4346 Ind. Eng. Chem. Res.

2000, 39, 4346-4356

Absorption of Carbon Dioxide into Aqueous Blends of

Diethanolamine and Methyldiethanolamine
Edward B. Rinker, Sami S. Ashour, and Orville C. Sandall*
Department of Chemical Engineering, University of California, Santa Barbara, California 93106-5080

In this work, a comprehensive model is developed for the absorption of carbon dioxide into
aqueous mixtures of primary or secondary alkanolamines with tertiary alkanolamines. The
model, which is based on penetration theory, incorporates an extensive set of important reversible
reactions and takes into account the coupling between chemical equilibrium, mass transfer,
and chemical kinetics. The reaction between CO2 and the primary or secondary amine is modeled
according to the zwitterion mechanism. The key physicochemical properties that are needed for
the model are the CO2 physical solubility, the CO2 and amine diffusion coefficients, and the
reaction rate coefficients and equilibrium constants. Data for carbon dioxide absorption into
aqueous mixtures of diethanolamine and methyldiethanolamine are compared to model
predictions.

Introduction Chakravarty et al.1 suggested that, by mixing a

primary or secondary amine with a tertiary amine, the
Removal of acid gas impurities, such as CO2, is CO2 selectivity in the presence of H2S could be improved
important in natural gas processing. Natural gas, and regeneration costs minimized. These blended amine
depending on its source, can have varying concentra- solutions also offer the advantage of setting the selectiv-
tions of CO2. Some of the CO2 is often removed from ity of the solvent toward CO2 by judiciously mixing the
natural gas because, when present at high levels, it amines in varying proportions, which results in an
reduces the heating value of the gas, and because it is additional degree of freedom for achieving the desired
costly to pump this extra volume when it does not have separation for a given gas mixture. This approach could
any heating value. dramatically reduce capital and operating costs while
Aqueous solutions of alkanolamines are the most providing more flexibility in achieving specific purity
widely used solvents for removing CO2. The most com- requirements. Because of the need to exploit poorer
monly used alkanolamines are the primary amine quality crude and natural gas coupled with increasingly
monoethanolamine (MEA), the secondary amine dietha- strict environmental regulations, highly economical and
nolamine (DEA), and the tertiary amine methyldietha- selective acid gas treatment is more important today
nolamine (MDEA). Primary and secondary amines react than at any time in the past. As a result, there has been
rapidly with CO2 to form carbamates. By the addition a resurgence of interest in improved alkanolamine sol-
of a primary or secondary amine to a purely physical vents and particularly in aqueous blends of alkanol-
solvent such as water, the CO2 absorption capacity and amines.
rate is enhanced manyfold. However, because there is Design methods for acid-gas-treating processes em-
a relatively high heat of absorption associated with the ploying aqueous blends of alkanolamines vary widely
formation of carbamate ions, the cost of regenerating in their effectiveness at predicting process performance.
primary and secondary amines is high. Primary and Many acid-gas-treating processes are still designed by
secondary amines also have the disadvantage of requir- experience and heuristics, resulting in overdesign,
ing 2 mol of amine to react with 1 mol of CO2; thus, excessive energy consumption, and often failure to meet
their loadings are limited to 0.5 mol of CO2/mol of purity requirements entirely (Chakravarty et al.1).
amine. Tertiary amines lack the N-H bond required to Another common method uses equilibrium stage models
form the carbamate ion and therefore do not react modified by tray efficiencies. This method, however,
directly with CO2. However, in aqueous solutions, requires the use of existing plant data and lumps all
tertiary amines promote the hydrolysis of CO2 to form uncertainties about the finite reaction rates of the gases
bicarbonate and the protonated amine. Amine-promoted in the solvent into one parameter, the tray efficiency.
hydrolysis reactions are much slower than the direct Such a model cannot be predictive and will not capture
reactions of primary and secondary amines with CO2, the essential interplay of mass transfer, chemical kinet-
and therefore, the kinetic selectivity of tertiary amines ics, and chemical thermodynamics that occur in complex
toward CO2 is poor. However, the heat of reaction chemical solvents such as aqueous blends of alkanola-
associated with the formation of bicarbonate ions is mines. The third method of design is to develop models
much lower than that associated with carbamate forma- based on the chemistry and physics of the process, which
tion, and thus, the regeneration costs are lower for accounts for rates of mass transfer coupled with chemi-
tertiary amines than for primary and secondary amines. cal kinetics and thermodynamics. These models, while
Another advantage with tertiary amines is that the still requiring some experimental hydrodynamic infor-
stoichiometry is 1:1, which allows for very high equi- mation specific to different types of contacting devices,
librium CO2 loadings. are capable of predicting column performance, thus
minimizing the costs of design, equipment, and energy
* Corresponding author: [email protected]. consumption.
10.1021/ie990850r CCC: $19.00 © 2000 American Chemical Society
Published on Web 09/27/2000
 Ind. Eng. Chem. Res., Vol. 39, No. 11, 2000 4347

