Carbon

Capture using Membrane and Adsorp3on Processes

Jennifer Wilcox Department of Energy Resources Engineering RECS Summer School June 5th, 2012

Clean Energy Conversions Team - 2012

Panithita Rochana (PhD) Yangyang Liu (PhD) Ekin Ozdogan (PhD) Jiajun He (PhD) Kyoungjin Lee (PhD) Abby Kirchofer (PhD) Ana Suarez Negreira (PhD, ChemE)

Tao Narakornpijit (MS) Jeremy Homan (UG, Chem) Reza Haghpanah (Post-doc) Dong-Hee Lim (Post-doc) Mahnaz Firouzi (Post-doc) Dawn Geatches (Post-doc) Erik Rupp (Research Assistant)

Agenda
Work and Cost of Carbon Capture N2-selecWve membrane for carbon capture Carbon-based sorbents for carbon capture

Minimum Work for Separa3on

combined rst and second laws

B B CO CO C C W min = RT [n CO2 ln(y CO2 ) + n B CO2 ln(y B CO2 )] + RT [nC 2 ln(yC 2 ) + nC CO2 ln(yC CO2 )] B B A A RT [ n CO2 ln(y CO2 ) + n A CO2 ln(y A CO2 )] A A

Minimum Work for Separa3on

Published in APS Report, Feasibility of DAC with Chemicals (2011)

Sherwood Plot for Flue Gas Scrubbing

CalculaWons carried out using IECM, all cases assume 500-MW plant burning Appalachian bituminous, NGCC (477-MW) O&M + annualized capital costs are included in the cost esWmates

1Cost and Scale

Process CO2-PCC CO2-NGCC SOx (MS) SOx (LS) NOx Hg Price [$/kg] 0.045 0.059 0.66 2.1 1.1 22000 Concentra3on [mole frac3on] 0.121 0.0373 0.00127 Emissions Cost [kg/day] [1000s $/day] 8.59 x 106 3.01 x 106 8.94 x 104 392 178 59.6 50.4 12.5 21.6

0.000399 (399 ppm) 2.32 x 104 0.000387 (387 ppm) 1.11 x 104 5 x 10-9 (ppb) 0.951

the recovery of potentially valuable solutes from dilute solution is dominated by the costs of processing large masses of unwanted materials.2 -Edwin Lightfoot

1These can change based upon coal-type burned and scrubbing methods; 2EN Lighdoot, MCM Cockrem, What Are Dilute

SoluWons, Sep. Sci. Technol., 22(2), 165 (1987)

2nd-Law Eciency Drops with Concentra3on

2nd
Wmin = Wreal

House, K.Z. et al., Proc. Nat. Acad. Sci., 108(51), 20428-20433 (2011)

How to Increase the 2nd-Law Eciency?

Current State-of-the-Art Technology: Taking a closer look at ABsorpWon via MEA as an example:

1. 2. 3. 4.

RegeneraWon Compression Blower/Fan Pumping

Improvements:
Regenera9on: Consider separaWon processes that do not involve solvents Compression: Consider separaWon processes that incorporate compression

Benets of Adsorp3on and Membrane Processes

Both processes are based primarily upon physical separaWon processes, with CO2 maintaining its linear form throughout separaWon Water does not need to be unnecessarily heated in either process; most solvents are aqueous-based w/ the chemical ~ 30 % Membrane Process: Major challenge w/ CO2-selecWve polymers: lack of driving force in ue gas w/ CO2 concentraWon ~ 12 % - consider N2-selecWve membrane instead Membranes have fairly small footprints and require no regeneraWon Adsorp9on Process: Mesoporous carbons are scalable and can be cost-eecWve Carbons have opWmal heat properWes Major challenge w/ MOFs and zeolites: water compeWWon and acid gases consider chemistries in which H2O assists in the capture mechanism, recall ue gas water concentraWon is ~ 10 %

Agenda

Work and cost of carbon capture N2-Selec3ve Membrane for Carbon Capture Carbon-based sorbents for carbon capture

N2-Selec3ve Membrane for Carbon Capture

Flux:
Q = permeability, L = membrane thickness, Feed Permeate Residue (retentate)

