechan- 15000 DGEBF
ics. Here, it defines the energy stored in the bonds when it is 10000 TGDDM
subjected to some external force. The main objective of MD simu-
5000
lation is to determine the positions ri(t), i=1. . .N and momenta pi(t),
0
i=1. . .N of these N atoms over time. From the positions, other param-
0 5 10 15 20
eters like displacement and strain can be extracted. -5000
Extension (Å)
The atoms are assumed to behave classically and hence obey
Newton’s Second Law of motion defined as, Fig. 11. Force–displacement of epoxy resins.
 B.S. Sindu, S. Sasmal / Computational Materials Science 96 (2015) 146–158 151

atoms that are restrained to move along desired degrees of U is the potential energy,
freedom and they are called fixed atoms. SMD can be carried out k is the spring constant,
by two methods: constant force pulling and constant velocity v is the pulling velocity,
pulling. In constant force pulling, the steered atoms are pulled with t is the time,
a constant force along a specified direction. In constant velocity ~
r is the actual position of SMD atom,
pulling, the steered atoms are attached to dummy atoms via virtual ~
r0 is the initial position of SMD atom,
spring which is actually pulled with a constant velocity along ~
n is the direction of pulling.
the specified direction. The force between both the atoms are mea-
sured by, 2.2. Molecular dynamics
F_ ¼ rU ð8Þ
The MD simulations were carried out using an open source MD
where U ¼ 12 k½v t  ð~
r  r~0 Þ  ~
n code available online. The initial atomic positions were described.

Fig. 12. Atomic structure of epoxy resins (a) DETA, (b) DDS and (c) DETDA.
152 B.S. Sindu, S. Sasmal / Computational Materials Science 96 (2015) 146–158

In order to describe the inter-particle interactions, Chemistry at From Fig. 4 it is clear that the major contribution of total energy
HARvard Molecular Mechanics (CHARMM) force-field was used. is from bond energy whereas some contribution comes from angle
According to CHARMM force-field, energy also.
To get a clear picture of the deformation process of CNT, the
U inter-atomic ¼ U bonded þ U non-bonded ð9Þ deformation structure of armchair nanotube of diameter 10 Å at
where different loading levels are shown in Fig. 5 below. Since it is a typ-
ical deformation process, the time step details are not given.
U bonded ¼ kbond ðr  r0 Þ2 þ kangle ðh  h0 Þ2 þ kimproper ðx  x0 Þ2 In order to get a picture of the stiffness and the strength of var-
ious types of CNTs, force vs. extension plots is given in Fig. 6 (chi-
þ ktorsion ð1 þ cosðn/  /0 ÞÞ ð10Þ
ralities mentioned in the legend).
where From Fig. 6 it can be seen that as the diameter increases, both
kbond, kangle, ktorsion and kimproper are the force constants for bond the strength and stiffness of the CNT increases. This phenomenon
stretching, in plane bending, out of plane bending and torsion is observed in both the chiral tubes.
respectively, From the force vs. extension curves, the Young’s modulus of
r0, h0, x0 and /0 are the equilibrium bond length, in plane angle, CNT can be obtained using the following equation,
out of plane angle and dihedral angle respectively and
r, h, x and / are the instantaneous bond length, in plane angle, r F=A0
Y¼ ¼ ð12Þ
out of plane angle and dihedral angle respectively. e DH=H0
" 12  6 # where
qi qj rij rij F is the total force applied,
U non-bonded ¼ þ 4ij  ð11Þ
Drij r ij r ij Ao is the cross-sectional area,
Ho is the initial length and
where DH is the elongation.
qi and qj are the partial charges on atoms i and j separated by a
distance of rij and The cross-sectional area is given by,
ij is the depth of the potential well, rij is the (finite) distance at
which the potential is zero as and rij is the vdW radius. Ao ¼ pDt ð13Þ

The atomic system is first subjected to minimization, i.e., where D is the diameter of the CNT and t is the thickness of the CNT.
brought to its least energy state or stable state using Conjugate It is taken as 3.4 Å which is the inter-layer spacing in graphite.
gradient method. It is then equilibrated using Langevin dynamics.

