International Journal of Thermal Sciences 49 (2010) 2363e2375

Contents lists available at ScienceDirect

International Journal of Thermal Sciences

journal homepage: www.elsevier.com/locate/ijts

Three-dimensional numerical simulation of uid ow with phase change

heat transfer in an asymmetrically heated porous channel
H.Y. Li a, K.C. Leong a, *, L.W. Jin a, J.C. Chai b
a
b

School of Mechanical and Aerospace Engineering, Nanyang Technological University, 50 Nanyang Avenue, Singapore 639798, Republic of Singapore
Mechanical Engineering Department, The Petroleum Institute, Abu Dhabi, United Arab Emirates

a r t i c l e i n f o

a b s t r a c t

Article history:
Received 13 February 2010
Received in revised form
17 June 2010
Accepted 27 July 2010

Fluid ow with phase change heat transfer in a three-dimensional porous channel with asymmetrically
heating from one side is numerically studied in this paper. The modied Kirchhoff method is used to
deal with the spatial discontinuity in the thermal diffusion coefcient in the energy equation. The
velocity and temperature elds, as well as the liquid saturation eld on the heated section of the wall
with different Peclet and Rayleigh numbers are investigated. The results show that the liquid ow
bypasses the two-phase zone, while the vapor ows primarily to the interface between the sub-cooled
liquid zone and the two-phase zone. An increase in the Peclet number decreases the two-phase region
while an increase in the Rayleigh number helps to spread the heat to a larger region of the domain. The
distribution of the liquid saturation on the heated section of the wall indicates that the minimum liquid
saturation increases with the increase of both the Peclet and Rayleigh numbers.
2010 Elsevier Masson SAS. All rights reserved.

Keywords:
Three-dimensional
Two-phase mixture model
Phase change
Porous media

1. Introduction
Fluid ow with phase change heat transfer in porous media
occurs in many engineering applications such as the drying
processes in the food [1] and paper industries [2], heat exchangers
[3] and chemical catalytic reactors [4]. In many of these systems,
two-phase ow with phase change in porous media is encountered.
Understanding the uid ow with phase change heat transfer in
such systems is of fundamental interest. Therefore, extensive
studies have been carried out by many investigators since the late
1970s [5e11].
Numerous studies have been performed to investigate singlecomponent uid ow with phase change heat transfer in porous
media both experimentally and numerically [12e19]. Comprehensive reviews of these studies are well documented in the books of
Nield and Bejan [20], Kaviany [21], Ingham and Pop [22], Vafai [23]
and Vadasz [24]. As numerical simulation can provide useful
information and preliminary results in complex systems with
phase change, it is thus helpful in engineering design and prediction. However, the difculties encountered in the numerical study
of uid ow with phase change heat transfer in porous media is due
to the strongly nonlinear and coupled nature of the governing

* Corresponding author. Tel.: 65 6790 5596; fax: 65 6792 4062.

E-mail address: mkcleong@ntu.edu.sg (K.C. Leong).
1290-0729/$ e see front matter 2010 Elsevier Masson SAS. All rights reserved.
doi:10.1016/j.ijthermalsci.2010.07.018

equations. The discontinuity in the thermal properties at the phase

change boundary requires careful consideration. In the numerical
simulation of the phase change problem, a front tracking approach
was needed to capture the evolving irregular interfaces as phase
change occurs. This may lead to the problem of singularity when
the interface touches the bottom or the top boundary. The model
developed by Wang [14], called the two-phase mixture model
addresses this difculty. Recently, Li et al. [25] used the two-phase
mixture model to study the transient behavior of uid ow with
phase change heat transfer in porous media. In their work, the
effects of heat ux locations on the ow and heat transfer elds
were investigated. The results showed that the liquid and vapor
ow elds, as well as the temperature and liquid saturation elds
exhibit distinctly different features for different heating locations.
In addition, the discontinuous diffusion coefcient in the energy
equation was handled with the modied Kirchhoff method [26]
in their work. The performance of this method and the harmonic
mean method [27] were compared. It was found that the modied Kirchhoff method was better in handling the rapid change in
the diffusion coefcient.
Generally, in the literature [12e19], numerical simulations are
based only on a two-dimensional (2D) setting which cannot
capture the three-dimensional (3D) effects which may be important under certain circumstances. Furthermore, large discrepancies
between numerical prediction and the experimental data may arise
from a 2-D simulation. Although the assumption of a 2D setting in
the numerical simulation is applicable in some very specic cases,

