Chemical Engineering Science 61 (2006) 6511 – 6529

www.elsevier.com/locate/ces

Simulations of chemical absorption in pilot-scale and industrial-scale packed

columns by computational mass transfer
G.B. Liu a , K.T. Yu a , X.G. Yuan a,∗ , C.J. Liu a , Q.C. Guo b
a State Key Laboratory for Chemical Engineering (Tianjin University), Chemical Engineering Research Center and School of Chemical Engineering and
Technology, Tianjin University, Tianjin, 300072, China
b Institute of photoelectronics thin film devices and technique, Nankai University, Tianjin, 300071, China

Received 28 December 2005; received in revised form 22 May 2006; accepted 22 May 2006
Available online 3 June 2006

Abstract
A complex computational mass transfer model (CMT) is proposed for modeling the chemical absorption process with heat effect in packed
columns. The feature of the proposed model is able to predict the concentration and temperature as well as the velocity distributions at once
along the column without assuming the turbulent Schmidt number, or using the experimentally measured turbulent mass transfer diffusivity.
The present model consists of the differential mass transfer equation with its auxiliary closing equations and the accompanied formulations of
computational fluid dynamics (CFD) and computational heat transfer (CHT). In the mathematical expression for the accompanied CFD and
CHT, the conventional methods of k.
and t 2 .
t are used for closing the momentum and heat transfer equations. While for the mass transfer
equation, the recently developed concentration variance c2 and its dissipation rate
c equations (Liu, 2003) are adopted for its closure. To test the
validity of the present model, simulations were made for a pilot-scale randomly packed chemical absorption column of 0.1 m ID and 7 m high,
packed with 1/2 ceramic Berl saddles for CO2 removal from gas mixture by aqueous monoethanolamine (MEA) solutions (Tontiwachwuthikul
et al., 1992 ) and an industrial-scale randomly packed chemical absorption column of 1.9 m ID and 26.6 m high, packed with 2 stainless
steel Pall rings for CO2 removal from natural gas by aqueous MEA solutions (Pintola et al., 1993). The simulated results were compared
with the published experimental data and satisfactory agreement was found between them in both concentration and temperature distributions.
Furthermore, the result of computation also reveals that the turbulent mass transfer diffusivity Dt varies along axial and radial directions. Thus
the common viewpoint of assuming constant Dt throughout the whole column is questionable, even for the small size packed column. Finally,
the analogy between mass transfer and heat transfer in chemical absorption is demonstrated by the similarity of their diffusivity profiles.
䉷 2006 Elsevier Ltd. All rights reserved.

Keywords: Computational mass transfer (CMT); Turbulent mass transfer diffusivity; Simulation; Packed bed; Chemical absorption; Mathematical modeling

1. Introduction especially the diffusion of species in the liquid phase, is strongly

coupled with the extent of turbulent flow as well as the heat
In the chemical industry, packed columns have been widely effect if involved. As the process is complicated, the behavior
used in separation and purification processes involving gas and and prediction of the mass transfer in packed column have been
liquid contact such as distillation and absorption due to its high an active research area for decades in both experimental work
efficiency, high capacity and low pressure drop. Despite the suc- and mathematical modeling. It is known that, in most case,
cess of applying structured packing in the recent years (Gualito the conventional plug-flow assumption is not applicable due
et al., 1997; Spiegel and Meier, 2003), the randomly packed to the presence of non-uniform packing and the creating tur-
columns are still commonly used in the separation processes. bulence, especially in the case of column-to-particle diameter
The mass transfer process undertaking in a packed column, ratios ( = aspect ratios) lower than about 10 (Wen et al., 2001;
Giese et al., 1998; Ziolkowska and Ziolkowski, 1988; Lerou
∗ Corresponding author. Tel.: +86 22 27404732; fax: +86 22 27404496. and Froment, 1977; Bey and Eigenberger, 1997; Zhang, 1986;
E-mail address: yuanxg@tju.edu.cn (X.G. Yuan). Zhang and Yu, 1988; Yuan et al., 1989). Thus the investigation

0009-2509/$ - see front matter 䉷 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2006.05.035
6512 G.B. Liu et al. / Chemical Engineering Science 61 (2006) 6511 – 6529

on the non-uniform distributions of velocity and temperature simulated results on the distributions of temperature and con-
in the packed column have been received much attention in the centration along the axial and radial directions were compared
past. The notable advance is coming from the use of compu- with the published experimental data.
tational fluid dynamics (CFD) and computational heat transfer
(CHT) to help the solution of this problem, as the velocity and 2. The reaction mechanism of CO2 absorption by aqueous
temperature profiles in the packed column can be predicted by MEA solution
such methodology (Yin et al., 2000, 2002; Hjertager et al., 2002;
Liu, 2001, Jiang et al., 2002a,b). However, for the prediction of When CO2 is being absorbed and reacts with aqueous MEA
concentration profile, it is necessary to use the empirical turbu- solutions, the following three overall reactions are taken place
lent Schmidt numbers or the experimentally determined disper- at the condition of carbonation ratio (or CO2 loading) less than
sion coefficients obtained by tracer technique as an approximate 0.5 mol of CO2 per mole of MEA. (Danckwerts, 1979; Astarita
substitution to the unknown turbulent mass transfer diffusivity et al., 1964):
(Yin et al., 2000; de Lemos and Mesquita, 2003). Obviously,
the use of such empirical method for finding the concentration CO2,g −→ CO2,L + HA , (1)
distribution is not dependable, especially for the cases where k2
the empirical relationship or the experimental coefficients for CO2,L + 2BNH2 −→ BNHCOO− + BNH+
3 + HR , (2)
the investigated system are unavailable. The best way of over-
coming this difficulty is to eliminate the turbulent mass transfer CO2,L + BNHCOO− + 2H2 O −→ BNH+ −
3 + 2HCO3 , (3)
diffusivity from the differential mass transfer equation in the where letter B denotes the group HOCH2 CH2 –, step (1) repre-
model computation. The recently developed c2 .
c model pro- sents the physical absorption of CO2 by water, HA is the accom-
posed by Liu (2003) is suitable for this purpose. With the idea panied heat of solution. At very short times of the liquid–gas
of using such model, Sun et al. (2005) successfully obtained the encountered in industrial absorbers, the reaction (3) can be ne-
concentration profiles on the sieve trays of a commercial-scale glected, and only reaction (2) affects the absorption rate of CO2
distillation column without relying on the empirical correlation (Sada et al., 1976). Reaction (2) takes place in two steps:
of turbulent mass transfer diffusivity and showed satisfactory
agreement with the published experimental measurement. CO2,L + BNH− − +
2 −→ BNHCOO + H , (4)
The chemical absorption has been extensively used in gas
purification as well as in many chemical processes. For in- BNH2 + H+ −→ BNH+
3. (5)
stance, the absorption of CO2 in randomly packed columns by Reaction (4) could be considered as second order, and is the
alkaline solutions such as NaOH, monoethanolamine (MEA), rate controlling step, because reaction (5) is a proton transfer
diethanolamine (DEA) and methyldiethanolamine (MDEA) reaction and virtually instantaneous. Therefore, the absorption
aqueous solutions is commonly adopted in acid gas treatments. of CO2 in MEA solutions can be regarded as gas absorption
Among them, the use of aqueous MEA solutions as absorbent accompanied by an irreversible second-order reaction with a
takes more share in the industrial process of CO2 removal, and stoichiometric coefficient of 2, and the overall reaction is rep-
the research concerned have been reported (Danckwerts, 1979; resented by reaction (2). The reaction rate Rc can be expressed
Hikita et al., 1979; Tontiwachwuthikul et al., 1992; Astarita by the following equation:
et al., 1964; Aboudheir et al., 2003). However, the investiga-
tions were mostly based on overall mass balances and plug Rc = k2 [CO2 ][MEA]. (6)
flow, without considering the non-ideal behaviors of flow, heat
and mass transfer in the randomly packed bed. The problem 3. The experimental data available
of close modeling is remained to be solved.
With the increasing power of computer and rapid devel- The validity of the proposed model for the chemical absorp-
opment of numerical computational method, it is possible to tion of CO2 by MEA solution can be testified by the compar-
solve the problem by establishing a complex model to describe ison between the model prediction and the experimental data.
in-depth the transport phenomena of the chemical absorption There are two sets of data available:
process with a set of differential equations. Thus, a complex (1) The experimental data of a pilot-scale column were re-
computational mass transfer model (CMT) is proposed in this ported by Tontiwachwuthikul et al. (1992). They conducted the
paper for modeling the chemical absorption process with heat experiment for the absorption of CO2 from air by using aque-
effect. The present model consists of the differential mass trans- ous MEA solution at total pressure of 103.15 kPa in a column
fer equation, closing with the c2 .
c equations, and the accom- of 0.1 m ID and packed with 1/2 (∼ 1.27 cm) ceramic Berl
panied formulations of CFD and CHT. In the mathematical saddles with a total packing height of 6.55 m. The column con-
formulation of the accompanied CFD and CHT, the conven- sisted of six equal-height sections, and the samples were taken
tional methods of k.
and t 2 .
t are used for closing the mo- at the inlet and outlet of each section for analyzing the concen-
mentum and heat transfer equations. To testify the validity of tration.
the proposed model, simulation was made to a pilot-scale and (2) The operating data of an industrial column were reported
an industrial-scale randomly packed columns undertaking the by Pintola et al. (1993). The column was used for reducing the
chemical absorption of CO2 by aqueous MEA solution. The CO2 content in the natural gas from approximately 2% to less
 G.B. Liu et al. / Chemical Engineering Science 61 (2006) 6511 – 6529 6513

