You are on page 1of 8

Article

pubs.acs.org/Langmuir

Mechanisms of Gas Permeation through Single Layer Graphene


Membranes
Lee W. Drahushuk and Michael S. Strano*
Department of Chemical Engineering, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, United States

ABSTRACT: Graphene has enormous potential as a unique


molecular barrier material with atomic layer thickness, enabling
new types of membranes for separation and manipulation.
However, the conventional analysis of diusive transport
through a membrane fails in the case of single layer graphene
(SLG) and other 2D atomically thin membranes. In this work,
analytical expressions are derived for gas permeation through
such atomically thin membranes in various limits of gas
diusion, surface adsorption, or pore translocation as the rate-
limiting step. Gas permeation can proceed via direct gas-phase interaction with the pore, or interaction via the adsorbed phase on
the membrane exterior surface. A series of van der Waals force elds allows for the estimation of the energy barriers present for
various types of graphene nanopores. These analytical models will assist in the understanding of molecular dynamics and
experimental studies of such membranes.

INTRODUCTION
Graphene is interesting because of its unique electronic
of the membrane that is orthogonal to the transport direction is
almost always neglected in the analysis of conventional
properties, strong mechanical strength, and nanoscale dimen- membranes. In a recent MD study of H2/N2 separation via a
sions.1,2 Single layer graphene consists of a monolayer of graphene membrane, it was shown that, for certain pore sizes,
covalently bonded sp2 hybridized carbon atoms in the the nitrogen permeation rate could exceed that of hydrogen
conventional hexagonal planar array, creating a 2D electronic due to increased adsorption of the larger molecules.15 This is
system. There is interest in using single layer graphene (SLG) contrary to the expectation that the smaller molecules should
and other graphene derivatives3 as membranes for a variety of have a higher permeation rate. However, as we show in this
applications. Pristine SLG is impermeable,4 but can function as work, the external surface is expected to dominate molecular
a membrane after the layer is suitably treated to create atomic transport in many cases involving SLG membranes. Other
scale pores in the otherwise crystalline carbon lattice. DNA has molecular dynamics simulations have also shown an importance
been observed to pass though graphene pores, which is a rst of the adsorbed phase.19,23 In this work, we derive analytical
step toward potential application to DNA sequencing.57 There expressions for several relevant limiting cases of gas permeation
are also simulations of SLG interacting with DNA,8 water,9 through SLG membranes involving both strongly and weakly
ions,1012 and gases.1325 Recent theoretical work has adsorbing gases. These mechanisms are then used to predict
suggested such membranes could have large separation factors the results of a separation of hydrogen and nitrogen for pores of
for gas separations.13,16,17,24 Some have used theory to explore various sizes. We show that for larger pores, the more strongly
tunneling eects with graphene membranes to achieve isotopic adsorbing nitrogen permeates faster than weakly adsorbing
separations.14,18,2022 Theoretical work has also shown that hydrogen. We also show and explain that the permeation rate is
graphene membranes could act as membranes for desalina- expected to be relatively constant across various larger
tion.11 However, experimental realization of SLG membranes nanopores in SLG membranes.


as gas separation barriers is in its infancy. Recent experimental
work with micrometer-sized membranes has shown that pores RESULTS AND DISCUSSION
in SLG and bilayer graphene are capable of molecular sieving
gases on the basis of molecular size.26 Also, a millimeter scale There are several analogies between molecular permeation
membrane has been demonstrated in aqueous phase membrane through an SLG membrane and catalytic chemical reaction of a
experiments.27 molecule at an active site on the surface. Both can proceed by a
Graphene represents a fundamentally new type of membrane preliminary adsorption step to the surface, followed by surface
active layer, because it is only a single carbon atom thick. While diusion to the active site (or open pore) or by direct
a typical membrane possesses a mass transport resistance that interaction of the molecule with the active site from the gas
scales inversely with thickness, single layer graphene represents
the extreme limit of negligible thickness. Hence, continuum Received: August 28, 2012
equations that are frequently employed for membrane analysis Revised: October 24, 2012
do not apply to the case of SLG. Moreover, the external surface Published: October 26, 2012