The objective of the work presented here is to develop K15

a comprehensive model for the absorption of CO2 into HCO3- + H3O+ 798 H2CO3 + H2O (15)
aqueous blends of tertiary and primary or secondary
amines. Experiments to test the model using methyldi- K16
ethanolamine (MDEA) as the tertiary amine and di- 2H2O 798 OH- + H3O+ (16)
ethanolamine (DEA) as the secondary amine are de-
scribed.
For MDEA, we have R(1) ) -CH3 and R(2) ) R(3) )
-CH2CH2OH, and for DEA, we have R(4) ) R(5) -CH2-
CH2OH. Ki, ki, and k-i are the equilibrium constant, the
Reaction Mechanism forward rate coefficient, and the reverse rate coefficient
for reaction i. Reactions 1-11 are considered to be
When CO2 is absorbed into an aqueous solution of a reversible with finite reaction rates. Whereas reactions
tertiary alkanolamine, R(1)R(2)R(3)N, and a primary or a 12-16 are considered to be reversible and instantaneous
secondary alkanolamine, R(4)R(5)NH, the following reac- with respect to mass transfer and at equilibrium, since
tions may occur: they involve only proton transfers.
Note that not all of the reaction equilibrium constants
are independent. Only eight equilibrium constants (K2,
K1,k1
CO2 + H2O 798 H2CO3 (1) K4, K5, K12, K13, K14, K15, and K16) are independent. The
remaining eight can be obtained by appropriate combi-
K2,k2 nations of the independent equilibrium constants.
CO2 + OH- 798 HCO- (2) The interaction between the protonated and unpro-
tonated amines according to the reaction
K3,k3
CO2 + R(1)R(2)R(3)N + H2O 798 K17
R(4)R(5)NH2+ + R(1)R(2)R(3)N 798
R(1)R(2)R(3)NH+ + HCO3- (3)
K4,k4,k-4
R(4)R(5)NH + R(1)R(2)R(3)NH+ (17)
(4) (5) (4) (5) + -
CO2 + R R NH 798 R R NH COO (4)
involves only a proton transfer and is considered to be
K5,k5,k-5 instantaneous and at equilibrium. Reaction 17 is im-
R(4)R(5)NH+COO- + R(4)R(5)NH 798 plicitly included in the reaction scheme above, as it can
be obtained by properly combining the instantaneous
R(4)R(5)NH2+ + R(4)R(5)NCOO- (5) equilibria reactions 12 and 13. Hence, we have K17 )
K6,k6,k-6
K13/K12.
R(4)R(5)NH+COO- + R(1)R(2)R(3)N 798 The proposed mechanism for the reaction between
CO2 and tertiary alkanolamines, R(1)R(2)R(3)N, indicates
R(1)R(2)R(3)NH+ + R(4)R(5)NCOO- (6) that they do not react directly with CO2. Instead,
K7,k7,k-7
tertiary alkanolamines act as bases that catalyze the
R(4)R(5)NH+COO- + H2O 798 hydration of CO2 (Donaldson and Nguyen,2 Haimour et
al.,3 Versteeg and van Swaaij,4 Littel et al.,5 Rinker et
H3O + R(4)R(5)NCOO- (7) al.6). In contrast, the proposed mechanism for the
reaction between CO2 and a primary or secondary
K8,k8,k-8 alkanolamine, R(4)R(5)NH, involves the formation of a
R(4)R(5)NH+COO- + OH- 798 zwitterion, R(4)R(5)NH + COO- (see reaction 4), followed
by the deprotonation of the zwitterion by a base to
H2O + R(4)R(5)NCOO- (8) produce carbamate, R(4)R(5)NCOO-, and protonated base
K9,k9,k-9 (see reactions 5-10) (Caplow,7 Danckwerts,8 Blauwhoff
R(4)R(5)NH+COO- + HCO3- 798 et al.,9 Versteeg and van Swaaij,10 Versteeg and
Oyevaar,11 Versteeg et al.,12 Glasscock et al.,13 Littel et
H2CO3 + R(4)R(5)NCOO- (9) al.,14 Rinker et al.15). Any base present in the solution
K10,k10,k-10
might contribute to the deprotonation of the zwitterion.
R(4)R(5)NH+COO- + CO32- 798 The contribution of each base would depend on its
concentration as well as its strength. Hence, the main
HCO3- + R(4)R(5)NCOO- (10) contribution to the deprotonation of the zwitterion in
an aqueous solution of a mixture of a primary or
K11,k11 secondary alkanolamine, R(4)R(5)NH, and a tertiary
R(4)R(5)NCOO- + H2O 798 alkanolamine, R(1)R(2)R(3)N, would come from R(4)R(5)NH,
R(1)R(2)R(3)N, and to a lesser extent OH- and H2O. There
R(4)R(5)NH + HCO3- (11)
are two limiting cases in the zwitterion mechanism.
K12 When the zwitterion formation reaction is rate-limiting,
R(1)R(2)R(3)NH+ + OH- 798 R(1)R(2)R(3)N + H2O (12) the reaction rate appears to be first-order in both the
amine and CO2 concentrations. In the case of mono-
K13 ethanolamine (MEA), a primary alkanolamine, the
R(4)R(5)NHH2+ + OH- 798 R(4)R(5)NH + H2O (13) formation of the zwitterion has been shown to be the
rate-determining step (Danckwerts,8 Sada et al.,16 Ver-
K14
HCO3- + OH- 798 CO32- + H2O (14) steeg and van Swaaij,10 Littel et al.14). On the other
hand, when the zwitterion deprotonation reactions are
4348 Ind. Eng. Chem. Res., Vol. 39, No. 11, 2000

rate-limiting, the overall reaction rate appears to be of All reactions at equilibrium

order 2 in the amine concentration. For the secondary (only independent equilibrium constants)
alkanolamines diethanolamine (DEA) and diisopropanol-
amine (DIPA), the rate-determining step is the depro- u04
tonation of the zwitterion (Littel et al.13). Several K2 ) (22)
u01 u05
authors have reported some rate coefficients for this
limiting case of the zwitterion mechanism for DEA and u07 u010
DIPA for a few temperatures (Blauwhoff et al.,9 Ver- K4K5K13K16 ) (23)
steeg and van Swaaij,16 Versteeg and Oyevaar,10 Glass- u01 u08
cock et al.,12 Rinker et al.14). Similarly, if neither
reaction in the zwitterion mechanism is rate-limiting, u02
the reaction rate exhibits a fractional order between 1 K12 ) (24)
and 2 with respect to the amine concentration. However, u03 u05
the rate expression is more complicated than the limit-
ing cases. Fractional orders are usually only observed u08
K13 ) (25)
for reactions between CO2 and secondary amines (Sada u09 u05
et al.,16 Versteeg and Oyevaar,11 Littel et al.14); however
fractional orders have also been observed for sterically u06
hindered primary amines such as 2-amino-2-methyl- K14 ) (26)
propanol (AMP) (Bosch et al.,17 Alper18). u04 u05

u011
Mathematical Model K15 ) (27)
u04 u07
For convenience, the concentrations of the chemical
species are renamed as follows: K16 ) u05 u07 (28)

