Chemical Engineering Science 60 (2005) 6462 6471

www.elsevier.com/locate/ces

Rigorous modelling of NOx absorption in tray and packed columns

Bernhard Hpen, Eugeny Y. Kenig
Department of Biochemical and Chemical Engineering, University of Dortmund, 44227 Dortmund, Germany
Received 15 November 2004; received in revised form 19 April 2005; accepted 19 April 2005
Available online 23 June 2005

Abstract
The absorption of NOx into pure water or nitric acid is usually a kinetically controlled operation in which complex reactions in both
gas and liquid phases are combined with multicomponent mass transfer. To be able to describe this process adequately, a rigorous general
rate-based model for the reactive absorption in tray and packed columns has been developed based on earlier works (see, e.g. [Kenig et
al., 2001. Chemical Engineering Science 56, 343350; Kenig et al., 2003. Chemical Engineering and Technology 6, 631646]). Heat and
multicomponent mass transfer in the lm phases, chemical reactions and thermodynamic non-idealities are taken into account. Furthermore,
the inuence of column internals and hydraulics is considered via relevant correlations. For tray columns, a special consideration of the
empty sections above the gasliquid layer on a tray is included. The model is implemented into the simulation environment Aspen Custom
ModelerTM and validated by comparison with experimental data for a pilot scale packed column and an industrial sieve tray column
equipped with cooling coils. The simulation results obtained for both column types agree well with the experimental measurements. In
addition, the inuence of cooling on the process behaviour is studied.
2005 Elsevier Ltd. All rights reserved.
Keywords: NOx absorption; Rate-based model; Cooling; Absorption; Simulation; Modelling; Transport processes

1. Introduction
Absorption of nitrous gases represents a widespread and
important operation in the chemical process industries. It is
mainly used in the production of nitric acid and in the purication of exhaust gas streams, e.g. in the fertiliser industry.
One the other hand, it is a highly complex process due to
the interaction of several components (e.g. NO, NO2 , N2 O4 ,
HNO3 ), both in the liquid and gas phases, parallel and consecutive chemical reactions and simultaneous absorption and
desorption.
The design of columns for NOx absorption is mostly
based on experience and/or on usage of simplied models
which cannot reproduce the process behaviour with sufcient accuracy. As a consequence, possible design errors can
lead to serious problems (e.g. exceeding emission limits for

Corresponding author. Tel.: +49 231 755 2357; fax: +49 231 755 3035.

E-mail address: e.kenig@bci.uni-dortmund.de (E.Y. Kenig).

0009-2509/$ - see front matter 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2005.04.060

nitrogen oxides). A rigorous and reliable process description

would help to avoid these difculties.
Modelling and simulation of gasliquid reactor units have
been usually based on the equilibrium stage concept in which
a unit is subdivided into several axial segments (stages). The
column proles are determined assuming the equilibrium
state between the streams leaving each stage and by using
the HETP-values (Height Equivalent to a Theoretical Plate)
(Taylor and Krishna, 1993). However, the application of the
equilibrium stage concept is difcult for processes driven
by rate controlled phenomena. This is just the case with
reactive absorption operations, in which reaction kinetics
and mass and heat transfer resistances represent the key
issues (Schneider et al., 1999).
The rate-based approach is a more consistent modelling
method to consider reaction and transport kinetics. Here,
gas and liquid phases are balanced separately, whereas
the mass and heat uxes across the interface are taken
into account directly (Taylor and Krishna, 1993). In addition, mutual inuence of chemical reactions and mass

B. Hpen, E.Y. Kenig / Chemical Engineering Science 60 (2005) 6462 6471

transfer can be properly taken into account (Kenig et al.,

2001).
Absorption of NOx is one of the most complex gasliquid
reaction operations. This is the reason why various simplications reducing the accuracy and predictivity of the ratebased approach have been made, e.g. consideration of mass
transport of the key components NO2 and N2 O4 only (Emig
et al., 1979; Greve and Bode, 1994; Wiegand et al., 1990),
assumption of equal component diffusivities (Emig et al.,
1979; Ramanand and Phaneswara Rao, 1996; Suchak et al.,
1991), application of enhancement factors (de Paiva and
Kachan, 2004), or neglecting the reactions in the gas lm
(Patwardhan et al., 2002; Suchak et al., 1991).
The present work is free of such simplications. We developed a general rate-based model which is valid for both
packed and tray columns and applicable for a broad spectrum of NOx absorption operations. The model is an extension of previous models (see Kenig et al., 1997, 2001, 2003;
Schneider et al., 1999) developed for the description of reactive absorption in packed columns. The model involves the
lm model equations combined with relevant reaction and
diffusion kinetics for all reactions and components, in both
the gas and liquid phases, and it is extended to include an
additional elementthe so-called gas areawhich considers the inuence of the gas-phase reactions on the composition in the empty section above the gasliquid layer on a
tray. Further, the model includes the description of cooling
effects through cooling coils installed in tray columns.
The model is realised with the help of the simulation environment Aspen Custom ModellerTM and successfully validated with NOx absorption experimental data for columns
of both packed and tray type.

