You are on page 1of 10

Ryoichi S.

Amano
e-mail: amano@uwm.edu
Mohsen M. Abou-Ellail
e-mail: abouellail@hotmail.com
Mechanical Engineering Department,
University of Wisconsin-Milwaukee,
Milwaukee, WI 53201
S. Kaseb
Mechanical Engineering Department,
Cairo University,
Cairo 12613, Egypt
e-mail: kaseb@pathways.edu.eg
Numerical Predictions of
Hydrogen-Air Rectangular
Channel Flows Augmented by
Catalytic Surface Reactions
Catalytic combustion of hydrogen-air boundary layers involves the adsorption of hydro-
gen and oxygen into a platinum-coated surface, chemical reactions of the adsorbed spe-
cies, and the desorption of the resulting products. Re-adsorption of some produced gases
is also possible. This paper presents numerical computations of laminar momentum
transfer, heat transfer, and chemical reactions in rectangular channel ows of hydrogen-
air mixtures. Chemical reactions are included in the gas phase as well as on the solid
platinum surfaces. In the gas phase, eight species are involved in 26 elementary reac-
tions. On the platinum hot surfaces, additional surface species are included, which are
involved in 16 additional surface chemical reactions. The platinum surface temperature
distribution is prespecied, while the properties of the reacting ow are computed. The
results show very good agreement with the measured data. [DOI: 10.1115/1.4005202]
Keywords: catalyst, catalytic, channel ow, surface reaction, hydrogen, platinum
1 Introduction
Platinum and palladium are common catalysts used to assist
combustion. The catalytic reactions can be benecial in porous
burners and catalytic reactors that use low equivalence ratios. In
this case the porous burner ame can be stabilized at low tem-
peratures to prevent any substantial gas emissions, such as nitro-
gen oxides. Cattolica and Schefer [1,2] experimentally
investigated the combustion of hydrogen-air mixtures in bound-
ary layer ows over catalytic and noncatalytic surfaces. In cata-
lytic combustion of hydrogen-air boundary layers, there involves
the adsorption of hydrogen and oxygen into a platinum-coated
surface, chemical reactions of the adsorbed species, and the de-
sorption of the resulting products. Re-adsorption of some pro-
duced gases is also possible. The catalytic reactions are found to
be very benecial in porous burners and catalytic reactors that
use low equivalence ratios. In this case the porous burner ame
can be stabilized at low temperatures to prevent any substantial
gas emissions, such as nitrogen oxides. Combustion of several
fuels, i.e., NH
3
, CH
4
, C
3
H
8
and NH
3
/CH
4
mixture, on a platinum
surface in a stagnation ow eld has been studied experimentally
by Williams et al. [3]. In addition, combustion of methane on a
platinum surface has been studied numerically by Song et al. [4].
In their work [4], the surface reactions of methane over platinum
have been modeled by a global reaction mechanism. Similarly,
combustion of hydrogen on a platinum surface in a stagnation
ow eld and a boundary layer ow eld has been studied
numerically by Warnatz et al. [5]. In their work, the elementary
reaction mechanism of the oxidation of hydrogen on a platinum
surface, including reactions of absorption/desorption of reactants
and products and reactions of surface radical recombination, has
been introduced. They also compared the computed [OH] con-
centration with experimental data from Cattolica and Schefer
[1,2]. However, Warnatz et al. [5] predicted values of OH one to
two orders of magnitude lower than those measured by Cattolica
and Schefer [1,2], at 1170 K surface temperature. Thermochemi-
cal data of surface species involved in the surface reactions have
also been published by Warnatz et al. [5]. Later, the elementary
surface reaction mechanism of methane with platinum has been
established by Deutschmann et al. [6]. The detailed surface reac-
tion mechanism has been used by Deutschmann et al. [6] to
numerically simulate the experiments of surface combustion of
methane over platinum conducted by Williams et al. [3]. It was
found that [6] methane is ignited at surface temperatures around
1000 K and the reaction is fast between surface temperatures of
1000 K and 1300 K; for surface temperatures below 1000 K, the
surface reaction is slow and the methane could not be ignited.
Surface coverage of surface species and mole fractions of gas-
phase species have been calculated by Deutschmann et al. [6] in
their work. Raja et al. [7] used the detailed surface mechanisms
for methane combustion [6] to model the catalytic honeycomb
monolith combustion.
In the present paper, the combustion of hydrogen-air mixtures
over catalytic surfaces is studied by numerical methods. The test
case, used to validate the present solution procedure, is that of a
reacting ow of hydrogen-air mixture in a catalytic rectangular
channel. The test case ow conguration is shown in Fig. 1. A
parallel ow of hydrogen-air mixture passes through the inlet sec-
tion of a rectangular channel coated with platinum. The equiva-
lence ratio of the hydrogen-air mixture is 0.32. The gas mixture
velocity at the channel inlet section is 1.6 m s
1
while the gas tem-
perature is 313 K. A two-dimensional reacting ow model, with
surface reactions, is developed for the rectangular channel geome-
try since the channel width is nearly 16 times its height. In this
case the ow along the central section, perpendicular to the chan-
nel width, is two-dimensional in nature. A multistep reaction
mechanism is adopted for the gas-phase reactions. It involves 24
elementary reactions and 8 chemical species for hydrogen-air
mixtures. The multistep surface reaction mechanism is adopted
from the mechanism developed by Deutschmann et al. [6]. This
surface reaction mechanism consists of 16 elementary surface
reactions involving four gas-phase species and ve surface spe-
cies. The surface reaction model developed by Coltrin et al. [8] is
adopted as a general layout for the present work. The present
Contributed by the Heat Transfer Division of ASME for publication in the
JOURNAL OF HEAT TRANSFER. Manuscript received October 29, 2010; nal manuscript
received September 24, 2011; published online February 13, 2012. Assoc. Editor:
Sujoy Kumar Saha.
Journal of Heat Transfer APRIL 2012, Vol. 134 / 041201-1 Copyright VC
2012 by ASME
Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 04/12/2013 Terms of Use: http://asme.org/terms
numerical predictions are compared with recent experiment and
numerical data.
2 Mathematical Model and Solution Procedure
The numerical model for laminar ow, heat transfer, gas-
phase combustion, and catalytic surface reactions is presented.
The ow conguration investigated in the present paper is that
of a rectangular channel burner. Finite-volume equations are
obtained by formal integration over control volumes surround-
ing each grid node. Hybrid differencing is used to ensure that
the nite-difference coefcients are always positive or equal to
zero to reect the real effect of neighboring nodes on a typical
central node. The nite-volume equations of the reacting gas
ow properties are solved by a combined iterative-marching
algorithm. On the platinum surfaces, surface species balance
equations, under steady-state conditions, are solved numeri-
cally. A nonuniform computational grid is used, concentrating
most of the nodes in the boundary sublayer adjoining the cata-
lytic surfaces. The ow conguration is shown in Fig. 1. The
numerical model can be classied into four main groups of
equations as follows.
2.1 Continuity and Momentum Equations. The mass conti-
nuity of the reacting ow may be written as
@q
@t

