1
Molecular Computer Simulations of
Graphene oxide intercalated with methanol:
Swelling Properties and Interlayer Structure
Taleb Baker Jader ([email protected]), Physics Department, Umeå University
April, 2017
2
Table of contents
Abstract 3 1. Introduction 4 1.1 Graphite oxide 5 1.2 Simulation method and codes 6 2. Simulation systems 6 3. Detailing of simulation process 8 3.1 Simulation concepts 13 3.1.1 Periodic boundary conditions (PBC) 13 3.1.2 Short-range interactions 14 3.1.3 Long-range interactions 14 3.1.4 Temperature and pressure 14 3.1.5 Constraints 14 3.1.6 Cutoffs 15 4. Results and Discussion 16 4.1 Swelling curves 16 4.2. Solvation 23 4.3 The enthalpy 25 4.4 Interlayers distribution 25 5. Conclusion 29 6. References 31
3
Abstract Molecular dynamics (MD) simulations has emerged as a very useful technique to study many types of properties for layered nanomaterials, such as graphene oxide (GO), which is of broad interest in different areas such as biotechnology, nanoscience, biomedicine and so forth. Moreover, its physical, chemical, and electronic properties have attracted significant attention from researchers to investigate its properties.
The purpose of the present work was to probe the expansion of GO by intercalation of methanol between GO Nano-layers using molecular computer simulations. Atoms and molecule trajectories and their mutual interaction were modelled on the length and time scales of nanometres and nanoseconds, using the GROMACS MD code computing all electrostatic and van der Waals interactions. Analysis of the MD simulations versus methanol loading revealed that graphite
oxide’s layers could be intercalated with methanol. Subsequently, it changes the
distance between graphene oxides interlayers, which have different functional
groups, such as epoxy, and hydroxyl by oxidation. These properties revealed the
apparent swelling that corresponds to solvation of the GO monolayers, which
helps to illustrate the energetic origin of crystalline swelling processes in
graphene oxide. The results will be valuable for researchers by increasing their
knowledge about graphite oxides interlayer’s properties. The results will also
help researchers to unravel the range and location of graphite oxides
applications in our life. They can also serve the readers with new information
about graphene oxides properties.
4
1. Introduction To be mindful about a material such as graphene oxide (GO), which has
remarkable relevance in various fields, requires studying it under specific
conditions. Because GO has unique structural and dynamic properties. Which
can be used to investigate unanticipated physical or chemical properties of
atoms and molecules as the present project with the graphene oxide (GO)
molecule (it has many crystalline properties e.g. the phenomenon of swelling
due to strong oxidation) under periodical boundary conditions1.
This study concerns molecular dynamics (MD) simulation of graphite oxide
(GO) and how its interlayers change by intercalation of methanol as a solvent.
Why graphite oxide? It has many interesting chemical and physical properties
which encourage detailed investigations. To mention some of that explored
properties, there are similarities between GO and clay material about their
swelling behaviour in water and hydrophilic solvents because they are
hydrophilic materials. Just as with clays, the modulation of the GO structure
can be changed by the physicochemical environments properties (such as pH,
salinity, etc.)2.
Solvation plays an important role in the regulation of the structural properties
e.gs. the interlayer distance of GO increases when immersed in water3,4. In pure
water both GO and graphite oxide membrane are solvated similarly5. The first
layered material, which was showing powerful swelling with water at high-
pressure level, was graphene oxide. Subsequently, phenomenon of hydrophilic
was demonstrated6,7. Graphene oxide (GO) has a tuneable conductivity and
electronic band gap, which opens possibilities for a huge number of potential
applications. GO is chemically stable with a high surface area up to 2630
(m2/g)8. This plays a huge role in the field of absorption applications, i.e. it
affects the heat transfer’s process during the porous material9. The interlayer
limitation of graphite oxide’s membrane for absorption depends on the type of
liquid as a solvent, chemical nature of the solutes, and interaction of dissolved
species into the graphite oxide structure. Limitation of GO for Nano-filtrations
application for some solution (basic salts, and amines etc.) showed a broad
variation of results10,11 .
Moreover, the motivation behind the study was to know more about GO. To
know how its interlayers structure changes, how its crystalline properties
change by swelling, considering the effect of intercalation of methanol into GO
interlayer structure. The aim of the study was to investigate swelling of
graphene oxide layers and compute the solvation, immersion, and potential
energy. Molecular dynamics (MD) simulation was used for this purpose. After
building of six different systems with different ratios of –OH and -epoxide of
GO functional groups and preparing a suitable environment factors, such as
constant temperature, pressure, with periodic boundary conditions (PBC). The
results are that we could show by swelling the possibility of intercalation of
5
graphene oxides interlayers structure12 of methanol and recognizing the
solvation, immersion, and potential energy by analysis of the MD
simulations13,14.
Why is this topic important? Graphene oxide displays a great role in the field of
filtration. One of the greatest challenges, which people have just now in the
world, is clean water. The population increases year after year, accordingly they
need clean water. Such a challenge requires development and optimization of
industrial processes aimed at filtering and desalinating (removing salt from
seawater) using the most promising filtration technology. Graphene oxide acting
as Nano-sponges (NS) has been prepared and intensively studied for this
purpose. They can absorb many times its own weight, and can have both
hydrophilic and hydrophobic properties, which are extremely useful to separate
liquids like oil from water, e.g. when we treat oil spills in the oceans15.
Additionally, nowadays, the expanding of chemical products in agriculture, the
industry’s applications or in general the modern life leads to many new
dangerous contaminants into our environment. Moreover, not long ago the
removing of these contaminants such as bacteria, viruses and other natural
organic matter (NOM) were difficult with the traditional absorbents e.g. ozone
or ultraviolet light. That is because they were very expensive and not active.
Therefore, it was and hitherto necessary to search after a new generation of
absorbent materials and their properties which extend the limits of that
removing materials and absorbents. As we explained above GO has an
enormous role in the field of absorption applications. Therefore, the aim of this
study is to discover something new about its properties, such as the immersion
and solvation energy. We hence think our results will be a value in the field of
adsorption16.