2. Process description
A typical simplied ow diagram of NOx absorption
process is shown in Fig. 1, with three columns connected
counter-currently. The inlet of the raw gas consisting of
nitrogen oxides and air is at the bottom of the rst column. The solvent is fed at the top of the third column.
The nitrogen oxides are absorbed by the solvent owing

Fig. 1. Simplied ow diagram of a NOx absorption process.

6463

counter-currently to the gas and form nitric acid which is

drawn off at the bottom of the rst column. These columns
can be either of a packed or of a tray type. Tray columns
are often additionally equipped with cooling coils on trays
in order to achieve higher absorption rates via cooling.
The reaction network of the NOx system is extremely
complicated. Altogether, there are more than 40 reactions
known. For the simulation of the process, a relevant set of reactions was taken from Suchak et al. (1991). It contains ve
gas-phase and four liquid-phase reactions presented below:
gas-phase reactions
2NO + O2 2NO2 ,

(R1)

2NO2 N2 O4 ,

(R2)

NO + NO2 N2 O3 ,

(R3)

NO + NO2 + H2 O 2HNO2 ,

(R4)

3NO2 + H2 O 2HNO3 + NO;

(R5)

liquid-phase reactions
2NO2 + H2 O HNO2 + HNO3 ,

(R6)

N2 O3 + H2 O 2HNO2 ,

(R7)

N2 O4 + H2 O HNO2 + HNO3 ,

(R8)

3HNO2 HNO3 + H2 O + 2NO.

(R9)

For the process performance, the oxidation of nitrogen

monoxide to nitrogen dioxide (R1) is especially signicant,
because of the low solubility of nitrogen monoxide in water.
Compared to nitrogen monoxide, the solubility of nitrogen
dioxide is one order of magnitude higher. Reaction (R1)
is a kinetically controlled reaction and it is assumed to be
irreversible for temperatures below 350 C. In reactions
(R2)(R5), nitrogen dioxides instantaneously react to intermediate products (e.g. N2 O3 or HNO2 ) whose solubility
increases with increasing molecular weight.
The liquid-phase reactions are kinetically controlled and
assumed to be irreversible at nitric acid concentrations below
35 wt% (Joshi et al., 1985). Via reactions (R6)(R8) the
dissolved nitrogen oxides are transformed into nitric and
nitrous acids. NO produced in reaction (R9) is desorbed due
to its low solubility.
The instantaneous gas-phase reactions (R2)(R5) are
taken into consideration by means of the mass action law

G
p,r
p (yp P )
G
Kr =
G
e,r .
(1)
e (ye P )
The chemical equilibrium constants Kr are given in Table 1.
The reaction rates of the irreversible, kinetically controlled reactions are dened by the following

6464

B. Hpen, E.Y. Kenig / Chemical Engineering Science 60 (2005) 6462 6471

Table 1
Chemical equilibrium and forward reaction rate constants of reactions (R1)(R9)
Reaction

Equilibrium and rate constants

Reference

2 s1 )
ln k1 = 1468
T 10.9043 (kPa
6893
ln K2 = T 25.865 (kPa1 )
1
ln K3 = 4869
T 21.61 (kPa )
2051.17
ln K4 = T
8.7385 (kPa1 )
2003.8
ln K5 = T 10.763 (kPa1 )
log kc6 = 4.67209 (m2 mol1 s1 )
log kc7 = 4.23044 (s1 )
1
log kc8 = 4139
T + 16.3415 (s )
2 m 9 kmol3 s1 )
+
20.1979
(atm
log kc9 = 6200
T

R1
R2
R3
R4
R5
R6
R7
R8
R9

expressions:

G
rrG = kr,for
(yeG P )
e,r

for (R1),

(2)

for (R6).(R9).