2364

H.Y. Li et al. / International Journal of Thermal Sciences 49 (2010) 2363e2375

Nomenclature
Cp
Ds
f(s)
g
hfg
h
H
j
J(s)
krl
krv
keff
K
L
p
Pe
q00
Ra
s
S
t
T
u

specic heat of uid (J/kg K)

capillary diffusion coefcient (m2/s)
hindrance function
gravity vector (m/s2)
latent heat of phase change (J/kg)
enthalpy (J/m3)
height of the simulation domain (m)
diffusive mass ux (kg/m2 s)
capillary pressure function
relative permeability of liquid
relative permeability of vapor
effective thermal conductivity (W/m K)
permeability of the porous medium (m2)
length of the simulation domain (m)
pressure (Pa)
Peclet number, Pe uinH/a
heat ux (W/m2)
Rayleigh number, Ra KHgrlCpl/ylkeff
liquid saturation
energy source term (W/m3)
time (s)
temperature ( C)
velocity vector (m/s)

a 3D numerical simulation is more desirable since it is more realistic for comparison with actual experimental data. It is especially
important for understanding and predicting the development of
complex ow structures and the two-phase zone for cases which
cannot be simplied to a 2D setting such as the case shown in
Fig. 1a. However, to the best of the knowledge of the authors, there
is no 3D numerical study of phase change heat transfer in porous
media based on the two-phase mixture model.
In the present study, a 3D numerical code is developed by
adopting the two-phase mixture model. The primary objective of
the current article is to study the uid ow with phase change heat
transfer in porous media in a 3D domain based on the developed 3D
code. It is a continuation of the authors previous study on phase
change heat transfer in porous media in a 2D setting [25]. The
velocity and temperature elds under different Rayleigh and Peclet
numbers in a 3D domain with asymmetrically heated from one side
are analyzed. In this article, the discontinuity in the thermal
diffusion coefcient at the phase change boundary is also treated
with the modied Kirchhoff method [26]. This 3D code can be
applied to simulate the uid ow and heat transfer for a large class
of phase change ow problems.

Greek symbols
a
thermal diffusivity of porous media (m2/s)
b
thermal expansion coefcient (1/K)
g
advection correction coefcient
G
diffusion coefcient
Gh
diffusion coefcient in enthalpy equation
Dr
rl  rv (kg/m3)
3
porosity
l
relative mobility
m
dynamic viscosity (kg/m s)
y
kinematic viscosity (m2/s)
r
density (kg/m3)
s
surface tension (N/m)
U
effective heat capacitance ratio
Subscripts
l
liquid
in
inlet
o
initial state
s
solid
sat
saturation
v
vapor

section and thus a two-phase zone is formed. Further increase in

the heat ux will lead to a superheated vapor zone. This could cause
dry-out in which no liquid exists on the heated section. The vapor
layer blankets the heated section of the wall and greatly reduces the
heat transfer between the heated section and the uid. Consequently, the local temperature at the dry-out location increases
sharply. This is unfavorable to the cooling system. Hence, the
present work considers only the case without any occurrence of
a superheated vapor zone.

2.2. Governing equations

The present model considers an isotropic and homogeneous
porous medium. The local thermal equilibrium model is used here.
This model assumes that the solid and uid phases are at the same
temperature. The governing equations obtained by adopting the
two-phase mixture model [14] are:
Conservation of mass:

vr
V$ru 0
vt

2. Mathematical formulations
2.1. Problem description
The schematic diagram of the problem is shown in Fig. 1. A 3D
channel with dimensions L  W  H is lled with a porous medium.
A nite heat source with constant heat ux is applied on one side
wall as shown in Fig. 1a. The rest of the channel wall is perfectly
insulated. An external pressure difference drives the sub-cooled
water with low temperature Tin through the channel. The owing
liquid is heated as it ows past the heated section of the wall. When
the heat ux is increased sufciently, boiling occurs at the heated

width of the simulation domain (m)

(1)

Conservation of momentum:

K
u  Vp  rk  ro g

(2)

Conservation of energy:


K Drhfg
vh
g
V$gh uh V$Gh Vh V f s
nv
vt

(3)

The mixture variables and the properties in Eqs. (1)e(3) are listed
in Table 1. The temperature and liquid saturation s can be calculated
from the enthalpy as:

H.Y. Li et al. / International Journal of Thermal Sciences 49 (2010) 2363e2375

8
h 2rl hvsat
>
>
h  rl 2hvsat  hlsat
>
>
rl Cpl
<
T
Tsat
 rl 2hvsat  hlsat < h  rv hvsat
>
>
h rv hvsat
>
>
: Tsat
 rv hvsat < h
rv Cpv

2365

(4)

8
>
>
>
<

1
h  rl 2hvsat  hlsat
h rv hvsat
 rl 2hvsat  hlsat < h  rv hvsat
s 
rl hfg rl  rv hvsat
>
>
>
:
0
 rv hvsat < h
(5)
Subscripts l and v in Eqs. (4) and (5) refer to the liquid and vapor,
respectively. The individual velocities of the liquid and vapor can be
recovered from

rl ul ll ru j

(6)

rv uv lv ru  j

(7)

where j is the total mass ux which is expressed as

j rl DsVs f s

K Dr

nv

(8)

2.3. Boundary conditions

The initial and boundary conditions for the present problem are
given as follows:
At t 0,



h rl Cpl Tin  2hvsat
u v w 0
Fig. 1. Schematic of the problem and the coordinate systems (a) the simulation
domain; (b) selected planes.

