International Journal of Heat and Mass Transfer 75 (2014) 624636

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer

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

Higher order compact computations of transient natural convection

in a deep cavity with porous medium
Swapan K Pandit , Anirban Chattopadhyay
Integrated Science Education and Research Centre (ISERC), Visva-bharati, Santiniketan 731235, India

a r t i c l e

i n f o

Article history:
Received 5 July 2013
Received in revised form 9 March 2014
Accepted 29 March 2014
Available online 6 May 2014
Keywords:
Natural convection
Higher order compact method
DarcyBrinkmann extended model
DarcyForchheimer model
Porous medium

a b s t r a c t
In this paper, an extension of the fourth order compact scheme on nonuniform grids (Pandit et al. (2007)
[28]) is proposed for solving two dimensional (2D) unsteady natural convection ows in a rectangular
cavity (with different aspect ratios) lled with a uid saturated porous medium. The bottom wall of
the cavity is uniformly and non-uniformly heated and the top wall is adiabatic while the vertical walls
are cold maintained at constant temperature. We have used streamfunction (w)vorticity (f) formulation
of NavierStokes equations with the consideration of Brinkmann-extended Darcy model to simulate the
momentum transfer in the porous medium. The streamfunctionvorticity and the energy equations are
all solved as a coupled system of equations for the ve eld variables consisting of streamfunction, vorticity, two velocities and temperature. In this wf formulation, the temperature gradient source term also
has been treated as fourth order compact. The higher order compact scheme adopted in the present study
yields consistent performance for a wide range of key parameters e.g. Rayleigh number Ra (from 103 to
108 ), Darcy number Da (from 105 to 103 ). Results are presented in the form of streamline and isotherm
plots as well as the plots of Nusselt number at the heat source surface under different conditions. The
present scheme is not only robust as evidenced from computations at higher Ra, but also accurate as
is seen from comparisons with reliable existing results.
2014 Elsevier Ltd. All rights reserved.

1. Introduction
Natural convection in a square cavity has been studied widely
due to a large number of engineering and technical applications
and has attracted many researchers interest. In recent years, there
has been considerable attention to study the ow in a thermally
driven square cavity through porous media due to large number
of technical applications such as packed sphere beds, chemical catalytic reactors, grain storage, geothermal reservoirs, solidication
of casting, crude oil production etc. In this context, the readers
are refereed to [25,912,17,18,21,23] (as well as the references
therein). Experimentally, uid ow deviations from Darcys law
have long been observed. Various terms, such as Darcy ow, nonDarcy ow, turbulent ow, inertial ow, high velocity ow, etc.,
have been used to describe this behavior. For very high velocities
in porous media, inertial effects can also become signicant and
in such situations, non-darcy behavior is important for describing
uid ow. Vafai and Tien [20] presented an in-depth analysis of
Corresponding author. Tel.: +91 9475850133.
E-mail addresses: swapankumar.pandit@visva-bharati.ac.in
animath81@rediffmail.com (A. Chattopadhyay).
http://dx.doi.org/10.1016/j.ijheatmasstransfer.2014.03.079
0017-9310/ 2014 Elsevier Ltd. All rights reserved.

(S.K

Pandit),

boundary and inertia effects. It may be mentioned here that Amiri

and Vafai [8] conducted a comprehensive transient analysis of
incompressible ow through a packed bed, taking into the account
the inertial and viscous forces. They found that excluding the inertial effects would shorten the time needed to reach steady-state.
This was expected since the absence of the inertia effects would
cause a higher ow rate, which results in a higher rate of heat
transfer between the uid and the solid, achieving steady state
conditions in a relatively shorter time. In addition, for a given ow
driving force (pressure gradient or buoyancy) the results have
shown that each of the three effects, i.e. inertia, boundary and
velocity-square, reduces the heat transfer rate [22]. The buoyancy
driven convection in a porous cavity heated differentially in the
horizontal side has been analyzed by Walker and Homsy [10] by
using different techniques and found fairly good agreement with
each other as well as experimental results. However, most of the
existing studies in the literature dealt primarily with the mathematical simplications based on Darcys law, which neglect the
effects of the solid boundary, inertial forces, and variable porosity
on ow through porous media. When relaxing some assumptions
one gets more complex models. For instance, assuming that the
frictional effects in the uid are not negligible one gets the

S.K Pandit, A. Chattopadhyay / International Journal of Heat and Mass Transfer 75 (2014) 624636

625

Nomenclature
AR
Da
Ra
Pr
g
Nu
Nu
K
J
H
L
p
P
T
T
T0
Th
Tc
u; v
x; y
t

aspect ratio (H=L)

Darcy number
Rayleigh number
Prandtl number
acceleration due to gravity (ms2)
local Nusselt number
average Nusselt number
permeability of the porous medium
Jacobian
cavity height (m)
cavity width (m)
pressure (Pa)
dimensionless pressure
temperature (K)
dimensionless temperature
reference temperature (K)
temperature of hot bottom wall (K)
temperature of cold vertical wall (K)
dimensionless velocities in x; y directions respectively
dimensionless distances along x and y coordinate
respectively
time

Brinkman model. Further, if one relaxes the hypothesis of negligible inertial effects one gets a series of nonlinear models, depending
on the way interaction forces are modelled. The aforesaid discussions motivate to study natural convection without considering
the Forchheimer inertia term.
Many numerical methods, including nite difference, nite element, nite volume, and lattice boltzman methods have been used
to investigate the steady natural convection in a square cavity.
Most of these numerical schemes are either rst-order or secondorder accurate in space, particularly the central difference ones
have been used in a large number because of their straightforwardness in application. In keeping with this trend, in most of the previous attempts to tackle the problem of natural convection, the
schemes were at most second-order accurate in space. Also, whenever there has been attempts to solve for the transient ows, they
are conned invariably to uniform space grids. However, a very little work has been done for the case of unsteady and transient ow
situations.
Lauriat and Prasad [23], and Pop et al. [9,29] have made fruitful contributions to this area. However, their scheme could not
fully exploit the advantages associated with using nonuniform
grids, particularly that of mesh grading to resolve smaller scales
in the regions of large gradients in the physical domain. The
model commonly used of a porous cavity with both the vertical
walls maintained at constant temperatures, while the horizontal
walls are adiabatic. Some investigations have been made in the
past [14,15] to focus on natural convection in porous medium
due to uniform heating from below. Recently, Basak et al. [16]
have studied the uniform and nonuniform heating of the bottom
wall. They have considered the Brinkmann-extended Darcy model
i.e. by neglecting the inertia term from Forchheimer Darcy model
to study natural convection using Penalty-Galerkin nite-element
method. Almost all of these works presented results for Ra 103
to Ra 106 . However, note that most methods are available to
solve this problem in moderate Rayleigh numbers until 106 , and
therefore it is still necessary to develop accurate and efcient
numerical methods suitable for a wide range of Rayleigh numbers. To the best of our knowledge, no studies have been found
in the literature on transient natural convection in a uid saturated porous medium using higher order compact scheme on

h; k

step lengths in horizontal and vertical directions

respectively

Greek symbols
m
kinematic viscosity m2 =s
q
density (kg=m3 )
a
thermal diffusivity (m2 =s)
b
thermal expansion coefcient (1/K)
w
stream function
f
vorticity
n
horizontal coordinate in a unit square in computational
plane
g
vertical coordinate in a unit square in computational
plane
Subscripts
i; j
cell faces
Superscript
n
time level

