See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/258806611

Natural Convection Heat Transfer Enhancement in a Square

Cavity Periodically Cooled from Above

Article in Numerical Heat Transfer Applications · April 2013

DOI: 10.1080/10407782.2013.733248

CITATIONS READS

47 540

4 authors, including:

A. Raji Mohammed Hasnaoui

Faculté des Sciences et Techniques de Beni Mellal Cadi Ayyad University
96 PUBLICATIONS 1,053 CITATIONS 298 PUBLICATIONS 3,964 CITATIONS

SEE PROFILE SEE PROFILE

Mouaouia Firdaouss
Sorbonne Université
22 PUBLICATIONS 1,941 CITATIONS

SEE PROFILE

All content following this page was uploaded by Mohammed Hasnaoui on 12 July 2015.

The user has requested enhancement of the downloaded file.

Numerical Heat Transfer, Part A, 63: 511–533, 2013
Copyright # Taylor & Francis Group, LLC
ISSN: 1040-7782 print=1521-0634 online
DOI: 10.1080/10407782.2013.733248

NATURAL CONVECTION HEAT TRANSFER

ENHANCEMENT IN A SQUARE CAVITY
PERIODICALLY COOLED FROM ABOVE

A. Raji1, M. Hasnaoui2, M. Firdaouss3,4, and C. Ouardi1

1
Physics Department, Laboratory of Flows and Transfers Modelling
(LAMET), Sultan Moulay Slimane University, Faculty of Sciences and
Technics, Béni-Mellal, Morocco
2
Physics Department, Laboratory of Fluid Mechanics and Energetics
(LMFE), Cadi Ayyad University, Faculty of Sciences Semlalia,
Unit affiliated with CNRST, Marrakech, Morocco
3
CNRS, UPR 3251, LIMSI, Orsay, France
4
UPMC Univ. Paris, Paris, France

The present article reports numerical results of natural convection within an air filled
square cavity with its horizontal walls submitted to different heating models. The tem-
perature of the bottom horizontal surface (hot temperature) is maintained constant, while
that of the opposite surface (cold temperature) is varied sinusoidally with time. The
remaining vertical walls are considered adiabatic. The parameters governing the problem
are the amplitude (0
a
0.8) and the period (s  0.001) of the variable temperature,
the Rayleigh number (103
Ra
7  106), and the Prandtl number (Pr ¼ 0.71). In con-
stant cooling conditions (a ¼ 0), up to three different solutions (monocellular flow MF,
bicellular vertical flow BVF, and bicellular horizontal flow BHF) are obtained. Their
existence ranges are delineated and, in the limits of the existence range of each solution,
the transitions observed are identified and described. In the variable cooling conditions,
the effect of the amplitude and the period of the exciting temperature on fluid flow and
heat transfer is examined in the case of the MF, and BHF for specific values of Ra.
Results are presented in terms of Wmax(t), Wmin(t), Nu(t) and streamlines, heatlines,
and isotherms during the evolutions of selected flow cycles. In comparison with the
constant heating conditions, it is found that the variable cooling temperature could lead
to a drastic change in the flow structure and the corresponding heat transfer, especially
at specific low periods of the cold variable temperature. This leads to a resonance
phenomenon characterized by an important increase in heat transfer by about 46.1%
compared to the case of a constant cold temperature boundary condition.

Received 15 May 2012; accepted 15 September 2012.

Address correspondence to M. Hasnaoui, Cadi Ayyad University, Faculty of Sciences Semlalia,
Physics Department Laboratory of Fluid Mechanics and Energetics (LMFE), Unit affiliated with
CNRST (URAC, 27), B.P. 2390, Marrakech, Morocco. E-mail: [email protected]; hasnaouim@
yahoo.fr

511
512 A. RAJI ET AL.

NOMENCLATURE
a amplitude of the periodic b thermal expansion coefficient of fluid,
0
temperature, a ¼ a’ = ðT 0H  T C Þ 1=K
g acceleration due to the gravity, m=s2 dt dimensionless time step
H0 height of the cavity, m k thermal conductivity of fluid,
L0 length of the cavity, m W=K  m
Nu mean Nusselt number n kinematic viscosity of fluid, m2=s
2
Pr Prandtl number, Pr ¼ n=a X dimensionless vorticity, X ¼ X0 H0 =a
Ra Rayleigh number, W dimensionless stream function,
3
Ra ¼ g b ðT 0H  T 0C ÞH 0 =an W ¼ W0 =a
t dimensionless time, t ¼ t0 a=H 0 s dimensionless period of the cooling
2
T dimensionless fluid temperature, temperature, s ¼ s0 a=H 0
T ¼ ðT 0  T 0C Þ=ðT 0H  T 0C Þ Subscripts
T0 C temperature of the cold wall, K C cooled surface
T0 H temperature of the heated wall, K H heated surface
u, v dimensionless horizontal and vertical max maximum value
velocities, ðu; vÞ ¼ ðu0 =H 0 =a; v0 =H 0 =aÞ min minimum value
x, y dimensionless coordinates,
ðx; yÞ ¼ ðx0 =H 0 ; y0 =H 0 Þ Superscripts
0
a thermal diffusivity of fluid, m2=s dimensional variable

1. INTRODUCTION
Natural convection flows generated by buoyancy forces in rectangular closed
cavities have been the object of numerous studies in the past. Actually, considerable
efforts are still devoted in the area where this basic geometry remains attractive. In
some situations, the need to change the dynamical behavior of complex technologi-
cal systems may be necessary by adding partitions [1–3], by dispersing nano-scale
particles [4–8], or by imposing external constraints (boundary conditions) variable
in time but with controllable parameters to allow these systems to operate in better
conditions. In fact, in practical applications involving natural convection as a
removal heat transfer mechanism, the energy provided to the system is variable in
time and gives rise to unsteady natural convection flow. Solar collectors and printed
circuit boards are examples of such systems submitted to variable thermal boundary
conditions. In addition, thermal and dynamical behaviors of a fluid subjected to time
dependent thermal conditions are impossible to predict on the basis of the results
obtained with constant temperature or heat flux conditions due to the strong non
linear aspect of the governing equations. Moreover, by an appropriate choice of
the parameters characterizing the variable excitation (amplitude and period of a per-
iodic heating or cooling temperature, for example), it becomes possible to establish a
variety of dynamical regimes (periodic, quasi-periodic, intermittent, chaotic etc.)
instead of a stationary regime. Consequently, the thermal exchanges within the con-
sidered system could be improved or reduced compared to the stationary regime.
Hence, many studies have been performed with oscillating temperature boundary
conditions. In an earlier investigation, Lage and Bejan [9] studied numerically and
theoretically the problem of natural convection in enclosures with one side heated
with a pulsating heat flux. They showed that the buoyancy-induced flow resonates
 NATURAL CONVECTION IN A SQUARE CAVITY 513

