
Full Citation 
Material Information 

Title: 
17.6.4  Continuum simulations of CO2 capture by dry regenerable Potassium based sorbents Reactive Multiphase Flows 

Series Title: 
7th International Conference on Multiphase Flow  ICMF 2010 Proceedings 

Physical Description: 
Conference Papers 

Creator: 
Garg, R. Shahnam, M. Huckaby, E.D. 

Publisher: 
International Conference on Multiphase Flow (ICMF) 

Publication Date: 
June 4, 2010 
Subjects 

Subject: 
multiphase reactive flows EMMS CO2 capture dry sorbents 
Notes 

Abstract: 
Concerns over increased atmospheric CO2 concentration has led to the investigation of technologies which can
provide cost effective options to reduce CO2 emissions while meeting global energy demand. Due to the large supply
of coal and coalbased power production capacity, systems and/or modifications to existing systems are needed to
extract energy from coal while allowing for CO2 capture and sequestration. These technologies include: oxyfuel
combustion, chemical looping combustion, precombustion CO2 separation (gassification), postcombustion CO2
separation (NETL 2009), membranes and lowerenergy solvent processes (see Wall (2007) for a general discussion
of the emerging technologies). Current commercially available systems for postcombustion separation are based on
amine liquid solvent systems which have a significant energy penalty. Solidsorbent based systems (NETL 2009) offer
the possibility of CO2 separation with lower energy costs (Yi et al. 2007). In this work we perform CFD simulations
of CO2 capture by dry regenerable solidbased sorbents (particularly Potassium based) using the opensource CFD
solver MFIX (Syamlal et al. 1993). The objective of this numerical study is to compare with the continuous CO2
capture and regeneration experiments of Yi et al. (2007). Additionally several modeling and numerical challenges for
high fidelity CFD simulations of such systems are identified. 

General Note: 
The International Conference on Multiphase Flow (ICMF) first was held in Tsukuba, Japan in 1991 and the second ICMF took place in Kyoto, Japan in 1995. During this conference, it was decided to establish an International Governing Board which oversees the major aspects of the conference and makes decisions about future conference locations. Due to the great importance of the field, it was furthermore decided to hold the conference every three years successively in Asia including Australia, Europe including Africa, Russia and the Near East and America. Hence, ICMF 1998 was held in Lyon, France, ICMF 2001 in New Orleans, USA, ICMF 2004 in Yokohama, Japan, and ICMF 2007 in Leipzig, Germany. ICMF2010 is devoted to all aspects of Multiphase Flow. Researchers from all over the world gathered in order to introduce their recent advances in the field and thereby promote the exchange of new ideas, results and techniques. The conference is a key event in Multiphase Flow and supports the advancement of science in this very important field. The major research topics relevant for the conference are as follows: BioFluid Dynamics; Boiling; Bubbly Flows; Cavitation; Colloidal and Suspension Dynamics; Collision, Agglomeration and Breakup; Computational Techniques for Multiphase Flows; Droplet Flows; Environmental and Geophysical Flows; Experimental Methods for Multiphase Flows; Fluidized and Circulating Fluidized Beds; Fluid Structure Interactions; Granular Media; Industrial Applications; Instabilities; Interfacial Flows; Micro and NanoScale Multiphase Flows; Microgravity in TwoPhase Flow; Multiphase Flows with Heat and Mass Transfer; NonNewtonian Multiphase Flows; ParticleLaden Flows; Particle, Bubble and Drop Dynamics; Reactive Multiphase Flows 
Record Information 

Bibliographic ID: 
UF00102023 

Volume ID: 
VID00432 

Source Institution: 
University of Florida 

Holding Location: 
University of Florida 

Rights Management: 
All rights reserved by the source institution and holding location. 