u1 ) [CO2] u2 ) [R(1)R(2)R(3)N] We have 11 unknowns (u01, ... , u11

0
) and 11 nonlinear
+ -
(1) (2)
u3 ) [R R R NH ] (3)
u4 ) [HCO3 ] algebraic equations that we can solve for the liquid
bulk concentrations. We have found that Newton’s
u5 ) [OH-] u6 ) [CO32-]
method did not converge unless the initial guesses
u7 ) [H3O+] u8 ) [R(4)R(5)NH] for the liquid bulk concentrations were very close to
u9 ) [R(4)R(5)NH2+] u10 ) [R(4)R(5)NCOO-] the (unknown) solution. We, therefore, used the
Newton homotopy continuation method (Hanna and
u11 ) [H2CO3] u12 ) [H2O]
Sandall19), which exhibited better convergence be-
havior.
Liquid Bulk Concentrations of All Chemical The Partial Differential and Nonlinear Alge-
Species. The liquid bulk concentrations of all chemical braic Equations That Describe the Diffusion/
species can be estimated from the initial concentrations Reaction Processes. Higbie’s penetration model (Hig-
of R(1)R(2)R(3)N and R(4)R(5)NH; the initial CO2 loading bie,20 Danckwerts21) was used to set up the diffusion/
of the solution, L1; and the assumption that all reactions reaction partial differential equations that describe the
are at equilibrium. Because the concentration of water
absorption/desorption of CO2 into/from aqueous solu-
is much larger than the concentrations of all other
tions of tertiary amines, R(1)R(2)R(3)N, and primary or
chemical species, changes to its concentration over very
secondary amines, R(4)R(5)NH, in a laminar-liquid-jet
short contact times are negligible, and we assume that
absorber or a stirred-cell absorber. All reactions were
its concentration remains constant. Hence, we only need
to solve for the concentrations of the remaining 11 treated as reversible reactions. The first 11 reactions
chemical species. We have the following equations for have finite reaction rates, which are given by the
following reaction rate expressions, where Ri is the
the liquid bulk concentrations u01, ... , u11
0
:
reaction rate expression for reaction i:

Overall tertiary amine, R(1)R(2)R(3)N, balance k1

R1 ) -k1u1 + u (29)
u02 + u03 (1) (2) (3)
)[R R R N]initial (18) K1 11

Overall primary or secondary amine, R(4)R(5)NH, k2

balance R2 ) -k2u1u5 + u (30)
K2 4
u08 + u09 + u010 )[R(4)R(5)NH]initial (19) k3
R3 ) -k3u1u5 + uu (31)
Overall carbon (from CO2) balance K3 3 4

u01 + u04 + u06 + u010 + u011 + L1{[R(1)R(2)R(3)N]initial +

R4,...,10 )
[
-k4 u1u8 - (BA)u ]
10
(32)
[R R NH]initial} (20)
(4) (5)
1
1+
Electroneutrality balance B

u03 + u07 + u09 - u04 - u05 - 2u06 - u010 ) 0 (21) where

 Ind. Eng. Chem. Res., Vol. 39, No. 11, 2000 4349

A) ( )
k5 u 9
k-4 K4K5
+( ) ( )
k6 u3
k-4 K4K6
+
k7 u7
k-4 K4K7
+
Electroneutrality balance
∂u3 ∂u7 ∂u9 ∂u4 ∂u5 ∂u6 ∂u10

( ) ( ) ( )
k8 u12 k9 u11 k10 u4 + + - - -2 - )
∂t ∂t ∂t ∂t ∂t ∂t ∂t
+ + (33)
k-4 K4K8 k-4 K4K9 k-4 K4K10 ∂2u3 ∂2u7 ∂2 u 9 ∂2 u 4 ∂2u5
D3 2 + D7 2 + D9 2 - D4 2 - D5 2 -
∂x ∂x ∂x ∂x ∂x
and 2
∂ u6 2
∂ u10
2D6 2 - D10 2 (40)

( ) ( ) ( ) ( )
k5 k6 k7 k8 ∂x ∂x
B) u + u + u + u +
k-4 8 k-4 2 k-4 12 k-4 5 Carbamate, R(4)R(5)NCOO-, balance

( ) ( ) k9
u +
k-4 4
k10
u (34)
k-4 6
∂u10
∂t
∂2u10
) D10 2 + R11 - R4,...10
∂x
(41)

k11
R11 ) -k11u10 + uu (35) Instantaneous reactions assumed to be at equilibrium
K11 4 8
u2
K12 ) (42)
u3u5
Note that eq 32 was derived with the assumption of
a pseudo-steady-state approximation for the zwitterion u8
reaction intermediate, R(4)R(5)NH+COO-. The partial
K13 ) (43)
u9u5
differential equations that describe the diffusion/reac-
tion processes were combined so as to eliminate the very u6
large reaction rates for the instantaneous reactions 12-
K14 ) (44)
u4u5
16. Because these reactions are assumed to be at
equilibrium, their equilibrium constant expressions u11
were used to complete the equations that are required
K15 ) (45)
u4u7
to solve for the concentration profiles of all chemical
species. Furthermore, the diffusion coefficients of the K16 ) u5u7 (46)
ionic species were assumed to be equal. With this
assumption, the electrostatic potential gradient terms We have 11 partial differential/algebraic equations that
in the diffusion/reaction partial differential equations we can solve for the concentrations of the 11 chemical
for the ionic species can be neglected, while the elec- species, u1, ... , u11.
troneutrality of the solution is preserved. The more Initial Condition and Boundary Condition at x ) ∞.
rigorous approach of taking into account the electro- At t ) 0 (for all x ) 0) and at x ) 8 (for all t ) 0), the
static potential gradient terms with unequal diffusion concentrations of all chemical species are equal to their
coefficients for the ionic species requires much greater liquid bulk concentrations.
computational effort with essentially intangible effects
on the predicted rates of absorption. The following ui ) u0i , i ) 1, ... , 11 (47)
equations describe the diffusion/reaction processes:
Boundary Condition at Gas-Liquid Interface (x ) 0).
CO2 balance At x ) 0 (gas-liquid interface), the fluxes of the
nonvolatile chemical species are equal to zero, which
∂u1 ∂2u1 leads to the following equations:
) D1 2 + R1 + R2 + R3 + R4,...10 (36)
∂t ∂x ∂ui
) 0 at x ) 0, t > 0 (48)
Total carbon (from CO2) balance ∂x