InspiraWon ARPA-E brainstorm session in 2010 Capture CO2 on the high-pressure side of the membrane may lead to cost savings in terms of compression energy System/OpWmizaWon will be crucial, but lets see if its possible rst
PhD students: Ni Rochana, Ekin Ozdogan, Kyoungjin Lee

Mo3va3on for N2-Selec3ve Membrane

Poten3al Applica3ons: Carbon capture Ammonia synthesis Methane/N2 mixtures Air separaWon (selecWve O2) (IGCC, oxy-combusWon)
N2
Step 1 Adsorp3on Step 2 Dissocia3on

Membrane

N2

N N

Step 3 Bulk Diusion

N H2

N N

NH3

Goals: Use theoreWcal modeling to provide insight into tuning the electronic structure of materials for enhanced nitrogen reacWvity Benchmark DFT predicWons with UHV experiments on single-crystal surfaces Perform permeaWon tests on the Group V materials

N2 Dissocia3on is Dicult!
Bond dissociaWon energies N2 ~ 225 kcal/mol; 944 kJ/mol; 9.7eV O2 ~ 119 kcal/mol; 498 kJ/mol; 5.1 eV H2 ~ 104 kcal/mol; 435 kJ/mol; 4.4eV Common N2 dissociaWon catalysts (H-B, ammonia synthesis) Fe, Ru d-band center model (Hammer and Nrskov) provides insight

Density of States
The density of states (DOS) of a system describes the number of states at each energy level that are available to be occupied. Fermi level

occupied

unoccupied

TransiWon metal reacWvity is disWnguished by its d-states, with each transiWon metal having a characterisWc d-band center

d-band Center Model

When bonding and anW-bonding states are formed, bond strength depends on the relaWve occupancy of states Bonding states lled strong bonds; anW-bonding states lled weakening d-band center increases from R to L of periodic table (transiWon metals) both bonding and anW-bonding states are higher from R to L Strength of adsorbate-metal bond increases Why use Fe and Ru for ammonia synthesis? Why not Group V? answer volcano

Hammer and Nrskov, Nature 376 238 (1995); Hammer and Nrskov, Adv. Catal. 45 71-129 (2000)

N and O Diusivity in Vanadium

Permeability = Diusivity Solubility

1Keinonen et al. Appl. Phys. A 34, 39 (1984); 2Nakajima et al. Philosophical Magazine A 67, 557 (1993). 3Holleck, J. Phys.

Chem. 74, 503 (1970); 4 Fukai and Sugimoto, Adv. In Phys. 34, 263 (1985)

Scope of Work
1. Surface ac3vity N2 adsorpWon mechanism N2 dissociaWon pathway Comparison to other typical ammonia synthesis catalysts 2. Solubility and Diusivity Atomic N binding mechanism Comparison to atomic H binding 3. Eect of alloying Ru Eect on binding ImplicaWons for permeability

Computa3onal Methodology

VASP (Vienna ab iniWo SimulaWon Package) Density funcWonal theory (DFT) Projector-augmented wave (PAW) potenWal GGA PBE
Bulk vanadium Lattice constant [] 2.98 2.93-2.941 3.0212 3.0243

This study Previous calculation

Experiment

1Mehl and Papaconstantopoulos, Phys. Rev. B 54, 4519 (1996); 2Vitos et al., J., Surf. Sci. 411, 186 (1998); 3Online CRC

Handbook of Chemistry and Physics, 91st ediWon, 2010-2011

Molecular N2 Adsorp3on Energy

V(110) Eads (eV/molecule) = E(surf+N2) [E(surf)+E(N2)] n(N2)

1-top 2-short-bridge (SB) 3-long-bridge (LB) 4-three-fold (TF)

V(111)

strength of N2-metal bond increases

1-top, 2-hcp 3-fcc, 4-bridge
1Grunze, et al., Appl. Phys. A 44, 19 (1987); 2 Bozso, et al. J. Catal. 49, 18 (1977); Ertl et al., Surf. Sci. 114, 515 (1982); 3Shevy

et al., J. Phys. Chem. C 112, 17768 (2008)