@qu
i

@x
i
0 (1)
The momentum equation for the laminar reacting ow may be
written in Cartesian tensor notations as [9]
@ qu
j
_ _
@t

@
@x
i
qu
j
u
i
_ _

@
@x
i
l
@u
j
@x
i
_ _

@ P
@x
j

@
@x
i
l
@u
i
@x
j

2
3
@u
k
@x
k
d
ij
_ _ _ _
(2)
where q is the gas density; u
j
is the gas velocity along coordinate
x
j
; l is the dynamic viscosity.
2.2 Energy Equations. The energy equation for the reacting
ow is
@qh
g
@t

@
@x
i
qu
i
h
g
_ _

@
@x
i
C
h
@h
g
@x
i
_ _

j
DH
j
_ _
W
j
(3)
where h
g
is the gas sensible enthalpy; C
h
is the thermal diffusivity;
DH
j
and W
j
are the enthalpy of reaction and reaction rate of chem-
ical reaction j of the reaction mechanism involving hydrogen-air
mixtures; in the above equations, the subscript g denotes the gas
phase. The gas temperature at any point is directly related to the
local gas sensible enthalpy (h
g
), gas species mass fractions (Y
j
),
and constant-pressure specic heats, namely
h
g

_

j
C
pj
Y
j
_ _
dT (4)
The lower limit of the above integration is 298 K, while the upper
limit equals the local gas temperature, T
g
.
2.3 Species Mass Fraction Equations. The species mass
fractions for the reacting channel ow may be written as
@qY
l

@t

@ qu
i
Y
l

@x
i

@
@x
i
C
Yl
@Y
l
@x
i
_ _
M
l

lm
W
m
(5)
where Y
l
is the mass fraction of species l while C
Yl
is its molecular
diffusivity; M
l
is the molecular mass of species l;
lm
is the stoi-
chiometric coefcient of species l in reaction m while W
m
is the
reaction rate of the elementary chemical reaction m of the present
reaction mechanism. It consists of 24 elementary reactions. The
present hydrogen-air chemical kinetic mechanism involves eight
species, namely, O
2
, O, OH, H, H
2
, H
2
O, H
2
O
2
, and HO
2
. The ele-
mentary reaction mechanism for hydrogen-air mixtures adopted in
the present work is the same as given by Tong et al. [10].
It should be mentioned here that
lm
is taken as a positive value
for products and negative for reactants as required for proper sum-
mation of the effect of each reaction on the net production of a
particular species.
2.4 Surface Reactions. For a steady-state problem, the solu-
tion has no change with respect to time. Thus, surface coverage of
any surface species with respect to time is zero. The variation of
the surface coverage with respect to time can be computed from
the net production of each surface species. The conservation of
the surface coverage of surface species k may be written as [5]
dz
k
dt

_ s
k
Z
(6)
where z
k
is the surface coverage of surface species k, Z is the total
surface site density, and the surface density used in the present pa-
per is 1.63 10
15
cm
2
[6]; _ s
k
is the net surface production rate,
in mol cm
2
s
1
, of the surface species k.
Since the present reacting ow includes gas-surface interac-
tions, the mass transfer between the gas phase and the catalytic
surface needs to be included while solving the species mass frac-
tions of the gas species in the ow. The mass uxes transferred
through the convection and the diffusion processes at the gas-
surface interface of any gas-phase species are balanced by the net
production or depletion rates of that species by surface reactions.
The surface boundary condition of each gas-phase species k based
on the mass balance is given by Coltrin et al. [8] as
n qY
k
V
k
u _ s
k
M
k
(7)
where n is the unit normal vector pointing outward with respect to
the surface, u is the bulk uid velocity, V
k
is the diffusion veloc-
ity, M
k
is the molecular weight of species k, _ s
k
is the net produc-
tion or depletion rates of gas-phase species k involved in surface
reactions.
The production rate of each species _ s
k
, either gas-phase species
or surface species, may be written as [11]
_ s
k