2012 American Chemical Society 16671 dx.doi.org/10.1021/la303468r | Langmuir 2012, 28, 1667116678
Langmuir Article

phase (the EleyRideal mechanism in catalysis). Adsorption the molecular regime and Poisseuille ow is negligible. The
into and transport across the graphene nanopore can be solution for gas transport through an innitely thin orice was
controlled by an activation energy if the constriction is small described by Knudsen28 and is simply the impingement rate, n,
enough, analogous to the activation energy for chemical upon the area of the pore given by eq 1. The inclusion of
reaction at a catalytic active site. Because of this analogy, it Arrhenius dependence in eq 2 gives the ux, Ji,gas, accounting
will be useful to label the permeating species on the high for the energy barrier the molecule experiences passing through
pressure side the membrane as A, and the species on the low the pore.
pressure side R, with a transition from AR corresponding to
P
spatial transport through the SLG membrane along the n = A
chemical potential gradient through the pore. Similarly, the 2MRT (1)
pore will be considered a species and labeled as B.
Consider a single layer graphene (SLG) membrane 1 E / RT
Ji ,gas = CBsB e a,gas,i pi
separating two gas reservoirs held at pressures PH and PL, 2MRT (2)
with P = PH PL; the partial pressure of species 1 in the two
reservoirs will be called pA and pR, respectively. The SLG sB is the surface area per pore, M is the molecular weight, and
contains CB pores/area of radius rB equally spaced at a pitch of Ea,gas,i is the energy barrier height relative to the gas phase.
w nm in a 2D rectangular lattice, and the surface density of Adsorbed Phase Pathway. In the adsorbed phase
available adsorption sites, or the surface density of a monolayer pathway, shown by Figure 2, the species passes through four
of adsorbed gas, is CT. CT is assumed to be equivalent for all gas states on the surface of the graphene and ve individual steps.
species. As the high pressure and low pressure surfaces are two First, a gas-phase molecule adsorbed to a site (step 1) on the
faces of the same SLG, the density of pores, CB, and the density surface. Second, the molecule diuses to a pore (step 2) and
of possible adsorption sites, CT, is identical for both sides. On associates into a potential well positioned in the space above
the high pressure side of the SLG, a surface site that is occupied the pore. Third, the molecule associated over the pore passes
by an adsorbed gas molecule of the ith species is labeled as Xm,i, though the pore (step 3) to be associated with the pore on the
and an empty area of the same size is considered an unoccupied downstream side of the graphene. Fourth, the molecule
site and labeled as l. To distinguish between the two faces of disassociates from the area above the pore (step 4) onto the
downstream surface of the graphene. Fifth and nally, the
the SLG, on the low pressure side, occupied and unoccupied
molecule desorbs from the surface (step 5) and enters the gas
sites are labeled as Xm,i and m, respectively. We neglect the
phase on the downstream side of the graphene. For simplicity,
inuence of the hypothetical support underneath the SLG
we will derive analytical solutions for the limits of this model
membrane as having much lower mass transport resistance than
assuming only a single adsorbing component. The correspond-
the SLG membrane. We assume that the pore is symmetrical
ing results for multiple adsorbing species are given at the end. A
with respect to the plane of the graphene. In Figure 1, the pore
previous analytical model for the adsorbed phase was applied to
depicted is model pore type a from Figure 3.
the limit of maximum porosity.23 Table 1 summarizes the steps
Gas-Phase Pathway. For steric selectivity in gas separation,
in reaction kinetics notation; the rate constants for the forward
pores in graphene membranes should have dimensions near
and reverse for the ith step are given by ki and ki, respectively.
and below 1 nm. This is small enough that transport occurs in Table 4 summarizes the variables and constants used in these
derivations.
The denitions for equilibration of each of the ve
mechanistic steps are given in eqs 37. The denition of the
third equilibrium constant as unity is a result of the assumption
that the pore is symmetric about the plane of the graphene.
k1 [Al ]
K1 = =
k 1 pA [l] (3)

k2 [ABl ][l]
K2 = =
k 2 [Al ][Bl ] (4)

k3 [ABl ]
K3 = = =1
k 3 [RBm] (5)

k4 [R m][Bm]
K4 = = = 1/K 2
Figure 1. Schematic for a SLG gas separation membrane that divides a k 4 [RBm][m] (6)
reservoir of gas at high pressure PH containing species at partial
pressures pH,i and a reservoir at low pressure PL with partial pressures k5 p [m]
pL,i. The primary gas component is labeled A on the high pressure side K5 = = R = 1/K1
and R after permeation through to the low pressure side. The k 5 [R m] (7)
membrane contains CT total number of sites of gas adsorption on the
SLG surface per area, with unoccupied sites l and those occupied Xl,i A site balance on the surface area of the graphene, accounting
for species i for the high pressure side. The low pressure side has for area occupied by pores, yields eq 8, where sA is the surface
unoccupied sites m and occupied sites Xm,i for species i. Pores of radius area per adsorbed species. Similarly, a balance on the number of
rp have a pitch of w in a square tiling. free and occupied pores yields eq 9.
16672 dx.doi.org/10.1021/la303468r | Langmuir 2012, 28, 1667116678
Langmuir Article

Figure 2. Nitrogen passing through a decavacancy pore. (a) Top view. (b) Side view. (c) Calculated vdW energy coordinate. Interaction energy
calculated with Lennard-Jones potential.15,29

