You are on page 1of 15

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/319431904

Adsorption and Phase Behavior of Pure/Mixed Alkanes in Nanoslit Graphite


Pores: An iSAFT Application

Article in Langmuir · September 2017


DOI: 10.1021/acs.langmuir.7b02055

CITATIONS READS

47 1,518

5 authors, including:

Jinlu Liu Le Wang


Rice University Rice University
12 PUBLICATIONS 332 CITATIONS 17 PUBLICATIONS 469 CITATIONS

SEE PROFILE SEE PROFILE

Shun Xi Dilip Asthagiri


ANSYS Oak Ridge National Laboratory
11 PUBLICATIONS 179 CITATIONS 136 PUBLICATIONS 3,984 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Jinlu Liu on 14 January 2019.

The user has requested enhancement of the downloaded file.


Article

Cite This: Langmuir 2017, 33, 11189-11202 pubs.acs.org/Langmuir

Adsorption and Phase Behavior of Pure/Mixed Alkanes in Nanoslit


Graphite Pores: An iSAFT Application
Jinlu Liu, Le Wang, Shun Xi, Dilip Asthagiri, and Walter G. Chapman*
Department of Chemical and Biomolecular Engineering, Rice University, Houston, Texas 77005, United States
*
S Supporting Information

ABSTRACT: The prediction of fluid phase behavior in


nanoscale pores is critical for shale gas/oil development. In
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

this work, we use a molecular density functional theory (DFT)


to study the effect of molecular size and shape on partitioning
to graphite nanopores as a model of shale. Here, interfacial
statistical associating fluid theory (iSAFT) is applied to model
alkane (C1 − C8) adsorption/desorption/phase behavior in
Downloaded via RICE UNIV on October 31, 2018 at 22:35:34 (UTC).

graphite slit pores for both pure fluids and mixtures. The pure
component parameters were fit to the bulk saturated liquid
density and vapor pressure data in selected temperature
ranges. The potential of interaction between the fluid and graphite is modeled with a Steele 10-4-3 potential that is fit to the
potential of mean force from single-molecule simulations. Good agreement is found between theory and molecular simulation for
the density distributions of pure components in slit pores. The critical properties of methane, ethane, and their mixtures as well
as the shift in bubble point and dew point densities were studied, showing good agreement with simulation. The competitive
adsorption of mixtures of normal and branched alkanes in graphite pores was also studied. Heavier components more strongly
adsorb up to the point that the entropic penalty due to confinement reduces adsorption.

■ INTRODUCTION
The exploration and production of shale gas/oil requires a
SBA-16). Both of these experimental studies reveal the impact
of confinement on lowering the bubble point for pores of less
than 5 nm.
reliable prediction of fluid-phase behavior under reservoir
In addition to experimental work, molecular simulations have
conditions. The pores in shale are on the nanometer scale,
been used extensively to understand the effect of confinement
unlike that in conventional sandstones where the pore sizes are
on fluid properties. Severson and Snurr6 studied the single-
in micrometers. The nanoconfinement of hydrocarbons can
component adsorption of alkanes in carbon slit pores. Besides
affect the phase behavior and partitioning of components
the adsorption isotherms, the orientation profiles were also
between the matrix and production fractures as a result of the
reported, showing that short alkanes lay parallel to the walls
strong interaction between the fluid molecules and pore while longer alkanes have less distinct layering. To study how
surface. Successful reservoir simulation can be achieved only confinement affects criticality, Akkutlu and Didar7 used the
with an accurate model of hydrocarbon phase behavior, and a NVT-Gibbs Monte Carlo method and Pitakbunkate et al.8 used
number of studies have been performed to better understand the grand canonical Monte Carlo method to obtain the critical
fluids in confined geometry. The physics of nanoconfinement parameters of methane and ethane in graphite slit pores. Their
also affects other applications involving porous material, such as results, though different from each other, showed a large
gas storage using metal organic frameworks,1 catalysis in deviation in critical properties from the bulk. The suppressed
zeolites,2 and gas mixture separation.3 critical point was shown to cause a large compositional change
One of the primary applications in the shale gas/oil industry in gas production. In a series of molecular dynamics studies by
is to determine the density and phase behavior of the in-place Hu et al.,9−11 an activated kerogen model, namely, graphene
hydrocarbon mixture as a function of temperature and pressure fragments with oxidized functional groups, was proposed in
(ρ(T, P)). Standard adsorption and desorption experiments are comparison with the simple graphite model. They have studied
often used to characterize porous materials through capillary the behavior of water and hydrocarbons in model shale
condensation and hysteresis in nanoscale pores. In shale oil nanopores. The activated kerogen model shows mixed wetting
production, the bubble point is one of the key parameters that
ensures a continuous flow of liquid production. Luo et al.4 used
Special Issue: Tribute to Keith Gubbins, Pioneer in the Theory of
differential scanning calorimetry to measure the bubble-point
Liquids
temperature of octane and decane in controlled-pore glasses
(CPGs) with pore sizes of 4.3 and 38.1 nm. Cho et al.5 Received: June 15, 2017
measured the bubble-point pressures of octane−methane and Revised: August 29, 2017
decane−methane mixtures in mesoporous silica (SBA-15 and Published: September 1, 2017