Ks
j1

kj
k
fj

NgNs
k1
X
k

0
k
j
(8)
where K
s
is the total number of elementary surface reactions;
0
kj
is left hand side stoichiometric coefcients of the reaction
Fig. 1 Layout of the conguration of reacting hydrogen-air
mixture ow inside platinum-coated rectangular channel
041201-2 / Vol. 134, APRIL 2012 Transactions of the ASME
Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 04/12/2013 Terms of Use: http://asme.org/terms
equation; v
kj
is the right hand side minus left hand side stoichio-
metric coefcients of the reaction equation; k
fj
is the forward ki-
netic rate constants; [X
k
] is the species concentrations; the units of
gas-phase species and surface species concentrations are mol
cm
3
and mol cm
2
, respectively; N
g
is the number of gas-phase
species while N
s
is the number of surface species.
A detailed surface reaction mechanism is used to model the
gas-surface interaction between the fuel and the platinum-coated
surface. The surface reaction mechanism adopted in the present
work, for hydrogen-air mixtures reacting over platinum-coated
surfaces, is established by Deutschmann et al. [6]. The surface
reaction mechanism consists of 16 elementary reactions and
involves four gas-phase species, namely, O
2
, OH, H
2
, H
2
O, and
ve surface species, namely, O(s), OH(s), H(s), H
2
O(s), and Pt(s).
The reaction rate constants are described in terms of either the
Arrhenius expression or a sticking coefcient, c. The Arrhenius
expression form is
k
fj
A
j
exp
E
j
RT
_ _
(9)
The sticking coefcient can be converted to the usual kinetic rate
constants via the relation given by Coltrin et al. [8]
k
fj

c
1 c=2
1
Z
m

RT
2pM
_
(10)
where Z is total surface site concentration, m is sum of the surface
reactants stoichiometric coefcient, R is the universal gas con-
stant, T is the gas temperature at the catalytic surface, and M is the
molecular weight of the gas-phase species. The thermochemical
data needed to determine the rate coefcients for the reverse reac-
tions and enthalpies of the surface species are provided by War-
natz et al. [5].
In the present work, steady-state conditions of the owing gas
and the surface species are assumed. Equations (6) and (8), in this
case, can be used to compute the surface species mole fractions,
given the gas species mole fractions at the interface between the
catalytic surface and the owing gas phase. Equations (7) and (8),
on the other hand, are used to compute gas species uxes at the
gas-surface interface. Although these uxes do not appear explic-
itly in Eq. (5), they are used to impose the proper boundary condi-
tions on the owing gas species nite-volume equations.
It should be mentioned here that the gas physical properties are
computed from temperature dependent relations. At any point in
the owing gas phase, the density is computed from the ideal gas
equation of state, while the viscosity is computed from a tempera-
ture polynomial of the third order. The species specic heats are
computed from temperature polynomials of the sixth order, which
are valid for 300 K to 2500 K temperature range [12]. The remain-
ing physical properties are deduced from the above properties and
Prandtl and Lewis numbers specied as 0.7 and 1.0, respectively,
for most gas species. However, the Lewis numbers of H
2
and H
are taken as 0.3 and 0.2, respectively, to reect the faster diffusion
of these species with respect to the other heavier species such as
H
2
O and O
2
.
2.5 Numerical Solution Procedure. The two-dimensional
ow inside the catalytic rectangular channel, shown in Fig. 1, is
overlaid with a nite grid of nodes. Equations (2), (3), and (5) and
the pressure correction counterpart of Eq. (1) are formally inte-
grated over the control volume surrounding each grid node. The
time dependent terms of the governing equations were dropped
out since the reacting ow considered is under steady-state condi-
tions. The faces of the control volume bisect the distances
between the particular node and the four nearest neighbor nodes.
The formal integration is performed with due care to preserve the
physical meaning and overall balance of each dependent variable.
The nal form of the nite-volume equations are written as fol-
lows [9]:

a
n
S
P
_ _
W
p

a
n
W
n
S
u
(11)
where W stands for any of the dependent variables; namely, axial
and radial velocity components, gas sensible enthalpy, species
mass fractions, or the pressure correction which is used to satisfy
both mass continuity and momentum equations simultaneously.
The summation (R) is over the n four neighbors of a typical node
p. The above nite-difference coefcients a
n
are computed using
the hybrid method, such that these coefcients are always non-
negative to give the proper combined effects of convection and
diffusion. The above hybrid method is a combination of upwind
and central differencing schemes. Each scheme prevails for a par-
ticular range of the cell Peclet numbers, as given in detail by
Abou-Ellail et al. [9]. S
p
and S
u
are the coefcients of the inte-
grated source terms at node p. The ow in a rectangular channel
with dominant streamwise direction is parabolic in nature and
thus the hybrid scheme is accurate enough to eliminate any false
diffusion. In this case, the computational grid directions are
aligned along the main ow direction and normal to it.
The solution procedure is based on the, line-by-line, tridiagonal
matrix algorithm (TDMA). The nite-volume equations (Eq. (11))
for each dependant variable are modied at the boundaries of the
solution domain, shown in Fig. 1, to impose the conditions there.
On the catalytic surface where y 0.0 mm, the velocity compo-
nents vanish while all other dependent variables normal gradients
reect the mass and heat uxes due to surface reactions. Along
the axis of symmetry where y 3.5 mm and the exit section of the
ow, the normal gradient is equal to zero. Moreover, for the cata-
lytic channel ow, the transverse velocity is equal to zero at the
upper boundary where y 3.5 mm. At the inlet section, all varia-
bles are uniform. Within each computational loop of the gas
phase, Eq. (6) is solved, for all surface species concentrations,
iteratively at each transverse plane for steady-state conditions,
i.e., d(z
k
)/dt 0. The validation of the present numerical algo-
rithm is presented by Tong et al. [12] for heat and mass transfer in
impinging ows on a hot catalytic surface. Tong et al. [12] com-
pared their results for temperature and species mole fractions in
the reacting impinging jet with numerical data of Deutschmann
et al. [6]. The agreement they obtained conrms the accuracy of
the present numerical procedure [12]. Moreover, standard laminar
thermal boundary layer thickness and heat transfer data [13] were
also used for validation purposes by Tong et al. [14]. Further vali-
dation is given in the present work by comparing the present nu-
merical results with the experimental and numerical data of Appel
et al. [15].
3 Presentation and Discussion of Results
The catalytic surface reactions in channel ows are investigated
numerically in this section. The numerical results are compared
with the recent experimental measurements conducted by Appel
et al. [15]. The rectangular channel, used by Appel et al. [15], has
its upper and lower plates coated with platinum. The height,
width, and length of the channel are equal, respectively, to 7 mm,
110 mm, and 300 mm. The ratio of width of the channel to the
height of the channel is large enough so that the ow could be
considered two-dimensional in the vertical plane passing by the
centerline. The hydrogen-air mixture ows through the channel
and reacts on the catalytic surface. The schematic plot of the chan-
nel ow is shown in Fig. 1. The channel geometry shown in Fig.1
is used in the present numerical model. The grid used for the
channel ow, namely the number of grid nodes along the x-axis
and y-axis is 10,000 and 330, respectively. Due to symmetry, only
the lower half of the computational plane, indicated in Fig. 1,
needs to be overlaid with the above grid of nodes. Moreover, the
axial increment Dx is taken as 3.0 10
2
mm. The transverse grid
Journal of Heat Transfer APRIL 2012, Vol. 134 / 041201-3
Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 04/12/2013 Terms of Use: http://asme.org/terms
nodes are arranged such that 200 nodes lay inside the boundary
sublayer, i.e., in the range 0.0 <y <1.0 mm. The rest of the trans-
verse nodes lay outside the boundary sublayer, i.e., in the range
1.0 <y <3.5 mm. The above number of grid nodes is sufcient to
handle the fast changing gas density and viscosity, both axially
and transversely inside the rectangular channel, as well as insuring
a grid-independent solution. With a total number of 3.3 million
grid nodes, it was necessary to utilize the parabolic nature of the
ow. This is done through the use of a line-by-line TDMA scheme
to obtain numerical solutions in such a way to sweep the solution
domain from the inlet section to the exit section. The eld values
of the dependent variables are stored and then updated during the
subsequent iterations. The iterations are stopped when the nite-
volume equations of the dependent variables are satised with
errors less than 0.1%. This criterion is achieved with only three
iterations. The present numerical simulator is actually a combina-
tion of parabolic and elliptic procedures. It can thus be referred to
as a parabolic-elliptic procedure.
For the case being studied in the present work, the inlet velocity
of the gaseous mixture is 1.6 m/s, the inlet gas temperature is
313 K, and the equivalence ratio is 0.32. This case is referred to as