1.1. Graphite oxide According to the production method, there are two types of graphene oxide
Brodie (B-GO membranes) and Hummer (H-GO powders). However, the type
selected graphene model for the present project was inspired by Brodie. Studies
have shown the same results about their absorption of polar organic solvent
under constant temperature. Studying such property by different solvents and
under specific conditions e.g. constant volume and pressure allows intercalation
of graphene oxide interlayers by different molecules. Accordingly, producing
information about properties of both solute and solvent17.
One study showed that the change in interlayer distance of both types is similar
in pure water. However with intercalation of some aqueous liquids, such as
ethanol, the interlayer of graphene will change. Moreover, the interlayers
distance for H-GO powders will increase whereas for B-GO membranes
decrease. That is because the solvent will contract the lattice and the permeation
will slow down. Other studies have reported that the synthesized graphite oxide
(Brodie and Hummer) is different in solvation and exfoliation18.
6
Shije et al. (2012) reported that there is a significant expansion of H-GO in
alcohol with increasing temperature. Nevertheless at low temperature the phase
transitions differs more for B-GO than H-GO. The reason behind GO becoming
one of the most intensive subject of debate between researchers is the possibility
to synthesize it in different types with various degrees and number of functional
groups19. The possibility of intercalation of graphene oxide by alcohol (ethanol,
and methanol etc.) in different temperature opens other possibilities to
intercalate it by larger solvents; also producing varies composition of graphene
oxide20.
1.2. Simulation method and codes A number of methods and programs have been used for molecular or particle
simulation techniques. Molecular dynamics (MD) has been used in a wide range
of the applications in both science and engineering, because it can be used for
relatively large simulations on the molecular scale (Nano-scale) starting from
simple gases and liquids to complex materials. In classical MD simulations, the
system is described by Newton’s equation of motion, with the atom being the
basic particle. That is why magnetic and electronic properties cannot be
determined by this method, because interatomic potential will be used21. Typical
programs for MD simulations are LAMMPS, AMBER, DL-POLY, CHARMM,
and GROMACS, where the latter was utilized in this work22,23.
Another method is the Finite Element Analysis (FEA). In this method, the
system is divided into many small elements. FEA provides solutions for very
complex problems that have relation with material properties such as
permeation24. A third method is Monte Carlo (MC) simulations, which uses
random atomic movements accepted by a probability determined by the ratio of
a random number and the Boltzmann factor (the Boltzmann factor indicates
how probable is the new configuration), to solve problems and to find new
configurations25,26.
2. Simulation systems To investigate the crystalline swelling properties of graphene oxide (GO) with intercalation of methanol as a solvent, molecular dynamics (MD) simulations were used. The simulations systems consisted of constant number of particles (N) inside a periodically replicated rectangular simulation box. The initial simulation box size was approx. (3.40, 2.95, 12) nm in the x, y and z-axis directions, respectively, based on a graphene unit cell with 8 C atoms, and x and y dimensions of 0.4254 and 0.4912 nm. This GO layer and box size was required to disallow the methanol molecules not to ‘see’ itself, even at the lowest methanol loading.
7
The simulation boxes contained four GO layers based on 8x6x1 unit cells of graphene, having the atom types C (opls_141), C_OH (opls_159), Ce (opls_183), epoxide Oe (opls_180), hydroxyl Oh (opls_154) and H (opls_155), These four GO layers were initially intercalated with 360 methanol molecules each. The z-axis is taken perpendicular to the interlayers, and thus it decreases in length when removing the methanol. Hence, the initial systems consisted of 8x6 graphene of GO unit cells, with a basal spacing of 3 nm. Five different systems of GO and methanol molecules were used, differing in their epoxide O (Oe) to hydroxyl O (Oh) ratio. Each system consisted of 146 O atoms and 384 C atoms per each layer of graphene oxide. For reference one pure graphene system was also modelled.
Figure 1. Representative snapshots from the MD simulations showing representative graphene
and GO layers. From A-D the layers are: A) Graphene, B) 146 OH, C) 100 OH +46EpO, D)
146EpO, E) 73OH + 73EpO, and F) 46OH +100EpO.
Specifically, the five GO systems consisted of 146OH (Oe/Oh = 0), 146Epo
(Oe/Oh = 1), 100OH+46Epo (Oe/Oh = 0.31), 73OH+73Epo (Oe/Oh = 0.50),
and 46OH+100Epo (Oe/Oh = 0.69), where Oe represent epoxide O and Oh the
hydroxyl O. The pictures for the systems Graphene, 146 OH, 100 OH+46EpO,
A) B) C)
E) D) F)
8
146EpO, 73OH + 73EpO and 46OH +100EpO, with dimensions of (2.898nm x
3.346nm), is shown in figure 1 as well as an example for the full system of
46OH+100EpO in figure 2. The short-range van der Waals and the electrostatics
interactions were handled by the Lennard-Jones potential and Coulombs law,
resp., using a cut-off of 1 nm, whereas long-range electrostatic interactions were
treated with particle mesh Ewald (PME) method27,28.
Figure 2. Initial configuration for the 46OH+100EpO system with 360 methanol in each
interlayer.
3. Details of the simulation process All-atom optimized potentials for liquid simulations using the OPLS/aa force
field was used29. Before the simulation the system has to be equilibrated by
finding a minimum point of the systems potential energy. The equilibration was
performed first by energy minimization using the steepest descent algorithm,
secondly under NVT conditions in the canonical ensemble i.e. constant number
of particles, volume and temperature30. Volume optimization was performed
under NPT conditions in the isothermal-isobaric ensemble.
GROMACS (5.1.2) was used for all simulations. The energy minimizing by this
method started with steepest descent (an extension of Laplace's method for
approximating an integral) algorithm for flexible and constrained bonds. It
9
takes a step in the direction of the force without depending on the building of
the previous step. It adjusts the step size in order for the search to become
faster. Although it is not an efficient algorithm for searching, it is strong, simple,
and powerful to carry out31,32.
The time evolution of the particles interactions inside the simulation cell was
calculated by integrating Newton’s equations of motion over discrete time steps.
The initial velocities were assigned to the Maxwell-Boltzmann distribution
corresponding to a temperature of 298K. The output trajectories were saved
every 5 ps and the pressure was coupled to a Berendsen barostat.
