Molecular Simulations of Supercritical Fluid Permeation through

Subscriber access provided by Universitätsbibliothek | Technische Universität München

Letter

Molecular Simulations of Supercritical Fluid Permeation through Disordered Microporous Carbons Alexandru Botan, Romain Vermorel, Franz-Josef Ulm, and Roland Jean-Marc Pellenq Langmuir, Just Accepted Manuscript • DOI: 10.1021/la402087r • Publication Date (Web): 25 Jul 2013 Downloaded from http://pubs.acs.org on July 27, 2013

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Langmuir is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 19

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Langmuir

Molecular simulations of supercritical fluid permeation through disordered microporous carbons Alexandru Bo¸tan,∗,† Romain Vermorel,†,§ Franz-Josef Ulm,†,‡ and Roland J.-M. Pellenq∗,†,‡,¶ Department of Civil and Environmental Engineering, Massachusetts Institute of Technology, Cambridge, MA, USA, MSE2 , UMI 3466 CNRS-MIT, Massachusetts Institute of Technology, Cambridge MA, USA, and CINaM, Aix Marseille University, 13288 Marseille, FRANCE E-mail: [email protected]; [email protected]

Abstract Fluid transport through microporous carbon-based materials is inherent in numerous applications, ranging from gas separation by carbon molecular sieves to natural gas production from coal seams and gas shales. The present study investigates the steady-state permeation of supercritical methane in response to a constant cross-membrane pressure drop. We performed dual control volume grand canonical molecular dynamics (DCV-GCMD) simulations to mimic the conditions of actual permeation experiments. To overcome arbitrary assumptions regarding the investigated porous structures, the membranes were modeled after the CS1000a and CS1000 molecular models, representative of real microporous carbon materials. When ∗ To

whom correspondence should be addressed

† MIT ‡ MSE2 ¶ CINaM § Now

at LFC-R, University of Pau, 64013 Pau, FRANCE

1 ACS Paragon Plus Environment

Langmuir

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

adsorption induced molecular trapping (AIMT) mechanisms are negligible, we show that the permeability of the microporous material, while not significantly sensitive to the pressure gradient, monotonically decreases with temperature and reservoirs pressures consistently with diffusion theory. On the other hand, when AIMT occurs, the permeability increases with temperature in agreement with experimental data found in the literature.

Introduction The problem of fluid transfer through carbon-based disordered microporous materials is inherent in numerous filtration processes. Water and wastewater treatment, 1 gas separation, 2 energy storage, 3 carbon dioxide sequestration or gas production from coal seams and gas shales fall in that category of applications for which the prediction of the permeability is of utmost practical importance. Aside from thermodynamical parameters, such as temperature, bulk fluid pressure and pressure gradient, other structural parameters greatly impact the permeability of microporous systems, namely the nature of the permeating fluid, the chemical composition of the solid and the texture of the porous network. Except for idealized porous structures, 4–6 the effect of these factors is not analytically defined but can be investigated by experiments, including numerical ones. Molecular modeling has indeed been widely used to study permeation and selectivity in disordered microporous carbons assuming slit pore geometries. 7–10 In such idealized porous structures, adsorption does not prevent the fluid molecules to diffuse through all the porous volume. This results in a decrease of the molar flux with temperature, consistently with the decrease of the interstitial concentration. In actual experiments however, measurements performed on real disordered microporous carbons reveal the opposite behavior. 11,12 Two different arguments may therefore be invoked to explain these discrepancies: firstly, simply accounting for a more realistic texture of the pore walls might imply additional effects not captured by regular slit pores models; On the other hand, single slit pore models might fail because they disregard the effects of constrictions as only one percolating pore is considered. 13,14 The present work investigates both possibilities. In this paper, we study the permeability of CS1000a and CS1000 15 as models for disordered 2 ACS Paragon Plus Environment

