PCCP

View Article Online

PAPER View Journal

Alkane separation using nanoporous

graphene membranes
Published on 11 November 2014. Downloaded by University of Guelph on 21/11/2014 15:38:39.

Cite this: DOI: 10.1039/c4cp02745k

Krzysztof Nieszporek* and Mateusz Drach

Received 23rd June 2014,
Accepted 10th November 2014
DOI: 10.1039/c4cp02745k
www.rsc.org/pccp
We studied the permeability of graphene sheets with designed nanopores using the classical molecular
dynamics. To characterize the energy profile for transmission we calculated the potential of the mean
force. A high selectivity for methane + butane mixture with the hydrogen-passivated pore diameter of
0.32 nm was found where the volume exclusion mechanism governs the separation process. In the case
of a slightly larger pore diameter of 0.64 nm the same alkane mixture separates completely
unexpectedly: a larger butane molecule permeates much faster than a small methane one. The blocking
eect of the permeation path by a larger mixture component when it worked like a cork was also
observed. This is a promising perspective for using graphene to design intelligent membranes which can
maintain a constant composition of mixtures in the permeable area.

1. Introduction apparatus to introduce nanopores into suspended graphene sheets.8

Recent development in material science has shown a rapid
growth of interest in carbon-based materials. Fullerenes,1 carbon
nanotubes2 and graphene3,4 open a new age in adsorption,
accumulation and separation processes. It is due to their unique
properties such as a highly ordered structure, chemical inertness,
electrical conductivity and high durability.
One of the potential applications of graphene is its use as
membranes for gas separation. Besides cryogenic distillation
and separation by adsorption processes, the membrane separation
is the cheapest technique which is commonly used in laboratories,
automotive services or aircraft on-board gas apparatus. It is due to
low energy consumption, separation at ambient temperature,
low space requirement, continuous separation process and simple,
modular design. The last mentioned advantage makes for easy
adjustment of the apparatus for separating a particular gas
mixture. Since the permeance of a membrane is inversely propor-
tional to its thickness,5 the use of porous graphene gives an
unprecedented opportunity to increase the separation eciency.
It is due to the fact that the graphene sheet is one-atom thick.
However, the perfect graphene sheet is impermeable to gases as
small as helium6,7 because high electron density of aromatic rings
prevents the particles from passing across these rings. Thus to
design a suitable graphene membrane, it is necessary to drill holes
of a proper diameter or develop a method for production of porous
graphene with a well-defined pore diameter. Very promising is
the application of focused electron beam irradiation in the TEM
Faculty of Chemistry, Department of Theoretical Chemistry,
Maria Curie-Skodowska University, pl. M. Curie-Skodowskiej 3, 20-031 Lublin,
Poland. E-mail: krzysn@hektor.umcs.lublin.pl; Tel: +48-815375537
The obtained graphene structures are characterized by high stability
in time and a lack of undesirable distortions. Bell et al.9 reported
another attractive method to produce porous graphene by using a
modified helium ion microscope for lithography. An example of
chemical synthesis of graphene with periodically missing phenyl
rings is the surface-assisted arylaryl coupling reaction.10
In this work, based on the classical molecular dynamics simula-
tions, the ability of porous graphene to separate various alkanes
mixtures was investigated. Previously, inter alia, De-en Jiang et al.11
studied the selectivity of porous graphene for the H2/CH4 mixture.
Blankenburg et al.12 studied the porous graphene selectivity of H2
and He among other atmospheric gases. Li et al.13 calculated the
selectivity of porous graphene for H2/CO, H2/CO2 and H2/CH4.
Hauser and Schwerdtfeger14 studied the capability of functiona-
lized graphene nanopores to eciently separate methane from air.
These studies show that graphene with designed nanopores is a
highly selective membrane for mixture separation. The properties
of such a membrane can be characterized by calculating its
selectivity for dierent mixtures. Here we performed simulations
to compute the potential of the mean force (PMF) for various sets of
alkane/diameter pore size to characterize the molecule permeation.
In the results we showed that the separation process by single-layer
graphene could be connected not only with the simple volume
exclusion mechanism. We examined how separation changes with
temperature, pore size and length of the hydrocarbon chain.
2. Calculation method
All calculations presented here were performed using the
Gromacs simulation suite1517 employing the OPLS all-atom

This journal is the Owner Societies 2014 Phys. Chem. Chem. Phys.