Periodic boundary condition were used in all directions. The time step for each
production run was 2.0 fs, and the total simulation time for each system was 60
x 5 ns
In MD simulations all atoms interact with each other thus, the forces act on
particles are the negative derivatives of a potential energy.
𝐹𝑖 = −𝜕𝐸
𝜕𝑟𝑖 (1)
In molecular dynamics simulations the forces of the N particles with
coordinates 𝑟𝑖 are described by
𝑚𝑖 𝑑2𝑟𝑖
𝑑𝑡2= 𝐹𝑖 i = 1, . . . . . . . . . . . . . .,n
(2)
Where 𝑚𝑖 is the mass of particle i, and Fi is the force acting
Part of the total potential energy and interaction forces stem from the bonded
intramolecular interactions, not just between pair interactions, but also between
three and four atoms. The bonded interaction between two covalently bonded
atoms is called bond stretching, between three atoms is bond angle, and
between four atoms is called dihedral angle. There is also an improper dihedral
which is a special type of dihedral interaction (torsions) used to maintain
remaining atoms in a plane.
Electrostatic (long-ranged potential) and van der Waals interaction (short-
ranged interactions potential) as a non-bonded energy between pairs
interaction33. Figure 3 explains the non-bonded interactions between particles
during simulation’s process.
10
Figure 3. Types of interaction forces between particles during molecular dynamics simulation
which includes, bonded (bond stretching, angle, dihedral angle, and improper) and non-bonded
interactions (electrostatic, and van der Waals) from http://www.verachem.com/products/vm2/.
As we mentioned above, the molecule has the following configuration of energy
consists of the sum of bonding (𝐸𝑏) and non-bond (𝐸𝑛𝑏).
𝐸𝑡𝑜𝑡 = 𝐸𝑏 + 𝐸𝑛𝑏 (3)
The bonding energy is
𝐸 = 𝐸𝑏𝑜𝑛𝑑 + 𝐸𝑎𝑛𝑔 + 𝐸𝑑𝑖ℎ + 𝐸𝑡𝑜𝑟𝑡ℎ (4)
Where 𝐸𝑏𝑜𝑛𝑑 is the potential energy corresponding to covalent bonds, 𝐸𝑎𝑛𝑔is the
energy of the angular bonds, 𝐸𝑑𝑖ℎ is the energy of the dihedral bond’s angles
which involving more than three atoms, and 𝐸𝑡𝑜𝑟𝑡ℎ is the energy of the torsion,
which keep groups in planar.
The harmonic potential energy of the system due to covalent bonds, which is a
function of the particle separation, is described by the following equations34.
Bond potential (interactions between pairwise particles)
U (𝑟𝑖𝑗) = ∑1, 2 pairs 1
2𝐾0 (𝑏𝑖𝑗 -b0)2 (5)
Where b0 is the average bond length and K0 the bond spring constant.
Angle potential (interactions between three particles)
11
𝑈(𝜃𝑖𝑗𝑘) = ∑1
2𝑎𝑛𝑔𝑙𝑒𝑠 𝐾𝜃 (𝜃𝑖𝑗𝑘 − 𝜃0)2 (6)
Where 𝐾𝜃 is the force constant, 𝜃 is the average bond angle.
Dihedral potential (interactions between more than three particles)
For dihedral potential (interactions between mor than three particles), or for
torsion as it is symmetric beween (o-𝜋). For alkanes (methane, ethane, propane,
and higher members), the Ryckaert-Bellemans potential is often used. An
equivalent torsional potential is the Fourier cosine seriesexpansion35.
𝑈𝑡𝑜𝑟𝑠𝑖𝑜𝑛(∅) = ∑1
2 𝑛 𝑉𝑛 ( 1 + 𝑐𝑜𝑠(𝑛∅ − 𝛿𝑛) (7)
Where, Ø is diheral angle, Vn the torsional rotation, 𝛿𝑛 is the face factor and n is
multipilicity force constant or the number of function minima under rotation of
2𝜋.
The non-bonded interaction energy (𝐸𝑛𝑏 ) is the sum of van der Waals and
Coulomb potential energy. Due to the complex of calculation of pairwise
potential during simulation, the non-bonded interactions are the most time
consuming task, since each particle in non-bonded interactions has a large
number of interacting neighbours.
𝐸𝑛𝑏 = 𝐸𝑣𝑑𝑊 + 𝐸𝑐𝑜𝑙 (8)
The short-range van der Waals interactions energy between each pair of atoms
which located within a cutoff distance (in this study taken as 1.2nm), is
expressed by the Lennard-Jones potential energy.
𝐸 ( 𝑟𝑖𝑗) = ∑ 4휀𝑖𝑗𝑖,𝑗 [(𝜎𝑖𝑗
𝑟𝑖𝑗)
12
− (𝜎𝑖𝑗
𝑟𝑖𝑗)
6
] for rij < 1.2nm (9)
where rij is the distance between the pair of atoms, εij is the potential energy
minimum (the maximum interaction strength) whereas σij represents the
interatomic distance at which the potential U is zero36. The term 1/r6 describes
the attraction and the term 1/r12 describes the repulsion term37.
As for the short-range electrostatic contribution to Enb, it is expressed by
Coulomb’s law:
E(rij ) = 1
4𝜋𝜀𝑟𝜀0 𝑞𝑖 𝑞𝑗
𝑟𝑖𝑗 (10)
Where 𝑞𝑖 and 𝑞𝑗 corresponds charges, 휀0 is the permittivity of space and 휀𝑟 the
dielectric constant of the medium (here == 1). For efficiency reasons, the long-
12
range coulombic interactions beyond the cutoff distance are treated by Particle
Mesh Ewald (PME) method.
Matter strives to the lower its energy state. The balance between attraction and
repulsion of the particles will take place when they reach the distance of
equilibrium38 as it is explained by Lennard Jones potential in the figure 4.
Figure 4. The Lennard Jones potential between two particles as a function of the separation r,
displaying the attractive and repulsive contributions to the potential energy. From:
http://www.talkchannels.com/lennard-jones-potential-equation/
The leap-frog (Verlet) algorithm was used to calculate atomic trajectories by
integration of Newton’s equation of motion39. Its disadvantage is that the
position and velocity will be obtained at different time step. Gathering errors
produces fluctuations on the molecular dynamics simulation.
