International Journal of Thermal Sciences 95 (2015) 88e98

Contents lists available at ScienceDirect

International Journal of Thermal Sciences

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

Double-diffusive natural convective in a porous square enclosure lled

with nanouid
M. Dastmalchi*, G.A. Sheikhzadeh, A.A. Abbasian Arani
Department of Mechanical Engineering, University of Kashan, Kashan 87317-51167, Iran

a r t i c l e i n f o

a b s t r a c t

Article history:
Received 20 March 2014
Received in revised form
15 February 2015
Accepted 6 April 2015
Available online

In this work, effects of double-diffusive natural convection of Al2O3ewater nanouid on ow eld and
heat transfer in a porous square cavity are investigated. Homogeneous and two-component non-homogeneous of Buongiorno's model that includes the effects of Brownian motion and thermophoresis are
utilized, while the Darcy model is used for the porous medium. The governing equations are discretized
using the nite difference and control volume method. Properties of nanouid have been assumed
functions of temperature and volume fraction of nanoparticles. Since constant Rayleigh number could
not be used, simulations have been performed for various physical conditions such as temperature
difference between the hot and cold walls from 1 to 20
C, bulk volume fraction of nanoparticles from
0 to 0.04, and porosity between 0.1 and 0.5. Both models suggest that by increasing the bulk volume
fraction of nanoparticles heat transfer is reduced, but non-homogeneous model predicts a greater
reduction compared to the homogeneous model. Also non-homogeneous model predicts reduced heat
transfer with increased porosity.
2015 Elsevier Masson SAS. All rights reserved.

Keywords:
Double-diffusive
Thermophoresis
Brownian diffusion
Porous medium
Variable properties
Nanouid
Numerical study

1. Introduction
Nanouids contain small quantity of nano-sized particles
(usually less than 100 nm) that are suspended in a liquid. Researchers [1,2] have observed that nanouids can have anomalously higher thermal conductivities than that of the base uids
even for low solid volume fraction of nanoparticles in a mixture,
thus posing as a promising alternative for thermal applications.
Although the higher conductivity is encouraging, it is by no means
conclusive evidence of the cooling capabilities of such uids and
the sensitivity to the viscosity model and other seems to be undeniable and it plays a key role for heat transfer behavior.
Natural convection in pure uids is driven only by density variations due to temperature gradient. However, Double-diffusive
convection in nanouids is an important uid dynamics topic
describing a form of convection driven by two different density
gradients with different rates of diffusion [3]. These density variations may be caused by gradients in the volume fraction of the
nanouid or by temperature gradient.
One of the most common assumptions in studying the nanouids behavior is that there aren't any gradients in the volume
* Corresponding author. Tel.: 98 3155912446; fax: 98 3155912475.
E-mail address: majid.dastmalchi@gmail.com (M. Dastmalchi).
http://dx.doi.org/10.1016/j.ijthermalsci.2015.04.002
1290-0729/ 2015 Elsevier Masson SAS. All rights reserved.

fraction. This assumption results in a homogeneous mixture of

nanoparticles in the uid. Many researchers used this assumption
and they were not able to predict the decreasing behavior of Nusselt number with adding the nanoparticles volume fraction [4].
There are several mechanisms such as Brownian motion and
thermophoresis that cause heterogeneity in suspensions. Brownian
motion, named after Robert Brown, is the presumably random
drifting of particles suspended in a uid [5]. Temperature gradient
can cause mass ux by a process called either thermophoresis or
thermal diffusion and the Soret effect. This was rst reported by
Tyndall [6] in 1870. Concentration gradients can produce heat
transfer known as diffusion-thermo, or Dufour effect, which is
usually small and negligible [7]. Buongiorno [8] introduced seven
transport mechanisms which cause relative velocity between
nanoparticles and uid. By comparing the diffusion time scale of
transport mechanisms, he showed that the Brownian motion and
thermophoresis are the two most important mechanisms. Kuznetsov and Nield [9] studied natural convective boundary-layer
ow of a nanouid past a vertical plate, analytically. By using
similarity solution, they showed that the Brownian motion and
thermophoresis decreases the Nusselt number. They assumed the
volume fraction of nanoparticles at wall is constant, whereas
physically zero particle ux at the wall will happen. Pakravan and
Yaghoubi [10] investigated effects of Brownian motion,