than 100 ppm by the aqueous MEA solution absorption. The be solved by using the c2 .
c model (Liu, 2003) as follows.
column was 1.9 m in diameter and packed with 2 stainless steel According to the c2 .
c model, the Dt can be expressed by
Pall rings in three sections with total packing height of 14.1 m.
The operating pressure was ranging from 5.39 to 7.60 atm. All
1/2
k c2
their experimental data were given only at the top and the Dt = Cc0 k , (9)
bottom of column.

c

4. The computational mass transfer (CMT) model for the where the concentration variance c2 and its dissipation rate
c
chemical absorption concerned are defined by

4.1. Assumptions c2 ≡ cc, (10)

 
The following assumptions are made for the present CMT jc jc

c ≡ D . (11)
model for simulating the chemical absorption of CO2 by aque- jxj jxj
ous MEA solution in randomly packed columns:
1. The pseudo-single-liquid model is used and the liquid The detailed derivation of c2 .
c model was given by Liu
phase is assumed to be pseudo-continuous, and the gas phase (2003) and Sun et al. (2005). The model equations used in this
is considered to be uniform along the radial direction. The flow paper are given below.
is axis-symmetric. (2) The concentration variance c2 equation:
2. The gas absorption operation is steady, and the liquid is
   
incompressible, which means the liquid density does not vary Dt
with the liquid temperature and concentration. This assumption ∇ • (huc2 ) − ∇ • h D + ∇c2
c
is reasonable for the current cases where the changes of liquid
= 2hD t ∇ C̄∇ C̄ − 2h
c (12)
temperature and liquid concentration are not large enough to
make substantial variation of the liquid density. (3) The dissipation rate
c equation:
3. Only the CO2 component in the gas phase is absorbed    
by the aqueous MEA solutions, and the solvent water does not Dt
∇ • (hu
c ) − ∇ • h D + ∇
c
transfer to the gas phase. 
c
4. The heat of solution and reaction generated is all absorbed
c
2c

c
instantaneously by the liquid phase, and the effect of packing = Cc1 hD t ∇ C̄∇ C̄ −2 − Cc2 h −2 − Cc3 h . (13)
c c k
on the heat transfer is neglected. There is not heat transfer
between gas and liquid phases. The constants in Eqs. (9)–(13) are given as follows (Zhang,
5. The gas absorption process is adiabatic, which means that 2002):
there is no heat exchange between the column and environment.
This is reasonable and justified by Pandya (1983). Cc0 = 0.11, Cc1 = 1.8, Cc2 = 2.2, Cc3 = 0.8,
c = 1.0 and 
c = 1.0.
4.2. Model equations
4.2.2. The accompanied CFD equations
The CMT model consists of the differential mass transfer (1) The continuity equation:
equation with its auxiliary closing equations and the accompa-
nied formulations of CFD and CHT. ∇ • (hu) = M, (14)

4.2.1. The mass transfer equation and its auxiliary closing where  is the liquid density, h is the volume fraction of liquid
equations phase based on pore space, u is the interstitial velocity vector,
(1) The equation for time average MEA mass fraction C̄ in M is the source term of the continuity equation due to the
liquid phase: chemical absorption of CO2 from gas phase, which is equal to
the quantity of CO2 absorbed by the aqueous solutions per unit
∇ • (huC̄) = ∇ • (hDeff ∇ C̄) + MC, (7) volume and unit time.
(2) The momentum equation:
where MC is the MEA sink term which comes from the chem-
ical reaction between absorbed CO2 and MEA aqueous solu-
∇ • (huu) − ∇ • (heff (∇u + (∇u)T − 23 ∇ • uI))
tions, Deff is the effective diffusivity of MEA, which is defined
by = −h∇p + FLG + h(FLS + g), (15)

Deff = D + Dt , (8) eff =  + t , (16)

where D is the molecular diffusivity of MEA in the liquid phase, k2

t = C , (17)
and Dt is the turbulent diffusivity for mass transfer, which can

6514 G.B. Liu et al. / Chemical Engineering Science 61 (2006) 6511 – 6529

where , t and eff represent the molecular, turbulent and the model constants in t 2 .
t equations are (Nagano and Kim,
effective viscosities of the liquid, respectively. The turbulent 1988):
viscosity t is unknown and can be solved simultaneously by
the standard k-
model equations given below, and I is the unit Ct0 = 0.11, Ct1 = 1.8, Ct2 = 2.2, Ct3 = 0.8,
tensor. FLG is the interface drag force between gas phase and t = 1.0 and 
t = 1.0.
liquid phase, FLS is the flow resistance created by the random
The concentration and the temperature as well as the velocity
packing, which is considered to be the body force.
distributions in the absorption column can be obtained by the si-
(3) The turbulent kinetic energy k equation:
multaneous solution of the foregoing model Eqs. (7), (12)–(15),
   
t (18)–(22) without assuming the empirical Schmidt number or
∇ • (huk) − ∇ • h  + ∇k knowing the experimental mass transfer diffusivity. However,
k
before the computation of foregoing equation system, the terms
= heff ∇u • ((∇u + (∇u)T ) − 23 ∇ • uI) − h
. (18)
of h, M, FLG , FLS , MC, Q appearing in the equations must
(4) The turbulent dissipation rate
equation: be determined. In this paper, most of these terms are obtained
from existing correlations as shown in the following sections.
   
t
∇ • (hu
) − ∇ • h  + ∇


4.3. Evaluation of various terms in the model


= c1 heff ∇u • ((∇u + (∇u)T ) − 23 ∇ • uI)
k 4.3.1. The volume fraction of liquid phase

2 The volume fraction of liquid phase h based on pore space
− c2 h (19) can be determined from the total liquid holdup Ht and the
k
porosity  under the operating condition of the gas–liquid two
the parameters which appear in k.
model Eqs. (17)–(19) are phases flow:
customarily chosen to be
h = Ht /. (23)
C = 0.09, k = 1.0, 
= 1.3,
c1 = 1.44 and c2 = 1.92. The total liquid holdup can be obtained as follows:

Ht = Hs + Hop . (24)
4.2.3. The accompanied CHT equations
(1) The time averaged liquid temperature T equation: Shulman et al. (1955) reported the value of static holdup Hs
to be 0.0317 for 1/2-in ceramic Berl saddles. The static holdup
∇ • (cp huT ) = ∇ • (cp heff ∇T ) + Q, (20) Hs for stainless steel Pall rings could be determined from the
correlation by Engel et al. (1997):
where cp is the liquid phase specific heat and taken as a con-
 g 
stant; eff is the effective thermal diffusivity of liquid phase, Hs = 0.033 exp −0.22 2 . (25)
defined as eff =+t , in which  and t are the molecular and a
turbulent thermal diffusivities, respectively, and t can be cal- Otake and Okada obtained a correlation for the operating
culated by the t 2 .
t closing equation set given below; Q is the holdup Hop , as reported by Sater and Levenspiel (1966):
thermal source term including the heat of solution and reaction
and other thermal effects. Hop = 1.295(Re)0.676
L (Ga)−0.44
L (ad p ), (26)
(2) The temperature variance t 2 equation: The model equa-
tions of temperature variance t 2 and its dissipation rate
t pro- where (Re)L = dp L/ is the Reynolds number of liquid phase;
posed by Nagano and Kim (1988) were used. However, in this (Ga)L = dp3 g2 /2 is the Gallileo number of liquid phase; a
paper, some terms of the t 2 and
t equations were neglected is the surface area per unit volume of packed bed; dp is the
due to their unimportance as compared with other terms. The nominal diameter of the packed particle.
simplified model equations were given as follows: The porosity  of randomly packed bed is changed from a
constant around the center to a maximum in the neighborhood
   
t of the wall region, which were observed by many experimental
∇ • (hut 2 ) − ∇ • h  + ∇t 2 investigations (Giese et al., 1998; de Klerk, 2003; Roblee et al.,
t
1958). Thus the uneven porosity distribution should be consid-
= 2ht ∇ T̄ ∇ T̄ − 2h
t . (21)
ered and calculated by the following correlation given by Liu
(3) The dissipation rate
t equation: (2001):
   
t (1 − ∞ )
∇ • (hu
t ) − ∇ • h  + ∇
t  = ∞ + Er (1 − 0.3pd )

t 2



t
2

t 2 R−r
= Ct1 ht ∇ T̄ ∇ T̄ − Ct2 h t − Ct3 h , (22) × cos + 0.3pd , (27)
t 2 t 2 k c + 1.6Er 2 pd dp
 G.B. Liu et al. / Chemical Engineering Science 61 (2006) 6511 – 6529 6515

where ∞ is the porosity of an unbounded packing, R is the where the wetted surface area aw is obtained from the following
radius of the column, r is the position in radial direction, Er is correlation by Onda et al. (1968):
the exponential decaying function, which is given by
aw
  3/4  a