Table 1
Variables in the two-phase mixture model.
Variables

Expressions

Density

r rl s rv 1  s
ru rl ul rv uv

Enthalpy

rh rl shl rv 1  shv

Kinetic density

rk rl 1  bl T  To ll s rv 1  bv T  Tsat lv s

Viscosity
Advection correction
coefcient

Effective heat
capacitance ratio
Effective diffusion
coefcient

dT
U 3 rs Cps 1  3dH

Capillary diffusion
coefcient
Relative motilities

Ds

Hindrance function

f s

Gh

rl s rv 1  s
krl =nl krv =nv
rv =rl 1  s shvsat 1 ll  hlsat ll 
2hvsat  hlsat s rv hvsat =rl 1  s

1
dT

D keff dH
1 1  rv =rl hvsat =hfg
3K0:5 s

krl krv
J 0 s
nv =nl krl krv
krl =nl
krv =nv
ll s
; lv s
krl =nl krv =nv
krl =nl krv =nv

ml

krv krl =nl

krl =nl krv =nv

s3 ;

(10)

At the inlet x 0 and t > 0,



h rl Cpl Tin  2hvsat

(11)

u uin ;

(12)

v 0;

w 0

At the outlet x L and t > 0,

Velocity

gh

(9)

vh
0
vx

(13)

vu
vv
vw

0
vx
vx
vx

(14)

At the position of the wall that is heated with constant heat ux and
t > 0,

Gh vh
q00
r vy

(15)

u v w 0

(16)

At the non-heated portions of the side walls and t > 0,

Gh vh
0
r vy

(17)

u v w 0

(18)

Relative permeabilities

krl

Capillary pressure
function

Js 1:4171  s  2:1201  s2 1:2631  s3

krv 1  s

At the upper and lower walls and t > 0,

2366

K Drhfg
Gh vh
g 0
 f s
r vz
vv

H.Y. Li et al. / International Journal of Thermal Sciences 49 (2010) 2363e2375

(19)

u v w 0

(20)

2.4. Numerical procedure

The momentum equation [Eq. (2)] is different from the traditional NaviereStokes equation. In the present problem, the
momentum equation [Eq. (2)] is rst substituted into the continuity
equation [Eq. (1)] to obtain an equation for the pressure. The
resulting pressure equation is then solved by a line-by-line tridiagonal matrix algorithm. Upon obtaining the pressure, the
mixture velocity eld can be calculated using the momentum
equation [Eq. (2)]. The individual velocities of the liquid and vapor
can be obtained from the mixture velocity using Eqs. (6) and (7).
These velocities are stored at the interfaces of the control volumes.
The energy equation [Eq. (3)] can be written as a general convectiveediffusive equation of the form

!


vrf v ruj f
v
vf
G
S

vt
vxj
vxj
vxj

(21)

where f, G and S are dened as the dependent variables, diffusion

coefcient and source term. These equations are solved using the
nite-volume method [27]. The power-law is used to treat the
combined convection-diffusion term. Unlike the work of Wang [14],
the discontinuous diffusion coefcient in this problem is treated
with the modied Kirchhoff method [26]. With this, a smoother
interface between the liquid and two-phase regimes can be
captured. The performance of the modied Kirchhoff method has
been demonstrated in the authors previous work [25]. The
descriptions of this method are well documented in the work of
Voller and Swaminathan [26]. The current work will adopt the
same terminology of modied Kirchhoff method for easy reference. However, unlike the work of Voller and Swaminathan [26]
where the integral variable is temperature, the present study
adopts enthalpy h as the integral variable given the form of the
energy equation expressed in terms of enthalpy. The performance
of this method is compared with the harmonic mean method [27].
It shows that the former approach produces smoother isothermal
lines. These simulations were carried out on a desktop with Intel (R)
Core (TM) 2 Quad CPU Q9450 @ 2.66GHz and 3.25GB RAM. A
relative error of less than 106 is required for both the velocity and
temperature elds between successive iterations to achieve
convergence.
2.5. Code validation
A 2D code has been developed in a previous work by the authors
[25]. This 2D code was validated with the experimental results of
Easterday et al. [28]. With symmetric boundary conditions enforced
along the width direction in the 3D domain (uid ows along
length (x) direction and heating along height (z) direction), the
results from the present 3D code should reduce to its 2D counterpart with these boundary conditions. This comparison is given in
Fig. 2. The temperature distributions obtained from the 2D and 3D
codes were found to overlap at different locations. The validity of
the 3D code is therefore demonstrated.
3. Results and discussion
The present study focuses on uid ow with phase change heat
transfer in a 3D porous channel. For all cases studied, the width W

Fig. 2. Comparison of 3D results under symmetric boundary conditions with 2D

results (a) x 1/3L; (b) x 1/2L.