© 2017 American Chemical Society 11189 DOI: 10.1021/acs.langmuir.7b02055


Langmuir 2017, 33, 11189−11202
Langmuir Article

properties, and water tends to be trapped inside the pores. DFT, to name a few, include the interfacial tension of
They have also shown that a rough surface model adsorbs polymers,32 lipid assembly and confined block copolymers,33
alkanes more evenly while a smooth surface adsorbs alkanes in tethered polymers within or without solvent,34 surfactant in
layers. A more rigorous approach to modeling kerogen oil−water interfaces,35 the formation of micelles in oil/water/
molecules, which is based on an analysis of chemical structure, surfactant systems,36 tethered polymer-modified pores,37,38 and
elemental composition, and physical properties, was taken by polymer adsorption.39 For applications in shale gas/oil systems,
Ungerer et al.,12 and the adsorption isotherms of methane and other DFT approaches have also been proposed on the basis of
ethane on different-sized kerogen pores were generated.13 The some common bulk equations of state. For example, the
adsorption isotherms were modeled by an extended Langmuir adsorption and desorption behavior of hydrocarbon mixtures
model and ideal adsorbed solution theory, which can capture was studied by Li et al.40 using a Peng−Robinson equation of
the adsorption behavior at low pressure but only works well for state (EoS) free energy functional.
large pores at high pressure. Molecular simulation by atomistic A SAFT-41 or PC-SAFT-based42 free energy functional,
models also helps to improve our understanding of kerogen however, is more promising because molecular shape is
morphology and surface properties. Recent work by Vasileiadis included in the free energy functional. To enable iSAFT to
et al.14 has demonstrated the characterization of pore size and match the dispersion contribution of the bulk PC-SAFT EoS,
pore surface elemental analysis using Ungerer’s model. several modifications of the dispersion functional have been
Computer simulation is a powerful tool for providing proposed. Shen et al.43,44 used a WDA in the dispersion term to
detailed and accurate descriptions of real kerogen systems. model methane and carbon dioxide adsorption on porous
The high computing times of molecular simulation are carbon. Klink and Gross45 included the hard chain pair
rewarded by the accurate prediction of system behavior. correlation function in the dispersion functional and incorpo-
However, for some purposes such as understanding the phase rated a local correction term to absorb the difference from the
behavior of a multicomponent system, density functional theory PC-SAFT EoS chemical potential. Xu et al.46 decomposed the
is computationally more efficient. In particular, if one of the dispersion term to a local and nonlocal contribution where the
components in the mixture is of extremely low concentration, it surface-averaged density was used in a weighted-density-
becomes very challenging to obtain good statistics in molecular approximated local term and a mean field manner was used
simulation, resulting in increased computational cost. On the to account for the density variation. These developments can
other hand, a theoretical method such as density functional lead to an improved prediction of the bulk pressure, density,
theory can be validated versus molecular simulation and and chemical potential, which makes it easier to compare with
efficiently extend the results to multicomponent mixtures and experiment and simulation. The accuracy of these dispersion
trace components. terms that reduce iSAFT to PC-SAFT EoS in the bulk is not
The molecular density functional theory applied here fully proven for confined fluids. A WDA alone does not account
combines several approaches to describe short-range repulsion, for the long-range attraction sufficiently, and “local plus
long-range attraction, and molecular shape. Alkane molecules correction” approaches47 have been applied only to fluid−
are modeled as chains of spherical segments. The basic building fluid interfacial systems. On the other hand, an addition of hard
block of the theory is an inhomogeneous hard sphere fluid. The sphere repulsion and mean field attractive interactions is still
fundamental measure theory15 proposed by Rosenfeld and a the “simplest and very successful approach”48 to modeling
few variants thereafter realized a quantitative description of dispersion interactions.
inhomogeneous hard sphere mixtures. The bonding of spheres In our current work, we use an MFA in describing the
to form polyatomic molecules is accomplished through dispersion contribution in iSAFT. Model parameters are fit to
extensions of Wertheim’s perturbation theory16−19 for associat- pure component vapor pressure and saturated liquid densities.
ing and polymerizing spheres.20−22 Long-range attractions are Additionally, we have systematically evaluated the average
described using a mean field approximation (MFA) described interaction between an alkane molecule and a graphite slab by
later. The mean field approximation (MFA), weighted density the potential of mean force, which quantifies the interaction
approximation (WDA), and Wertheim perturbation force of a molecule with a surface. The predictions of
theory16−19 allow for modeling the molecular size and shape. adsorption/desorption, thermodynamic equilibrium, and crit-
Spherical molecules such as Lennard-Jones spheres and ical properties are performed for a series of alkanes depicting
associating spheres in confined geometry have been studied the hydrocarbons in a shale reservoir. The calculations for pure
with DFT by numerous groups such as Peterson et and binary systems are verified by molecular simulation data,
al.,23Sokolowski and Fischer,24 Segura et al.,21 Yu and Wu,25 and further studies on multicomponent systems suggest a
Tripathi and Chapman,26 and Haghmoradi et al.27 The competition between enthalpic and entropic effects in
adsorption isotherm kernel generated from DFT has been determining pore selectivity.


used to characterize nanoporous carbon and silica28−30 and has
become one standard analysis tool in commercial adsorption
instruments. Although small molecules (e.g., nitrogen, argon,
METHOD
and krypton) that are nearly spherical are mostly used in Model Description. The modeled slit pore system can be
nanoporous material characterization, polyatomic molecules in represented by the schematic shown in Figure 1. The two
confined media can be more fascinating in that practical infinitely large slabs of graphitic wall confine the hydrocarbons
systems often involve mixtures such as in polymer synthesis, inside. The graphitic wall is represented by layers of carbon
lubrication, and hydrocarbon storage in shale. The develop- atoms, and here for simplicity, we assume that the wall is flat
ment of inhomogeneous statistical associating fluid theory and smooth. The fluid inside the pore is connected and in
(iSAFT),22 modified iSAFT,31 and polymer-DFT25 has largely equilibrium with a bulk fluid reservoir, which means an
enhanced our ability to model inhomogeneous systems of equilibrium of temperature and chemical potential in each
complex fluids. The implementations of iSAFT or polymer- species.
11190 DOI: 10.1021/acs.langmuir.7b02055
Langmuir 2017, 33, 11189−11202
Langmuir Article
n mi
Aid [ρ(r)] = kT ∑ ∫ dr ∑ ρα(i)(r)[ln(ρα(i)(r)Λi3) − 1]
i=1 α=1
(4)
where k is the Boltzmann constant, mi is the segment number
in species i, Λi is the thermal de Broglie wavelength, and ρ(i)
α (r)
denotes the density of segment α in the ith molecule. A
correction of the atomic ideal gas free energy is included in the
chain term. For a homonuclear chain, all of the segments have
the same density, i.e., ρ(i)
1 (r) = ρ2 (r) = ... = ρmi (r) = ρi(r).
(i) (i)

Because Λi is temperature-dependent and the temperature is a


constant here, the ln(Λ3i ) term will be canceled by the
corresponding term in the bulk chemical potential in eq 1.
Hard Sphere Contribution. The hard sphere part accounts
for the excluded volume effect for an inhomogeneous fluid. We
use the original fundamental measure theory15 in which the
Figure 1. Schematic representation of the modeled graphite slit pore hard sphere free energy functional is
system (H, pore width; σs, carbon atom diameter; Δ, separation
between two graphite layers). ΔAhs[ρ(r)] = kT ∫ Φex,hs[nj(r)] dr (5)
and Φ [nj(r)] is the excess Helmholtz free energy density. It
ex,hs

is a functional of six fundamental measure densities nj(r) of the


For a fluid with fixed chemical potential (μ), temperature inhomogeneous fluid,
(T), and volume (V), the grand potential or grand free energy
nn
(Ω) is a minimum in the equilibrium state, with Ω determined Φex,hs[nj(r)] = −n0 ln(1 − n3) + 1 2
by the chemical potential of each species (μi), the intrinsic 1 − n3
Helmholtz free energy of the fluid (A[ρ(r)]), and the external n23 n v1·n v2 n (n · n )
potential from the graphite wall (Vext(r)), + 2
− − 2 v2 v2 2
24π (1 − n3) 1 − n3 8π (1 − n3) (6)
n
Ω[ρ(r)] = A[ρ(r)] − ∑ ∫ ρi (r)(μi − V iext(r)) dr where the nj(r) values are calculated by fundamental weights,
i=1 (1) n

where n is the total number of species, Vext


nj(r) = ∑ ∫ ρi (r)ωi(j)(r − r′) dr′; j = 0, 1, 2, 3, v1, v 2
i (r) is the external i=1
potential acting on each species, and ρi(r) is the molecular (7)
density of the ith species. With Ω being a free energy
functional, the equilibrium density distribution is obtained by At each position, the summation runs over all of the species
minimizing the grand potential, in the mixture. The weight functions are given by
r
ωi(2) = δ(R i − r ) ωi(3) = Θ(R i − r ) ωi(v2) = δ(R i − r )
δ Ω[ρ(r)] r
= 0 for i = 1, 2, ... n (8)
δρi (r)
equilibrium (2)
and
The average density is then determined from the equilibrium
ωi(2) ωi(2) (v1) ωi(v2)
density profile of each component. ωi(0) = ωi(1) = ωi =
Free-Energy Functional. In density functional theory, the 4πR i2 4πR i 4πR i (9)
main challenge is to approximate the Helmholtz free energy
where δ(r) and Θ(r) are the Dirac function and Heaviside
functional, and this functional can be formed through a
function, respectively, and Ri = di/2 is the radius of the segment
perturbation approach. In the iSAFT method proposed by
in component i. The temperature dependency of the hard
Tripathi and Chapman,22 molecules are constructed by
sphere diameter is determined by an empirical correlation49 of
bonding the spherical beads or segments to form chains that
the Barker−Henderson hard sphere diameter, which is
resemble the structure of molecules. In this work, our interest is
only in alkanes, so the molecules are represented by chains of ⎡ ⎛ 1/2 ⎤−1/6
T * + 1.1287T *2 − 0.05536T *4 ⎞ ⎥
homonuclear spheres. For a mixture of n molecular species, we d = 21/6⎢1 + ⎜1 + ⎟ σ
⎢⎣ ⎝ 0.0007278 ⎠ ⎥⎦
have n types of segments with mi segments in each chain (10)
molecule. The Helmholtz energy is a summation of different where T* is the dimensionless temperature reduced by fluid−
contributions, namely, an ideal gas (Aid), hard sphere (Ahs), fluid interaction energy ϵαα(T* = kbT/ϵαα) and σ is the segment
chain formation (Ach), and dispersion (Adisp) contributions, diameter. For the numerical implementation in a one-
dimensional system, we refer readers to the work by Gross,50
A[ρ(r)] = Aid [ρ(r)] + ΔAhs[ρ(r)] + ΔAch [ρ(r)] where the density symmetry in x and y coordinates is
+ ΔAdisp[ρ(r)] (3) considered.
Chain Contribution. The free energy due to chain formation
Ideal Gas Contribution. The exact ideal gas free energy for has been developed by Tripathi and Chapman,22 where the
an atomic mixture of n species is chain free energy functional is formed by the complete
11191 DOI: 10.1021/acs.langmuir.7b02055
Langmuir 2017, 33, 11189−11202
Langmuir Article