∂u1∂u4 ∂u6 ∂u10 ∂u11 ∂2u1 ∂2u4 for all i except i ) 1 (CO2). For the volatile component,
+ + + ) D1 2 + D4 2 + CO2, the mass transfer rate in the gas near the interface
∂t ∂t ∂t ∂t ∂t ∂x ∂x is equal to the mass transfer rate in the liquid near the
∂2u6 ∂2u10 ∂2u11 interface.
D6 2 + D10 2 + D11 2 (37)
∂x ∂x ∂x
∂u1
Total tertiary amine, R(1)R(2)R(3)N, balance -D1 ) kg,1[P1 - H1u1(0,t)] at x ) 0, t > 0 (49)
∂x
∂u2 ∂u3 ∂2 u 2 ∂2 u 3 H1 is the physical equilibrium constant (Henry’s law
+ ) D2 2 + D 3 2 (38)
∂t ∂t ∂x ∂x constant) of CO2, which is defined as the interfacial
partial pressure of CO2 in the gas phase, P/1, divided by
Total primary or secondary amine, R(4)R(5)N, the interfacial concentration of CO2 in the liquid, u/1.
balance For the case of pure CO2 in the gas phase, the interfacial
∂u8 ∂u9 ∂u10 ∂2u8 ∂2 u 9 ∂2u10 partial pressure of CO2, P/1, is the same as the bulk
+ + ) D8 2 + D9 2 + D10 2 (39) partial pressure of CO2, P1, and there is not any mass
∂t ∂t ∂t ∂x ∂x ∂x transfer resistance in the gas-phase (kg,1 f ∞). Hence,
4350 Ind. Eng. Chem. Res., Vol. 39, No. 11, 2000

the boundary condition for CO2 at the gas-liquid 17265.4

interface reduces to log10(k1) ) 329.85 - 110.54 log10(T) - (55)
T
2895
P1 log10(k2) ) 13.635 - (56)
u1(0,t) ) u/1 ) at x ) 0, t > 0 (50) T
H1
Table 1 gives the values of the equilibrium constants
The differential equations are integrated from t ) 0 used in the model calculations, with the exception of
to t ) τ, the contact time. For the laminar-liquid-jet K15. The value of K15 at 298 K was taken to be 2 × 10-4
absorber, the contact time is given by m3/kmol (Cotton and Wilkinson31). K15 was then cor-
rected for temperature dependence according to the
following equation:
πd2l

[ ] ( )
τ) (51)
4Q K15(T) -∆H0 1 1
ln ) - (57)
K15(298.15) R T 298.15
where Q is the volumetric liquid flow rate, and d and l
are the diameter and length of the laminar-liquid jet, where the standard enthalpy change of reaction, ∆H0,
respectively. For Higbie’s penetration model, the liquid- is assumed to be independent of T and is approximated
phase mass transfer coefficient for physical absorption 0
of CO2 is defined as by its value at 298.15 K, ∆H298.15 (Smith and Van
0
Ness32). ∆H298.15 values were calculated from values
reported in the CRC Handbook of Chemistry and

x
0
D1 Physics (Lide33).
kl,1 )2 (52)
πτ
Experimental Apparatus and Procedure
The average rate of absorption of CO2 per unit interfa-
cial area is then computed from the following equation: The rates of absorption of CO2 into aqueous solutions
of DEA and MDEA were measured in a laminar-liquid-
jet absorber and a stirred-cell absorber. A schematic
D1 ∂u1
RA1 ) -
τ
∫0τ ∂x
(0,t) dt (53) drawing of the laminar-liquid-jet absorber is shown in
Figure 1. The laminar-liquid-jet absorber and its opera-
tion are described in detail by Rinker et al.15 The stirred-
The enhancement factor of CO2 is determined accord- cell absorber is shown schematically in Figure 2. The
ing to the following equation: absorption chamber is made of a 30.5 cm long, 10.1 cm
i.d. Pyrex cylinder and is enclosed in a constant-
RA1 temperature heating jacket constructed from a 31 cm
E1 ) (54) long, 24 cm i.d. Lucite cylinder. The ends of the Lucite
0
kl,1 (u/1- u01) cylinder are sealed with rubber O-rings between two
anodized aluminum flanges, and the glass cylinder is
where u/1 and u01 are the interfacial and bulk concen- sandwiched between two stainless steel flanges with the
trations of CO2 in the liquid, respectively. ends sealed by Teflon gaskets. Cooling or heating water
Method of Solution for the Partial Differential/ is supplied to the jacket and recycled to a constant-
Algebraic Equations. The method of lines was used temperature circulating bath. Two separate stainless
to transform each partial differential equation into a steel coils are placed in the heating/cooling jacket and
system of ordinary differential equations in t (Hanna are used to control the temperatures of the liquid and
and Sandall19). The systems of partial differential/ gas feeds. The liquid supply is introduced into the
algebraic equations were transformed into larger sys- chamber by a 0.635 cm o.d. stainless steel tube that can
tems of ordinary differential/algebraic equations, which slide in the vertical direction through the bottom flange
were then solved by using the code DDASSL (Petzold22) and can be locked in position by a Swagelok nut with
in double-precision Fortran on an HP-735 computer. Teflon ferrules. The end of this tube is plugged, and the
liquid is discharged into the absorption chamber from
Model Parameters perforations on the side of the tube near the plugged
end. This assembly makes it possible to discharge the
The nitrous oxide analogy method was used to esti- liquid at different heights if so desired. The absorption
mate the CO2 solubility (Rinker and Sandall23) and the chamber is also equipped with four flat stainless steel
CO2 diffusivity (Ashour et al.24) in the aqueous amine baffles, which help in reducing vortex formation and
solutions. The kinetics for CO2/MDEA were measured promoting better mixing of the liquid phase. The baffles
by Rinker et al.,6 and the kinetics for the CO2/DEA are 12.7 cm long, 1.0 cm wide, and 0.1 cm thick and are
reaction were determined by Rinker et al.15 The diffu- pinned to the bottom flange, and their top ends are
sion coefficients of MDEA and DEA were estimated from connected by a wire ring. The distance between the
the diffusivity data of Hikita et al.25 baffles and the glass wall of the chamber is about 0.35
We include here correlations that were used to cm. When the liquid height in the chamber is 11.0 cm,
estimate various other reaction rate coefficients and the baffles extend about 1.7 cm above the liquid surface.
equilibrium constants that were obtained from the The absorber has two concentric shafts that protrude
literature. Values for the forward rate coefficient of into the chamber through the top flange. The inner shaft
reactions 1 and 2, k1 and k2, respectively, were calcu- is 0.6 cm in diameter and 36.5 cm long and is made of
lated from the following correlations, which were re- stainless steel. This shaft extends to a Teflon bushing
ported by Pinsent et al.26 for the temperature ranges of in the bottom flange and is supported at the top end by
273-311 K and 273-313 K, respectively: a pin bearing held in a cup on a crossbar. There are
 Ind. Eng. Chem. Res., Vol. 39, No. 11, 2000 4351

Figure 1. Schematic drawing of the laminar-liquid-jet absorber.