R−r  0.75  L 0.1 L2 a −0.05  L2 0.2

Er = exp −1.2pd , (28) = 1−exp −1.45
ct
.
dp  a 2 g a
(35)
where pd is the period of oscillation normalized by the nominal
particle size and pd = 0.94 for Berl saddles, pd = 0.94 × (2 + According to the research result by Onda et al. (1968),
1.414)/3 for Pall rings; c is a constant depending on the ratio the w in Eq. (35) is considered to be equal to eff , and the
of the particle size to column size: flow rate of liquid is related to the local liquid velocity by
L = h|u|.
  3/4  The viscosity  of aqueous MEA solutions can be calcu-
2R R
c = − 1.6 exp −2.4pd , (29) lated from the correlation developed from experimental data by
n  pd d p dp Weiland et al. (1998):

 
 100C̄(2373 + 2118.6C̄)[rc (−2.2589 + 0.0093 × T + 1.015C̄) + 1.0]
= exp , (36)
H2 O T2
where rc is the carbonation ratio of aqueous MEA solutions.
where The surface tension  of aqueous MEA solutions is deter-

mined by following equation (Vazquez et al., 1997):
2 R
n = int . (30)  = H2 O − (H2 O − MEA )
1 + 1.6 exp[−2.4pd (R/dp ) ] d p
3/4 p d

(0.63036 − 1.3 × 10−5 (T − 273.15))xH2 O
× 1+ xMEA .
1 − (0.947 − 2 × 10−5 (T − 273.15))xH2 O
4.3.2. The source term M in continuity equation
(37)
The source term M, the rate of mass transfer per unit volume,
can be determined from the following equation based on the The pure liquid surface tensions of water and MEA are obtained
well-known two-film model: as follows:
H2 O = 76.0852 − 0.1609(T − 273.15), (38)
M = KG aeff MCO2 (pCO2 − pe,CO2 ), (31)
MEA = 53.082 − 0.1648(T − 273.15), (39)
where the overall coefficient of gas phase KG is related to the where xH2 O and xMEA denote the molar fractions of water and
individual film coefficients as follows: MEA in aqueous MEA solutions, respectively.
The enhancement factor E, which is defined as the ratio of the
1 1 1
= + . (32) mass transfer coefficient kR,L for the absorption with chemical
KG kG kL EH e reaction to the mass transfer coefficient kL for the physical
absorption (Danckwerts, 1970), varies along the column height.
In the foregoing equations, kL , kG are the film coefficients
For the irreversible second-order reaction such as CO2 –MEA
of mass transfer of liquid phase and gas phase respectively; eff
reaction, Wellek et al. (1978) derived the following explicit
is the effective area for mass transfer between the gas phase
correlation for the calculation of enhancement factor E with
and liquid phase; MCO2 is the molecular weight of CO2 ; pCO2
deviation less than 3%:
is the partial pressure of CO2 in main body of gas; pe,CO2 is
partial pressure of CO2 in equilibrium with the solution; E E = 1 + ((Ei − 1)−1.35 + (E1 − 1)−1.35 )−1/1.35 , (40)
is the enhancement factor for the mass transfer process with
chemical reaction; He is the Henry’s constant. where
The liquid phase and gas phase mass transfer coefficients kL , DX MEA
Ei = 1 + , (41)
kG are determined from the correlations by Onda et al. (1968): 2DCO2 ,L Xi,CO2
√
 1/3    −1/2 Ha
g L 2/3  E1 = √ , (42)
kL = 0.0051 (ad p )0.4 , tanh H a
 aw  DCO2 ,L
(33) DCO2 ,L k2 XMEA
Ha = . (43)
(kL )2
 0.7  1/3  −1
G G RG T
kG = 5.23 (ad p )−2 , (34) Here XMEA denotes the molar concentration of MEA in liquid
aG G D G aD G phase; DCO2 ,L represents the molecular diffusivity of CO2 in
6516 G.B. Liu et al. / Chemical Engineering Science 61 (2006) 6511 – 6529

aqueous MEA solutions; and k2 is the second-order reaction where the excess Henry’s quantity ℵMEA,H2 O for the present
rate constant for CO2 –MEA reaction. system can be determined by the correlation of Tsai et al.
Due to the chemical reaction of CO2 with MEA, the molec- (2000):
ular diffusivity of CO2 in the MEA aqueous solution DCO2 ,L
can be determined by the use of N2 O analogy as follows: ℵMEA,H2 O = − vMEA vH2 O (4.793 − 7.446 × 10−3 T
− 2.201vH2 O ), (55)
DCO2 ,L = DN2 O,L (DCO2 /DN2 O )w . (44)
where vMEA and vH2 O denote the volume fractions of MEA
The diffusivities of CO2 and N2 O in the pure water are deter- and water in aqueous MEA solutions respectively; the Henry’s
mined from the correlations by Versteeg and Vanswaaij (1988): constant (H eN2 O )MEA for N2 O in pure MEA liquid is obtained
from the correlation developed from the experimental data by
(DCO2 )w = 2.35 × 10−6 exp(−2119/T ), (45) Wang et al. (1992):

(DN2 O )w = 5.07 × 10−6 exp(−2371/T ). (46) (H eN2 O )MEA = (1.207 × 105 exp(−1136.5/T ))−1 . (56)

The diffusivity of N2 O in aqueous MEA solutions is deter- In order to calculate the gas phase film coefficient kG , some
mined by the correlation reported by Ko et al. (2001): physical properties of gas phase, such as molecular diffusiv-
ity DG of CO2 in gas phase, viscosity G and density G of
DN2 O,L = (5.07 + 0.865XMEA + 0.278XMEA
2
) gas phase, should be known. In this paper, DG in the gas mix-
  ture is obtained from the Blanc’s law and the binary molecular
−2371.0 − 93.4XMEA
× exp × 10−6 . (47) diffusivity was calculated by empirical correlation reported by
T
Poling et al. (2001). The gas mixture viscosity G is calcu-
The diffusivity of MEA molecule in aqueous MEA solutions lated by Bromley and Wilke correlation reported by Perry and
is calculated by using the correlation given by Snijder et al. Green (2001). The viscosity of air and pure CO2 is obtained
(1993): from Geankoplis (2003), and the vapor viscosity of pure hy-
drocarbons at low pressure is predicted by the method of Stiel
D = exp(−13.275 − 2198.3/T − 0.078142XMEA ). (48) and Thodos reported by Perry and Green (2001). The density
of gas mixture and pure components can be obtained according
The second-order reaction rate constant k2 for CO2 –MEA to the method of Lee and Kesler, which was reported by Perry
reaction is obtained from the correlation by Hikita et al. (1979): and Green (2001).
log k2 = 10.99 − 2152/T . (49)
4.3.3. Interface drag force FLG
According to Pohorecki and Moniuk (1988), the molar con- For irrigated packing, the pressure drop is greater than the
centration of CO2 at interface Xi,CO2 can be expressed in terms dry bed pressure drop
pd because of the presence of liquid
of Henry’s law: adhered to the packing surface. The part of increased pressure
drop,
pL , represents the interfacial drag force between gas
Xi,CO2 = H ep t yCO2 , (50) and liquid phases. Robbins (1991) correlated the total pressure
drop including
pd and
pL for irrigated packing as follows:
where He is the Henry’s constant for CO2 in MEA solutions; pt