case 2 in the experimental work of Appel et al. [15]. The surface
temperature prole of the channel wall, along the x-axis was also
measured by Appel et al. [15]. Figure 2 shows the catalytic wall
temperature prole along the streamwise direction. As shown in
Fig. 2, the surface wall temperature is not uniform. The wall tem-
perature at channel entrance is about 1060 K. It gradually
increases to a temperature around 1255 K then decreases to
1200 K near the channel exit section. This temperature prole was
maintained by Appel et al. [15] with the help of guard heaters and
ceramic insulation. They also allowed the rst 60 mm of the chan-
nel to radiate to the cold surroundings to prevent the wall tempera-
ture from reaching super-adiabatic levels as a result of the faster
diffusion of the low- Lewis-number fuel, namely H
2
.
The curve tting equations that accurately describe the surface
temperature prole (T
s
) are given below as
T
s
1062 0:0089x
2:57
; 0 < x < 20 (12a)
T
s
1079 1:80x 20
1:05
; 20 < x < 80 (12b)
T
s
1208 3:20x 80
0:66
; 80 < x < 130 (12c)
T
s
1248:5 0:23x 130
0:76
; 130 < x < 210 (12d)
T
s
1254:3 0:029x 210
2:02
; 210 < x < 265 (12e)
Equations (12a)(12e) are used in the present work as the bound-
ary condition for the catalytic surface temperature axial prole of
the channel ow of Appel et al. [15]. Therefore, the heat gener-
ated by catalytic chemical reactions on the surface is not explicitly
considered in the present numerical computations. Figure 3 shows
the present numerical temperature transverse proles together
with the experimental data of Appel et al. [15], at x 25, 85, 105,
165, and 235 mm. However, no experimental data for the 265-mm
section can be found. The present temperature proles for the
measured ve sections are in good agreement with the experimen-
tal data of Appel et al. [15]. Numerical computations were also
conducted by Appel et al. [15] using a fully elliptic computer
code that required 20,000 to 30,000 iterations depending on the
initial guess. They used 340 nodes along x-axis and 90 nodes
along the y-axis. They overpredicted the temperature of most of
the sections reported with a maximum error of 110 K between
their numerical and experimental data. However, as it can be seen
from Fig. 3, the experimental data of the gas temperature are well
predicted in the present work. This can probably be attributed to
the much ner grid adopted in the present work. However, while
the present computer code uses a parabolic-elliptic solution proce-
dure that requires few iteration loops, Appel et al. [15] fully ellip-
tic procedure converged in 20,000 to 30,000 iterations depending
on the initial guess.
Figure 4 shows the mole fractions of H
2
and H
2
O at different
streamwise locations of 25, 85, 105, 165, and 235 mm. The mole
fraction of H
2
is multiplied by a factor of 2 for the purpose of
clarity. The predicted mole fraction streamwise proles are com-
parable with the measured data of Appel at al. [15]. The stream-
wise mole fraction proles of H
2
are in good agreement with the
experimental data up to a distance of 165 mm along the x-axis.
However, for x >165 mm, the H
2
mole fraction is slightly under-
predicted by the present solution procedure. Moreover, the nu-
merical results of Appel et al. [15] slightly underpredicted their
experimental data for all reported transverse planes, although
they used a costly fully elliptic numerical procedure. The pre-
dicted mole fraction transverse proles of H
2
O are compared to
Fig. 2 Catalytic surface temperature prole for the channel
ow of Appel et al. [15]
Fig. 3 Transverse temperature proles at streamwise distances of 25mm, 85mm, 105mm, 165mm, 235mm, and 265mm
041201-4 / Vol. 134, APRIL 2012 Transactions of the ASME
Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 04/12/2013 Terms of Use: http://asme.org/terms
the experimental data of Appel et al. [15], as depicted in Fig. 4.
The predicted mole fraction proles of H
2
O are in good agree-
ment with the experimental data for the rst three transverse
planes. However, the numerical mole fraction proles of H
2
O
overpredict the experimental data for the last two planes, as
shown in Fig. 4. This is an indication of an overprediction of the
H
2
O production rate that the adopted surface reaction mechanism
computes at the downstream locations where most of the plati-
num is inactive. Moreover, the numerical data of Appel et al.
[15] slightly overpredicted the experimental data for the last two
panes with x 165 and 235 mm. The difference in the level of
agreement between the present results and Appel et al. [15] pre-
dictions could hardly be attributed to the difference in the solu-
tion procedures. However, this better agreement is attributed to
the modication in Eq. (10) that Appel et al. [15] adopted to
compute k
fj
in their numerical model. This modication reduces
the specic reaction rate constants of the H
2
and O
2
adsorption
reactions at lower values of Pt(s) prevailing in the downstream
area of the inlet section.
In order to assess the merits of the present numerical work, the
ratio of the execution times of the present and Appel et al. [15] so-
lution procedures is thus estimated. The TDMA is used in the
present work and will be hypnotized as the solution scheme for
Appel et al. [15]. In the present work, TDMA is performed only
along the transverse direction and is repeated for all transverse