Eect of Ru Addi3on
Pure Vanadium Distance (N-Ru)= 0.5 Distance (N-Ru)= 0.71

-0.09 -0.254 +2.836 -0.255 -0.141 -0.257 -0.255

-0.374 -0.292 +2.710 -0.292 -0.292 -0.292 Ru +3.347

-0.372

-0.235

+3.075 -0.214 -0.174 Ru +3.075 -0.372

Lattice Constant= 3.01 Eb= -2.132 eV Lattice Expansion= 1.01%

Lattice Constant= 3.02 Eb= -0.889 eV Lattice Expansion= 1.34%

Lattice Constant= 3.01 Eb= -1.48 eV Lattice Expansion= 1.01%

H binding in V: O-site = -0.076eV; T-site = -0.280eV

Aboud and Wilcox, J. Phys. Chem. C, 114(24) 10978-10985 (2010); Pauling-Scale Electronegativities: N = 3.04; V = 1.63; Ru = 2.2

Flux Measurements
Test Temperatures: 500C -1000C
Retentate Sweep Gas Feed Gas

Test Temperatures: 20 90 psi

Inside of Membrane Holder

Permeate

(Group V metals) (uniformly rigidized sheet of

Membrane Foils

Porous Support (Hastelloy X) Diffusion Barrier

alumina ber and binder)

Membrane Defect Correc3ons

Flux measurements: Argon gas used to correct for pinhole and general leaks in the membrane system Each pure foil is tested at a temperature range of 500C-1000C. At each temperature, feed pressure is changed between 23.4-93.4 psig. Retentate Pressure is kept at 3.4psig Use Knudsen diusion for correcWons:

Gas Mixtures
Niobium (P=90 psi)
Natural gas ue gas
3.00E-04 2.50E-04 N2 Flux (mole/ms) 2.00E-04 1.50E-04 1.00E-04 5.00E-05 0.00E+00 0.001 0.0011 0.0012 0.0013 1/T (K-1) 1.00E-06 5.00E-07 0.00E+00 0.0014

Coal ue gas
2.50E-06 CO2 Flux ((mole/ms) 2.00E-06 1.50E-06 2.50E-04 2.00E-04 1.50E-04 1.00E-04 5.00E-05

4 mol% CO2-96 mol% N2

N2 CO2

15 mol% CO2-85 mol % N2

N2 CO2

8.00E-06 7.00E-06 6.00E-06 5.00E-06 4.00E-06 3.00E-06 2.00E-06 1.00E-06 CO2 Flux ((mole/ms)

N2 Flux (mole/ms)

0.00E+00 0.001

0.0011

0.0012 1/T (K-1)

0.0013

0.00E+00 0.0014

Next Steps
ConWnue DFT calculaWons to predict alloys for enhanced N2 separaWon InvesWgate subsurface and bulk diusion predicWons of various alloys Surface study experiments at SSRL to benchmark DFT Repeat experiments with N2 ux in pure foils and invesWgate the potenWal of ammonia synthesis with H2 as a sweep gas Work with SwRI to spuver deposit alloys of VRu and NbRu on porous stainless steel supports (Mov) Measure N2 and CO2 uxes of alloys and compare to pure

Agenda
Work and cost of carbon capture N2-selecWve membrane for carbon capture Carbon-Based Sorbents for Carbon Capture

Finding Ways Not to Bend CO2

Also cover design of Carbon Capture, Springer (2012) ISBN 978-1-4614-2214-3

How to Increase the 2nd-Law Eciency?

Taking a closer look at ABsorpWon as an example:
Slow kineWcs Highly exothermic

Envision a separa3on process that does not involve bending CO2

Closer Look at Heat Proper3es

Assume: Heat of regeneraWon = CpT + H hea9ng up all material in system from T1 to T2 + breaking the CO2 interac9on

CCS Applica3ons of Carbon Materials

SorpWon mechanisms in carbon-based sorbents and nanoporous natural systems for sequestraWon are similar Cost-eecWve (carbon) and scalable (chemistry) sorbents Maximize sorbent capacity by surface chemistry Appreciate the importance of transport kineWcs, e.g., 500-MW power plant emits ~ 11,000 tons of CO2 per day KineWcs: one of the main dierences bet/ PCC and DAC