and height H of the channel are prescribed as 20 mm. The length of

the heated section on the side wall is 20 mm. The lengths of the
unheated sections of the side wall namely, l1/H and l2/H are set to 1.
These lengths are chosen because the inlet and exit boundary
conditions have no effects on the solution. Water enters the domain
with a uniform velocity and at a constant temperature of 22  C. A
constant heat ux q00 200 kW/m2 is imposed at the heated section
of the side wall. The porous medium considered in the current
simulation is the graphite foam developed at Oak Ridge National
Laboratory (ORNL), USA in 1997 [29]. This material consists of
predominantly spherical pores with small openings between the
ligaments. The porosity of this graphite foam is 0.75 and its thermal
properties are well documented in the report of Klett et al. [30]. The
effective thermal conductivity for this graphite foam is calculated
based on the model proposed by Tee et al. [31].
A grid independency study was carried out and the results show
that a mesh of 60  20  20 control volumes with Dt 1.0 s
produces a grid-independent solution. All subsequent computations were performed using this time-step and mesh size.

H.Y. Li et al. / International Journal of Thermal Sciences 49 (2010) 2363e2375

The effects of the Rayleigh and Peclet numbers are discussed in

the present article. There are two Rayleigh numbers based on
different ow regimes [13,32]. One is for the liquid regime and is
based on the temperature difference. The second Rayleigh number is
for the two-phase regime and involves the density difference
between the liquid and vapor phases. Detailed information can be
found in the works of Ramesh and Torrance [13,32]. As the primary
interest in the current work is to study the uid ow and heat transfer
behavior in the two-phase regime in the porous media, the authors
have chosen the form of the Rayleigh number based on the twophase regime. The mathematical expression of this Ra number is:

Kg rl Cpl H
Rah
nl keff

(22)

The Peclet number is dened as:

u H
Peh in

(23)

3.1. Effects of Peclet number on uid ow and heat transfer

In the current work, 3D uid ow with phase change heat
transfer in a porous channel (Fig. 1a) with asymmetrically heating
from one side is studied. In order to describe the characteristics of
the uid ow clearly, results for a series of planes are shown. Fig. 1b
shows the selected planes. Although transient uid ow with phase
change heat transfer was carried out as part of the simulation, only
the steady-state results are discussed in the following sections.

2367

For the ow elds in the present article, the interface between the
sub-cooled liquid and the two-phase zones, viz. the condensation
front is shown as the solid curve.
Fig. 3 shows the ow elds for Ra 226 and Pe 0.06. As the
liquid ows into the channel with a uniform velocity prole, it
absorbs heat when it ows past the heated section of the wall. Its
temperature increases, leading to a lower density. Acted upon by
the buoyancy force, the heated liquid ows in a slightly upward
manner. The liquid upstream of the heated section of the wall,
which is cooler and therefore denser, tends to ow downward to
replace the upward ow of the heated liquid. As the buoyancy force
becomes important, this combination of downward ow of the
cooler liquid from the upstream of the heated section and upward
ow of the hotter liquid downstream results in a circulatory ow.
This circulation ow is located at the leading edge of the two-phase
zone (Fig. 3a). The recirculation hinders the liquid upstream from
penetrating the circulation cell. Therefore, the liquid in the subcooled liquid zone from the inlet ows decisively downward to
bypass the recirculation (Fig. 3a). When it approaches the
condensation front, it shows a slightly upward motion. This slightly
upward motion of the liquid is due to a decrease in the permeability
which causes the sub-cooled liquid to ow through the two-phase
zone. The two-phase zone is occupied by both the liquid and vapor.
The presence of the generated vapor in the void space reduces the
permeability for the sub-cooled liquid to ow through the twophase zone. The liquid therefore has difculty in penetrating the
void space in the porous media which also contains vapor. When
the incoming ow is weak, the liquid in the sub-cooled liquid zone

Fig. 3. Liquid velocity in different planes for Pe 0.06 and Ra 226 (a) the xz plane with y/H 0.5; (b) the xy plane with z/H 0.5; the yz plane at (c) x/H 0.1; (d) x/H 0.3;
(e) x/H 0.5.