Figure 5. The leap-frog algorithm process between the position and velocity as function of
time. From: https://www.physics.drexel.edu/students/courses/Comp_Phys/Integrators/
leapfrog/
13
The leapfrog algorithm is the standard MD integrator in Gromacs molecular
dynamics simulations. The positions, velocities and accelerations of particles
were be known at the same time step. Thus, it is stable, works perfect for both
short and long time steps and the time is changeful and precise. Figure 5 shows
the positions and velocities, which leap-frog algorithm over each other.
Leap-frog integration is a particular differential equation for the two-coupled
first order which can be written:
Xn+1 – X n / Δ t = Vn+1 /2 → Xn+1 = X n + Vn+1 /2 Δ t (11)
Where, Δ t → 0. Χ is sampled at time points t = 0… n Δt and V is sampled at t =
Δ t /2, 3t Δt/ 2 ,………… n Δt/2. When the time changes the position jumps over,
so on for all other orders40.
3.1 Simulation concepts
3.1.1. Periodic boundary conditions (PBC)
As mentioned above periodic boundary conditions were used during the
simulations. The idea behind periodic boundary conditions is to decrease the
number of interactions and to avoid edge artefacts (since the interactions
between the particles at the edge of the box would be different from the other
places). Another idea for of periodic boundary conditions is that when a particle
leaves the simulation cell on the right, it will re-enter it as an image of itself on
the left41 as it seems in figure 6 .
Figure 6. Explanation of how particles move under PBC, particle that leaves the domain at the
right of the box will re-enter the domain from the left (dashed). From:
https://astronomy.swin.edu.au/sao/guest/knebe/
In such situation, translated copies of itself will be surround the box, as it
appears in figure 7, i.e. there will be no actual limiting boundaries of the
system. Thus, implementation of molecular dynamics simulation requires a
suitable simulation box in order to decrease the number of interaction between
14
particles, to be a suitable for the study number of particles in the box and to
keep only track of the original image as a representative of all other images42.
Figure 7. Illustration of the particle images inside the simulations box. From:
http://www.ch.embnet.org/MD_tutorial/pages/MD.Part3.html
3.1.2. Short-range interactions
For calculation of short-ranged interactions between neighbouring simulation
cells, the minimum image convention (MIC) combined with particle boundary
conditions (PBC) are used in MD simulations, in such a way that only one image
of each particle is considered for each pair interaction calculation. The role of
the cut-off is to avoid inefficiency as it decreases the number of pairwise
potentials that needs to be calculated43,44.
3.1.3 Long-range interactions The long-range electrostatic interactions are calculated in the reciprocal space
using both PBC and PME. Using a cut-off for this version is to decrease errors
produced from a large number of interactions between particles 45.
3.1.4. Temperature and pressure
In general there are very few instances under which the micro canonical NVE-
ensemble, i.e. constant number of particles N, volume V, and energy E usually
are studied by molecular dynamics. It is often more relevant to study and probe
other types of ensembles such as the canonical NVT ensemble, which
corresponds to constant number of particles N, volume V and temperature T, or
the isobaric-isothermal ensemble, having constant number of particles N,
pressure P and temperature T, GROMACS supports all common baro-stats and
thermostats46,47.
3.1.5 Constraints
By applying constraints on the bond lengths and angles of the atoms, creates
possibilities for removing the fastest motion (degree of freedom) from the
system, and use of longer time steps and subsequently more efficient
simulations. Because the fastest degree of freedom is that of to the light atoms,
15
here hydrogen atoms, which in fact cannot be well described by a harmonic
oscillation, it is constrained using the so-called LINCS algorithm48.
Since the hydrogen bonds are important for efficiency of the simulations, we
will mention some facts about them. Hydrogen bonds is an intermolecular bond
formed between two different electronegative atoms, such as N, O, and F, via an
intermediate H atom. Different geometric criteria for hydrogen bonds exist. In
GROMACS it’s considered that when the angle between the donor-hydrogen-
acceptor is less than 30° and the donor-acceptor distance is less than 0.35nm,
they share enough electron density to make a hydrogen bond. In hydroxyl
groups, the oxygen becomes negative charged because it takes electron density
since it has more charge density around it compared to hydrogen49. Figure 8 is
illustrated what mentioned above.
Figure 8. Shows hydrogen bond between two different molecules bond. From:
https://www.khanacademy.org/science/biology/water-acids-and-bases/hydrogen-bonding-in-
water/a/hydrogen-bonding-in-water.
Constraints in GROMACS can include: positions, angles, distances, orientations
and dihedrals. They are based on the fixed lists and are useful when increasing
the time step for molecular dynamics simulation. The most obvious constraint
among them is the position constraints, which are used to avoid severe
rearrangement when the system equilibrates 50.
3.1.6. Cut-offs
To perform the simulations, the non-bonded van der Waals and Columbic forces
in every atom have to be calculated. Particles within a cut-off radius (𝑅𝑐) and
neighbour list (𝑟𝑙) considered for the force calculation as illustrated in figure 9.
During the simulations, the number of solvent molecules should be somewhat
small to reduce the computational cost. Due to the minimum-image convention
(MIC), each particle only interacts with the nearest (image-) particles of the
remaining system. Because of this the non-bonded interactions must have a cut-
off radius (𝑅𝑐) that does not exceed half of the box’s length (𝑅𝑐 ≤ L/2)51,52 , since
else some atom-atom interactions would be accounted twice.
16
Figure 9. Illustration showing the particles that have to be accounted for when performing (MD) simulation those particles inside both the cut-off radius (𝑟𝑐) and neighbour list (𝑟𝑙), which in principle can be identical. From: htt://silas.psfs.mit.edu./22.15/lectures/chap 10 .xml
4. Results and Discussion
4.1. Swelling curves The properties of graphene oxide were investigated in order to compute the
swelling energetics and layers spacing. Simulations were performed with six
different energy systems. The basal spacing and the solvation, immersion and
potential energies were calculated. The basal spacing for the graphene, 146OH,
100OH 46EpO, 73OH 73EpO, 46OH 100EpO and 146EpO systems are plotted
in Figure 10. The curves except graphene show its first plateau approximately at
one methanol layers per GO unit cell which identical start of swelling. The
simulation reproduce that plateaus with very small spacing during the process.