Table 1. Full Model of Five Elementary Steps of Adsorbed Phase Pathway and Corresponding Rate of Each Step
step no. step rate
1 adsorption k1 k1pA [l] k 1[Al ]
l + A Al
2 association k2 k 2[Al ][Bl ] k 2[ABl ][l]
Al + Bl ABl + l
3 translocation k3 k 3[RBm] k 3[ABl ]
ABl RBm
4 dissociation k4 = k2 k4[RBm][m] k 4[R m][Bm]
RBm + m HoooooooI R m + Bm
5 desorption k5= k1 k5[R m] k 5pR [m]
R m HoooooooI m + R

CBsB With these results, it is possible to describe the surface


C T = 1/sA = [l] + [Al ] = [m] + [R m] concentrations in terms of equilibrium concentrations and the
sA (8)
bulk pressure. Equations 1114 describe the surface in
CB = [Bl ] + [ABl ] = [Bm] + [RBm] (9) equilibrium with the upstream surface of graphene. As the
graphene surfaces (H and L) are symmetric, equivalent
By applying the surface site balance eq 8 along with the expressions can be written for surface concentrations in
equilibrium denition for the adsorption eq 3, the equilibrium equilibrium with the downstream bulk. To derive tractable
surface concentrations can be derived as in eqs 11 and 12. This expressions for the rates of permeation, we will derive limits
is the result using the Langmuir model of adsorption, which assuming various rate-limiting steps, with all other steps in
assumes adsorption is limited to a single monolayer, the energy equilibrium.
of adsorption is independent of coverage, and that all Assuming Step 1, Adsorption to the Graphene
adsorption sites are equivalent. It is often convenient to use Surface, Is Rate Limiting. The resulting expression for
the variable A, the fraction of monolayer coverage, given by eq adsorption to the graphene surface rate limiting is eq 16.
10.
J1 = k1pA [l] k 1[Al ]
K1pA
A = J1 = k1pA ([m]) k 1([R m])
1 + K1pA (10) (15)

K1pA C C Kp
[Al ] = C TA = C T J1 = k1pA T k 1 T 1 R
1 + K1pA (11) 1 + K1pR 1 + K1pR (16)
1 Assuming Step 2, Association to a Pore Site, Is Rate
[l] = C T(1 A) = C T
1 + K1pA (12) Limiting. The resulting expression for association to a pore site
rate limiting is eq 18.
If equilibrium is achieved for surface species associating to
the pore, then eqs 9, 11, and 12 can be used to derive the J2 = k 2[Al ][Bl ] k 2[l][ABl ]
number of occupied and unoccupied pore sites. These results
are given by eqs 13 and 14. J2 = k 2[Al ]((CB [RBm])) k 2[l]([RBm]) (17)
CB(1 A)
[Bl ] = CB(1 R )
(1 + (K 2 1)A) (13) J2 = k 2C TA k 2C T(1 A)
(1 + (K 2 1)R )
CBK 2A CBK 2R
[ABl ] =
(1 + (K 2 1)A) (14) (1 + (K 2 1)R ) (18)

16673 dx.doi.org/10.1021/la303468r | Langmuir 2012, 28, 1667116678


Langmuir Article

Figure 3. Six model pores formed by removing (a) 10, (b) 11, (c) 12, (d) 13, (e) 22, and (f) 32 carbon atoms from a graphene lattice, along with
energy coordinate for permeation through the pore (see Figure 2) for both hydrogen and nitrogen.

Table 2. Summary of Activation Energies for the Six Pores Shown in Figure 3a
H2 (eV) N2 (eV)
pore Ea,2 Ea,3 Ea,4 Ea,5 Ea,gas Ea,2 Ea,3 Ea,4 Ea,5 Ea,gas
a 0.017 0.019 0.017 0.076 0.029 0.344 0.028 0.141 0.205
b 0.021 0.002 0.021 0.076 0.036 0.165 0.032 0.141 0.028
c 0.020 0.024 0.076 0.035 0.053 0.028 0.141
d 0.021 0.021 0.076 0.036 0.054 0.141
e 0.022 0.008 0.076 0.039 0.021 0.141
f 0.033 0.001 0.076 0.057 0.009 0.141
a
Energies are calculated via Lennard-Jones potential.15,29