3. MD on CNTs

The mechanical properties of CNTs are not well established. In

order to use it in engineering applications, it is utmost important
to understand its mechanical properties and various parameters
affecting the same. For this, MD simulations were carried out on
armchair and zigzag nanotubes with various diameters. The influ-
ence of diameter and the chirality of CNT on its mechanical prop-
erties have been observed. Fig. 2 shows the atomic structure of
armchair and zigzag nanotube generated for carrying out MD
simulations.
MD simulations were carried out using NPT ensemble with the
time step of 1 fs. The temperature was maintained at 300 K and
pressure at 1 atm using Langevin Dynamics. CNTs were fixed at
one end and subjected to constant velocity pulling using SMD on
the other end. The spring constant of the virtual spring was
7 kcal/mol Å2 and the velocity with which it was pulled was
0.1 Å/ts. The total energy plots and the force–extension plots have
been obtained which shows clearly the effects of various parame-
ters on its mechanical properties.
Typical plots of various energies for both chiral tubes of same
diameter, say 10 Å, are shown in Fig. 3.
From the above plots it is seen that the maximum contribution
to total energy comes from bond energy. Even the trend of total
energy curve is that of bond energy. Dihedral energy and van der
Waals energy has no contribution. This is because there are no
dihedrals in CNTs and since the study is limited to Single Walled
Nanotubes, the contribution of van der Waals energy is not there.
In both the tubes, there is some contribution of angle energy and
kinetic energy.
To get a better insight on how the energy stored varies for dif-
ferent types of CNTs, the plot of energy vs. diameter at an exten-
sion of 10 Å for both chiral tubes is given in Fig. 4. Fig. 13. Simulated structure of epoxy hardeners (a) DETA, (b) DDS and (c) DETDA.
 B.S. Sindu, S. Sasmal / Computational Materials Science 96 (2015) 146–158 153

Using the above equation, the Young’s modulus of CNTs of var- armchair, the Young’s modulus increases as the diameter increases
ious radius and chirality are calculated and shown in Fig. 7. but after a certain radius, it gets converged and remains constant.
From the plot it can be seen that the Young’s modulus of zigzag In order to validate the results obtained in the present study, it
tube is almost a constant value for all the diameters whereas in is compared with the values reported in the literature in Fig. 8.

(a) CNT – DETA (initial) (b) CNT – DETA (final)

(c) CNT – DDS (initial) (d) CNT – DDS(final)

(e) CNT – DETDA (initial) (f) CNT – DETDA (final)

(g) CNT – DGEBA (initial) (h) CNT – DGEBA (final)