Figure 10. Swelling curves from simulation between the number of methanol contents per GO
layer and basal spacing (d001) for the systems Graphene, 146OH, 100OH+46EpO, 73OH+73EpO,
46OH+100EpO, and 146EpO. Note that all GO curves were shifted for clarity. Error bars are
omitted since the block-averaged standard deviations of the d001 values were less than 1pm.
0 2 4 6 8
0
0.5
1
1.5
2
2.5
3
3.5
4
Num. of MET per GO unit cell
Bas
al s
pac
ing
[nm
]
Graphene 146 OH
100 OH 46 EpO 73 OH 73 EpO
46 OH 100 EpO 146 EpO
+0.8 nm +0.6 +0.4 +0.2 +0.0
17
The computed basal spacing are functions of the methanol layer contents. From
figure 10, it can be seen that the swelling curves of the GO systems increase
monotonically above approx. 3 and sometimes 2 methanol layers per GO unit
cell whereas the graphene system display a non-linear expansion up to much
higher methanol loadings.
The density maps in two and three dimensions for Methanol in two interlayers
between GO layers as a function of the basal spacing values for the six systems
can be observed in the figures (11-16), where the blue colour corresponds to low
density, and yellow high density of Methanol. The basal spacing decrease when
moving from left to wright.
Figure 11. Shows density maps in two and three dimensions for Methanol in two interlayers between GO layers as a function of the d001 values for the 146OH system
18
Figure 12. Shows density maps in two and three dimensions for Methanol in two interlayers between GO layers as a function of the d001 values for the 100OH+46Epo system
19
Figure 13. Shows density maps in two and three dimensions for Methanol in two interlayers between GO layers as a function of the d001 values for the 73OH+73Epo system
20
Figure 14. Shows density maps in two and three dimensions for Methanol in two interlayers between GO layers as a function of the d001 values for the 46OH+100Epo system
21
Figure 15. Shows density maps in two and three dimensions for Methanol in two interlayers between GO layers as a function of the d001 values for the 146Epo system
22
Figure 16. Shows density maps in two and three dimensions for Methanol in two interlayers between GO layers as a function of the d001 values for the graphene reference system
23
4.2. Solvation
In general the solvation energy comes from intercalation of solute by solvent
that causes to change the layers and interlayers properties. Moreover, Solvation
considers the change of GO swelling energy and reflect the energy of adsorbed
methanol. The solvation energy can be defined by the following equation.
ΔU= (U (N) – U (0)/N (12)
Where, N is the number of the interlayer methanol molecules, U(N) is the
ensemble average of the potential energy and U(0) is the potential energy of the
dry GO respectively53.
The calculated solvation energy of GO for the graphene, 146OH,
100OH+46EpO, 73OH+73EpO, 46OH+100EpO and 146EpO systems is
displayed in figure 17.
Figure 17. Solvation energy, from simulation as a function of the number of methanol content
per GO layers for the six systems.
For the GO systems the common feature is that when the methanol content per
GO unit cells increases, the negative (i.e. favourable) solvation energies
approach the less negative average bulk of methanol energy. Among the
different GO systems, increased ratios of hydroxyl groups generally lower
solvation energy. From figure 17, the solvation energy of the graphene system
shows that the graphene does not swell since its solvation energy is positive.
The determination of immersion energy will supply a complementary and
efficient method for investigation of discrete swelling properties. The immersion
0 2 4 6 8
-35
-30
-25
-20
-15
-10
-5
0
5
10
15
20
Num. of MET per GO unit cell
Solv
atio
n e
ner
gy [
KJ/
mo
l]
Graphene 146 OH
100 OH 46 EpO 73 OH 73 EpO
46 OH 100 EpO 146 EpO
X1/10
24
energy was found to be most useful uncover an apparently global minimum in
the enthalpy swelling energy and is displayed in figure 18, which shows many
clear minima as a meta-stable points implying that there is an energetic
stabilization of the discrete layers. These immersion energies data provide
computational verification of discrete swelling process in GO. Many
simulation’s processes for swelling have done without effecting of the pressure
i.e. the pressure was constant, as in present study. However, the evidence for
swelling process is the oscillation pressure inside the simulation box. The
oscillation uncovers behaviour of methanol through layers and interlayers
equilibration and formation of solvation.
𝐅𝐢𝐠𝐮𝐫𝐞 18. Immersion energy, from simulation as a function of methanol content per layer of
GO, which shows the stable points that do not tend to solvate, that have minimum immersion
energy.
From figure 18, with increasing methanol content, all the GO systems show a
decrease in the immersion energy, except 146EpO. Since it does want to solvate
initially. For graphene, the fully desolvated state is by far the most stable which
means also it does not tend to solvate at all, although several meta-stable
minima are found at higher methanol loadings. The states of methanol content
layers for the systems graphene, 146OH, 100OH+46EpO, 73OH+73EpO,
46OH+100EpO, and 146EpO, of above 0, 3, 3 or 4, 4, 3 and 3 or 3.5,
respectively correspond to the global immersion energy minimum.
The maximum immersion energies values, for the 146OH, 100OH 46EpO, 73OH
73EpO, 46OH 100EpO, and 146EpO systems, calculated from the most stable
0 2 4 6 8
0
50
100
150
200
250
300
350
Num. of MET per GO unit cell
Imm
ersi
on
en
ergy
[J/
g]
Graphene 146 OH
100 OH 46 EpO 73 OH 73 EpO
46 OH 100 EpO 146 EpO
+80 J/g +60 +40 +20 +0
25
solvation state minus the value of starting point, were -186.5, -202.6, -205.0, -
200.6, and -213.9 (J/g). This suggests that the system 146OH has minimum
immersion energy. Note that for graphene, the immersion energy would be
positive upon solvation.
