Você está na página 1de 13

Analysis of the effect of substrate material on the steady-state and

transient performance of monolith reactors


Santhosh R. Gundlapally
1
, Vemuri Balakotaiah
n
Department of Chemical and Biomolecular Engineering, University of Houston, Houston, TX 77204, USA
H I G H L I G H T S
c Ignited state exists over a wider range
of parameters for metallic substrates.
c Cumulative emissions have a mini-
mum with respect to gas velocity.
c Cumulative emissions are not mono-
tonic with substrate conductivity.
c For back-end ignition, it is shown
that metallic monoliths are superior.
c For front ignition, metallic monoliths
are superior at higher gas velocities.
G R A P H I C A L A B S T R A C T
Comparison of cold-start cumulative emissions of a monolith with metallic [50 W/(m K)], ceramic
[1 W/(m K)] and substrates with intermediate conductivity. The cold-start emissions from a monlith
are not monotonic with inlet gas velocity with metallic substrates leading to lower emissions at higher
gas velocities. There is also an intermediate conductivity around 5 W/(m K) that leads to lowest cold-
start emissions.
a r t i c l e i n f o
Article history:
Received 12 November 2012
Received in revised form
22 January 2013
Accepted 27 January 2013
Available online 4 February 2013
Keywords:
Catalyst support
Chemical reactors
Reaction engineering
Transient response
Light-off
Cumulative emissions
a b s t r a c t
We study the effect of the substrate material (ceramic versus metallic) on the steady-state and transient
performance of monolith reactors using a one-dimensional two-phase model with position dependent
transfer coefcients. When the operation of the reactor is on the ignited branch, it is shown that monoliths
with metallic substrate clearly lead to a superior steady-state performance compared to those with
ceramic substrate. In such cases, the ignited branch extends to lower inlet gas temperatures (typically 60
100 1C in after-treatment applications) for the same catalyst loading. For transient operation where the
time to light-off is important, it is shown that for the case of back-end ignition (corresponding to low inlet
temperatures or low catalyst loading), metallic substrates are again superior. However, for the case of
front-end ignition, ceramic converters may lead to lower cumulative emissions at lower inlet gas velocities
while the converse is true at higher velocities. It is shown that the transient heating time and hence the
cumulative emissions decrease with decrease in the channel hydraulic diameter and thermal capacitance
of the substrate but are not monotonic with the inlet gas velocity. We present mathematical analysis and
simulations to support these conclusions. Some novel results on the effect of substrate conductivity on the
number of steady-states and upstream propagation of temperature fronts are also presented.
& 2013 Elsevier Ltd. All rights reserved.
1. Introduction
Monoliths consists of a honeycomb structure with large
number of straight parallel channels of small diameter
(0.56 mm) through which the uid ows. Catalytic monoliths
are widely used in pollution abatement, such as the oxidation of
CO and unburnt hydrocarbons, and reduction of NOx from
Contents lists available at SciVerse ScienceDirect
journal homepage: www.elsevier.com/locate/ces
Chemical Engineering Science
0009-2509/$ - see front matter & 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ces.2013.01.051
n
Corresponding author. Tel.: 1 713 743 4318; fax.; 1 713 743 4323.
E-mail address: bala@uh.edu (V. Balakotaiah).
1
Current address: Gamma Technologies, Inc., 601 Oakmont Lane, Westmont,
IL 60559, USA.
Chemical Engineering Science 92 (2013) 198210
gasoline and diesel engines, and in the combustion of VOCs in
chemical industries. Due to their several advantages compared to
pellet type reactors, monoliths are exclusively used in the exhaust
after-treatment of automobiles, for example, three-way conver-
ters (TWCs), diesel oxidation catalysts (DOCs), lean NOx traps
(LNTs) and selective catalytic reduction units (SCRs). The sub-
strate of the monoliths used in the automobiles is mainly made
up of cordierite (2MgO 2Al
2
O
3
5SiO
2
), which has lower thermal
conductivity (0.82 W/m K). However, metallic monoliths are
gaining wider use in recent years. In the TWC, nearly 6085% of
the cumulative emissions are before the converter lights-off
which takes about 1080 s depending on the catalyst loading,
location and other factors. This light-off time also depends on the
thermal capacitance of the converter and the exhaust gas tem-
perature and velocity at the inlet of the converter. Metallic
substrates have a higher heat capacity compared to ceramic for
the same volume but the total thermal capacitance of the metallic
monolith may be comparable or lower compared to ceramic
monoliths due to thin walls. While high substrate conductivity
leads to superior steady state performance, it may lead to longer
transient times. Thus, the effect of substrate material on the
overall performance of a monolith is not clear. There are few
studies, mostly experimental, in the literature comparing the
performance of ceramic and metallic monoliths. We review here
briey the literature related to this topic.
Nishizawa et al. (1989) found that the conversion efciency
of metallic converter is better than that of ceramic converter of
equal cell density. In contrast, Jasper et al. (1991) reached the
opposite conclusion, while Pfalzgraf et al. (1996) concluded that
the type of substrate does not signicantly affect the conversion
efciency of the converter. Pannone and Mueller (2001) com-
pared the ceramic and metallic converters with the same volume
and cell density, and concluded that the ceramic substrate has
superior conversion efciency relative to the equivalent cell
density metallic substrates of either equal volume or equivalent
surface area. Santos and Costa (2008) compared the performance
of the ceramic and metallic converters for several vehicle
operating conditions. They concluded that ceramic substrate
performs better compared to metallic at low space velocities,
and vice versa for high space velocities. From these few studies
in the literature, it is clear that there is no general agreement
among the authors about the effects of substrate material on the
converter performance.
We also note that there are a few modeling studies comparing
the effect of substrate material on the steady-state and transient
performance of monoliths. Ramanathan et al. (2004a) examined
the effect of solid conduction on the hysteresis locus and showed
that the precious/Platinum group metal (PGM) loading need to
obtain light-off in a metallic monolith is about a factor of 10 times
smaller compared to a ceramic monolith (with all other para-
meters xed). In a separate study, Ramanathan et al. (2004b) also
presented simulations of the transient behavior and showed that
for the case of back-end ignition, the cumulative emissions may
be lower for metallic monoliths. However, for the case of front-
end ignition, the light-off time is shorter for ceramic monoliths
leading to lower cumulative emissions compared to metallic
monoliths. These as well as other prior modeling studies in the
literature were mainly concerned with emissions from automo-
biles. Recently, there is also interest in using monoliths in
catalytic combustion, partial oxidations as well as other hetero-
geneous catalysis applications in place of conventional reactors
like packed-beds. For example, in recent work, Maestri et al.
(2005) compared various supports (monoliths, foams and packed-
beds) for the conversion of methane to syngas by partial oxida-
tion. In such cases, the transients usually play a minor role
compared to the steady-state performance.
The main goal of this work is to analyze in some detail the
impact of substrate material on the steady-state and transient
behavior of the monolith reactors used in the after-treatment as
well as other applications cited above. This paper is organized as
follows. In the next section, we present a one-dimensional two-
phase model of the monolith reactor with position dependent
transfer coefcients. Since metallic monoliths used in applications
have sinusoidal channels for which the transfer coefcients are not
readily available in the literature, we present some computed new
results on the asymptotic Nusselt and Sherwood numbers for
sinusoidal channels of various aspect ratios. In Section 3, we present
a detailed analysis of the steady-state behavior of the monolith and
the effect of substrate conductivity on the ignition and extinction
points. In Sections 4 and 5, we analyze the transient behavior of the
monolith and the dependence of the cold-start or cumulative
emissions on the substrate properties. In Section 6, we present
numerical simulations that are guided by the analysis. The main
results of this work are summarized in Section 7.
2. Mathematical model
We consider a one dimensional two-phase mathematical model
of a straight channeled, washcoated catalytic monolith with the
following assumptions: (i) conduction/diffusion in the axial direc-
tion is negligible in the uid phase, (ii) the ow in the channel is
laminar (either fully developed or developing), (iii) the pressure
drop in the channel is small so that its effect on the concentration
and velocity may be neglected, (vi) constant physical properties,
(v) uniform channel cross-sectional area, (vi) the reaction occurs
only in the washcoat and catalyst activity may be assumed to be
constant within the time frame of the simulations, (vii) the inlet gas
velocity is constant (independent of time), (viii) heat losses are
neglected and the reactor is considered adiabatic, and (ix) the
kinetics is linear in the concentration or mole fraction of the
limiting reactant. The last assumption, though not necessary, allows
us to solve the diffusion-reaction equation in the washcoat analy-
tically to obtain the effective reaction rate in terms of cup-mixing
uid phase concentration. The gas velocity is not constant in a real
TWC (and varies typically between 0.5 m/s and 10 m/s during the
FTP cycle) but this assumption is made to simplify the analysis.
With the above assumptions, the mathematical model is described
by the following equations and boundary conditions:
@C
f
@t
= /uS
@C
f
@x