Page 2 of 19

Page 3 of 19

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Langmuir

microporous carbons. Due to the presence of a wide percolating pore, the CS1000a molecular model can be thought of as a slit pore with enhanced pore walls texture. In contrast, the CS1000 exhibits a pore size distribution conducive to Adsorption Induced Molecular Trapping (AIMT) effects as a result of numerous pore constrictions. The steady-state permeation of a single component fluid through a disordered microporous membrane is addressed. We use dual control volume - grand canonical molecular dynamics (DCV-GCMD) to mimic the experimental conditions of supercritical methane (CH4 ) flow driven by a constant pressure gradient for a wide range of temperatures and pressures. Our goal consists in overcoming the assumptions of previous studies by considering a realistic porous carbon molecular structure. In the case of CS1000a, a model based on Maxwell-Stefan diffusion theory 16,17 coupled to a Langmuir adsorption isotherm reproduces the simulated permeabilities, thus confirming that diffusion and adsorption are the dominating transport mechanisms as long as AIMT can be neglected. However, only the results obtained for CS1000 turn out to be consistent with experimental data found in the literature, which suggests that the description of transport properties in real microporous carbons does involve strong AIMT mechanisms.

Model and simulation schemes Principles of the DCV-GCMD numerical experiment Our numerical experiment consists in reproducing an experimental set-up used for permeability measurements, as illustrated in Figure 1. For that purpose we use the DCV-GCMD method. 18–20 High and low fugacity reservoirs are imposed at each end of the membranes. This allows the application of periodic boundary conditions. While the fugacities in these reservoirs are controlled by means of Grand Canonical Monte Carlo simulation, molecular motions are described using Molecular Dynamics simulation. Once the system has reached the steady-state regime, the molar flux J is estimated by counting the number of molecules N crossing the membrane of cross section


Langmuir

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Low fugacity region

Page 4 of 19

High fugacity region

l

Low fugacity region

l

4l flux

4l l

l

2l Membrane

Membrane

Figure 1: Schematic representation of DCV-GCMD method. In our simulations the membrane thickness stands as one CS1000a unit cell: l = 2.5 nm. S during a time interval ∆t: N . S∆t

J=

(1)

The permeability is then defined as the flux per unit of fugacity gradient (∆ f /l) across the membrane: Pe =

Jl . ∆f

(2)

Membrane and fluid models As microporous membranes, we chose two models of porous carbons, CS1000a and CS1000, with main characteristics listed in Table 1. These models were obtained by Jain and co-workers 15 as results of Hybrid Reverse Monte Carlo (HRMC) simulations, where arrangement of the atoms was driven by systematic refinement of simulated radial distribution functions to match X-ray scattering experimental data and simultaneous energy minimization. The pore size distributions and pair correlation functions shown in Figure 2(c) clearly indicate that both models are disordered polydisperse microporous materials. Furthermore, the CS1000 model was recently used to model CH4 and CO2 adsorption induced swelling in coal, in quantitative agreement with experimental data. 21 We can therefore consider these molecular models as fairly representative of disordered microporous carbons. In the following of this paper, we study the flow of pure CH4 through CS1000a and CS1000. 4 ACS Paragon Plus Environment

Page 5 of 19

CS1000a

(a)

CS1000

(b)

2.5 nm

2.5 nm CH4 diameter

(c)

1

10 8 6 4 2 0 0

g (r)

CS1000

0.8 p (H)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Langmuir

0.6

2

CS1000a

0.4

4 r (Å)

6

8

0.2 0 0

5

10

15

H (Å)

Figure 2: (a), (b) Snapshots of CS1000a and CS1000 cubic unit cells respectively (2.5 nm). Carbon and hydrogen atoms appear in cyan and grey respectively. (c) Pore size distributions of CS1000a (red solid line) and CS1000 (black dashed line). The vertical black solid line indicates the diameter of CH4 molecules. Inset figure: Radial distribution functions of CS1000a (red solid line) and CS1000 (black dashed line).