Paper
Published on 11 November 2014. Downloaded by University of Guelph on 21/11/2014 15:38:39.
Fig. 1 Hydrogenated graphene pores of the computational systems used in this paper. The notation H1 to H6 corresponds to the approximate number
of phenyl groups removed from the graphene sheet. Pore areas and diameters are denoted in the figure.
force field. This force field is used by other authors studying
graphene separation properties18,19 or effects occurring on
carbon surfaces.20 As molecular structures and parameters of
investigated alkanes were also taken from the OPLS-AA to verify
its reliability we calculated the potential of mean force for
system containing one methane molecule and graphene sheet
without a pore. The obtained equilibrium distance and binding
energy for CH4 adsorption on graphene are 0.3568 nm and
11.76 kJ mol1 respectively, and are in agreement with experi-
mental and theoretical data.21
Rigid graphene membranes with various nanopore diameters
were modeled, ranging from 0.32 to 0.76 nm (Fig. 1). Pore
diameters were obtained from the open pore area measurements
p
by using the well-known formula d 2 A=p. Pores were passi-
vated with hydrogen atoms and centered on the 4.96 nm 
4.79 nm graphene sheet. Bonds between carbon and hydrogen
atoms are flexible to allow possible deformations of the edge of the
nanopores. In the case it is not pointed out, the temperature was
maintained at 300 K using velocity rescaling with a stochastic
term22 with a time constant of 0.1 ps. The constant volume of
the system was imposed; the simulation box sizes are 5.1 nm 
4.9 nm  16 nm. The designed pores used during calculations are
shown in Fig. 1.
To figure out the changes of potentials connected to pore
gating we calculated the potential of the mean force (PMF) of
alkane permeation across the designed nanopore. It is crucial
for the understanding of the selectivity and flux of the designed
porous graphene.
To perform the umbrella sampling calculations the porous
graphene sheet was horizontally located at the position z = 8 nm.
Phys. Chem. Chem. Phys.
View Article Online
PCCP
The alkane molecule was inserted in vacuum upright above a
graphene in a distance of 3 nm, vertically aligned with the center of
the pore. Next a series of initial configurations was generated, each
corresponding to a dierent alkane center of the mass distance
(COM) from the COM of graphene sheet along a reaction coordi-
nate z with a step of 0.05 nm. These configurations were used for
the umbrella sampling windows. Next separate umbrella simula-
tions were made by using a harmonic potential (force constant
k1 = 500 kJ mol1 nm2) and 5 ns period of time (5  106 steps).
In some cases such a value of the force constant was too small
to push a big molecule eq. butane across a small pore eq. H1.
It resulted in some discontinuity of the potential of the mean force
(PMF) curves close to z = 0.
We note that our model of molecular deformation is some-
what biased since we only allow the rearrangement of CH
bonds, which covers only a fraction of the total deformation.
3. Results
The permeation of methane, ethane, propane, butane, isobutane
and their two-component mixtures was simulated using all
pore diameters shown in Fig. 1. Fig. 2 shows the snapshots of
independent pulling simulation (CH4/H1, the rate 0.01 nm s1,
k1 = 50 000 kJ mol1 nm2). Dierent configurations of hydrogen
molecules which passivate a nanopore can be seen. The flexibility
of CH bonds allows for the membrane leaflet asymmetry
i.e. different configurations on one side than on the other.
The leaflet eect can be shown more clearly by monitoring
the time dependence of the average angle of the CH bonds at
the edge of the nanopore and the normal of the graphene sheet
This journal is the Owner Societies 2014
 View Article Online

PCCP Paper

Fig. 2 Snapshots of pulling methane molecule across the graphene

nanopore H1.23 There can be seen deformations of the angles of CH
Published on 11 November 2014. Downloaded by University of Guelph on 21/11/2014 15:38:39.

bonds at the edge of nanopore.