k
c
(x)
R
O
(C
f
C
s
), (1)
@T
f
@t
= /uS
@T
f
@x

h(x)
r
f
c
pf
R
O
(T
s
T
f
), (2)
k
c
(x)(C
f
C
s
) =a
/
(x)d
c
ZR
v
(C
s
,T
s
), (3)
(r
c
d
c
c
pc
r
s
d
s
c
ps
)
@T
s
@t
= (d
c
k
c
d
s
k
s
)
@
2
T
s
@x
2
h(x)(T
s
T
f
) a
/
(x)d
c
Z(DH
R
)R
v
(C
s
,T
s
), (4)
C
f
=C
in
(t); T
f
=T
in
(t) at x =0, (5)
@T
s
@x
=0 at x =0, L, (6)
C
f
(x) =C
f 0
(x); T
f
(x) =T
f 0
(x); T
s
(x) =T
s0
(x) at t =0: (7)
The diffusion and reaction in the washcoat are described by the
following expressions:
D
e
@
2
C
@y
2
=R
v
(C,T
s
); 0oyod
c
, (8)
S.R. Gundlapally, V. Balakotaiah / Chemical Engineering Science 92 (2013) 198210 199
C =C
s
at y =0;
@C
@y
=0 at y =d
c
: (9)
In the above equations, C
f
(T
f
) is the concentration (temperature) in
the uid phase, C
in
(T
in
) is the inlet uid concentration (tempera-
ture), C
s
is the concentration at the uidwashcoat interface, T
s
is
the solid temperature, /uS is the average uid velocity, R
O
is the
effective transverse length scale (R
O
=D
h
=4 =A
O
=P
O
, where A
O
is
the channel cross-sectional area open to ow and P
O
is the wetted
perimeter), L is the channel length, r
f
(r
c
,r
s
) and c
pf
(c
pc
,c
ps
) are the
density and specic heat capacity of the uid (catalyst and
substrate), k
c
(x)[h(x)] is the position dependent mass [heat] transfer
coefcient, a
/
(x) is the local catalyst activity, k
s
(k
c
) is the thermal
conductivity of the substrate (catalyst), d
c
is the thickness of
washcoat layer, d
s
is the half-thickness of the substrate wall,
DH
R
( o0) is the heat of the reaction, and Z is the effectiveness
factor.
In this work, we consider only the case of linear kinetics (rst-
order reaction) and hence take the reaction rate to be given by
R
v
(C,T
s
) =A
0
exp
E
R
g
T
s
_ _
C,
where A
0
is the pre-exponential factor, E is the activation energy
of the reaction, and R
g
is the universal gas constant.
The above equations are non-dimensionalized with the follow-
ing dimensionless variables and parameters:
t =
t/uS
L
, c
m
=
C
f
C
ref
, g =
E
R
g
T
ref
, y =g
TT
ref
T
ref
_ _
,
s =
(r
c
c
pc
d
c
r
s
c
ps
d
s
)
r
f
c
pf
R
O

r
w
c
pw
d
w
r
f
c
pf
R
O
,
Pe
h
=
/uSLr
f
c
pf
R
O
(k
c
d
c
k
s
d
s
)

/uSLr
f
c
pf
R
O
k
w
d
w
, DT
ad
=
(DH
R
)C
ref
r
f
c
pf
, B =g
DT
ad
T
ref
,
Da =
Ld
c
R
v
(C
ref
,T
ref
)
/uSC
ref
R
O
, L=
D
m
d
c
D
e
R
O
, f
2
s
=
R
O
d
c
R
v
(C
ref
,T
ref
)
D
m
C
ref
=PDa,
P =
/uSR
2
O
LD
m
, Le
f
=
k
f
r
f
c
pf
D
m
,
Sh
O
(z) =
k
c
(Lz)R
O
D
m
=
Sh
H2
(z)
4
, Nu
O
(z) =
h(Lz)R
O
k
f
=
Nu
H2
(z)
4
:
The state variable y
f
represents the dimensionless temperature of
the uid and c
m
is the dimensionless concentration (or mole
fraction) of the reactant in uid phase, while y
s
is the dimension-
less solid temperature. The function a(z) is the dimensionless
axial catalyst activity distribution. The axial Peclet number (Pe
h
)
is one of the most important dimensionless groups that inuences
the steady-state as well as the transient behavior of the monolith.
We note that Pe
h
is a monotonically decreasing function of the
product k
s
d
s
. The transverse Peclet number (P), which is the ratio
of transverse diffusion time to the convection (residence) time,
determines the conversion in the monolith in the mass transfer
controlled regime. It also determines the interphase temperature
and concentration gradients within the monolith channels and
the nature of ignition (e.g. front versus back end). The parameter
L determines the magnitude of diffusion limitations in the
washcoat, while f
2
s
determines the importance of external mass
transfer compared to reaction. Other important parameters are
the dimensionless adiabatic temperature rise (B), the dimension-
less activation energy (g), the heat capacity ratio (s), the
Damk ohler number (Da), the Sherwood number (Sh
O
) and the
Nusselt number (Nu
O
). The last two parameters describe the
variation of the heat and mass transfer coefcients with position
along the monolith. The heat capacity ratio (s) determines the
speed of the temperature fronts as well as the transient time.
[Here, T
ref
is the reference temperature, which in this work is
taken to be 500 K when the inlet temperature is varying. For the
case of constant inlet uid temperature, T
ref
can be taken as the
inlet uid temperature. The reference concentration C
ref
is taken
as the inlet concentration of the limiting reactant when it is
constant. It should also be noted that while Nu
O
(z)=Sh
O
(z) appear
naturally in the model equations, the correlations given below
and in the literature are for Nu
H2
(z)=Sh
H2
(z)].
By introducing the above dimensionless variables and para-
meters, the model may be expressed as
@c
m
@t
=
@c
m
@z
a(z)Dar
e
(c
m
,y
s
), (10)
@y
m
@t
=
@y
m
@z

Le
f
Nu
O
(z)
P
(y
s
y
m
), (11)
s
@y
s
@t
=
1
Pe
h
@
2
y
s
@z
2

Le
f
Nu
O
(z)
P
(y
s
y
m
)a(z)BDar
e
(c
m
,y
s
), (12)
with inlet, initial and boundary conditions:
c
m
(0,t) =c
in
(t), y
m
(0,t) =y
in
(t), (13)
c
m
(z,0) =c
0
(z), y
m
(z,0) =y
m0
(z); y
s
(z,0) =y
s0
(z), (14)
@y
s
(0,t)
@z
=
@y
s
(1,t)
@z
=0: (15)
Here, r
e
(c
m
,y
s
), the effective dimensionless reaction rate, obtained
by solving the diffusion-reaction equation in the washcoat analy-
tically. It is given by the following expression:
r
e
(c
m
,y
s
) =
c
m
X
F(z)
tanh F(z)