Fig. 14. Snapshots showing interaction of CNT with epoxy polymers (front view).
154 B.S. Sindu, S. Sasmal / Computational Materials Science 96 (2015) 146–158
From Fig. 8, it has can be seen that the results obtained in the
present study are within the range of values reported in literature.
4. Molecular dynamics on epoxy resins
There is no straight forward method to evaluate the properties
of epoxy resins. Generally, they are cured with some hardeners and
only then their properties are determined. To have a better under-
standing on the behavior of epoxy resins, MD simulations have
been carried out on three epoxy resins, namely, Diglycidyl ether
of Bisphenol A (DGEBA), Diglycidyl ether of Bisphenol F (DGEBF)
and Tetraglycidyl ether of diamino diphenyl methane (TGDDM)
whose atomic structures with partial charges are given in Fig. 9.
The simulated structures are given in Fig. 10.
MD simulations were carried out using NPT ensemble. The tem-
perature was maintained at 300 K and pressure at 1 atm using
Langevin Dynamics. Epoxy resins were fixed at one end and sub-
jected to constant velocity pulling using SMD on the other end.
The spring constant of the virtual spring was 7 kcal/mol Å2 and
the velocity with which it was pulled was 0.001 Å/ts.
The force–displacement curve for all three epoxy resins is
shown in Fig. 11.
It can be seen from the plot that DGEBA has higher stiffness. The
stiffness of TGDDM and DGEBF is almost the same.
5. MD on CNT–epoxy polymers
In order to arrive at a good nano composite, it is desirable to
have an efficient load transfer mechanism from the matrix to
CNT. The load transfer is achieved through interfacial binding
energy. Since covalent functionalization of CNT introduces defects
and builds internal stresses in CNTs, introducing non-covalent
(i) CNT – DGEBF(initial)
(k) CNT – TGDDM (initial)
Fig. 14 (continued)
interactions between CNT and the matrix may lead to good inter-
facial binding.
The load transfer between CNT and polymer matrix is deter-
mined by various methods like force-separation methods, shear-
lag models and through interaction energy. Awasthi et al. [20]
modeled the interfacial behavior of graphene sheet and polymer
matrix using molecular dynamics force-separation model. The
polymer matrix was kept near the graphene sheet at an equilib-
rium distance initially. Load was applied to pull the graphene sheet
away from the polymer matrix. As the graphene sheet was pulled
away from the matrix, the load increased upto a peak value and
then reduced to zero. This peak load is called the peak traction
force and the area of the force–displacement curve gives the sepa-
ration energy. Using these parameters cohesive law has been
established to describe the interface between graphene and poly-
mer matrix. Similar procedure is followed to determine the shear
traction and separation energy by applying shear load to the
graphene sheet. Li and Seidel [21] used similar procedure to char-
acterize the interface between CNT and polymer. The effect of cur-
vature of CNT and temperature on the interfacial behavior was also
taken into account. Frankland et al. [22] used shear lag model to
determine the shear strength of the interface between CNT–poly-
mer matrix. CNT was placed at the center of the polymer matrix
and pulled out from it. The total work done in pulling CNT out of
the matrix is determined as the shear strength of the interface.
The effect of functionalization of CNT on the shear strength of
the interface is investigated by providing chemical cross-links
between CNT and polymer matrix. Chowdhury and Okabe [23] also
used the shear lag model to determine the interfacial shear
strength of the CNT–polymer matrix. The effect of matrix density,
chemical cross-links and geometrical defects on the interfacial
shear strength was also investigated.
Lordi and Yao [24] carried out MD simulations to understand
the factors that govern the strength of the interface. It has been
(j) CNT – DGEBF (final)
(l) CNT – TGDDM (final)
 B.S. Sindu, S. Sasmal / Computational Materials Science 96 (2015) 146–158 155

found out from their studies that the binding energy and frictional
forces play a minor role in governing the strength of the interface
whereas the steric interactions like wrapping of polymer on CNT
plays a major role. Chen et al. [25] observed that the interaction
energy of functionalised CNT with polymer was more compared
to the unfunctionalised one and hence concluded that the func-
tionalised CNT had a better load transfer behaviour. Zaminpayma
[26] carried out MD simulations on different polymer matrices
and demonstrated that there is a direct relationship between the
mechanical properties of CNT incorporated polymer matrix and
interaction energy between CNT and polymer.
Thus, it is evident that the load transfer between CNT and poly-
mer matrix has been determined by many methods such as force-
separation methods, shear-lag models and through interaction
energy. But it has been proved that the interaction energy plays
a major role in the load transfer between CNT and polymer and
also that there is a direct relationship between the interaction
energy and the load transfer behavior. Hence, in this study the

(a) CNT – DETA (initial) (b) CNT – DETA (final)

(c) CNT – DDS (initial) (d) CNT – DDS (final)

(e) CNT – DETDA (initial) (f) CNT – DETDA (final)

(g) CNT – DGEBA (initial) (h) CNT – DGEBA (final)

(i) CNT – DGEBF (initial) (j) CNT – DGEBF (final)

(k) CNT – TGDDM (initial) (l) CNT – TGDDM (final)

Fig. 15. Snapshots showing interaction of CNT with epoxy polymers (side view).
156 B.S. Sindu, S. Sasmal / Computational Materials Science 96 (2015) 146–158

(a) 0.6 tions were carried out using NVT ensemble with an infinite vol-
ume. The temperature was maintained at 300 K using Langevin
0.4
Dynamics. The interaction of epoxy polymers with CNT under the
0.2 above-mentioned conditions is monitored. The snapshots given
Affinity Index

in Figs. 14 and 15 show the interaction mechanism very well.