2368

H.Y. Li et al. / International Journal of Thermal Sciences 49 (2010) 2363e2375

is blocked as it hits the condensation front. The blocked liquid thus

reverses its ow direction to the inlet as the two-phase zone
presents a region of high ow resistance. This can be seen in the xy
plane with z/H 0.5 (Fig. 3b). Further away from the heated section
of the wall, less vapor is generated. The ow resistance caused by
the presence of the vapor to the owing liquid is therefore smaller.
Hence, the liquid away from the heated section in the sub-cooled
liquid zone ows across the condensation front into the two-phase
zone (Fig. 3b). The volume fraction occupied by the liquid in the
two-phase zone is called the liquid saturation [21], denoted as s.
The vapor volume fraction is given by 1-s. In the locations near the
heated section of the wall, more liquid vaporizes, leading to
a smaller liquid saturation. In contrast, at locations away from the
heated section, the liquid saturation is higher. This difference in the
liquid saturation causes a liquid saturation gradient in the twophase zone. Such a liquid saturation gradient gives rise to the
capillary-induced force. The liquid is driven by the capillary force to
ow from the locations with high volume fraction of the liquid to
the locations with low volume fraction of the liquid; viz. the
capillary force drives the liquid in the two-phase zone towards the
heated section of the wall. This can be seen from the motion of
the liquid in the vicinity of the heated section of the wall shown in
the xy plane with constant z/H 0.5 (Fig. 3b). In the region with low
liquid saturation gradient, the liquid ows approximately parallel
to the outlet. Under the combined effect of the buoyancy and
capillary forces, the liquid from the inlet divides into two streams.
While one stream ows backwards toward the inlet in a circulatory
manner, the other stream ows to the outlet. In the yz plane with
constant x/H 0.1 (Fig. 3c), the two-phase zone occupies a small
portion of the plane in the upper-left corner. Given the buoyancy
force and the weak ow from inlet, the vapor rises and spreads to
the upstream of the heated section. At certain elevation, the
buoyancy force assists the liquid to ow upward while the liquid
below such elevation ow directly downward (Fig. 3c). Driven by
the capillary force, this downward ow was dragged to ll the
empty space caused by the upward ow in a slightly lateral manner

to the heated section. This is different to the case of single phase

natural convection in a cavity with heating from one side. As the
uid ows into the middle of the domain with the heated section at
x/H 0.3, an extensive region of the plane is covered by the twophase zone. The sub-cooled liquid zone is pushed to a small area
located in the lower right corner of the plane (Fig. 3d). This twophase zone grows to become almost symmetrical to the diagonal of
the plane. When the liquid leaves the heated section at x/H 0.5
(Fig. 3e), the two-phase zone expands. As most of the uid ows
parallel to the outlet of the domain (Fig. 3a and b), the liquid ow in
the yz plane becomes insignicant.
Fig. 4 shows the characteristics of the vapor ow in different
planes. The vapor is rst generated on the heated section of the wall
and then it moves away from the heated source in different
directions. A portion of the vapor is carried downstream in the form
of a vaporeliquid mixture while the other portion ows upward
due to buoyancy force (Fig. 4a and b). As a result, the upper portion
of the channel is progressively richer in vapor along the ow
direction. On the other hand, the lower portion of the channel is
relatively rich in liquid. This phase separation is caused by the large
density difference between the liquid and vapor. The quantity of
the vapor increases along the ow direction since more liquid
vaporizes (Fig. 4d). It is noted that the vapor ows toward the
condensation front where it is condensed.
Figs. 3e8 demonstrate the effects of Peclet number, Pe on the
velocity elds for a xed Rayleigh number of 226. The Peclet
number is directly related to the inlet velocity. At low Pe (or low
inlet velocity), the buoyancy force is dominant. The circulatory ow
caused by the buoyancy force is therefore strengthened. The eye of
the recirculatory ow is shifted upstream. Generally, a smaller Pe
suggests a weaker convection effect, i.e. a longer residence time of
the uid within the channel. The uid absorbs more heat, leading to
the generation of more vapor. Therefore, a larger two-phase zone is
developed in the channel. The interface shape between the subcooled liquid zone and the two-phase zone is a direct result of the
strong upward vapor ow. At Pe 0.12, the increase in the relative

Fig. 4. Vapor velocity in different planes for Pe 0.06 and Ra 226 (a) the xz plane with y/H 0.5; (b) the xy plane with z/H 0.5; the yz plane at (c) x/H 0.1; (d) x/H 0.3;
(e) x/H 0.5.

H.Y. Li et al. / International Journal of Thermal Sciences 49 (2010) 2363e2375

2369

Fig. 5. Liquid velocity in different planes for Pe 0.12 and Ra 226 (a) the xz plane with y/H 0.5; (b) the xy plane with z/H 0.5; the yz plane at (c) x/H 0.1; (d) x/H 0.3;
(e) x/H 0.5.

strength of the inlet ow reduces the circulatory ow (Fig. 5a). The

inertial of the liquid increases so that it gradually changes the
direction for the back ow of the liquid near the condensation front
in the xy plane (Fig. 5b). When Pe is sufciently high (Pe 0.24), the
strong incoming liquid pushes the liquid near the condensation
front in the xy plane decisively downstream (Fig. 7b). With the
increase of the inlet velocity, the heat transfer from the heated

section of the wall to the uid is increased. A small amount of vapor

is generated and the two-phase zone shrinks substantially. It is
observed that the two-phase zone recedes appreciably away from
the inlet (Fig. 7b). The vapor in the two-phase zone ows partially
to the condensation front and partially to the outlet.
Fig. 9 shows the isotherms within the porous medium. The
temperature difference between successive isotherms is 15  C. The

Fig. 6. Vapor velocity in different planes for Pe 0.12 and Ra 226 (a) the xz plane with y/H 0.5; (b) the xy plane with z/H 0.5; the yz plane at (c) x/H 0.1; (d) x/H 0.3;
(e) x/H 0.5.