association of spheres with directional association sites. The ⎧0 0 < r < σ αβ



free energy functional for mixtures of associating spheres was ⎪− ε − u LJ
σ αβ < r < rmin
developed by Chapman20 and Segura et al.21 based on ⎪ αβ cut
att
uαβ =⎨
Wertheim’s thermodynmaic perturbation theory (TPT1).16−19 ⎪ ⎡⎛ αβ ⎞12 ⎛ αβ ⎞6⎤
σ σ
A schematic representation of chain formation is shown in ⎪ 4εαβ ⎢⎜ ⎟ −⎜ ⎟ ⎥ − ucut
LJ
rmin < r < rcut
⎪ ⎢⎝ r ⎠ ⎝ r ⎠ ⎥⎦
⎩ ⎣ (15)
Figure 2, where spheres with directional association sites are
in which uLJ
cut is the Lennard-Jones pair potential evaluated at
the cutoff distance, and here we use rcut = 4σαβ. For mixtures,
the Lorentz−Berthelot mixing rule σαβ = (σα + σβ)/2 and
εαβ = εααεββ is applied.
Grand Potential. The formalism to calculate the equilibrium
density profile can be found directly from the original paper by
Tripathi and Chapman.22 To calculate the grand potential, we
use the form in modified iSAFT developed by Jain et al.31 that
is more convenient and does not assume homonuclear
molecules. Although modified iSAFT is more rigorous and
predicts the exact ideal chain distribution, it requires an integer
number of segments whereas original iSAFT (like the bulk
SAFT EoS) allows noninteger values. In our tests of both
methods using integer numbers of segments (m = 1, 2, 3...), we
find that there is hardly any difference between the converged
density profiles for a wide range of temperatures and pressures,
Figure 2. Schematic of the ith molecule in the mixture formed from mi which ensures the reliability to transfer a grand potential
associating spheres. functional from modified iSAFT to original iSAFT.
n
β Ω[ρ(r)] = ∑ ∫ [ρi (r)(miDi(r) − 1)] dr + β ΔAhs
i=1
forced to bond in a specific order to form chainlike molecules. disp
The resulting chain free energy functional for homonuclear + β ΔA (16)
chains is n αα ′
δ ln ycontact [ρi̅ (r1)] δ ΔAhs δ ΔAdisp
ch n ⎡ 1 δ(|r1 − r2| − σ αα ′
)
Di(r) = ∑ (1 − mi) ∫ ρi (r1) dr1 − −
ΔA δρi (r) δρi (r) δρi (r)
= ∑ 2(mi − 1)∫ dr1ρα(i)(r1)⎢− ln ∫ dr2 y αα ′ i=1
kT i=1
⎢⎣ 2 4π(σ αα ′)2 (17)
1⎤ Note that the terms that involve mi are just simplified from eq
(r1, r2)ρα(i)(r2) + ⎥
2 ⎥⎦ (11) 32 in the modified iSAFT grand potential31 with identical
segment densities ρ(i)
α (r) in each chain.
δ(| r1 − r2 | −σ αα ′) External Potential. As we see from eq 1, the effect of the
where enforces tangential bonding. The cavity
4π (σ αα ′)2 fluid−solid interaction is modeled only by the external
correlation function yαα′(r1, r2) is needed only at a contact potential. For a graphitic wall, this external field is modeled
distance of σαα′, and we evaluate it by the bulk reference fluid by a Steele 10-4-3 potential,51 and for slit pores, the potential
radial distribution function at some averaged density ρα̅ (i) (r1), from both walls can be added to give Vext j (z) = Vj
10‑4‑3
(z) +
i.e., Vj (H − z),
10‑4‑3

⎡ ⎛ σ ⎞10 ⎛ σ ⎞4 σsj4 ⎤
αα ′
ycontact (ρα(i)(r1), σ αα ′) = ycontact
HS
(ρα̅ (i) (r1), σ αα ′) (12) V 10 −4−3 2 sj
(z) = 2πρs Δs σsj2εsj⎢ ⎜ ⎟ − ⎜ ⎟ −
sj

⎢⎣ 5 ⎝ z ⎠ ⎝z⎠
j
3Δ(z + 0.61Δ)3 ⎥⎦
Here we use the simplest weight within a sphere, i.e., (18)
ρα̅ (i) (r1) =
3
∫|r −r |<σ ρα(i)(r2) dr2 where ρs = 0.114 Å−3 is the carbon atom density, Δs = 3.35 Å is
4π (σα)3 1 2 α (13) the separation distance between two graphite layers, σsj and εsj
are the segment diameter and interaction energy that are
An alternative free energy functional for heteronuclear chains is estimated from the Lorentz−Berthelot mixing rules, σsj = (σs +
given in the work of Jain et al.31 σj)/2 and εsj = εsεj , where εs and εj are the interaction
Dispersion Contribution. For dispersion attractions, we use
the mean field approximation that assumes the pair correlation energies for solid−solid and fluid−fluid interactions and σs and
function between segments equals 1 beyond the hard sphere σj are the carbon atom diameter and segment diameter of the
contact. The dispersion contribution to the free energy component, respectively. The commonly used parameters for
functional is written as graphite are σs = 3.345 Å and εs/k = 28 K.52
n n
However, this external potential was derived for a spherical
ΔAdisp 1 molecule interacting with a surface assuming that the molecule
= ∑∑∫ dr1 dr2u att
αβ (|r2 − r1|) · mi ρi (r1) · has only one interaction site. For a chain molecule with
kT 2 i=1 j=1 |r2 − r1|> σαβ
multiple segments, the interacting force at each position is also
mjρj (r2) determined by the configurational change within the molecule.
(14)
Previous studies using an external potential in density
and the pair potential is given by a WCA-type perturbation, functional theory usually treat εsj as an adjustable parameter
11192 DOI: 10.1021/acs.langmuir.7b02055
Langmuir 2017, 33, 11189−11202
Langmuir Article