4.3. The enthalpy The enthalpy of the system equals the internal potential energy U and the
smaller product of pressure and volume. As mentioned the driving forces for
solvation GO depend on the deferential enthalpy. A plot between methanol
contents layer and the immersion energy (Q) in figure 18 obviously shows the
enthalpy minima. Q can be define by the equation:
𝑄 = ⟨𝑈(𝑁)⟩ − ⟨𝑈(𝑁0)⟩ − (𝑁 − 𝑁0) 𝑈𝑏𝑢𝑙𝑘 (13)
Where, ⟨U(N°)⟩ is the average energy of the GO at some standard solvation state
N° (usually taken from the minimum, i.e. the most stable solvation state) and
(U) bulk = -10.432 KJ/mole is the mean bulk Methanol-Methanol interaction
potential energy, obtained from corresponding bulk simulations.
If the dry state is taken to be the reference state the immersion and solvation
energy will be related by the equation.
𝑄𝑑 = 𝑁 [𝛥 𝑈 (𝑁) − 𝑈𝑏𝑢𝑙𝑘] (14)
Where, d is the dry reference state of GO during immersion and the primary
difference between 𝑄𝑑 and ∆U is the factor N; the solvation energy is expressed
per mole of methanol while the immersion energy is expressed by J/g GO.
4.4. Interlayer distribution
The distribution of solution’s layers and interlayers indicates interaction
between the solvent and GO. As shown in Figure 12, the distribution of
interlayers for all the systems except 146EpO is single-peaked at two-layer
solute and splits into two smaller peaks at the four-layer solute etc. The splitting
of the peaks distribution relate to solvation of methanol in different GO
configuration.
Figure 19 shows the calculated potential energy for the different systems. The
data show that the potential energy is a function of the methanol content. The
changes of the curves (small plateaus) show swelling property of GO, especially
graphene system.
26
Figure 19. Graphene oxide potential energy from simulation, as a function of the number of
methanol content per GO layers.
The calculated relative changes of x and y dimension for the graphene, 146OH,
100OH 46EpO, 73OH 73EpO, 46OH 100EpO, and 146EpO systems are
presented in figures 20-25. All the plotted curves for the six systems are
between the number of methanol content per GO layers and the norm size. For
the graphene system shown in figure 20, and due to the plateaus of the curves,
the expansion or oscillation of the system before 3 methanol per unit cell, the
relative changes in x dimension (red) are higher than the y dimension (blue).
Figure 20. The norm. Size for the relative changes in x (red) and y dimension (blue), as a
function of the number of methanol molecules per GO unit cell for the graphene system.
0
500
1000
1500
2000
2500
3000
3500
4000
4500
5000
0 2 4 6 8
Pote
nti
al e
ner
gy [
KJ/
mo
l]
Num. of MET per GO unit cell
Graphene 146 OH100 OH 46 EpO 73 OH 73 EpO46 OH 100 EpO 146 EpO
0.999
0.999
0.999
1.000
1.000
1.000
1.000
1.000
1.001
0 1 2 3 4 5 6 7 8
No
rm. s
ize
Num. of MET per GO layer
Graphene rel_X_dim
Graphene rel_Y_dim
27
Figure 21. The norm. Size for the relative changes in x (red) and y dimension (blue), as a
function of the number of methanol molecules per GO unit cell for the 146OH system.
From figure 21, due to the plateaus and minimum points of the relative changes
in x and y dimension, the system oscillates regularly, except the last part after
about 5.5 number methanol content per GO layer, the relative changes in y
dimension (blue) are higher than x dimension.
Figure 22. The norm. Size for the relative changes in x (red) and y dimension (blue), as a
function of the number of methanol molecules per GO unit cell for the 100OH+46EpO system.
For the 100OH+46EpO system shown in figure 22, by comparing the relative
changes in x and y dimension. Due to the plateaus and minimum points, the
changes in x dimension are almost higher than y dimension during the
simulation process.
0.997
0.998
0.999
1.000
1.001
1.002
1.003
0 1 2 3 4 5 6 7 8
No
rm. s
ize
Num. of MET per GO layer
146 OH rel_X_dim
146 OH rel_Y_dim
0.997
0.998
0.999
1.000
1.001
1.002
1.003
0 1 2 3 4 5 6 7 8
No
rm. s
ize
Num. of MET per GO layer
100 OH 46 EpOrel_X_dim
100 OH 46 EpOrel_Y_dim
28
Figure 23. The norm. Size for the relative changes in x (red) and y dimension (blue), from
simulation with the number of methanol content per GO layers for the 73OH+73EpO system.
The relative changes in x and y dimension for the 73OH+73EpO system are
shown in figure 23. It is seen that the changes in x dimension (red) are higher
than y dimension during the time of simulations. After 3 methanol molecules
per GO unit cell, and due to the plateaus, the relative changes in y dimension are
a straight line, which means the changes are monotonically.
Figure 24. The norm. Size for the relative changes in x (red) and y dimension (blue), from
simulation with the number of methanol content per GO layers for the 46OH+100EpO system.
Figure 24 show that the relative changes in x and y dimensions for the
46OH+100EpO system starts at the same point, and the plotted data show
obvious differences in the y dimension (blue) after the first methanol molecule
per GO unit cell are higher until number 4.5 of methanol content, after that the
changes occur regularly.
0.997
0.999
1.001
1.003
1.005
1.007
1.009
0 1 2 3 4 5 6 7 8
No
rm. s
ize
Num. of MET per GO layer
73 OH 73 EpO rel_X_dim
73 OH 73 EpO rel_Y_dim
0.997
0.998
0.999
1.000
1.001
1.002
1.003
0 1 2 3 4 5 6 7 8
No
rm. s
ize
Num. of MET per GO layer
46 OH 100 EpO rel_X_dim
46 OH 100 EpO rel_Y_dim
29
Figure 25. The norm. Size for the relative changes in x (red) and y dimension (blue), from
simulation with the number of methanol content per GO layers for the 146EpO system.
The relative changes (oscillations) in x and y dimensions for the 146EpO system
is shown in figure 25, displaying changes in both dimensions. With note due to
the plateaus and minimum pointe (oscillation), it seems that the changes in x
dimension (red) are higher during simulation.