(3)

Miller (1987)
Miller (1987)
Hisatsune (1961)
Joshi et al. (1985)
Joshi et al. (1985)
Lee and Schwartz (1981)
Corriveau (1971)
Wendel and Pigford (1958)
Wendel and Pigford (1958)

L
rrL = kr,for


e

e,r
(xeL L
e)

The data on the forward reaction rate constants kr,for are

listed in Table 1. For reactions (R6)(R9), these rate constants are based on concentrations. During the simulation,
they are converted to activity-based constants.
3. Rigorous rate-based model
In the investigated system, several components are present
in both phases. All components participate in complex gasand liquid-phase consecutive and parallel reactions, which
lead to a coupling of multicomponent mass transfer and
multiple reaction mechanism. As a consequence, a specially
developed rigorous mathematical model which considers the
reaction system and mass transfer resistances as well as the
column conguration is required (Kenig et al., 2003).
In Section 1, some previous studies of NOx absorption
based on different simplications are discussed. In this work,
we developed a more general model in which the abovementioned assumptions are not necessary.
The model is based on a description of a single stage,
which is related either to a tray or to a packing segment of a
column unit. Gas and liquid phase are balanced separately,
whereas mass and heat transfer resistances are considered
according to the lm theory, by the direct account of the
interfacial uxes. The lm model equations are combined
with the relevant reaction and diffusion kinetics. The model
is further extended in order to properly describe empty sections in tray columns using the ideal plug ow reactor concept (see Emig et al., 1979). On each tray, such a section is
located above the gasliquid layer (see Fig. 2). In the following, this empty section is called gas area. The gasliquid
layer is considered as homogeneously mixed. This is reasonable, because of the low liquid ow rate and, at the same

Fig. 2. Model structure of an axial segment for a tray column.

time, the high liquid hold-up on the trays of the investigated

column due to a weir height of 250 mm. This leads to a
strong mixing of the liquid on trays and, therefore, to relatively low concentration gradients over the liquid ow path.
An important advantage of this model is that the hydrodynamics in the column can be directly involved using correlations for hold-up, pressure drop, interfacial area and mass
transfer coefcients (Schneider et al., 1999).
3.1. Bulk-phase equations
In the liquid bulk balance equations, the axial change of
the total molar stream and the composition are considered

j Lb
int col
RiL Acol L ;
(x L) nLb
i a A
jz i
i = 1, . . . , n.

0=

(4)

The differential heat balance of the liquid bulk is

0=

j
(H Lb L) q Lb a int Acol + QLb
loss .
jz

(5)

The term QLb

loss is implemented into the heat balance to consider the heat effect achieved by cooling. In addition, the
summation condition for the molar fractions in the liquid
phase is valid
n

i=1

xiLb = 1.

(6)

B. Hpen, E.Y. Kenig / Chemical Engineering Science 60 (2005) 6462 6471

Similar equations can be applied for the gas bulk phase

j Gb
int col
(y G) + nGb
RiG Acol Gb ;
i a A
jz i
i = 1, . . . , n,

3.3. Interface

0=

j
0 = (H Gb GGb ) + q Gb a int Acol + QGb
loss ,
jz
n

i=1

yiGb = 1.

6465

The thermodynamic equilibrium at the gasliquid interface is usually calculated as follows:

(7)
(8)

(9)

3.2. Equations for the lm region

In the liquid lm region, mass transfer and chemical reaction occur simultaneously. Therefore, a differential component balance has to be considered

yiint = Ki xiint ;

i = 1, . . . , n,

(18)

where the distribution coefcient Ki comprises fugacities in

both phases and activity coefcients in the liquid phase. To
determine Ki of very diluted systems, the Henry law for the
dissolved components and the Raoult law for the solvent are
used.
Gas and liquid phases are connected by the following
boundary conditions:
Gf

ni |Gf =Gf = nLb

i |Lf =0 ;

i = 1, . . . , n,

q Gf |Gf =Gf = q Lf |Lf =0 .

(19)
(20)

Lf

0=

jni
Lf
Ri ;
jLf

i = 1, . . . , n.

(10)

In dilute systems, the diffusional interactions can usually

be neglected, and the MaxwellStefan equations, generally
applied for the description of mutlicomponent diffusion, are
replaced by the effective diffusivity description (Taylor and
Krishna, 1993)
Lf

ni

jx i
Lf Lf
+ x i nt ;
jLf

i = 1, . . . , n.