Table 1. Parameters for DFT Modeling of the Studied Components


AAD%
species m σ(Å) εj/k(K) vapor pressure liquid density temperature range (K) fluid−solid interaction εsj/k (K)
methane 1.0000 3.6754 202.8 2.2 3.1 100−180 63.9
ethane 1.5969 3.4557 243.0 5.6 2.2 200−300 65.7
n-propane 2.0049 3.4915 260.0 5.7 1.7 250−350 66.6
n-butane 2.3650 3.5923 275.5 5.3 2.1 250−350 67.5
n-pentane 2.6636 3.6808 290.0 6.0 2.1 250−350 69.1
iso-pentane 2.5321 3.7448 289.9 5.6 2.1 250−350 68.9
neo-pentane 2.3459 3.8459 281.7 5.7 2.1 260−360 64.1
n-hexane 3.0334 3.7295 296.1 7.9 2.3 245−365 70.7
n-heptane 3.4502 3.7357 298.5 7.4 2.1 250−350 70.4
n-octane 3.9598 3.7034 296.2 8.1 2.1 250−350 69.7

to fit the experimental data. For example, in the work by Shen The averaged density inside the slit pore is defined as
et al.,43 the interaction energy between solid and fluid was H − σs /2
adjusted along with pore size to achieve the best fit of the 1
experimental isotherm on porous carbon. However, such an
ρiavg =
H − σs
∫σ /2
s
ρi (z) dz
(21)
approach can mask the deficiency of the free energy functional
and confound the interpretation of other contributions to the which is used to calculate the adsorption isotherms.
functional. Numerical Procedures. An advantage of the original
Here we determine the interacting force between a single iSAFT is the simplicity in the chain free energy that requires no
molecule and a surface by molecular simulation and apply the recursive integrals. This method shows stable and satisfactory
simulated potential of mean force to the DFT model. The convergence performance with a simple Picard iteration starting
adaptive biasing force method53 is selected in which a force from a uniform density profile in the pore. The discretization
profile is applied to the center of mass of the molecule, and resolution we use is dz = 0.01σ, and a symmetric density profile
eventually the molecule is allowed to move freely in the is assumed to impose a reflective boundary condition. In
specified domain as if there is no attractive surface. In this way, calculating the adsorption or desorption isotherms, we take the
the force is averaged for all possible configurations of the converged density profile from the last pressure (Pi) as an initial
molecule, which has a different magnitude from a simple guess for calculation at the next pressure (Pi+1).The initial guess
summation of the interacting forces on each site with only one plays an important role because the metastable state in
configuration. Detailed information on the potential of mean hysteresis can be established only with an appropriate initial
guess density profile.


force simulation can be found in Supporting Information A.
Equilibrium Density Profile. Solving eq 2 results in the
equilibrium density profiles for n species in the system. On the RESULTS AND DISCUSSION
basis of the functional contributions described above, the Parameters for n-Alkanes and External Potentials.
density profiles are obtained from Within the framework of iSAFT, the alkane molecules is
described using three parameters, which are the chain length or
⎡μ ext {0,1,2,3, v1, v 2} ⎤
⎢ i − V i (r) − dr ∂Φex,hs ⎥ number of segments (m), the segment size (σ), and the
⎢ mi mi
∫ 1 ∑ ωi(j)(|r − r1|)
∂nj(r1) ⎥
⎢ j

⎢ n ⎥
⎢ −∑ ∫ att
dr1uij (|r1 − r|)mjρj (r1) ⎥
⎢ j=1
|r − r1|> σij ⎥
ρi (r) = exp ⎢ ⎥
⎢ m⎡ δ ln λ (r) ⎤⎥
⎢ + i ⎢ln(y cont (r)λii(r)) − 1 + dr1ρ (r1) ∫ ii ⎥⎥
⎢ 2 ⎢⎣ ii i
δρi (r) ⎥⎦ ⎥
⎢ ⎥
⎢ mi − 1 δ ln yiicont (r1) ⎥
⎢ + ∫dr1ρi (r) ⎥
⎢⎣ mi δρi (r) ⎥⎦
(19)
i = 1, 2, ...n

where
δ(|r1 − r| − σi)
λii(r) = ∫ dr1 4π (σi)2
ρi (r1)
(20)

It should be noted that eq 19 has been simplified from Tripathi


and Chapman.22 For the ith chain, even though the two end
segments would have different chain contribution from the Figure 3. Methane reduced density profiles from theory (solid lines)
middle segment, we take the average over all of the segments and Monte Carlo simulations (circles), at bulk pressures of (a) 5.0 bar,
and assume that all of the segment density distributions are (b) 7.2 bar, and (c) 10.2 bar in a slit pore at 155 K. σ is the segment
identical. diameter.

11193 DOI: 10.1021/acs.langmuir.7b02055


Langmuir 2017, 33, 11189−11202
Langmuir Article

100
Nref qical − qiref
AAD % = ∑
Nref i=1 qiref

from saturated liquid densities and vapor pressures is


minimized to find the three model parameters. As mentioned
in Dominik et al.,32 the chain length and segment diameter
were not significantly changed in comparison to the parameters
regressed by Gross and Sadowski,42 so we first fix the chain
length and segment diameter parameters by keeping them the
same as the PC-SAFT parameters and adjust only the attraction
energy. We then adjust the chain length and segment diameter
to obtain the smallest error to the reference data. The fitted
parameters and fitting errors are presented in Table 1. As we
see, the errors are fairly large in comparison to the PC-SAFT
EoS,42 especially for the vapor pressures. This is expected
because the dispersion term has only one parameter while the
PC-SAFT EoS dispersion term is correlated to simulation
Figure 4. Heptane reduced segment density profile from theory (solid results of linear alkanes. Overall we consider this mean field
line) and molecular dynamics simulation (circles) in a slit pore at a approximated equation of state accurate enough for the current
bulk pressure of 1 atm and a temperature of 298 K.
study.
For fluid−solid interactions, we find that the potential of
dispersion energy of each segment (εk). As was described mean force (PMF) predicted by molecular dynamics (MD)
above, the dispersion contribution is solely accounted by the simulation nearly conforms to the Steele 10-4-3 potential, but
mean field approximation, which means that the radial with a different well depth. Considering that the Steele
distribution function is set equal to 1 outside of the hard potential has an analytical formula, we use this equation after
sphere contact. It is therefore expected that the dispersion calibrating the interaction energy parameters according to the
energy between molecules will not be the same as parameters potential well values from simulation. The solid−fluid
from the bulk SAFT. To retain both efficiency in inhomoge- interaction energies are also listed in Table 1.
neous calculation and accuracy in bulk property prediction, we Pure Component Adsorption, Desorption, and Phase
refitted parameters for alkanes based on their saturated liquid Equilibrium. Before predicting the phase behavior of different
densities and vapor pressures. Unlike in modified iSAFT31 and components in nanopores, we first verify our model in
polymer DFT,25 the formalism of the chain contribution predicting density profiles by comparing with grand canonical
according to Tripathi and Chapman22 allows us to have Monte Carlo (GCMC) and MD simulations. As previous work
noninteger numbers for the chain length, which has the presented, this model shows good performance in model
advantage of more accurately modeling the phase behavior of a molecules like hard chains.22 In this work, we mainly focus on
full-range of alkanes. The reference data is from the online the modeling of real molecules, in which case the dispersion
library of National Institute of Standards and Technology contribution plays an important role. In particular, the mean
(NIST).54 The absolute average deviation field approximation may bring up some inaccuracy, so it is

Figure 5. (a) Adsorption and desorption isotherms of ethane in a 3 nm pore at 230 K. (b) Corresponding grand potential change with the chemical
potential for adsorption and desorption isotherms, with the intersection being the equilibrium transition, which is located at the circles in plot a.

11194 DOI: 10.1021/acs.langmuir.7b02055


Langmuir 2017, 33, 11189−11202
Langmuir Article

Figure 6. (a) Density pressure (ρ−P) diagram. (b) Temperature−density (T−ρ) diagram of methane confined in a 3 nm slit pore in comparison to
the bulk phase. Circles show the equilibrium densities at each temperature, which read 105, 110, 115,120, 125, 130, 135, 140, 145, 150, 155, 160,
165, 170, and 177 K, respectively.