Resource Identifier: 
1764GargICMF2010.pdf 

Full Text 
7th International Conference on Multiphase Flow,
ICMF 2010, Tampa, FL, May 30 June 4, 2010
Continuum simulations of CO2 capture by dry regenerable Potassium based
sorbents
R. Garg *, M. Shahnam and E. D. Huckaby *
Computational Science Division, National Energy Technology Laboratory, Morgantown, WV 26507, USA
Rahul.Garg@gmail.com, Mehrdad.Shahnam@netl.doe.gov and E.David.Huckaby@netl.doe.gov
Keywords: multiphase reactive flows, EMMS, CO2 capture, dry sorbents
Abstract
Concerns over increased atmospheric CO2 concentration has led to the investigation of technologies which can
provide cost effective options to reduce CO2 emissions while meeting global energy demand. Due to the large supply
of coal and coalbased power production capacity, systems and/or modifications to existing systems are needed to
extract energy from coal while allowing for CO2 capture and sequestration. These technologies include: oxyfuel
combustion, chemical looping combustion, precombustion CO2 separation (gassification), postcombustion CO2
separation (NETL 2009), membranes and lowerenergy solvent processes (see Wall (2007) for a general discussion
of the emerging technologies). Current commercially available systems for postcombustion separation are based on
amine liquid solvent systems which have a significant energy penalty. Solidsorbent based systems (NETL 2009) offer
the possibility of CO2 separation with lower energy costs (Yi et al. 2007). In this work we perform CFD simulations
of CO2 capture by dry regenerable solidbased sorbents (particularly Potassium based) using the opensource CFD
solver MFIX (Syamlal et al. 1993). The objective of this numerical study is to compare with the continuous CO2
capture and regeneration experiments of Yi et al. (2007). Additionally several modeling and numerical challenges for
high fidelity CFD simulations of such systems are identified.
1 Introduction and Background
The U.S. Department of Energy (DOE) has embarked on
a Carbon capture and sequestration (CCS) initiative that
by 2020 aims to develop energy systems which capture
greater than I' of the produced CO2 with less than a
:.'. increase in the cost of energy. Various CO2 capture
technologies are being developed and they can broadly
be classified as oxyfuel combustion, chemical looping
combustion, precombustion CO2 separation (gassifica
tion), postcombustion CO2 separation (NETL 2009).
These technologies differ, among other process details,
on what stage of the fuel oxidation process the CO2is
captured. Interested readers are referred to Wall (2007)
for a more comprehensive discussion of these emerging
technologies.
Amine liquid solvent systems are the only proven
commercially available technology for postcombustion
CO2 capture. These systems, however, introduce a se
vere energy efficiency penalty (Rao and Rubin 2002).
This high energy penalty has motivated the development
of CO2 capture technology using dry regenerable solid
sorbents.
A team of researchers at the Korean Institute of En
ergy Research (KIER) and Korean Electric Power Re
search Institute (KEPRI) developed a dry regenerable
potassium carbonate, which the group has named Sorb
KX35 (Ryu et al. 2005). The feasibility of using this sor
bent was later demonstrated in a benchscale fluidized
bed reactor system (Yi et al. 2007) in which a CO2 re
moval capacity of 507 .'. was measured over the 20
hour continuous operation of the system. In the first or
carbonation step of this process (shown by the schematic
in Fig. 1), synthetic flue gas (containing N2, CO2, and
steam) is mixed with dry sorbent to absorb CO2 in a
fast fluidizedbed reactor. In the second or regeneration
step, the spent sorbent from carbonator is sent through
a bubbling fluidized bed in order to removed CO2 and
regenerate the sorbent. The regenerated sorbent is re
turned to the carbonator resulting in a continuous CO2
removal. In this study, we perform the EulerianEulerian
simulations of the operation of the carbonator using the
open source MFIX (Syamlal et al. 1993) CFD code for
reactive multiphase flows, equations for the mass and
momentum conservation in both this work our primary
objective is to identify the modeling
7th International Conference on Multiphase Flow,
ICMF 2010, Tampa, FL, May 30 June 4, 2010
CO2 lean flue gas
t
I I 
A.CO,(s)
(COz &H O)
n
I:JHeat
Flue Gas (CO, N,, H0) Fuidizing Gas (COj1H0)
Figure 1: A schematic of the CO2 capture unit. The
flue gas is fluidized with solid sorbent (rep
resented by A(s)) in carbonator (left reactor)
resulting in exothermic adsorption of CO2
from flue gas by the solid sorbent. The CO2
rich solid sorbent is then fluidized with regen
eration gas in the regenerator (right reactor)
along with external heat in order to produce
pure CO2, steam and CO2 free solid sorbent.
This regenerated solid sorbent is circulated
between the carbonator.
1.1 KIER Experiments Review