Assuming Step 3, Translocation through a Pore, Is CBK 2A


Rate Limiting. We assume that the association of a species on J4 = k4C T(1 R ) k 4R
(1 + (K 2 1)A)
one face of the pore is independent of the state of the other
face. We have thus far not distinguished pores that have a CB(1 A)
molecule trapped on both sides of the pore entrance. If there (1 + (K 2 1)A) (23)
are molecules on both sides of the pore, they may inhibit
passage. The subsequent equations will show that accounting Assuming Step 5, Desorption from the Graphene
for this blocking results in the same overall rate for a single Surface, Is Rate Limiting. The resulting expression for
species. adsorption to the graphene surface rate limiting is eq 16.
J5 = k5[R m] k 5pR [m]
J3 = k 3[ABl ] k 3[RBm] (19)
J5 = k A([Al ]) k 5pR ([l]) (24)
First, modify eq 19 to subtract any pores that have molecules
on both faces of the pore from the rate expression. By then C TK1p C
employing the fact that the forward and reverse rate constants J5 = k5 A
k p T
are identical for this step, eq 20 shows that the overall rate is 5 R
1 + K1pA 1 + K1pA (25)
the same.
The above derivation assumes that there is only a single species.
It is possible to extend these results to multiple species,
assuming that all species have the same rate-limiting step. The
resulting expressions are given by eqs 2731.
K1, i pH/L, i
The resulting expression for translocation through the pore as H/L, i =

rate limiting is eq 21. 1 + K1, j pH/L, j (26)
CBK 2A CBK 2R C K p
J3 = k 3 k3 CT
J1 = k1pA k
T 1, A R
(1 + (K 2 1)A) (1 + (K 2 1)R ) (21)
1 + K p 1
1 + K p
1, j L, j 1, j L, j (27)
Assuming Step 4, Disassociation from the Pore Site, Is
Rate Limiting. The resulting expression for disassociation CBA(1 H, i)
from the pore site rate limiting is eq 23. J2 = k 2C T k 2C T
(1 + (K 2, i 1)H, i)
J4 = k4[m][RBm] k 4[R m][Bm] CBK 2, AR(1 H, i)
J4 = k4[m]([ABl ]) k 4[R m]([Bl ]) (22) (1 + (K 2, j 1)L, j) (28)

16674 dx.doi.org/10.1021/la303468r | Langmuir 2012, 28, 1667116678


Langmuir Article

Table 3. Comparison of the Equations Derived for Each Rate Limiting Step to the Numerical Result of the Full Model for Pores
(a) and (c) from Figure 3a
Ea (eV) ki N2 permeation rate (mol/m2/s)
limit pore (a) pore (c) pore (a) pore (c) pore (a) pore (c)
1: eq 27, adsorption to surface 1 106 1/Pa/s 1 106 1/Pa/s 2.5 105 2.5 105
2: eq 28, association to pore 0.029 0.035 1 1017 m2/mol/s 7 1016 m2/mol/s 3.2 104 2.5 104
3: eq 29, translocation through pore 0.344 0.053 2 107 1/s 1 1012 1/s 4.9 101 3.4 104
4: eq 30, dissociation from pore 0.028 0.028 1 1017 m2/mol/s 1 1017 m2/mol/s 3.2 104 2.8 104
5: eq 31, desorption from surface 0.141 0.141 4 1010 1/s 4 1010 1/s 1.5 105 1.8 105
numerical solution 4.9 101 8.7 103
a
Assuming single species, nitrogen, P_up = 0.2 MPa, P_down = 0.0 MPa, and K1 = 3 1/MPa.30