nonuniform grids. This motivates to use higher order compact

scheme for heat transfer problems, where the main objective is
to obtain the transient solutions for different aspect ratios and
also to examine the numerical performance of our proposed
scheme. In computational uid dynamics eld of study, highorder compact formulations are becoming more popular as compact formulations provide more accurate, efcient and stable
solutions in a compact stencil [24,27,28,31].
In this work, we have studied the problem on natural convection ow in a rectangular cavity (with aspect ratio 0.5, 1, 2, 4
and 6) lled with uid saturated porous medium by considering
the Brinkmann-extended Darcy model with our proposed scheme
which is an extension of the fourth order accurate compact scheme
[28] on nonuniform grids. In this extension, the temperature gradient source term in the vorticity equation is not explicitly known
and has been discretized as fourth order compact and smoothly
integrated to the solution procedure. Another attractive feature
of the computation is the use of hybrid BiCGStab [6,30] algorithm
for solving the resultant linear algebraic equations and this
improves the convergence behavior of the algorithm. As expected,
the results obtained using higher order compact scheme are accurate and even excellent agreement of our coarse grid results with
the results available in the literature.
The remainder of this paper is organized as follows. Section 2
gives formulation of the problem. Section 3 describes the discretization of the governing equations with related issues. Section 4
analyzes the solution procedure of algebraic systems. Section 5
deals with the results and discussions and Section 6, the
conclusions.
2. Problem formulation
2.1. The problem
The problem considered in this paper is the two-dimensional
transient natural convection of uid in a rectangular enclosure
(lled with a uid saturated porous medium) of height H and
width L having aspect ratios AR HL 0.5, 1.0, 2.0, 4.0 and 6.0.
The ow is driven by a uniform and nonuniform temperature prole applied along the bottom wall of the enclosure, as illustrated in

626

S.K Pandit, A. Chattopadhyay / International Journal of Heat and Mass Transfer 75 (2014) 624636

@u
@u
@u
@p
Pr
u;
u
v
 Prr2 u 
@t
@x
@y
@x
Da

@v
@v
@v
@p
Pr
u
v
 Prr2 v 
v RaPrT;
@y
Da
@t
@x
@y

@T
@T
@T
u
v
r2 T;
@t
@x
@y

Adiabatic Wall

Porous medium

where u and v are the velocity components in the x and y directions

respectively, p is the pressure, and T is the temperature. Here, all
distances are normalized by width L, all velocities are normalized

by aL (where a is the thermal diffusivity of the uid), time by aL2 ,

H
Tc

Tc

pressure by q aL2 (q is the uid density). The temperature is normalized as T TT

L
Th or
(Th- Tc)sin( x /L)+Tc

T 0

h T c

, where T  ; T 0 ; T h and T c are the dimensional, refer-

ence, hot bottom wall and cold vertical walls temperature

respectively. The dimensionless parameters appearing in the Eqs.
(2)(4) are the Prandtl number Pr am , Darcy number Da LK2 and

Fig. 1. Schematic view of the heated rectangular cavity.

Fig. 1. The top wall is well insulated and the vertical walls are
cooled to a constant temperature. A Boussinesq approximation of
the uid buoyancy is employed, whereby density differences in
the uid are neglected with the exception of the gravity contribution. Further, it is assumed that the temperature of the uid phase
is equal to the temperature of the solid phase everywhere in the
porous region, and local thermal equilibrium model is applicable
in the present investigation [2]. Several widely used models have
been introduced in the literature to study the ow problems in
porous media, such as Darcy, the Brinkman-extended Darcy, and
the Forchheimer-extended Darcy models. A recent achievement
in modelling ow in porous media is the so-called generalized
model or BrinkmanForchehimer-extended Darcy model, in which
all uid forces and the solid drag force are considered in the
momentum equation. Among these models, the Brinkman
extended Darcy model with the convective terms has also been
extensively used in modelling the ow and heat transfer in nite
porous enclosures. In the present study, the Brinkman-extended
Darcy model with the inclusion of convective and transient terms
has been adopted in the governing equations of the problem. In
addition, the momentum equation consists of the Brinkmann term,
which describes viscous effects due to the presence of solid body.
Although the viscous boundary layer in the porous medium is very
thin for most engineering applications, inclusion of this term is
essential for heat transfer calculations. The Forchheimer inertia
term in the momentum equations is neglected and a brief discussion of the exclusion of this term can be found in Basak et al.
[16]. Contextually, the Brinkman-extended Darcy model i.e, the
Forchheimer-extended Darcy model without the inertia term,
adopted in the present study, has been used in a large number of
investigations for natural convection in annular and rectangular
porous enclosures [16,19,23].

the Rayleigh number Ra gbL mTahT c , where m is the kinematic viscosity of the uid, b is the thermal expansion coefcient of the uid, g
is the gravitational acceleration and K is the permeability of the porous medium.
Now introducing dimensionless streamfunction wx; y; t and
the vorticity fx; y; t, dened by

@w
;
@y

v 

By employing the aforementioned approximations, the governing equations for unsteady two-dimensional natural convection
ow in the porous cavity using conservation of mass, momentum
and energy in non-dimensional primitive variable formulation
can be written as

@w
;
@x

@ v @u

:
@x @y

the NS equations (1)(4) can be written as

@2w @2w

f;
@x2 @y2

1 @f @ 2 f @ 2 f u @f v @f
1
@T



f Ra :
Pr @t @x2 @y2 Pr @x Pr @y Da
@x

@T @ 2 T @ 2 T
@T
@T


u
v
0:
@t @x2 @y2
@x
@y

If the reference temperature T 0 is taken as being equal to T c , the

dimensionless boundary conditions for the present problem are
specied as follows:
u v w 0; T 0 at left wall,
u v w 0; T 0 at right wall and
u v w 0; T 1 (for uniform heating) or sinpx (for nonuniform heating) at bottom wall and
u v w 0; @T
0 at top wall.
@y
The Nusselt number often describes the heat transfer characteristics across the cavity. In this problem we mainly pay attention to
the Nusselt number on the hot wall, which is computed by

Nu T y 0;j

and the average Nusselt number on that wall is computed by

Nu 

Z
0

2.2. Governing equations

@u @ v

0;
@x @y

T y 0;j dx

10

3. Discretization and related issues

3.1. Discretization of the governing equations
To discretize the governing equations (6)(8) on nonuniform
grids, we have rst considered a general 2D transient convectiondiffusion-reaction equation for a transport variable / in some
domain as

627

S.K Pandit, A. Chattopadhyay / International Journal of Heat and Mass Transfer 75 (2014) 624636

@/
@2/
@2/
@/
@/
ax; y; t 2 bx; y; t 2 cx; y; t
mx; y; t
@t
@x
@y
@x
@y
xx; y; t/ hx; y; t;
11

where l is constant and h is the source term.