Table 1: Structural properties of CS1000a and CS1000 models. 15 property linear size (nm) carbon density (atoms/Å3 ) H/C ratio pore diameter (nm) porosity

CS1000a CS1000 2.5 2.5 0.03621 0.0743 0.091 0.149 0.3-1.3 0.31-0.57 0.667 0.088


Langmuir

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 6 of 19

The CH4 molecules, as well as the CS models carbon and hydrogen atoms are described as simple Lennard-Jones (LJ) interaction sites: U(ri j ) = 4εi j

[( σ )12 ij

ri j

−

( σ )6 ] ij

ri j

(3)

with parameters shown in Table 2. Here ri j is the distance between sites i and j, σi j and εi j are LJ parameters deduced from the conventional Lorentz-Berthelot combining rules. 22 Because the CH4 molecule is symmetric and non-polar, its representation by a single LJ site is possible. Furthermore, the microporous adsorbent is considered infinitely rigid. Therefore, we neglect the internal degrees of freedom of the CH4 molecules as well as the many-body interactions that would arise from the flexibility of the adsorbent. Aside from quantitative corrections on the simulated diffusion coefficients and permeabilities, 23 we do not expect these additional contributions to significantly change the phenomenology observed in our numerical experiments. Table 2: Lennard-Jones potential parameters. Site type ε (kJ/mol) methane 1.230 carbon 0.232 hydrogen 0.125

σ (Å) 3.73 3.36 2.42

The DCV-GCMD set-up, described in the previous section, allowed us to investigate the permeation of CH4 through the CS models in the three directions of space. The permeability measured for the different spatial orientations of a given membrane was the same within a proportionality constant lower than 2. In this paper, we therefore report the results obtained for the orientation that yields intermediate values of permeability only. Furthermore, as observed in Figure 2(a), the CS1000a exhibits a 1.3 nm wide pore percolating through the sample. As the pore width is larger than several CH4 molecular diameters (0.38 nm), this guarantees that AIMT mechanisms have little influence on the overall permeability of the system. On the other hand, the average pore size of CS1000 (0.41 nm) is close to the molecular diameter of CH4 and no large pore percolates through the sample as can be seen in Figure 2(b). This means that the CS1000 structure is conducive to 6 ACS Paragon Plus Environment

Page 7 of 19

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Langmuir

AIMT effects with respect to methane molecules.

Simulation details All simulations presented in this study are based on two types of techniques: Grand Canonical Monte Carlo simulations (GCMC) and Molecular Dynamics. 22 Specifically, GCMC insertion/deletion moves are used to maintain a constant chemical potential difference between the low and high fugacity reservoirs during the DCV-GCMD simulation. The chemical potential of CH4 in the reservoirs, which is an input parameter for DCV-GCMD simulations, is estimated thanks to the Widom test particle method 24 performed during NVT simulations of bulk methane. Molecular Dynamics simulations were used to move particles as no MC displacement steps were performed. The simulation box in the DCV-GCMD method (see Figure 1) is divided into five regions: two low fugacity reservoirs 5 nm thick (2 CS unit cells), two membranes each 2.5 nm thick (one CS unit cell) and one central high density reservoir 10 nm thick (4 CS unit cells). The transverse dimensions of the simulation box are 25 nm (10 unit cells) each. In order to maintain constant temperature throughout the simulation box, each region is coupled to its own (Nose-Hoover) thermostat with a damping constant of 100 fs. The temperature was then defined by subtracting the streaming velocity from the x component (flux direction) of the velocities. Typically, we imposed ten GCMD insertion or deletion steps every one MD step. The permeability calculations were run for 5 ns and we excluded the first 0.5 ns corresponding to the transient regime.

Determination of the collective diffusivity and adsorption isotherm In addition, we performed equilibrium molecular dynamics (EMD) to calculate the collective diffusivity Do and adsorption isotherm of CH4 molecules in the CS1000a membrane, both of which are required for the theoretical prediction of the permeability as discussed further. At the temperature T = 150 ◦ C, the collective diffusivity was found almost constant over a range of pressures p = 10 − 600 bar. Fitting therefore the collective diffusivity by a constant value yields Do = (3.65 ± 0.4) × 10−8 m2 .s−1 , which is of the same order of magnitude as the data reported by 7 ACS Paragon Plus Environment