to a certain frequency of the pulsating heat input and the resonance phenomenon is
characterized by maximum fluctuations observed in the heat transfer evolution with
the period of the time-depending temperature. Two-dimensional natural convection
in a horizontal channel heated from below with a temperature varying sinusoidally
with time and provided with rectangular adiabatic blocks distributed on its lower
wall was studied numerically by Douamna et al. [10]. Results of the study show that
the three different routes leading to chaos were identified by progressively varying
the amplitude and the period of the heating variable temperature. Transient convec-
tive heat transfer in a square cavity subjected to a sinusoidal heat flux modulation on
the bottom wall was investigated numerically by Ishida et al. [11]. Results of the
study show the existence of different types of resonant modes, depending on the
amplitude and the angular frequency of the vibration. The numerical results
obtained by Lakhal et al. in the case of a square cavity partially [12] heated from
below with periodic variable temperatures, showed that the resulting flow structure
and heat transfer are strongly dependent on the amplitude and the period of the vari-
able temperature. Later on, Lakhal et al. [13] studied numerically transient natural
convection in a square cavity isothermally cooled from above and partially heated
from the side. The heating temperature was varied sinusoidally in time with a pulsat-
ing manner. The frequencies leading to the resonance phenomenon were determined
for both types of considered excitations. The pulsating temperature was found to
lead to a better enhancement of the heat transfer in comparison with that engendered
by the sinusoidal varying temperature. Antohe and Lage [14] investigated theoreti-
cally and numerically the case of clear fluid and fully saturated porous medium dif-
ferentially heated enclosures with a time periodic pulsating heat flux. The numerical
simulations indicate that the natural convection activity within the enclosure reaches
several local maxima for certain values of the heating frequency referred to as reson-
ance frequency. In the continuity of this work, the same authors [15] performed a
study where the effect of the Prandtl number on the natural convection flow was
examined. The resonance frequency of a fluid or a porous medium system was found
to vary as Pr1=2. Kwak and Hyun [16] and Kwak et al. [17] studied also such a res-
onant phenomenon of natural convection. It was clearly noted from their studies
that the resonant frequency might be well predicted by the Brunt-Väisälä frequency,
which is characterized by the thermal stratification of a system. The case of convec-
tion heat transfer was considered by Abourida et al. in a square cavity filled with air
and submitted to variable thermal boundary conditions on its horizontal [18] and
vertical [19] walls. Different modes of heating were investigated. Using initial con-
ditions favorable to the monocellular flow, results reported in [18] show that, at high
Rayleigh number, the periodic heating can be used to enhance (slightly) or to reduce
(notably) the mean heat losses with respect to the case of constant temperatures. In
addition, fundamental differences in terms of behaviors are observed by changing
the heating mode. For a cavity with vertical active walls [19], it is reported that
the buoyancy driven flow presents a tendency to enter in resonance with the
periodic heating only when the cold temperature was maintained constant. In a rela-
tively recent work, Zhao et al. [20] analyzed numerically and theoretically transient
laminar natural convection induced by two discrete heating elements flush mounted
on one vertical wall of a square enclosure. One heater provides the system
with time-dependent temperature and the other maintains constant the heating
514 A. RAJI ET AL.

temperature. The resonance frequencies are obtained numerically and predicted by

mean of theoretical analysis. The transient interaction of the two kinds of heat
sources was investigated with the objective to maximize the overall heat transfer rate.
The problem of transient natural convection in an inclined enclosure filled with
water was studied experimentally and numerically by Wang et al. [21]. The tempera-
ture of one wall of the enclosure is varied stepwise periodically with time while the
opposite wall temperature is kept at a constant time-averaged value. The obtained
results show that an upper wall temperature oscillating leads to a periodically varied
heat flux across the enclosure. Also, the net heat flux is from the lower wall to the
upper one. The coupling between natural convection and surface radiation was
examined by El Ayachi et al. [22] in the case of a square cavity with one of its vertical
walls submitted to a heating temperature varying periodically with time. The results
obtained show that the heat transfer could be significantly enhanced, with respect to
the case of a constant heating temperature, by a proper choice of the parameters
related to the periodic temperature and the emissivity of the walls. In order to aug-
ment the natural convection cooling, Florio and Harnoy [23] have conducted a
numerical study of natural convection in a vertical channel provided with a trans-
verse oscillating thin short plate located in close proximity of a rectangular heat
source. Results of the study show that, by an appropriate choice of the controlling
parameters, a significant enhancement of the heat transfer can be obtained in
comparison with the standard natural convection case. Finally, it is interesting to
underline that the imposition of boundary conditions periodically variables with
time allows, by means of an appropriate choice of the control parameters, to
suppress the chaos in some conditions where it exists normally, as reported by Lima
and Pettini [24] and Kim and Stringer [25], or to establish it where it would not be
existing, as reported by Xia et al. [26]. These possibilities, offered by variable heating
conditions, could be exploited properly according to the application needs.
The majority of the published works consider cavities with the driving heating
walls customarily held at constant (or periodically varying) temperatures or heat
fluxes. However, no attempt has considered the inverse case where the cavities are
heated with a constant temperature and submitted to a time-dependent cooling. This
intermittent cooling generates convection in an unsteady manner and a good under-
standing of free convective flows under such thermal boundary conditions is neces-
sary to avoid overheating phenomena, which are frequently encountered in practical
situations. Hence, the objective of the present study consists to investigate the effect
of the periodic cooling temperature on the cooling process and its impact on the flow
structure and temperature distribution inside the cavity. Some answers elements to
these interrogations are given in the present study which is devoted to investigate
numerically the natural convection heat transfer in a square cavity filled with air
and submitted to a time dependent cooling. Then, the effect of the parameters related
to the variable temperature (amplitude a and period s) on the dynamical and thermal
behaviors is examined for selected values of Rayleigh number.