pt =
pd +
pL , (57)
is the total pressure of gas phase; yCO2 is the volume fraction
of CO2 in gas phase. Similar to the calculation of DCO2 ,L , He
pd = p1 G2f × 10p2 Lf , (58)
 0.1
is also determined by the use of N2 O analogy in the following Lf
form:
pL = 0.774 (p1 G2f × 10p2 Lf )4 , (59)
20000
H e = H eN2 O (H eCO2 /H eN2 O )w . (51) where p1 = 0.04002, p2 = 0.0199, Gf is the gas loading fac-
tor, and Lf is the liquid loading factor. Gf and Lf can be
The Henry’s constant for CO2 and N2 O in water can be ob- determined by the following correlations:
tained from the following correlations proposed by Versteeg
   
and Vanswaaij (1988): 1.2 0.5 Fpd 0.5
Gf = G for pt
1.0 atm, (60)
G 65.62
(H eCO2 )w = (2.82 × 106 exp(−2044/T ))−1 , (52)    
1.2 0.5 Fpd 0.5
(H eN2 O )w = (8.552 × 106 exp(−2284/T ))−1 . (53) Gf = G × 100.0187G for pt > 1.0 atm,
G 65.62
The Henry’s constant for N2 O in aqueous MEA solutions is (61)
  0.5
calculated from the semi-empirical model developed by Wang 1000 Fpd
et al. (1992): Lf = L 0.2 for Fpd 15, (62)
 65.62
  0.5
ln H eN2 O = vMEA ln(H eN2 O )MEA 1000 65.62
Lf = L 0.1 for Fpd < 15. (63)
+ vH2 O ln(H eH2 O )w + ℵMEA,H2 O , (54)  Fpd
 G.B. Liu et al. / Chemical Engineering Science 61 (2006) 6511 – 6529 6517

The term Fpd is a dry packing factor, specified for a given

packing type and size. For 1/2 (∼ 1.27 cm) ceramic Berl sad-
dles, Fpd = 900 m−1 , and for 2 stainless steel Pall rings,
Fpd = 79 m−1 . Since the total operating pressure of chemical
absorption in this study is more than 1.0 atm, Eqs. (61) and (62)
are chosen to evaluate Lf and Gf .
Finally, the interface drag force FLG can be calculated by

pL
FLG = uslip , (64)
|uslip |

where uslip is the slip velocity between gas and liquid phases,
which is defined by

uslip = uG − u. (65)

As the quasi-single-liquid-phase CFD method is applied in the

present work, thus the average uG in axial direction is deter- Fig. 1. The simulation domain and boundary conditions arrangement.
mined from gas phase flow rate G, and the uG in radial direc-
tion is set to be zero.
4.4. The boundary conditions
4.3.4. Body force FLS
The computational domain and boundaries are shown in
The resistance of liquid flow due to the presence of random
Fig. 1. The boundary conditions for the above equation set are
packing is considered as body force, which can be calculated
specified as follows:
by Ergun (1952) equation with replacing the mean porosity ∞
(1) Inlet: At the top of the column, the boundary condition
by the porosity distribution function .
for the liquid phase is set to be u = uinlet , vinlet = 0, T = Tinlet ,

 C̄ = Cinlet , kinlet = 0.003u2inlet ,
inlet = 0.09(kinlet
1.5 /d ) (Khalil
H
(1 − )2 (1 − )
FLS = − 150 2 2 + 1.75 |u| u, (66) et al., 1975). The term dH denotes the hydraulic diameter of
 de de random packing (Bird et al., 2002), which can be calculated by
dH = 4∞ /a(1 − ∞ ).
where de is the equivalent diameter of random packing, which
The boundary condition for temperature variance t 2 is
is defined by
taken from the work of Tavoularis and Corrsin (1981a,b) and
6(1 − ∞ ) Ferchichi and Tavoularis (2002). After averaging their results,
de = . (67) we get the following equation:
a

4.3.5. The sink term MC in C̄ equation

2
tinlet = (0.082
T )2 . (70)
The rate of MEA consumed per unit volume for absorbing As the change of temperature due to the chemical absorption
CO2 from the gas phase, MC, can be calculated according to at the inlet is infinitesimal or very small, the
T is set to be
the chemical reaction (2) as follows: 0.1 K for the convenience of computation.
M There are no existing experimental measurements or empir-
MC = − × 61 × 2. (68) ical correlations for the inlet condition of the concentration
44
variance c2 . However, by the analogy between heat and mass
4.3.6. The heat effect Q transfer, we may assume that
The temperature of liquid phase will be increased due to the
heat of absorption and reaction, and as a result, the parameters
2
cinlet = (0.082Cinlet )2 . (71)
such as H e, k2 and DCO2 ,L as well as the absorption rate of
The inlet condition for
c is a complicated problem and could
CO2 will be changed. Therefore, the increase of temperature in
be obtained by using the basic relationship of turbulent mass
liquid must be properly evaluated. The total heat effect Q can
transfer diffusivity Dt and the time scale ratio R :
be calculated as follows:
k/

Q=
M
(HA + HR ), (69) R ≡ . (72)
MCO2 c2 /
c

where HA is the heat of physical absorption, HA = 1.9924 × Combining Eqs. (10) and (56), we get
107 J kmol−1 CO2 absorbed (Danckwerts, 1970), and HR de-  2
k2 1
notes the heat of chemical reaction, HR =8.4443×107 J kmol−1 R = Cc0 . (73)
CO2 (Kohl and Riesenfeld, 1985).
Dt
6518 G.B. Liu et al. / Chemical Engineering Science 61 (2006) 6511 – 6529

Different values of the ratio R were reported in the literatures. The simulation was done under the condition of steady two-
According to the viewpoint of Launder (1976), it may not be dimensional flow of axial and radial symmetry. The well-known
a universal constant and may strongly depend on the nature SIMPLEC algorithm is used to solve the pressure–velocity cou-
of the flow field. As an approximation, we may let R to be pling problem in the momentum equations.
a fixed value only at the inlet boundary, and such value can The grid arrangement for the simulation of pilot-scale col-
be determined from the experimental dispersion coefficient ob- umn of 6.55 m height and 0.05 m radius is as follows: there
tained by using inert tracer technique, which is applicable to are 1310 nodes uniformly distributed along the column height
the “end conditions” of C̄ = 0. The diffusivity Dt at C̄ = 0 for a and 80 nodes non-uniformly distributed along the radial direc-
column randomly packed with 1/2 (∼ 1.27 cm) ceramic Berl tion with high grid resolution at the near wall region, so the
saddles was measured by using the inert tracer technique by total numbers are about 104 800 quadrilateral cells. While the
Sater and Levenspiel (1966), and correlated the “coefficient of grid arrangement for simulating the commercial-scale column
dispersion” of liquid phase , i.e. equivalent to the Dt , by the of 14.1 m height and 0.95 m radius, there are 1000 nodes uni-
following equation: formly distributed along the column height and 75 nodes non-
   0.703±0.238 uniformly distributed along the radial direction with higher grid
Udp −3 dp L resolution at the near wall region, so the total numbers are
= 7.58 × 10 , (74)
Dt  about 75000 quadrilateral cells. The one-order upwind spatial
discretization scheme was used for all differential equations.
where U is the liquid superficial velocity. Later, Michell and
Furzer (1972) correlated the dispersion coefficient according to
5. Comparison between simulated results and
their own and some published data in the following form:
experimental data
 
−0.32
|u|dp uinter dp  0.7 dp3 g2
= 1.00 . (75) 5.1. Comparison between simulated and experimental results
Dt  2 for pilot-scale column
According to Eqs. (74) and (75), the order of magnitude for
Twelve sets of experimental data for chemical absorption of
Dt is about 10−3 . By taking their average value of Dt , and
CO2 from mixture of CO2 and air by aqueous MEA solutions in
substituting into Eq. (73), the value of R was found to be about
a pilot-scale column were reported by Tontiwachwuthikul et al.
0.4. Therefore, the boundary condition at the inlet for
c can
(1992), including the variation of radial averaged carbonation
be given as follows:
ration ( = CO2 loading), the temperature in the liquid phase and
 

inlet the radial averaged CO2 concentration in the gas phase along

c,inlet = 0.4 c2 . (76) the column height. In this paper, only the simulated results on
kinlet inlet
experiments T13, T14, T17, and T22 are used as examples for
By the analogy between heat and mass transfer, the boundary comparison.
condition for dissipation rate
t of temperature variance is set In Figs. 2–5, the square and circle symbols represent the
to be experimental data, and the solid lines indicate the simulated
 