Figure 7. (a) Density pressure (ρ−P) diagram. (b) Temperature−density (T−ρ) diagram of ethane confined in a 3 nm slit pore in comparison to
the bulk phase. Circles show the equilibrium densities at each temperature, which read 205, 225, 230, 235, 240, 245, 250, 255, 265, and 275 K,
respectively.

necessary to compare the density profile to molecular respect to the diameter of the molecule in each case. The
simulation, i.e., grand canonical Monte Carlo (GCMC) success is mainly attributed to fundamental measure theory
simulation or molecular dynamics (MD) simulation. We have because the structure is primarily determined by repulsion. The
performed GCMC simulation for vapor methane adsorption theory and GCMC simulation both show growing layers and
and MD simulation with NAMD55 for liquid heptane then capillary condensation of vapor methane as the pressure
adsorption. The parameters in the simulation are from a increases, which is due to the strong attraction from the
united-atom model TraPPE force field,56 which is widely used graphite surface. However, it is much more challenging to
for alkanes. In MD simulation, the system setup follows the obtain excellent agreement between simulation and theory for a
work by Hu et al.,11 where the simulation box includes both the chain molecule. As we see the density distribution of heptane in
confined alkane between two blocks of graphite and the bulk a slit pore of H = 3 nm at 298 K and 1 atm in Figure 4,
region. although simulation and theory show similar magnitudes of
We start with methane, a spherical molecule, in which there peaks and valleys, the wavelength of oscillation is much
is no contribution from the chain free energy. Figure 3 different. The chain modeled in DFT is much too flexible
compares the density profiles of adsorbed methane with those relative to that modeled by atomistic simulations, thus the
from GCMC simulation for a slit pore of H = 5 nm at 155 K distance between two peaks is a little larger than the segment
and different bulk pressures. Although different molecular diameter, whereas in density functional theory the distance can
models are used in the two methods, the agreement between be even smaller than one segment diameter. Given the fact that
them is satisfactory. We have normalized the position with the density profile cannot be fully captured by theory as a result
11195 DOI: 10.1021/acs.langmuir.7b02055
Langmuir 2017, 33, 11189−11202
Langmuir Article

Figure 8. Critical temperature and pressures at different pore sizes for methane (a, b) and ethane (c, d) from iSAFT prediction and a Monte Carlo
simulation of the literature.8,7 The bulk critical points predicted by this mean-field-approximated EoS are 202 K, 54.8 bar for methane and 325 K,
58.3 bar for ethane, whereas the experimental values are 190.6 K, 46.1 bar for methane and 305.3 K, 49 bar for ethane.

of different assumptions, we compare the average density inside two local minima of the grand potential. To determine the
the pore, which is our major interest of study. The DFT- stable state at the corresponding pressure, we need to examine
predicted excess adsorption differs from the simulation result the grand potential change along each isotherm and the
by only 1.8%, which shows that theory still predicts the excess thermodynamic transition that is located at the intersection of
adsorption in a convincing manner. the two grand potential isotherms. In Figure 5, the adsorption/
Because methane and ethane are the major components in desorption isotherms of ethane in a 3 nm pore at 230 K are
shale gas, we study the phase behavior of pure methane and shown and the thermodynamic transition point is located in
ethane first. Consider the adsorption and desorption shown in Figure 5b.
Figure 5a for ethane in a 3 nm slit pore. When we have a The same routine was followed at other temperatures for
gaseous phase outside the pore, increasing bulk pressure would methane and ethane in a 3 nm slit pore, and the vapor−liquid
lead to an increasing number of layers adsorbed in the pore, so equilibria are displayed in Figures 6 and 7. The bulk phase
part of the pore is filled with adsorbed fluid that has a liquidlike density−pressure diagrams are also calculated with this MFA
density and in the middle the density is still vaporlike, which is type EoS, which are inserted into the confined phase diagrams.
called free gas. As the pressure increases, a first-order vapor− In the confined system, the critical point is reached when
liquid transition or capillary condensation can be seen when the adsorption−desorption hysteresis disappears and the disconti-
density everywhere is liquidlike. This phase transition happens nuity in density diminishes. For methane, we find the critical
below the bulk saturation pressure and depends strongly on the temperature and pressure in the 3 nm slit pore to be around
pore size. However, a decrease in pressure from the saturated 177 K and 12.48 bar, respectively. Compared to the bulk critical
adsorption produces a phase transition at a lower pressure. This point based on MFA type EoS of 202 K and 54.8 bar,
hysteresis is a prominent feature of confined fluids. In this way, confinement dramatically shifts the critical point. Similarly for
two different density profiles can be obtained corresponding to ethane, the critical point is also shifted by more than 20 K in
11196 DOI: 10.1021/acs.langmuir.7b02055
Langmuir 2017, 33, 11189−11202
Langmuir Article

Figure 9. (a) Adsorption/desorption isotherms of a methane/ethane mixture in different bulk compositions at T = 250 K with a slit pore size of H =
5 nm. Isotherms from left to right correspond to bulk ethane fractions of 0.9, 0.75, 0.65, 0.5, 0.4, 0.35, 0.3, 0.27, 0.25, 0.23, and 0.22. The critical
point is predicted at a bulk ethane concentration of ∼0.22 and a bulk pressure of ∼50 bar. (b) Equilibrium pressure vs ethane fraction diagram in
comparison with GCMC simulation and the bulk phase.

temperature and more than 30 bar in pressure. Overall the bar at a bulk ethane fraction of 0.22 as adsorption/desorption
phase boundaries are narrower in the pore than in the bulk. hysteresis disappears when the bulk ethane concentration
The high density of the dew curve is due to the strong decreases from 0.23 to 0.22. The equilibrium transition was
attraction from the wall that causes the first layer of fluid near determined from the grand potential of adsorption/desorption
the wall to have a liquidlike density. This liquidlike density isotherms. The equilibrium dew point and bubble-point curves
region occupies a fairly large fraction of the pore space, so the of ethane are also shown in the same figure in comparison with
average density in the pore is no longer comparable to a bulk the bulk ethane phase boundaries. Because the exact bulk
vapor phase. As the pressure increases, the fraction of the liquid concentration corresponding to each isotherm was not reported
part becomes larger, which makes the average density in the in Pitakbunkate et al.,8 we compare only the predicted bubble/
pore deviate more and more from vaporlike density. dew curves of ethane with their simulated ones. We see that our
Critical properties can influence the gas compressibility, calculated equilibrium phase boundary is in very good
viscosity, and formation volume factor predictions in reservoir agreement with the GCMC adsorption isotherm. We present
simulators and thus impact the production scheme.7 In Figure the bulk phase ethane density to better illustrate the shift of
8, we show the iSAFT-predicted critical properties in phase boundaries and suppression of critical point in the pore.
comparison with literature data for different pore sizes. The Because of fluctuations in the simulations, it is difficult to trace
predictions are qualitatively consistent with Monte Carlo the full adsorption curve. Thus, we suspect that the adsorption
simulations. For methane, the theory predicts critical temper- phase boundary from simulation is close to the equilibrium
atures closer to the NVT-Gibbs MC, and the predicted critical state. Finally, we compute the composition of the methane/
pressure lies between the two simulation studies. It is known ethane mixture at each equilibrium point and obtain the
that a bulk equation of state overpredicts the critical point, and pressure−fraction diagram in the pore. Again we compare this
the difference between theory and simulation critical points result with that from adsorption simulation and that of the bulk
appears to be due, in part, to differences in bulk critical points. phase. From Figure 9b, we see that the theory predicts a higher
The theory predicts the trend in the critical point with respect critical pressure and lower ethane concentrations. This is
to pore size consistent with simulated data. The data set in consistent with the performance of the iSAFT bulk EoS as we
Figure 8 is listed in Table 1 and Table 2 in Supporting see that it overpredicts the critical pressure and underestimates
Information B. the ethane concentration when comparing to bulk PC-SAFT
Binary Systems. The prediction of hydrocarbon mixture EoS predictions.
partitioning between the kerogen matrix and fractures is As we can see from the last example, ethane is always
essential to production management in shale reservoirs. preferred over methane in the 5 nm graphite pore because of
Pitakbunkate et al.8 performed a GCMC study of the pore- the larger affinity of ethane for the graphite surface. However,
filling process of a methane/ethane mixture at 250 K in a 5 nm as the chain becomes longer, the entropy loss of entering the
graphite pore, and the phase boundaries and critical point were confined space will compete with the enthalpy benefit so that
determined from simulation data. Here we calculate the same the longer chain partitions to the bulk. Here we study the
properties with iSAFT and compare the results with their adsorption of a Cn/Cn+1 binary mixture and characterize the
simulation data. The adsorption and desorption isotherms with pore preference of the longer alkane Cn+1 by the ratio of the
different bulk ethane concentrations are shown in Figure 9a. As pore fraction and bulk fraction. In Figure 10a, gas mixtures of
the bulk ethane concentration decreases, the phase-transition different compositions at 298 K, 1 atm show different selectivity
pressure increases, and the hysteresis disappears as the pore behavior in a 3 nm pore. For the C4/C5 mixture, the maximum
fluid reaches a critical point. The calculated critical point is 50 C5 fraction is 0.5, which is below the mixture saturation
11197 DOI: 10.1021/acs.langmuir.7b02055
Langmuir 2017, 33, 11189−11202
Langmuir Article