A detailed description of the KIER CO2 capture unit
(carbonator and regenerator) can be found in Yi et al.
(2007). In this study we focus on modeling the carbon
ator. The schematic of the carbonator used in KIER ex
periments is shown in Fig. 2. This fastfluidized reactor
is made up of a lower mixing zone (60 cm tall and 3.5 cm
internal diameter) which tapers off to an upper riser zone
(540 cm tall and 2.5 cm internal diameter). In the CFD
simulations described later the carbonator geometry is
simplified. Among other things, the smooth transition
between the lower mixing zone and riser is replaced by
an abrupt change in the diameter. The labels DP1, DP2,
DP3, and DP4 in the schematic (Fig. 2) reference differ
ential pressure measurement sections. Likewise, labels
Tl, T2, T3, and T4 correspond to the measurement po
sitions of the reactor temperature.
The KX35 sorbent has a bulk density of 1.1 g/cm3
and a mean particle diameter of 98 pm. The sorbent of
:.\\ K2CO3 as the active component and the remain
der I..\\t' is activated carbon, which is provided for
mechanical strength and pore structure. The reported
solids circulation rate ranged from 8 to 35 kg/m2 s corre
sponding to approximately 5 to 20 g/s to corresponding
based on the riser diameter (2.5 cm). Similarly, the su
perficial gas velocity ranged from 1.6 to 3.0 m/s, corre
sponding to gas mass flow rate of 0.8 to 1.4 g/s based on
Riser
2.5 cm I.D.
Figure 2: Schematic (not drawn to scale) of carbonator
used in KIER experiments. Differential pres
sure measurements were reported across sec
tions labeled as DP1, DP2, DP3, and DP4.
Bed temperature measurements were reported
at points labeled as T1, T2, T3, and T4.
the riser diameter.
Governing Equations
The MFIX CFD code used to simulate the system
is based on the discretization of multifluid conserva
tion equations (based on the continuum hypotheses) for
mass, momentum, and energy of multicomponent gas
and solids phases. Depending on treatment of the gas
pressure field, two different formulations have been pro
posed which are commonly referred to as Model A and
Model B (Gidaspow 1994). In Model A, the gas pres
sure field acts on both the gas and solids phases, while
in model B, the gas pressure fields acts solely on the gas
phase. Conservation equations corresponding to model
A are presented assuming a single solids phase. Using
the suffix notation, the mass and momentum conserva
tion equations in the gas phase are
a(gag)
a gpg) ( agpgUg;) i= .
a~t 0xzi
m
I Heat I
Heat
(EgpgUgi) + 9x (EgpgUgjUgi)
at OX!
ap,
g 9x
aT
+ ai+f,., +/3(Usi Ugi), (2)
Oxi
where pg is the thermodynamic density of the gas phase
which is calculated using the ideal gas law, Pg and Ugi
are gasphase pressure and velocity fields, respectively,
and ; is the net rate of gas production (production
consumption) due to chemical reactions. In the momen
tum conservation, Tgij is the gasphase shear stress ten
sor which is assumed to be Newtonian and is given by
Tgij ftg (aUgi augj
2 9Ug
3 OTk
where 6ij is the Kronecker delta, pg is the fluid dyna
coefficient of viscosity, and the bulk coefficient of
cosity has been assumed to be 2/3 times dynamic co
cient of viscosity. Furthermore, 3 in the momentum c
servation equation is the interphase momentum tran
coefficient. Closure of this term is of particular iml
tance in this study and will be addressed in further de
later in the section.
Similar to the gasphase, the mass and moment
conservation equations for the solid phase are
0a
t (EsPs) ax (EspsUsi) = rm;
8 8x '
a a
a 5p5U5i)+ x E. a. [T T
a p, aP
s axi ax
(7xT 8.,
+ + spsg (Ui Ugi), (5)
where ps is the particle density, Ps is the solidphase
pressure, and ihs is the net rate of production of solid
phase that accounts for interphase mass transfer due to
chemical reactions. Additionally, Tsij is the shear stress
tensor in the solidphase which is given by
(9U + U ) + 2 9Usk,
Tsij = Ps + + I. s 6ij,
9xj 9x 3 ) 9xk
(6)
where ps is the dynamic coefficient of viscosity and Pb
is the modified bulk solid viscosity. Depending on the
volume fraction, the constitutive relations for Ps, Pb and
Ps are obtained either from one of the several kinetic the
ories (for Es < 0.5) or from one of the frictional theories
(Es > 0.5) that are available in MFIX. Since the consti
tutive relations for transport coefficients (such as ps) are
dependent on solid granular temperature, an additional
conservation equation for solid granular temperature is
mic
vis
effi
on
sfer
3or
7th International Conference on Multiphase Flow,
ICMF 2010, Tampa, FL, May 30 June 4, 2010
also solved. Interested readers are referred to the MFIX
theory manual (Syamlal et al. 1993) for additional de
tails.
For reactive multicomponent multiphase flows, each
phase is composed of several species which are each
tracked individually. The mass conservation equation
for the .' gasphase species is
a a n)
at) ( gpgUg
+ Rgn (7)