Table 1. Equilibrium Constant Correlations Used for Model Calculations

equilibrium temp range
constant equation (K) reference
K16 log10(K16) ) 8909.483 - 142613.6/T 293-573 Olofsson
-4229195 log10(T) + 9.7384 T and Hepler27
-0.012 963 8T2 + (1.150 68 × 10-5)T3
-(4.602 × 10-9)T4
K2K16 log10(K2K16) ) 179.648 + 0.019 244T 273-523 Read28
-67.341 log10(T) - 7 495.441/T
K14K16 log10(K14K16) ) 6.498 - 0.023 8T - 2 902.4/T 273-323 Danckwerts
and Sharma29
K4K5K13K16 log10(K4K5K13K16) ) -10.549 2 + 1 526.27/T 298-333 Barth et al.30
K13K16 log10(K13K16) ) -4.030 2 - 1 830.15/T + 0.0043T 298-333 Barth et al.30
K12K16 log (K12K16) ) -14.01 + 0.018T 298-333 Barth et al.30

two liquid-phase impellers mounted on the shaft, one introduced into the absorption chamber through the
with a diameter of 7.5 cm mounted at a height of 8.0 bottom flange and is discharged into the chamber by
cm from the bottom flange, and another with a diameter seven perforations about 8 cm from the bottom of the
of 6.0 cm mounted at a height of 3.0 cm from the bottom chamber. The liquid leaves the chamber through the
flange. The second shaft, which is a hollow stainless bottom flange and goes to a liquid-leveling reservoir.
steel tube 20.0 cm long with a 1.1 cm i.d., is supported The leveling device is similar to that used in the
by a concentric tubular extension welded onto the top laminar-liquid-jet absorber.
flange and extends 10.0 cm into the chamber. Mounted The gases pass through two sets of regulators (high
on the second shaft are two impellers for stirring the and low pressure) and then through mass flow control-
gas phase, a 7.5 cm diameter impeller at 10.5 cm from lers in order to maintain constant gas feed rates. The
the top flange, and a 9.5 cm diameter impeller at 17.5 gases subsequently pass through soap bubble meters so
cm from the top flange. When the liquid height in the that their volumetric flowrates can be measured. The
chamber is 11.0 cm, the bottom edge of the second gas gases are next mixed in a T-fitting and fed to a
impeller is 2.0 cm away from the gas-liquid interface. saturator. After the saturator, the overall gas flow rate
Two sets of ball bearings, Teflon packings, and spring- is measured with a bubble meter. The temperature of
loaded Teflon seals support the tubular shaft. Two the gas mixture is measured and recorded at this point
similar assemblies support the liquid shaft. The two with a type-J thermocouple. The gas mixture then
shafts are driven independently by two variable-speed passes through the stainless steel heating/cooling coil
motors. in the constant-temperature jacket before being intro-
The liquid feed is pumped to a surge tank and then duced into the absorption chamber through the tube in
through a calibrated rotameter and through a stainless the top flange. The gas mixture is sampled just before
steel coil in the constant-temperature jacket. It is then it enters the absorption chamber and is analyzed with
4352 Ind. Eng. Chem. Res., Vol. 39, No. 11, 2000

Figure 2. Schematic drawing of stirred-cell absorber.

a gas chromatograph. The gas is discharged just above Table 2. CO2 Absorption Data in 10 wt % Blends of DEA
the larger impeller through perforations near the end and MDEA Obtained in the Laminar-Liquid-Jet Absorber
at 25 °C
of the feed tube. The gas exits the chamber through a
fitting in the top flange and is sampled for composition RCO2 × 105 RCO2 × 105
by a gas chromatograph. The volumetric flow rate of the [MDEA] [DEA] measured predicted
exit gas is measured with a soap bubble meter and the (kmol/m3) (kmol/m3) (kmol/m2s) (kmol/m2s)
temperature of the exit gas is measured with a type-J 0.844 0 1.22 1.22
thermocouple. 0.808 0.040 1.29 1.35
The gas samples were analyzed using a model 5890 0.692 0.173 1.45 1.44
0.586 0.293 1.65 1.65
Hewlett-Packard gas chromatograph equipped with a 0.532 0.355 1.71 1.76
thermal conductivity detector. The column used in the 0.232 0.696 2.40 2.69
GC was a 6 ft. long, 0.085 in. i.d., stainless steel column 0 0.959 3.28 3.32
packed with Poropak Q with a mesh size of 80/100. The
run conditions were 70 °C for 1.5 min, ramp to 90 °C at deprotonation of the zwitterion. Littel et al.14 report that
5 °C per minute, cool, and equilibrate for 2 min. MDEA makes a measurable contribution to the depro-
tonation of the zwitterion for CO2 absorption into
Experimental Results aqueous blends of DEA and MDEA.
To test the hypothesis that MDEA contributes to the
(A) Absorption Measurements at Short Contact deprotonation of the zwitterion, the rates of CO2 absorp-
Times Using a Laminar-Liquid-Jet Absorber. Ac- tion into aqueous blends of DEA and MDEA were
cording to the zwitterion mechanism, any base can measured in the laminar-liquid-jet absorber at 25 °C.
contribute to the deprotonation of the zwitterion. For The total weight percents of the aqueous DEA/MDEA
the aqueous DEA case, the potential bases are R2NH, blends were 10, 30, and 50 wt %, and the molar ratios
OH-, H2O, HCO, and CO. In an earlier study (Rinker of DEA to MDEA were varied from zero to infinity. The
et al.15), the contributions of OH-, H2O, HCO, and CO absorption data are listed in Tables 2-4. The rates of
to the deprotonation of the zwitterion were found to be CO2 absorption were predicted using the numerical
insignificant for the reaction of CO2 with aqueous DEA model developed in this work without taking into
in the laminar-liquid-jet absorber. However, if CO2 is account the contribution of MDEA to the deprotonation
absorbed into aqueous blends of DEA and MDEA, it is of the zwitterion. The predicted and measured rates of
possible that MDEA could contribute significantly to the CO2 absorption are compared in the parity plot shown
 Ind. Eng. Chem. Res., Vol. 39, No. 11, 2000 4353