Figure 10. iSAFT prediction of (a) a binary gas mixture adsorption in different compositions, (b) C1/C2 mixture adsorption in different pore sizes,
(c) binary liquid mixture adsorption in different compositions, and (d) C7/C8 mixture adsorption in different pore sizes. The temperature and
pressure are fixed at 298 K and 1 atm.

mixture lies between C2/C3 and C3/C4. This implies that the
entropy loss of pentane plays a more important role than in the
C3/C4 mixture.The selectivity has a completely different trend
for liquid alkane mixtures under ambient conditions, as shown
in Figure 10c with C7/C8 < C6/C7 < C5/C6. For the C5/C6
mixture, C6 still partitions to the pore more than C5 does.
However, for the C6/C7 mixture, the pore fluid has almost the
same composition as that in the bulk. What is surprisingly
predicted is that for the C7/C8 mixture the preference to
longer alkane no longer holds as the fraction of C8 in the pore
is smaller than that in the bulk. The results demonstrate the
competition between the enthalpy benefit from surface
attraction and entropy loss due to geometric confinement.
We are not aware of experimental studies that would provide
insight into long alkane adsorption in nanopores. For longer
alkanes, the advantage in surface affinity becomes less dominant
Figure 11. Simulated potential of mean force and radius of gyration as the long alkanes lose more configurational entropy upon
for n-, iso-, and neo-pentane in vacuum. entering the pore space. In Figure 10b,d, the pore size effect is
studied for a C1/C2 gas mixture and a C7/C8 liquid mixture.
composition. The preference for the longer alkane partitioning As the pore size decreases from 8 to 2 nm, an increasing
to the pore is ranked as C1/C2 < C2/C3 < C4/C5 < C3/C4. selectivity for the larger molecule is seen for C1/C2 whereas
As we notice, an interesting transition is seen in that C4/C5 the opposite is observed in C7/C8. This suggests that in short
11198 DOI: 10.1021/acs.langmuir.7b02055
Langmuir 2017, 33, 11189−11202
Langmuir Article

Figure 12. Selectivity of n-pentane and iso-pentane to neo-pentane in H = 1.5 nm at T = 300 K when the bulk is (a) a vapor phase or (b) a liquid
phase with equimolar composition.

Figure 13. Ratio between the pore fraction and bulk fraction of C1/C2/C3/C4/C5 in H = 1.5 nm at T = 300 K when the bulk is (a) a vapor phase
or (b) a liquid phase with molar composition C1/C2/C3/C4/C5 = 5/4/3/2/1.

alkane adsorption the selectivity is mainly determined by the This can be verified by the potential of mean force calculation
surface affinity; however, for long alkanes in the liquid state, from single-molecule simulations shown in Figure 11. The
adsorption is a competition between attraction to the surface radius of gyration of each isomer is also presented because we
and the loss of configurational entropy. see that n-pentane is mostly stretched and thus has the
Competitive Adsorption of Multicomponent Systems. strongest interaction with the surface. In partitioning to the
The competitive adsorption of a multicomponent (n > 3) pore space, we see in Figure 12 that n-pentane is most strongly
system is of interest both for shale reservoir characterization adsorbed, with iso-pentane being next. The calculation is
and other applications such as gas separation and chromatog- conducted both for vapor and liquid bulk phases, with the
raphy. Jiang et al.57 performed a configurational-bias Monte vapor phase showing a larger selectivity of linear pentane to
Carlo simulation of linear (C1−nC5) and branched (C5) neo-pentane. Here we show the selectivity as the molar density
alkane mixtures on single-walled carbon nanotube bundles. ratio between one species to neo-pentane. A similar trend is
They discussed the different competitive adsorption behavior at seen in the adsorption in carbon nanotubes. At higher pressure,
low and high pressures, where short alkanes partition to the decreased selectivity is due to the packing limit when
pore more strongly at high pressures. In this section, we have approaching adsorption saturation. Although n-pentane is
performed a similar study and find a similar trend in a slit pore. preferentially adsorbed by the pore, as the pressure increases
We first study the adsorption of an equimolar mixture of C5 the configurational entropy loss of entering the pore would
isomers, namely, n-/iso-/neo-pentane = 1:1:1, at 300 K in a 1.5 hinder the continuous increase in the stretched molecule in
nm slit pore. Because of the different arrangements of carbon pores. If we take a look at a mixture of more dispersed chains,
atoms, the surface attraction force for each molecule is different. as in the next example, a mixture of C1/C2/C3/C4/C5 = 5/4/
11199 DOI: 10.1021/acs.langmuir.7b02055
Langmuir 2017, 33, 11189−11202
Langmuir Article

3/2/1, this pressure effect is more clearly seen. Figure 13a,b surface affinities but also by the change in configurational
⎛ x pore ⎞ entropies, which is also mentioned by Jiang et al.57 Further
shows the partition coefficient ⎜ ibulk ⎟ of this mixture in a 1.5 studies on the selectivies of multicomponent systems, which are
⎝ yi ⎠
nm pore at 300 K with the bulk being a vapor in Figure 13a and branched pentanes and C1−C5 mixtures, have shown similar
a liquid in Figure 13b. Compared to the bulk composition, we trends as reported by simulation57 and also imply the same
see in (a) that pentane strongly adsorbs in the pore with a large mechanism. Selectivities from molecular simulation will be
difference from the bulk concentration. Before capillary considered for future work. Because fluid in shale is a complex
condensation in the pore, the n-pentane fraction continuously mixture confined in nanoscale pores, the partitioning of
increases. However, the n-pentane fraction decreases after different components between fractures and the pore matrix
condensation whereas other molecules increase in adsorption. should be considered to obtain a realistic fluid composition in a
The calculation is stopped before the mixture’s bulk dew point reservoir. As reported by other works,7,8,40 an adjustment of the
pressure, and the fraction of each species approaches a fluid thermodynamic properties is inevitable in reservoir
simulation.


constant. For a liquid mixture in the bulk, the liquid density
changes slowly with the increase in pressure as a result of the
ASSOCIATED CONTENT
packing limit. In Figure 13b, we observe a smaller enhancement
of pentane in the pore compared to the bulk concentration *
S Supporting Information

than in Figure 13a. The greater than 1 partition coefficients of The Supporting Information is available free of charge on the
propane, n-butane, and n-pentane indicate a preference for ACS Publications website at DOI: 10.1021/acs.lang-
these large molecules, but in the case of a bulk liquid, methane muir.7b02055.
is a major component in the pore, similar to its bulk Potential of mean force simulation and critical point data
concentration. Again, this result is consistent with the (PDF)


simulation study on carbon nanotubes, which implies that our
model produces reasonable results. This study also suggests AUTHOR INFORMATION
that a quite different gas composition from produced gas is Corresponding Author
expected to be in the reservoir fluid and the production *E-mail: wgchapman@rice.edu.
simulation should take into account a reasonable molar
ORCID
composition of the reservoir fluid.