M. Dastmalchi et al. / International Journal of Thermal Sciences 95 (2015) 88e98

Nomenclature
cp
Cs
d
DB
DT
g
h
JP
K
k
kB
L
n
NBT
Nu
p
Pr
Ra
Re
ST
T
x, y
X,Y

specic heat capacity, J/kg K

non-continuum constant
diameter, m
Brownian coefcient, m2/s
thermophoresis coefcient, m2/s K
gravitational acceleration, m/s2
local heat transfer coefcient, W/m2 K
particle ux vector, kg/m2 s
permeability of the porous medium
thermal conductivity, W/m K
Boltzmann constant, J/K
length, m
normal vector
ratio of Brownian and thermophoretic diffusivities
Nusselt number
pressure, Pa
Prandtl number
Rayleigh number, dened by Eq. (27)
Reynolds number
thermophoresis parameter
temperature, K
dimensional coordinates, m
dimensionless coordinates

Greek symbols
a
uid thermal diffusivity, m2/s

thermophoresis and Dufour in natural convection heat transfer,

analytically. They reported good agreement for Nusselt number by
comparing their results with various experimental results. However they didn't study some effects such as double diffusive natural
convection and change of properties due to presence of volume
fraction gradient. Aminfar and Haghgoo [11] and Pakravan and
Yaghoubi [12] investigated the effects of Brownian motion and
thermophoresis on natural convection heat transfer of aluminaewater nanouid in a square vertical cavity using the twocomponent model, numerically. They concluded that the use of
single phase homogeneous method does not seem reasonable for
modeling this class of natural convection. Haddad et al. [13] studied
nard convection
numerically, Cuo-water nanouid Rayleigh-Be
considering the role of Brownian and thermophoresis effects and
made comparison with the case where both effects were neglected.
They noticed higher heat transfer when Brownian and thermophoresis effects were considered. Ho et al. [14] investigated natural
convection of Al2O3ewater nanouid in square enclosures of three
different sizes, experimentally. They explained the unusual increase or decrease of heat transfer cannot be explained solely based
on relative changes in thermophysical properties of the nanouid,
and other factors such as nanoparticle transport mechanisms
which change the homogeneity of the volume fraction of nanoparticles in the domain are also important. In our previous numerical works [15,16] we utilized Buongiorno's nanouid model
that includes the effects of Brownian motion and thermophoresis to
the case square enclosure (without porous matrix). We showed
homogeneous model is not competent to predict the heat transfer
features of nanouids and our predictions were in better agreement with experimental results.
Natural convection ow in porous media, due to thermal
buoyancy alone, has been widely studied (Combarnous and Bories,
1975) and well-documented in the literature (Cheng [17], 1978;
Bejan [18], 1984) while only a few works have been devoted to

m
r

n
j
J

89

thermal expansion coefcient, K1

porosity
nanoparticle volume fraction
normalized nanoparticle volumetric fraction
dynamic viscosity, N.s/m2
nanouid density, kg/m3
dimensionless temperature
kinematic viscosity, m2/s
dimensional stream function, kg/m s
dimensionless stream function

Subscripts
0
reference
b
bulk or overall, bed
B
Brownian
C
cold
eff
effective
f
uid
fr
freezing point of the base liquid
H
hot, homogeneous
nf
nanouid
p
Particle
s
solid
T
thermophoresis
Superscripts
average

double-diffusive natural convection in porous media. The problem

of natural convection in a porous square enclosure studied by
Osvair et al. [19]. This problem is further investigated in Beckermann et al. [20] and Basak et al. [21].
A rst extension to the case of a natural convective in a porous
medium saturated by a nanouid, based on a model presented by
Buongiorno [8] was made by Nield and Kuznetsov [22]. In their
paper it was assumed that one could control the value of the
nanoparticles volume fraction at the boundary in the same way as
the temperature there could be controlled, but no indication was
given of how this could be done in practice. Khan and Aziz [23] used
the Buongiorno model [8] and similarity method to study the
double-diffusive natural convection from a vertical plate to a porous
medium saturated with a nanouid. They showed the highest
values of reduced Nusselt numbers are achieved in mono diffusion
in a regular uid and the lowest values occur with double diffusion
in nanouids. Nield and Kuznetsov [24] employing the Darcy model
for the momentum equation, presented a similarity solution. They
assumed the simplest possible boundary conditions, namely those
in which both the temperature and the nanoparticles volume fraction are constant along the wall. They found that with increasing the
thermophoresis parameter, Brownian motion parameter, Dufour
parameter and with decreasing the buoyancy ratio, the reduced
Nusselt number decreases. Yadav et al. [25] examined the effect of
boundaries and constant internal heat source on the onset of DarcyBrinkman convection in a nanouid saturated porous layer heated
uniformly from below and cooled from above. Kuznetsov and Nield
[26] used the Buongiorno model [8] to study the double-diffusive
natural convection from a vertical plate to a porous medium saturated with a nanouid using similarity method. They revisited their
previous model and extended it to the case when the zero nanoparticle ux is imposed on the boundary. They found that the
reduced Nusselt number is almost independent of the Brownian
motion parameter. Tham et al. [27] studied the problem of steady