CBK 2, AA For the forward rate constant of the second step, the 2D
J3 = k 3 ideal gas collision theory can provide an estimation, given by eq
(1 + (K 2, i 1)H, i)
33, where Ps is a steric factor between 0 and 1, AB is the sum of
CBK 2, AR the two radiuses, and and AB is the moment that simplies to
k3
(1 + (K 2, j 1)L, j) the mass of the gas molecule in this case. For nitrogen and a 1
(29)
nm pore, this results in a pre-exponential factor of 1017 m2/s/
mol. The equilibrium coecient for the third reaction is 1 as
CBK 2, AA(1 L, j) the membrane is symmetric.
J4 = k4C T k 4C TR
(1 + (K 2, i 1)H, i)
kBT E
CB(1 L, j) k = 2PsAB exp a
2AB RT (33)
(1 + (K 2, j 1)L, j) (30)
Using these models, we estimate nitrogens parameters for
C K p two of the pores, (a) and (c), in Figure 3 that were used in a
CT recent molecular dynamics simulation examining H2 and N2.15
J5 = k5 k 5p
T 1, A A
R As a simplifying assumption, we will set the value of equilibrium
1 + K1, i pH, i 1 + K1, i pH, i (31)
constant to 1 for the second and fourth steps, which are the
Model Pores. We examine six model pores, along with their association and dissociation of a molecule on the surface to the
van der Waals energy coordinates for nitrogen and hydrogen pore. We expect the true value of the step 2 equilibrium
molecules. The pores and energy coordinates are depicted in constant to be less than 1, as a molecule suers an entropic loss
Figure 3; the values for activation energies for the various steps by associating to the pore from the 2D ideal gas, in addition to
depicted in Figure 3 are summarized in Table 2. Without edge a slight potential energy dierence between those two states. A
termination, these model pores are unlikely to be stable in a real value of 105 mol/m2 is used for CT, from the inverse of
graphene sheet. However, we will use these simple pores to nitrogens surface area per molecule. A value of 8.8 108 mol/
represent real pores over a range of sizes. m2 is used for CB to match the setup of literature molecular
Model Parameters. We have presented a model for the dynamic simulation.15 For comparison, a recent ion beam
permeation through graphene membranes via an adsorbed exposure technique achieved a nanopore density of 8 109
phase, and presented equations describing the permeation rate mol/m2 in SLG.32 The results modeling single species N2
for various limits of our model. To make use of the equations, permeation are given in Table 3.
estimates for the various parameters are necessary. Without In addition to the analytical results made by assuming a rate-
intensive eort, it is possible to make estimates that should be limiting step, it is also possible to generate a steady-state
within a few orders of magnitude and can provide insight on numerical solution by solving for the dierential equations
which of the steps is most likely to be limiting. derived from the rate expression for the ve individual steps
Adsorption studies have not been focused on a single (see Table 1). For pore (a) in Table 3, the limit corresponding
monolayer of graphene; however, there are ample studies of to the translocation through the pore (step 3) perfectly
adsorption to carbon nanotubes and graphite, which would be matches the numerical solution result, indicating that step three
expected to be reasonable analogues for monolayer graphene. is rate limiting. For pore (c) in Table 3, no equation perfectly
The adsorption equilibrium properties of some gases on matches the numerical solution, and multiple limits provide
monolayer graphene have also been calculated.30 A simple rates of a comparable order of magnitude. This indicates that
kinetic model for the desorption is given by eq 32.31 One no single step is rate limiting, and while none of the analytical
interpretation of this estimation is the vibration frequency solutions apply exactly, they do still provide an answer within
normal to the surface; therefore, this rough estimation can also an order of magnitude of the numerical result.
be applied for the rate at which a species associated directly The conditions used here, while not exactly the same, were
above the pore will pass over the energy barrier through the similar to a previous molecular dynamics study.15 The nitrogen
pore. An order of magnitude estimate for the terms in the pre- permeation seen in the original MD results15 is several orders
exponential factor is 1013 1/s.31 of magnitude smaller for both pores considered. In general, the
simple models used here carry a higher level of uncertainty than
kBT q E a more detailed investigation. Ignoring steric factors for all
k= exp a reaction rate constants is one reason the rate constants here are
h qa RT (32) likely to be overestimates.
16675 dx.doi.org/10.1021/la303468r | Langmuir 2012, 28, 1667116678
Langmuir Article