where Xn is the mass fraction of .' species, Rgn is
the corresponding net rate of production, and Dgn is the
diffusion coefficient. Similar species conservation equa
tions are solved for the solidphase.
For nonisothermal flows an energy equations is
solved for each phase. The energy conservation equa
tion discretized by MFIX for gasphase is
g pg [arg gai rTg aqgi
EgpgC t U xi x
'gm (Ts Tg) AHg, (8)
;tail where Tg and Ts are the temperature fields in the gas and
solid phases, respectively, AHg is the heat of reaction in
tum the gasphase, qg is fluidphase conductive heat flux. A
similar energy conservation is solved in the solidphase
except that the conductive heat flux term (first term on
) RHS of Eq. 8) is zero.
(4)
The interphase heattransfer coefficient in the absence
of mass transfer, ". is calculated from the Nusselt
number, Nu, as
SNu (9)
7gm D2 (9)
P
where kg is the thermal conductivity of the gasphase
and Dp is the particle diameter. In this study we use the
Nusselt number correlation proposed by Gunn (1978)
which is of the form
Nu (7 10s + 2) 1 + 0.7Re.2Pr3) +
+ (1.33 2.4g + 1.2E ) Reo.Pr1/3, (10)
where Pr = gCpg/kg is the Prandtl number, and Re =
gpgDp Ug Ug
gpgDp Ug g is the Reynolds number based on
Pg
particle diameter and mean slip between gas and solid
phases. This expression is valid for the gas voidage
range 0.35 < Eg < 1.0 and from Stokes flow limit up to
Re 105. The heattransfer coefficient 7gm corrected
for mass transfer is calculated using the following ex
pression (Bird et al. 1962)
gm (11)
exp Rm 1
a (Dg +
9xi O xi
where R, is a non dimensional ratio of mass transfer to
heat transfer R pggm that has been introduced
to simply the expression. Lastly, Cpg is the specific heat
capacity of the gasphase.
1.2 Interphase drag force
Typically in continuum simulations of gassolids flows
the closures for gassolids momentum exchange (or drag
force) are based on semiempirical correlations, such as
the well known Wen & Yu (Wen and Yu 1966) and Er
gun correlations (Ergun 1952), or obtained from detailed
numerical simulations of gassolids flows, such as the
Lattice Boltzmann simulations (Hill et al. 2001; van der
Hoef et al. 2005). These relations are based on a spa
tially homogeneous distribution of solids which is not
always observed in working systems.
In fluid catalytic cracking (FCC) particles it has been
observed in the fluidization literature (Arastoopour and
Gidaspow 1979; McKeen and Pugsley 2003) that use
of above closures for drag force results in large quan
titative discrepancies between simulations and experi
ments for quantities such as bed pressure drop or bed
expansion height. In Eulerian simulations of riser flows
by Qi et al. (2007) it was observed that the particles fed
into the riser are immediately transported out without
any recirculation near the walls. This discrepancy be
tween simulations and experiments for such small parti
cles is attributed to drag reduction due to cluster forma
tion which the above semiempirical correlations do not
account for. The recent drag closures from LBM simu
lations have been obtained for homogeneous assemblies
and, therefore, fail to account for strong heterogeneity
that is observed in such flows due to formation of local
clusters (Fortes et al. 1987; Liu et al. 2005). Several cor
rections to the standard drag correlations have been pro
posed. For example, Arastoopour and Gidaspow (1979)
and McKeen and Pugsley (2003) propose using an ef
fective cluster size (which is of the order of few par
ticle diameters) in the standard drag correlations in or
der to improve predictive capability of the simulations.
Recently, Li and Kwauk (2003) proposed the energy
minimization multiscale (EMMS) model to calculate
gassolids drag force for locally heterogeneous condi
tions.
The sorbent particles (diameter of 100 pm, 1.1 g/cm3)
used in this study are similar to FCC particles in size
and density and fall under the Geldart typeA classifi
cation. We explore two drag force correlations in this
study. First is the standard Gidaspow drag model which
is a blend of Wen & Yu and Ergun correlation. The sec
ond drag correlation is a curve fit calculated from a sim
plified form of the EMMS approach and accounts for
7th International Conference on Multiphase Flow,
ICMF 2010, Tampa, FL, May 30 June 4, 2010
clustering effects by multiplying the Wen & Yu drag cor
relation with a heterogeneity factor H(g, Re). Both the
drag correlations can be conveniently expressed by a sin
gle expression of the form
3 Pg1 SE~~g 75 g g Usl.65
CD E H(E,, Re),
1 gDp Dp
(12)
where CD is the single particle drag coefficient (Schiller
and Naumann 1933) given by
e (1 + 0.15Re0.687) if 0 < Re < 103 ,and
CD = Re 
0.44 ifRe > 103.
(13)
Reynolds number is defined based on mean slip veloc
gpgDp Ug Ug 
ity as Re gpgDp In the expression for
,3 (Eq. 12), the first expression is the Wen & Yu drag
correlation and the second expression is the Ergun drag