sections. However, for a fully elliptic solution procedure the
TDMA is normally applied along the streamwise (N
x
) and the
transverse (N
y
) directions. It should be mentioned here that N
x
and N
y
are the number of nodes along the x- and y-coordinates.
The ratio between the execution time of Appel et al. [15] and the
present work is estimated as the ratio between the two numbers of
iterations times the ratio between the two execution times per iter-
ation. The execution time of the TDMA(N
x
) together with the
execution time for computing the nite-difference equation coef-
cients is commonly assumed to vary with N
x
to an exponent hav-
ing a value between 1.0 and 2.0. If this exponent is taken as 1.5,
the estimated execution time of Appel et al. [15] is roughly 2
orders of magnitude higher than the present parabolic-elliptic so-
lution procedure execution time. As the TDMA algorithm is con-
sidered one of the most economical nite-difference equation
solvers, the above result should be taken as a nominal value rather
than an exact one.
The predicted oxygen mole fraction transverse proles at differ-
ent streamwise locations are depicted in Fig. 5 for streamwise dis-
tances of 25 mm, 85 mm, 105 mm, 165 mm, 235 mm, and 265 mm.
The maximum value at each section occurs at the channel center-
line, as the oxygen is mostly consumed at the walls. At
x 265 mm, the proles are completely at indicating the end of
most surface and gas reactions. Since the equivalence ration of
this ow is 0.32, then residual oxygen is expected at the exit sec-
tion as can be seen from Fig. 5.
Figure 6 shows the transverse proles of the streamwise veloc-
ity at distances along the x-axis of 25 mm, 85 mm, 105 mm,
165 mm, 235 mm, and 265 mm. The streamwise velocity shows
nearly a uniform prole near the inlet section changing to a para-
bolic prole near the exit section as it would be for ow between
parallel plates. However, because of the excessive heating of the
gas mixture as it ows along the channel axis, the centerline ve-
locity increases from 1.6 m/s at the inlet section to approximately
9 m/s at the exit section.
Figure 7 depicts the numerical transverse proles of the trans-
verse velocity at streamwise sections of 25 mm, 85 mm, 105 mm,
165 mm, 235 mm, and 265 mm. The transverse velocity is highest
near the inlet section peaking to a value of about 2.2 cm s
1
.
Fig. 4 Mole fractions of H
2
and H
2
O at streamwise distances of 25mm, 85mm, 105mm, 165mm, 235mm, and 265mm
Fig. 5 Mole fraction of O
2
at streamwise distances of 25mm, 85mm, 105mm, 165mm, 235mm, and 265mm
Journal of Heat Transfer APRIL 2012, Vol. 134 / 041201-5
Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 04/12/2013 Terms of Use: http://asme.org/terms
Because of the symmetry of the channel ow, the transverse ve-
locity is asymmetric with respect to the x-axis.
The OH mole fraction transverse proles are shown in Fig. 8,
for streamwise distances of 25 mm, 85 mm, 105 mm, 165 mm,
235 mm, and 265 mm. For x <105 mm, the highest OH mole frac-
tions occur at the walls. However, at x 105 mm a peak for OH is
immerging which indicates the onset of a gas-phase ignition. The
peak OH value continues to increase at the subsequent sections
until it starts to decrease after x 235 mm.
Figure 9 shows the H mole fraction transverse proles at differ-
ent streamwise sections. Near the inlet section the H mole fraction
is highest at the walls. However, a peak starts to appear away
from the wall for streamwise distances greater than 85 mm. This
peak increases with streamwise distances up to 165 mm where it
decreases with further increase in the streamwise distance. This
behavior is concurrent with that of OH. The O mole fraction trans-
verse proles are depicted in Fig. 10 for streamwise distances of
25 mm, 85 mm, 105 mm, 165 mm, 235 mm, and 265 mm. Here the
trend of the O mole fraction proles is similar to both mole frac-
tion proles of OH and H. It is possible to conclude that a form of
high gas-phase reaction rates occur in the range of streamwise dis-
tances of 85 mm105 mm. This could be attributed to local gas-
phase ignition.
The mole fraction transverse proles of H
2
O
2
are depicted in
Fig. 11 for streamwise distances of 25 mm, 85 mm, 105 mm,
165 mm, 235 mm, and 265 mm. Near the inlet section, the H
2
O
2
Fig. 6 Streamwise velocity, u, at streamwise distances of 25mm, 85mm, 105mm, 165mm, 235mm, and 265mm
Fig. 7 Transverse velocity, v, at streamwise distances of 25mm, 85mm, 105mm, 165mm, 235mm, and 265mm
Fig. 8 Mole fraction of OH at streamwise distances of 25mm, 85mm, 105mm, 165mm, 235mm, and 265mm
041201-6 / Vol. 134, APRIL 2012 Transactions of the ASME
Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 04/12/2013 Terms of Use: http://asme.org/terms
mole fraction essentially vanishes. It starts to form two clear peaks
for x ranging from 85 mm to 165 mm. The H
2
O
2
mole fraction
proles start to form one peak at the channel centerline for values
of x >165 mm where the temperature reaches its minimum value,
at each section. This behavior is dictated by two of the main ele-
mentary reactions that produce H
2
O
2
. These two reactions (R11
and R12f) increase with decreasing temperatures either because of
negative activation energy or negative temperature exponent [10].
Figure 12 shows the mole fraction transverse proles of HO
2
for