90

M. Dastmalchi et al. / International Journal of Thermal Sciences 95 (2015) 88e98

mixed convection boundary layer ow past a horizontal circular

cylinder with a constant surface temperature and embedded in a
porous medium saturated by a nanouid with Darcy model and
Buongiorno nanouid equation model.
Sheremet et al. [28] studied the free convection ow in shallow
and slender porous cavities. Also Sheremet and Pop [29] and
Sheremet et al. [30] respectively study 2D and 3D natural convection heat transfer in porous enclosure lled with a nanouid using
the Buongiorno model. They found that at low Rayleigh numbers,
low Lewis numbers, low Brownian motion parameter, and high
thermophoresis parameter, a non-homogeneous model is more
appropriate for description of the system.
For experimental study of convection ow in porous media
using nanouids, few studies have been carried out. Hajipour and
Molaei Dehkordi [31] were the rst who experimentally and
numerically investigated the nanouid miconvective heat transfer
inside a vertical channel partly lled with porous metal foam. The
obtained results show that nanouid ow in the presence of porous
metal foams can improve the heat-transfer rate and may have applications in industrial. Nazari et al. [32] investigated the forced
convective heat transfer due to ow of nanouid through a circular
tube lled with a aluminum open-cell porous media with porosity
of 50% experimentally. Casting around space holder with an
average size of 3e5 mm diameter materials method was used for
fabricating the metal foam. They found that using the porous medium inside the tube leads to a signicant enhancement of heat
transfer in comparison with the empty tube.
The aim of the present investigation is to study numerically a
double-diffusive natural convective of variable properties ow in a
square cavity lled with Al2O3ewater nanouid saturated
aluminum porous medium. Simulations have been performed for
several physical conditions with various values of temperature
difference between the hot and cold walls, bulk volume fraction of
nanoparticles, and porosity. The Buongiorno's two-component
non-homogeneous model [8] has been employed. In this paper it
is consider a realistic boundary condition that there is no nanoparticle ux at the walls.
2. Nanouid and porous properties
Many studies have evaluated the nanouids properties
depending on both properties of base uid and nanoparticles and
according to them, several models have been proposed. In 2010, Ho
et al. [14] measured thermophysical properties of Al2O3ewater
nanouid, experimentally. The size of nanoparticles was 33 nm and
ultra-pure Milli-Q water was used as the base uid. The properties
were measured for different temperatures as well as volume fractions of nanoparticles. Khanafer and Vafai [33] developed a correlation for the density of Al2O3ewater nanouid using the
experimental data of Ho et al. [14] as a function of temperature
ranged between 5 and 40
C and volume fraction of nanoparticles
between 0 and 0.04, as:

rnf 1001:064 2738:61914  0:2095T

(1)

bnf obtain as:

1 vrnf
0:2095T0

bnf 
rnf vT
rnf

(2)

The specic heat of nanouid can be determined by assuming

thermal equilibrium between the nanoparticles and the base uid
[33] as:








rcp nf 1  4 rcp f 4 rcp p

(3)

Experimental data show that classical models such as those of

Maxwell [34] and Hamilton-Crosser [35] for predicting thermal
conductivity and Einstein [36,37], Brinkman [38], and Batchelor
[39] for predicting the viscosity of nanouids do not lead to accurate results [40]. These models include only the effect of the
nanoparticle concentration and don't consider important mechanisms of heat transfer such as Brownian motion and don't employ
temperature and size of nanoparticles. The heat transfer characteristics of nano mechanisms of uids are dependent on effective
viscosity.
Using regression analysis and based on different valid experimental data, Corcione [40], proposed the following empirical correlation for thermal conductivity with a standard deviation of less
than 1.86%:

knf
T
1 4:4Re0:4 Pr 0:66
Tfr
kf

!10

kp
kf

!0:03
40:66

(4)

In Eq. (3) Pr and Re are:

Pr

mf
2r k T
; Re f 2 B
rf af
pm dp

(5)

Corcione model for thermal conductivity has been proposed for

nanouids consisting of alumina, copper oxide, titanium and copper nanoparticles with diameter in the range between 10 and
150 nm suspended in water or ethylene glycol (EG) with volume
fraction in the range from 0.002 to 0.09 and temperature in the
range between 294 and 324 K.
Corcione model for viscosity is:

mnf

mf

1
. 0:3
41:03
1  34:87 dp df

(6)

where df is the equivalent diameter of a base uid molecule [40].