Using transformation

x xn; g;

y yn; g:

12

b
b
b
b
b
b
@/
@2 /
@2 /
@2 /
@/
@/
b f;
a 2 g
b 2 c
d
p/
@t
@n@ g
@g
@n
@g
@n

13

where the coefcients l; a; b; c; d; g; p are the transformed coefcients and f is the transformed part of the source function in the
computational plane. The earlier HOC formulation in w  f forms
on nonuniform grids of the 2D transient incompressible viscous
ows governed by NS equations given in Pandit et al. [28], forms
the core of the discretization of our present study. In that formulation it was assumed that the forcing function f and its derivatives
are known analytically or their discrete approximations are known.
It may be mentioned here that the source term f in (13) is in the
form of a derivative for the transformed vorticity equation (7).
Now, assuming the transformed domain to be rectangular and
constructing on it a uniform rectangular mesh of steps h and k in
the n- and g-directions respectively, the fourth order HOC approximation of (13) at the (i; j)th node is given by [28]

(
l

bn
dt /
i;j

)
2
 k2 

h  2 bn

bn
2 bn
bn

d d / ci;j dn dt / i;j
d d / di;j dg dt / i;j
12 n t i;j
12 g t i;j

b ij
A0i;j d2n /

b i;j
G0i;j dn dg /

b i;j
B0i;j d2g /

b i;j
C 0i;j dn /

b i;j H0 /
b
D0i;j dg /
i;j i;j


n

o
1 h 2
2
b i;j h2 di;j k2 2dg ai;j d ai;j d2 dg /
b i;j
k ai;j h bi;j d2n d2g /
n
i;j
12
n

o
i
2
2
b i;j
k ci;j h 2dn bi;j c bi;j dn d2 /

i;j

4 4
F 0i;j ODt;h ;k ;

14
Details of the coefcients appearing in the above relations are available in [28].
It may be mentioned here that the temperature gradient @T
in
@x
the source term of the vorticity equation (7) is not explicitly
known. It needs to be discretized in a higher order
 compact form.
In the computational plane the source term Ra @T
of Eq. (7) is
@x

Ra
@T
@T
yg
 yn
J
@n
@g

15

where J xn yg  xg yn .
In the following, we proceed to obtain a compact fourth order
accurate approximation of this term applying the same mechanism
of using the original partial differential equation. Now,

"
#
2
@T
h @3T
4
ji;j dn T 
Oh
@n
6 @n3
i;j

18

i;j

h
e1 h
e2 k
2dn e1 c e1
dn a0 c0
dg a0
12
12a0
12b
2
2
2
k
e2 k
e1 h

0
0
0
B00
2dg e2 d e2
dn b
0 dg b d
12
12a0
12b
2
2
2
h
k
e1 h

d2n e1 c dn e1
d2g e1 d dg e1
dn c0 p0
C 00 e1
12
12
12a0
2
e2 k
0

0 dg c
12b
2
2
2
h
k
e2 k

0
0
d2n e2 c dn e2
d2g e2 d dg e2
D00 e2
0 dg d p
12
12
12b
2
e1 h
0
dn d

12a0
2
2
2
2
h
k
e1 h 0 e2 k 0

G00
2dn e2 c e2
2dg e1 d e1
d
0c
0
12
12
12a
12b
2
2
2
0
2
2
e2 h
a0 e2 k
e1 k
b e1 h
h
E00

F 00

; H00
e1 dn p0 ;
0 ;
0
12
12
12a
12a0
12b
A00

I00

k
0
0 e2 dg p ;
12b

h e1 l
;
12a0

K 00

k e2 l
0
12b

We now introduce a weighted average parameter l through the

b
approximation of the time derivative @@t/
such that
tl 1  lt n ltn1 where 0 6 l 6 1. Varying l yields different
schemes of different time accuracies. With these, Eq. (14) can be put
in the form
1
1
X
X

b
wik1 ;jk2 /
ik1 ;jk2
n1

k1 1k2 1

1
1
X
X

b
w0ik1 ;jk2 /
ik1 ;jk2

 h
1
0
0
dn a0 c0 d2n T  d dn dg T  dn b d2g T
a0
!#
i
0
0
4
Oh
b dn d2g T  dn c0 dn T  dn d dg T  p0 dn T  Tdn p0  ldn T t
2

@T
h
j dn T 
6
@n i;j

i;j

17

k1 1k2 1

24Dt

lF i;jn1 1  lF n
i;j ;

19

where

wik1 ;jk2

pik1 ;jk2 qik1 ;jk2

2 2

h k

w0ik1 ;jk2 l  1

Dt
2 2

h k

pik1 ;jk2 qik1 ;jk2

with
0

16

Similarly for @@Tg ji;j . Differentiating the transformed equation of (8)

w.r.t n in computational plane and substituting in Eq. (16), we get

"

h
F 0i;j A00 d2n T B00 d2g T C 00 dn T D00 dg T G00 dn dg T E00 d2n dg T
i
F 00 dn d2g T H00 dn pT I00 dg pT K 00 dn T t L00 dg T t
Such that

the Eq. (11) can be written in the transformed plane as

where a0 ; b ; c0 ; d ; g 0 ; f 0 ; p0 are the coefcients in the transformed

y Ra
equation of the temperature equation (8). Assuming e1 gJ and
yn Ra
e2  J , the source term (15) in transformed vorticity equation
will be

pi1;j1 6hkGi;j 2X i;j  kY i;j  hZ i;j ;

pi;j1

2
24h B0i;j

pi1;j1

 12h

0
6hkGi;j

pi1;j 24k

A0i;j

0
kDi;j

 12hk

 4X i;j 2kY i;j ;

2X i;j  kY i;j hZ i;j ;

2

C 0i;j

qi1;j1 0;


qi;j1 l2  kdi;j ;

qi1;j1 0;

 4X i;j 2hZ i;j ;

2 2

qi1;j l2  hci;j ;

pi;j 48k A0i;j  48h B0i;j 24h k H0i;j 8X i;j ;

pi1;j 24k
pi1;j1

A0i;j

12hk

0
6hkGi;j

C 0i;j

 4X i;j  2hZ i;j ;

2X i;j kY i;j  hZ i;j ;

qi;j 16l;


qi1;j l2 hci;j ;

qi1;j1 0;

2
2
0
pi;j1 24h B0i;j 12h kDi;j  4X i;j  2kY i;j ; qi;j1
0
pi1;j1 6hkGi;j 2X i;j kY i;j hZ i;j ; qi1;j1 0;

l2 kdij

628

S.K Pandit, A. Chattopadhyay / International Journal of Heat and Mass Transfer 75 (2014) 624636


in which the coefcients A0 ; B0 ; C 0 ; D0 ; G0 ; H0 ; F 0 ; X; Y; Z; c ; d appearing

in the above relations are available in [28].
For vorticity equation, the expression for the source term within
n1
n
the parenthesis (in the r.h.s) in (19) i.e lF i;j
1  lF i;j will be
1
X

1
X

k1 1

k2 1

1
X

1
X

k1 1

k2 1

n1

sik1 ;jk2 T ik1 ;jk2

s0ik1 ;jk2 T ik1 ;jk2 ;

20

where
sik1 ;jk2 4hl2 k2 r ik1 ;jk2 r 0ik1 ;jk2
1l
0
s0ik1 ;jk2 4h
with such that
2 2 r ik1 ;jk2  r ik ;jk
1
2
k

00

00

00

00

00

00

00

00

r i1;j1 hkG  2kE 2hF ri1;j 4hk C 4k A 4hF

2

2 2

00

00

00

00

00

00

00

00

00

00

00

r i1;j 2hk C 4k A  4hF r i1;j1 hkG 2kE  2hF

2

r i;j1 4h B00 2h kD  4kE r i1;j1 hkG 2kE 2hF

r 0i;j 0; r 0i1;j

0;j

Using the above results and v 0;j =0 (as the vertical velocity on
the left wall is zero) with Dx xn dn xg dg hxn (*xg 0; dn h
in the computational plane), (21) can be written in the computational plane for the third order accurate discretization of the vorticity boundary as follows:

1
1
1 2
3
d w  hxn f0;j  h xn dn f0;j Oh ;
xn n 0;j 2
6

22

employed [1], which is given by the following expression

L00
K 00
; r 0i1;j1 0; r 0i1;j 
;
2kDt
2hDt

00


@ 3 v 


@x@y2 
0;j
0;j






2
@f
@ f
f
@T 
@ 3 v 
  2    Ra 

@t 0;j @y 
Da 0;j
@x 0;j @x@y2 

Vorticity at the left wall (f0;j ) can be easily obtained from the Eq.
(22). The vorticity at the other boundaries can be approximated in
a similar way. It is noted here that one can nd the fourth order
accurate discretization of the vorticity from Eq. (21).
In this simulation, a fourth order one sided approximation to
 
boundary condition has been
the Neumann temperature @T
@y

and

r 0i1;j1 0; r 0i;j1 


@ 2 f
 2
@x 

0;j

0

00

r i;j 8k A00  8h B00 4h k H 00

2

0;j

00

r i1;j1 hkG  2kE  2hF r i;j1 4h B00  2h kD 4kE

00


@ 4 w

@x4 

"

  #
@T
T i;j 0:04 48T i;j1  36T i;j2 16T i;j3  3T i;j4  12h
@y i;j

00

K
L
; r0
0; r0i;j1 0; r 0i1;j1
;
2hDt i1;j1
2kDt

r 0i1;j1 0;

Oh :
4

Thus Eq. (19) together with (20) become the ODts ; h ; k HOC
nite difference approximation for (7) on the transformed plane.
It may be mentioned here that Eq. (19) with F 0 represents the
discretization of Eqs. (8) and (19) with steady-state form demonstrates the discretization of (6) in the computational plane. It should
be noted that for l 0, the computational stencil requires nine
points at the nth and ve points at (n 1)th time level resulting
in what may be termed as a (9, 5) scheme. Similarly l 0:5 and
l 1 yields a (9, 9) and a (5, 9) scheme respectively. The temporal
order of accuracy s is two for the (9, 9) scheme and one for the other
two. Throughout our computations, we have used the (9, 9) scheme.

3.3. Stability analysis

In this subsection, we present the von Neumann linear stability
analysis for the nite difference scheme (13) by assuming the coefcients l; a; b; c; d; f to be constants and p 0 whereas the coefcient of mixed derivative term (=g) in Eq. (13) to be zero due to
p
b n Bn eiIhn ejIhg , where I 1; Bn is the
use of orthogonal grid. If /
ij


amplitude at time level n, and hn 2Kp1h and hg 2Kp2k are phase
angles with wavelengths K1 and K2 respectively. For stability,
n1

v BBn is the amplication factor which has to satisfy the relation

3.2. HOC wall boundary condition

jvj2  1 6 0:

23

We now proceed to develop HOC wall boundary conditions for

vorticity f in the following manner:
In the physical plane, the velocity on the left wall, where the
index for x-direction is 0 and j is the y-direction index varying from
0 to ymax , on using forward difference, we have

Now v can be obtained by substituting the expression for vnij and

vn1
in (19). The stability criteria of our proposed schemes becomes
ij

v 0;j 

It is seen that the proposed scheme is conditionally stable for

0 6 l < 0:5 while the scheme is unconditionally stable for
0:5 6 l 6 1 as the denominator is always positive because the coefcients a < 0 and b < 0.


@w
;
@x 0;j

dx w0;j




Dx @ 2 w
Dx2 @ 3 w
Dx3 @ 4 w




2 @x2 
6 @x3 
24 @x4 
0;j

0;j

ODx ;

21

where Dx is the distance between the left wall and the rst point
closest to it in the physical plane. Using the fact that
2

@2 w
@y2

0,

@@yv2 0 on a vertical wall, from the Poisson equation (6), we

have


@ 2 w

@x2 

f0;j ;

0;j


@f
@x0;j

24

4. Solution of algebraic systems

In this section, the solution of algebraic systems associated with
the nite difference approximations has been discussed. Because of
the presence of large gradients near the walls of the stated problem, we generate a centro-symmetric grid with clustering near
the walls using the stretching functions

k
sin2pn;
2p
k
sin2pg;
yg
2p

xn

0;j


@ 3 w

@x3 

2l

:
2
2
Dt ca 8a
 db 8b

h2
k2

0;j

@3 w
@x@y2

1  2l 6



1 
Dx @ 2 f
 dn f

xn
2 @x2 
0;j

0;j

where k 0:6 is the stretching parameter.

25
26

629

S.K Pandit, A. Chattopadhyay / International Journal of Heat and Mass Transfer 75 (2014) 624636

The coefcient matrix A of Eq. (19) is an asymmetric sparse

matrix. For a grid of size m  n; A has dimension mn. The next step
now is to solve Eq. (19) with iterative methods. As the coefcient
matrix A is not generally diagonally dominant, conventional iterative methods such as GaussSeidel cannot be used. On uniform
grids, some of the associated matrices are symmetric and positive
denite, which allows algorithms like conjugate-gradient (CG) [6]
to be used. As nonuniform grid invariably leads to non-symmetric
matrices, in order to solve these systems the biconjugate gradient
stabilized method (BiCGStab)[6,7,30] is used here without
preconditioning.
The system of algebraic equations resulting from discretization
of temperature equation, vorticity equation and stream-function
equation are solved in sequence using a decoupled algorithm in
an outer-inner iteration procedure. In a typical outer temporal
cycle, we solve the transformation of (8) using (19) with
l 1; a 1; b 1; c u, m v ; x 0; h 0 in (11). Then we
solve the transformation of (7) using (19) and (20) with
v , x 1 , h Ra @T in (11), and
l Pr1 ; a 1, b 1; c Pru , m Pr
Da
@x
the transformation of (6) using the steady-state form of (19) with
l 0; a 1, b 1; c 0; m 0; x 0; h f in (11). For each
system of equations, BiCGStab is used, which constitutes
the inner iterations. Once (6) is solved, u and v in (5) can be

approximated compactly up to fourth order accuracy using the

following HOC form

"






1

dn f  dn a c d2n w  dn g d dn dg w
a








g d2n dg w  dn b d2g w  b dn d2g w  dn cdn w  dn d dg w
(
)
2



xn
k 1

 pdn w  wdn p
dg w 
dg f  dg b d
J
6 b

 

  



 d2g w  dg g c dn dg w  g dn d2g w  dg a d2n w 
2

h
6

Nonuniform, Da=0.00001

Uniform, Da=0.00001
2
5
10

0.8

-2

0
0

0.2

0.4

0.6

0.8

0.1979

0.2003

0.2014

Ra 105

3.1306

3.2043

3.2254

Ra 106

15.0042

14.7798

14.7251

0.8

0.8

0.5

0.5

0.6

7  104

3  105

Da 104

5  105

5  105

0.6

0.8

0.2

0.4

0.6

0.8

0.
4

0
0.1 .2

0.8

0
0

0.2

0.4

0.6

75

0.125

0.2

0.3

25
0.6

y
0.4

-29.0106
-25.8062
-20.5551
-14.6574
2.57314
9.61657
20.2843
33.6241

0.6
0.75

0.2

0
0

(c)

0.875

0.2

0.4

0.8

9.616
20.28 57
33.6241 43

0.4

0.2

0.6

33.6241
20.2843
9.61657

33.6241
20.2843
9.61657

0.6

0.4

0
0

(b)

0.8

-1.08565

0.6

0.4

(b)
1

-0.0152131
-0.161859
-0.329179
-0.543821
-0.740909
-0.931196

0.8

0.2

0.2

Fig. 3. (a) Streamline and (b) temperature contours for Pr 0:71, and Ra 108 in
both the cases of uniform and nonuniform bottom wall heating.