(11)

jq Lf
.
jLf

(12)

= 

Lf

jT Lf  Lf Lf
+
(ni Hi );
jLf

i = 1, . . . , n. (13)

i=1

Similar equations for the mass and heat transport are valid
in the gas lm:

Gf

= ct Dieff

0=

jq Gf
,
jGf

q Gf = Gf

i = 1, . . . , n,

(14)

Gf

ni

j Ga
(y G) RiG Acol Ga ;
jz i

i = 1, . . . , n.

(21)

Further, the summation condition for the molar fractions has

to be fullled
n


yiGa = 1.

(22)

The energy balance of the gas area is

0=

j
(H Ga GGa ).
jz

(23)

Eqs. (21)(23) are used in combination with the equations

for the gasliquid layer.
3.5. Model parameters

Gf

jni
Gf
0 = Gf
Ri ;
j
Gf

0=

i=1

The heat ux in the liquid lm consists of a conductive and

a convective term
Lf

For the description of tray columns, some additional consideration is necessary to describe the gas area (cf. Fig.
2). Whereas the gasliquid layer is still described by Eqs.
(4)(20), the gas area is covered in a different way.
The change of the gas composition in the gas area is
considered through the following differential equation:

Lf

Lf

= ct Dieff

The differential energy balance of the liquid lm is as follows:

0=

3.4. Gas area equations

jyi
Gf Gf
+ y i nt ;
Gf
j

i = 1, . . . , n,

(15)

(16)
n

jT Gf  Gf Gf
+
(ni Hi ).
jGf
i=1

(17)

For the determination of the mass transfer coefcients,

specic gasliquid interfacial area and volumetric hold-up,
empirical correlations are used, which govern the inuence
of column internals and hydraulics. These correlations have
to be incorporated into the whole system of model equations.
The relevant correlations are taken from Billet and Schultes
(1999) for packed columns and from Zuiderweg (1982) for
sieve tray columns. For the description of the hydraulics in
sieve tray columns, we included the equations for the spray
and emulsion ow regime. The decision, which equation
should be used, is taken automatically during the simulation

6466

B. Hpen, E.Y. Kenig / Chemical Engineering Science 60 (2005) 6462 6471

Table 2
Physical property models according to Aspen Properties Manual (2001)
Property

Gas phase

Liquid phase

Enthalpy
Molar density
Dynamic viscosity
Thermal conductivity
Diffusion coefcients
Surface tension
Activity coefcients

Via the heat capacity of an ideal gas

Ideal gas
ChapmannEnskogBrokaw
MasonSaxena
ChapmannEnskogWilkeLee

Watson
Rackett
Andrade
SatoRiedel
WilkeChang
HakimSteinbergStiel
NRTL

Table 3
NRTL-parameters bij and bj i
Components

Parameters

Components

Parameters

bij

bj i

bij

bj i

H2 O
H2 O
H2 O
H2 O
H2 O
H2 O
H2 O
H2 O
H2 O
HNO3
HNO3
HNO3
HNO3
HNO3
HNO3
HNO3
HNO3
HNO2
HNO2
HNO2
HNO2
HNO2
HNO2

HNO3
HNO2
NO
NO2
N2 O3
N2 O4
O2
N2
CO2
HNO2
NO
NO2
N2 O3
N2 O4
O2
N2
CO2
NO
NO2
N2 O3
N2 O4
O2
N2

0
0
751.8109
751.8109
0
751.8109
101.8304
35.8278
291.4104
0
0
0
0
0
0
0
0
0
0
0
0
0
0

0
0
1892.6936
1892.6936
0
1892.6936
174.5380
142.4195
177.4012
0
0
0
0
0
0
0
0
0
0
0
0
0
0

HNO2
NO
NO
NO
NO
NO
NO
NO2
NO2
NO2
NO2
NO2
N 2 O3
N 2 O3
N 2 O3
N 2 O3
N 2 O4
N 2 O4
N 2 O4
O2
O2
N2

CO2
NO2
N 2 O3
N 2 O4
O2
N2
CO2
N2 O3
N2 O4
O2
N2
CO2
N2 O4
O2
N2
CO2
O2
N2
CO2
N2
CO2
CO2

0
128.8337
0
128.8337
448.7655
358.1516
48.3459
0
128.8337
448.7655
358.1516
48.3459
0
0
0
0
448.7655
358.1516
48.3459
4.2683
246.3538
179.4870