Corcione model for viscosity was proposed for nanouids consisting of alumina, titanium, silica oxides and copper nanoparticles
with diameter ranging between 25 and 200 nm, suspended in
water, ethylene glycol (EG), propylene glycol (PG) or ethanol (Eth)
with nanoparticle volume fraction in the range from 0.0001 to
0.071 and temperature in the range between 293 and 333 K. When
water is used as the base uid its viscosity as a function of temperature is:

mf 562:77lnT 62:7568:9137

(7)

The thermo-physical properties of the base uid and Al2O3

nanoparticles at 295 K are presented in Table 1 [41]. In Fig. 1, the
viscosity and thermal conductivity of Al2O3ewater predicted by
Corcione models have been compared with the experimental data
of Ho et al. [14] and predictions of classical models. Good agreement is observed in Fig. 1a between Corcione models predictions
and experimental data of Ho et al., but the Maxwell model does not
predict the thermal conductivity of nanouid correctly, even at
room temperature. Fig. 1b shows that Brinkman model predicts the
viscosity correctly only for low volume fractions (up to 0.01).

Table 1
Thermophysical properties of base uid and nanoparticles at 295 K [34,35].
Physical properties

Water

Al2O3

Porous media

cp (J/kg K)
r(kg/m3)
K (W/m K)
dp  109 (m)

4179
997.8
0.59
0.384

765
3970
40
33

e
e
100
e

M. Dastmalchi et al. / International Journal of Thermal Sciences 95 (2015) 88e98

In this study the Corcione models [40] for conductivity and

viscosity functions of temperature, volume fraction and nanoparticle size and the Khanafer and Vafaei model [33] for density
function of temperature and volume fraction are used for Al2O3ewater nanouid.
The values for the permeability K in the momentum equations
for the porous layer are given by Ergun [42] for packed beds of
beads of diameter db and porosity

d2b 3
1751  2

(8)

In addition, models for the effective properties of the porous

medium are needed. It has been found that taking meff mnf provides good agreement with experimental data and is adopted in the
present work. Various models for the effective thermal conductivity keff have been proposed. In the present study, a nonlinear
equation is used, which is claimed to be universally applicable for a
medium with randomly distributed spherical inclusions [20].

keff

1
0
k

k
1=3
@ b 1=3 nf Akeff  kb 0
knf








rcp eff rcp nf 1  rcp s

(10)

3.1. Brownian motion

The random motion of nanoparticles within the base uid is
called Brownian motion which results from continuous collisions
between the nanoparticles and the molecules of the base uid.
Brownian motion in microscopic scale leads to macroscopic scale
diffusion ux. Nanoparticle ux in Brownian motion is dened as:

(11)

where the Brownian diffusion coefcient, DB, is given by the EinsteineStokes's equation [8] as:

DB

kB T
3pmf dp

The mechanism by which nanoparticles transport due to temperature gradient under the inuence of the thermophoretic force
is called thermophoresis [44]. Aitken [45] proved, in a series of
experiments on dusty air, that the particles must be driven away
from the heated surface by differential bombardment of the gas
molecules due to temperature gradient. The molecules moving
close to the hot surface carry greater kinetic energy than those
moving close to the cold surface, resulting in a net force on the
particles. The resulting net force is called thermophoretic. For the
nanoparticle diameter between 1 and 100 nm, the Knudsen number is relatively small, thus the continuum assumption is reasonable. The particles ux due to thermophoresis is:

Jp;T rp DT VT

(12)

(13)

where thermophoresis coefcient [8,10] is:

(9)

3. Transport mechanisms

Jp;B rp DB V4

3.2. Thermophoresis

DT ST

The effective heat capacity of the solideuid mixture, (rcp)eff, is

dened as [43].