Table 3. CO2 Absorption Data in 30 wt % Blends of DEA

and MDEA Obtained in the Laminar-Liquid-Jet Absorber
at 25 °C
RCO2 × 105 RCO2 × 105
[MDEA] [DEA] measured predicted
(kmol/m3) (kmol/m3) (kmol/m2s) (kmol/m2s)
2.581 0 1.12 0.870
2.118 0.529 1.97 1.49
1.376 1.376 3.08 2.83
0.712 2.135 3.77 3.80
0 2.947 4.86 4.75
2.581 0 1.06 0.850
2.118 0.529 1.73 1.51
1.376 1.376 3.06 2.82
0.712 2.135 3.72 3.80
0 2.947 4.65 4.71

Table 4. CO2 Absorption Data in 50 wt % Blends of DEA

and MDEA Obtained in the Laminar-Liquid-Jet Absorber Figure 3. Comparison of predicted and measured rates of CO2
at 25 °C absorption into aqueous blends of DEA and MDEA at 25 °C in
the laminar-liquid-jet absorber. The total amine concentrations
RCO2 × 105 RCO2 × 105 were 10, 30, and 50 mass %.
[MDEA] [DEA] measured predicted
(kmol/m3) (kmol/m3) (kmol/m2s) (kmol/m2s) CO2 into carbonate. The carbonate was precipitated
4.217 0.202 0.753 0.692 with barium chloride, and the remaining NaOH was
4.217 0.202 0.868 0.783 titrated with 0.05 M HCl. This method was reproducible
3.722 0.772 1.10 1.11 to within 1%, and it was accurate enough to measure
3.719 0.771 1.15 1.12
3.722 0.772 1.09 1.09
the small rates of CO2 absorption over liquid stirring
3.034 1.517 1.67 1.72 speeds of 30-140 rpm. For the pure PEG400 experi-
3.034 1.517 1.50 1.74 ments, the samples were injected into twice their
volume of 0.05 M NaOH in order to convert the CO2 to
in Figure 3. From this plot, it is clear that there is good carbonate and dilute the PEG400 so that the samples
agreement between the measured rates of CO2 absorp- were mostly water (on a molar basis) for accurate pH
tion and the rates of CO2 absorption predicted by the determination. Pure PEG400 was used to check the
model neglecting the contribution of MDEA to the viscosity dependence of the mass transfer coefficient as
deprotonation of the zwitterion. The average deviation it has a viscosity of 0.9527 P at 25 °C and water has a
of the predicted rates from the measured rates is 6.8%. viscosity of 0.00895 P at 25 °C.
On the basis of these experiments, it appears that The physical mass transfer coefficients were calcu-
MDEA does not significantly contribute to the depro- lated from the measured data using the following
tonation of the zwitterion. equations:
A possible explanation for the difference in the results
of this study and the work of Littel et al.14 is that we RA1 ) k°l,1(C/A- CA0) (58)
used a laminar-liquid-jet absorber whereas they used
a batch stirred-tank reactor. The laminar-jet absorber C0A L
operates at steady state, with a gas-liquid contact time
RA1 ) (59)
As
of about 0.005 s, whereas the reactor used by Littel et
al. operates under transient conditions and over much where C/A is the interfacial concentration of CO2, C0A is
longer gas-liquid contact times. For very short contact the bulk or effluent concentration of CO2, L is the
times, the deprotonation reaction with MDEA may not volumetric liquid flowrate, As is the area of the gas-
have time to contribute significantly to the overall rate liquid interface (79.64 cm2), and k is the physical liquid-
of absorption. phase mass transfer coefficient. The mass transfer
(B) Absorption Measurements at Long Contact coefficients were correlated in dimensionless form as
Times Using a Stirred-Cell Absorber. To predict follows:
rates of CO2 absorption obtained using the stirred-cell
apparatus, accurate values of the liquid-phase mass Sh ) 0.0193Re0.845Sc0.5 (60)
transfer coefficient for physical absorption must be
determined from experimental absorption data obtained where
for a model system. The liquid-phase physical mass
transfer coefficient was determined by measuring the k°l,1da FNLdi2
concentration of dissolved CO2 in the liquid effluent µ
Sh ) Re ) Sc ) (61)
from the stirred-cell absorber using wet-chemical analy- DA µ FDA
sis. The liquids used for these experiments were pure
deionized water and pure poly(ethylene glycol) (PEG400). The dimensionless mass transfer coefficients used to
The gas phase was pure CO2 saturated with the vapor obtain the correlation given above are plotted in
pressure of the liquid. The liquid feedstocks were Figure 4.
degassed at elevated temperatures. The concentration (i) CO2 Absorption into Aqueous MDEA. Rates of CO2
of CO2 dissolved in the liquid was determined by absorption into aqueous solutions of MDEA were mea-
titration of samples of effluent. The water samples were sured in the stirred-cell absorber at 25 °C. The CO2
drawn slowly into a syringe and injected into an equal partial pressure was varied from 0.15 to 0.5 atm, and
volume of 0.05 M NaOH to convert all of the dissolved the diluent was N2. The measured absorption rates are
4354 Ind. Eng. Chem. Res., Vol. 39, No. 11, 2000

Figure 7. Comparison of predicted and measured rates of CO2

Figure 4. Liquid-phase physical absorption mass transfer coef- absorption into aqueous 50 mass % blends of DEA and MDEA at
ficient correlation. 25 °C in the stirred-cell absorber.