0
114.7814
0
114.7814
345.7854
289.1752
125.2540
0
114.7814
345.7854
289.1752
125.2540
0
0
0
0
345.7854
289.1752
125.2540
5.1213
132.3010
91.2366

of the absorption process. It can be changed for each tray

within the same column and depends on the following ratio:
 
uL L 0.5
.
(24)
uG G
In addition, pressure drop along the column is allowed for
via the relevant pressure drop correlations (see Zuiderweg,
1982; Billet and Schultes, 1999).

essary for the calculation of the physical properties. These

models are listed in Table 2.
The activity coefcients are calculated by the NRTLmodel for multicomponent mixtures. The NRTL-parameters
aij , aj i , dij , eij , ej i , fij and fj i for each binary pair are set
to the default value zero, whereas cij to the default value 0.3.
The estimated NRTL-parameters bij and bj i are collected
in Table 3.

3.6. Physical properties

4. Model validation
The physical properties are calculated with the help of the
programme package Aspen PropertiesTM (Aspen Properties
Manual, 2001) which is directly linked to the simulation environment Aspen Custom ModellerTM , whereas the NRTLmethod is chosen as a basis method. According to the Aspen
PropertiesTM structure, this method species all models nec-

4.1. Model implementation

In Kenig et al. (1997), an industrial NOx absorption problem was solved by splitting the rate-based equation set onto
parts describing lm and bulk phenomena separately and by

B. Hpen, E.Y. Kenig / Chemical Engineering Science 60 (2005) 6462 6471

simplifying the lm equations in order to obtain an analytically solvable matrix boundary value problem. Linearisation of the diffusion and reaction terms of the lm equation
(cf. Eqs. (10) and (14)) allowed an analytical solution in a
closed matrix form (see Kenig and Grak, 1995; Kenig et
al., 1997). The composition boundary values need to be determined from the total system of equations describing the
process.
In this work we applied a fully numerical algorithm which
does not require any additional linearisation. To obtain a
numerical solution of such problems, a discretisation in
regard to the axial and normal co-ordinates was performed
to yield an algebraic equation system. The number of
equations can reach up to 200,000 depending on the axial discretisation steps of the column. To solve this large
equation system, the model was implemented into the commercial simulation environment Aspen Custom ModelerTM
(see http://www.aspentech.com/). The simulation tool offers
a direct link to the software package Aspen PropertiesTM
for the calculation of all required physical properties.

6467

Fig. 3. Axial prole of the total gas-phase NOx concentration of column

1 and measured values.

4.2. Validation for a packed column

The validation of the suggested model is performed by
comparison of simulation results with experimental data.
The data for the packed column is taken from Suchak et al.
(1991) where three pilot-scale columns connected countercurrently are presented. We carried out simulations for two
of these columns. The simulated units have a simple conguration, with one liquid inlet stream at the top and one gas
inlet stream at the bottom. Both columns have a diameter
of 0.254 m and are lled with a random packing (Pall rings,
16 mm, steel), with a packing height of 6 m. In column 1,
a gas stream containing 20 mol% of nitrogen oxides is absorbed by an aqueous solution with 5 mol% of nitric acid.
The inlet gas stream of the column 2 contains 10.2 mol% of
nitrogen oxides and is treated by an aqueous solution containing 2.65 mol% nitric acid.
Figs. 3 and 4 show the simulated axial proles of the
total NOx concentration in the gas phase for columns 1 and
2. The outlet values are compared with the measurements.
To demonstrate the accuracy of the simulation results, the
experimental values are plotted with 5% error bars.
For column 1, the calculated total NOx concentration at
the top of the column show good agreement with the experiments, the maximum deviation being less than 5%.
In Figs. 5 and 6, the simulated axial proles of the liquidphase temperature and measured values for both columns
are presented. Similar to Figs. 3 and 4, 5% error bars are
plotted to demonstrate the accuracy of the simulation results.
The liquid temperature proles reveal a maximum in the
lower section of each column. This is typical for NOx absorption processes and can be explained as follows. The
NOx concentration of the inlet gas stream is higher than in
the upper part of the column. Therefore, in the lower part,

Fig. 4. Axial prole of the total gas-phase NOx concentration of column

2 and measured values.