2370

H.Y. Li et al. / International Journal of Thermal Sciences 49 (2010) 2363e2375

Fig. 7. Liquid velocity in different planes for Pe 0.24 and Ra 226 (a) the xz plane with y/H 0.5; (b) the xy plane with z/H 0.5; the yz plane at (c) x/H 0.1; (d) x/H 0.3;
(e) x/H 0.5.

isotherm of 100  C is not smooth. This is caused by the interpolation error in plotting the data, a limitation of the plotting software.
The 100  C isotherm is the interface between the sub-cooled liquid
and two-phase zone. Within the two-phase zone, the liquid-vapor
mixture remains at the saturation point of 100  C. The results from
3D simulation are more realistic compared with those from a 2D
simulation since the heat transfer eld can be clearly seen in the

whole domain. For different Peclet numbers, the temperature

contours are all inclined towards the outlet due to the incoming
sub-cooled liquid ow. The presence of the circulatory ow in the
upper left-hand corner (Figs. 3a, 5a and 7a) of the plane actually
hinders cooler liquid from reaching the heated section. This results
in poor heat transfer from the heated section to the surrounding
liquid. Therefore, denser temperature contours are observed in the

Fig. 8. Vapor velocity in different planes for Pe 0.24 and Ra 226 (a) the xz plane with y/H 0.5; (b) the xy plane with z/H 0.5; the yz plane at (c) x/H 0.1; (d) x/H 0.3;
(e) x/H 0.5.

H.Y. Li et al. / International Journal of Thermal Sciences 49 (2010) 2363e2375

2371

Fig. 9. Temperature distributions in the 3D domain for Ra 226 and (a) Pe 0.06;
(b) Pe 0.12; (c) Pe 0.24.

region with circulatory ow. This indicates a large temperature

gradient. For large Pe, the isotherms are prominently inclined
towards the outlet due to the strong incoming ow. The increase of
Pe increases the convection heat transfer between the heated
section of the wall and the uid. The region covered by the twophase zone is therefore reduced greatly.
The variation of the liquid saturation s on the wall with heated
section under different Peclet numbers is shown in Fig. 10. The
liquid saturation s is kept at a constant value of 1 at the entrance for
different inlet velocities, indicating a sub-cooled liquid at this
region. However, it begins to drop as the liquid approaches the
heated section of the wall, reaches a minimum somewhere within
the heated section and then increases at the point in which the
liquid leaves the heated section as a liquidevapor mixture.
Generally, s reduces along the direction of the height (z direction) as
the vapor ows upward; indicating that the upper wall is relatively

Fig. 10. Liquid saturation on the wall with heated section for Ra 226 and
(a) Pe 0.06; (b) Pe 0.12; (c) Pe 0.24.

rich in vapor. Special attention should be paid to such problems in

engineering applications. An implication of this result could be that
in thermal systems with heat sources located at the side walls,
a dry-out zone could appear near the upper wall within the heated
section. The temperatures at these locations would increase
signicantly and local hot spots may be induced. This could be
potentially disastrous for sensitive electronic devices and clearly
demonstrates the need for detailed 3D studies to accurately capture
the locations of minimum liquid saturation. The increase of Pe

2372

H.Y. Li et al. / International Journal of Thermal Sciences 49 (2010) 2363e2375

Fig. 11. Liquid velocity in different planes for Pe 0.06 and Ra 113 (a) the xz plane with y/H 0.5; (b) the xy plane with z/H 0.5; the yz plane at (c) x/H 0.1; (d) x/H 0.3;
(e) x/H 0.5.

pushes the two-phase zone downstream. As a result, the location

where s drops from 1 recedes away from the inlet. The minimum
liquid saturation also increases with the increase of the inlet
velocity.
3.2. Effects of Rayleigh number on uid ow and heat transfer
Attention is now focused on the effects of Rayleigh number on
the velocity, temperature and liquid saturation elds. The Rayleigh

number, Ra is associated with buoyancy driven ow. According to

the denition of Ra adopted in the present work, a change in Ra
while keeping the other parameters constant implies a variation in
the permeability of porous media. In the current study, three
different Rayleigh numbers, viz. three different permeabilities of
the porous media, are studied. For all cases, the Peclet number is set
to 0.24.
The ow elds for both liquid and vapor under different Rayleigh numbers are shown in Figs. 11e14. Generally, a large Ra

Fig. 12. Vapor velocity in different planes for Pe 0.06 and Ra 113 (a) the xz plane with y/H 0.5; (b) the xy plane with z/H 0.5; the yz plane at (c) x/H 0.1; (d) x/H 0.3;
(e) x/H 0.5.