with an average deviation of 10.2% from the measured

rates.
(iii) Absorption of CO2 into Aqueous Blends of DEA
and MDEA. Rates of absorption of CO2 into aqueous
blends of DEA and MDEA were measured in the stirred-
cell absorber at 25 °C. The total amine concentration
was approximately 50 wt %, and the molar ratios of
DEA to MDEA were 0.050, 0.207, 0.250, and 0.500. The
model predictions are compared to the measured ab-
sorption rates in Figure 7. The average deviation of the
predictions from the measured rates of absorption is
17.6%. The predicted rates of absorption were made by
neglecting the contribution of MDEA to the deprotona-
tion of the zwitterion. Some of the absorption rates are
Figure 5. Comparison of predicted and measured rates of CO2 overpredicted and some are underpredicted when the
absorption into aqueous MDEA at 25 °C in the stirred-cell MDEA contribution is neglected. As a result, the CO2
absorber. absorption data obtained in the stirred-cell absorber
support the findings for CO2 absorption into blends in
the laminar-liquid-jet absorber that MDEA does not
significantly contribute to the deprotonation of the
zwitterion.

Conclusions
The model developed in this work for the rates of
absorption of carbon dioxide into a aqueous mixed amine
solutions was found to agree reasonably well with the
experiments. The reaction between carbon dioxide and
the secondary amine, DEA, was described by the zwit-
terion mechanism in this model. For our experiments
with gas-liquid contact times varying from approxi-
mately 0.005 to 10 s, it appears that the tertiary amine,
MDEA, does not contribute significantly to the depro-
Figure 6. Comparison of predicted and measured rates of CO2 tation of the zwitterion. The only species that contrib-
absorption into aqueous DEA at 25 °C in the stirred-cell absorber. utes significantly to the deprotonation of the zwitterion
is DEA, and thus, eqs 6-10 in the reaction model could
compared to the model predictions in the parity plot be deleted according to the results obtained in this
shown in Figure 5. The average deviation of the model study.
predictions from the measured rates is 24%. The key physicochemical property needed for the
(ii) CO2 Absorption into Aqueous DEA. Rates of CO2 model calculations is the physical solubility of CO2. Any
absorption into 10 wt % aqueous DEA were measured uncertainty in this property translates to an equivalent
in the stirred-cell absorber at 25 °C. The CO2 partial error in the predicted absorption rate.
pressure was varied from 0.088 to 0.712 atm, and the
diluent was N2. The measured and predicted rates of Acknowledgment
CO2 absorption are compared in the parity plot shown
in Figure 6. In this case, there is fairly good agreement This work was sponsored by the Gas Research Insti-
between the model predictions and the measurements, tute and the Gas Processors Association.
 Ind. Eng. Chem. Res., Vol. 39, No. 11, 2000 4355