inlet 2 results which are obtained after radial average. As seen from

t,inlet = 0.4 t . (77) the figures, the simulated and experimental results on the CO2
kinlet inlet
loading in liquid phase and the CO2 volume percentage in gas
The specific values calculated from the foregoing relation- phase are in satisfactory agreement.
ships for the inlet boundary condition of the present simulation It is interesting to compare the simulated results between
are listed in Appendix A. using the present model without knowing the coefficient of dis-
(2) Outflow: The liquid flow at the bottom exit of the column persion and using the conventional model with published ex-
is considered to be close to the fully developed condition, so perimental coefficient. The conventional simulated model used
the “outflow” boundary condition of zero normal gradient for is one-dimensional with axial mixing, and the axial dispersion
all flow variables except pressure is chosen under the platform coefficient is calculated by Eq. (74). Take run T17 as an ex-
of the software FLUENT 6.1. ample, the simulated results by the conventional model on gas
(3) Axis: Under the assumption that all variables are axially CO2 concentration and liquid CO2 loading profiles are shown
symmetrical, we have j /jr = 0 at r = 0. in Fig. 4(a) by the dash dot lines along the axial direction. It
(4) Wall: The no-slip condition is applied to the wall, and should be pointed out that the column is taller than it is needed
the flow behavior in the near wall region is approximated by for the separation concerned as seen on the axial concentration
using the “standard wall functions”. The zero flux condition is profile, in which the change of liquid phase concentration is
applied for other variables. becoming very small in the upper part of the column. It means
that the separation is approaching to the equilibrium state at
4.5. Numerical procedure some distance from the column bottom and thus results almost
the same concentration regardless what model is being used
The model equations were solved numerically by using the for the simulation. However, the difference is clearly seen at
commercial software FLUENT 6.1 with finite volume method. the lower part of the column, as the simulation by using the
 G.B. Liu et al. / Chemical Engineering Science 61 (2006) 6511 – 6529 6519

Fig. 3. Predicted (solid lines) and experimental (points) results for Run
Fig. 2. Predicted (solid lines) and experimental (points) results for Run T14: (a) gas CO2 concentration and liquid CO2 loading profiles; (b) liquid
T13: (a) gas CO2 concentration and liquid CO2 loading profiles; (b) liquid temperature and enhancement factor.
temperature and enhancement factor.

present model give better prediction than the conventional. In scale column were reported by Pintola et al. (1993), including
order to display clearly the difference of the simulated results, the CO2 volume percentage in the gas phase and the CO2 load-
a table for comparison is given in Appendix B, in which the ex- ing with temperature in the liquid phase. In this paper, only the
perimental measurements, the predictions by the present model simulated results on experimental Run 115 was presented as an
and the conventional one for run T17 are listed. It is seen that example for comparison.
the present model give better prediction. As shown in Fig. 6, the simulated top and bottom concentra-
As shown in Figs. 2(b)–5(b), the predicted temperature pro- tions by using the proposed model are closely checked by the
files along the column show somewhat higher than the exper- reported measurement. Furthermore, our simulation is better
imental measurement, especially at the column bottom. The than that given by Pintola et al. (1993), as shown in Appendix C.
discrepancy may be due to the following reasons: firstly, the For Run 115, the typical CO2 volume percentage in gas
present modeling neglects the heat transfer between gas and phase, the CO2 loading, the temperature of liquid phase and
liquid phases, that is the cooling of descending liquid by the the enhancement factor along the column are shown in Fig. 6.
entering gas; secondly, the evaporation of solvent water in liq- It is seen from the figure that 99% of the CO2 was absorbed
uid phase is also neglected, so that the liquid temperature is by aqueous MEA solutions in the bottom two sections of the
overestimated; thirdly, the assumptions of adiabatic operation column, and only little CO2 was removed in the top section.
and neglecting the heat transfer between the solid packing and Figs. 7–10 show the radial variation of the CO2 concentration
the fluid are also the causes of producing error. in gas phase, the CO2 loading, the liquid temperature and the
free MEA concentration of liquid phase at different height of
5.2. Comparison between simulated and measured results for the column.
industrial-scale column As seen from Figs. 7–9, the CO2 concentration in gas phase,
the CO2 loading and the temperature of liquid phase are de-
Fifteen sets of data for chemical absorption of CO2 from creased from column center to column wall at different height
natural gas mixture by aqueous MEA solutions in an industrial- of the column. It could be explained that the flow of liquid
6520 G.B. Liu et al. / Chemical Engineering Science 61 (2006) 6511 – 6529

Fig. 4. Comparisons between the predicted values by present model (solid

lines) and by one-dimensional with axial mixing model (dash dot lines) with Fig. 5. Predicted (solid lines) and experimental (points) results for Run
experimental (points) results for Run T17: (a) gas CO2 concentration and T22: (a) gas CO2 concentration and liquid CO2 loading profiles; (b) liquid
liquid CO2 loading profiles; (b) liquid temperature and enhancement factor. temperature and enhancement factor.

5.4. Turbulent mass transfer diffusivity Dt

phase is slow down near the wall in the randomly packed
The turbulent mass transfer diffusivity or dispersion coeffi-
column, resulting in worse contact with the gas phase, and
cient Dt is often neglected or assumed to be constant in the
consequently less CO2 to be absorbed, producing less exother-
packed column design. However, according to the calculated
mic heat of solution and reaction. It may be also noted from
Dt by using the proposed model, it varies from top to bottom
Fig. 10 that, at a fixed axial position, the radial difference of
and from center to wall of the columns as shown in Figs. 13(a),
free MEA concentration in the liquid phase is larger at the bot-
14 (a) and 15 (a). Furthermore, the calculated Dt for the in-
tom of the column than at the top.
dustrial column are very closer to the reported experimental
Fig. 11 shows the axial distribution of the CO2 concentration
measurements. From Fig. 14(a), the turbulent diffusivity Dt ob-
in the gas phase, the CO2 loading, the free MEA molar con-
tained by the present model is ranging from 6.0 × 10−4 m2 /s to
centration and the liquid temperature. It can be seen that most
1.5 × 10−3 m2 /s, while the values calculated from the empiri-
absorptions are taken place at the bottom part of the column,
cal correlations obtained by using inert tracer technique with no
and the top part has a little effect on the CO2 removal.
mass transfer by Sater and Levenspiel (1966) and by Michell
and Furzer (1972) are 0.001011 and 0.00116 m2 /s, respectively.
5.3. Liquid velocity profile Although they are practically in the same order of magnitude,
the difference in values between them demonstrates that the Dt
Due to the non-uniform packing structure and higher poros- is changed somewhat due to the process effect of mass transfer.
ity near the wall region, the fluid flow deviates from the plug The variations of Dt with radius at different axial positions
flow. As seen from Fig. 12, in both the pilot-scale and the are presented in Figs. 13(a) and 14(a). Generally speaking, Dt
industrial-scale columns, the “wall flow” effect is clearly seen, is found to be substantially constant around the center region of
and the flow behaves relatively uniform only about 2dp apart the packed bed, and increased to a maximum near the column
from the wall. This phenomenon has been confirmed by many wall, and then decreased sharply to the wall surface for both the
investigators (Liu, 2001; Giese et al., 1998). pilot-scale column and the industrial-scale column. Such phe-
 G.B. Liu et al. / Chemical Engineering Science 61 (2006) 6511 – 6529 6521

Fig. 8. The CO2 loading profiles along the radial direction at different height.

Fig. 6. Predicted (solid lines) and experimental (points) results for Run
115: (a) gas CO2 concentration and liquid CO2 loading profiles; (b) liquid
temperature and enhancement factor.

Fig. 9. The temperature profiles of liquid along the radial direction at different
height.

Fig. 7. Gas CO2 concentration profiles along the radial direction at different
height.

nomenon is consistent with the experimental results by Fahien

and Smith (1955) and Dorweiler and Fahien (1959) on using
gas phase tracer technique. The uneven Dt distribution is re-
lated to the mal-distributed fields of flow and concentration as Fig. 10. The free MEA concentration profiles along the radial direction at
shown in Figs. 11(c) and 12. different height.
6522 G.B. Liu et al. / Chemical Engineering Science 61 (2006) 6511 – 6529

Fig. 11. Axial distributions of: (a) CO2 concentration; (b) CO2 loading; (c) free MEA concentration; and (d) liquid temperature for Run 115.