PhD Students: Yangyang Liu, Abby Kirchoer, and Jiajun He

Research Associates: Mahnaz Firouzi and Erik Rupp

Molecular Simula3on
1. Shale characterizaWon (XPS, SEM, Quantachrome, FTIR, etc.) for building accurate pore models 2. Electronic structure theory - decorate pore surfaces with accurate chemistry, i.e., clay, carbon, dissociated water, defect sites 3. Grand Canonical Monte Carlo predict adsorpWon isotherms and compare to experiment; How do uid densiWes change at the nanoscale? 4. Molecular Dynamics predict transport properWes, e.g., permeability How does viscosity change at the nanoscale? Permeability?

Molecular Simula3on
1. Shale characterizaWon (XPS, SEM, Quantachrome, FTIR, etc.) for building accurate pore models 2. Electronic structure theory - decorate pore surfaces with accurate chemistry, i.e., clay, carbon, dissociated water, defect sites 3. Grand Canonical Monte Carlo predict adsorpWon isotherms and compare to experiment; How do uid densiWes change at the nanoscale? 4. Molecular Dynamics predict transport properWes, e.g., permeability How does viscosity change at the nanoscale? Permeability?

Poten3al Models (L-J and TraPPE)

SchemaWc plot of one-center Lennard-Jones potenWal model of CO2 in slit-pore

CO2
2

= 3.75 A = 240.00K

= 3.40 A C
k = 28.00K

CO2 C
2

= 3.571 A = 81.9756K

CO
k

CO C
k

Dening Adsorp3on
Total Adsorp3on Direct results from GCMC Modeling Excess Adsorp3on Direct results from Lab Measurements Convert from Total to Excess AdsorpWon Total Adsorbed Bulk = Excess

Adsorp3on Isotherm Predic3on Based on PSD

Measured PSD predict adsorp3on isotherm Original PSD of AC sample Assume the total isotherm consists of a number of individual single pore isotherms mulWplied by their relaWve distribuWon over a range of pore sizes. The set of isotherms for a given system can be obtained by GCMC simulaWons.

PSD truncated at 20 nm

T = 305 K

Eect of surface func3onali3es

Perfect graphite: the basic slit-pore surface Chemical heterogeneity: the possible funcWonal groups1 and the mono vacancy site in the environment of volaWle components environment (e.g., water2) have been invesWgated
epoxy epoxy2 hydroxyl H2O dissociate on the mono-vacancy

carbonyl

carbonyl - hydroxyl

hydroxyl - carbonyl

pore width

1Bagri, A. et al. J. Phys. Chem. C 2010; Kudin, K. N. et al. Nano LeW. 2008. 2Kostov, M.K. et al. Phys. Rev. LeW. 2005.

Eect of surface func3onali3es

Electronic properWes and parWal charge distribuWons by Density FuncWonal Theory (DFT)

parWal charge distribuWon

30
Total Loading [mmol/cm3]

30 Total Loading [mmol/cm3] 25 20 15 10 5 0 0 50 100 150 200 Pressure [bar] @ 298 K (pore width = 9.2 ) 250 0 50 100 150 200 Pressure [bar] @ 298 K (pore width = 20 ) 250
Perfect graphite Epoxy functionalized Hydroxyl functionalized Carbonyl functionalized Carbonyl_Hydroxyl functinoalized Hydroxyl_Carbonyl functionalized Carboxyl functionalized Hydrated graphite

25 20 15 10 5 0
Perfect graphite Epoxy functionalized Hydroxyl functionalized Carbonyl functionalized Carbonyl_Hydroxyl functinoalized Hydroxyl_Carbonyl functionalized Carboxyl functionalized Hydrated graphite

In general, oxygen-containing funcWonal groups increase the adsorbed CO2 density in micropores, especially in the cases of hydroxyl and carbonyl-funcWonalized slit pores;