Thus, from the results of the simulated basal spacings, the relative changes in x
and y dimensions, solvation energy and the immersion energy, the latter
perhaps best illustrates the solvation behaviour, and potential energy for the six
different systems were calculated. From figure 18, among the GO systems, the
146EpO has minimum immersion energy. I think the curves in figure 18 show
the swelling property of graphene more obviously among the other systems. Due
to the curves carbon system does not like to solvate.
I think we could get better results by increasing the number of investigated
systems. Search for the better interlayer distance (d001) larger or smaller than
the one taken in this project (2.8 nm). Search for the most suitable size of
simulation’s box. In order to prevent or decrease number of interactions
between particles inside solvent with copies of neighbouring boxes (unite cells).
Moreover, a longer time step for integration will give better results.
5. Conclusion Simulation of six different systems of pure graphene as well as graphene oxide
(GO) layers (based upon a 384 atom graphene layer) denoted 146EpO,
46OH+100EpO, 73OH+73EpO, 100OH+46EpO, and 146OH were performed,
where the numbers denote the amount of hydroxyl groups or epoxide groups
per 384 carbon atoms. The graphene and GO layers were initially intercalated
with 7.5 methanol molecules per GO unit cell (defined with 8 C atoms), which
gradually were removed or ‘evaporated’ through 60 sequential simulations
0.997
0.998
0.999
1.000
1.001
1.002
1.003
0 1 2 3 4 5 6 7 8
No
rm. s
ize
Num. of MET per GO layer
146 EpO rel_X_dim
146 EpO rel_Y_dim
30
steps. The interpretation of the association of methanol with GO stems from the
presence of hydroxyl and epoxide groups.
The technique of MD simulations offers atomistic depiction of the GO oxide
hydroxyl and epoxide groups, with the possibility to simulate them separately.
The MD simulations allowed the determination of interlayers swelling without
competition of any other effects or changes in the graphene/GO layers
themselves. The subsequent post-analysis indicated an energetic origin for
discrete swelling in graphene and GO, providing a unique computational
verification for swelling of GO crystallite.
The swelling behaviour of GO will open new possibilities for intercalation or
functionalization of this material, using different types of molecules and
synthesis of new GO framework materials. Swelling is also a useful step for
manufacturing functionalized graphene that can potentially facilitate better
mechanical, thermal and electronically properties. The expansion of the GO
layers thus supplies an opportunity for applications such as molecular filters
with tuneable size, which adjusts to specific conditions during absorption
processes.
Since different types of GO were investigated, the sorption of solvent to the
different GO layers was as expected not the same. Accordingly, these results also
lead to new possibilities to separate GO structures by inserting some other
solvents. Thus, we explored the absorption or swelling properties of GO
intercalated with methanol for the first time. We hope these results will help
others in their research and serve the readers with new information about
applications in nanoscience.
Acknowledgements
First and foremost my sincerest acknowledgements goes to my supervisor,
Assistant Professor Michael Holmboe, for giving me the opportunity to work
with the project, who has always made time to answer all my questions. I am so
grateful to him for his kind guidance. I give my warmest thanks to my family, for
their consistent support especially my sons Dolovan, Bahdin and Matin, which
this work is dedicated to.
31
6. References
1. Hougardy, J. NMR study of adsorbed water. I. Molecular orientation and protonic motions in the two-layer hydrate of a Na vermiculite. J. Chem. Phys. 64, 3840 (1976).
2. Ketolainen, T., Havu, V. & Puska, M. J. Enhancing conductivity of metallic carbon nanotube networks by transition metal adsorption. J. Chem. Phys. 142, 54705 (2015).
3. Talyzin, A. V, Hausmaninger, T., You, S. & Szabó, T. The structure of graphene oxide membranes in liquid water, ethanol and water-ethanol mixtures. Nanoscale 6, 272–81 (2014).
4. Talyzin, A. V., Luzan, S. M., Szabó, T., Chernyshev, D. & Dmitriev, V. Temperature dependent structural breathing of hydrated graphite oxide in H2O. Carbon N. Y. 49, 1894–1899 (2011).
5. Klechikov, A., Yu, J., Thomas, D., Sharifi, T. & Talyzin, A. V. Structure of graphene oxide membranes in solvents and solutions. Nanoscale 7, 15374–84 (2015).
6. Chen, Y. et al. Aerosol synthesis of cargo-filled graphene nanosacks. Nano Lett. 12, 1996–2002 (2012).
7. Vorobiev, A. et al. Graphene oxide hydration and solvation: An in situ neutron reflectivity study. Nanoscale 6, 12151–12156 (2014).
8. Tummala, N. R. & Striolo, A. Role of counterion condensation in the self-assembly of SDS surfactants at the water-graphite interface. J. Phys. Chem. B 112, 1987–2000 (2008).
9. Klechikov, A. et al. Hydrogen storage in high surface area graphene scaffolds. Chem. Commun. 51, 3–6 (2015).
10. Korobov, M. V et al. Sorption of polar organic solvents and water by graphite oxide: Thermodynamic approach. Carbon N. Y. 102, 297–303 (2016).
11. Bi, H. et al. Spongy graphene as a highly efficient and recyclable sorbent for oils and organic solvents. Adv. Funct. Mater. 22, 4421–4425 (2012).
12. Kim, D. et al. Intercalation of gas molecules in graphene oxide interlayer: The role of water. J. Phys. Chem. C 118, 11142–11148 (2014).
13. Joshi, R. K. et al. Precise and ultrafast molecular sieving through graphene oxide membranes. Science 343, 752–754 (2014).
14. Gómez-Navarro, C. et al. Electronic transport properties of individual chemically reduced graphene oxide sheets. Nano Lett. 7, 3499–3503 (2007).
15. Smith, S. C. & Rodrigues, D. F. Carbon-based nanomaterials for removal of chemical and biological contaminants from water: A review of mechanisms and applications. Carbon N. Y. 91, 122–143 (2015).
16. Zhu, Y. et al. Graphene and graphene oxide: Synthesis, properties, and applications. Adv. Mater. 22, 3906–3924 (2010).
17. Dreyer, D. R., Park, S., Bielawski, C. W. & Ruoff, R. S. The chemistry of graphene
32
oxide. Chem. Soc. Rev. 39, 228–240 (2010).