Table 4. Description of All Variables and Constants Used in Figure 3. For the smallest pore, (a), the separation favors
the Derivation of the Model and Limits hydrogen because of the high energy barrier for passing though
the membrane (step 3). However, for larger pores, nitrogen is
PH total pressure on high pressure side
favored over the smaller hydrogen, because the nitrogen
pH,i partial pressure of ith species on high pressure side
absorbs more strongly to the graphene surface. Furthermore,
pA partial pressure of species i = 1 on high pressure side
for the larger pores, (c)(f), the absolute permeation rates are
PL total pressure on low pressure side
relatively constant. This is because the molecules no longer
pL,i partial pressure of ith species on low pressure side
experience a signicant energy barrier for passing though the
pR partial pressure of species i = 1 on low pressure side
plane of the membrane (step 3), but they still experience a
CT total surface sites per area
relatively constant energy barrier for entering the pore (step 2)
[l] surface concentration of empty sites on high pressure side of SLG
and/or leaving pore (step 4), as shown in Table 2.
[Xl,i] surface concentration of ith species on high pressure side of SLG
Implications. The gas-phase pathway, in which a molecule
[Al] surface concentration of species i = 1 on high pressure side of SLG
in the gas phase strikes the pore, is perhaps the more intuitive
[m] surface concentration of empty sites on low pressure side of SLG
mechanism for transport through a graphene membrane.
[Xm,i] surface concentration of ith species on low pressure side of SLG
However, the adsorbed phase pathway has some important
[Rm] surface concentration of species i = 1 on low pressure side of SLG
w pitch of pores in SLG for square tiling
implications on the performance of graphene membranes.
n impingement rate at a surface from gas phase
One implication is that the permeation rate, while linear at
A active area for impingement from gas phase
low pressures, may not be so at pressures high enough to
P pressure drop from high pressure side to low pressure side
saturate the surface. For the adsorbed phase pathway, the
M molecular weight
permeation rate in the expected limits becomes independent of
R ideal gas constant
pressure at higher pressures. In addition, at higher pressures,
T temperature
the pores may be saturated with adsorbed molecules that block
Ji,gas ux of ith species through a graphene membrane via gas-phase
gas-phase molecules incident on the pore. Potentially, there
impingement could be an upper limit to permeation rate; however,
CB surface density of pores unaccounted eects, such as coverage over a monolayer, may
sB surface area per pore somewhat alter such simple behavior at high pressures.
Ea,gas,i activation energy for ith species to pass through the pore from the Another important implication of the adsorbed phase model
gas phase is that the observed permeation rate of a given species is
[Bl] surface concentration of pores that are empty on the high pressure inuenced by the partial pressures of all of the components of
side
the mixture. This is in contrast to the gas-phase pathway, in
[ABl] surface concentration of pores with species i = 1 associated on the
high pressure side which the permeation rate for a given species is only dependent
[Rm] surface concentration of pores that are empty on the low pressure on its own partial pressure. This distinction can provide a
side simple ngerprint for identifying whether or not the adsorbed
[RBm] surface concentration of pores with species i = 1 associated on the phase pathway is signicant in future experimental graphene
low pressure side membranes. Even gas-phase molecules that are too large to pass
kj forward rate constant of the jth step, for species i = 1 through the pores in the membrane can adsorb to the surface
kj reverse rate constant of the jth step, for species i = 1
and alter the permeation rate of smaller species.
Kj equilibrium constant of the jth step, for species i = 1
Which path dominates depends on the adsorption character-
sA surface area per adsorbed species i = 1
istics of the gases and the energy barrier of the pores. Smaller
A fraction of monolayer coverage on high pressure side of SLG
dened by Langmuir adsorption model, for species i = 1 pores with higher energy barriers are most likely to give high
R fraction of monolayer coverage on low pressure side of SLG dened separation factors. Because a molecule in the adsorbed phase
by Langmuir adsorption model, for species i = 1 starts at a lower potential energy, the height of the energy
Jj ux via adsorbed phase assuming jth step is rate limiting, for species barrier it must overcome is necessarily greater than that of a
i=1 molecule in the gas phase. This makes smaller pores with high
H/L,i fraction of monolayer coverage on high/low pressure side of SLG energy barriers to be more likely dominated by the gas-phase
dened by Langmuir adsorption model, for ith species
pathway. However, the adsorbed phase may still play a role by
Kj,i equilibrium constant of jth step, for ith species
Ea,j activation energy for jth step
blocking the pore from incident gas-phase molecules. Larger
pores are more likely to have smaller energy barriers that enable
the adsorbed phase to contribute signicantly to permeation.
Our analysis demonstrates that the most signicant steps However, for large enough pores, it is less likely that the
inuencing transport are 2, 3, and 4 as these are most likely to adsorbed phase will block the pore, so both pathways will be
be rate limiting. For H2/N2 separation, the simulation in ref 15 present. Depending on the graphene synthesis and membrane
demonstrates that, for large pores (1632 atom vacancies), the fabrication techniques, experimental graphene membranes may
nitrogen permeation rate is nearly constant, while hydrogen have a distribution of dierent pore sizes. Of the early
permeation increases linearly with pore area. This indicates that experimental realizations of SLG and few layer graphene
the permeation of nitrogen is dominated by the adsorbed phase membranes, one is thought to be dominated by small,
pathway and limited by association to the pore (step 2), and subnanometer, UV-induced oxidative etched pores,26 while
hence a larger pore has little eect. Hydrogen permeation, on another is thought to have intrinsic pores over a range of sizes,
the other hand, is dominated by the gas-phase pathway and has 115 nm.27 In general, the pathway having the least mass
the expected linear increase with the pore area. transfer resistance will dominate, but the modeling included in
Figure 4 shows the results for the full multicomponent model this work allows for a combination of adsorbed phase and gas-
predictions for the separation of an equimolar mixture of phase dominance when the respective pore populations are
hydrogen and nitrogen for all six model pores depicted in balanced.
16676 dx.doi.org/10.1021/la303468r | Langmuir 2012, 28, 1667116678
Langmuir Article

Figure 4. (a) Predicted mixture separation factors from full numerical computation and (b) total permeation rates for a hydrogen/nitrogen
separation using the six model pores depicted in Figure 3. Permeation rates and according separation factors are the sum of the gas-phase
permeation, calculated using eq 2, and the permeation from the adsorbed phase pathway calculated from the multicomponent numerical solution of
the process described by Table 1. Modeled equimolar mixture of hydrogen and nitrogen with T = 298 K, total upstream pressure set at 2 atm, and
downstream set as vacuum.