Langmuir

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 8 of 19

Lim and Bhatia, who studied the permeation of CH4 molecules through carbon slit pores. 14 For comparison, pulsed field gradient NMR experiments performed on carbon molecular sieves (CMS) membranes yield a typical self diffusivity of 1 × 10−11 m2 .s−1 . 25 This difference results from the CS1000a pore size distribution, which presents significantly wider pores than usual CMS membranes materials. Simulation of the CH4 sorption isotherm in CS1000a was carried out using the DCV-GCMD technique, in which upstream and downstream fugacities were set equal to each other. This approach guarantees the fluid is adsorbed into accessible pores only. The simulated adsorption isotherm for T = 150 ◦ C agrees with the Langmuir model (see eq 6) for the range of investigated fugacities. The best-fitting parameters are cs = 10.49 mmol.m−3 and b = 4.52 × 10−3 bar−1 . More details about these simulations are provided in the accompanying supplementary information.

Theoretical framework From the thermodynamical point of view, a system of fluid particles sorbed in a immobile porous medium deviates from equilibrium when a gradient in molecules chemical potential exists. Under isothermal conditions and on the assumption that transport mechanism is diffusive, the local molar flux ⃗j (number of mols of fluid per unit surface per unit time) through the fixed porous solid satisfies the Maxwell-Stefan equation 16,26 : ⃗j = − cDo ⃗∇µ , RT

(4)

where T is the temperature, R the ideal gas constant, c the average interstitial concentration (number of moles per material unit volume), while Do stands as the collective diffusivity of the sorbed fluid, as previously discussed. Moreover, using the definition of the chemical potential

µ ( f , T ) ≡ µ0 (T ) + RT ln( f / f0 ), one can alternatively consider the gradient in fugacity as the driving force of fluid motion. Hence the rearranged expression of the local molar flux: ⃗j = − cDo ⃗∇ f , f


(5)

Page 9 of 19

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Langmuir

where f is the fluid fugacity. In order to estimate the concentration in the microporous membrane, we refer to the classical Langmuir model, commonly used to describe adsorption isotherms of fluids in microporous adsorbents: c = cs

bf . 1+bf

(6)

Here cs is the complete filling concentration and b is an equilibrium adsorption constant, which can be interpreted as the inverse of a characteristic filling pressure. We stress that we use the Langmuir model for its ability to reproduce the adsorption isotherms simulated in our membrane models (see supplementary information) and its convenient analytical form. Let us now consider a microporous membrane of thickness l in the x direction and separating two infinite bulk fluid reservoirs exhibiting a difference in chemical potentials. Under these conditions eq 5 describes the local motion of interstitial fluid in the membrane. In an actual experiment, one can only measure the total molar flux J⃗ as a function of the fugacity drop ∆ f across the membrane. Hence the averaged transport equation, obtained from the average of eq 5 along the thickness of the membrane: 1 ∆f ⃗ex = − J⃗ = −Pe l l

∫ x=l x=0

Do c dx f

(

∂f ∂x

) ⃗ex ,

(7)

in which Pe is referred to as the permeability of the membrane. It should be stressed that this definition of the permeability differs from the classical definition deriving from Darcy’s law, which considers the viscous flow of a Newtonian interstitial fluid. In the present case, for the sake of generality we define the permeability as the transport coefficient relating the molar flux to the driving force of fugacity gradient, as found in the literature. 17 Assuming a constant Do , we deduce the overall permeability Pe of the membrane from eq 7 as Do Pe = ∆f

∫ fd

df fu

c , f


(8)

Langmuir

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 10 of 19

in which fd and fu are the downstream and upstream fugacities respectively. Finally, combining eq 8 with eq 6 the permeability is given as a function of ∆ f and fd ( ) D o cs b∆ f Pe = ln 1 + . ∆f 1 + b fd

(9)