Fig. 5. Axial prole of the liquid temperature of column 1 and measured

values.

the exothermic reactions are enhanced. In addition, the absorption rate of NOx is higher due to high NOx concentration gradients. Both effects lead to an increase of the liquid
temperature. On the other hand, at the bottom of the column, the liquid is cooled down again by the cold inlet gas

6468

B. Hpen, E.Y. Kenig / Chemical Engineering Science 60 (2005) 6462 6471

Fig. 6. Axial prole of the liquid temperature of column 2 and measured

values.

Fig. 7. Axial prole of the total NOx concentration in the gas phase and
the nitric acid concentration in the liquid phase and measured values.

Table 4
Design parameters of the industrial sieve tray column

Fig. 7 shows the calculated axial proles of the total NOx

concentration in the gas phase and the nitric acid concentration in the liquid phase. The inlet and outlet values determined with a measurement error of 15% are given, too. The
agreement between the experimental and simulated values
of the total NOx has a maximum deviation of 8.5%. The
deviation of the simulated HNO3 concentration from the experimental value is in a range of 10%. Thus, the agreement
is good, as all deviations lie within the measurement error.
The simulated outlet temperature of the cooling water is
24.8 C. This value agrees very well with the measured value
of 24.9 C, the deviation being less than 1%.
From the results of both pilot plant and industrial applications, it can be concluded that the suggested model demonstrates a good accuracy for the highly complex NOx absorption processes studied.

Column diameter
Number of trays
Plate spacing
Weir height
Weir length
Flow path
Number of holes
Hole diameter
Distance of holes (pitch)
Hole diameter

3.8 m
20
0.9 m
0.25 m
3.015 m
2.3 m
54,000
2.2 mm
13 mm
2.2 mm

stream. These two opposite effects result in an extremum of

the axial temperature prole.
For both columns, the calculated liquid temperature at the
bottom differs from the measured value by more than 5%.
The absolute deviation is 4.5 C for column 1 and 3.8 C
for column 2. These deviations should be attributed to heat
losses through the column wall. Suchak et al. (1991) mentioned that their columns were not isolated, however they
did not quantify this phenomenon, and thus, it could not be
properly estimated in the calculations.
4.3. Validation for a tray column
The experimental data for an industrial scale onepass sieve tray column is provided by the Phosphoric
Fertilisers Industry (Greece) in the context of the European research project OPT-ABSO (G1RD-CT-2001-00649,
http://www.opt-abso.org). The design details of the tray column are given in Table 4. Each tray has a height of 0.9 m. In
the column, a gas stream containing 0.77 mol% of nitrogen
oxides is treated by an aqueous solution with 0.68 mol%
of nitric acid. To cool down the gas and liquid within the
column, it is equipped with cooling coils on 7 trays in the
lower part. The coils are fed with 152 m3 /h water, with an
average temperature of 23.6 C.

5. Sensitivity analysis
5.1. Internals-related parameters
The validated model has been used to carry out comprehensive sensitivity studies for the packed columns in order
to investigate the inuence of model parameters on the simulation results. The values of the internals-related parameters, such as the mass transfer coefcients k L and k G , the
gasliquid interfacial a int area and the liquid hold-up L ,
have been increased by 10%. The inuence of these parameters is quantied by the change of the total absorption rate
of NO, NO2 , N2 O4 and total NOx .
The results presented in Fig. 8 show a substantial inuence
of the interfacial area a int on the absorption rate of NO2 . An
increase of a int by 10% leads to a 7.2% higher absorption
rate. The inuence of a int on the absorption rate of the other
components and the inuence of the other model parameters
k L , k G and L is less signicant. Changes by 10% lead to
deviations less than 3%.

B. Hpen, E.Y. Kenig / Chemical Engineering Science 60 (2005) 6462 6471

Fig. 8. Inuence of a 10% increase of the model parameters L , k L , k G

and a int on the absorption rate of NO, NO2 , N2 O4 and total NOx .

Fig. 9. Effect of cooling on the absorption rate of NOx in the industrial

sieve tray column.