CONCLUSION
A SLG membrane has been shown to be capable of high
rD =
Di
k A(1 + KApA ) (37)
selectivity and high permeance in a way that is not described by
conventional membrane diusion treatments due to its atomic If the spacing between pores is larger than the length scale
layer thickness. Instead, we apply a surface science approach to for diusion, then the dierential equation can be solved with
successfully derive analytical expressions for the gas and an analytical solution. The apparent rate constant is dened
adsorbed phase uxes through several classes of model with eq 38. The solution to eq 34, written in terms of modied
graphene pores under dierent rate-limiting steps. The Bessel functions of the second kind, is given by eq 39.33
inclusion of the adsorbed phase portion adds important
CAl|rAB
considerations, including altered behavior for mixtures and kapp = ks
pore saturation/blocking. These models provide predictions CAl ,eq (38)
and limits for future modeling, simulation, and experimental
work with graphene membranes. K (r / r )

ks2Di(rAB/rD) K1(rAB / rD )
0 AB D
kapp =
APPENDIX K (r / r )
ks + 2Di(rAB/rD) K1(rAB / rD )
0 AB D (39)
Diusion-Limited Case
In step 2, the molecule must both diuse to the pore and As rD rAB, eq 39 simplies to eq 40.
overcome an activation barrier to position itself above the pore.
The rate constant for this step is aected by both the diusion ks(2 k A(1 + KApA )Di rAB)
properties and the activation energy. It also may exhibit kapp
dierent behavior depending on which process dominates. If ks + (2 k A(1 + KApA )Di rAB) (40)
diusion is the limiting process, then the concentration gradient
This compares to the more commonly used result for diusion-
can be described by the dierential eq 34 as a function of
limited reactions in solutions given by eq 41.
radius, r, in cylindrical coordinates, which includes surface
diusion and adsorption/desorption. The boundary conditions ks(4DrAB)
are given by eqs 35 and 36. Di is the self-diusion coecient, kapp,solution =
ks + (4DrAB) (41)
and rAB is the sum of the radius of the pore and the adsorbed
species.
D d dC
0 = i r Al + (kApA Cl k ACAl)
AUTHOR INFORMATION
Corresponding Author
r dr dr (34) *E-mail: strano@mit.edu.
Notes
CAl(r ) = CAl,eq (35) The authors declare no competing nancial interest.

ksCAl(rAB) = 2rABDi
dCAl
dr rAB (36)
ACKNOWLEDGMENTS
We are grateful for funding from the Institute for Soldier
Nanotechnologies at the Massachusetts Institute of Technology
With the inclusion of adsorption and desorption, the length supported by the U.S. Army Research Oce under contract
scale, rD, for the competition of diusion with ad-/desorption is W911NF-07-D-0004, and for funding from the 2009 U.S.
given by eq 37. Oce of Naval Research Multi University Research Initiative
16677 dx.doi.org/10.1021/la303468r | Langmuir 2012, 28, 1667116678
Langmuir Article

(MURI) on Graphene Advanced Terahertz Engineering (23) Schrier, J. Carbon dioxide separation with a two-dimensional
(GATE) at MIT, Harvard, and Boston University. polymer membrane. ACS Appl. Mater. Interfaces 2012, 4, 37453752.