Fig. 4 Schematic PMF profile characterizing alkane transport across the

graphene nanopore (here methane/H4, T = 300 K).

Fig. 3 Time evolution centered on the CH4 passing event across the H1
pore: j is the average angle between the CH bonds and the normal of
the graphene plane, RmetH is the distance between center of mass of CH4
and center of mass of hydrogen atoms passivating the graphene pore.

(see Fig. 3). At the time of methane permeation across the pore
H1, a large-amplitude angular jump is observed, and the
average angle between the graphene plane and CH bonds is
close to the right angle. This effect vanishes with the increasing
difference between the pore and molecule size and in the case
of methane disappears for the pore H5. In the case of small
pores and big molecules the assumption of rigid graphene may
lead to very unrealistic CCH bond angles. Thus the presented
results are rather qualitative.
Fig. 5 PMF profile of methane/H3 nanopore at T = 300 K.
The potential of mean force (PMF) was extracted by using
the Weighted Histogram Analysis Method (WHAM),24 included
in the GROMACS suite.1517 Fig. 4 shows a schematic profile suggests the influence of adsorption eects on the energy
determined during simulations. profile for transmission. Fig. 46 show how the energy profile
We observed the rapid change of PMF with a varying changes if the pore size gets smaller.
distance between an alkane molecule and a graphene sheet. At first, there is no transmission barrier at all (Fig. 4,
The potential well located at z E 0 nm corresponds to the methane at H4 i.e. big pore and small molecule). Simulta-
transition of a molecule across a nanopore and is connected neously the observed asymmetry of potential well is probably
with the strongest attractive interactions between the graphene connected with the leaflet eect of CH bonds at the edge of the
nanopore and the permeating molecule. The negative sign of nanopore. While considering methane transmission through
PMF value at z E 0 nm means that a system requires energy the H3 nanopore i.e. small molecule and medium nanopore
supply to separate the alkane from the graphene sheet and (Fig. 5) the energy barrier corresponding to methane permeation

This journal is the Owner Societies 2014 Phys. Chem. Chem. Phys.

Paper
Published on 11 November 2014. Downloaded by University of Guelph on 21/11/2014 15:38:39.
Fig. 6 PMF profile of methane/H1 at T = 300 K when the alkane got stuck
in the configurations close to z = 0 nm.
appears but it is still compensated through adsorption. As a
result, the double well is observed. Their asymmetry still testifies
the occurrence of leaflet effect. Finally, Fig. 6 shows the case of
very small pore and small molecule (methane and H1 nanopore).
In this case the energy barrier is huge and the complete profile
cannot be derived anymore from the umbrella sampling. Trans-
mission of methane is practically impossible with this pore size.
Undoubtedly, except pore size the kind of atoms used to
passivate the graphene pore as well as the length and flexibility
of bonds between carbons and passivating groups have an
influence on the energy profile for transmission. Here we used
hydrogen but probably in other cases such an eect will be
greater. According to the leaflet eect, if the flexibility of
passivating groups does not aect the energy profile of molecule
transmission, it has a symmetrical shape.
While the dierence between the pore diameter and alkane
size decreases, the simulation system comes to the point where,
for a given k1, it is not possible to push the molecule across the
pore. It concerns the alkane position close to z = 0. Such a case
can be recognized by analyzing the histograms of the configura-
tions within the umbrella sampling windows. When they do not
overlap, the alkane molecule gets stuck in a given umbrella
window and the potential force is too small to move the molecule
(see Fig. 6 as an example).
The performed calculations indicate that for k1 =
500 kJ mol1 nm2 the complete energy profile for trans-
mission can be derived from sampling in the following cases:

methane/H3H5 nanopores

ethane/H3H5 nanopores

propane/H4H6 nanopores

butane/H5H6 nanopores
In the other studied systems the dierences between the
alkane molecular size and the pore diameter are too small to
Phys. Chem. Chem. Phys.
View Article Online
PCCP
Fig. 7 Symmetrized PMF profile of methane/H2 at T = 300 K calculated
assuming k1 = 10 000 kJ mol1 nm2. Still the complete energy profile
cannot be derived (summit of the peak, configurations close to z = 0).
determine complete PMFs the energy barrier accompanying
molecule transmission is too high (e.g. Fig. 6).
To estimate the importance of neglected pore flexibility
(rigid graphene and CH bonds deformation allowed),
we separately calculated the PMF curve for CH4/H2 assuming
k1 = 10 000 kJ mol1 nm2 (Fig. 7). Such extremely high values
of the force constant makes it possible to calculate the energy
profile almost in the whole range of methane configurations.
It allows us to estimate the value of energetic barrier for CH4
transmission through the H2 nanopore, it is equal to approx.
70 kJ mol1 (E17 kcal mol1). On the other hand, ab initio
barrier energies for the same system calculated by Hauser and
Schwerdtfeger14 are, in the case of fully frozen geometry, equal
to 90.24 kJ mol1 (21.553 kcal mol1) and, allow full relaxation
of nanopore, 34.69 kJ mol1 (8.283 kcal mol1). It follows that
the pore relaxation reduces the barrier heights and our model
of molecular deformation is visibly biased. It seems that a valid
estimate of the energy can only be assumed in the cases where
the difference between the actual pore diameter and the
molecular size is larger than about 0.2 nm. In such cases a
relatively low value of the force constant k1 = 500 kJ mol1 nm2
is sufficient to determine the complete energy profile for alkane
transmission.
To study the permeability and selectivity of graphene sheets
with the designed nanopores we simulated the separation of
gaseous mixture composed of 500 + 500 alkane molecules.
We used the same box size with two graphene sheets located
at z = 4 nm and z = 12 nm and an imposed constant volume of the
simulation box. The gaseous mixture was located below and above
graphene (retentate area) and the permeate area is situated in
the range 4 nm o z o 12 nm (initially there is a vacuum).
This journal is the Owner Societies 2014

PCCP
Published on 11 November 2014. Downloaded by University of Guelph on 21/11/2014 15:38:39.
Fig. 8 Separation of the mixture CH4 (yellow) and C2H6 (orange) across the H4 graphene sheet. The image is rotated to 90 degrees.
Fig. 9 The time evolution of the number of molecules in the permeate
area for the nanopore H4 at T = 300 K and the alkane mixtures indicated in
the figure: solid line methane, dashed line second component.
Fig. 10 Temperature influence on the permeation rate of methane and butane across the H4 nanopore.
This journal is the Owner Societies 2014
View Article Online
Paper
Periodic boundary conditions are used to avoid problems with
boundary eects caused by the finite size of the simulation box.
The snapshot of the simulation box is shown in Fig. 8.
We considered the following mixtures: methane + ethane/
H4, methane + propane/H4, methane + butane/H4, methane +
isobutane/H4 and methane + butane/H6. Fig. 9 shows the time
evolution of the number of alkane molecules permeated across
the nanopore H4.
One can see in Fig. 9 that in the case of the H4 nanopore the
size exclusion mechanism probably governs the separation
process. With the increasing dierence in the molecular sizes
of separated mixtures, the size exclusion eect increases and
the mixtures CH4 + C4H10 can be practically totally separated.
Simultaneously, if the molecular size of the second mixture
component gets larger, the rate of methane permeation
decreases. It is particularly evident in the case of the mixtures
containing propane and butane and permeation times below
12 ns. Probably the cork eect is responsible for it. Such an
eect can be observed when the molecule size fits the nanopore
diameter. In that case a molecule can fall into the potential well
at the center of nanopore and close down the permeation path
for the second mixture component. In the case of methane/H4
and butane/H4, the potential wells determined from PMFs are
14 kJ mol1 and more than 21 kJ mol1 respectively, thus
Phys. Chem. Chem. Phys.