0 DETA CNT
0 20000 40000 60000 80000 By comparing the initial and final step, it is evident that all
DDS CNT
-0.2 epoxy polymers interact well with CNT. The epoxy polymers tend
DETDA CNT to come closer to CNT and try to align along the length or circum-
-0.4
ference of CNT. This is because of the non-bonded interaction
-0.6 between both the molecules. This shows that there is a good inter-
facial binding between CNT and epoxy polymers.
-0.8
Timesteps In order to quantify the interfacial interactions, a parameter
called ‘affinity index’ has been investigated. Affinity index can be
(b) 0.4
defined as the difference in the average distance between each
0.2 atom of the polymer to the geometric center of the CNT between
0
the current time step and the initial time step. It is given as
0 20000 40000 60000 80000 follows:
Affinity Index

-0.2
" # " #
DGEBA CNT XN XN
-0.4
DGEBF CNT Affinity index ¼ ðr i  C NT Þ  ðr i  C NT Þ ð14Þ
-0.6 TGDDM CNT i¼1 t i¼1 0

-0.8 where ri is the coordinate of ith atom of the polymer and CNT is the
-1 geometric center of the CNT.
In this study, the initial distance between polymers and CNT has
-1.2
Timesteps not been maintained constant. In order to remove the effect of ini-
tial placement of polymers from CNT on the affinity index, the sec-
Fig. 16. Time evolution of affinity index of (a) hardeners and (b) epoxy resins.
ond term of the equation has been given.
The time evolution of affinity index is shown in Fig. 16.
For the sake of clarity, the lowest value of the affinity index for
Table 1 every polymer is reported in Table 1.
Least value of affinity index. From the above plots and the table, it has been found that
Polymer Least affinity index (Å) TGDDM epoxy resin and DETA hardener has highest affinity with
CNT compared to others and usage of this particular epoxy resin
DGEBA 0.60973
DGEBF 0.61816 with hardener for nano modification will ensure better perfor-
TGDDM 0.96898 mance of structures towards their repair and retrofitting.
DDS 0.3427
DETA 0.60774
DETDA 0.24997
5.2. Simulations of pull out of CNT in epoxy polymer matrix