2. PROBLEM FORMULATION
A schematic of the physical problem and the associated boundary conditions
is shown in Figure 1. It consists of a two-dimensional square enclosure with
 NATURAL CONVECTION IN A SQUARE CAVITY 515

Figure 1. a) Geometry of the problem, and b) thermal excitations (color figure available online).

adiabatic vertical walls and thermally active horizontal walls. Specifically, the hori-
zontal upper wall of the enclosure is cooled with a temperature varying sinusoid-
ally with time, while the horizontal lower wall is heated at a constant temperature.
The third dimension of the cavity (direction normal to the plane of the diagram) is
assumed large enough so that the fluid motion can be considered two-dimensional.
The flow is conceived to be laminar and incompressible with negligible viscous
dissipation. All the thermophysical properties of the fluid are assumed constant
except the density in the buoyancy term which is assumed to vary linearly
with temperature (Boussinesq approximation); such a variation gives rise to the
buoyancy forces. Taking into account the above-mentioned assumptions, the
non-dimensional governing equations, written in the vorticity stream-function for-
mulation (X-W), are as follows.

" #
qX qðuXÞ qðvXÞ q2 X q2 X qT
þ þ ¼ Pr 2
þ 2  Ra Pr ð1Þ
qt qx qy qx qy qx

qT qðuTÞ qðvTÞ q2 T q2 T
þ þ ¼ þ 2 ð2Þ
qt qx qy qx2 qy

q2 W q 2 W
þ 2 ¼ X ð3Þ
qx2 qy

The dimensionless stream function and the vorticity are related to the non-
dimensional velocity components by the following expressions.

qW qW qu qv
u¼ ;v ¼ ; and X¼  ð4Þ
qy qx qy qx
516 A. RAJI ET AL.

2.1. Boundary Conditions

The hydrodynamic boundary conditions are such that the stream function and
the velocity components are zero on the rigid walls of the cavity (u ¼ v ¼ W ¼ 0). The
dimensionless thermal boundary conditions associated to the equations are as follows.
8
. > T ¼1 for y ¼ 0 and 0
x
1ðheated wallÞ
>
>
<
. T ¼ a sin ð2pt=sÞ for y ¼ 1 and 0
x
1ðcooled wallÞ
>
> qT
. >
: ¼0 for x ¼ 0; 1 and 0
y
1 ðadiabatic wallsÞ
qx
The parameters a and s are the amplitude of the cooling temperature and its period,
respectively.

2.2. Heat Transfer

At each time step, the mean Nusselt number, characterizing the heat transfer
through the cooled wall, is evaluated as follows.

Z

1
qT


NuðtÞ ¼ dx ð5Þ
0 qy
y¼1

The evolutions of Nu(t) with time, as well as those of Wmax(t) and Wmin(t), provide
useful information about the averaged instantaneous heat losses, and the nature of
the flow (periodic or nonperiodic, monocellular or multicellular flow structures, sim-
ple or complex periodicity, etc).
The mean Nusselt number, characterizing the heat transfer rate through the
cooled wall during one flow period, sC, is evaluated as as follows.
Z sC
1
Nu ¼ NuðtÞdt ð6Þ
sC 0

The flow period, sC, represents the period of the temporal variation of the Nusselt
number and also characterizes the flow period when it is identical to that of the extre-
mum stream function (which is the case in general), but it could be different from
that of the exciting cold temperature.

2.3. Heatlines Formulation

The visualization of the paths followed by the heat flows through the enclosure
requires the use of the heatlines concept (lines of constant heatfunction H). Such
lines are defined through the first derivatives of the function H, as follows [27].

qH qT qH qT
¼ u T  ; ¼vT ð7Þ
qy qx qx qy
 NATURAL CONVECTION IN A SQUARE CAVITY 517

The dimensionless heatfunction equation can be derived easily from Eq. (7) as
follows.

q2 H q2 H qðvTÞ qðuTÞ
þ ¼ þ ð8Þ
qx2 qy2 qx qy

Equation (8) is a conduction type problem with a source term becoming zero
if the fluid flow subsides (fluid at rest). The solution of Eq. (8) yields the values of
the heatfunction in the nodes of the computational domain. The contour plots of
the heatfunction values provide heatline patterns.
The H field is defined through its first-order derivatives and only differences of
H values are required instead of its intrinsic values. Note, the similarity in nature
between the heatline field and pressure field in incompressible fluid flows. Thus,
we have the freedom to state that H(0, 0) ¼ 0 (an arbitrary reference value for the
heat function H).
The boundary conditions for the dimensionless heatfunction, Eq. (8), are
obtained by integrating Eq. (7) along the considered boundary. The following equations
can be used to determine the values of heatfunction at the walls of the cavity.
Z y
qT
Hð0; yÞ ¼  dy for x ¼ 0
0qx
Z x
qT
Hðx; 1Þ ¼ Hð0; 1Þ þ dx for y ¼ 1
qy
Z0 y ð9Þ
qT
Hð1; yÞ ¼ Hð1; 1Þ  dy for x ¼ 1
qx
Z0 x
qT
Hðx; 0Þ ¼ Hð1; 0Þ þ dx for y ¼ 0
0 qy

3. NUMERICAL METHOD
The nonlinear partial differential governing equations (1)–(3), written in con-
servative form for the non-linear convective terms, were discretized using a second
order accurate finite difference method. The integration of Eqs. (1) and (2) was
ensured by the alternate direction implicit method (ADI). This classical method is
well adapted to the study of transient convection problems. Values of the stream
function at all grid points were obtained with Eq. (3) by using the point successive
over-relaxation method (PSOR) with an optimum over-relaxation coefficient
equal to 1.92 for the grid 81 
81 adopted
in the
present
study. At each time step,
P
kþ1 k
P
kþ1
5
the convergence criterion i;j
Wi;j  Wi;j
= i;j
Wi;j
< 10 has been satisfied
for W, where Wki;j is the value of the stream function at the node (i, j) for the kth iter-
ation level. The velocities at all grid points were determined with Eq. (4) using
updated values of the stream function. To check the effect of the grid size, prelimi-
nary tests were conducted for different combinations of the governing parameters.
The grid of 81  81 was found to be appropriate for the present study since it permits
518 A. RAJI ET AL.