Dilip Asthagiri: 0000-0001-5869-0807
Walter G. Chapman: 0000-0002-8789-9041
CONCLUSIONS
Notes
The promising development in shale gas/oil makes it critical to The authors declare no competing financial interest.


better understand the fluid properties in a shale reservoir. The
nanoconfinement effect will greatly shift the fluid thermody- ACKNOWLEDGMENTS
namic properties, which makes the exploration and production
The authors thank The Robert A. Welch Foundation (grant no.
challenging. Because the pore shape, pore size, and solid
C-1241) for financial support.


composition can all affect the fluid behavior, the prediction of
fluid properties in shale networks is an active area of research. REFERENCES
In this work, we present a model to provide a physical
(1) Fu, J.; Tian, Y.; Wu, J. Classical density functional theory for
understanding of the confined alkane properties in shale. As a methane adsorption in metal-organic framework materials. AIChE J.
computationally efficient method, iSAFT has been applied to 2015, 61, 3012−3021.
predict the phase behavior of pure and mixed hydrocarbons in (2) Davis, M. Zeolite-based catalysts for chemicals synthesis.
graphite slit pores. In quantifying the fluid−solid interactions, Microporous Mesoporous Mater. 1998, 21, 173−182.
we have introduced the potential of mean force from molecular (3) Yang, R. T. In Gas Separation by Adsorption Processes; Yang, R. T.,
simulation to evaluate the external potential in a nonlocal DFT Ed.; Butterworth-Heinemann, 1987; pp 1−8.
framework. (4) Luo, S.; Nasrabadi, H.; Lutkenhaus, J. L. Effect of confinement on
To summarize, we have examined the density distribution of the bubble points of hydrocarbons in nanoporous media. AIChE J.
adsorbed molecules in the adsorption of methane and heptane 2016, 62, 1772−1780.
by comparison to molecular simulation. The pore size (5) Cho, H.; Bartl, M. H.; Deo, M. Bubble Point Measurements of
Hydrocarbon Mixtures in Mesoporous Media. Energy Fuels 2017, 31,
dependency of pure methane and pure ethane critical 3436−3444.
properties has been studied, and the phase diagrams were (6) Severson, B. L.; Snurr, R. Q. Monte Carlo simulation of n-alkane
established for methane and ethane in a 3 nm pore. The adsorption isotherms in carbon slit pores. J. Chem. Phys. 2007, 126,
mixture phase behavior was studied by taking the methane and 134708.
ethane mixture as a showcase. The predicted properties are (7) Didar, B. R.; Akkutlu, I. Y. Pore-Size Dependence of Fluid Phase
comparable to molecular simulation, with the errors mainly Behavior and Properties in Organic-Rich Shale ReservoirsSPE
arising from the mean field approximation in treating the International Symposium on Oilfield Chemistry; April 8−10, 2013; The
dispersion energy. After the initial verification of the theory on Woodlands, TX, USA, doi: 10.2118/164099-MS.
pure and binary systems, we predicted the selective adsorption (8) Pitakbunkate, T.; Balbuena, P. B.; Moridis, G. J.; Blasingame, T.
A. Effect of Confinement on Pressure/Volume/Temperature Proper-
of a series of binary (Cn/Cn+1) mixtures up to n-octane. The
ties of Hydrocarbons in Shale Reservoirs. SPE J. 2016, 21, 621−634.
selectivities have been shown in different sized pores, where we (9) Hu, Y.; Devegowda, D.; Striolo, A.; Ho, T. A.; Phan, A.; Civan, F.;
see for short alkanes in the gaseous state that Cn+1 always Sigal, R. F. A Pore Scale Study Describing the Dynamics of Slickwater
partitions more to the pore whereas for longer alkane mixtures Distribution in Shale Gas Formations Following Hydraulic Fractur-
in the liquid state this is not always true. This observation ing.SPE Unconventional Resources Conference-USA; April 10−12, 2013,
implies that selective adsorption is determined not only by The Woodlands, TX, USA; doi: 10.2118/164552-MS.