concept of interaction energy has initially been used to determine
the load transfer between epoxy polymers and CNT and the
findings are further assessed and confirmed by the simulated pull
out test using the molecular dynamics.
In order to observe these interfacial interactions and to arrive at
the best possible epoxy polymer feasible for nano modification,
Molecular Dynamics (MD) studies have been carried out to inves-
tigate the interaction of Carbon Nanotube (CNT) with different
types of epoxy polymers. Three resins, viz., Diglycidyl ether of
Bisphenol A (DGEBA), Diglycidyl ether of Bisphenol F (DGEBF)
and Tetraglycidyl ether of diamino diphenyl methane (TGDDM)
and three hardeners, viz., Diethylene triamine (DETA), Diamino
diphenyl sulfone (DDS) and Diethyl toluene diamine (DETDA), are
taken up for the study. The atomic structures with partial charges
and the simulated structures of hardeners are shown in Figs. 12
and 13 respectively.
Partial charges, calculated by Mulliken method using Marvin-
View, were defined for epoxy polymers. For the sake of clarity for
the readers, the partial charges are denoted in their atomic struc-
tures shown in Figs. 9 and 12.
5.1. Determination of affinity from interaction energy
MD simulations were carried out to determine the resin and
hardener that has better affinity with CNT. CHARRM force-field
was used to monitor the inter-particle interactions. The simula-
Pull out simulations are also carried out in order to ascertain the
efficiency of load transfer between CNT and epoxy polymers.
Though three resins and three hardeners have been taken up for
the affinity test, only the best resin and worst resin which are eval-
uated through the interaction studies have been considered for
pull out simulations. Hence, TGDDM and DGEBA have been taken
up for this study.
Simulation has been carried out in a box of size
50  50  150 Å. CNT is embedded inside the epoxy resin with a
density of 1.3 g/cm3. COMPASS (Condensed-phase Optimized
Molecular Potentials for Atomistic Simulation Studies) force field
has been used to define the interatomic interactions in this study.
In order to simulate the pull out test, for each epoxy resin, two dif-
ferent states have been simulated, i.e., (1) CNT fully embedded
inside the epoxy resin and (2) CNT placed fully outside the epoxy
resin as shown in Fig. 17.
Both the states are subjected to energy minimization and their
corresponding energies at both states, i.e., fully embedded and
after completion of pull out simulation are calculated. Pull out
energy, Epull-out is defined as the difference in energy between fully
embedded state, E1 and fully pulled out state, E2 which can be rep-
resented as
Epull-out ¼ E1  E2 ð15Þ
It can be observed from Table 2 that the pull out energy for
CNT-TGDDM matrix is significantly greater than CNT-DGEBA
matrix which signifies that CNT has more affinity towards TGDDM
and thereby it demands more energy to be pulled out of the matrix
 B.S. Sindu, S. Sasmal / Computational Materials Science 96 (2015) 146–158
(a) Fully embedded (DGEBA-CNT)
(c) Fully embedded (TGDDM-CNT)
Fig. 17. Two different states during pull out simulation of both epoxy resins, DGEBA and TGDDM.
Table 2
Pull out energy.
Resin Epull-out (kcal/mol)
DGEBA 395.37
TGDDM 10444.42
when compared to DGEBA matrix. It is significant to mention that
the observation from the pull out test is well corroborated with
that observed from the studies using interaction energy which is
used to develop the affinity index.
6. Conclusion
Epoxy polymers are widely used in construction industry for
repair and rehabilitation of existing concrete structures because
of its high bond strength and excellent compatibility with con-
struction materials. The major failure mechanism in these
strengthened structures is debonding of the strengthening mate-
rial under lower load leading to brittle failure of the strengthened
structure which is more dangerous. While epoxy polymers have
high bond strength, they fail easily due to impact, fatigue and
vibration loading. Hence improving the properties of epoxy poly-
mers is utmost important. Nano modification has proven to be a