Table 1. Effect of the grid size on the results of MF and BHF for Ra ¼ 106

MF BHF

Grid Wmax Wmin Nu Wmax Wmin Nu

61  61 79.732 6.1836 6.5911 39.7225 39.7225 4.4979

81  81 79.661 6.1814 6.4722 39.8366 39.8366 4.4527
101  101 79.589 6.1732 6.4102 39.8724 39.8724 4.4319

a good compromise between the computational cost (a significant reduction of the

execution time) and the accuracy of the obtained results. The values of time step sizes
considered for the simulations were selected in the range 107
dt
105 depending
on the values of the governing parameters. More precisely, the small values of dt are
used for high values of Ra and small values of s. Results concerning the sensitivity
of the solutions vis-à-vis of the grid are presented in Table 1 for Ra ¼ 106 in the case
of steady MF and BHF. This table shows that the results obtained with the grid of
81  81 differ by less than 0.97% = (0.47%) (as a maximum relative difference
observed in the case of Nu corresponding to the MF=(BHF)) from those correspond-
ing to the grid of 101  101. The accuracy of the numerical model was checked by
comparing results from the present investigation against those previously published
by De Vahl Davis [28] in the case of a differentially heated square cavity. As it can be
seen in Table 2, the agreement observed is excellent and the maximum relative devi-
ation registered remains less than 1.4 and 2.3%, respectively, in terms of Wmax and
Nu. The numerical code was also validated against the study of Ridouane et al.
[29] in the case of a square cavity heated from below. The comparison shows a good
agreement with a maximum deviation of 2% in terms of Nu. Finally, we proceeded to
a supplementary test of validation by checking the energy balance of the system. In
fact, it was carefully verified that the energy transmitted to the fluid by the heated
wall during one flow cycle, leaves the cavity through the cold one; this energy balance
was verified for all the computations conducted with a maximum difference of 1%.
The calculations were performed on a Pentium 4 (Processor: Genuine Intel(R) CPU
1.3 GHz, Memory: 3 GB RAM). The CPU time required for obtaining sustained
oscillatory solutions depends on the parameters a, s, and Ra. The minimum CPU
time required in the simulations is of about 1200s for small periods. This computa-
tional effort is approximately four times less than that needed for large values of s.

Table 2. Validation of the numerical code in terms of maximum stream

function and Nusselt number

Present study Benchmark [28]

Ra Wmax Nu Wmax Nu
3
10 1.174 1.118 1.174 1.117
104 5.072 2.248 5.098 2.238
105 9.620 4.546 9.612 4.509
106 16.908 9.023 16.750 8.817
 NATURAL CONVECTION IN A SQUARE CAVITY 519

4. RESULTS AND DISCUSSION

The main parameters governing the present problem are the amplitude of the
variable temperature (0
a
0.8), its period (s  0.001), the Rayleigh number
(103
Ra
7  106), and the Prandtl number (Pr ¼ 0.71). In the following, the
results obtained are first presented in terms of Nu variations with Ra for all the solu-
tions obtained (case of constant heating conditions), then, in the variable regime,
temporal variations of maximum (Wmax(t)) and minimum (Wmin(t)) stream functions
and Nusselt number (Nu(t)) are presented in the unsteady regime in order to learn
how the system replies once excited with a cooling temperature periodically varying
with time.

4.1. Constant Active Temperatures (a ¼ 0)

In the case of a constant cold temperature boundary condition, the obtained
results are in general steady state. The preliminary tests conducted show that the
multiplicity of solutions is strongly dependent on the Rayleigh number and the initial
conditions imposed to the system. The objective in this paragraph consists to identify
the different solutions that may exist in this configuration and the determination of
the existence range for each solution obtained. The streamlines corresponding to the
three modes of convection are reported in Figure 2. Since these multiple solutions
may exist for the same set of governing parameters, the choice of initial conditions
is determinant in this study. Hence, by using the conduction regime to initiate
the numerical code for Ra ¼ 105 (moderate convection), the final steady state is a
dominating positive (or negative) monocellular flow (MF) with two small negative
(or positive) vortices located in the vicinity of two opposite corners of the cavity
(Wmax ¼ 26.04 and Wmin ¼ 0.12). Note that, for the same value of Ra, the flow with
Wmin ¼ 26.04 and Wmax ¼ 0.12 is also a solution of the problem since the flow

Figure 2. Variations of the Nusselt number versus the Rayleigh number for all solutions obtained in the
steady state regime (color figure available online).
520 A. RAJI ET AL.

rotation can either be clockwise or counter-clockwise without any effect on the

amount of heat evacuated through the cold surface. By increasing progressively
Ra and using as initial conditions (at each computation) the results corresponding
to the last Ra, the flow remains monocellular and the increase of the Rayleigh num-
ber is seen to be favorable to the secondary flow (to the small vortices). In fact, for
Ra ¼ 2.5  106 (value marking the threshold above which the monocellular flow tran-
sits towards another type of flow), Wmax ¼ 119.66 and Wmin ¼ 15.21 which indicates
that the increase rate of the small cells with Ra is more important than that of the
main flow. Now, by decreasing Ra below 105 and using initial conditions favorable
to the monocellular flow, the latter is maintained until the transition towards the
conductive regime which occurs at Ra  2.5  103 (Wmax ¼ 0.01846, Wmin ¼ 0.00001
and Nu ¼ 1.00003).
The second kind of solution obtained is the bicellular horizontal flow (BHF),
which is characterized by the formation of two horizontal cells with comparable
intensities. The flow structure is organized such that each cell is in contact with only
one active horizontal boundary. The obtaining of this kind of solution was facilitated
by the transition of the MF observed at Ra ¼ 2.6  106; this transition occurs
towards the BHF. Once obtained, the BHF was used to start the numerical code
and its existence in the steady regime was delineated by varying Ra in the range
[3.5  105, 6.8  106]. Below this range, the solution obtained is monocellular even
if the initialization of the numerical code is in favor of the BHF. Above this range,
it was verified that for Ra ¼ 7  106, the BHF is unsteady periodic; its evolution with
time is characterized by regular oscillations. Note that, in the steady state regime, all
the solutions corresponding to the BHF are found to be symmetrical with respect to
the horizontal plane y ¼ 0.5.
The last kind of solution obtained for this configuration is the bicellular
symmetrical vertical flow (BVF) which is a bicellular ascending flow, presenting a
symmetry with respect to the vertical plane x ¼ 0.5. In order to start the numerical
code with this bicellular flow for the case Ar ¼ 1, results obtained with an aspect ratio
Ar ¼ 2 were used since this value of the aspect ratio permits to reach easily this flow
structure. Another alternative to also obtain this flow pattern for Ar ¼ 1 consists of
using a stream function field defined by W(x, y) ¼ A  sin(2px)  sin(py) and a con-
duction temperature field for an adequate Rayleigh number. In the steady state
regime, this solution was maintained for Ra varying in the range [1.1  104,
4.1  104]. Below this range of Ra, the transition occurs directly towards the MF.
However, a window of periodic oscillatory BVF (OS BVF) was observed in the range
[4.2  104, 5.2  104] and, above this range, the transition occurs directly towards the
MF as it was verified for Ra ¼ 5.3  104. The arrows in Figure 2 indicate the tran-
sition undergone by each solution when Ra is progressively increased (blue arrows)
or decreased (red arrows).
The variations of the Nusselt number versus the Rayleigh number are presented
in Figure 2 for all solutions obtained. In this figure, it is seen that the existence of
the BHF and the BVF are restricted respectively, to high and relatively intermediate
Rayleigh numbers while the MF is obtained in a wide range of this parameter. In
addition, in the steady regime the MF is seen to be the most favorable to the heat
transfer as it permits a better thermal interaction between the horizontal active
boundaries. As indicated, for Ra ¼ 2  106 a quantitative comparison between the
 NATURAL CONVECTION IN A SQUARE CAVITY 521