18. Coff, C. Incorporation, Digestion and Incarnation. History 1–9 doi:10.1016/j.aqpro.2013.07.003
19. Botas, C. et al. The effect of the parent graphite on the structure of graphene oxide. Carbon N. Y. 50, 275–282 (2012).
20. You, S., Sundqvist, B. & Talyzin, A. V. Enormous lattice expansion of hummers graphite oxide in alcohols at low temperatures. ACS Nano 7, 1395–1399 (2013).
21. Åkesson, P. Molecular Dynamics of the Adsorption of Organic Molecules on Organic Substrates. 10 (2013).
22. Keilen, K. B. Polymer failure modeled by Molecular Dynamics. (2014).
23. Todorov, I. T., Smith, W., Trachenko, K. & Dove, M. T. DL_POLY_3: new dimensions in molecular dynamics simulations via massive parallelism. J. Mater. Chem. 16, 1911 (2006).
24. Zettervall, N. Multi-scale methods for stochastic. (2012).
25. Sattarnusart, W. Investment Real Options in Real Estate Development Investment Warut Sattarnusart. 66 (2012).
26. Parrinello, M. & Rahman, A. Polymorphic transitions in single crystals: A new molecular dynamics method. J. Appl. Phys. 52, 7182–7190 (1981).
27. Wohlert, J. Atomistic computer simulations of lipid bilayers. (2006).
28. Essmann, U. et al. A smooth particle mesh Ewald method. J Chem Phys 103, 8577–8593 (1995).
29. GROMACS (5.1.2) was used for all simulations. The energy minimizing by this method started with steepest descent (an extension of Laplace’s method for approximating an integral) algorithm for flexible and constrained bonds. It takes a step in the directio, 37. Molecular modelling of oxygen and water permeation in polyethylene. Polym. (United Kingdom) 54, 2988–2998 (2013).
30. Bussi, G., Donadio, D. & Parrinello, M. Canonical sampling through velocity rescaling. J. Chem. Phys. 126, (2007).
31. Juréus, A. & Langel, Ü. Galanin and galanin antagonists. Acta Chim. Slov. 43, 51–60 (1996).
32. Shih, C. J., Lin, S., Sharma, R., Strano, M. S. & Blankschtein, D. Understanding the pH-dependent behavior of graphene oxide aqueous solutions: A comparative experimental and molecular dynamics simulation study. Langmuir 28, 235–241 (2012).
33. Brandt, E. G. Molecular Dynamics Simulations of Fluid Lipid Membranes E r i k G . B r a n dt. (2011).
34. Iupinov, A. Implementation of the Particle Mesh Ewald method on a GPU. (2016).
35. Kholmurodov, K., Chulkova, A. & Yasuoka, K. Molecular Dynamics Study of the Effect of Electrostatic Interactions on the Biphenyl Structure in the Active NO 3 Solvent. 10–16 (2010).
33
36. Kenney, J. F., Kutcherov, V. a, Bendeliani, N. a & Alekseev, V. a. The evolution of multicomponent systems at high pressures: VI. The thermodynamic stability of the hydrogen-carbon system: the genesis of hydrocarbons and the origin of petroleum. Proc. Natl. Acad. Sci. U. S. A. 99, 10976–10981 (2002).
37. Wensink, E. J. W., Hoffmann, A. C., Van Maaren, P. J. & Van Der Spoel, D. Dynamic properties of water/alcohol mixtures studied by computer simulation. J. Chem. Phys. 119, 7308–7317 (2003).
38. Huang, Y., Nam, K. & Westlund, P.-O. The water R1(ω) NMRD profiles of a hydrated protein from molecular dynamics simulation. Phys. Chem. Chem. Phys. 15, 14089–97 (2013).
39. Steneteg, P. Development of Molecular Dynamics Methodology for Simulations of Hard Materials. (2012).
40. Young, P. The leapfrog method and other ‘symplectic’ algorithms for integrating Newtons’s laws of motion. Lect. notes Univ. california, St. cruz 1–15 (2009).
41. Gast, A. P. Boundary Conditions. Sci. Am. 306, 14–14 (2012).
42. M.J. Abraham, D. van der Spoel, E. Lindahl, B. Hess, and the GROMACS development team, GROMACS User Manual version 5.1.4, www.gromacs.org (2016).
43. Darden, T., York, D. & Pedersen, L. Particle mesh Ewald: An N⋅log(N) method for Ewald sums in large systems. J. Chem. Phys. 98, 10089 (1993).
44. Li, X. Applications of Molecular Dynamics in Atmospheric and Solution Chemistry. (2011).
45. Dauxois, T., Ruffo, S., Arimondo, E. & Wilkens, M. Dynamics and Thermodynamics of Systems with Long-Range Interactions: An Introduction. Lect. Notes Phys. 602, 1–19 (2002).
46. Temperature, M. dynamics simulations at constant pressure and/or. Molecular dynamics simulations at constant pressure and/or temperature. J. Chem. Phys. 72, 2384 (1980).
47. Martyna, G. J., Tuckerman, M. E., Tobias, D. J. & Klein, M. Explicit reversible integrators for extended systems dynamics. Mol. Phys. 87, 1117–1157 (1996).
48. Hartmann, C. & SCHÜTTE, C. A geometric approach to constrained molecular dynamics and free energy. Commun. Math. 1–20 (2005).
49. Andersson, K. Structure , Bonding and Chemistry of Water and Hydroxyl on Transition Metal Surfaces. Chemical Physics (2006).
50. Cheung, D. L., Anton, L., Allen, M. P. & Masters, A. J. Computer simulation of liquids and liquid crystals. Computer Physics Communications 179, 61–65 (2008).
51. Trokhymchuk, A. Computer simulations of liquid / vapor interface in Lennard-Jones fluids : 111, (1999).
52. Springer, P., Ismail, A. E. & Bientinesi, P. A scalable, linear-time dynamic cutoff algorithm for molecular dynamics. Lect. Notes Comput. Sci. (including Subser. Lect. Notes Artif. Intell. Lect. Notes Bioinformatics) 9137, 155–170 (2015).
34
53. Smith, D. E. Molecular computer simulations of the swelling properties and interlayer structure of cesium montmorillonite. Langmuir 14, 5959–5967 (1998).