(24) Qin, X.; Meng, Q.; Feng, Y.; Gao, Y. Graphene with line defect
as a membrane for gas separation: Design via a first-principles
REFERENCES modeling. Surf. Sci. 2013, 607, 153158.
(1) Geim, A. K.; Novoselov, K. S. The rise of graphene. Nat. Mater. (25) Hauser, A. W.; Schwerdtfeger, P. Methane-selective nanoporous
2007, 6, 18391. graphene membranes for gas purification. Phys. Chem. Chem. Phys.
(2) Geim, A. K. Graphene: status and prospects. Science 2009, 324, 2012, 14, 132928.
15304. (26) Koenig, S. P.; Wang, L.; Pellegrino, J.; Bunch, J. S. Selective
(3) Nair, R. R.; Wu, H. A.; Jayaram, P. N.; Grigorieva, I. V.; Geim, A. molecular sieving through porous graphene. Nat. Nanotechnol. 2012, 7,
K. Unimpeded permeation of water through helium-leak-tight 728732.
graphene-based membranes. Science 2012, 335, 4424. (27) OHern, S. C.; Stewart, C. A.; Boutilier, M. S. H.; Idrobo, J.-C.;
(4) Bunch, J. S.; Verbridge, S. S.; Alden, J. S.; van der Zande, A. M.; Bhaviripudi, S.; Das, S. K.; Kong, J.; Laoui, T.; Atieh, M.; Karnik, R.
Parpia, J. M.; Craighead, H. G.; McEuen, P. L. Impermeable atomic Selective molecular transport through intrinsic defects in a single layer
membranes from graphene sheets. Nano Lett. 2008, 8, 245862. of CVD graphene. ACS Nano 2012, DOI: 10.1021/nn303869m.
(5) Garaj, S.; Hubbard, W.; Reina, A.; Kong, J.; Branton, D.; (28) Knudsen, M. The molecular current of gases through openings
Golovchenko, J. A. Graphene as a subnanometre trans-electrode and the effusion. Ann. Phys. (Berlin, Ger.) 1909, 28, 75130.
membrane. Nature 2010, 467, 1903. (29) The values for the parameters for Lennard-Jones potentials were
(6) Merchant, C. A.; Healy, K.; Wanunu, M.; Ray, V.; Peterman, N.; taken from ref 15, and we note that there is some ambiguity as to
Bartel, J.; Fischbein, M. D.; Venta, K.; Luo, Z.; Johnson, A. T. C.; whether the values are atomic or molecular. For the small, diatomic
Drndic, M. DNA translocation through graphene nanopores. Nano species considered in this work, the dierence does not substantially
Lett. 2010, 10, 291521. change the conclusions.
(7) Schneider, G. F.; Kowalczyk, S. W.; Calado, V. E.; Pandraud, G.; (30) Schrier, J. Fluorinated and nanoporous graphene materials as
Zandbergen, H. W.; Vandersypen, L. M. K.; Dekker, C. DNA sorbents for gas separations. ACS Appl. Mater. Interfaces 2011, 3,
translocation through graphene nanopores. Nano Lett. 2010, 10, 44518.
31637. (31) Kolasinski, K. W. Surface Science: Foundations of Catalysis and
(8) Saha, K. K.; Drndic, M.; Nikolic, B. K. DNA base-specific Nanoscience; Wiley: Chichester, England; Hoboken, NJ, 2008.
modulation of microampere transverse edge currents through a (32) Russo, C. J.; Golovchenko, J. A. Atom-by-atom nucleation and
metallic graphene nanoribbon with a nanopore. Nano Lett. 2012, 12, growth of graphene nanopores. Proc. Natl. Acad. Sci. U.S.A. 2012, 109,
505. 59537.
(9) Suk, M. E.; Aluru, N. R. Water transport through ultrathin (33) Freeman, D. L. The influence of diffusion on surface reaction
graphene. J. Phys. Chem. Lett. 2010, 1, 15901594. kinetics. J. Chem. Phys. 1983, 78, 6002.
(10) Sint, K.; Wang, B.; Kral, P. Selective ion passage through
functionalized graphene nanopores. J. Am. Chem. Soc. 2008, 16448
16449.
(11) Cohen-Tanugi, D.; Grossman, J. C. Water desalination across
nanoporous graphene. Nano Lett. 2012, 12, 36028.
(12) Hu, G.; Mao, M.; Ghosal, S. Ion transport through a graphene
nanopore. Nanotechnology 2012, 23, 395501.
(13) Li, Y.; Zhou, Z.; Shen, P.; Chen, Z. Two-dimensional
polyphenylene: experimentally available porous graphene as a
hydrogen purification membrane. Chem. Commun. 2010, 46, 36724.
(14) Schrier, J.; McClain, J. Thermally-driven isotope separation
across nanoporous graphene. Chem. Phys. Lett. 2012, 521, 118124.
(15) Du, H.; Li, J.; Zhang, J.; Su, G.; Li, X.; Zhao, Y. Separation of
hydrogen and nitrogen gases with porous graphene membrane. J. Phys.
Chem. C 2011, 115, 2326123266.
(16) Jiang, D.; Cooper, V. R.; Dai, S. Porous graphene as the ultimate
membrane for gas separation. Nano Lett. 2009, 9, 401924.
(17) Blankenburg, S.; Bieri, M.; Fasel, R.; Mullen, K.; Pignedoli, C.
A.; Passerone, D. Porous graphene as an atmospheric nanofilter. Small
2010, 22662271.
(18) Hauser, A. W.; Schwerdtfeger, P. Nanoporous graphene
membranes for efficient 3 He/4 He separation. J. Phys. Chem. Lett.
2012, 3, 209213.
(19) Shan, M.; Xue, Q.; Jing, N.; Ling, C.; Zhang, T.; Yan, Z.-F.;
Zheng, J. Influence of chemical functionalization on the CO(2)/N(2)
separation performance of porous graphene membranes. Nanoscale
2012, 4, 547782.
(20) Hauser, A. W.; Schrier, J.; Schwerdtfeger, P. Helium tunneling
through nitrogen-functionalized graphene pores: Pressure- and
temperature-driven approaches to isotope separation. J. Phys. Chem.
C 2012, 116, 1081910827.
(21) Schrier, J. Helium separation using porous graphene
membranes. J. Phys. Chem. Lett. 2010, 1, 22842287.
(22) Hankel, M.; Jiao, Y.; Du, A.; Gray, S. K.; Smith, S. C.
Asymmetrically decorated, doped porous graphene as an effective
membrane for hydrogen isotope separation. J. Phys. Chem. C 2012,
116, 66726676.

16678 dx.doi.org/10.1021/la303468r | Langmuir 2012, 28, 1667116678

You might also like