amounts of heat exchanged shows that the convective heat transfer for the MF
(Nu ¼ 7.374) is enhanced by about 29.14% in comparison with the one corresponding
to the BHF (Nu ¼ 5.710). This difference in heat transfer can be easily explained by
using the heatfunction concept and its contour lines (known as heatlines) which allow
a visualization of the heat transport process induced by convection. In Figure 3, we
can observe some similitude in the shapes of heatlines and streamlines indicating a
predominant effect of convection in the heat transfer process (large values of the
streamfunction). In the case of MF, Figure 3a shows that the heat provided by the
bottom heated wall to the confined fluid is largely evacuated directly through
the top cold wall as a consequence of the existence of a main unicellular flow in con-
tact with the thermally active surfaces. The heat released at the level of the right part
of the lower wall is aspired by the small convective cell before being transported by the
main convective cell. This direct evacuation of the heat released by the heated wall
leads to a good convective heat exchange between the heated surface and the confined
fluid. The examination of the bicellular horizontal flow heatlines, presented in
Figure 3b, shows that the heat transport process is different from that corresponding
to the unicellular flow. In fact, the heat dissipated by the heated wall is constrained to
follow a complicated trajectory before reaching the cold wall (the heat follows an ‘‘S’’
trajectory). This path followed by the heat before its evacuation leads, obviously, to a
reduction in terms of heat exchange by convection between the horizontal active walls
and the fluid.

4.2. Variable Cooling Temperature (a 6¼ 0 and s 6¼ 0)

When the cavity is periodically cooled from above with a temperature varying
sinusoidally with time, results of the simulations indicate that the natural-convection
activity within the enclosure becomes periodic. The evolutions with time correspond-
ing to the fluid circulation and heat transfer are characterized by sustained periodic
variations reached after a transient period which strongly depends on the governing
parameters (a, s, and Ra). Hence, combined effects of the amplitude and the period
of the cooling temperature on Wext and Nu are examined in the following subsections
for selected values of Ra. Only, the MF and BHF are concerned here since the

Figure 3. Heatlines for Ra ¼ 2  106. a) MF structure and b) BHF structure (color figure available online).
522 A. RAJI ET AL.

important changes in the presence of variable cooling conditions occur, in general, at

relatively high values of Ra.

4.2.1. Effect of the amplitude a of the cooling temperature

4.2.1.1. Case of MF. The effect of the amplitude on the monocellular flow
structure is first tested for Ra ¼ 105 (a moderate value of Ra) and s ¼ 0.5 (relatively
large value of the period). The reference case corresponding to a constant cooling
temperature (a ¼ 0) is also presented to better highlight the effect of the variable
cooling. The obtained results are presented in Figures 4a–4c respectively, in terms
of temporal variations of Wmax(t), Wmin(t), and Nu(t). It can be noted easily in
Figure 4b, that the small cells corresponding to a ¼ 0 (constant cold temperature)
are of very low intensity (and volume, also; the corresponding streamlines are not
presented here). In the same figure, it is also seen that the increase of the amplitude
of the cooling temperature is evidently favorable to the small cells; it leads in terms of

Figure 4. Effect of the amplitude a for Ra ¼ 105, s ¼ 0.5. a) Wmax(t), b) Wmin(t), and c) Nu(t) (color figure
available online).
 NATURAL CONVECTION IN A SQUARE CAVITY 523

mean values to its intensification. The oscillations that characterize the variations of
Wmin(t) with time are periodic and more complex than the sinusoidal nature imposed
to the excitation. The temporal evolutions of Wmax(t) (characterizing the intensity of
the main cell) and Nu preserve the sinusoidal form of the variable temperature, and
are characterized by increasing amplitudes following the increase of the parameter a.
However, it is to be specified that the imposed thermal cooling excitation is seen to
be without any noticeable effect on the mean values of these quantities, since they
remain almost identical to their corresponding values obtained without thermal
modulation (a ¼ 0) for this moderate value of the Rayleigh number.
The nature of the oscillations of the monocellular flow becomes complicated by
the increase of Ra up to 106. Indeed, for s ¼ 0.5 the monocellular flow is preserved
only for small amplitudes of the cooling temperature (a
0.176), as illustrated in
Figures 5a–5c. Oscillations are around average values different from those corre-
sponding to the reference case; they are in the benefit of Wmin(t) but in the disadvan-
tage of Wmax(t) and Nu(t) in the case of monocellular dominating flows. In fact, the
increase of Ra is accompanied by an improvement of the intensity of the small cell and
leads consequently-to time-averaged values of heat transfer and intensity of the main
flow lower than those corresponding to the reference case. It was verified that for
a  0.177, the monocellular flow disappears in favor of the bicellular horizontal oscil-
lating flow with a loss of the sinusoidal nature of the oscillations; the latter become
more complex by increasing the amplitude a above this threshold (see 4.2.1.2.).
In Figures 6a–6d, the effect of the variable cooling on the dynamical, thermal,
and heat transfer behaviors of the fluid is illustrated by presenting streamlines, iso-
therms, and heatlines over one flow cycle for Ra ¼ 106, s ¼ 0.5, and a ¼ 0.175 at selec-
ted instants indicated by the letters a–d in Figure 6 where the monocellular flow
remains visibly dominating throughout the cycle. During the evolution of the cycle,
the flow passes through various states characterized by the presence of a big domi-
nating cell and the existence of two small cells located at two diametrically opposite
corners of the cavity.