F(z)
2
LSh
O
(z)
, (16)
where F(z) =f
s

a(z)LX
_
and X =exp(y
s
=(1(y
s
=g))). Unless
stated otherwise, in the present work, we conne only to the
case of uniform activity and hence take the function a(z) to
be unity.
2.1. Heat and mass transfer correlations
Heat and mass transfer coefcient values must be accurate to
capture all the qualitative features of a monolith using one
dimensional two-phase models. It should be pointed out that
the heat and mass transfer correlations in the presence of a
chemical reaction on the wall are bounded by two limiting cases
(Gupta and Balakotaiah, 2001; Ramanathan et al., 2003). When
the reaction is very slow, the dimensionless transfer coefcients
(or the Nusselt/Sherwood numbers) approach the constant ux
limit denoted by Nu
H
=Sh
H
(for circular or parallel plate channels)
or Nu
H
2
=Sh
H
2
for channels of non-circular cross-section. The so
called axially constant ux limit, denoted by Nu
H
1
=Sh
H
1
, may be
relevant when the axial and peripheral conduction/diffusion in
the solid are signicant. In general, except for symmetric channel
geometries (circular and parallel plate), Nu
H
2
aNu
H
1
. For the most
common non-symmetric channel geometries of interest (e.g.
triangular, sinusoidal and square or rectangular), Nu
H
1
4Nu
H
2
and the difference between these two values is an indication of
the non-uniformity of ignition in the circumferential direction. In
this work, we ignore the variations in the circumferential direc-
tion and use the constant ux transfer coefcients Nu
H
2
=Sh
H
2
in
the model. [This may be justied by noting that the addition of
washcoat reduces the corner effect and hence the difference
between Nu
H
1
and Nu
H
2
becomes smaller.] In addition to these
slow reaction limits, when the reaction is fast, the heat and mass
transfer coefcients reach the so called constant wall temperature
or concentration limit denoted by Nu
T
=Sh
T
. Thus, depending on
S.R. Gundlapally, V. Balakotaiah / Chemical Engineering Science 92 (2013) 198210 200
the values of the reaction and other parameters, the Nusselt
and Sherwood numbers are bounded between Nu
H
2
=Sh
H
2
and
Nu
T
=Sh
T
.
Several theoretical and experimental correlations are proposed
in the literature for estimating the values of heat and mass
transfer coefcients in the entry region. Recently Gundlapally
and Balakotaiah (2011) proposed simple and accurate correla-
tions for estimating the heat and mass transfer coefcients for
fully developed as well as simultaneously developing ows in an
arbitrary geometry. For the case of fully developed velocity
prole, but developing concentration and temperature proles,
they proposed the following correlations:
Sh
T
(z) =Sh
To

0:068(f Re)
1=3
o
P
z
10:063
P
z
_ _
2=3
, (17)
Sh
H2
(z) =Sh
H2o

0:108(f Re)
1=3
o
P
z
10:083
P
z
_ _
2=3
, (18)
Here, f Re is the product of friction factor times Reynolds number
for the channel. When the velocity prole is also developing along
with concentration and temperature proles, the following cor-
relations are proposed:
Sh
T
(z) =Sh
To

0:44 Sc
1=6
P
z
10:314
P
z
_ _
1=2
, (19)
Sh
H2
(z) =Sh
H2o

0:98 Sc
1=6
P
z
10:512
P
z
_ _
1=2
: (20)
Similar expressions may be written down for Nu
H2
(z) and Nu
T
(z)
with P replaced by P=Le
f
and the Schmidt number (Sc) by the
Prandtl number (Pr). Gundlapally and Balakotaiah (2011) com-
pared the above correlations with the most widely used correla-
tions in the literature and numerical data, and found them to be
very accurate. It was shown that the bifurcation behavior, such as
number of steady states and number and location of ignition/
extinction points is sensitive to the ow conditions in the entry
region, especially for large values of B and PZ1. However, for
small values of P (r0:05) and B(r10), typical of after-treatment
applications, the use of constant asymptotic values for Nu and Sh
may be justied. We use the above correlations corresponding to
simultaneously developing ows in the simulations presented in
this work but use constant values in the analysis of simplied
cases.
It should also be pointed out that the transfer coefcients
always start with the value corresponding to the constant ux
boundary condition (slow reaction) at the channel inlet, and after
ignition jump to a value corresponding to constant temperature/
concentration boundary condition (fast reaction) as we move in
the axial direction. In order to incorporate the jump in the values
of transfer coefcients during the numerical simulations with the
one dimensional two-phase model, one has to keep track of the
magnitude of the second term in the denominator of Eq. (16).
Further, in the ignited region the monolith may attain the
external mass transfer limited asymptote if the inlet temperature
of the gas or the adiabatic temperature rise is high. In this region,
the exit conversion may be approximated by
z =1c
m
(1) -a
1
exp
Sh
To
4P
_ _
; (P51), (21)
where a
1
and Sh
T
are the Fourier coefcient and asymptotic
Sherwood number, respectively. Since these values are not avail-
able in the literature for sinusoidal channels of different aspect
ratios and most metallic monoliths are of this type, we have listed
some computed values in Table 1. We have also listed the
asymptotic Nusselt (Sherwood) numbers Nu
To
, Nu
H1o
, Nu
H2o
,
and the friction factor. It should be noted that these values are
accurate only for the limiting case of very thin washcoat or when
the channel shape or hydraulic diameter is not impacted by the
addition of the washcoat. When the addition of washcoat changes
the hydraulic diameter as well as the shape of the ow area
substantially (e.g. from sinusoidal to a rounded triangle), these
values need to be adjusted accordingly. [Remark: Most values of
a
1
and Sh
T
listed in earlier literature, including those by
Ramanathan et al. (2003) were for another type of sinusoidal
channel, which is not common in practice. The values listed in
Table 1 are for the common sinusoidal channels whose shape is
Table 1
Asymptotic values of Nusselt/Sherwood numbers and friction factors for commonly used channel geometries.
Channel shape D
h
(4R
O
) (f Re)
o
Nu
H2o
(Sh
H2o
) Nu
H1o
(Sh
H1o
) Nu
T
(Sh
T
) a
1
(Sc = o)
2a 16 4.364 4.364 3.656 0.8190
2a 14.23 3.089 3.608 2.977 0.8074
1.5419a 14.574 0.95 3.311 2.496 0.5810
1.418a 14.023 1.38 3.267 2.536 0.6094
1.2103a 13.023 1.55 3.102 2.476 0.6378
1.0466a 12.234 1.34 2.916 2.353 0.6485
0.8118a 11.207 0.90 2.617 2.110 0.6490
0.4673 10.123 0.33 2.213 1.719 0.6129
S.R. Gundlapally, V. Balakotaiah / Chemical Engineering Science 92 (2013) 198210 201
bounded by the line y=0 and y(z) =2b sin
2
(pz=2a),0rzr2a,
where b=a is the channel aspect ratio].
3. Effect of substrate material on the steady-state behavior
3.1. Steady-state bifurcation diagrams
Before we present numerical simulations, we analyze the
effect of solid conduction on the steady-state behavior of the
monolith using simplied versions of the above general model.
We consider rst the limiting case of zero solid conduction and
uniform catalyst activity. Further we make the assumption of
constant asymptotic values for the transport coefcients, use the
positive exponential approximation (g-o) and neglect washcoat
diffusional limitations (L-0). With these assumptions, the
steady-state model may be simplied to the following pair of
differential-algebraic system of equations:
dy
m
dz
=
Nu
Oo
Le
f
P
(y
s
y
m
); y
m
(0) =y
m,in
, (22)
Nu
Oo
Le
f
P
(y
s
y
m
) =
Da(By
m
y
m,in
) exp[y
s
]
1
PDa exp[y
s
]
Sh
Oo
_ _ : (23)
In the limit of P-0, the difference between the solid and uid
temperatures vanishes (y
s
=y
m
=y) and this model reduces
further to the homogeneous plug ow reactor model:
dy
dz
=Da(Byy
m,in
)exp[y]; y(0) =y
m,in
: (24)
A plot of exit temperature y(z =1) (or conversion, z = (y(1) y
m,in
)=B)
versus Da of Eq. (24) has an inexion point at (Balakotaiah, 1996)
y(1) -B1y
m,in
, (25)
BDa exp(y
m,in
) -1: (26)
It may be shown that for nite but small values of the solid
conductivity (1=Pe
h
51) and interphase gradients (P51) the
above model exhibits an ignition and extinction point provided
Pe
h,eff
o(2=B)exp(B2), where the effective heat Peclet number is
dened by
1
Pe
h,eff
=
1
Pe
h

P
Nu
Oo
Le
f
: (27)
Further, in this case, the light-off (ignition point) location as well
as the extinction point are very close to the above inexion point
of the plug ow model. [For a more detailed analysis of the
bifurcation behavior of the model dened by Eqs. (22) and (23),
we refer to the articles by Balakotaiah, 1996, Dommeti et al.,
1999, and Gupta et al., 2001.]
At the other extreme case of innite solid conduction, with the
same assumptions as stated above, the steady-state model of the
monolith may be simplied to an algebraic equation for the solid
temperature:
y
m,in
=y
s