correlation. In the standard Gidaspow drag correlation,
heterogeneity factor H( g, Re) is unity. Also, the Wen
& Yu drag correlation is used for solids volume fraction
up to 0.2 (i.e., 0 < Es < 0.2), and the Ergun correlation
is used for >s > 0.2. However, in the EMMS correction
proposed by Lu et al. (2007), Wen & Yu drag correla
tion is used for solids volume fraction up to 0.6 (i.e.,
0 <
a narrow range from Es = 0.6 to closepacking. For the
heterogeneity factor H( g, Re) in the EMMS corrected
drag correlation, we use the correlations presented by
Lu et al. (2007). In the results section we show the com
parison of pressure drop obtained from the above two
drag correlations.
1.3 Reaction Model
The absorption of CO2 by potassium carbonate is ex
pected to occur by the following exothermic reaction
K2CO3(s)+CO2(g)+H20(g) * 2KHCO3(s)(+Heat)
(14)
For this initial modeling study of CO2 capture in flu
idized beds, a simple homogeneous reaction model sim
ilar to that proposed by Park et al. (2006) is used:
dCK2 CO3 K, psXK2CO3 (PgXco,\ (15)
dt \ WK2CO \) WC02 )
where Es is the solid volume fraction, pg and ps are
the densities of the gas and sorbent particles, respec
tively, W (g/mol) is the molecular weight, and CK2CO3
(mol/cm3) is the molar concentration of K2CO3. The
rate constant, k, was tuned to 500 cm3/s to approxi
mately match the outlet concentration for the baseline
operating condition of 15 g/s solids and 1 g/s gas.
This model obviously neglects several processes, but
as will be shown later, the model qualitatively captures
much of the operation of the reactor. Further, as will
be discussed in subsequent sections, the simulations us
ing a homogeneous drag model show a large discrepancy
with the reported pressure drop measurements. This dif
ference implies that the solids inventory in the reactor is
much different than the experiments. The solids inven
tory is a first order effect and is somewhat independent
of the other details in the reaction model. Therefore, we
focus on the drag model, prior to further refinements of
the reaction model.
1.4 Numerical Modeling
Both 2D and 3D simulations of the KIER CO2 absorp
tion reactor (Fig. 1) are performed. For 2D simulations
the reactor geometry is approximated by two rectangu
lar channels of equal depth (Fig. 3a). For the lower mix
ing zone, the width, Dmix is 3.5 cm, the same as the
diameter of the KIER unit. The depth of the reactor
(Lz z 2.75 cm) is specified such that the mixing zone
has the same cross sectional area as the experimental re
actor. Similarly, the width (xextent) of the upper riser
section is Dis 1.78 cm, which is calculated such that
the crosssectional area of the 2D reactor is same as the
2.5 cm diameter experimental riser.
For the 3D simulations, the geometry is approximated
by two channels with square cross sections (Fig. 3b).
The crosssectional areas of the experimental apparatus
is used to determine the width and depth of the two chan
nels. Using this procedure, depth and width of the lower
mixing zone, Lmix is 3.10 cm. Similarly, the depth and
width of upper riser zone, L,,, is 2.21 cm.
As mentioned previously, the particle diameter and
density are equal to 0.01 cm and 1.1 g/cm3, respec
tively. Solids are injected from a slot in the mixing zone
which extends 1 cm and 5 cm from the bottom along the
side wall of the mixing zone. The solids are injected at
350 K. Gas is injected from the bottom of the mixing
zone at a temperature of 350 K. A constant value of
1.8 x 104 poise is used for the gasphase dynamic vis
cosity /g. The gas density is calculated using the ideal
gas law. The injected gas is composed of three species:
H2O (steam), CO2, and N2 having mass fractions of
0.15, 0.10, and 0.75, respectively. Similarly, the injected
sorbent is composed of two species: K2COs and C (the
activated carbon supporting material) having mass frac
tions of 0.35 and 0.65, respectively. In addition to the
above two species, the product of carbonation, KHCO3,
is included as a component in the solidphase.
7th International Conference on Multiphase Flow,
ICMF 2010, Tampa, FL, May 30 June 4, 2010
For the velocity field, no slip and partial slip bound
ary conditions are applied at the walls for gas and solid
phases, respectively. For the species and temperature
fields, a zero flux boundary condition is applied at walls
for both phases. Mass conservation (for both phases)
and granular energy evolution equations are discretized
using the firstorder upwinding scheme. Secondorder
accurate Superbee scheme is used to discretize the mo
mentum, species, and temperature evolution equations.
2 Results
2.1 Pressure Drop and Solids
Inventory
Two simulations were performed at a baseline operating
condition using the Gidaspow and EMMS drag model.
At this operating condition, the solid is injected at 15 g/s
and gas flow rate is 1 g/s. A uniform grid was used com
posed of cells which were 0.206 cm wide and 1 cm high.