Notation (5) Littel, R. J.; van Swaaij, W. P. M.; Versteeg, G. F. Kinetics

of carbon dioxide with tertiary amines in aqueous solution. AIChE
As ) gas-liquid interface area, m2 J. 1990, 36, 1633-1640.
CA ) diameter of the laminar-liquid jet, m (6) Rinker, E. B.; Ashour, S. S.; Sandall, O. C. Kinetics and
Di ) coefficient of species i in the aqueous alkanolamine modelling of carbon dioxide absorption into aqueous solutions of
solution, m2/s N-methydiethanolamine. Chem. Eng. Sci. 1995, 50, 755-768.
(7) Caplow, M. Kinetics of carbamate formation and breakdown.
DEA ) diethanolamine J. Am. Chem. Soc. 1968, 90, 6795-6803.
DIPA ) diisopropanolamine (8) Danckwerts, P. V. The reaction of CO2 with ethanolamines.
E1 ) factor of CO2, defined by eq 54 Chem. Eng. Sci. 1979, 34, 443-446.
h0 ) spacing next to the gas-liquid interface for the (9) Blauwhoff, P. M. M.; Versteeg, G. F.; van Swaaij, W. P. M.
discretized spatial variable x, m A study on the reaction between CO2 and alkanolamines in
H1 ) physical equilibrium constant (Henry’s law constant) aqueous solutions. Chem. Eng. Sci. 1984, 39, 207-225.
(10) Versteeg, G. F.; van Swaaij, W. P. M. On the kinetics
for CO2, H1 ) P/1/u/1, (kPa m3)/kmol between CO2 and alkanolamines both in aqueous and nonaqueous
kapp ) rate coefficient for the reaction between CO2 and a solutions I. Primary and secondary amines. Chem. Eng. Sci. 1988,
secondary or a primary alkanolamine, defined by eq 70, 43, 573-585.
m3/(kmol s) (11) Versteeg, G. F.; Oyevaar, M. H. The reaction between CO2
kg,1 ) gas-phase mass transfer coefficient for CO2, kmol/ and diethanolamine at 298 K. Chem. Eng. Sci. 1989, 44, 1264-
(kPa m2 s) 1268.
(12) Versteeg, G. F.; Kuipers, J. A. M.; van Beckum, F. P. H.;
k0l,1 ) mass transfer coefficient for CO2 in the liquid phase, van Swaaij, W. P. M. Mass transfer with complex reversible
defined by equation (52), m/s chemical reactions-II. Parallel reversible chemical reactions. Chem.
ki ) rate coefficient of reaction i, s-1 for first-order Eng. Sci. 1990, 45, 183-197.
reactions, m3/(kmol s) for second-order reactions (13) Glasscock, D. A.; Critchfield, J. E.; Rochelle, G. T. CO2
k-i ) rate coefficient of reaction i absorption/desorption in mixtures of methyldiethanolamine with
Ki ) constant for reaction i monoethanolamine or diethanolamine. Chem. Eng Sci. 1991, 46,
2829-2845.
l ) length of the laminar-liquid jet, m (14) Littel, R. J.; Versteeg, G. F.; van Swaaij, W. P. M. Kinetics
L1 ) CO2 loading of the aqueous amine solution, (kmol of CO2 with primary and secondary amines in aqueous solutions
CO2)/(kmol amine) II. Influence of temperature on zwitterion formation and depro-
MDEA ) N-methyldiethanolamine tonation rates. Chem. Eng. Sci. 1992, 47, 2037-2045.
MEA ) monoethanolamine (15) Rinker, E. B.; Ashour, S. S.; Sandall, O. C. Kinetics and
NL ) impeller speed, rev/s Modelling of Carbon Dioxide Absorption into Aqueous Solutions
of Diethanolamine. Ind. Eng. Chem. Res. 1996, 35, 1107-1114.
P1 ) pressure of CO2 in the gas phase, kPa (16) Sada, E.; Kumazawa, H.; Han, Q.; Matsuyama, H. Chemi-
P/1) partial pressure of CO2 in the gas phase, kPa cal kinetics of the reaction of carbon dioxide with ethanolamines
Q ) volumetric flow rate of the liquid, m3/s in nonaqueous solvents. AIChE J. 1985, 31, 1297-1303.
Ri ) reaction rate of reaction i, kmol/(m3 s) (17) Bosch, H.; Versteeg, G. F.; van Swaaij, W. P. M. Kinetics
RA1 ) rate of absorption of CO2 per unit interfacial area, of the reaction of CO2 with the sterically hindered amine 2-amino-
methylpropanol at 298 K. Chem. Eng. Sci. 1990, 45, 1167-1173.
defined by equation (53), kmol/(m2 s)
(18) Alper, E. Reaction mechanism and kinetics of aqueous
Re ) Reynolds number defined by eq 80 solutions of 2-amino-2-methyl-1-propanol and carbon dioxide. Ind.
Sc ) Schmidt number defined by eq 80 Eng. Chem. Res. 1990, 29, 1725-1728.
Sh ) Sherwood number defined by eq 80 (19) Hanna, O. T.; Sandall, O. C. Computational Methods in
t ) independent time variable, s Chemical Engineering; Prentice Hall: Englewood Cliff, NJ, 1994.
T ) absolute temperature, K (20) Higbie, R. The rate of absorption of a pure gas into a still
liquid during short periods of exposure. Trans. Am. Inst. Chem.
ui ) concentration of species i in the liquid phase (which
Eng. 1935, 31, 365.
is a function of x and t), kmol/m3 (21) Danckwerts, P. V. Gas-Liquid Reactions; McGraw-Hill:
0
ui ) liquid bulk concentration of s pecies i (which is a New York, 1970.
constant), kmol/m3 (22) Petzold, L. R. A Description of DASSL: A differential/
u/1 ) interfacial concentration of CO2 in the liquid, kmol/ algebraic system solver. In Scientific Computing; Stepleman, R.
S., Carver, M., Peskin, R., Ames, W. F., Vichnevetsky, R., Eds.;
m3
North-Holland; Amsterdam, 1983; pp 65-68.
x ) independent spatial variable, m (23) Rinker, E. B.; Sandall, O. C. Solubility of nitrous oxide in
aqueous solutions of methyldiethanolamine, diethanolamine and
Greek letters
mixtures of methyldiethanolamine and diethanolamine. Chem.
µ ) viscosity of the aqueous alkanolamine solution, kg/(m Eng. Commun. 1996, 144, 85-94.
s) (24) Ashour, S. S.; Rinker, E. B.; Sandall, O. C. Correlations
F ) density of the aqueous alkanolamine solution, kg/m3 for estimating the diffusivities of nitrous oxide in aqueous solutions
of diethanolamine and n-methyldiethanolamine and the solution
τ ) gas-liquid contact time, defined by eq 51 for a laminar- densities and viscosities. Chem. Eng. Commun. 1997, 161, 15-24.
liquid-jet absorber, s (25) Hikita, H.; Ishikawa, H.; Uku, K.; Murakami, T. Diffu-
sivities of mono-, di-, and tri-ethanolamines in aqueous solutions.
Literature Cited J. Chem. Eng. Data 1980, 25, 324-325.
(26) Pinsent, B. R. W.; Pearson, L.; Roughton, F. J. W. The
(1) Chakravarty, T.; Phukan, U. K.; Weiland, R. H. Reaction kinetics of combination of carbon dioxide with hydroxide ions.
of acid gases with mixtures of amines. Chem. Eng. Prog. 1985, 81 Trans. Faraday Soc. 1956, 52, 1512-1520.
(4), 32-36. (27) Oloffson, G.; Hepler, L. G. Thermodynamics of ionization
(2) Donaldson, T. L.; Nguyen, Y. N. Carbon dioxide reaction of water over wide ranges of temperature and pressure. J. Solution
kinetics and transport in aqueous amine membranes. Ind. Eng. Chem. 1975, 4, 127.
Chem. Fundam. 1980, 19, 260-266. (28) Read, A. J. The first ionization constant of carbonic acid
(3) Haimour, N.; Bidarian, A.; Sandall, O. C. Kinetics of the from 25 to 250° and to 2000 bar. J. Solution Chem. 1975, 4, 53-
reaction between carbon dioxide and methyldiethanolamine. 70.
Chem. Eng. Sci. 1987, 42, 1393-1398. (29) Danckwerts, P. V.; Sharma, M. M. The absorption of carbon
(4) Versteeg, G. F.; van Swaaij, W. P. M. On the kinetics dioxide into solutions of akalis and amines (with some notes on
between CO2 and alkanolamines both in aqueous and nonaqueous hydrogen sulphide and carbonyl sulphide). Chem. Eng. 1966,
solutions II. Tertiary amines. Chem. Eng. Sci. 1988, 43, 587-591. October, 44, CE244-CE280.
4356 Ind. Eng. Chem. Res., Vol. 39, No. 11, 2000

(30) Barth, D.; Tondre, C.; Lappai, G.; Delpuech, J.-J. Kinetic (33) Lide, D. R., Ed. CRC Handbook of Chemistry and Physics;
study of carbon dioxide reaction with tertiary amines in aqueous 71st ed.; CRC Press: Boca Raton, FL, 1990.
solutions. J. Phys. Chem. 1981, 85, 3660-3667.
(31) Cotton, F. A.; Wilkinson, G. Advanced Inorganic Chemistry; Received for review November 22, 1999
Interscience Publishers: New York, 1966. Revised manuscript received July 17, 2000
(32) Smith, J. M.; Van Ness, H. C. Introduction to Chemical Accepted July 20, 2000
Engineering Thermodynamics, 4th ed.; McGraw-Hill: New York,
1987. IE990850R