5.2. Reaction rate

Further, the sensitivity of the reaction rate of the important oxidation reaction (R1) on the simulation results for
the packed columns has been investigated. Simulations are
carried out based on rate constants obtained by four different approaches (Sherwood et al., 1975; Emig et al., 1979;
Koukolik and Marek, 1968; Miller, 1987). The simulations
show a strong inuence of the reaction rate. The different
approaches for this reaction lead to a variation of 12% of the
overall absorption rate of NOx . Similar analysis performed
for the other reactions lead to deviations less than 1%.
Thus, for both validation cases, the reaction rate constants
for (R1) and the gasliquid interfacial area appear to be the
most signicant model parameters, and hence, considerable
attention should be drawn to their proper determination.
5.3. Cooling
Finally, the effect of cooling on the process behaviour of
the industrial sieve tray column has been analysed. Fig. 9

6469

Fig. 10. Effect of cooling on the absorption rate of NOx for an increased
NOx inlet gas stream concentration.

shows the simulated absorption rate of NOx and the cooling

duty for different cooling water ows. A strong dependency
of the absorption rate of NOx on the cooling duty is visible.
Increasing heat duty leads to a growing absorption rate. The
increase of the cooling water ow up to 100 m3 /h leads to
an enhanced NOx absorption rate by 0.15% and to a cooling
duty of 0.18 MW. A further increase of the cooling water ow
does not have any signicant inuence on both parameters.
Despite the strong dependency on the cooling duty, the
absolute changes of the absorption rates caused by the cooling are small. This effect can be understood if one bears in
mind the low NOx fraction in the inlet gas stream. A higher
NOx concentration in the inlet gas stream would result in
a more signicant enhancement of the exothermic reactions
and thus to a temperature increase and to a drop of the absorption rate. In essence, the aim of gas and liquid cooling
is just to prevent these effects, and its efciency should be
more clear at high inlet NOx concentrations. An example is
shown in Fig. 10 for an inlet gas stream containing 7.8 mol%
of NOx , which is about ten times more than in the previously considered industrial case.

6. Conclusions
In this work, a general rigorous rate-based model for the
NOx absorption in tray and packed columns has been developed. The model is based on the lm theory and is able to
describe the strong interplay of the reaction, mass and heat
transport. This ability is based on the consideration of the
relevant reaction and diffusion kinetics for all reactions and
components in the lm equations. An additional issue has
been the effect of the gas-phase reaction on the gas composition in the empty sections of a tray column. This effect is
described using the model of an ideal plug ow reactor. Furthermore, the model of a tray unit includes the description
of cooling through cooling coils installed on trays.
The model is veried against experimental data of a pilot
scale packed column and of an industrial sieve tray column

6470

B. Hpen, E.Y. Kenig / Chemical Engineering Science 60 (2005) 6462 6471

equipped with cooling coils. Good agreement between experimental and simulated data is obtained for both processes.
This should be regarded as a success, since such processes
are usually difcult to describe adequately due to their extremely high complexity. Thus, the suggested model can be
judged as accurate enough for the description of the NOx
absorption processes in tray columns and as both accurate
and predictive for the NOx absorption in packed columns.
A comprehensive parameter sensitivity study reveals that
the reaction rate constants of the oxidation reaction (R1)
and the gasliquid interfacial area have the most signicant
inuence on the process performance. Further, the investigation of the effect of column cooling on the process behaviour shows that there is a clear dependency between the
cooling duty and the NOx absorption rate.
The application of the suggested model enables an accurate design and operation of NOx absorption processes
and can be used for other relevant gasliquid reaction
systems.

Notation
a int
Acol
c
Dieff
G
H
kr
K
Kr
L
ni
n
P
q
Qloss
R
T
u
x
y
z

specic gasliquid interfacial area, m2 /m3

column cross-section, m2
molar concentration, mol/m3
effective diffusion coefcient, m2 /s
total molar gas ow, mol/s
molar enthalpy, J/mol
reaction rate constant of reaction r, diverse, depending on the kinetic expression
phase distribution coefcient, dimensionless
equilibrium constant of reaction r, dimensionless
total molar liquid ow, mol/s
molar ux of component i, mol/m2 s
number of components
pressure, Pa
heat ux, J/m2 s
heat loss, J/s
total reaction rate, mol/m3 s
temperature, K
velocity, m2 /s
liquid phase molar fraction, dimensionless
gas phase molar fraction, dimensionless
axial co-ordinate directed from the top to the bottom, dimensionless

Greek letters






activity coefcient, dimensionless

lm thickness, m
lm coordinate, m
thermal conductivity, W/m K
stoichiometric coefcient, dimensionless




density, kg/m3
volumetric hold-up, m3 /m3

Subscripts
c
e
for
i
p
r
t

concentration-based
reactant
forward reaction
component index
product
reaction index
total

Superscripts
b
f
G
Ga
int
L

bulk
lm
gas phase
gas area
interface
liquid phase

Acknowledgements
The nancial support of the European Commission in the
context of the Project OPT-ABSO (Contract no. G1RD-CT2001-00649) is highly appreciated.