Figure 4a shows a contour plot of gasvoidage Eg in
the mixing zone and the lower section of the riser af
ter 500 s. Figure 4b shows the same data with different
contour levels. Figure 4c shows contour plot for Eg ob
tained using the EMMS corrected Gidaspow drag cor
relation. From the contour plots (Figs. 4a and 4b), it
can be observed that the standard Gidaspow drag corre
lation results in less hold up of solid particles than the
EMMS correlation. The maximum solid volume frac
tion obtained from Gidaspow drag correlation is about
0.1. However, simulations using the EMMS corrected
drag predict a highly unsteady state with solid volume
fractions locally approaching the closepacking limit.
Comparison of the differential pressure (in mm of
H2O) for the four subsections of the reactor provides
a more quantitative assessment of the two models. Ta
ble 1 reports these values for the experiments (Yi et al.
2007), simulations using the EMMS drag model at two
different mesh resolutions and simulations using the Gi
daspow drag model. The significant under prediction of
differential pressure (10 mm vs. 100200 mm for the
mixing section) by the Gidaspow model would seemed
to be due to the lack of solids holdup as indicated by
the void fraction field. The simulation using the EMMS
corrected drag correlation predict a differential pressure
(80 mm), which is much closer to the experiment.
2.2 Mesh Resolution and Model
Dimensionality
In Table 1, simulations results for two mesh resolu
tions are shown from EMMS corrected drag correla
tion. Doubling the mesh resolution in both directions,
6x = 0.206 cm by 6y = 1.0 cm to 6x = 0.103 cm by
7th International Conference on Multiphase Flow,
ICMF 2010, Tampa, FL, May 30 June 4, 2010
Dri=1.78 cm
x
x
Dm,= 3.5 cm
Sorbent
injection
T 3.0 cm
__t 3.5cm
t Flue gas inlet
'p
51
Ln=3.1 cm
Sorbent
injection
Flue gas inlet
Figure 3: Schematic (not drawn to scale) of computational domain in (a) 2D and (b) 3D MFIX simulations.
Pressure differ
ence
(mm of H2 0)
KIER
(Yi et al. 2007)
Standard Gi
daspow
6x = 0.206 cm,
6y = 1.0 cm
DP1
DP2
DP3
DP4
EMMS
EMMS
S0.206 cm, 6x
1.0 cm 6y
103
250
226
176
Table 1: Comparison of the average pressure difference across sections DP1, DP2, DP3, and DP4 (cf Fig. 2) obtained
from 2D MFIX simulations using the standard Gidaspow drag correlation and the EMMS corrected drag
correlation for two different mesh resolutions with the KIER results (Yi et al. 2007).
6y 0.5 cm, increases the average pressure difference
values across the four sections by approximately 1l' .
However, this resulted in a significant increase in com
putational time. The coarse grid simulations required 4
days on 4 processors compared to 24 days using 8 Intel
Xeon L5335 series processors required for the fine grid
simulations
A threedimensional simulations was also performed.
As mentioned early, the cylindrical geometry is repre
sented by square duct of equivalent crosssectional area
(cf. schematic in Fig. 3b). The same gas and solids mass
flow rates are the same as those mentioned in the previ
ous section. Uniforms cells 0.206cm wide, 1.0 cm high
and 0.206 cm deep were used.
Figure 5a shows the evolution of the differential pres
sure across different sections in 3D MFIX simulation
over the entire course of simulation. The simulation at
tains a stationary state after about 100 seconds. The last
= 0.103 cm,
0.5 cm
7th International Conference on Multiphase Flow,
ICMF 2010, Tampa, FL, May 30 June 4, 2010
EP g
1.0
10.0
EPg
1.00
0.97
0.95
0.93
10.90
AY
EP_g
1.0
0.7
0.5
0.2
0.0
Figure 4: Contour plots of the gasvoidage Eg from 2D MFIX simulations showing the effect of drag closure. (a)
& (b): standard Gidaspow drag correlation; (c): EMMS corrected Gidaspow drag correlation. Due to very
skewed aspect ratio (600 cm tall and 3.5 cm maximum width), only the lower mixing zone and a subsection
of the riser zone are shown here.
500 DP1 DP1
: 450 DP2 :I450 DP2
E 400 DP3 E 400 DP3
E DP4 E DP4
350 350
U U
300 300 
250 250 
O 200 0 200 
150 150
100 100
a. a. q5
Figure 5: Evolution of the pressure difference across sections DP1, DP2, DP3, and DP4 obtained from 3D MFIX
simulation for.
40 seconds are shown in Figure 5b. The average dif
ferential pressure values, obtained by averaging between
100 and 400 seconds, are equal to 140, 291, 224, and 121
mm of H20 across sections DP1, DP2, DP3, and DP4,
respectively. These average differential pressure values
are on an average within _'..'. of those obtained from
the coarse grid 2D simulation reported in Table 1. The
computational cost to simulate 500 seconds, 3D simu
lation takes nearly 34 days of wall time using 16 Intel
Xeon L5335 processors compared to the coarse mesh 2
D simulation that takes only 4 days using 4 processor.
Both the 3D and finemesh 2D simulations require
significantly more computational resources than the
coarsemesh simulation. Further, the fact that the results