Paper
the butane molecule is more strongly bonded at the center of
pore rim.
Fig. 10 shows the influence of temperature on the time
evolution of the number of permeated molecules for CH4 +
C4H10/H4. It can be seen that the temperature increase causes a
rapid increase of the methane permeation rate. The number of
Published on 11 November 2014. Downloaded by University of Guelph on 21/11/2014 15:38:39.
Fig. 11 The influence of temperature on the average time spent by the
permeating butane molecule in the center of the nanopore H4 during
separation of the CH4 + C4H10 mixture.
Phys. Chem. Chem. Phys.
View Article Online
PCCP
Fig. 12 The time evolution of the number of alkane molecules in
the permeate area for the nanopore H6 and the mixture CH4 + C4H10 at
T = 300 K.
permeated butane molecules varies in the range 39 and rather
insignificantly depends on the system temperature.
The confirmation of the cork eect can be the average time
spent by a butane molecule in the pore center. Fig. 11 shows the
time evolution of the average butane distance from COM of
hydrogen molecules passivating the graphene pore centered at
the stuck event, temperatures in the range 150350 K. We
defined the permeation incident as the case in which the
distance between COMs of butane and hydrogen molecules
which passivate the nanopore is smaller than 0.05 nm.
Fig. 11 indicates that at the lowest temperature T = 150 K a
butane molecule spends much more time in the center of the
H4 nanopore than at temperature 350 K. For that reason
the number of permeated methane molecules is the smallest
at T = 150 K (Fig. 10).
A surprising case is found when the mixture consisting of
methane and butane is separated by the nanopore H6. Fig. 12
shows that the permeation rate of butane is much higher than
methane. Here the adsorption eect plays a crucial role: butane
adsorbed at graphene is oriented horizontally and its interaction
energy is much higher than that of methane. As a result, butane
blocks the access to the graphene sheet for methane molecules.
4. Conclusion
The application of graphene as the membrane for gas separa-
tion is much more puzzling than it seems. The shape of PMFs
shows that the leaflet asymmetry also exists in the case of
one-atom-thin porous membrane passivated with hydrogen.
While considering the separation of methane + butane by
porous graphene we showed that for small pore diameters the
sieving mechanism works as expected. But in the case of larger
This journal is the Owner Societies 2014