Eect of surface func3onali3es

Perfect graphite slit-pore

Eect of surface func3onali3es

Local CO2 density distribuWon
epoxy funcWonalized Perfect graphite slit pore hydroxyl funcWonalized

carbonyl funcWonalized

carbonyl_hydroxyl funcWonalized

hydroxyl_carbonyl funcWonalized

carboxyl funcWonalized

1. Local density distribuWon is not homogeneous in the slit pores; 2. adsorbed layer has high density (> dry ice) higher packing eciency

hydrated graphite slit pore

Eect of surface func3onali3es

Compared to MOFs:
At 1.0 bar, the loading is up to ~12 mmol/g, compared to current state-of-the-art MOFs that range between 2~12 mmol/g at similar T and P The pore size and surface funcWonality of GCMC simulaWons is easily tunable to control adsorpWon

Supercritical CO2 @ 298K 250 bar

Solid CO2 (dry ice)*

Adsorbed CO2 in COOH @ 298K 1 bar

*CO2 Crystal structure data from AMCSD

Next Steps
ConWnue GCMC to provide insight into opWmal funcWonality for enhanced adsorpWon Surface funcWonalized (Zn-based) sorbents to catalyze the bending of CO2 using controlled mesoporous carbons (collaboraWon w/ Bao and Stack) ConWnue PSD experiments and benchmarking w/ Quantachrome and GCMC Carry out adsorpWon experiments with Rubotherm microbalance InvesWgate adsorpWon and breakthrough experiments using real ue gas condiWons, i.e., water vapor, NOx, and SO2

triblock copolymers to template ordered mesoporous silica

Sorp3on and Transport at the Nanoscale Coal and Shale

Shale consists of organic (kerogen) and clay components with porosity on the nanoscale Molecular simulaWon can determine the mechanisms of sorpWon and transport of uids (CO2, methane, water) under nanoconnement Fluid properWes of interest may include: Density Viscosity Surface tension We hypothesize that uid properWes are dierent at the nanoscale vs macroscale due to: Pore size Pore chemistry (clay vs carbon vs surface funcWonality, e.g., dissociated water) Improvements in understanding sorpWon and transport may inuence capacity esWmates and recovery, respecWvely

Research Outline
Coal / Gas Shale / Carbon-based Sorbents Micro and Mesopores with Structural and Chemical Heterogeneity Slit Pores with FuncWonalized GraphiWc Surfaces

Characteriza3on Quantachrome, FTIR, XPS, SEM, etc. Chemical ComposiWon, Carbon- and FuncWonal Groups, Clay-based PSD, etc. Materials AdsorpWon/Transport Measurements

Modeling Electronic Structure, GCMC, and MD

AdsorpWon Isotherms Capacity EsWmates SorpWon Energy Permeability Compare Transport Mechanism Slippage Factors AdsorpWon Isotherms, PermeabiliWes

CO2 sta9onary source in the U.S.

Sources and Sinks

Unmined coalbeds (ECBM) Oil and Gas Reservoirs Deep Saline Aquifers Gas Shale Reservoirs (EGR)

Overlap between sources and ECBM/EGR eorts

Making the Connec3on between Length Scales

Molecular Simula3on
1. Shale characterizaWon (XPS, SEM, Quantachrome, FTIR, etc.) for building accurate pore models 2. Electronic structure theory - decorate pore surfaces with accurate chemistry, i.e., clay, carbon, dissociated water, defect sites 3. Grand Canonical Monte Carlo predict adsorpWon isotherms and compare to experiment; How do uid densiWes change at the nanoscale? 4. Molecular Dynamics predict transport properWes, e.g., permeability How does viscosity change at the nanoscale? Permeability?

Understanding Transport in Micropores

Micropore < 2nm
Transport of equimolar binary mixture of CH4 and CO2 has been modeled using NEMD simulaWons in a slit pore model The pore wall is assumed smooth and the interacWon between molecules and pore wall was modeled by the Steele and uid-uid by the LJ potenWals Verlet algorithm was used to solve the equaWons of moWon

Upstream Pressure

Transport

Downstream

Pressure

Length = 15.2 nm [152 ] Width = micro to mesopore range

Upstream pressure = 3 atm, Downstream pressure = 1 atm, Temperature = 298 K

CH4/CO2 Velocity Proles in Micro and Mesopores

In small pores the velocity prole is plug ow and becoming parabolic at approximately 4 nm pores for CH4 and greater than 10 nm pores for CO2