B
1exp
Nu
Oo
Le
f
P
_ _ _ _ 1exp
Da exp[y
s
]
1
PDa exp[y
s
]
Sh
Oo
_ _
_

_
_

_
_

_
_

_
:
(28)
In this limit, the solid temperature is independent of axial
position in the monolith but the uid temperature and conversion
vary along the length. This model in which plug ow is assumed
for the moving phase and no gradients exist in the stationary
phase is referred to as the lumped thermal model in the literature.
It is a very good approximation to describe metallic monoliths.
[Remark: A bifurcation diagram y
s
versus y
m,in
of Eq. (28) may be
plotted easily as it is explicit in the inlet uid temperature.] In the
limit of negligible interphase gradients or P-0, this lumped
thermal model simplies further to the homogeneous limit:
y
m,in
=yB[1exp{Da exp y]]: (29)
This model has been analyzed in the literature in the context of
multiple steady-states in bubble column reactors (Huang and
Varma, 1981; Luss and Balakotaiah, 1984). The hysteresis locus
(where the ignition and extinction points coincide or the bound-
ary between light-off and no light-off) of this model may be
determined to be
B =e =2:718, (30)
y =B1y
m,in
, (31)
BDa exp(y
m,in
) =exp(2e) -0:488: (32)
However, unlike the low conduction case, the ignition and
extinction points of this model (for B4e) are well separated
and may be written in parametric form:
y
m,in
=ln
u
Da
_ _

(e
u
1)
u
, (33)
B =
e
u
u
, 0ouoo: (34)
The ignition locus (for u-0) may be approximated by
BDa exp(y
m,in
) =e
1
=0:368, (35)
while the extinction locus may be expressed as
BDa exp(y
m,in
) =exp(uB); B =
e
u
u
: (36)
Fig. 1 shows a plot of the bifurcation set (ignition and extinction
locus) given by Eqs. (33) and (34) for Da=0.14 for the two
extreme cases of zero (Pe
h
-o) and innite solid conduction
(Pe
h
-0). [As stated earlier, for the case of Pe
h
= o, the ignition
and extinction points coincide at the inexion point given by Eq.
(26).] Comparing Eqs. (35) and (26), we note that the ignition/
light-off point moves to lower temperatures (by about one unit in
y
m,in
, which corresponds to about 20 1C] as the solid conduction
increases. However, the most important result from the above
10 8 6 4 2 0
2
4
6
8
10
12
14

m,in
B
Da=0.14
Extinction locus
Ignition locus
Pe
h
=
Pe
h
=0
Fig. 1. Bifurcation set of the lumped thermal model of the monolith in the inlet
temperature (y
m,in
)-inlet concentration (B) plane. The dashed line is the light-off
locus of the zero conduction model.
S.R. Gundlapally, V. Balakotaiah / Chemical Engineering Science 92 (2013) 198210 202
analysis is the impact of solid conduction on the extinction point,
or how far the ignited (high conversion) branch extends in the
steady state diagram of conversion versus inlet uid temperature
for a xed precious group metal (PGM) loading, or equivalently,
the PGM loading needed to obtain an ignited (high conversion)
branch for the same inlet temperature. This can be seen better in
Fig. 1 which shows a plot of the bifurcation set given by Eqs. (33)
and (34) for Da=0.14. For values of B that are typical in the after-
treatment applications (814), the separation between the igni-
tion and extinction points is about 36 units in dimensionless
temperature (Fig. 1). Thus, the above analysis leads to the
following conclusions: (i) for the case of high solid conduction,
an ignited branch exists for temperatures that are lower by about
60120 1C compared to the case of very low solid conduction,
(ii) for the same inlet temperature, the PGM loading needed to
obtain an ignited branch is about 20 ( -e
3
) to 400 ( -e
6
) times
smaller for the case of high conduction (metallic monoliths)
compared to the low conduction case (ceramic monoliths).
[Remark: The ignition and extinction loci shown in Fig. 1 are for
a xed reactor size or residence time (Da). If the reactor size or
residence time changes, all the curves in Fig. 1 shift horizontally
by an amount Ln(0:14=Da). For example, if the residence time is
reduced by a factor 20 and other parameters are unchanged, the
curves shift about three units to the right in dimensionless
temperature.]
The above analysis and conclusions may be veried using
numerical simulations of the full model. We list in Table 2 the
numerical values used in the simulations. The corresponding
values for the various dimensionless groups are listed in
Table 3. In selecting these values, we have assumed a cell density
of 400 cpi (or a channel hydraulic diameter of 1105 mm) and a
mean washcoat thickness of 25 mm. Thus, the hydraulic diameter
of the open channel is 1055 mm or R
O
-264 mm. The values
selected for the half-thickness of the wall (d
s
), thermal conduc-
tivity (k
s
) and heat capacity per unit volume (r
s
c
ps
) correspond to
ceramic and metallic substrates, respectively. [Remark: The k
s
value of 50 W/(m K) used for metallic substrate is an upper bound
which is attained by using some alloys or steels.]
Fig. 2 shows steady-state bifurcation diagrams of solid tem-
perature and exit uid conversion as a function of inlet uid
temperature for two different values of the heat Peclet number
(Pe
h
=279 and 23:5), corresponding to ceramic and metallic
monoliths, respectively. In this specic case, the ceramic monolith
does not exhibit any ignition or extinction points but the
temperature and conversion rise sharply at the approximate
location of the inexion point predicted by Eq. (26). In contrast,
the metallic monolith exhibits ignition and extinction points but
the separation between them is only about 0.5 unit in the
dimensionless temperature (which corresponds to about 10 1C).
We note that for the parameters selected, the washcoat diffusion
has a strong impact on the ignited steady-state branch. When the
washcoat diffusion is reduced (by changing the value of L from
1.89 to 0.001), the light-off or ignition point does not move but
the extinction point (and hence the ignited branch extends to
lower inlet gas temperatures (Fig. 3). [Note: There is also a very
small region in Fig. 3 in which ve steady-states exist.] If the heat
Peclet number is reduced further by reducing the monolith length
by about a factor ve but increasing the cross-sectional area by
Table 2
Typical values of parameters used in the simulations.
Variable Value
/uS 2 m/s
L 0.076 m
R
O
264 mm
ds
25 mm (metallic), 82:55 mm (ceramic)
dc
25 mm
cpc 1000 J/kg K
cps 460 J/kg K (metallic), 1000 J/kg K (ceramic)
r
c
1300 kg/m
3
r
s
7500 kg/m
3
(metallic), 2500 kg/m
3
(ceramic)
r
f
0.7 kg/m
3
n
3:78 10
5
m
2
=s
c
pf
1068 J/kg K
k
f
0.0386 W/m K
ks
50 W/m K (metallic), 1 W/m K (ceramic)
kc
1 W/m K
Dm 5 10
5
m
2
=s
De
Dm
20
A
0 10
11
s
1
E=Rg 10 825 K
(DH
R
) 300 kJ/mol
C
in
0.6 mol/m
3
T
s0
298 K
T
ref
500 K
Table 3
Values of dimensionless groups used in the simulations.
Dimensionless parameter Value
g 21.7
B 10.4
f
2
s
0.0052
L 1.89
Le
f
1.03
P 0.037
Pe
h
23.5 (metallic), 279.0 (ceramic)
s 602.0 (metallic), 1210.0 (ceramic)
y
s,0
8.74
Pr 0.73
Sc 0.76
Nu
H2o
3.6
Sh
H2o
3.6
1 0.5 0 0.5
0
5
10