Results and Discussion Effects of downstream pressure and pressure drop We investigated the permeation of methane through the CS1000a membrane for downstream pressures in the range of 0 to 300 bar at fixed temperature T = 150 ◦ C and constant pressure drop. Figure 3(a) compares the theoretical prediction obtained for Do = 3.65 ± 0.4 × 10−8 m2 .s−1 with the permeability measured in DCV-GCMD numerical experiments according to eq 2. If the model captures the evolution of permeability with downstream pressure, we observe a systematic overestimation that can be corrected upon multiplying eq 9 by an adjusted proportionality coefficient equal to 0.85. Accounting for the uncertainty from the estimation of Do during EMD simulations, the overestimation of the permeability by the theory falls in the range of 4% to 23%. We attribute this discrepancy to AIMT involved in the narrowest pores of the CS1000a structure. As a result, some fluid particles present in the membrane are trapped in the smallest pores and do not contribute to the molar flux as opposed to the assumption leading to eq 9. However, while AIMT occurs in the narrowest pores, the permeability of the CS1000a is mostly driven by the diffusion and adsorption in the widest percolating pore, which explains the fair agreement with the theory. Furthermore, in Figure 3(a) the permeability of CS1000a and CS1000 are plotted versus downstream pressure. Vieira-Linhares and Seaton report permeabilities as high as 2 × 104 Barrers. 27 However, usual values of CH4 permeabilities in real disordered microporous carbon membranes range from 0.1 to 1 × 103 Barrers. 28,29 The simulated permeabilities of CS1000a therefore exceed the experimental data by one to two orders of magnitude. This result is not surprising: the thickness of the simulated membrane (l = 2.5 nm) is comparable to the size of the widest pore (1.3 nm), 10 ACS Paragon Plus Environment

Page 11 of 19

×10 4

5

(a)

3 6 ×10

CS1000a

Permeability, (Barrer)

0

20

40

60

80

3

2

· 1 0 (b)

CS1000

4 2 0

4

9

50

· 100 150 200 250 Downstream Pressure, (bar)

300

×10 4 CS1000a

7

Arrhenius fit : Po = 8470 Barrer E p = -5.2 kJ/mol

5 Permeability, (Barrer)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Langmuir

3 4

×10 3 CS1000

3 2 Arrhenius fit : Po = 12950 Barrer E p = 6.5 kJ/mol

1 300

400 500 Temperature, (K)

600

Figure 3: (a) Downstream pressure dependence of the permeability of CS1000a at T = 150 ◦ C for two fixed pressure gradients: ∆p = 10 bar (black hollow squares, dashed black line) and ∆P = 25 bar (red dots, solid red line). The lines stand for the theoretical predictions (eq 9) with a correction factor adjusted to 0.85. Inset: data obtained for CS1000 in the same conditions and ∆P = 25 bar. The dashed line is a guide for the eye. (b) Temperature dependence of the permeability of CS1000a (red dots, solid red line) and CS1000 (black hollow squares, dashed line) for pu = 50 bar and pd = 25 bar. The lines show the fits of the Arrhenius equation (see eq 11).


Langmuir

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 12 of 19

the latter providing a percolating path for the fluid molecules. As a result, simulated permeabilities are significantly higher than measurements performed on 15 nm to 2 µ m thick microporous carbon membranes in which no percolating micropore should be observed. On the other hand, the permeabilities obtained for CS1000, if somewhat high, are consistent with experimental data found in the literature. This result suggests that AIMT cannot be neglected in actual experiments. For both models, we observe a monotonic decrease of the permeability with downstream pressure as already observed with slit pores models. 7–10 While the same behavior was observed for CMS hollow fibers membranes, 30 experiments performed on other microporous carbons evidence that the evolution of permeability with pressure strongly depends on the nature of the sorbed fluid, more precisely when AIMT becomes important. 12 Furthermore, we observe no significant effect of the pressure drop on the permeability of CS1000a in the conditions of the simulations. Noting that the typical order of magnitude of b∆ f ≃ 10−2 − 10−1 is small compared to unity, from eq 9 we deduce the following approximation of the permeability: Pe ≃

bDo cs , 1 + b fd

(10)

which is indeed independent of ∆ f , and therefore independent of the pressure drop across the membrane. In addition, the model predicts a permeability decaying to zero for very high reservoir pressures. However, the range of reservoir pressures investigated in the present work did not allow us to observe this behavior.