4.2.1.2. Case of BHF. It should be recalled that for a ¼ 0 and Ra below the
threshold of the Hopf’s bifurcation, the BHF is characterized by the presence of
two horizontal cells with equal size and intensity. The effect of a on this type of flow
is illustrated in Figure 7 for Ra ¼ 106 and s ¼ 0.5. As can be seen in this figure, an
increase of a leads to an oscillatory nature increasingly complex and obviously dif-
ferent from the sinusoidal nature of the imposed excitation. In addition, a visual
inspection of the curves reveals a clear deviation of the averaged values from the ref-
erence ones obtained with constant cold temperature. For values of the amplitude
larger than 0.68, the BHF disappears in favor of the MF structure. To measure
the impact of the excitation on the flow structure, we keep in mind the reference case
corresponding to a ¼ 0 and presented in Figure 2. In Figure 8, six instants indicated
by the letters a–f are identified in the flow cycle, and the corresponding streamlines,
isotherms and heatlines are presented in Figures 8a–8f. Note, that each half cycle of
the flow is favorable to the intensification of one cell to the detriment of the other
which loses in intensity and volume but remains resistant. The maximum ratio in
terms of flow intensity of the two cells reaches 1.92 during each half cycle alterna-
tively in favor of one cell or the other.
524 A. RAJI ET AL.

Figure 5. Effect of the amplitude a for Ra ¼ 106, s ¼ 0.5. a) Wmax(t), b) Wmin(t), and c) Nu(t) (color figure
available online).

4.2.2. Effect of the period s of the cooling temperature. The effect of the
period is illustrated in Figure 9 for Ra ¼ 105, a ¼ 0.4, and different values of s. It is to
underline that only the MF was obtained in the steady regime for moderate values of
Ra and a. It is seen in Figure 9b that the nature of the secondary cells oscillations is
different from the sinusoidal nature imposed to the variable cooling temperature. The
oscillations of the secondary cells are below the reference case (a ¼ 0) for s ¼ 0.2, but
pass above this reference during a part of the cycle by progressively increasing the
period as it is observed for s ¼ 0.5 and 0.8. However, the maximum (in absolute value)
of Wmin (Wmin characterizes the small cells in these cycles) is in favor of the low periods;
it decreases (in absolute value) by increasing s. In the case of the main flow and heat
 NATURAL CONVECTION IN A SQUARE CAVITY 525

Figure 6. Streamlines, isotherms, and heatlines over one flow cycle corresponding to a monocellular struc-
ture for Ra ¼ 106, a ¼ 0.175, and s ¼ 0.5 (at instants reported in Figures 6a–6d). a) Wmax ¼ 71.08, Wmin ¼
8.47, b) Wmax ¼ 75.23, Wmin ¼ 5.69, c) Wmax ¼ 80.91, Wmin ¼ 11.46, and d) Wmax ¼ 82.73, Wmin ¼ 6.53
(color figure available online).

transfer, represented in this figure, respectively, by Wmax and Nu, it can be seen
in Figures 9a, and 9c that the sinusoidal nature of the excitation is relatively well
reproduced in the temporal variations of these quantities (Wmax and Nu). The oscilla-
tions are around average values close to those of the reference case (a ¼ 0 or s ! 1);
This behavior is expected since the important variations and transformations under-
gone in the presence of the variable heating or cooling conditions occur in general at
526 A. RAJI ET AL.

Figure 7. Effect of the amplitude a in the case of the bicellular horizontal flow for Ra ¼ 106, s ¼ 0.5. a)
Wmax(t), b) Wmin(t), and c) Nu(t) (color figure available online).

low periods of the exciting temperatures. In fact, it can be seen in Figure 10, plotted
for a ¼ 0.6, Ra ¼ 106 and various values of s, that the important changes in the flow
structure are obtained with the small values of the period, particularly for s ¼ 0.005,
where a double periodicity is observed. For this small value of s, the flow structure
changes drastically during the evolution of the cycle; it passes by states where the
negative cell dominates during one half of the flow cycle and others where the positive
one dominates during the remaining half of the cycle. In addition, the main period-
icity becomes different (larger than that of the excitation) from that of the cold tem-
perature; the period of the latter appears in the cycle only as a secondary periodicity.
However, when the system is excited with low imposed frequencies, it replies by
 NATURAL CONVECTION IN A SQUARE CAVITY 527

Figure 8. Streamlines, isotherms, and heatlines over one flow cycle corresponding to a bicellular flow struc-
ture for Ra ¼ 106, a ¼ 0.4, and s ¼ 0.5 (at instants reported in Figures 8a–8f. a) Wmin ¼ 39.80, Wmax ¼ 50.09;
b) Wmin ¼ 47.23, Wmax ¼ 47.18; c) Wmin ¼ 49.16, Wmax ¼ 32.23; d) Wmin ¼ 46.13, Wmax ¼ 17.23; e)
Wmin ¼ 30.46, Wmax ¼ 30.42, and f) Wmin ¼ 32.23, Wmax ¼ 49.16 (color figure available online).
528 A. RAJI ET AL.