promising way to enhance the properties of epoxy polymers. Since,
the research on nano materials and nano modification is limited in
construction industry, an attempt has been made to study the
mechanical characteristics of CNTs and epoxy polymers and their
interactions using MD. At first, the mechanical characteristics of
CNTs of different chiralities and diameter have been investigated.
The energy contribution and the force–displacement behavior have
been observed. Next, the force–displacement characteristics of
epoxy resins have been investigated. Then, the interaction of epoxy
resins and hardeners with CNT has been evaluated. A parameter
called affinity index has been proposed to quantify the interaction
of CNT with epoxy polymer. In order to verify the results obtained
from affinity test using the interaction energy, pull out simulations
have also been carried out. It has been found that pull out test
results strongly supports the results obtained from the study using
interaction energy in choosing the best and worst performing res-
157
(b) Fully outside (DGEBA-CNT)
(d) Fully outside (TGDDM-CNT)
ins for nano engineering using CNT. It can be underscored that
affinity test as performed in the present study can be used as a sim-
ple means to determine the compatibility of polymers for CNT
incorporation. From the present study, it has been identified that
TGDDM and DETA interacts well with CNT and hence the nano
modification of these epoxy polymers with CNTs might lead to
an enhanced polymeric composite that can be used in the repair
and rehabilitation of structures. The work carried out in the pres-
ent study will lead to the development of enhanced epoxy poly-
mers that can actually be used as a binding agent in
strengthening of existing concrete structures.
Acknowledgement
This paper is being published with the kind permission of the
Director, CSIR-Structural Engineering Research Centre, Chennai,
India.
References
[1] Y.V. Suleimanov, R. Collepardo-Guevara, D.E. Manolopoulos, J. Chem. Phys. 134
(2011) 044131-1–044131-10.
[2] T. Ishiyama, H. Takahashi, A. Morita, J. Phys.: Condens. Matter 24 (2012) 1–7.
[3] W. Sekkal, B. Bouhafs, H. Auorag, M. Certier, J. Phys.: Condens. Matter 10
(1998) 4975–4984.
[4] E.N. Gilbert, B.S. Hayes, J.C. Seferis, Polym. Eng. Sci. 43 (5) (2003) 1096–1104.
[5] L. Wang, K. Wang, L. Chen, Y. Zhang, C. He, Composites: Part A 37 (2006) 1890–
1896.
[6] M. Zamanian, M. Mortezaei, B. Salehnia, J.E. Jam, Eng. Fract. Mech. 97 (2013)
193–206.
[7] D. Mao, SAMPE 9 (2009) 18–21.
[8] N. Kordani, A. Fereidoon, M. Ashoori, J. Electron. Sci. Technol. 1 (2010) 007.
[9] M. Yang, V. Koutsos, M. Zaiser, J. Phys. Chem. B 109 (20) (2005) 10009–10014.
[10] M. Al-Haik, M.Y. Hussaini, H. Garmestani, J. Appl. Phys. 97 (074306) (2005) 1–
5.
[11] J. Gou, B. Fan, G. Song, A. Khan, Int. J. Nanosci. 5 (01) (2006) 131–144.
[12] W. Liu, C.L. Yang, Y.T. Zhu, M.S. Wang, J. Phys. Chem. C 112 (6) (2008) 1803–
1811.
[13] J. Pang, G. Xu, Comput. Mater. Sci. 65 (2012) 324–330.
[14] L.C. Saha, S.A. Mian, J. Jang, Bull. Korean Chem. Soc. 33 (3) (2012) 893–896.
[15] E. Zaminpayma, K. Mirabbaszadeh, Comput. Mater. Sci. 58 (2012) 7–11.
[16] S. Haghighatpanah, K. Bolton, Comput. Mater. Sci. 69 (2013) 443–454.
[17] V.N. Popov, V.E. Van Doren, M. Balkanski, Phys. Rev. B 61 (4) (2000) 3078.
[18] X. Guo, J.B. Wang, H.W. Zhang, Int. J. Solids Struct. 43 (5) (2006) 1276–1290.
[19] H.W. Zhang, J.B. Wang, X. Guo, J. Mech. Phys. Solids 53 (9) (2005) 1929–1950.
[20] A.P. Awasthi, D.C. Lagoudas, D.C. Hammerand, Modell. Simul. Mater. Sci. Eng.
17 (1) (2009) 015002.
[21] Y. Li, G.D. Seidel, Modell. Simul. Mater. Sci. Eng. 22 (2) (2014) 025023.
158 B.S. Sindu, S. Sasmal / Computational Materials Science 96 (2015) 146–158

[22] S.J.V. Frankland, A. Caglar, D.W. Brenner, M. Griebel, J. Phys. Chem. B 106 (12) [25] H. Chen, Q. Xue, Q. Zheng, J. Xie, K. Yan, J. Phys. Chem. C 112 (42) (2008)
(2002) 3046–3048. 16514–16520.
[23] S.C. Chowdhury, T. Okabe, Composites Part A 38 (3) (2007) 747–754. [26] E. Zaminpayma, Polym. Compos. (2014).
[24] V. Lordi, N. Yao, J. Mater. Res. 15 (12) (2000) 2770–2779.