Effect of temperature Figure 3(b) illustrates the typical effect of temperature on methane permeability, for fixed upstream and downstream reservoir pressures, set to pu = 50 bar and pd = 25 bar respectively. Depending on the membrane model we find different trends: permeation of CH4 molecules decreases with temperature for CS1000a, whereas we observe the opposite behavior with CS1000. Classically, the temperature dependency of membranes permeability is described with an empirical Arrhenius


Page 13 of 19

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Langmuir

law:

( ) Ep Pe = P0 exp − , RT

(11)

where E p is the permeation activation energy. The fit by the empirical Arrhenius law reported in Figure 3(b) yields a negative permeation activation energy E p = −5.2 kJ.mol−1 for CS1000a. This result is consistent with other numerical studies on graphitic slit pores, 14 but contradicts numerous experiments performed on real microporous carbons. On the contrary, CS1000 exhibits a positive activation energy E p = +6.5 kJ.mol−1 in good agreement with experimental data found in the literature. 11 This result further demonstrates that the AIMT mechanism, also referred to as the pore constriction or pore mouth effect by some authors, 13,14 does stand as the limiting process in actual permeation experiments. 11,12,31 As a result of AIMT, fluid particles diffusion is confined to local areas of the membrane. Therefore, fluid molecules need to overcome energy barriers in order to permeate through the constrictions and explore adjacent pores. Consecutively, rising the temperature increases the flux by releasing trapped molecules in the flow, which is consistent with models based on transition state theory. 13 In the case of CS1000a such effects do not prevail, which explains the decrease of permeability with temperature as further discussed.

Discussion on the diffusive/adsorptive regime Table 3: Estimation of the contribution from temperature and pressure on the permeability of CS1000a. Pe∗ was computed assuming the molar flux is given by the product of the interstitial fluid concentration c and their mean velocity ν in the flow direction. T (C) pd (bar) c (kmol/m3 ) 50 150 6.868 50 25 2.391 350 25 0.379

ν (m/s) 1.9 8.4 18.9

Pe∗ (Barrer) 62420 73263 16139

Regarding the CS1000a, we observed the monotonic decrease of the permeability of the membrane with respect to the reservoir pressures and temperature. We can intuitively understand this behavior by arguably neglecting the effect of AIMT mechanisms and considering the influence of adsorption and diffusion only. Because of the importance of fluid adsorption in the microp13 ACS Paragon Plus Environment

Langmuir

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

orous membrane, the two main contributions to the flux of fluid particles through the microporous membrane, namely the interstitial fluid concentration c and the average transport velocity ⃗ν , are intertwined. For instance, as adsorption weakens with increasing temperature or reducing pressure, the interstitial concentration decays, which in turn favors the mobility of the fluid molecules. Reducing the temperature or raising the reservoir pressure triggers the exact opposite mechanism. Table 3 reports the permeability Pe∗ resulting from the approximation J ≈ cν , for a fixed pressure drop ∆p = 25 bar and different values of reservoir pressures and temperatures. This approximation qualitatively reproduces the direct measurements reported in Figure 3. For a given pressure drop ∆p, we can describe the influence of reservoir pressures and temperature on permeability as follows: 1. At fixed temperature, upon increasing the reservoir pressures the loss in mobility of the fluid particles is more important than the gain in interstitial concentration. Consecutively, the permeability decreases with reservoir pressure. Data from Table 3 obtained at T = 50 ◦ C illustrates well this behavior. 2. If the temperature is raised while the pressures in each reservoirs are held constant, the loss in concentration is greater than the gain in particle mobility, thus leading to a decrease of the permeability, as observed for the data obtained at pd = 25 bar and reported in Table 3.