Figure 9. Effect of the period s in the case of an MF structure for Ra ¼ 105 and a ¼ 0.4. a) Wmax(t), b)
Wmin(t), and c) Nu(t) (color figure available online).

conserving the period imposed to the excitation. Hence, the small value s ¼ 0.005,
which leads to drastic changes in the flow structure, is selected to illustrate the impor-
tant flow modifications during the cycle evolution. The eight instants selected in the
cycle to illustrate the flow structure and temperature distribution inside the cavity
are indicated by the letters a–h in Figure 11. For the first selected instant, the flow
is mainly monocellular dominated by a negative cell, while the positive flow is repre-
sented by two small vortices located on two diametrically opposite corners. As the
cycle evolves, during its first half the dominant negative cell loses more and more
of its importance to the advantage of the secondary positive ones. One of the two lat-
ter, secondary at the beginning of the cycle, ends by becoming dominant during some
period in the second half of the cycle, where the two negative vortices at the bottom of
 NATURAL CONVECTION IN A SQUARE CAVITY 529

Figure 10. Effect of the period s in the case of a bicellular flow structure for Ra ¼ 106 and a ¼ 0.6. a)
Wmax(t), b) Wmin(t), and c) Nu(t) (color figure available online).

the cavity play similar roles and undergo similar transformations than those played
by the positive vortices during the first half of the flow cycle. This competition
between cells of, opposite signs leads to repetitive and drastic changes in the flow
structure during the cycles scrolling.
Based on a set of preliminary tests, the time-averaged values of the Nusselt num-
ber Nu, obtained for various periods, are seen to be different from those correspond-
ing to the case of a constant cooling temperature. These differences are in favor of
the variable cooling temperature, which indicates a probable existence of the reson-
ance phenomenon within the cavity. The latter is characterized by maximum fluctua-
tions in the response of the system to an external excitation with a well-defined period.
530 A. RAJI ET AL.

Figure 11. Streamlines over one flow cycle corresponding to a bicellular flow structure for Ra ¼ 106,
s ¼ 0.005, and a ¼ 0.6 (at instants reported in Figures 11a–11h. a) Wmin ¼ 96.27, Wmax ¼ 10.66; b)
Wmin ¼ 51.27, Wmax ¼ 36.22; c) Wmin ¼ 44.11, Wmax ¼ 46.96; d) Wmin ¼ 36.26, Wmax ¼ 53.29; e) Wmin ¼
7.43, Wmax ¼ 87.35; f) Wmin ¼ 33.66, Wmax ¼ 55.22; g) Wmin ¼ 44.29, Wmax ¼ 46.00, and h) Wmin ¼
54.62, Wmax ¼ 35.30 (color figure available online).

The resonance phenomenon has been reported in the past by several authors [9, 13–
16, 19, 22]. The determination of the resonance period was done by progressively
varying the period of the cooling temperature until reaching maximum values of
 NATURAL CONVECTION IN A SQUARE CAVITY 531

Figure 12. Effect of the period s on the time averaged Nusselt number for Ra ¼ 106 and various values of a
in the case of a bicellular flow structure (color figure available online).

the heat transfer. It is important to note that this resonance period depends strongly
on the boundary conditions supplied to the studied configuration. To illustrate this
resonance phenomenon, we present in Figure 12 the variations of Nu versus the per-
iod s of the variable temperature in the case of the BHF for Ra ¼ 106 and various
values of the cooling temperature amplitude. The values corresponding to the case
of the isothermal cooling source (a ¼ 0) are also presented in the same figure as refer-
ences. In Figure 12, we note the important increase of Nu engendered by the increase
of the period s in the case of moderate amplitudes, even for small values of s, while in
the case of small amplitudes the onset of the increase of Nu is delayed. This indicates
an insensitivity of the system vis-à-vis the imposed thermal excitation only in the case
of small periods and amplitudes (Nu remains close to the value corresponding to the
reference case in the case of weak periods and amplitudes). The maximum values of
Nu are obtained for a critical value scr  0.01 independent of the amplitude a. Further
increase of s above scr leads to a continuous decrease of Nu toward the reference case
value from s  0.1=0.7 for a ¼ 0.2=0.4. For a ¼ 0.6, the decrease of Nu occurs toward
an asymptotic value higher than the reference case one. Hence, the importance of the
resonance phenomenon is favored by increasing the amplitude of the cooling tem-
perature and the range of s for which Nu is higher than the Nusselt number corre-
sponding to a ¼ 0 (constant cooling condition) is enlarged by increasing a. It is
important to underscore that the invariability of the resonance period scr by increas-
ing the amplitude a is in accordance with previous numerical findings [9, 14, 19, 22].
Quantitatively, the enhancement (with respect to the case of a constant cooling tem-
perature) observed in terms of heat transfer at scr ¼ 0.01 (period of resonance) is
about 29.6%, 44.51%, and 46.1%, respectively, for a ¼ 0.2, 0.4, and 0.6. Finally, the
resonance phenomenon was not clearly observed in the case of the MF for
Ra ¼ 106, since this flow structure disappears from a ¼ 0.177 (the resonance phenom-
enon is supported by the increase of a in general).
532 A. RAJI ET AL.

5. CONCLUSION
Natural convection in a square cavity, heated from below with a constant tem-
perature and cooled from above with a time periodic temperature, was studied
numerically. The results obtained in the steady regime show the existence of three
types of solutions, and their existence is found to be strongly dependent on the Ray-
leigh number. It is also found that during a flow cycle, the flow structure could
present a complex behavior and, depending on the Rayleigh number, the amplitude
and the period of the exciting temperature. Also, the effect of the parameters related
to the periodic cooling temperature generates different oscillatory convection types
which give rise to enhanced heat transfer with respect to the case of a constant cool-
ing temperature. The existence of the resonance phenomenon is proved in the present
study, and it is found to be more intensified by increasing the amplitude of the excit-
ing temperature. The system enters in resonance with the exciting temperature for a
critical period which is independent of the amplitude. For this critical period, the
results reported show that the periodic cooling can be used to notably enhance the
heat losses in comparison with the case of constant cooling temperature; the
maximum enhancement observed is about 46.1% and it is obtained for scr ¼ 0.01
and a ¼ 0.6.