(a)
1

0.4

0.3

y
Da 103

0
0

0.4

0.5

2
0. 1
0.

Ra (Nonuniform) Present and

[16]

0.6

Ra (Uniform) Present and

[16]

0.2 .1
0

0.2

0.6

0.3

0.4 0.3

0.4

Variables

Nonuniform, Da=0.00001

Table 2
Maximum value of Rayleigh number for which conduction dominant heat transfer
occurs for AR 1:0.

0.1

Ra 104

Uniform, Da=0.00001

0.8

0.
3

0.0185

0.6

0.
2

81  81

0.0184

0.4

(a)

41  41

0.0183

0.2

4
0.

21  21

0
0

Da 103

Ra 103

-15

17

0.2

0.5

Variables

-5
-10

15

0.4

0.2

Table 1
Thermal-driven cavity ow problem: grid independence study of the steady-state
data of wmin for uniform bottom wall heating with Da 103 and AR 1:0.

-2

10

0.6

-18
-1 9

19

0.8

-1 5

18

0.4

2
5

-5

-10

15

0.6

dn w 

xg
J

-1

ui;j 

0.2

0.8

0
0

0.2

0.4

0.6

0.8

Fig. 2. Benchmark problem of natural convection [2527]: (a) stream-function contours (wmin 1:174), (b) vorticity contours (fmax 51:25), (c) isotherms distribution
across the cavity for Ra 103 on a grid 61  61.

S.K Pandit, A. Chattopadhyay / International Journal of Heat and Mass Transfer 75 (2014) 624636

-3

2
0.05
0.1

0.4

0.1

7
00
0. 0.01

0.0

5
0.
1

0.2

1.5

0.02

0.
1

Uniform, Da=0.001

0.2
0.3
0.4
0.5 0.6
0.7 0.
8
0.9
0
0 0.2 0.4 0.6 0.8 1

0.5

0.5

0.6

0.5

(i)

0.1

0.4

-3

0
0 0.2 0.4 0.6 0.8 1

0.3

0
0 0.2 0.4 0.6 0.8 1

1.5

0.3

0.3

-2.5

0
0 0.2 0.4 0.6 0.8 1

-2

2
0.

0.5

5
-1 .

-0.1
15
-0 .
-0.2

-1

-0.05

-0.5

2.5

05
0.

0.7 0.6 0.5

Uniform, Da=0.0001

0.5

0.5

0.2

1.5

2
-0.0

Uniform, Da=0.001

1.5

1.5

0
0 0.2 0.4 0.6 0.8 1

-0.008

02
0.

0.1

0.3

1.5

0.9
0
0 0.2 0.4 0.6 0.8 1

-0.003

Nonuniform, Da=0.001
2

0.5

Uniform, Da=0.0001

00
0.

0.2

0.2
0. 0
8 .7 0.6

0.003

Uniform, Da=0.001
2

0.1

0.9
0
0 0.2 0.4 0.6 0.8 1

0.2

0.5

0
0 0.2 0.4 0.6 0.8 1

0.2 1
0.

0.5

15
0.5

0.3

0.5

0.05
0.1

0.1
0.2

0.4

-16

0
0 0.2 0.4 0.6 0.8 1

Nonuniform, Da=0.0001

1.5

16

0.5

-12

12

-1 5

-7
-9

-4

0.2

-2.5

0
0 0.2 0.4 0.6 0.8 1

0.3

2.5

-1

0.
7

0.5

0.1

12

1.5

.5

0.5

1.5

1.5

4
7

0.3

Uniform, Da=0.0001

-1
-1
.5

-2

1.5

7
9

4 1

1.5

0
0 0.2 0.4 0.6 0.8 1

-0.5

0.5

1.5

0.5

Nonuniform, Da=0.001

-1
-4
-7
-9

0.1

-3

0.4

3.5

-2.
5

-3

Uniform, Da=0.001

Nonuniform, Da=0.0001

-2

0.5
1

2.5

1.5

1.5

-0.5
-1
.5
-1

-12

Uniform, Da=0.0001

630

0.8
0.9
0
0 0.2 0.4 0.6 0.8 1

(ii)

Fig. 4. For AR = 2.0, steady-state solutions of (a) streamline contours (b) temperature contours in a rectangular cavity with Ra 106 (c) (i) streamline contours (ii)
temperature contours with Ra 105 .

)#
 2
  
 



 
a dn dg w  dg c dn w  dg d dg w  p dg w  w dg p
i;j
4

Oh ; k ;
"
(
2


yg
h 1

v i;j 
dn w 
dn f  dn a c d2n w 
J
6 a








dn g d dn dg w  g d2n dg w  dn b d2g w  b dn d2g w 

)


dn cdn w  dn d dg w  pdn w  wdn p


(
2
 




yn
k 1h
dg w 
dg f  dg b d d2g w  dg g c dn dg w 
J
6 b

  


  
g dn d2g w  dg a d2n w  a d2n dg w  dg c dn w
 



 
4
4
27
 dg d dg w  p dg w  w dg p gi;j Oh ; k ;

631

S.K Pandit, A. Chattopadhyay / International Journal of Heat and Mass Transfer 75 (2014) 624636

t=0.02

t=0.03
0.4

0.2
0
0

0.4

0.4

0.2

0.2

0.4

0.6

0.8

0
0

0.2

0.2

0.6

0.8

0.4

0.6

0.8

0
0

0.4

0.6

0.8

0.6

0.8

0
0

0.4

0.6

0.8

0.6

0
0

0.8

0
0

0.6

0.8

0.2

0.4

0.6

0.8

0.6

0.8

t=1.0
0.4
y

y
0.2

0.2

0.4

0.4

0.2

0.4

0.2

0.2

t=0.08

0.2

0.4

t=0.05
0.4

0.8

t=0.03

0
0

0.2

0.4

0.6

y
0.2

0.4

0.2

0.2

t=0.02

0.4

0.4

0.4

t=0.005

0.2

0.2

t=2.0

0.2

0
0

0.2

0.2

0
0

0.4

0.4

0
0

0.4

t=0.12

t=0.11

0
0

t=0.1

0.2

0.2

0.4

0.6

0.8

0
0

0.2

0.4

Fig. 5. Evolution of streamlines at different time stations for the natural convection ow for Pr 0:71; Ra 106 ; Da 103 and AR = 0.5 in both the cases of (a) uniform and
(b) nonuniform bottom wall heating.

where the expressions for a; b; c; d; f ; g; p can be calculated from

(13). This completes one outer iteration cycle. We utilize a relaxation parameter c for the inner iteration cycles for all T; x and w.
For larger values of Rayleigh number, we need smaller values of c.
All of our computations were carried out on a Pentium 4 based PC
with 2 GB RAM.