The predicted Dt along the column height varies little for the gas phase. The axial distribution of Dt for the industrial column
pilot-scale column as shown in Fig. 13(a), which results from is shown in Fig. 15(a).
the lower liquid Reynolds number and higher volume fraction
of CO2 in gas phase. While for the industrial column, the Dt is 5.5. Analogy between turbulent mass transfer and heat transfer
gradually decreased as approaching to the bottom of the column
as shown in Fig. 14(a), which is probably due to the higher In applying the present model, both turbulent mass trans-
liquid Reynolds number and lower volume fraction of CO2 in fer diffusivity Dt and turbulent heat transfer diffusivity t can
 G.B. Liu et al. / Chemical Engineering Science 61 (2006) 6511 – 6529 6523

Fig. 12. Relative axial velocity profile of the liquid along the radial direction for Run T22 and Run 115.

be solved and show in Figs. 13–15. From these figures, It 5.7. Study on the influence of grid number on the accuracy of
can be clearly seen the similarity between them in spite of numerical simulation
having a little difference. For example, the volume average
values of Dt and t calculated for Run 115 of the industrial Generally, the number of grid to be used in simulation has
column are 0.001049 and 0.001107 m2 /s, respectively, while considerable effect on the numerical result. Regarding this as-
those in the Run T22 of the pilot-scale column are 0.001144 and pect, it is necessary to test whether the use of 75 grid points in
0.001169 m2 /s; the approximate equality of Dt and t demon- radial simulation is sufficient for the industrial scale absorption
strates the analogy between turbulent mass transfer and heat column with 1.9 m in diameter. Taking Run 115 of the large
transfer. column concerned, it was simulated with 75 and 225 points
separately in the radial direction, and using the second order
upwind discretisation scheme for the differential equations in-
5.6. Distribution of radial average enhancement factor E stead of first order upwind scheme. The simulated radial gas
along the column CO2 concentration profiles with 75 and 225 radial grid points
are given in Fig. 16.
The enhancement factor E is an important parameter for It can be seen from Fig. 16 that the difference of the simu-
chemical absorption, and its accurate estimation is essential for lated results is small. It could be understood that, the numerical
proper modeling the mass transfer process. The enhancement accuracy of the present simulation with 75 radial grid points is
factor is influenced by many aspects, such as the physical prop- sufficient in practical sense, and it has very weak improvement
erties of liquid and gas, the MEA concentration in liquid phase, on the increasing number of grids beyond 75.
the CO2 concentration in gas phase, the flow field, the reaction However, there is no doubt that the numerical accuracy could
rate and the others; therefore it is a varying parameter along be further improved by precise simulation with very large num-
the column as shown in Figs. 2 (b)–6 (b). ber of grid points.
6524 G.B. Liu et al. / Chemical Engineering Science 61 (2006) 6511 – 6529

Fig. 13. Turbulent diffusivity profiles along radial direction for Run T22: (a) turbulent mass transfer diffusivity profiles; and (b) turbulent heat transfer diffusivity
profiles.

Fig. 14. Turbulent diffusivity along radial direction for Run 115: (a) turbulent mass transfer diffusivity profiles; and (b) turbulent heat transfer diffusivity profiles.

Fig. 15. Distribution of turbulent diffusivity for Run 115: (a) turbulent mass transfer diffusivity; and (b) turbulent heat transfer diffusivity.
 G.B. Liu et al. / Chemical Engineering Science 61 (2006) 6511 – 6529 6525

aeff effective area for mass transfer between

the gas phase and liquid phase, 1/m
aw wetted surface area, 1/m
c2 concentration variance, dimensionless
cp liquid phase specific heat, J/kg/K
c a constant depending on the ratio of the
particle size to column size, dimension-
less
C̄ average concentration of mass fraction,
dimensionless
C , c1 , c2 model parameters in k–
model equa-
tions, dimensionless
Cc0 , Cc1 , Cc2 , Cc3 model parameters in c2 .
c model equa-
tions, dimensionless
Ct0 , Ct1 , Ct2 , Ct3 model parameters in t 2 .
t equations,
Fig. 16. The simulated gas CO2 concentration profiles with 75 and 225 grid dimensionless
points in the radial direction at x = 13.0 m. [CO2 ], [MEA] molar concentration of CO2 and MEA
in solution, respectively, kmol/m3
6. Conclusions
de equivalent diameter of random pack-
The following conclusions can be drawn from the present ing, m
work: dH hydraulic diameter of random pack-
ing, m
(1) A complex model of computational mass transfer is pro- dp nominal diameter of the packed parti-
posed for simulating the concentration, temperature and cle, m
velocity distributions for a chemical absorption process D molecular diffusivity of MEA molecu-
with heat effect. The proposed model is applied to a pilot- lar in aqueous MEA solutions, m2 /s
scale packed column undertaking the chemical absorption DCO2 ,L molecular diffusivity of CO2 in aque-
of CO2 from air by MEA aqueous solutions, and to an ous MEA solutions,m2 /s
industrial-scale packed column for the chemical absorption (DCO2 )w , (DN2 O )w diffusivity of CO2 and N2 O in the pure
of CO2 from natural gas mixture by aqueous MEA solu- water, respectively, m2 /s
tion. The simulated results are in satisfactory agreement Deff effective diffusivity of MEA, m2 /s
with the experimental data reported by Tontiwachwuthikul DG molecular diffusivity of CO2 in gas
et al. (1992) and by Pintola et al. (1993). phase, m2 /s
(2) The proposed model is able to predict the turbulent mass DN2 O,L diffusivity of N2 O in aqueous MEA so-
transfer diffusivity Dt , which shows varying along the axial lution, m2 /s
and radial directions. Dt turbulent diffusivity for mass transfer,
(3) The computed profiles of Dt for mass transfer and t for m2 /s
heat transfer are found to be very similar, and the calculated E enhancement factor, dimensionless
volume average values of Dt and t are nearly equal, which FLG interface drag force between gas phase
demonstrates the analogy between turbulent mass and heat and liquid phase, N/m3
transfer. FLS flow resistance created by the randomly
(4) The enhancement factor of chemical absorption predicted packing, N/m3
from the proposed model varies significantly along the Fpd dry packing factor, 1/m
column. g acceleration due to gravity, m/s2
(5) The proposed model is useful for simulating chemical ab- G gas phase flow rate per unit cross-
sorption and relevant process with heat effect without re- section area, kg/m2 /s
lying on the experimental measurement of turbulent mass Gf gas loading factor, kg/m2 /s
transfer and heat transfer diffusivities, especially for cases (Ga)L Gallileo number of the liquid phase, di-
where the empirical turbulent Schmidt number or the dis- mensionless
persion coefficients is not available. h volume fraction of liquid phase based
on pore space, dimensionless
HA physical absorption heat of mol CO2
Notation absorbed, J/kmol
Hop operating holdup, dimensionless
a surface area per unit volume of packed HR chemical reaction heat of mol CO2 ab-
bed, 1/m sorbed, J/kmol
6526 G.B. Liu et al. / Chemical Engineering Science 61 (2006) 6511 – 6529

Hs static holdup, dimensionless t2 temperature variance , dimensionless

Ht total liquid holdup, dimensionless T liquid temperature, K
He Henry’s constant for CO2 in MEA so- u interstitial velocity vector of liquid, m/s
lutions, kmol/m3 /kPa uG gas velocity vector, m/s
(H eCO2 )w ,(H eN2 O )w , Henry’s constant for CO2 and N2 O in uslip slip velocity vector between gas phase
water, respectively, kmol/m3 /kPa and liquid phase, m/s
H eN2 O Henry’s constant for N2 O in aqueous U liquid superficial velocity, m/s
MEA solutions, kmol/m3 /kPa vMEA , vH2 O volume fraction of water and MEA in
(H eN2 O )MEA Henry’s constant for N2 O in pure MEA aqueous MEA solutions, respectively,
liquid, kmol/m3 /kPa dimensionless
k2 second-order reaction rate constant, xH2 O , xMEA molar fraction of water and MEA in
m3 /kmol/s aqueous MEA solutions, respectively,
kG gas phase mass transfer coefficient, dimensionless
kmol/m/s/kPa XCO2 molar concentration of CO2 in the liq-
kL liquid phase mass transfer coefficient uid bulk, kmol/m3
without chemical reaction, m/s Xi,CO2 molar concentration of CO2 at inter-
kR,L liquid phase mass transfer coefficient face, kmol/m3
with chemical reaction, m/s XMEA molar concentration of MEA in liquid
K turbulent kinetic energy, m2 /s2 phase, kmol/m3
KG overall coefficient of gas phase, yCO2 volume fraction of CO2 in gas
kmol/m2 /s1 /kPa phase,dimensionless
L liquid flow rate per unit cross-section
area, kg/m2 /s Greek letters
Lf liquid loading factor, kg/m2 /s
M CO2 quantity absorbed by the aqueous , eff , t molecular, turbulent and effective ther-
solution per unit volume and unit time, mal diffusivities, respectively, m2 /s
kg/m3 /s  porosity distribution of the random
MCO2 molecular weight of CO2 , kg/kmol packing bed along the radial direction,
MC sink term of MEA conservation equa- dimensionless
tion, kg/m3 /s ∞ porosity in an unbounded packing, di-
p1 , p2 constants, dimensionless mensionless
pd period of oscillation normalized by the
turbulent dissipation rate, m2 s−3
nominal particle size, dimensionless
c turbulent dissipation rate of concentra-
pe,CO2 partial pressure of CO2 in equilibrium tion fluctuation, 1/s
with solutions, kPa
t turbulent dissipation rate of tempera-
pt total pressure of gas phase, kPa ture fluctuation, 1/s