Conclusion In this study, we have used DCV-GCMD to investigate pure methane permeation through two molecular models of disordered microporous carbon membranes, namely CS1000a and CS1000. In the case of CS1000a, we identified the decrease of interstitial fluid molecules mobility and concentration as the two main mechanisms responsible for the decay of permeability with pressure and temperature respectively. This phenomenology was already observed with simpler slit pores models and is characteristic of purely diffusive/adsorptive permeation as AIMT can be disregarded. A theoretical model based on the simplified Maxwell-Stefan equation and the Langmuir adsorption 14 ACS Paragon Plus Environment

Page 14 of 19

Page 15 of 19

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Langmuir

isotherm indeed captures the evolution of permeability with respect to reservoirs pressures. In the case of CS1000, which exhibits significantly narrower pores along with more tortuous flow paths, AIMT comes into play. As a result, the permeability of CS1000 increases with temperature consistently with most of the experiments performed on real microporous carbons.

Acknowledgement The authors thank Shell and Schlumberger for financial support through the MIT X-Shale research project. Romain Vermorel acknowledges financial support from the Fulbright Program through the Fulbright Aquitaine Program.

Supporting Information Available “Estimation of the collective diffusivity and adsorption isotherm by means of equilibrium molecular dynamics simulations, investigation of the influence of the ratio of MC to MD steps in DCV-GCMD simulations.”

This material is available free of charge via the Internet at http:

//pubs.acs.org/.

References (1) Pollard, S.; Fowler, G.; Sollars, C.; Perry, R. Low-cost adsorbents for waste and wastewater treatment: a review. Sci. Total Environ. 1992, 116, 31 – 52. (2) Ismail, A. F.; David, L. I. B. A review on the latest development of carbon membranes for gas separation. J. Membr. Sc. 2001, 193, 1–18. (3) Zhang, L. L.; Zhao, X. S. Carbon-based materials as supercapacitor electrodes. Chem. Soc. Rev. 2009, 38, 2520–2531. (4) Nicholson, D.; Petropoulos, J. Influence of adsorption forces on the flow of dilute gases through porous media. J. Colloid and Interface Sc. 1973, 45, 459 – 466. 15 ACS Paragon Plus Environment

Langmuir

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

(5) Burggraaf, A. Single gas permeation of thin zeolite (MFI) membranes: theory and analysis of experimental observations. J. Membr. Sc. 1999, 155, 45 – 65. (6) Thornton, A.; Hill, A.; Hill, J. A mathematical investigation into nanoscale gas separation: Effects of pore size and temperature. Proceedings of the 2010 International Conference on Nanoscience and Nanotechnology 2010, 87–90. (7) Nicholson, D. Simulation studies of methane transport in model graphite micropores. Carbon 1998, 36, 1511 – 1523. (8) Travis, K. P. Computer simulation investigation of diffusion selectivity in graphite slit pores. Mol. Phys. 2002, 100, 2317–2329. (9) Lim, Y.-I.; Bhatia, S. K.; Nguyen, T. X.; Nicholson, D. Prediction of carbon dioxide permeability in carbon slit pores. J. Membr. Sc. 2010, 355, 186 – 199. (10) Firouzi, M.; Wilcox, J. Molecular modeling of carbon dioxide transport and storage in porous carbon-based materials. Micropor. Mesopor. Mat. 2012, 158, 195 – 203. (11) Centeno, T. A.; Fuertes, A. B. Supported carbon molecular sieve membranes based on a phenolic resin. J. Membr. Sc. 1999, 160, 201–211. (12) Katsaros, F. K.; Steriotis, T. A.; Stubos, A. K.; Mitropoulos, A.; Kanellopoulos, N. K.; Tennison, S. High pressure gas permeability of microporous carbon membranes. Microporous Mater. 1997, 8, 171–176. (13) Nguyen, T. X.; Bhatia, S. K. Pore accessibility of N2 and Ar in disordered nanoporous solids: theory and experiment. Adsorption 2007, 13, 307–314. (14) Lim, Y.-I.; Bhatia, S. K. Simulation of methane permeability in carbon slit pores. J. Membr. Sc. 2011, 369, 319 – 328. (15) Jain, S. K.; Pellenq, R. J.-M.; Pikunic, J. P.; Gubbins, K. E. Molecular Modeling of Porous Carbons Using the Hybrid Reverse Monte Carlo Method. Langmuir 2006, 22, 9942–9948. 16 ACS Paragon Plus Environment