5. Results and discussions of the transient natural convection in

a deep cavity with porous medium
In this section we now present the time-marching steady-state
and the transient solutions produced by our proposed compact
scheme for this problem in Tables 1 and 2 and Figs. 29. The working uid is chosen with Prandlt number Pr = 0.71. The computations were performed for both the cases of uniform and
nonuniform bottom wall heating with AR 0:5; 1:0, 2.0, 4.0 and
6.0 and for various Rayleigh numbers. It is mentioned here that
the nonuniform heating bottom wall removes the singularities at
the edges of the bottom wall in contrast to uniform heating bottom
wall.
In order to obtain grid independent solution, a grid renement
study has been performed in Table 1 for the steady-state data of
wmin in the case of uniform bottom wall heating. It is observed that
grid independence is achieved with a grid of size 41  41. This grid
resolution is therefore used for all subsequent computations for

AR 1. For taller cavities with AR > 1 and shorter cavity with

AR < 1, a proportionately large number of grids and small number
of grids respectively in the vertical direction are used while keeping the number of grids in the horizontal direction xed at 41.
To asses the accuracy of our proposed numerical scheme, we
have also studied the benchmark problem for the differentially
heated square enclosure with a heated left side wall while bottom
and top walls are adiabatic similar to the case reported by Kalita
et al. [27], De Vahl Davis and Jones [25,26]. The results in Fig. 2
are in excellent agreement with their work.
In Table 2, we have compared our results with those of Basak
et al.[16] of maximum value of Rayleigh number for which conduction dominant heat transfer occurs in the case of square enclosure.
Our computed results exactly match with the numerical values
given in [16].
The power of the scheme has been better realized in Fig. 3,
which reveals that even at Ra 108 , our proposed scheme is capable to produce results on coarser grids (41  41) in contrast to the
lower order scheme (Second order central difference scheme). In
this context, it can be concluded here that the higher order compact scheme is capable to make the full system run more stably
and perform more efciently.
The steady-state solutions of streamlines and isotherms at
Ra 106 for both the uniform and nonuniform bottom wall heating are shown in Fig. 4(a) and (b) with aspect ratio AR = 2.0. Visual
examination of the streamlines does not reveal any signicant

S.K Pandit, A. Chattopadhyay / International Journal of Heat and Mass Transfer 75 (2014) 624636

t=0.01

0.8

0
0

0.2

0.4

0.4

0.2
0
0

0.6
0.4
0.2
0
0

0
0

0.2

0.4

0.6

0.001

0.8

0
0

-0.005

0.02

-0.02

0.05
0.1

-0.05

0.2

-0.2

0.4
0.6

-0.4
-0.6

0.2

0.8
0.6

-0.1

0.4

0.6

0.8

0.4
0.2
0
0

0.8

-1 6

0.6

16

0.6

0.8

-5

-10
-1 2
-1 4

0.6

0.2

0.4

0.6

0.8

0
0

0.005

-0.005

0.03

-0.03

0.2

-1 2
4
-1

0.4

0.1

-0.1

0.2

-0.2
-0.4

0.4

-1

-2

0.2

0.4

0.6

0.8

0.04
0.2

0.8

0.7

0.6

0
0

0
0

0.2

-0.04
-0.2
-0.7

-3

-8
-12

0.4

10

-10
-12

0.6

10
12

0.6

0.6

0.8

1
0.8

t=1.0

-5

0.4

0.8

0.2

-2

0.2

0.6

8
12

2
5

12

0.4

0.2

0.8

-8

t=0.02

0.2

0.6

-0.7
-3

14

0.2

0.8

0.2

0.4

0.4

0.6

12

0.4

0.4

0.2

0.8

5
10
12
14

0.4

0
0

0.7
3
8
8

t=0.1
1

-0.3
-3
-8

0.3
3
8

-8

t=0.03
1

0.4

t=0.01
1

-0.001

0.005

0.2

t=1.0

0.2

0.8

0.8

-1 -2

t=0.005
1

0.6

0.4

0.6
-10
-1 2

0.6
10
12

0.8

-5

0.005

-0.5

-2

0.5

0.8

0.4

t=0.1
-0
.5

-0.005

0.5

-8.99328

0.2

t=0.08
1

0.6

0.2

0
0.6

-6.27644

0.4

-2

-5

0.2

-3

-0.4

-1

-2

0.4

0.4

-0.2

-3

-0.3

-0.1

0.2

0.3

0.8

8 3
1

0.2

0.6

0.3

0.9

-0.9

-8

0.4

0.8

-0.015
-0.03

0.015
0.03
0.055
0.1

0.6

-0.3

-0.005

0.005

0.8

0
0

t=0.05

t=0.03
1

632

0.8

0
0

0.2

-2

-5

-1 0

-1 2

0.4

0.6

0.8

Fig. 6. Evolution of streamlines at different time stations for the natural convection ow for Ra 106 ; Da 103 and AR = 1.0 in both the cases of (a) uniform and (b)
nonuniform bottom wall heating.

difference among the different cases. As expected, two counter

rotating vortices are formed. In each case the ow descends downwards along the cold side walls. The ow then rises along the vertical symmetry axis gets blocked at the adiabatic top wall, which
turns the ow horizontally towards the cold side walls. Thus a pair
of counter rotating cells is formed in the ow domain. The location
of the centre of the vortices rises slightly upward for both the cases

of uniform and nonuniform heating bottom wall for higher Da (see

Fig. 4). However, noticeable difference is observed in the isotherm
plots for different Darcy numbers. So, it is the parameter of focus in
the analysis for all cases. It is seen that the solution is symmetric
about the vertical centerline due to the symmetry of the problem
geometry and boundary conditions. It is also seen that the ow is
very weak for the lower values of Da. The isotherm lines change

633

S.K Pandit, A. Chattopadhyay / International Journal of Heat and Mass Transfer 75 (2014) 624636

t=0.02

t=0.04

0.6

0.6

0.3

0.8

0
0

0.2

0.4

0.6

0.6

0
0

0.2

0.2

0.4

t=0.08
1

0.8

0.8

0.2 0.3

0.6

0.8

0
0

0.4
0.
3

0.3

0.2

t=0.005

0.2

0.4

0.6

0.8

0
0

0.8

0.8

0.6

0.6

0.6

y
0.4

0.2

0.4

0.05

0.2

0.6

0.8

0
0

0.2

0.4

0.6

0.8

0.2

0.6

0
0

0.8

0.8

0.8

0.2

0.4

0.2
0.1

0.4

0.6

0.2

0.8

0
0

0.6

0.3
0.2

0.2
0.1

0.2

0.4

0.6

0.6

0.8

0.8

0.4

0.4

0.3

0.4

0.1

0.3

y
0
0

0.4

0.2

0
0

0.3

0.3

0.2

0.2

0.5

0.2

0.2

4
0.

0.6

0.2

0.8

t=1.0
1

0.1

0.6

0.1 0.2
0.5

0.4

0.4

t=0.1

t=0.03

0.3

0.2

0.45
0.4
0.35
0.3
5
0.2

0.4

0.6

0.8

0.05

0.8

t=0.02

0
0

0.4
0.35
0.3
0.25

t=0.01

0.4

0.6

0.5

0.45

0.2

0.5

0.6

0.2

0.4

0.8

0.2

0.4
0.5

y
x

0.2

t=1.0

0.2

0.4

0
0

0.55

0.3

0.2

0.8

0.4

0.5

2
0.