pd dry-bed pressure drop per meter pack- , t , eff , liquid molecular, turbulent and effec-
ing, N/m3 tive viscosity, respectively, kg/m/s

pL wet-bed pressure drop per meter pack- G gas phase viscosity, kg/m/s
ing, N/m3  liquid density, kg/m3

pt total pressure drop per meter packing, G gas phase density, kg/m3
N/m3  surface tension of aqueous MEA solu-
PCO2 partial pressure of CO2 in main body tions, dynes/cm, or N/m
of gas, kPa ct critical surface tension of the pack-
Q thermal source term of temperature ing material, N/m, for ceramic, ct =
equation, J/m3 /s 0.061 N/m, for stainless steel, ct =
rc carbonation ratio (CO2 loading) of the 0.071 N/m
aqueous MEA solutions, mol CO2 / mol c , 
c model parameters in c2 .
c model equa-
MEA tions, dimensionless
R position in radial direction, m H2 O pure water surface tension, dynes/cm
R radius of the column, m k ,
model parameters in k–
model equa-
RG gas universal constant, kJ/kmol/K tions, dimensionless
(Re)L Reynolds number of liquid phase, di- MEA pure MEA surface tension, dynes/cm
mensionless t , 
t model parameters in t 2 .
t model equa-
Rc the rate of reaction, kmol/m3 /s tions, dimensionless
R velocity to concentration time scale ra- ℵMEA,H2 O excess Henry’s quantity, kmol/m3 /kPa
tio, dimensionless variable, dimensionless
 G.B. Liu et al. / Chemical Engineering Science 61 (2006) 6511 – 6529 6527

Table 1
The specific values for the inlet boundary condition of the present simulation

u (m/s) v (m/s) h C T (K) k (m/s)2 ε(m2 /s3 ) c2 εc (s−1 ) t 2 (K 2 ) ε(s−1 )

T13 0.0387 0 0.1563 0.121878 292.15 4.492E − 6 6.106E − 8 9.988E − 5 4.073E − 7 6.724E − 5 2.742E − 7
T14 0.0387 0 0.1563 0.093115 292.15 4.492E − 6 6.106E − 8 5.830E − 5 2.377E − 7 6.724E − 5 2.742E − 7
T17 0.0387 0 0.1563 0.121805 293.15 4.492E − 6 6.106E − 8 9.976E − 5 4.068E − 7 6.724E − 5 2.742E − 7
T22 0.0304 0 0.1399 0.182817 292.15 2.777E − 6 2.967E − 8 2.247E − 4 7.204E − 7 6.724E − 5 2.742E − 7
Run 115 0.1372 0 0.0319 0.219380 312.95 5.651E − 5 4.351E − 7 3.236E − 4 9.967E − 7 6.724E − 5 2.071E − 7

Table 2
Comparison between experimental, two and one dimensional simulation results for Run T17

T17 CO2 concentration in gas phase (%) CO2 loading in liquid phase

Exp. Two dim. One dim. Exp. Two dim. One dim.

x = 0.0 m 0.0 0.0265 0.0389 0.237 0.237 0.237

x = 1.05 m 0.0 0.1045 0.1101 (0.237) 0.238 0.238
x = 2.15 m 0.8 0.4596 0.3288 0.243 0.241 0.240
x = 3.25 m 2.0 1.4343 0.9797 0.256 0.252 0.247
x = 4.35 m 5.3 4.1456 2.8740 0.296 0.282 0.267
x = 5.45 m 10.2 9.7249 7.7675 0.350 0.347 0.324
x = 6.55 m 15.6 15.60 15.60 0.428 0.427 0.427

Table 3 Acknowledgments
Comparison of CO2 concentration in gas phase exit between the results of
experimental, by the proposed model and by Pintola et al. (1993)
The authors acknowledge the financial support by the Na-
Run Exp. CO2 Simu. CO2 (proposed) Simu. CO2 (Pintola et al. tional Natural Science Foundation of China (Contract no.
No. (ppm) (ppm) 1993) (ppm) 20136010) and the assistance from the staff in the State Key
101 8.0 9.1 65.2 Laboratories for Chemical Engineering (Tianjin University).
102 11.5 2.4 17.5
103 14.5 27.5 13.0 Appendix A
104 8.7 10.0 30.6
105 9.7 4.7 19.1
106 12.0 2.1 13.1 The specific values for the inlet boundary condition of the
107 6.5 8.1 39.1 present simulation are given in Table 1.
108 15.0 7.4 45.3
109 8.7 5.0 45.4 Appendix B
110 16.0 8.0 55.5
111 13.0 4.8 37.4
112 5.3 9.1 54.7 Comparison between experimental, two and one dimen-
113 28.0 9.3 60.4 sional simulation results for Run T17 are given in Table 2.
114 74.0 2.2 41.5
115 4.8 8.5 41.5
Appendix C

Comparison of CO2 concentration in gas phase exit be-

tween the results of experimental, by the proposed model and
by Pintola et al. (1993) are given in Table 3.

Subscripts References
eff effective, dimensionless Aboudheir, A.A., Tontiwachwuthikula, P., Chakmab, A., Idema, R., 2003.
G gas, dimensionless Kinetics of the reactive absorption of carbon dioxide in high CO2 -loaded
I interface, dimensionless concentrated aqueous monoethanolamine solutions. Chemical Engineering
L liquid, dimensionless Science 58, 5195–5210.
Astarita, G., Marrucci, G., Gioia, F., 1964. The influence of carbonation ratio
R reaction, dimensionless
and total amine concentration on carbon dioxide absorption in aqueous
slip slip, dimensionless monoethanolamine solutions. Chemical Engineering Science 19, 95–103.
S solid, dimensionless Bey, O., Eigenberger, G., 1997. Fluid flow through catalyst filled tubes.
 time scale, dimensionless Chemical Engineering Science 52, 1365–1376.
6528 G.B. Liu et al. / Chemical Engineering Science 61 (2006) 6511 – 6529