Upstream Pressure

Transport

Downstream Pressure

Height = 1.1 nm
1.5 1.0 0.5 Z* 0.0 -0.5 -1.0 -1.5 0 0.3 0.6 0.9 1.2 1.5 Velocity 5 3 1 -1 -3

Height = 3.8 nm
CH4 CO2

Height = 7.6 nm
10 6 2 -2 -6

CH4 CO2

-5 -0.3 0 0.3 0.6 0.9 1.2 1.5 Velocity

-10 -0.3 0 0.3 0.6 0.9 1.2 1.5 Velocity

Pure CH4, CO2 Velocity Proles in Mesopores

As pore sizes increase to 10 nm reduced wall interactions take place in the center of the pore
Height = 10 nm Upstream Pressure Transport Downstream Pressure

60 40
Z ()

60 40 20 0 -20 -40 -60 0 20 40 60 80 100 0 20 40 60 80 100 Velocity x 10-5 (cm/sec) Velocity x 10-5 (cm/sec)

60 40 20 0 -20 -40 -60 0 20 40 60 80 100 Velocity x 10-5 (cm/sec)

20 0 -20 -40 -60

CH4, CO2 Permeability Versus P

Gas permeability is enhanced for both components in micropores The permeability of CO2 is larger than CH4 due to the anity of CO2 for carbon

surfaces and lager density of CO2 in the pore and shielding eects

Permeability ((grmole.cm)/(min.cm2.atm))x10- 3

10 8 6 4 2 0 0 2 4 6 8 10 12 14 P (atm)

1.0 0.8 0.6 0.4 0.2 0.0 0 2 4 6 8 P (atm)

CH4-76.2 CO2-76.2

Height =11

Height =76

10

12

14

M. Firouzi et al., Chemical Engineering Science 62, 2777 (2007)

CH4, CO2 Slippage Factor

The slippage factor and k for CO2 is larger than CH4 as expected As the pore becomes smaller the slippage factor and k becomes larger due to the eect of the walls

kg= k(1+b/pm)
Permeability ((grmole.cm)/(min.cm2.atm))x10- 3 10 8 6 4 2 0 0.0 0.1

Height =11
0.6
CH4: K = 0.84, b = 2.1
CO2: K = 1.74, b = 4.3

Height =76
CH4: CO2: K = 0.19, b = 0.29 K = 0.23, b = 0.84
y = 0.1969x + 0.2348

CH4-76.2 CO2-76.2

0.5 0.4

y = 7.523x + 1.7447

0.3
y = 0.0552x + 0.1922

0.2
y = 1.7728x + 0.8412

0.1 0.0 0.6 0.0

0.2

0.3

0.4

0.5

0.1

Inverse mean pressure (atm-1)

0.2 0.3 0.4 0.5 0.6 Inverse mean pressure (atm-1) 50

Pure CH4, CO2 Slippage Factor

The viscosity near solid surfaces are spaWally varying and needs to be calculated using molecular dynamics:
kg= k(1+b/pm) Slit Pore: Height =11
30 Permeability (nd) 25 Permeability (md) 20 15 10 5 0 0.0 0.4 0.8 1.2 1.6 -1) Inverse mean pressure (atm 2.0
y = 19.104x + 3.8256 R = 0.95475 y = 34.924x + 5.0398 R = 0.98012

q/A=k/ (-dp/dL) Pore network: Average pore size = 2 nm, Porosity = 20%
CH4 CO2
y = 4840.6x + 85.585 R = 0.98903

CH4 CO2

500 400 300

y = 3950.5x + 84.295 R = 0.83745

200
CH4: K = 3.8, b = 5.0 CO2: K = 5.0, b = 6.9

100 0 0.02

CH4: K = 84, b = 47 CO2: K = 85, b = 56

0.03

0.04

0.05

0.06

0.07

Inverse mean pressure (atm-1)

Temperature was held at 298 K. The time step for integrating the equation of motion was 1 fs. Simulations of at least 50 ps long were needed to establish the ow of water under a pressure gradient and to reach steady state, after which the rst 50 000 time steps were ignored and then the numerical data were collected over a minimum time of 25 ps. Equal numbers of Si and C atoms were, of course, used. Water viscosity in the nanotubes was estimated from the Einstein relation,