H.Y. Li et al. / International Journal of Thermal Sciences 49 (2010) 2363e2375

2373

Fig. 13. Liquid velocity in different planes for Pe 0.06 and Ra 56 (a) the xz plane with y/H 0.5; (b) the xy plane with z/H 0.5; the yz plane at (c) x/H 0.1; (d) x/H 0.3;
(e) x/H 0.5.

suggests signicant buoyancy effect. In this case, buoyancy works

against convection. At high Rayleigh number, Ra 226, as shown in
Fig. 3, the buoyancy force produces a strong circulatory ow in the
xz plane. The recirculation deects the liquid from inlet to ow
downwards. Such a downward ow is accelerated in a narrow
region adjacent to the lower wall. The circulatory ow reduces

convection heat transfer near the upper wall while the accelerated
liquid increases convection near the lower wall. Therefore, the twophase zone blankets the upper wall in a larger region compared
with that on the lower wall, leading to a more inclined condensation front. This is different from what was observed at low Rayleigh
number. With the decrease of Ra, the buoyancy force becomes

Fig. 14. Vapor velocity in different planes for Pe 0.06 and Ra 56 (a) the xz plane with y/H 0.5; (b) the xy plane with z/H 0.5; the yz plane at (c) x/H 0.1; (d) x/H 0.3;
(e) x/H 0.5.

2374

H.Y. Li et al. / International Journal of Thermal Sciences 49 (2010) 2363e2375

Fig. 15. Temperature distributions in the 3D domain for Pe 0.06 and (a) Ra 113; (b)
Ra 56.

insignicant. This reduces the circulatory ow. At Ra 56 (Fig. 13),

the recirculation near the upper wall is almost curtailed by the
incoming ow. The leading edge of the two-phase zone on the
upper wall is pushed further downstream. The condensation front
becomes steeper (Fig. 13a). Unlike the case for Ra 226, no twophase zone is observed in the yz plane with constant x/H 0.1 at
Ra 113 and Ra 56 (Figs. 11c and 13c). The vapor ows in a similar
fashion for different Ra, having a direction which is primarily
normal to the condensation front.
The temperature contours under Ra 113 and Ra 56 are
shown in Fig. 15(a) and (b), respectively. Comparisons are made
against Ra 226 in Fig. 3. As indicated by the temperature
contours, a large region of the domain is blanketed by the twophase zone. The buoyancy force assists the vapor moving from the
heated section of the wall to expand to a wide region. The heat is
not only expelled from the heated source downstream, but also
swept upstream. An implication of this result in engineering
applications could be that the region which is rich in vapor can be
signicantly far away from the heat source. Hence, a detailed study
of two-phase heat transfer becomes necessary.
Fig. 16 shows the distribution of the liquid saturation on the wall
with heated section for Ra 113 and Ra 56. A similar prole is
observed for different Rayleigh numbers. For ow with large Ra,
a larger s is to be expected. For example, s reaches a minimum of
about 0.2 for Ra 56, but 0.3 for Ra 113 and 0.4 for Ra 226. Of
particular interest is the effect of the buoyancy force, although it
helps to spread the heat to a large region of the domain, it also, to
some extent, helps to promote mixing of the uid, reducing the
liquid saturation in some local spots. Therefore, the minimum
liquid saturation occurs at low Rayleigh number.

Fig. 16. Liquid saturation on the wall with heated section for Pe 0.06 and (a)
Ra 113; (b) Ra 56.

4. Conclusions
A 3D numerical study has been carried out for uid ow with
phase change heat transfer in a porous channel with asymmetrically heating from one side. The velocity, temperature and liquid
saturation elds under different Peclet and Rayleigh numbers were
considered. The liquid ow in the xz plane shows a circulatory ow
at the leading edge of the two-phase zone. Such a recirculation is
reduced with the increase of Pe and decrease of Ra. The liquid in the
xy plane bypasses the two-phase zone. In the two-phase zone, the
capillary force drives the liquid towards the heated section while
vapor ows primarily to the condensation front. An increase of
Peclet number decreases the size of the two-phase zone, while the
increase of Rayleigh number helps to spread the heat to a larger
region of the domain. The minimum liquid saturation s on the wall
with heated section increases with the increase of both Pe and Ra.
Acknowledgment
The authors gratefully acknowledge the nancial support
provided under Defence Science and Technology Agency, Singapore
Grant no. DSTA-NTU-DIRP/2005/01 for the work described in this
paper.
References
[1] P.P. Tripathy, S. Kumar, Modeling of heat transfer and energy analysis of
potato slices and cylinders during solar drying, Appl. Therm. Eng. 29 (2009)
884e891.