s
(
1
)
1 0.5 0 0.5
0
0.2
0.4
0.6
0.8
1

m,in
E
x
i
t

c
o
n
v
e
r
s
i
o
n
Metallic
Metallic
Ceramic
Ceramic
Fig. 2. Bifurcation diagrams of exit solid temperature and reactant conversion
versus inlet uid temperature showing the effect of substrate material on the
region of multiple solutions. Parameter values used are listed in Table 3.
S.R. Gundlapally, V. Balakotaiah / Chemical Engineering Science 92 (2013) 198210 203
the same factor so that the volume remains constant (but the gas
velocity through the channels is reduced by a factor ve so that
the residence time or space velocity as well as other parameters
listed in Table 3 remain unchanged), the region of multiple
steady-states expands. For example, when Pe
h
is reduced from
23.5 to 1, as shown in Fig. 4, the region of multiple solutions
expands to about four units in the dimensionless temperature
(about 80 1C ), even in the presence of strong washcoat diffusion
(and expands even further by 1.5 units if washcoat diffusion is
eliminated). When this is contrasted with the case of Pe
h
=279
(also shown in Fig. 4), the profound impact of the substrate
material conduction and monolith aspect ratio on the region of
multiplicity, predicted by the above analysis is clearly conrmed.
Fig. 5 shows the temperature proles of the solid for three
different inlet uid temperatures on the bifurcation diagrams
shown in Fig. 2. We note that for y
m,in
= 0:5, only a quenched
steady-state exists for ceramic substrate while an ignited state
with about 80% of the length in the high temperature region
exists for the metallic substrate. Even when an ignited state exists
for both cases (for y
m,in
=0:5), we note that only about 70% of the
length is ignited for ceramic case while 100% of the length is
ignited for the metallic case.
To further illustrate the impact of solid conduction and
entrance region effects on the number of steady-states, we con-
sider a different set of dimensionless parameters, namely the
case of B =15,g =25,Le
f
=1,f
2
s
=0:0015,L=0:001,P =0:5 and
Pe
h
=300, for which the steady-state bifurcation diagram of exit
conversion versus inlet temperature is shown in Fig. 6. We note
the existence of ve steady-states, of which three are stable. The
solid temperature proles of the three stable steady-states for the
case of y
m,in
=2 are shown in the bottom diagram of Fig. 6. We
note that there are two ignited states, one in which the entire
monolith is ignited and another where only about 80% of the
length is ignited. The existence of more than one ignited branch
for the case of an adiabatic monolith reactor with rst-order
kinetics and uniform catalyst loading is a novel result, which has
not been reported in the prior monolith modeling studies in the
literature. From the proles shown in Fig. 6, it is clear that the
existence of additional steady-states is due to the position
dependency of the heat and mass transfer coefcients.
4. Analysis of transient behavior and temperature
front propagation
In this section, we review the conditions for back end and front
ignition in monoliths and analyze the impact of conduction in the
substrate material on the temperature front propagation.
Fig. 4. Bifurcation diagram of exit solid temperature versus inlet uid tempera-
ture showing the effect of monolith substrate material and aspect ratio on the
region of multiple solutions. The case Pe
h
=1 corresponds to a high aspect ratio
metallic monolith while Pe
h
=279 to that of a low aspect ratio ceramic monolith
with same volume. Other parameter values used are listed in Table 3.
0 0.2 0.4 0.6 0.8 1
2
0
2
4
6
8
10
12
z

m,in
=0

m,in
=0.5

m,in
=0.5

m,in
=0

m,in
=0.5

m,in
=0.5
Fig. 5. Steady-state temperature proles of the solid for different inlet uid
temperatures. Parameter values are same as in Fig. 2. Solid lines denote the
ceramic substrate and dashed lines denote the metallic substrate.
1.5 1 0.5 0 0.5
2
0
2
4
6
8
10
12

m,in

s
(
1
)
Ceramic
Metallic
Fig. 3. Bifurcation diagrams of exit solid temperature versus inlet uid tempera-
ture showing the effect of substrate material on the region of multiple solutions
for L=0:001 (negligible washcoat diffusional effect). Other parameter values used
are listed in Table 3.
S.R. Gundlapally, V. Balakotaiah / Chemical Engineering Science 92 (2013) 198210 204
4.1. Impact of substrate material on light-off
Ramanathan et al. (2003) derived an ignition/light-off criterion
from the one-dimensional two-phase model of the monolith.
In terms of the present notation (and when the reference
temperature is not equal to the feed temperature), this criterion
may be expressed as
1
f (j)
g(Pe
h,eff
)Bf
2
s
w
P

4eBf
2
s
w
Le
f
Nu
H,o
_ _
41, (37)
where
f (j) =
1, jo0:5,
2j, j40:5,
_
j=f
s

Lw
_
and (38)
w =exp
y
m,in
1
y
m,in
g
_
_
_
_
_
_
_
_
, (39)
g(Pe
h,eff
) = 1
Pe
h,eff
2
_ _
2=Pe
h,eff
: (40)
The rst term in Eq. (37) represents the ignition locus of the
pseudohomogeneous model and the second term represents the
heterogeneous contribution (when the interphase temperature
gradients are signicant). Light-off will occur at the back end of
the monolith if the rst term in Eq. (37) just exceeds unity and the
second term is much less than unity. If the rst term is much
larger than unity while the second term is much less than unity
(typically in the homogeneous limit of f
2
s
and P51 but the ratio
Da =f
2
s
=P is nite), ignition will occur closer to the inlet at a
dimensionless location z
ig
-1=BDaw for the case of no conduction
in the substrate (Pe
h
-o) or ignition is nearly uniform along the
length when the conduction in the substrate is dominant
(Pe
h
-0). However, if the second term in Eq. (37) exceeds unity,
light-off will occur at the front end irrespective of the magnitude of
the rst term. Middle ignition will occur if the sum exceeds unity
with the second term is less than unity and not negligible compared
to the rst term. The function f (j) describes the inuence of
diffusional limitations in the washcoat. Here, j is the washcoat
Thiele modulus. The function g(Pe
h,eff
) describes the inuence of
inter and intraphase heat transfer on light-off and changes from 1
(for Pe
h,eff
= o) to e =2:718 (for Pe
h,eff
=0). [Note: The function
g(Pe
h,eff
) is modied to a simpler form than that given by
Ramanathan et al., 2003. Also, Pe
h
is replaced by Pe
h,eff
to account
for interphase communication near the homogeneous limit.]
As pointed out by Ramanathan et al. (2003), the above
criterion was developed by assuming linear kinetics, negligible
reactant consumption (or Bb1), positive exponential approxima-
tion (gb1), and constant transfer coefcients. For typical values
of g and B encountered in after-treatment applications (e.g.
B =10:4,g =21:7), the accuracy of the predicted light-off bound-
ary can be improved by replacing the unity on the r.h.s of Eq. (37)
by (1(1=B)(2=g)). [This error is usually small and translates a
temperature difference of about 3 1C in the predicted light-off
temperature.] We also note that the above ignition criterion
describes the steady-state behavior and hence is independent of
initial solid temperature. In the homogeneous limit (P-0) and
high values of solid conduction (Pe
h
-o), the location of ignition
is independent of the initial solid temperature, but in all other
cases it depends on the initial solid temperature.
It follows from Eq. (37) that for the case of front-end ignition,
the substrate material has no impact on the light-off temperature.
However, for the case of back-end ignition, high solid conduction
can reduce the inlet temperature (y
m,in
) required for ignition by
about one unit (about 20 1C) for the case of no washcoat diffusion
and 0.5 unit (or 10 1C) when washcoat diffusion is strong.
For the parameter values listed in Table 2, the ignition criterion
given by Eq. (37) predicts back end ignition for metallic substrate at
y
m,in
- 0:5 (and for ceramic at y
m,in
- 0:4), which is veried by
the numerical calculations presented in Fig. 2. Further, for both
cases, front end ignition is predicted when y
m,in
42:0. Thus, for the
typical parameter values selected, the window of inlet tempera-
tures over which the light-off behavior changes from back to front
end is about 2.5 dimensionless units (about 50 1C).
4.2. Impact of substrate material on temperature front propagation
When the condition for back end ignition is satised, the
temperature front may or may not propagate toward the front of
the monolith. Here, we provide some guidelines to determine the
conditions under which the temperature front propagates to the front
of the monolith.
First, we note that when Pe
h
b1 and ignition occurs at the
back end and on the ignited segment, the solid temperature
attains values very close to the adiabatic temperature. However,
as the Pe
h
decreases or the solid conduction increases, the
temperature in the ignited region is below the adiabatic value
and this leads to propagation of the temperature front toward
the inlet of the monolith. When washcoat diffusion effect is
negligible (L51), there can be a second ignition at the inlet,
which raises the temperature to the adiabatic value and forward
propagation of a smaller temperature front. However, when
washcoat diffusion is strong, this second ignition and forward
propagation may be eliminated. By analyzing the model in the
traveling wave coordinate, the following expression may be
derived for the dimensionless speed of the upstream propagat-
ing temperature front (for the practical case of sb1):
v
wb
-
[B(y
i
y
m,in
)]
s(y
i
y
m,in
)
: (41)
1 1.5 2 2.5 3
0
0.5
1