Bird, R.F., Stewart, W.E., Lightfoot, E.N., 2002. Transport Phenomena, second Pandya, J.D., 1983. Adiabatic Gas-Absorption and Stripping with Chemical-
ed. Wiley Inc., New York. Reaction in Packed Towers. Chemical Engineering Communications 19,
Danckwerts, P.V., 1970. Gas–Liquid Reactions. McGraw-Hill, New York. 343–361.
Danckwerts, P.V., 1979. Reaction of CO2 with Ethanolamines. Chemical Perry, R.H., Green, D.W., 2001. Perry’s Chemical Engineers’ Handbook,
Engineering Science 34, 443–446. seventh ed. McGraw-Hill, New York.
de Klerk, A., 2003. Voidage variation in packed beds at small column to Pintola, T., Tontiwachwuthikul, P., Meisen, A., 1993. Simulation of pilot plant
particle diameter ratio. A.I.Ch.E. Journal 49, 2022–2029. and industrial CO2 MEA absorbers. Gas Separation and Purification 7,
de Lemos, M.J.S., Mesquita, M.S., 2003. Turbulent mass transport in saturated 47–52.
rigid porous media. International Communications in Heat and Mass Pohorecki, R., Moniuk, W., 1988. Kinetics of reaction between carbon-dioxide
Transfer 30, 105–113. and hydroxyl ions in aqueous-electrolyte solutions. Chemical Engineering
Dorweiler, V.P., Fahien, R.W., 1959. Mass transfer at low flow rates in a Science 43, 1677–1684.
packed column. A.I.Ch.E. Journal 5, 139–144. Poling, B.E., Prausnitz, J.M., O’connell, J.P., 2001. The Properties of Gases
Engel, V., Stichmair, J., Geipel, W., 1997. A new model to predict liquid and Liquids, fifth ed. McGraw-Hill, New York.
holdup in packed columns—using data based on capacitance measurement Robbins, L.A., 1991. Improve pressure-drop prediction with a new correlation.
techniques. Institute Chemical Engineering Symposium Series, No. 142, Chemical Engineering Progress 87, 87–90.
or Part 2, pp. 939–947. Roblee, L.H.S., Baird, R.M., Tierney, J.W., 1958. Radial porosity variations
Ergun, S., 1952. Fluid flow through packed columns. Chemical Engineering in packed beds. A.I.Ch.E. Journal 4, 460–464.
Progress 48, 89–94. Sada, E., Kumazawa, H., Butt, M.A., 1976. Chemical absorption kinetics
Fahien, R.W., Smith, J.M., 1955. Mass transfer in packed beds. A.I.Ch.E. over a wide-range of contact time - absorption of carbon-dioxide into
Journal 1, 28–37. aqueous-solutions of monoethanolamine. A.I.Ch.E. Journal 22, 196–198.
Ferchichi, M., Tavoularis, S., 2002. Scalar probability density function and Sater, V.E., Levenspiel, O., 1966. Two-phase flow in packed beds. Industrial
fine structure in uniformly sheared turbulence. Journal of Fluid Mechanics and Engineering Chemistry Research Fundamentals 5, 86–92.
461, 155–182. Shulman, H.L., Ullrich, C.F., Wells, N., 1955. Performance of packed columns.
Geankoplis, J.C., 2003. Transport Processes and Separation Process Principles I. Total, static, and operating holdups. A.I.Ch.E. Journal 1, 247–253.
(Includes Unit Operations), fourth ed. Prentice-Hall PTR, Upper Saddle Snijder, E.D., Teriele, M.J.M., Versteeg, G.F., Vanswaaij, W.P.M., 1993.
River, NJ. Diffusion-coefficients of several aqueous alkanolamine solutions. Journal
Giese, M., Rottschafer, K., Vortmeyer, D., 1998. Measured and modeled of Chemical and Engineering Data 38, 475–480.
superficial flow profiles in packed beds with liquid flow. A.I.Ch.E. Journal Spiegel, L., Meier, W., 2003. Distillation columns with structured packings
44, 484–490. in the next decade. Trans IchemE 81 (A1), 39–47.
Gualito, J.J., Cerino, F.J., Cardenas, J.C., Rocha, J.A., 1997. Design method Sun, Z.M., Liu, B.T., Yuan, X.G., Liu, C.J., Yu, K.T., 2005. New turbulent
for distillation columns filled with metallic, ceramic, or plastic structured model for computational mass transfer and its application to a commercial-
packings. Industrial and Engineering Chemistry Research 36, 1747–1757. scale distillation column. Industrial and Engineering Chemistry Research
Hikita, H., Asai, S., Katsu, Y., Ikuno, S., 1979. Absorption of carbon-dioxide 44, 4427–4434.
into aqueous monoethanolamine solutions. A.I.Ch.E. Journal 25, 793–800. Tavoularis, S., Corrsin, S., 1981a. Experiments in nearly homogeneous
Hjertager, L.K., Hjertager, B.H., Solberg, T., 2002. CFD modelling of fast turbulent shear-flow with a uniform mean temperature-gradient. 2. The
chemical reactions in turbulent liquid flows. Computers and Chemical fine-structure. Journal of Fluid Mechanics 104, 349–367.
Engineering 26, 507–515. Tavoularis, S., Corrsin, S., 1981b. Experiments in nearly homogenous
Jiang, Y., Khadilkar, M.R., Al-Dahhan, M.H., Dudukovic, A.P., 2002a. CFD turbulent shear-flow with a uniform mean temperature-gradient. 1. Journal
of multiphase flow in packed-bed reactors: I. k-fluid modeling issues. of Fluid Mechanics 104, 311–347.
A.I.Ch.E. Journal 48, 701–715. Tontiwachwuthikul, P., Meisen, A., Lim, C.J., 1992. CO2 Absorption by
Jiang, Y., Khadilkar, M.R., Al-Dahhan, M.H., Dudukovic, M.P., 2002b. CFD NaOH, monoethanolamine and 2-amino-2-methyl-1-propanol solutions in
of multiphase flow in packed-bed reactors: II. Results and applications. a packed-column. Chemical Engineering Science 47, 381–390.
A.I.Ch.E. Journal 48, 716–730. Tsai, T., Ko, J., Wang, H., Lin, C., Li, M., 2000. Solubility of nitrous oxide
Khalil, E.E., Spalding, D.B., Whitelaw, J.H., 1975. Calculation of local flow in alkanolamine aqueous solutions. Journal of Chemical and Engineering
properties in 2-dimensional furnaces. International Journal of Heat and Data 45, 341–347.
Mass Transfer 18, 775–791. Vazquez, G., Alvarez, E., Navaza, J.M., Rendo, R., Romero, E., 1997. Surface
Ko, J., Tsai, T., Lin, C., Li, M., 2001. Diffusivity of nitrous oxide in aqueous tension of binary mixtures of water plus monoethanolamine and water plus
alkanolamine solutions. Journal of Chemical Engineering and Data 46, 2-amino-2-methyl-1-propanol and tertiary mixtures of these amines with
160–165. water from 25◦ C to 50◦ C. Journal of Chemical and Engineering Data 42,
Kohl, A.L., Riesenfeld, F.C., 1985. Gas Purification, fourth ed. Gulf Pub. 57–59.
Co., Book Division, Houston. Versteeg, G.F., Vanswaaij, W.P.M., 1988. On the kinetics between CO2 and
Launder, B.E., 1976. Heat and mass transport. Topics in Applied Physics alkanolamines both in aqueous and non-aqueous solutions. 1. Primary and
Turbulence. Springer, Berlin. secondary-amines. Chemical Engineering Science 43, 573–585.
Lerou, J.J., Froment, G.F., 1977. Velocity temperature and conversion Wang, Y.W., Xu, S., Otto, F.D., Mather, A.E., 1992. Solubility of N2O in
profiles in fixed-bed catalytic reactors. Chemical Engineering Science 32, alkanolamines and in mixed-solvents. Chemical Engineering Journal and
853–861. the Biochemical Engineering Journal 48, 31–40.
Liu, B.T., 2003. Study of a new mass transfer model of CFD and its application Weiland, R.H., Dingman, J.C., Cronin, D.B., Browning, G.J., 1998. Density
on distillation tray. Ph.D. Thesis, Tianjin University, Tianjin, China. and viscosity of some partially carbonated aqueous alkanolamine solutions
Liu, S.J., 2001. A continuum model for gas-liquid flow in packed towers. and their blends. Journal of Chemical and Engineering Data 43, 378–382.
Chemical Engineering Science 56, 5945–5953. Wellek, R.M., Brunson, R.J., Law, F.H., 1978. Enhancement factors for gas-
Michell, R.W., Furzer, I.A., 1972. Mixing in trickle flow through packed absorption with 2nd-order irreversible chemical-reaction. Canadian Journal
beds. The Chemical Engineering Journal 4, 53–63. of Chemical Engineering 56, 181–186.
Nagano, Y., Kim, C., 1988. A 2-equation model for heat-transport in wall Wen, X., Shu, Y., Nandakumar, K., Chuang, K.T., 2001. Predicting liquid
turbulent shear flows. Journal of Heat Transfer-Transactions of the ASME flow profile in randomly packed beds from computer simulation. A.I.Ch.E.
110, 583–589. Journal 47, 1770–1779.
Onda, K., Takeuchi, H., Okumoto, Y., 1968. Mass transfer coefficients between Yin, F.H., Sun, C.G., Afacan, A., Nandakumar, K., Chuang, K.T., 2000.
gas and liquid phases in packed columns. Journal of Chemical Engineering CFD modeling of mass-transfer processes in randomly packed distillation
of Japan 1, 56–62. columns. Industrial and Engineering Chemistry Research 39, 1369–1380.
 G.B. Liu et al. / Chemical Engineering Science 61 (2006) 6511 – 6529 6529

Yin, F.H., Afacan, A., Nandakumar, K., Chuang, K.T., 2002. CFD simulation Zhang, Z.T., 1986. Study on liquid flow characteristics and mass transfer
and experimental study of liquid dispersion in randomly packed metal Pall model in packed column. Ph.D. Thesis, Tianjin University, Tianjin, China.
rings. Chemical Engineering Research and Design 80, 135–144. Zhang, Z.T., Yu, K.T., 1988. Stochastic simulation of liquid flow distribution
Yuan, X.J., Li, F.S., Yu, K.T., 1989. Distribution of liquid phase in a large in packed column. Chinese Journal of Chemical Engineering 39, 162–169.
packed column. Chinese Journal of Chemical Engineering 40, 686–692. Ziolkowska, I., Ziolkowski, D., 1988. Fluid-flow inside packed-beds. Chemical
Zhang, Y.Sh., 2002. Turbulence. National Defense Industry Press, Beijing. Engineering and Processing 23, 137–164.