Page 16 of 19

Page 17 of 19

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Langmuir

(16) Maginn, E. J.; Bell, A. T.; Theodorou, D. N. Transport diffusivity of methane in silicalite from equilibrium and nonequilibrium simulations. J. Phys. Chem. 1993, 97, 4173–4181. (17) Makrodimitris, K.; Papadopoulos, G. K.; Theodorou, D. N. Prediction of Permeation Properties of CO2 and N2 through Silicalite via Molecular Simulations. J. Phys. Chem. B 2001, 105, 777–788. (18) Heffelfinger, G. S.; van Swol, F. Molecular modeling of carbon dioxide transport and storage in porous carbon-based materials. J. Chem. Phys. 1994, 100, 7548. (19) MacElroy, J. M. D. Nonequilibrium molecular dynamics simulation of diffusion and flow in thin microporous membranes. J. Chem. Phys. 1994, 101, 5274–5280. (20) Cracknell, R. F.; Nicholson, D.; Quirke, N. Direct molecular dynamics simulation of flow down a chemical potential gradient in a slit-shaped micropore. Phys. Rev. Lett. 1995, 74, 2463–2466. (21) Brochard, L.; Vandamme, M.; Pellenq, R. J.-M.; Fen-Chong, T. Adsorption-Induced Deformation of Microporous Materials: Coal Swelling Induced by CO2-CH4 Competitive Adsorption. Langmuir 2012, 28, 2659–2670. (22) Frenkel, D.; Smit, B. Understanding Molecular Simulation, Second Edition: From Algorithms to Applications (Computational Science), 2nd ed.; Academic Press, 2001. (23) Fritzsche, S.; Wolfsberg, M.; Haberlandt, R. The importance of various degrees of freedom in the theoretical study of the diffusion of methane in silicalite-1. Chemical Physics 2003, 289, 321–333. (24) Widom, B. Some Topics in the Theory of Fluids. J. Chem. Phys. 1963, 39, 2808–2812. (25) Mueller, R.; Kanungo, R.; Kiyono-Shimobe, M.; Koros, W. J.; Vasenkov, S. Diffusion of Methane and Carbon Dioxide in Carbon Molecular Sieve Membranes by Multinuclear Pulsed Field Gradient NMR. Langmuir 2012, 28, 10296–10303. 17 ACS Paragon Plus Environment

Langmuir

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

(26) Hanley, H. J. M. Transport phenomena in fluids. J. Macromol. Sci. Phys. 1969, 3, 365–365. (27) Vieira-Linhares, A.; Seaton, N. Non-equilibrium molecular dynamics simulation of gas separation in a microporous carbon membrane. Chem. Eng. Sci. 2003, 58, 4129?4136. (28) Rao, M. B.; Sircar, S. Nanoporous carbon membrane for gas separation. Gas. Sep. Purif. 1993, 7, 279 – 284. (29) Menendez, I.; Fuertes, A. B. Aging of carbon membranes under different environments. Carbon 2001, 39, 733 – 740. (30) Vu, D. Q.; Koros, W. J.; Miller, S. J. High Pressure CO2/CH4 Separation Using Carbon Molecular Sieve Hollow Fiber Membranes. Industrial & Engineering Chemistry Research 2002, 41, 367–380. (31) Rutherford, S. W.; Nguyen, C.; Coons, J. E.; Do, D. D. Characterization of Carbon Molecular Sieves Using Methane and Carbon Dioxide as Adsorptive Probes. Langmuir 2003, 19, 8335– 8342.


Page 18 of 19

Page 19 of 19

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Langmuir

Graphical TOC Entry Numerical Permeation Experiment Incoming Flux

Outgoing Flux

High Fugacity Reservoir

Microporous Carbon Membrane

Low Fugacity Reservoir


Molecular Simulations of Supercritical Fluid Permeation through

Recommend Documents