H.Y. Li et al. / International Journal of Thermal Sciences 49 (2010) 2363e2375

[2] A. Watzl, M. Rckert, Industrial through-air drying of nonwovens and paper
basic principles and applications, Drying Technol. 16 (1998) 1027e1045.
[3] G.F. Naterer, C.H. Lam, Transient response of two-phase heat exchanger with
varying convection coefcients, ASME J. Heat Transfer 128 (2006) 953e962.
[4] R.H. Nibbelke, J.H.B.J. Hoebink, G.B. Marin, Kinetically induced multiplicity of
steady states in integral catalytic reactors, Chem. Eng. Sci. 53 (1998) 2195e2210.
[5] C.H. Sondergeld, D.L. Turcotte, An experimental study of two-phase convection in a porous medium with applications to geological problems, J. Geophys.
Res. 82 (1977) 2045e2053.
[6] H.H. Bau, K.E. Torrance, Boiling in low-permeability porous material, Int.
J. Heat Mass Transfer 25 (1982) 45e55.
[7] H.H. Bau, K.E. Torrance, Thermal convection and boiling in a porous medium,
Lett. Heat Mass Transfer 9 (1982) 431e441.
[8] H.H. Bau, K.E. Torrance, Low Rayleigh number thermal convection in a vertical
cylinder lled with porous materials and heated from below, ASME J. Heat
Transfer 104 (1982) 166e172.
[9] C.Y. Wang, C. Beckermann, A two-phase mixture model of liquid-gas ow and
heat transfer in capillary porous media, I. Formulation, Int. J. Heat Mass
Transfer 36 (1993) 2747e2758.
[10] Z.Q. Chen, P. Cheng, T.S. Zhao, An experimental study of two phase ow and
boiling heat transfer in bi-dispersed porous channels, Int. Comm. Heat Mass
Transfer 22 (2000) 293e302.
[11] C.H. Sondergeld, D.L. Turcotte, Flow visualization studies of two-phase thermal
convection in a porous layer, Pure Appl. Geophys. 117 (1978) 321e330.
[12] P.S. Ramesh, K.E. Torrance, Stability of boiling in porous media, Int. J. Heat
Mass Transfer 33 (1990) 1895e1908.
[13] P.S. Ramesh, K.E. Torrance, Numerical algorithm for problems involving
boiling and natural convection in porous materials, Numer. Heat Transfer, Part
B: Fundamentals 17 (1990) 1e24.
[14] C.Y. Wang, A xed-grid numerical algorithm for two-phase ow and heat
transfer in porous media, Numer. Heat Transfer, Part B: Fundamentals 32
(1997) 85e105.
[15] T.S. Zhao, P. Cheng, C.Y. Wang, Buoyancy-induced ows and phase-change
heat transfer in a vertical capillary structure with symmetric heating, Chem.
Eng. Sci. 55 (2000) 2653e2661.
[16] T.S. Zhao, Q. Liao, Mixed convective boiling heat transfer in a vertical capillary
structure heated asymmetrically, J. Thermophys. Heat Transfer 13 (1999)
302e307.

2375

[17] M. Najjari, S.B. Nasrallah, Effects of latent heat storage on heat transfer in
a forced ow in a porous layer, Int. J. Therm. Sci. 47 (2007) 825e833.
[18] M. Najjari, S.B. Nasrallah, Heat transfer between a porous layer and a forced
ow: inuence of layer thickness, Drying Technol. 27 (2009) 336e343.
[19] K. Yuki, J. Abei, H. Hashizume, S. Toda, Numerical investigation of thermouid
ow characteristics with phase change against high heat ux in porous media,
ASME J. Heat Transfer 130 (2008) 1e12.
[20] D.A. Nield, A. Bejan, Convection in Porous Media, third ed. Springer, New York,
2006.
[21] M. Kaviany, Principles of Heat Transfer in Porous Media, second edition.
Springer-Verlag, New York, 1995.
[22] D.B. Ingham, I. Pop (Eds.), Transport Phenomena in Porous Media III, Elsevier,
Oxford, 2005.
[23] K. Vafai, Handbook of Porous Media, second ed. Taylor & Francis, New York,
2005.
[24] P. Vadasz, Emerging Topics in Heat and Mass Transfer in Porous Media.
Springer, New York, 2008.
[25] H.Y. Li, K.C. Leong, L.W. Jin, J.C. Chai, Transient two-phase ow and heat
transfer with localized heating in graphite foams, Int. J. Therm. Sci. 49 (2010)
1115e1127.
[26] V.R. Voller, C.R. Swaminathan, Treatment of discontinuity thermal conductivity in control-volume solutions of phase change problems, Numer. Heat
Transfer, Part B: Fundamentals 24 (1993) 161e180.
[27] S.V. Patankar, Numerical Heat Transfer and Fluid Flow. Hemisphere Publishing
Corporation, 1980.
[28] O.T. Easterday, C.Y. Wang, P. Zheng, A numerical and experimental study of
two-phase and heat transfer in a porous formation with localized heating
from below, Proc. ASME Heat Transfer Fluid Eng. Division 321 (1995)
723e732.
[29] J.W. Klett, Process for making carbon foam, US Patent 6033506 (2000).
[30] J.W. Klett, R. Hardy, E. Romine, High thermal conductivity, mesophase-pitchderived carbon foam: effect of precursor on structure and properties, Carbon
38 (7) (2000) 953e973.
[31] C.C. Tee, J.W. Klett, D.P. Stinton, N. Yu, Thermal conductivity of porous carbon
foam, Proc. of the 24 th Biennial Conference on Carbon, Charleston, USA, 1999.
[32] C.Y. Wang, C. Beckermann, C. Fan, Numerical study of boiling and natural
convection in capillary porous media using the two-phase mixture model,
Numer. Heat Transfer, Part A 26 (1994) 375e398.