are more sensitive to the drag model than the mesh reso
lution support the use of the coarse mesh 2D simulations
as an efficient platform for the development of a model
ing approach for sorbent based CO2capture systems.
2.3 C02 Capture
At the baseline operating condition, the CO2 removal of
the 2D coarse grid, 2D fine grid and 3D simulations are
compared. For all 3 simulations the EMMS drag closure
was used. The CO2 removal is normalized based on the
CO2 mass fraction locally and at the inlet, Xco2, ,
%CO2 removal
Xco2 Xco2
2, X 2x 11" 1. (16)
Xc02,,,
Figure 6a shows the evolution of % CO2 removal at
outlet obtained from these simulations. It can be seen
that all the three simulations predict approximately I,' .
CO2 removal at the outlet.
Figure 6b shows the time averaged % CO2 removal at
different axial locations along the carbonator. From the
figure, it can be seen that approximately ".II'. of CO2
removal I'' of the 1,I'. net removal) occurs in the
mixing zone which extends from y = 0 to y = 60 cm
(shown by the dashed line in the figure). The similar
ity of the results from the three simulations provide fur
ther confirmation that the coarse mesh 2D simulations
are sufficiently mesh independent.
Five additional 2D coarse grid simulations were per
formed to investigate the sensitivity of the reactor per
formance to solid circulation and gas flow rate. The re
sults of the baseline and the additional simulations are
shown in Figures 7a and 7b which show the compari
son of % CO2 removal at outlet as a function of solids
and gas mass flow rates, respectively. The MFIX simu
lations predict the sensitivity of the reactor performance
with respect to solids circulation rates quite accurately
(1.96 vs. 1.92 %C02/(g/s)). The simulations also
predict the leveling off CO2 removal capacity as the
solids mass flow rate approaches 20 g/s similar to what
was measured experimentally. The results for sensitiv
ity of the reactor performance with respect to the gas
7th International Conference on Multiphase Flow,
ICMF 2010, Tampa, FL, May 30 June 4, 2010
mass flow rate (Fig. 7b) are not as favorable. The simu
lated CO2 capture is much more sensitive ( 118 vs. 37
%C02/(g/s)) to the gas flow rate than observed in the
KIER experiments.
The ability of the simulations to predict the solid cir
culation rate sensitivity but not the gas flow rate sensi
tivity could be due to any number of assumptions and
simplifications in our model. Below we briefly discuss
aspects of the model of the reactor which may explain
this behavior:
Reaction Model The homogeneous reaction model
used in this study is independent of the steam con
centration, the temperature and gassolids slip ve
locity. Of these three, it is likely that including
a dependence on the gassolid slipvelocity via a
film diffusion model would be the only aspect of
the reaction model which would provide a sensitiv
ity to the gas flow rate. Including the other two as
pects would make the reaction model applicable for
a wider range of carbonator operating temperatures
and flue gas compositions.
Drag Model The EMMS drag correlation used in
the study (Lu et al. 2007) was developed for par
ticles with a diameter and density equal to 54 pm
and 930 kg/m3, respectively. In contrast, the diam
eter and density of the particles used in this study
are 100 pm and 1100 kg/m3, respectively. Despite
these differences the model does seem to provide a
necessary correction to the homogeneous drag law,
however these differences may explain the inability
of the simulations to predict the sensitivity to the
gas flow rate.
3 Summary
In this study we explored the feasibility applying a con
tinuum model to simulate the CO2 capture from flue gas
by dry solid sorbents. Recent experiments conducted
at KIER (Yi et al. 2007) of continuous CO2 capture
by Potassium based dry sorbent were used to compare
with simulations using the open source CFD software
MFIX (Syamlal et al. 1993). It was observed that sim
ulations using the standard Gidaspow drag correlation
severely under predicted the pressure drop across the
bed, and hence the solid hold up. An EMMS corrected
drag correlation, which accounts for clustering effects
improved the prediction of the pressure drop.
2D MFIX simulations with EMMS corrected drag
correlation and a homogeneous reaction model were
found to qualitatively predict the operation of the KIER
reactor. The sensitivity of CO2 removal with respect to
solid mass flow rate obtained was found to be within 2'.
of the sensitivity observed experimentally. However, the
7th International Conference on Multiphase Flow,
ICMF 2010, Tampa, FL, May 30 June 4, 2010
Inlet CO2 mass fraction =10%
60n a 2D, 5x=0.206, Sy=1
A 2D, 5x=0.103, Sy=0.5
o 3D, 5x=6z=0.206, Sy=1
50
40
30
0 100 200 300 400
Time (sec)
0 200 ) 400
y (cm)
Figure 6: (a) Comparison of temporal evolution of percentage CO2 removal (Eq. 16) at the outlet from 2D and 3