streamwise distances of 25 mm, 85 mm, 105 mm, 165 mm,
235 mm, and 265 mm. Near the channel walls, two HO
2
peaks
start to form for streamwise distances of 85 mm, 105 mm, and
165 mm. They collapse into a central peak for streamwise distan-
ces in the range of 235 mm265 mm. The production/consumption
rates of HO
2
are controlled by reactions R (5f) and R (5 b). The
rst reaction increases with decreasing gas temperatures while R
(5 b) is controlled by two temperature-conicting-effect terms
[10].
Figure 13 depicts the streamwise proles of the production rate
of H
2
O and consumption rates of H
2
and O
2
. It is interesting to
notice that the production rate of H
2
O is nearly equal to the con-
sumption rate of H
2
. Moreover, the consumption rate of H
2
is
nearly equal to twice the consumption rate of O
2
. This suggests
that the surface reactions can be summed up to a one step gas-
phase reaction as H
2
2 O
2
! H
2
O. In this case, the important
parameter would be the reaction rate in the above reaction. A cor-
relation for this reaction rate can be arrived at either experimen-
tally or from numerical data similar to the present data [12,14].
The H
2
O production rate decreases sharply in a short streamwise
distance of about 10 mm, as can be seen from Fig. 13. Figures 14
and 15 depict the surface coverage streamwise proles of surface
species Pt(s), O(s), OH(s), H
2
O(s), and H(s). The surface platinum
Fig. 9 Mole fraction of H at streamwise distances of 25mm, 85mm, 105mm, 165mm, 235mm, and 265mm
Fig. 10 Mole fraction of O at streamwise distances of 25mm, 85mm, 105mm, 165mm, 235mm, and 265mm
Fig. 11 Mole fraction of H
2
O
2
at streamwise distances of 25mm, 85mm, 105mm, 165mm, 235mm, and 265mm
Journal of Heat Transfer APRIL 2012, Vol. 134 / 041201-7
Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 04/12/2013 Terms of Use: http://asme.org/terms
decreases sharply from 1.0 to about 0.03 in a short distance of
about 5 mm; concurrently the O(s) surface coverage increases
sharply from 0.0 at the entrance to 0.97 at a streamwise distance
of about 5 mm, as can be seen from Fig. 14. Pt(s) then remains
nearly constant for most of the channel length, as can be seen
from Fig. 15. However, for x >150 mm the platinum surface cov-
erage gradually decreases to a value of 0.008. The reduction in the
platinum surface coverage implies that the adsorption reaction
rates must decrease. However, if the modications in the surface
reaction rate constant equation were included in the present work
then these reaction rates would further decrease. The channel sur-
face remains mostly covered with O(s) for the remaining length of
the channel, as can be seen from Fig.15. Figure 15 depicts also the
surface overages of OH(s), H
2
O(s), and H(s). Along the channel
walls, the surface species OH(s), H
2
O(s), and H(s) increase
sharply from 0.0 to 4.0 10
4
, 5.0 10
8
, and 1.5 10
8
,
Fig. 12 Mole fraction of HO
2
at streamwise distances of 25mm, 85mm, 105mm, 165mm, 235mm, and 265mm
Fig. 13 Surface production rate of H
2
O and Surface consumption rates of H
2
and O
2
Fig. 14 Surface coverage of surface species for 0<x<5 mm
041201-8 / Vol. 134, APRIL 2012 Transactions of the ASME
Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 04/12/2013 Terms of Use: http://asme.org/terms
respectively. These surface species continue to decrease along the
channel wall. At the channel exit section, the reductions in OH(s),
H
2
O(s), and H(s) are of orders of magnitudes ranging between
one and two. The streamwise pressure difference (p p
in
) prole
is shown in Fig. 16. It is clear that the pressure decreases linearly
along the streamwise direction. The pressure difference stream-
wise gradient is approximately equal to 133 N m
3
. This pres-
sure gradient could be useful in designing rectangular channel
catalytic burners.
4 Conclusions
The present parabolic-elliptic solution procedure is tested
against recent experimental and numerical data for reacting ows
in catalytic rectangular channels. Through this study several con-
clusions emerged:
1. The agreement between the present results and the experi-
mental data is generally good and in some cases surpasses
the published numerical data for the same ow. Moreover,
the present solution procedure is roughly two orders of mag-
nitude faster than the published fully elliptic numerical pro-
cedure for the same ow conditions. The H
2
O is
overpredicted for sections near the exit section and is attrib-
uted to higher values of the adsorption reaction constants in
the downstream area, where the surface platinum is mainly
covered with O(s). The rst 0.7 L of the channel is vital in
building up the surface species and the depletion of the
active platinum sites. The numerical results show high pro-
duction rates of H
2
O in the vicinity of streamwise distance
of about 15 L.
2. The channel ow computational results are compared with
recent detailed experimental data for similar geometry and
showed that the present numerical results for the gas temper-
ature, water vapor mole fraction, and hydrogen mole fraction
agree well with the experimental measurements, especially
in the rst 105 mm. However, some differences are observed
in the vicinity of the exit section of the rectangular channel.
The numerical results show that the production of water
vapor is very fast near the entrance section owed by a
much slower reaction rate. Gas-phase ignition is noticed