PCCP
pores butane transmission is preferred. Thus, besides the
simple volume exclusion mechanism, the separation by porous
graphene can be also controlled by blocking the access to the
nanopore of a given component or the cork eect. Due to
adsorption the permeating molecule can get stuck at the center
of nanopore and cause a decrease of flux of the second mixture
component. Although such an eect is reported in other
studies of molecules moving through pores and channels, it
deserves special attention. Hence adsorption accompanying
Published on 11 November 2014. Downloaded by University of Guelph on 21/11/2014 15:38:39.
separation by nanoporous graphene plays a very important role
which should not be underestimated.
Our paper represents an eort to demonstrate the possibility
of using porous graphene as an intelligent membrane for
mixture separation. The accurate design of the pore diameter
and the adequate choice of the method of pore functionaliza-
tion as well as temperature and pressure can create a great
perspective. It can be used as a membrane in fuel cells, gas
sensors or medical equipment which automatically doses the
amount of drug into the bloodstream. Of course, such outlooks
can be treated more or less realistically, and the obtained
promising results deserve further studies.
References
1 H. W. Kroto, J. R. Heath, S. C. OBrien, R. F. Curl and
R. E. Smalley, C60: Buckminsterfullerene, Nature, 1985, 318,
162163.
2 S. Iijima, Helical microtubules of graphitic carbon, Nature,
1991, 354, 5658.
3 R. P. Wesoowski and A. P. Terzyk, Pillared Graphene As a
Gas Separation Membrane, Phys. Chem. Chem. Phys., 2011,
13, 1702717029.
4 K. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, Y. Zhang,
S. V. Dubonos, I. V. Grigorieva and A. A. Firsov, Electric Field
Eect In Atomically Thin Carbon Films, Science, 2004, 306,
666669.
5 A. H. C. Neto, F. Gwinea, N. M. R. Peres, K. S. Novoselov and
A. K. Geim, The Electronic Properties of Graphene,
Rev. Mod. Phys., 2009, 81, 109162.
6 S. T. Oyama, D. Lee, P. Hacarlioglu and R. F. Paraf, Theory of
Hydrogen Permeability in Nonporous Silica Membranes,
J. Membr. Sci., 2004, 244, 4553.
7 J. S. Bunch, S. S. Verbridge, J. S. Alden, A. M. van der Zande,
J. M. Parpia, H. G. Craighead and P. L. McEuen, Imperme-
able Atomic Membranes From Graphene Sheets, Nano Lett.,
2008, 8, 24582462.
8 M. D. Fischbein and M. Drndic, Electron Beam Nanosculpting
of Suspended Graphene Sheets, Appl. Phys. Lett., 2008, 93,
113107113110.
9 D. C. Bell, M. C. Lemme, L. A. Stern, J. R. Williams and
C. M. Marcus, Precision Cutting And Patterning of Graphene
With Helium Ions, Nanotechnology, 2009, 20, 455301455305.
This journal is the Owner Societies 2014
View Article Online
Paper
10 M. Bieri, M. Treier, J. Cai, K. At-Mansour, P. Rueux,
O. Groning, P. Groning, M. Kastler, R. Rieger, X. Feng,
K. Mullen and R. Fasel, Porous Graphenes: Two-Dimensional
Polymer Synthesis With Atomic Precision, Chem. Commun.,
2009, 69196921.
11 D. Jiang, V. R. Cooper and S. Dai, Porous Graphene As The
Ultimate Membrane For Gas Separation, Nano Lett., 2009, 9,
40194024.
12 S. Blankenburg, M. Bieri, R. Fasel, K. Muellen, C. A.
Pignedoli and D. Passerone, Porous Graphene As an Atmo-
spheric Nanofilter, Small, 2010, 6, 22662271.
13 Y. Li, Z. Zhou, P. Shen and Z. Chen, Two-Dimensional
Polyphenylene: Experimentally Available Porous Graphene
as a Hydrogen Purification Membrane, Chem. Commun.,
2010, 46, 36723674.
14 A. W. Hauser and P. Schwerdtfeger, Methane-selective
nanoporous graphene membranes for gas purification,
Phys. Chem. Chem. Phys., 2012, 14, 1329213298.
15 H. J. C. Berendsen, D. van der Spoel and R. van Drunen,
GROMACS: A Message-Passing Parallel Molecular Dyna-
mics Implementation, Comput. Phys. Commun., 1995, 91,
4356.
16 E. Lindahl, B. Hess and D. van der Spoel, GROMACS 3.0:
A Package for Molecular Simulation And Trajectory Analy-
sis, J. Mol. Model., 2001, 7, 306317.
17 B. Hess, C. Kutzner, D. van der Spoel and E. Lindahl,
GROMACS 4: Algorithms for Highly Ecient, Load-
Balanced, and Scalable Molecular Simulation, J. Chem. Theory
Comput., 2008, 4, 435447.
18 H. Liu, S. Dai and D. Jiang, Insights into CO2/N2 Separation
Through Nanoporous Graphene from Molecular Dynamics,
Nanoscale, 2013, 5, 99849987.
19 P. Lazar, F. Karlicky, P. Jurecka, M. Kocman, E. Otyepkova,
K. Safarova and M. Otyepka, Adsorption of Small Organic
Molecules on Graphene, J. Am. Chem. Soc., 2013, 135,
63726377.
20 G. Gianese, V. Rosato, F. Cleri, M. Celino and P. Morales,
Atomic-Scale Modeling of the Interaction Between Short
Polypeptides and Carbon Surfaces, J. Phys. Chem. B, 2009,
113, 1210512112.
21 O. O. Adisa, B. J. Cox and J. M. Hill, Modeling the Surface
Adsorption of Methane on Carbon Nanostructures, Carbon,
2011, 49, 32123218.
22 G. Bussi, D. Donadio and M. Parrinello, Canonical Sampling
Through Velocity Rescaling, J. Chem. Phys., 2007, 126,
014101.
23 W. Humphrey, A. Dalke and K. Schulten, VMD: Visual
Molecular Dynamics, J. Mol. Graphics, 1996, 14, 3338.
24 J. S. Hub, B. L. de Groot and D. van der Spoel, G_wham-A
Free Weighted Histogram Analysis Implementation Including
Robust Error and Autocorrelation Estimates, J. Chem. Theory
Comput., 2010, 6, 37133720.
Phys. Chem. Chem. Phys.