m,in
E
x
i
t

c
o
n
v
e
r
s
i
o
n
0 0.2 0.4 0.6 0.8 1
5
10
15
z

s
Middle steadystate
Ignited steadystate
Quenched steadystate
Fig. 6. (a) Steady-state bifurcation diagram of conversion versus inlet uid tem-
perature, (b) solid temperature proles at three different stable steady-states for
y
m,in
=2. Parameter values: B=15, g =25,Le
f
=1,f
2
s
=0:0015,L=0:001,P =0:5,
and Pe
h
=300.
S.R. Gundlapally, V. Balakotaiah / Chemical Engineering Science 92 (2013) 198210 205
For example, for the typical values listed in Table 2 with B=10.4,
g =21:7, s =1210, Pe
h
=279, Da=0.1405 (P =0:037,f
2
s
=0:0052),
L=1:89 and y
m,in
=0, the solid ignites to y
i
-10:4 and the
temperature front (located at z
ig
-1=BDaw =0:68) does not propa-
gate upstream. (The corresponding steady-state prole is shown in
Fig. 5.) However, when the substrate material is changed to metallic
(Pe
h
=23:5, s =602), the back of the monolith ignites to y
i
-8:7
and the temperature front propagates all the way to the front and at
steady-state, the entire monolith is ignited (see Fig. 5 for the
corresponding steady-state prole). Eq. (41) predicts a front speed
of 0.00032 which is close to the value of 0.0003 obtained from
numerical simulations. [Note: As the front propagates upstream, y
i
increases from 8.2 to 9.2 and hence the front speed decreases as it
moves toward the inlet. The average value of y
i
-8:7 is used to
calculate the approximate front speed.] Thus, for the case of back
end ignition, the solid conduction has a strong impact on the
upstream propagation of the temperature front and hence cold-
start emissions.
It should be pointed out that for the case of front-end ignition,
the temperature front propagates in the ow direction at a
dimensionless speed given by
v
wf
=
1
1s
-
1
s
(for sb1): (42)
We note that this speed is independent of the reaction parameters
or the solid conductivity (but the width of the front depends on
the solid conductivity as well as reaction parameters). The speed
of the temperature front for the case of simple heating of the
monolith by the incoming gas (with no reaction in the monolith)
is also given by Eq. (42). For the numerical values listed in Table 2,
this speed is higher (by about a factor two) for metallic substrate
compared to the ceramic case.
5. Analysis of the effect of substrate on cold-start time
We now present a brief analysis of the effect of the substrate
material on the cold-start or heating time. As stated earlier, the
substrate material can inuence light-off time in two ways: through
the solid conduction and propagation of the temperature front
through solid heat capacitance. The reduction in the substrate wall
thickness reduces the thermal capacitance but also increases the
heat Peclet number. As shown above, for the case of back-end
ignition, the light-off front may not propagate upstream if the heat
Peclet number is high. However, for the case of front-end ignition,
the solid conduction has very little impact on the forward propaga-
tion of the temperature front and a lower solid heat capacitance
always leads to lower cumulative emissions. It should also be noted
that the initial heating of the cold monolith by the incoming hot gas
depends not only on the solid properties (thermal capacitance and
conductivity) but also on the interphase heat transfer. The latter
depends clearly on the inlet gas velocity as well as the channel
hydraulic diameter (or monolith cell density).
To simplify the transient analysis, we consider the limiting
case of small P, Pe
h
b1 and only the initial heating of the cold
monolith by the inlet gas (and neglect the reaction terms). We
also neglect the accumulation term in the gas phase balance and
assume constant values for the initial solid temperature and inlet
uid temperature. With these assumptions, the solid and uid
balances may be combined to obtain the following linear model
for the solid temperature:
s
@y
s
@t

@y
s
@z
=
1
Pe
h,eff
@
2
y
s
@z
2
; z40, t40, (43)
with initial and inlet conditions:
y
s
(z,0) =y
s0
, (44)
1
Pe
h,eff
@y
s
(0,t)
@z
=y
s
(0,t)y
m,in
: (45)
The solution of this model that describes the heating of the cold
monolith near the inlet may be approximated on a semi-innite
domain using Laplace or Fourier transform methods. The solid
temperature at the inlet may be expressed as
y
s
(0,t)y
s0
y
m,in
y
s0
= 1
l
2
_ _
erf

l
_
2
_ _

l
2

l
p
_
exp
l
4
_ _
; l =
tPe
h,eff
s
:
(46)
We can use this expression to estimate the heating time which is
an excellent approximation for the light-off time for the case of
front-end ignition. For example, the heating time required for the
front (z=0) solid temperature of monolith to reach a value that is
within 5% of the initial difference may be approximated by l -4:3
or
t =
/uSt
L
-
4:3s
Pe
h,eff
: (47)
Using this estimate, the cold-start emissions (CSE) through a
single channel may be expressed as
CSE =4pR
2
O
C
in
/uSt, (48)
CSE =17:2pC
in
r
c
c
pc
d
c
r
s
c
ps
d
s
r
f
c
pf
_ _
k
c
d
c
k
s
d
s
/uSr
f
c
pf

R
3
O
/uSr
f
c
pf
k
f
Nu
Oo
_ _
:
(49)
The following observations may be made from Eq. (49): (i) cold-
start emissions decrease with decrease in the channel hydraulic
diameter (for front-end ignition), (ii) cold-start emissions decrease
monotonically with decrease in the substrate thickness (d
s
),
thermal conductivity (k
s
) and volumetric heat capacity (r
s
c
ps
),
and (iii) cold-start emissions are not a monotonic function of the
inlet gas velocity (/uS). This last result is rather surprising but
may be explained by the fact that the effective conductivity of the
monolith due to both inter and intraphase gradients is not
monotonic with the gas velocity. For example, low gas velocities
lead to smaller interphase gradients but also smaller intraphase
(solid) gradients leading to longer heat up times. On the other
hand, high gas velocities lead to larger interphase as well as
intraphase gradients leading again to longer heat up times. The
minimum heating time is obtained when the interphase (gas to
solid) gradients are negligible but intraphase gradients (within
solid) are high (so that only a small front part of the solid needs to
be heated to the gas temperature before ignition). The optimum
gas velocity for heating may be estimated from Eq. (49) as
/uS
opt
=

(k
c
d
c
k
s
d
s
)
r
f
c
pf
R
3
O
k
f
r
f
c
pf
Nu
Oo
_ _

_
(50)
Substitution of the numerical values leads to /uS
opt
of about
0.5 m/s for ceramic and 1.8 m/s for metallic monoliths. We note
that these values fall within the range of the gas velocities
encountered in applications.
It should be pointed out that the existence of an optimum
velocity for the heating of the monolith has some design implica-
tions. For example, if the converter volume and gas inlet mass (or
volumetric) ow rate are xed, an optimum velocity implies that
there exists an optimum aspect ratio (diameter to length) of the
converter that leads to lowest cold-start emissions. This is in
contrast to the steady-state design where a high aspect ratio and
or high conductivity leads to a larger region of multiple solutions.
[Remark: This optimization of the monolith reactor aspect ratio
for the non-isothermal case is different from the optimization of
S.R. Gundlapally, V. Balakotaiah / Chemical Engineering Science 92 (2013) 198210 206
the channel aspect ratio for the isothermal case considered
recently by Mathieu-Potvin and Gosselin, 2012.]
6. Numerical simulation results on the effect of substrate
material on cumulative emissions
In this section, we present numerical results that validate the
analysis of the previous sections. The values of the various
parameters used in the numerical simulations are given in
Tables 2 and 3. As stated earlier, these values are typical in the
after-treatment applications.
Before presenting the simulations, it should be pointed out
that monoliths with ceramic substrate come with different
channel shapes with majority being the square, whereas, metallic
monoliths are usually made of sinusoidal shaped channels (and in
a few cases triangular shape). In the calculations of this section,
we assume that the channel shape (open ow area) for both the
substrates is smooth as it is known that washcoat tends to round
the corners. We also assume that Nu
Oo
=Sh
Oo
=0:9 (or
Nu
H2o
=3:6), corresponding to a rounded square or triangle.
Fig. 7 shows a plot of the cumulative emissions (for the rst
150 s) as a function of the inlet gas velocity for T
in
=500 K
(y
m,in
=0:0) and all other parameters as in Table 2. The cumulative
emissions (CE) are calculated based on the assumption of 12,400
channels (corresponding to a converter volume of 1.52 l), an
average reactant molecular weight of 28 and inlet reactant
concentration of 0.6 mol/m
3
. As discussed previously, for this inlet
temperature, we have back end ignition with the entire monolith
ignited for the case of metallic substrate while only the back 32%
is ignited for the ceramic case. Further, using the light-off criterion
given by Eq. (37), it may be calculated that the monolith with
ceramic substrate does not light-off for gas velocities greater than
about 3.5 m/s while that with metallic substrate ignites for gas
velocities as high as 9 m/s. Thus, for the case of back end ignition,
monoliths with metallic substrates are clearly superior (lower CE)
for gas velocities within the practical range.
Fig. 8 shows a plot of the cumulative emissions as a function of
inlet gas velocity for the two base cases metallic [50 W/(m K)] and
ceramic [1 W/(m K)] as well as two metallic substrates having
the same wall thickness but lower conductivities of 5 W/(m K)
and 15 W/(m K), respectively. Here, all other parameters (except
the inlet velocity) are as listed in Table 2 and the inlet gas
temperature is taken as T
in
=600 K (y
m,in
=4:34), leading to front
end ignition in all cases. First, we note that the CE are lower for
ceramic substrate for gas velocities below about 1.3 m/s while
they are lower for metallic monoliths at higher gas velocities. It is
clear from these computations that metallic substrates having
intermediate values of conductivity lead to lower cumulative
emissions (by about a factor two) in the entire practical range
of gas velocities (/uSZ2 m=s). [Remark: The monolith will
extinguish at sufciently high inlet gas velocities, which in this
case occurs at /uS4130 m=s for ceramic monolith and even
higher velocities for the metallic case.]
Fig. 9 shows a plot of the CE as a function of time for three
different inlet gas velocities of 0.5, 1.0 and 2.0 m/s, respectively. Here,
all other parameters (except the inlet velocity) are as listed in Table 2
and the inlet gas temperature is taken as T
in
=600 K (y
m,in
=4:34).
As discussed previously, this leads to front end ignition in both cases.
For the case of u=2 m/s, the heating time for ceramic substrate is
higher compared to metallic leading to higher CE. The converse is true
for the case of u=0.5 m/s and u=1 m/s leading to lower CE for
ceramic substrate. As noted by Ramanathan et al. (2004b), for this
inlet temperature, the cumulative emissions versus time curve
consists mostly of two asymptotes, namely the cold-start asymptote
and a steady-state asymptote of very small slope (as the conversion of
the reactant in the steady-state is nearly 100%). Thus, for the case of
front-end ignition (with high inlet temperature), the CE are mostly
due to the heating of the cold monolith by the incoming gas, and the
properties of the substrate material have a strong impact on the
emissions.
The simulations presented so far were limited to constant inlet
gas velocity and constant inlet temperature. Now we consider the
effect of varying inlet gas temperature and velocity and compare
the performance of ceramic and metallic substrates. Fig. 10 shows
the exhaust gas temperature and velocity at the inlet of a three-
way converter during the cold start driving cycle of FTP-75 test
0 1 2 3
0
10
20
30
40
50
60
70
Inlet Velocity (m/s)
C
u
m
u
l
a
t
i
v
e