Water Viscosity in Nanotubes

kB T The viscosity increases with increasing (9) , = 3 dDz nanotube sT is the temperature,alue Boltzmanns constant, dtthe ize and its v kB the is lower than hat where diameter of water, and D under bulk condiWon z its axial diffusivity that is estimated using the Green-Kubo relation: 1 Dz = N
N

Thus, the velocity ACF may exhibit one much damped oscillation (one major minimum) before decaying rapidly to zero, a description that is consistent with Fig. 2. Moreover, it has been illustrated that damping in MD calculation does not affect the dynamic properties of a system to within the statistical uncertainty.58 Fast dissipating behavior, similar to what we report, has also been commonly observed and used to calculate the properties of water in the CNTs. Figure 3 presents the computed viscosity of water and its dependence on the nanotubes diameter. For comparison,

v (t) vi (0) i , (10) The noWon of viscosity, ais used dtn classical 0 i=1 conWnuum vmechanics, may moleculee truly and N not b i at time t, where i (t) is the axial velocity of is the applicable total number of molecules. The bracketed quantity rep-

The computed values should be considered as apparent viscosity, as computed by the FIG. 3. Dependence of water viscosity on the diameter of SiC nanotubes. Einsteins and Green-Kubo relaWons For comparison the viscosity of water in CNTs (Ref. 47) is also given.

resents the velocity autocorrelation function (ACF). In small nanotubes the diffusivity Dz is dependent upon the tubes diameter. We did not assume that the atomistic structure of the nanotubes is rigid, so that the C and Si atoms could move in response to their environment in the presence of the water molecules and the applied pressure gradient according to the equation of motion.

Dependence of water viscosity modeled using MD on the diameter Downloaded 03 May 2012 to 171.67.34.69. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/about/rights_and_permissions of CNTs and SiC nanotubes

M. Khademi, S. Sahimi, The Journal of Chemical Physics 135, 204509 (2011)

3-D Pore Network Model

The dimensions of the system modeled are ~ 10 x 10 x 10 nm 3-D molecular pore network model based on the Voronoi tessellaWon method To generate the molecular pore network model: - Create a 3-D simulaWon box of structural atoms corresponding to porous structure - Tessellate the atomic structural box The pore space is created by specifying the desired porosity and # polyhedra total volume fracWon = specied porosity - pore space consists of interconnected pores of various shapes and sizes

Modeling Transport with MD

The pore network model previously described will be used

Non-equilibrium molecular dynamics (NEMD) simulaWons are carried out The system (pore network) is exposed to an external driving force (chemical
potenWal or pressure gradient) in a specied direcWon

Flux and permeability predicWons are carried out

Permeability of N2/CO2 and CH4/CO2 Mixtures

8

Permeability ((grmole.cm)/(min.cm 2.atm))x10- 7

7
6 5

4
3 2 1 0

0.88 N2 0.12 CO2 0.75 N2 0.25 CO2 0.5 N2 0.5 CO2 0.25 N2 0.75 CO2

9 8 7 6 5 4 3 2 1 0 10 15

0.75 CH4 0.25 CO2 0.50 CH4 0.50 CO2 0.25 CH4 0.75 CO2

10

15

20

25
Porosity

30

35

40

20

25 Porosity

30

35

40

Permeability of N2 / CO2 (lec) and CH4 / CO2 (right) mixtures with average pore diameter of 1.2 nm and 20%, 25%, 30% and 35% porosi9es With mixtures of N2, at high CO2 concentraWons, permeability is lower below a 30% porosity With mixtures of N2, 25% CO2 has the greatest permeability In gas mixtures of N2 and CH4, CO2 is always the more permeable species in 1.2 nm pores

Acknowledgements
CollaboraWons with Mark Zoback and Tony Kovsceck (shale research) CollaboraWons with Dan Stack and Zhenan Bao (sorbent research) Funding: Membrane: NSF Eager, Catalysis Division; EPA P3 (high-T furnace); Army Research Oce Sorp3on: BP; DOE-NETL; GCEP GCMC; MD: Stanford Center for ComputaWonal Earth & Environmental Science DFT: NSF Teragrid, UT AusWn

Ques3ons?