0
0

0.6

0.4

0.2

0.6

0.4

0.4
0.2

t=0.1

0.6

0.4

0.4

0.2

0.8

0.4

0.4

0.3

0.5

0.1
0.2
0.4
0.6

0.2

0.3

0.4

0.4

0.1 0.2

0.8

0.2

0.5

0.8

0.8

1
0.1
0
0.3 .2

0.
1

0.1

t=0.05

0.4

0.6

0.2

0.8

Fig. 7. Evolution of temperature contours at different time stations for the natural convection ow for Pr 0:71; Ra 106 , Da 103 and AR = 1.0 in both the cases of (a)
uniform and (b) nonuniform bottom wall heating.

value smoothly from the hot bottom wall to the cold vertical wall
and the heat transfer is dominated by conduction for low Darcy
numbers. As Da increases to 103 , the strength of the ow is
increased. The stronger circulation causes the temperature contours to be concentrated symmetrically near the side walls and
the edges of bottom wall which may result in greater heat transfer
rate due to convection. The convection region adjacent to the heat

source becomes thinner and denser producing higher temperature

gradient with the increase of Darcy numbers. It is noted here that
the temperature at the bottom wall is non-uniform and the maxima occurs at the center in contrast to the uniform heating bottom
wall. There is a boundary layer formed near the bottom and both
the side walls. At low values of Da, the conduction plays the main
role in the heat transport. For higher Da, the buoyancy begins to

634

S.K Pandit, A. Chattopadhyay / International Journal of Heat and Mass Transfer 75 (2014) 624636

t=0.01

t=0.1

t=0.2

1.5

1.5

1.5

0.5

0
0 0.2 0.4 0.6 0.8 1

0.5

t=0.005

0.5

0.5

0
0 0.2 0.4 0.6 0.8 1

1.5

t=0.05

0
0 0.2 0.4 0.6 0.8 1

t=0.1

0
0 0.2 0.4 0.6 0.8 1

1.5

1.5

0.5

0
0 0.2 0.4 0.6 0.8 1

0.5

0
0 0.2 0.4 0.6 0.8 1

t=1.0

1.5

1.5

0.5

t=1.0

0.5

0
0 0.2 0.4 0.6 0.8 1

0
0 0.2 0.4 0.6 0.8 1

Fig. 8. Evolution of streamlines at different time stations for the natural convection ow for Pr 0:71; Ra 106 ; Da 103 and AR = 2.0 in both the cases of (a) uniform and
(b) nonuniform bottom wall heating.

inuence the heat transport and the isotherms with high values
tend to concentrate near the heat source surface. It may be mentioned here that due to nonuniform bottom wall heating, the heating rate near the wall is in general lower which diminishes
buoyancy effects resulting in less thermal gradient throughout
the domain. In this context, it is noted here that we have also computed the results for Ra 105 with different Das shown in Fig. 4(c)
for aspect ratio AR = 2.0 in comparison with [13], though Haajizadeh et al. studied natural convection in a vertical porous enclosure with uniform internal heat generation and side wall cooling.
In their work, it is found that the two cells are mirror images to
each other and the centre of the vortices exist almost on the horizontal centerline and very strong boundary layer present near the
cold walls. Although in our problem, it is seen formation of two
cells which are mirror images of each other but the centres of
the vortices exist near the bottom wall. In addition, isotherms start
to atten (parabolic temperature distribution) at the bottom of the
cavity and shrink towards the top of the enclosure, while at the top
they are still almost parallel to the vertical walls which are in contrast to the isotherm patterns in [13].
In Figs. 58, we show the time-wise evolution of the streamlines and temperature contours in both the cases of uniform and
nonuniform bottom wall heating for aspect ratios AR 0:5; 1:0
and 2.0 keeping Da 103 and Ra 106 xed. The transient gures
shown for uniform and nonuniform bottom wall heating are more
exciting. These gures show that our scheme excellently captures
the formation of the different cells as time progresses. It is seen
that towards steady-state the streamline contours for both the
cases of uniform and nonuniform bottom wall heating look alike
although their evolution in different time stations are different.

In the early of the transient steps there are several cells cover
the whole domain. For AR 6 1, the evolution of streamlines is
always symmetric about the vertical centerline while for AR P 1
the streamlines are not symmetric at different time stations except
towards the steady-state solution. In Fig. 7, it is seen that at the
beginning of the uniform bottom wall heating, the liquid warms
up in the lower corners of the cavity and thermal gradients at
the bottom of the enclosure giving a growing up in bicellular thermal patterns. Then temperature increases in times and isotherms
show at shaped mushroom structure. Towards steady-state, in
that structures the isotherms 0.4 and 0.5 break down into some
symmetric contour lines about the vertical centerline and pushed
towards the side walls. On the contrary, it is seen that at the beginning of the nonuniform bottom wall heating, the liquid warms up
in the middle of the bottom wall of the cavity and thermal gradients at the bottom of the enclosure giving a growing up in unicellular thermal patterns. Then temperature increases in times and
isotherms show round shaped mushroom structure while towards
steady-state almost same phenomenon for uniform bottom wall
heating is observed. Almost similar behavior of evolution of temperature contours is also observed for different Darcy numbers
with these aspect ratios for which the plots are not prescribed here
for brevity. We note that a fundamental variation in the structure
of the ow and thermal patterns occurs as the height of the cavity
increasing.
Fig. 9(a), (b) and (c)) show that for uniform heating of the
bottom wall the heat transfer rate is very high at the edges of
the bottom wall. This is due to presence of discontinuity in the
temperature boundary condition at the edges while heat transfer
rate decreases towards the centre of the bottom wall. On the

S.K Pandit, A. Chattopadhyay / International Journal of Heat and Mass Transfer 75 (2014) 624636

Local Nusselt Number,bottom Wall

10

Uniform
8

AR=0.5
AR=1.0
AR=2.0
AR=4.0
AR=6.0

6. Conclusions

AR=0.5
AR=1.0
AR=2.0
AR=4.0
AR=6.0

Nonuniform

In this paper, we have extended the fourth order compact nite

difference formulation of [28] on nonuniform grids to solve transient natural-convection ow in a rectangular cavity lled with a
uid saturated porous medium. The momentum transfer in the
porous region is modelled by considering Brinkmann-extended
Darcy model. The robustness of the scheme is illustrated by its
applicability to the problem of varying physical complexities, represented among others, by Rayleigh numbers ranging from 103 to
108 and Darcy numbers ranging from 103 and 105 . The outcomes
of the present study are listed as follows:

0
0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Distance, X

Local Nusselt Number, Bottom Wall

20

Uniform
15

AR=0.5
AR=1.0
AR=2.0
AR=4.0
AR=6.0

AR=0.5
AR=1.0
AR=4.0
AR=6.0

Nonuniform AR=2.0

10

0
0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

635

0.9

(1) The transient ow phenomenon is more exciting than the

steady-state solutions.
(2) The overall heat transfer process improves as the aspect
ratio increases until the cavity becomes square. For taller
cavities (AR > 1), the improvement in heat transfer process
is not signicant.
(3) The temperature at the bottom wall is non-uniform in contrast to the uniform heating bottom wall. The nonuniform
heating exhibits greater heat transfer rates at the center of
the bottom wall than that with uniform heating case for different Rayleigh numbers.
(4) The streamline contours for both the cases of uniform and
nonuniform bottom wall heating look alike towards
steady-state although their evolution in different time stations are different.
(5) It is checked that the temperature decreases from the bottom to the top along the centerline of the cavity for a particular value of Da.