E
m
i
s
s
i
o
n
s

(
g
)
Metallic
Ceramic
Fig. 7. Plot of nal cumulative emissions as a function of gas velocity for metallic
and ceramic substrates for inlet uid temperature of 500 K. Other parameter
values used are listed in Table 2.
0 1 2 3 4
1
2
3
4
5
6
7
Inlet velocity (m/s)
C
u
m
u
l
a
t
i
v
e

e
m
i
s
s
i
o
n
s

(
g
)
Ceramic
Metallic
k
s
=15 W/(m.K)
k
s
=5 W/(m.K)
Fig. 8. Cumulative emissions as a function of gas velocity for ceramic and metallic
monoliths with different solid conductivities for inlet uid temperature of 600 K.
Solid line indicates the ceramic substrate (ks =1 W=m K) and dashed and dotted
lines indicate metallic substrate.
S.R. Gundlapally, V. Balakotaiah / Chemical Engineering Science 92 (2013) 198210 207
for a typical vehicle during the rst 100 s [Note: This data is taken
from the recent work of Kumar et al. (2012).] We see that the
exhaust gas temperature rises rapidly to about 700 K in about 30 s
and is nearly constant at about 650 K at longer times. The rapid
increase in the inlet temperature to above 600 K initiates front end
ignition in the monolith for the typical values (Table 2) used in this
work. While the gas velocity varies from about 0.9 to 4 m/s, for
calculating the cumulative emissions, we take an average value of
2 m/s, as the temperature effect is much stronger than the ow
rate. Further, we approximate the temperature variation as a
linear ramp from 300 K to the nal value. This approximation (of
constant inlet velocity and a linear temperature ramp) describes
the cold-start behavior of many after-treatment systems.
Fig. 11 shows the calculated cumulative emissions (with
/uS=2 m=s and linearly varying temperature from 300 to 500,
550, 650 K in 30 s) for the two substrates as a function of time.
[Remark: The high inlet gas temperature corresponds to a TWC while
the low value to that of a DOC.] In this gure solid lines denote the
ceramic substrate and dashed and dotted lines denote the metallic
substrates with thermal conductivity of 50 W/(mK) and 5 W/(mK),
respectively. For this case, the metallic substrate performs better than
the ceramic substrate during transient phase for all three nal
temperatures, and both have nearly 100% conversion at steady-
state. We see that the metallic substrate out performs the ceramic
substrate for most of transient time. For inlet gas temperature of
500 K, metallic substrate with 5 W/(mK) gives 35% lower cumulative
emissions than the ceramic substrate. The impact of substrate
material on the light-off behavior and cumulative emissions shown
in Fig. 11 can be explained better by examining the transient
temperature proles in the solid which are shown in Fig. 12. We
note that for low inlet gas temperatures (with back-end ignition), the
monolith with metallic substrate not only ignites earlier but also the
ignited length is larger compared to the ceramic case. In contrast, for
the case of high inlet gas temperatures (front end ignition), the
ignition time is nearly independent of the substrate conductivity but
the temperature front propagates slowly (and the gradient is higher)
for the ceramic case.
As a nal comparison, we note that the performance difference
between the ceramic and metallic substrates increases as the inlet
gas temperature ramp time reduces. Fig. 13 shows the calculated
cumulative emissions with linearly varying temperature from 300
to 500, 550, 650 K in 10 s for the two substrates as a function of
time. For inlet gas temperature of 500 K, metallic substrate with
5 W/m K gives 45% lower cumulative emissions than the ceramic
substrate. Based on the above comparisons, it may be concluded
that for low temperature applications (e.g. diesel oxidation
catalysts), monoliths with metallic substrates are superior to
ceramic substrates.
7. Conclusions and discussion
As stated in Introduction, the main goal of this work was to
investigate the effect of substrate material on the steady-state
and dynamic behavior of catalytic monoliths. If the objective is to
0 10 20 30 40
0
0.5
1
1.5
2
2.5
3
3.5
4
Time (sec)
C
u
m
u
l
a
t
i
v
e

e
m
i
s
s
i
o
n
s

(
g
)
u=0.5
u=2
u=1
Fig. 9. Cumulative emissions with time for different velocities for inlet uid
temperature of 600 K. Solid lines indicate ceramic substrate and dashed lines
indicate metallic substrate. Parameter values used are listed in Table 2.
300
400
500
600
700
800
I
n
l
e
t

T
e
m
p
e
r
a
t
u
r
e

(
K
)
Time (sec)
0 20 40 60 80 100
0
1
2
3
4
5
I
n
l
e
t

V
e
l
o
c
i
t
y

(
m
/
s
)
Fig. 10. Variation of the inlet gas temperature and velocity during the rst 100 s
of a FTP cycle.
0 20 40 60 80 100 120
0
5
10
15
20
25
30
Time (sec)
C
u
m
u
l
a
t
i
v
e