near the catalytic surface at a streamwise distance of about
120 mm. This gas-phase ignition manifests itself as a sudden
increase in the mole fractions of OH, H, and O.
Nomenclature
a
n
nite-difference coefcients due to combined convection
and diffusion, kg s
1
Cp constant-pressure specic heat, J kg
1
K
1
DH
j
enthalpy of reaction for reaction j, J mol
1
h heat transfer coefcient, W m
2
K
1
h
g
gas sensible enthalpy, J kg
1
k
c
mass transfer coefcient, kg m
2
s
1
L height of the computational domain, m
M molecular weight, kg kmol
1
N number of grid nodes
Nu Nusselt number
P gas pressure, N m
2
Pr Prandtl number
R universal gas constant, J mol
1
K
1
Re Reynolds number
R
s
surface reaction rate, mol cm
2
s
1
_ s
k
surface production or depletion rate of the surface species
k, mol cm
2
s
1
Sh Sherwood number
Fig. 15 Surface coverage of surface species for the total channel length (0<x >300mm)
Fig. 16 Streamwise pressure difference prole (p-pin)
Journal of Heat Transfer APRIL 2012, Vol. 134 / 041201-9
Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 04/12/2013 Terms of Use: http://asme.org/terms
SPR surface production rate, mol cm
2
s
1
T temperature, K
u
i
velocity in direction i, m s
1
U
1
uniform free stream axial velocity, m s
1
W
j
reaction rate of reaction j, mol cm
3
s
1
x streamwise distance, m
x
i
distance along direction i, m (here, x x
1
and y x
2
)
X mole fraction
[X] generalized concentration (gas phase: mol cm
3
, surface
phase: mol cm
2
)
y transverse distance, m
Y species mass fraction
Z surface site concentration, mol cm
2
z
k
surface site fraction of k species (surface coverage
fraction)
Greek Symbols
d thermal boundary layer thickness, m
C
h
thermal diffusivity, kg m
1
s
1
C
Yl
molecular diffusivity of species l, kg m
1
s
1
l dynamic viscosity, kg m
1
s
1
q density, kg m
3
c sticking coefcient
Subscripts
g gas
s surface
x at a distance x from the inlet section
y distance from the wall
References
[1] Cattolica, R. J., and Schefer, R. W., 1982, Effect of Surface Chemistry on the
Development of the (OH) in a Combustion Boundary Layer, 19th Symposium
(International) on Combustion, The Combustion Institute, Pittsburgh, PA,
pp. 311318.
[2] Cattolica, R. J., and Schefer, R. W., 1983, Laser Fluorescence Measurements
of the OH Concentration in a Combustion Boundary Layer, Combust. Sci.
Technol., 30, pp. 205212.
[3] Williams, W. R., Stenzel, M. T., Song, X., and Schmidt, L. D., 1991,
Bifurcation Behavior in Homogeneous-Heterogeneous Combustion: I. Experi-
mental Results Over Platinum, Combust. Flame, 84, pp. 277291.
[4] Song, X., Williams, W. R., Schmidt, L. D., and Aris, R., 1991, Bifurcation
Behavior in Homogeneous-Heterogeneous Combustion: II. Computations for
Stagnation-Point Flow, Combust. Flame, 84, pp. 292311.
[5] Warnatz, J., Allendorf, M. D., Kee, R. J., and Coltrin, M. E., 1994, A Model of
Elementary Chemistry and Fluid Mechanics in the Combustion of Hydrogen on
Platinum Surfaces, Combust. Flame, 96, pp. 393406.
[6] Deutschmann, O., Behrendt, F., and Warnatz, J., 1994, Modeling and Simula-
tion of Heterogeneous Oxidation of Methane on a Platinum Foil, Catal. Today,
21, pp. 461470.
[7] Raja, L. L., Kee, R. J., Deutschmann, O., Warnatz, J., and Schmidt, L. D., 2000,
Critical Evaluation of Navier-Stokes, Boundary-Layer, and Plug-Flow Models
of the Flow and Chemistry in a Catalytic-Combustion Monolith, Catal. Today,
59, pp. 4760.
[8] Coltrin, M. E., Kee, R. J., and Rupley, F. M., 1991, Surface Chemkin: A Gen-
eral Formalism and Software for Analyzing Heterogeneous Chemical Kinetics
at a Gas-Surface Interface, Int. J. Chem. Kinet., 23, pp. 11111128.
[9] Abou-Ellail, M. M., Gosman, A. D., Lockwood, F. C., and Megahed, I. E. A.,
1978, Description and Validation of a Three-Dimensional Procedure for
Combustion Chamber Flows, AIAA J. Energy, 2, pp. 7180 (also, published in
Turbulent Combustion, L. Kennedy, ed., AIAA, New York, Vol. 58,
pp. 163190).
[10] Tong, T. W., Abou-Ellail, M. M., Li, Y., and Beshay, K. R., 2004, Numerical
Computation of Reacting Flow in Porous Burners with an Extended CH
4
Air
Reaction Mechanism, Proceedings of HTFED04, 2004 ASME Heat Transfer/
Fluids Engineering Summer Conference, Charlotte, NC, Paper No. HT-FED
2004-56012.
[11] Deutschmann, O., Maier, L. I., Riedel, U., Stroemman, A. H., and Dibble, R.
W., 2000, Hydrogen Assisted Catalytic Combustion of Methane on Platinum,
Catal. Today, 59, pp. 141150.
[12] Tong, T. W., Abou-Ellail, M. M., and Li, Y., 2007, Mathematical Modeling of
Catalytic-Surface Combustion of Reacting Flows, AIAA J. Thermophys. Heat
Transfer, 21, pp. 512519.
[13] Incropera, F. P., and DeWitt, D. P., 2001, Fundamentals of Heat and Mass
Transfer, 5th ed., Wiley, New York.
[14] Tong, T. W., Abou-Ellail, M. M., and Li, Y., 2007, A Mathematical Model for
Heat and Mass Transfer in Methane-Air Boundary Layers With Catalytic Sur-
face Reactions, ASME J. Heat Transfer, 129, pp. 939950.
[15] Appel, C., Mantzaras, J., Schaeren, R., Bombach, R., Inauen, A., Kaeppeli, B.,
Hemmerling, B., and Stampanoni, A., 2002, An Experimental and Numerical
Investigation of Homogeneous Ignition in Catalytically Stabilized Combustion
of Hydrogen/Air Mixtures Over Platinum, Combust. Flame, 128, pp. 340368.
041201-10 / Vol. 134, APRIL 2012 Transactions of the ASME
Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 04/12/2013 Terms of Use: http://asme.org/terms

You might also like