D MFIX simulations. (b) Comparison of percentage CO2 removal along the reactor height from 2D and
3D MFIX simulations; dashed line represents the end of the mixing zone at y = 60 cm (cf. schematic in
Fig. 1). Gas and solids mass flow rates are equal to 1 g/s and 15 g/s, respectively. Grid size dimensions are
in centimeters.
simulations were approximately three times more sen
sitive to the gas flow rate than the experiments. Limi
tations in both the drag model and the reaction model
could explain these discrepancies. It is conjectured that
this could be addressed with a reaction model that is sen
sitive to gassolid slip velocity and/or a clustercorrected
drag model with a stronger dependence on local flow
conditions.
References
Arastoopour, H., Gidaspow, D., 1979. Analysis of
igt pneumatic conveying data and fast fluidization us
ing a thermohydrodynamic model. Powder Technology
22 (1), 77 87.
Bird, R. B., Stewart, W. E., Lightfoot, E. N., 1962.
Transport Phenomena. John Wiley & Sons.
Ergun, S., 1952. Fluid flow through packed columns.
Chem. Eng. Prog. 48, 8994.
Fortes, A., Joseph, D., Lundgren, T., 1987. Nonlinear
mechanics of fluidization of beds of spherical particles.
Journal of Fluid Mechanics 177, 467 483.
Gidaspow, D., 1994. Mutliphase flow and fluidization.
Academic Press.
Gunn, D. J., 1978. Transfer of heat and mass to particles
in fixed and fluidized beds. Intl. J. Heat Mass Transfer
21,467476.
Hill, R. J., Koch, D. L., Ladd, A. J. C., 2001. The first
effects of fluid inertia on flows in ordered and random
arrays of spheres. J. Fluid Mech. 448 (213241).
Li, J., Kwauk, M., 2003. Exploring complex systems
in chemical engineeringthe multiscale methodology.
Chemical Engineering Science 58 (36), 521 535.
Liu, X., Gao, S., Li, J., 2005. Characterizing particle
clustering behavior by pdpa measurement for dilute gas
solid flow. Chemical Engineering Journal 108 (3), 193 
202.
Lu, B., Wang, W., Li, J., Wang, X., Gao, S., Lu, W.,
Xu, Y, Long, J., 2007. Multiscale cfd simulation of
gassolid flow in mip reactors with a structuredependent
drag model. Chemical Engineering Science 62 (1820),
5487 5494, 19th International Symposium on Chemi
cal Reaction Engineering From Science to Innovative
Engineering ISCRE19.
McKeen, T, Pugsley, T., 2003. Simulation and experi
mental validation of a freely bubbling bed of fcc catalyst.
Powder Technology 129 (13), 139 152.
NETL, Accesssed September 2009. Innovations
for Existing Plants, PostCombustion C02 Control.
http://www.netl.doe.gov/technologies/coalpower/ew
Park, S. W., Sung, D. H., Choi, B. S., lee, J. W.,
Lumazawa, H., 2006. Carbonation kinetics of Potassium
Carbonate by Carbon Dioxide. Industrial & Engineering
Chemistry 12 (4), 522530.
0 o
E3 2D, 8x=0.206, y=1
A 2D, x=0.103, y=0.5
0 3D, x=6z=0.206, y=1
600
7th International Conference on Multiphase Flow,
ICMF 2010, Tampa, FL, May 30 June 4, 2010
Figure 7: Comparison of percentage CO2 removal at outlet with (a) solids mass flow rate and (b) gas mass flow rate
obtained from 2D MFIX simulations with the KIER experiments. In (a), a constant gas mass flow rate of
1 g/s is used. In (b), a constant solids mass flow rate of 10 g/s is specified. A constant 2D mesh resolution
of (6x = 0.206 cm, 6y = 1.0 cm) is used for all cases. Lines are linearleast squares fit to the data and the
magnitude of their slopes is indicated by the text boxes.
Qi, H., Li, F., Xi, B., You, C., 2007. Modeling of drag
with the eulerian approach and emms theory for hetero
geneous dense gassolid twophase flow. Chemical En
gineering Science 62 (6), 1670 1681.
Rao, A. B., Rubin, E. S., 2002. A Technical, Economic,
and Environmental Assessment of AmineBased CO2
Capture Technology for Power Plant Greenhouse Gas
Control. Environ. Sci. Technol 36 (20), 44674475.
Ryu, C. K., Lee, J. B., Eom, T H., Oh, J. M., Yi, C. K.,
2005. Development of na and kbased sorbents for co2
capture from flue gas. In: Proceedings of the 4th An
nual Conference on Carbon Capture and Sequestration.
Washington, DC.
Schiller, L., Naumann, A. Z., 1933. A Drag Coefficient
Correlation. Z. Ver. Deutsch Ing., 318320.
Syamlal, M., Rogers, W, O'Brien, T. J., 1993. Mfix
documentation: Theory guide. Tech. Rep. DOE/METC
95/1013, NTIS/DE95000031, NETL, Department of
Energy, uRL http://www.mfix.org.
van der Hoef, M. A., Beetstra, R., Kuipers, J. A. M.,
2005. LatticeBoltzmann simulations of lowReynolds
number flow past mono and bidisperse arrays of sphere:
results for the permeability and drag force. J. Fluid
Mech. 528, 233254.
Wall, T. F, 2007. Combustion processes for carbon cap
ture. Proceedings of the Combustion Institute 31 (1), 31
47.
Wen, C. Y, Yu, Y H., 1966. Mechanics of fluidization.
Chem. Eng. Prog. Symp. Series 62, 100111.
Yi, C. K., Jo, S. H., Seo, Y, Lee, J. B., Ryu, C. K.,
2007. Continuous operation of the potassiumbased dry
sorbent CO2 capture process with two fluidizedbed re
actors. International Journal of Greenhouse Gas Control
1 (1), 31 36.