E
m
i
s
s
i
o
n
s
(
g
)
Ceramic
550K
500K
650K
Fig. 11. Plot of cumulative emissions as a function of time for metallic (dashed
lines for 50 W/(m K) and dotted lines for 5 W/(m K) ) and ceramic substrates (solid
lines) for varying inlet uid temperature. Other parameter values used are listed in
Table 2.
S.R. Gundlapally, V. Balakotaiah / Chemical Engineering Science 92 (2013) 198210 208
operate the monolith in the region of multiple steady-states and
on the ignited branch, then we have shown that monoliths with
high conductivity substrates (metallic) and (or) high aspect ratio
(diameter to length) lead to the ignited branch at lower inlet gas
temperatures (typically 60100 1C) compared to ceramic sub-
strates. Equivalently, the PGM loading needed to obtain an ignited
steady-state is much lower for high aspect ratio metallic mono-
liths compared to low aspect ratio ceramic monoliths. In our view,
this is a signicant result as it implies that for steady-state
operation, it is possible to replace precious metals (such as Pt,
Pd, Rh) by base metals (Cu, Co, Fe, Ni) that have lower activity (by
about two to three orders of magnitude) by changing the
substrate and aspect ratio of the monolith.
The second main result of this work is the investigation of the
impact of substrate material on the transient behavior and hence
on cold-start emissions. We have shown that when the monolith
is much colder initially compared to the inlet gas (and front-end
ignition), the heat-up time, which determines the cumulative
cold-start emissions, decreases monotonically with the substrate
thickness and heat capacity but is non-monotonic with inlet gas
velocity. This heat-up time is higher at both low and high gas
velocities and is a minimum at some intermediate gas velocity.
This optimum gas velocity is lower for low conductivity (ceramic)
substrate and higher for high conductivity (metallic) substrate. In
general, metallic substrates lead to lower cumulative emissions at
higher gas velocities but may lead to higher emissions at lower
gas velocities (depending on the hydraulic diameter). These
results conrm and extend the earlier modeling studies of
Ramanathan et al. (2004a, 2004b, 2004c) and the recent experi-
mental observations of Santos and Costa (2008).
The calculations presented in Figs. 9, 11 and 13 lead to the
conclusion that thin walled metallic substrates with intermediate
values of conductivity (in the 515 W/(mK)) range can lead to lower
cumulative emissions for all gas velocities within the practical range.
Further, these calculations show that for a xed inlet gas velocity,
there is an optimum substrate conductivity that leads to lower
cumulative emissions. Thus, the properties of the substrate can have
substantial impact on the cumulative emissions.
Currently the industrial trend is making the monolith channels
with high cell density (cpi 41200), which reduces the open channel
diameter signicantly, and hence the transverse Peclet number.
While this leads to lower cold-start emissions for the case of front
end ignition, it may not be the best design for back end ignition. For
example, in the case of a TWC, the inlet gas temperatures are
generally higher (than that needed for front end ignition, typically
4600 K) and hence a high cell density monolith is always preferred
(as it leads to lower cumulative emissions). However, for the case of a
DOC and LNT applications, where the inlet gas temperature is not
high, using a high cell density monolith may push the ignition
location to the back-end and hence the temperature front has to
propagate upstream to ignite the entire monolith. As shown in this
work, for the case of low catalyst loading, or low conductivity, the
temperature front may fail to propagate upstream. Thus, depending
on the range of inlet temperatures and gas velocities, there appears to
be an optimumcell density and substrate conductivity for minimizing
cold-start emissions. The results of this work may be used to
determine these optimum conditions.
This work was conned to the case of a single catalyst brick with
uniform activity and conductivity in the axial direction. However, the
use of multiple bricks or a single brick with axially varying activity
and conductivity can also have a profound inuence on the transient
behavior. This case will be investigated in future work.
Nomenclature
Roman letters
A
O
cross sectional area of the channel
B dimensionless adiabatic temperature rise
c dimensionless concentration
C concentration
C
in
inlet uid concentration
c
p
heat capacity
D
e
effective diffusivity of reactant in the washcoat
D
h
channel hydraulic diameter
D
m
diffusion coefcient in the uid phase
Da Damk ohler number
E activation energy
f friction factor
h(x) local heat transfer coefcient
k
c
(x) local mass transfer coefcient
Fig. 12. Comparison of the light-off behavior at low and high inlet gas tempera-
tures for three different substrate conductivities, 1 (top), 50 (middle), and 5 W/
(m K) (lower).
0 20 40 60 80 100
0
5
10
15
20
25
Time (sec)
C
u
m
u
l
a
t
i
v
e

E
m
i
s
s
i
o
n
s

(
g
)
Ceramic
500K
550K
650K
Fig. 13. Plot of cumulative emissions as a function of time for metallic (dashed
lines for 50 W/(m K) and dotted lines for 5 W/(m K)) and ceramic substrates (solid
lines) for varying inlet uid temperature. Other parameter values used are listed in
Table 2.
S.R. Gundlapally, V. Balakotaiah / Chemical Engineering Science 92 (2013) 198210 209
k thermal conductivity
k
v
rst order rate constant per unit washcoat volume
L length of the monolith channel
Le
f
uid Lewis number
Nu(z) Local Nusselt number
P transverse Peclet number
Pe
h
heat Peclet number
P
O
perimeter of the channel
Pr Prandtl number
Re Reynolds number
R
g
universal gas constant
R
O
one-fourth the channel hydraulic diameter
Sc Schmidt number
Sh(z) local Sherwood number
t time
T uid temperature
T
f ,in
inlet uid temperature
/uS average uid velocity in the channel
x coordinate along the length of the channel
(dimensional)
y transverse coordinate (dimensional)
z dimensionless coordinate along the length of the
channel
Greek letters
r density
g dimensionless activation energy
d
c
effective thickness of the washcoat
d
s
half thickness of the support
d
w
effective wall thickness
y dimensionless temperature
t dimensionless time
f
s
Thiele modulus (local Damk ohler number)
L dimensionless washcoat thickness
Z effectiveness factor
Subscripts and superscripts
f uid phase
s solid phase
m cup-mixing
o asymptotic value
H2 constant wall heat ux
T constant wall temperature or concentration
Acknowledgements
This work was supported by a grant from the Robert A. Welch
foundation, grant #E-1152.
References
Balakotaiah, V., 1996. Chem. Eng. Educ., 234239.
Dommeti, S.M.S., Balakotaiah, V., West, D.H., 1999. Ind. Eng. Chem. Res. 38 (3),
767777.
Gundlapally, S.R., Balakotaiah, V., 2011. Heat and mass transfer correlations and
bifucation analysis of catalytic monoliths with developing ows. Chem. Eng.
Sci. 66 (9), 18791892.
Gupta, N., Balakotaiah, V., 2001. Heat and mass transfer coefcients in catalytic
monoliths. Chem. Eng. Sci. 56, 47714786.
Gupta, N., Balakotaiah, V., West, D.H., 2001. Bifurcation Analysis of a two-
dimensional catalytic monolith reactor model. Chem. Eng. Sci. 56, 14351442.
Huang, D.T.J., Varma, A., 1981. Steady-state multiplicity of a non-adiabatic bubble
column with fast reactions. AIChE J. 27, 111120.
Jasper, T.S., Robinson, K., Anderton, D., Cuttler, H.H., 1991. Monolith substrate
effects on catalyst light-off. In: Crucq, A. (Ed.), Catalysis and Automotive
Pollution Control II. Elsevier, pp. 523535.
Kumar, P., Makki, I., Kerns, J., Grigoriadis, K., Franchek, M., Balakotaiah, V., 2012. A low-
dimensional model for describing the oxygen storage capacity and transient
behavior of a three-way catalytic converter. Chem. Eng. Sci. 73, 373387.
Luss, D., Balakotaiah, V., 1984. Steady-state multiplicity features of chemical
reactors. In: Doraiswamy, L.K., Mashelkar, R.A. (Eds.), Frontiers in Chemical
Reaction Engineering. Wiley.
Maestri, M., Beretta, A., Groppi, G., Tronconi, E., Forzatti, P., 2005. Comparison
among structured and packed-bed reactors for the catalytic partial oxidation
of CH
4
at short contact times. Catal. Today 105, 709717.
Mathieu-Potvin, F., Gosselin, L., 2012. Threshold length for maximal reaction rate
in catalytic microchannels. Chem. Eng. J. 188, 8697.
Nishizawa, K., Masuda, K., Horie, H., Hirohashi, J., 1989. Development of Improved
Metal-Supported Catalyst. SAE Technical Paper 890188.
Pannone, G.M., Mueller, J.D., 2001. A Comparison of Conversion Efciency and
Flow Restriction Performance of Ceramic and Metallic Catalyst Substrates. SAE
Technical Paper 2001-01-0926.
Pfalzgraf, B., Rieger, M., Ottowitz, G., 1996. Close-Coupled Catalytic Converters for
Compliance with LEV/ULEV and EG III Legislation-Inuence of Support Material,
Cell Density and Mass on Emission Results. SAE Technical Paper 960261.
Ramanathan, K., Balakotaiah, V., West, D.H., 2003. Light-off criterion and transient
analysis of catalytic monoliths. Chem. Eng. Sci. 58, 13811405.
Ramanathan, K., Balakotaiah, V., West, D.H., 2004a. Bifurcation analysis of catalytic
monoliths with non-uniform catalyst loading. Ind. Eng. Chem. Res. 43, 288303.
Ramanathan, K., Balakotaiah, V., West, D.H., 2004b. Light-off and cumulative
emissions in catalytic monoliths with non-uniform catalyst loading. Ind.
Eng. Chem. Res. 43, 46684690.
Ramanathan, K., West, D.H., Balakotaiah, V., 2004c. Optimal design of catalytic
converters for minimizing cold-start emissions. Catal. Today 98, 357373.
Santos, H., Costa, M., 2008. Evaluation of the conversion efciency of ceramic and
metallic three way catalytic converters. Energy Convers. Manage. 49, 291300.
S.R. Gundlapally, V. Balakotaiah / Chemical Engineering Science 92 (2013) 198210 210

Você também pode gostar