11200 DOI: 10.1021/acs.langmuir.7b02055


Langmuir 2017, 33, 11189−11202
Langmuir Article

(10) Hu, Y.; Devegowda, D.; Sigal, R. F. Impact of Maturity on (32) Dominik, A.; Tripathi, S.; Chapman, W. G. Bulk and Interfacial
Kerogen Pore Wettability: A Modeling Study. SPE Annual Technical Properties of Polymers from Interfacial SAFT Density Functional
Conference and Exhibition; doi: 10.2118/170915-MS, October 27−29, Theory. Ind. Eng. Chem. Res. 2006, 45, 6785−6792.
2014, Amsterdam, The Netherland. (33) Jain, S.; Chapman, W. G. Effect of confinement on the ordering
(11) Hu, Y.; Devegowda, D.; Striolo, A.; Phan, A.; Ho, T. A.; Civan, of symmetric diblock copolymers: application of interfacial statistical
F.; Sigal, R. F. Microscopic Dynamics of Water and Hydrocarbon in associating fluid theory. Mol. Phys. 2009, 107, 1−17.
Shale-Kerogen Pores of Potentially Mixed Wettability. SPE J. 2014, 20, (34) Jain, S.; Jog, P.; Weinhold, J.; Srivastava, R.; Chapman, W. G.
112−124. Modified interfacial statistical associating fluid theory: Application to
(12) Ungerer, P.; Collell, J.; Yiannourakou, M. Molecular Modeling tethered polymer chains. J. Chem. Phys. 2008, 128, 154910.
of the Volumetric and Thermodynamic Properties of Kerogen: In fl (35) Emborsky, C. P.; Cox, K. R.; Chapman, W. G. Exploring
uence of Organic Type and Maturity. Energy Fuels 2015, 29, 91−105. parameter space effects on structure-property relationships of
(13) Collell, J.; Galliero, G.; Gouth, F.; Montel, F.; Pujol, M.; surfactants at liquid-liquid interfaces. J. Chem. Phys. 2011, 135, 084708.
Ungerer, P.; Yiannourakou, M. Molecular simulation and modelisation (36) Wang, L.; Haghmoradi, A.; Liu, J.; Xi, S.; Hirasaki, G. J.; Miller,
of methane/ethane mixtures adsorption onto a microporous molecular C. A.; Chapman, W. G. Modeling micelle formation and interfacial
model of kerogen under typical reservoir conditions. Microporous properties with iSAFT classical density functional theory. J. Chem.
Mesoporous Mater. 2014, 197, 194−203. Phys. 2017, 146, 124705.
(14) Vasileiadis, M.; Peristeras, L. D.; Papavasileiou, K. D.; (37) Borowko, M.; Patrykiejew, A.; Rysko, W.; Sokolowski, S.;
Economou, I. G. Modeling of Bulk Kerogen Porosity: Methods for Ilnytskyi, J. Complex phase behavior of a fluid in slits with
Control and Characterization. Energy Fuels 2017, 31, 6004−6018. semipermeable walls modified with tethered chains. J. Chem. Phys.
(15) Rosenfeld, Y. Free-energy model for the inhomogeneous hard- 2011, 134, 044705.
sphere fluid mixture and density-functional theory of freezing. Phys. (38) Borówko, M.; Patrykiejew, A.; Sokołowski, S. Unusual
Rev. Lett. 1989, 63, 980−983. mechanism of capillary condensation in pores modified with chains
(16) Wertheim, M. S. Fluids with highly directional attractive forces. forming pillars. J. Chem. Phys. 2011, 135, 054703.
I. Statistical Thermodynamics. J. Stat. Phys. 1984, 35, 19−34. (39) Hlushak, S. P.; Cummings, P. T.; Mccabe, C. Comparison of
(17) Wertheim, M. S. Fluids with highly directional attractive forces. several classical density functional theories for the adsorption of
II. Thermodynamic perturbation theory and integral equations. J. Stat. flexible chain molecules into cylindrical nanopores. J. Chem. Phys.
Phys. 1984, 35, 35−47. 2013, 139, 234902.
(18) Wertheim, M. S. Fluids with Highly Directional Attractive (40) Li, Z.; Jin, Z.; Engineering, R. Phase Behavior and Adsorption of
Forces.III. Multiple Attraction Sites. J. Stat. Phys. 1986, 42, 477−492. Pure Substances and Mixtures and Characterization in Nanopore
(19) Wertheim, M. S. Fluids with highly directional attractive forces.
Structures by Density Functional Theory. SPE J. 2014, 19, 1097−
IV. Equilibrium Polymerization. J. Stat. Phys. 1986, 42, 459−476.
1109.
(20) Chapman, W. Theory and Simulation of Associating Liquids.
(41) Chapman, W. G.; Gubbins, K. E.; Jackson, G.; Radosz, M. New
Ph.D. Dissertation, Cornell University, Ithaca, NY, 1988.
reference equation of state for associating liquids. Ind. Eng. Chem. Res.
(21) Segura, C. J.; Chapman, W. G.; Shukla, K. P. Molecular Physics
Associating fluids with four bonding sites against a hard wall: density 1990, 29, 1709−1721.
(42) Gross, J.; Sadowski, G. Perturbed-Chain SAFT: An Equation of
functional theory Associating fluids with four bonding sites against a
State Based on a Perturbation Theory for Chain Molecules. Ind. Eng.
hard wall: density functional theory. Mol. Phys. 1997, 90, 759−771.
(22) Tripathi, S.; Chapman, W. G. Microstructure of inhomogeneous Chem. Res. 2001, 40, 1244−1260.
polyatomic mixtures from a density functional formalism for atomic (43) Shen, G.; Ji, X.; Lu, X. A hybrid perturbed-chain SAFT density
mixtures. J. Chem. Phys. 2005, 122, 094506. functional theory for representing fluid behavior in nanopores. J. Chem.
(23) Peterson, B. K.; Heffelfinger, G. S.; Gubbins, K. E.; van Swol, F. Phys. 2013, 138, 224706.
Layering transitions in cylindrical pores. J. Chem. Phys. 1990, 93, 679− (44) Shen, G.; Ji, X.; Ö berg, S.; Lu, X. A hybrid perturbed-chain
685. SAFT density functional theory for representing fluid behavior in
(24) Sokolowski, S.; Fischer, J. Lennard-Jones mixtures in slit-like nanopores: Mixtures. J. Chem. Phys. 2013, 139, 194705.
pores: a comparison of simulation and density-functional theory. Mol. (45) Klink, C.; Gross, J. A Density Functional Theory for Vapor
Phys. 1990, 71, 393−412. Liquid Interfaces of Mixtures Using the Perturbed-Chain Polar
(25) Yu, Y.-X.; Wu, J. Density functional theory for inhomogeneous Statistical Associating Fluid Theory Equation of State. Ind. Eng.
mixtures of polymeric fluids. J. Chem. Phys. 2002, 117, 2368−2376. Chem. Res. 2014, 53, 6169−6178.
(26) Tripathi, S.; Chapman, W. G. A density functional approach to (46) Xu, X.; Cristancho, D. E.; Costeux, S.; Wang, Z.-G. Density-
chemical reaction equilibria in confined systems: Application to functional theory for polymer-carbon dioxide mixtures: A perturbed-
dimerization. J. Chem. Phys. 2003, 118, 7993−8003. chain SAFT approach. J. Chem. Phys. 2012, 137, 054902.
(27) Haghmoradi, A.; Wang, L.; Chapman, W. G. A density (47) Klink, C.; Gross, J. A Density Functional Theory for Vapor-
functional theory for colloids with two multiple bonding associating Liquid Interfaces of Mixtures Using the Perturbed-Chain Polar
sites. J. Phys.: Condens. Matter 2016, 28, 244009. Statistical Associating Fluid Theory Equation of State. Ind. Eng.
(28) Landers, J.; Gor, G. Y.; Neimark, A. V. Density functional theory Chem. Res. 2014, 53, 6169−6178.
methods for characterization of porous materials. Colloids Surf., A (48) Tarazona, P.; Cuesta, J.; Martínez-Ratón, Y. In Theory and
2013, 437, 3−32. Simulation of Hard-Sphere Fluids and Related Systems; Mulero, Á , Ed.;
(29) Gor, G. Y.; Thommes, M.; Cychosz, K. A.; Neimark, A. V. Springer: Berlin, 2008; pp 247−341.
Quenched solid density functional theory method for characterization (49) Tang, Y.; Lu, B. C.-Y. Analytical description of the Lennard-
of mesoporous carbons by nitrogen adsorption. Carbon 2012, 50, Jones fluid and its application. AIChE J. 1997, 43, 2215−2226.
1583−1590. (50) Gross, J. A density functional theory for vapor-liquid interfaces
(30) Ravikovitch, P. I.; Neimark, A. V. Density Functional Theory using the PCP-SAFT equation of state. J. Chem. Phys. 2009, 131,
Model of Adsorption on Amorphous and Microporous Silica 204705.
Materials. Langmuir 2006, 22, 11171−11179. (51) Steele, W. A. The physical interaction of gases with crystalline
(31) Jain, S.; Dominik, A.; Chapman, W. G. Modified interfacial solids. Surf. Sci. 1973, 36, 317−352.
statistical associating fluid theory: A perturbation density functional (52) Crowell, A. D.; Steele, R. B. Interaction Potentials of Simple
theory for inhomogeneous complex fluids. J. Chem. Phys. 2007, 127, Nonpolar Molecules with Graphite. J. Chem. Phys. 1961, 34, 1347−
244904. 1349.

11201 DOI: 10.1021/acs.langmuir.7b02055


Langmuir 2017, 33, 11189−11202
Langmuir Article

(53) Darve, E.; Rodríguez-Gómez, D.; Pohorille, A. Adaptive biasing


force method for scalar and vector free energy calculations. J. Chem.
Phys. 2008, 128, 144120.
(54) Thermophysical Properties of Fluid Systems. http://webbook.
nist.gov/chemistry/fluid/.
(55) Phillips, J. C.; Braun, R.; Wang, W.; Gumbart, J.; Tajkhorshid,
E.; Villa, E.; Chipot, C.; Skeel, R. D.; Kalé, L.; Schulten, K. Scalable
molecular dynamics with NAMD. J. Comput. Chem. 2005, 26, 1781−
1802.
(56) Martin, M. G.; Siepmann, J. I. Transferable Potentials for Phase
Equilibria. 1. United-Atom Description of n-Alkanes. J. Phys. Chem. B
1998, 102, 2569−2577.
(57) Jiang, J.; Sandler, S. I.; Schenk, M.; Smit, B. Adsorption and
separation of linear and branched alkanes on carbon nanotube bundles
from configurational-bias Monte Carlo simulation. Phys. Rev. B:
Condens. Matter Mater. Phys. 2005, 72, 45447.

11202 DOI: 10.1021/acs.langmuir.7b02055


Langmuir 2017, 33, 11189−11202

View publication stats

You might also like