References
Aspen Properties Manual, 2001. Physical property systems, physical
property methods and models 11.1. Aspen Technology Inc.
Billet, R., Schultes, M., 1999. Prediction of mass transfer columns with
dumped and arranged packings. Transactions of the Institution of
Chemical Engineers 77, 498504.
Corriveau, C.E., 1971. Master Thesis, University of California, Berkley.
Emig, G., Wohlfahrt, K., Hoffmann, U., 1979. Absorption with
simultaneous complex reactions in both phases, demonstrated by the
modelling and calculation of a countercurrent ow column for the
production of nitric acid. Computers and Chemical Engineering 3, 143
150.
Greve, A., Bode, J., 1994. Absorption von NOx in einer SalpetersureKolonne. Chemie Ingenieur Technik 66, 220223.
Hisatsune, I.C., 1961. Thermodynamic properties of some oxides of
nitrogen. Journal of Physical Chemistry 65, 22492253.
Joshi, J.B., Mahajani, V.V., Juvekar, V.A., 1985. Absorption of NOx gases.
Chemical Engineering Communications 33, 192.
Kenig, E.Y., Grak, A., 1995. A lm model based approach for simulation
of multicomponent reactive separation. Chemical Engineering and
Processing 34, 97103.
Kenig, E.Y., Wiesner, U., Grak, A., 1997. Modelling of reactive
absorption using the MaxwellStefan equations. Industrial and
Engineering Chemistry Research 36, 43254334.
Kenig, E.Y., Schneider, R., Grak, A., 2001. Reactive absorption: optimal
process design via optimal modelling. Chemical Engineering Science
56, 343350.

B. Hpen, E.Y. Kenig / Chemical Engineering Science 60 (2005) 6462 6471

Kenig, E.Y., Kucka, L., Grak, A., 2003. Rigorous modelling of reactive
absorption processes. Chemical Engineering and Technology 6, 631
646.
Koukolik, M., Marek, J., 1968. Proceedings of the 4th European
Symposium on Chemical Reaction Engineering. Pergamon, Oxford,
pp. 347359.
Lee, Y.N., Schwartz, S.E., 1981. Reaction kinetics of nitrogen dioxide
with liquid water at low pressure. Journal of Physical Chemistry 85,
840848.
Miller, D.N., 1987. Mass transfer in nitric acid absorption. A.I.Ch.E.
Journal 8, 13511358.
de Paiva, J.L., Kachan, G.C., 2004. Absorption of nitrogen oxides in
aqueous solutions in a structured packing pilot column. Chemical
Engineering and Processing 43, 941948.
Patwardhan, J.A., Pradhan, M.P., Joshi, J.B., 2002. Simulation of NOx
gas absorption under adiabatic condition and comparison with plant
data. Chemical Engineering Science 57, 48314844.
Ramanand, S.B., Phaneswara Rao, D., 1996. Modelling and simulation
of NOx absorption into water in a countercurrent ow packed column.
Computers and Chemical Engineering 20, 10591063.

6471

Schneider, R., Kenig, E.Y., Grak, A., 1999. Dynamic modelling of

reactive absorption with the MaxwellStefan approach. Transactions
of the Institution of Chemical Engineers 77, 633638.
Sherwood, J.K., Pigford, R.L., Wilke, C.R., 1975. Mass Transfer. Mc
Graw-Hill, New York.
Suchak, N.J., Jethani, K.R., Joshi, J.B., 1991. Modelling and simulation
of NOx absorption in pilot scale packed columns. A.I.Ch.E. Journal
37, 323339.
Taylor, R., Krishna, R., 1993. Multicomponent Mass Transfer. Wiley, New
York.
Wendel, M.M., Pigford, R.L., 1958. Kinetics of nitrogen tetroxide
absorption in water. A.I.Ch.E. Journal 4, 249256.
Wiegand, K.W., Schreibler, E., Thiemann, M., 1990. Computation of plate
columns for NOx absorption by a new stage-to-stage method. Chemical
Engineering and Technology 13, 289297.
Zuiderweg, F.J., 1982. Sieve trays: a view on the state of the art. Chemical
Engineering Science 37, 14411464.