Distance, X

Local Nusselt Number, Bottom Wall

20

Uniform
15

AR=0.5
AR=1.0
AR=2.0
AR=4.0
AR=6.0

Because of compactness and higher order accuracy of our proposed scheme, this treatment may be taken as the model for similar computations.

AR=0.5
AR=1.0
AR=2.0
AR=4.0
AR=6.0

Nonuniform

Conict of Interest
None declared.

10

References
5

0
0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Distance, X

Fig. 9. Variation of Local Nusselt number with distance at the bottom wall for
uniform and non-uniform bottom wall heating with Pr 0:71, AR = 0.5, 1.0, 2.0, 4.0
and 6.0 and for (a) Da 105 (b) Da 104 and (c) Da 103 .

contrary, for nonuniform heating bottom wall, heat transfer rate is

zero at the edges and increases towards the centre with its maximum value at the center (x = 0.5) for Da 105 , at x = 0.25 and
x = 0.75 for Da 104 and at x = 0.35 and x = 0.65 for Da 103 .
It is also observed that for nonuniform case the shape of the heat
transfer curves is a sinusoidal type for higher Das while for uniform
case the curve is almost parabolic in shape. It is mentioned here
that the overall heat transfer process improves as the aspect ratio
increases until the cavity becomes square. For taller cavities
(AR > 1), the improvement in heat transfer process is not
signicant.

[1] J.C. Tannehill, D.A. Anderson, R.H. Pletcher, Computational Fluid Mechanics
and Heat Transfer, Hemisphere Publishing Corporation, New York, 1984.
[2] D.A. Nield, A. Bejan, Convection in Porous Media, Springer Verlag, New York,
1999.
[3] K. Vafai, Handbook of Porous Media, Marcel Dekker, New York, 2000.
[4] D.B. Ingham, I. Pop (Eds.), Transport Phenomena in Porous Media, Pergamon,
Oxford, 1998 (vol. II, 2002; vol. III, 2005).
[5] A. Bejan, A.D. Kraus (Eds.), Heat Transfer Handbook, Wiley, NewYork, 2003.
[6] C.T. Kelley, Iterative Methods for Linear and Nonlinear Equations, SIAM
Publications, Philadelphia, 1995.
[7] Y. Saad, Iterative Methods for Sparse Linear Systems, PWS Publishing
Company, 1996.
[8] A. Amiri, K. Vafai, Transient analysis of incompressible ow through a packed
bed, Int. J. Heat Mass Transfer 41 (1998) 42594279.
[9] A.C. Baytas, I. Pop, Free convection in a square porous cavity using a nonequilibrium model, Int. J. Therm. Sci. 41 (2002) 861870.
[10] K.L. Walker, G.M. Homsy, Convection in porous cavity, J. Fluid Mech. 87 (1978)
449474.
[11] T.W. Tong, E. Subramanian, Boundary layer analysis for natural convection in
porous enclosure: use of the Brinkman-extended Darcy model, Int. J. Heat
Mass Transfer 28 (1985) 563571.
[12] D. Poulikakos, A. Bejan, The departure from Darcy ow in natural convection in
a vertical porous layer, Phys. Fluids 28 (1985) 34773484.
[13] M. Haajizadeh, A.F. Ozguc, C.L. Tien, Natural convection in a vertical porous
enclosure with internal heat generation, Int. J. Heat Mass Transfer 27 (10)
(1984) 18931902.
[14] R. Horne, M. Osullivan, Oscillatory convection in a porous medium heated
from below, J. Fluid Mech. 66 (1974) 339352.
[15] J.M. Straus, Large amplitude convection in porous media, J. Fluid Mech. 64
(1974) 5163.

636

S.K Pandit, A. Chattopadhyay / International Journal of Heat and Mass Transfer 75 (2014) 624636

[16] T. Basak, S. Roy, T. paul, I. Pop, Natural convection in square cavity lled with a
porous media: effects of various thermal boundary conditions, Int. J. Heat Mass
Transfer 49 (2006) 14301441.
[17] V. Prasad, F.A. Kulacki, Convective heat transfer in a rectangular porous cavity
effect of aspect ratio on ow structure and heat transfer, J. Heat Transfer 106
(1984) 158165.
[18] C.J. Ho, J.Y. Chang, A study of natural convection and heat transfer in a vertical
enclosure with two dimensional discrete haeting: effect of aspect ratio, Int. J.
Heat Mass Transfer 37 (6) (1994) 917925.
[19] M. Sankar, Youngyong Park, J.M. Lopez, Younghae Do, Numerical study of
natural convection in a vertical porous annulus with discrete heating, Int. J.
Heat Mass Transfer 54 (2011) 14931505.
[20] K. Vafai, C.L. Tien, Boundary and inertia effects on ow and heat transfer in
porous media, Int.J. Heat Mass Transfer 24 (1981) 195203.
[21] R.L. Fredrick, On the aspect ratio for which the heat transfer in differentially
heated cavities is maximum, Int. Commun. Heat Mass Transfer 26 (4) (1999)
549558.
[22] M. Kaviany, Non-Darcian effects on natural convection in porous media
conned between horizontal cylinders, Int. J. Heat Mass Transfer 29 (10)
(1986) 15131519.
[23] G. Lauriat, V. Prasad, Natural convection in vertical porous cavity: a numerical
study of Brinkman-extended Darcy formulation, J. Heat Transfer 109 (1987)
688696.

[24] Z. Tian, Y. Ge, A fourth-order compact nite difference scheme for the steady
stream functionvorticity formulation of the NavierStokes/Boussinesq
equations, Int. J. Numer. Methods Fluids 41 (2003) 495518.
[25] G. De Vahl Davis, Natural convection in a square cavity: a benchmark
numerical solution, Int. J. Numer. Methods Fluids 3 (1983) 249264.
[26] G. De Vahl Davis, I.P. Jones, Natural convection in a square cavity: a
comparison exercise, Int. J. Numer. Methods Fluids 3 (1) (1983) 227
248.
[27] J.C. Kalita, D.C. Dalal, A.K. Dass, Fully compact higher-order computation of
steady-state natural convection in a square cavity, Phys. Rev. E 64 (6) (2001).
066703-13.
[28] S.K. Pandit, J.C. Kalita, D.C. Dalal, A transient higher order compact schemes for
incompressible viscous ows on geometries beyond rectangular, J. Comput.
Phys. 225 (2007) 11001124.
[29] N.H. Saeid, I. Pop, Transient free convection in a square cavity lled
with a porous medium, Int. J. Heat Mass Transfer 47 (2004) 1917
1924.
[30] H. Van Der Vorst, BiCGSTAB: a fast and smoothly converging variant of BiCG
for the solution of nonsymmetric linear systems, SIAM J. Sci. Comput. 13
(1992) 631644.
[31] W.F. Spotz, Accuracy and performance of numerical wall boundary conditions
for the 2D incompressible stream-function vorticity, Int. J. Numer. Methods
Fluids 28 (1998) 737757.