REFERENCES
1. C. D. Sankhavara, and H. J. Shukla, Numerical Investigation of Natural Convection in a
Partitioned Rectangular Enclosure, Numer. Heat Transfer A, vol. 50, pp. 975–997, 2006.
2. E. Fontana, A. da Silva, V. C. Mariani, and F. Marcondes, The Influence of Baffles on the
Natural Convention in Trapezoidal Cavities, Numer. Heat Transfer A, vol. 58, pp. 125–
145, 2010.
3. W. Wu, and C. Y. Ching, Laminar Natural Convection in an Air-Filled Square Cavity
with Partitions on the Top Wall, Int. J. Heat Mass Transfer, vol. 53, pp. 1759–1772, 2010.
4. B. Ghasemi, and S. M. Aminossadati, Natural Convection Heat Transfer in an Inclined
Enclosure Filled with a Water–CuO Nanofluid, Numer. Heat Transfer A, vol. 55, pp.
807–823, 2009.
5. S. Kakaç, and A. Pramuanjaroenkij, Review of Convective Heat Transfer Enhancement
with Nanofluids, Int. J. Heat Mass Transfer, vol. 52, pp. 3187–3196, 2009.
6. E. Abu-Nada, and H. F. Oztop, Numerical Analysis of Al2O3=Water Nanofluids Natural
Convection in a Wavy Walled Cavity, Numer. Heat Transfer A, vol. 59, pp. 403–419, 2011.
7. A. H. Mahmoudi, M. Shahi, and F. Talebi, Entropy Generation due to Natural Convec-
tion in a Partially Open Cavity with a Thin Heat Source Subjected to a Nanofluid, Numer.
Heat Transfer A, vol. 61, pp. 283–305, 2012.
8. H. Bararnia, K. Hooman, and D. D. Ganji, Natural Convection in a Nanofluids-Filled
Portioned Cavity: The Lattice-Boltzmann Method, Numer. Heat Transfer A, vol. 59,
pp. 487–502, 2011.
9. J. L. Lage, and A. Bejan, The Resonance of Natural Convection in a Horizontal Enclos-
ure Heated Periodically from the Side, Int. J. Heat Mass Transfer, vol. 36, pp. 2027–2038,
1993.
10. S. Douamna, M. Hasnaoui, and B. Abourida, Two-Dimensional Transient Natural Con-
vection in a Repetitive Geometry Submitted to Variable Heating From Below: Numerical
Identification of Routes Leading to Chaos, Numer. Heat Transfer A, vol. 37, pp. 779–799,
2000.
 NATURAL CONVECTION IN A SQUARE CAVITY 533

11. H. Ishida, Y. Takagi, N. Kin, H. Yoshimura, and G. Kawahara, Resonant Thermal Con-
vections in a Square Cavity Induced by Heat-Flux Vibration on the Bottom Wall, Numer.
Heat Transfer A, vol. 58, pp. 20–40, 2010.
12. E. K. Lakhal, M. Hasnaoui, P. Vasseur, and E. Bilgen, Natural Convection in a Square
Enclosure Heated Periodically From Part of the Bottom Wall, Numer. Heat Transfer A,
vol. 27, pp. 319–333, 1995.
13. E. K. Lakhal, M. Hasnaoui, and P. Vasseur, Numerical Study of Transient Natural Con-
vection in a Cavity Heated Periodically with Different Types of Excitations, Int. J. Heat
Mass Transfer, vol. 42, pp. 3927–3941, 1999.
14. B. V. Antohe, and J. L. Lage, Amplitude Effect on Convection Induced by Time-Periodic
Horizontal Heating, Int. J. Heat Mass Transfer, vol. 39, pp. 1121–1133, 1996.
15. B. V. Antohe, and J. L. Lage, The Prandtl Number Effect on the Optimum Heating Fre-
quency of an Enclosure Filled with Fluid or with a Saturated Porous Medium, Int. J. Heat
Mass Transfer, vol. 40, pp. 1313–1323, 1997.
16. H. S. Kwak, and J. M. Hyun, Natural Convection in an Enclosure Having a Vertical Side-
wall with Time Varying Temperature, J. Fluid Mech., vol. 329, pp. 65–88, 1996.
17. H. S. Kwak, K. Kuwahara, and J. M. Hyun, Resonant Enhancement of Natural Convec-
tion Heat Transfer in a Square Enclosure, Int. J. Heat Mass Transfer, vol. 41, pp. 2837–
2846, 1998.
18. B. Abourida, M. Hasnaoui, and S. Douamna, Transient Natural Convection in a Square
Enclosure with Horizontal Walls Submitted to Periodic Temperatures, Numer. Heat
Transfer A, vol. 36, pp. 737–750, 1999.
19. B. Abourida, M. Hasnaoui, and S. Douamna, Convection Naturelle dans une Cavité
Carrée avec des Parois Verticales Soumises à des Températures Périodiques, Rev. Gén.
Therm., vol. 37, pp. 788–800, 1998. (In French.)
20. F. Y. Zhao, D. Liu, and G. F. Tang, Resonant Response of Fluid Flow Subjected to Dis-
crete Heating Elements, Energy Convers. Manage., vol. 48, pp. 2461–2472, 2007.
21. G. Wang, M. Zeng, Y. C. Ren, H. Ozoe, and Q. W. Wang, Transient Heat Flux Measure-
ment of Natural Convection in an Inclined Enclosure with Time-Periodically-Varying
Wall Temperature, Exper. Ther. and Fluid Sci., vol. 35, pp. 105–111, 2011.
22. R. El Ayachi, A. Raji, M. Hasnaoui, and A. Bahlaoui, Combined Effect of Radiation and
Natural Convection in a Square Cavity Differentially Heated with a Periodic Tempera-
ture, Numer. Heat Transfer A, vol. 53, pp. 1339–1356, 2008.
23. L. A. Florio, and A. Harnoy, Use of a Vibrating Plate to Enhance Natural Convection
Cooling of a Discrete Heat Source in a Vertical Channel, Appl. Therm. Eng., vol. 27,
pp. 2276–2293, 2007.
24. R. Lima, and M. Pettini, Suppression of Chaos by Resonant Parametric Perturbations,
Phys. Rev. A, vol. 41, pp. 726–733, 1990.
25. J. H. Kim, and J. Stringer, Applied Chaos, Wiley, New York, 1992.
26. Q. Xia, K. T. Yang, and D. Mukutmoni, Effect of Imposed Wall Temperature Oscilla-
tions on the Stability of Natural Convection in a Square Enclosure, J. Heat Transfer,
vol. 117, pp. 113–120, 1995.
27. S. Kimura, and A. Bejan, The Heatline Visualization of Convective Heat Transfer, Trans
ASME: J. Heat Transfer, vol. 105, pp. 916–919, 1983.
28. G. De Vahl Davis, Natural Convection of Air in a Square Cavity: A Bench-Mark Numeri-
cal Solution, Int. J. Numer. Methods Fluids, vol. 3, pp. 249–264, 1983.
29. E. H. Ridouane, M. Hasnaoui, and A. Campo, Effects of Surface Radiation on Natural
Convection in a Rayleigh-Bénard Square Enclosure: Steady and Unsteady Conditions,
Heat Mass Transfer, vol. 42, pp. 214–225, 2006.

View publication stats