91

mf
4
rf T

(14)

in which ST is thermophoresis parameter. For relatively large particles (dp  1 mm) it is reported to be [8,10]:

ST Cs

1
.
1 kp 2kf

(15)

Thermophoresis parameter depends only on thermal conductivity of both particles and base uid. Unfortunately, data for
thermophoresis parameter of nanouids are not available at this
time. In this study ST is obtained by a trial and error method using
experimental results of Ho et al. [14].
Nanoparticles ux can be written as combination of the effects
of Brownian motion and thermophoresis [8,10] as:

Jp Jp;B Jp;T rp DB V4  rp DT VT

(16)

4. Governing equations
A schematic view of porous square enclosure is shown in Fig. 2
(L H). The left wall is heated at a constant temperature (TH) and
the right wall is cooled at a constant temperature (TC). The horizontal walls are adiabatic.

Fig. 1. Al2O3ewater properties calculated using Corcione model compared with the experimental data of Ho et al. [6] and classical models. (a) conductivity and (b) viscosity.

92

M. Dastmalchi et al. / International Journal of Thermal Sciences 95 (2015) 88e98

The nanouid is treated as a two-component, continuous and

dilute Newtonian mixture which has variable physical properties.
The ow in the porous layer is modeled utilizing the Darcy equations. The compression work, dispersion and viscous dissipation are
assumed negligible in the energy equation and heat conduction is
represented by Fourier law. In developing the energy equation the
heat ux due to nanoparticles diffusion (Dufour effect) is ignored
[11,14]. Also order-of-magnitude estimates for the various terms of
the energy equation suggest that energy transfer by nanoparticle
dispersion is negligible, contrary to what is commonly stated in the
literature [7]. Homogeneity and local thermal equilibrium in the
porous medium is assumed and nanoparticles and porous media
are in thermal equilibrium with the base uid. There are not any
external force, heat source, chemical reaction and radiative heat
transfer in the problem. The ow is assumed to be slow so that an
advective term and a Forchheimer quadratic drag term do not
appear in the momentum equation.
According to the aforementioned assumptions, governing
equations including the momentum, the energy and the nanoparticle transport equations are the followings.
The continuity equation is:


 



1 vj v4 vj v4
v
v4
v
v4


DB

DB
vy vx vx vy
vx
vx
vy
vy



 
v
vT
v
vT
DT

DT

vx
vx
vy
vy

(21)

Eq. (21) states that nanoparticles can be transported in nanouid by convection (term on the left-hand side), but they also
possess a slip velocity relatively to the uid with a slip velocity
relative to the uid by Brownian diffusion (rst term on the righthand side) and also by thermophoresis (second term on the righthand side).
By considering no slip condition and zero ux of nanoparticles
(Jp n 0) at the solid walls, the boundary conditions for Eqs. 19e21
are:

j 0; T TH ;

v4
D vT
 T
vx
DB vx

at x 0 and 0  y  H

j 0; T TC ;

v4
D vT
 T
vx
DB vx

at x L and 0  y  H

j 0;

vT
v4
0;
0
vy
vy

at y 0; H and 0  x  L
(22)

vu vv

0
vx vy

(17)

Local convective heat transfer coefcient on either the left or

right wall is:

The momentum equations for Darcy ow through a porous

medium are [19]:

K vP
u
mnf vx


K vP
rnf g
v
mnf vx

h
(18)

Eliminating the pressure terms by cross differentiating Eqs (17)

and (18),





v
vj
v
vj
vr
mnf

mnf
Kg nf
vx
vx
vy
vy
vx

x0;L

(23)

DT

Average convective heat transfer coefcient is obtained by

integrating the local convective heat transfer coefcient along
these walls as:

1
L

ZL
hdy

(24)

(19)

The local and the mean Nusselt numbers, respectively, are:

Nu

The energy equation [18,30] is:








 vj vT vj vT
v
vT
v
vT


keff

keff
rcp nf
vy vx vx vy
vx
vx
vy
vy

knf vT
vx

(20)

The nanoparticle transport equation [8,26] in the homogenous

model is:

hH
hH
; Nu
kf
kf

(25)

To compare the effects of the Brownian and the thermophoresis

diffusivities, ratio of the Brownian diffusivity to the thermophoretic
diffusivity is dened as [8]:

NBT

KB rf 0
4b DB
T2

DT DT 3pST m2 dp DT
f

(26)

Rayleigh number is dened as:

Ra

gKbf 0 DTL
af 0 nf 0

(27)

The following dimensionless variables are used to present the

results.

x
X ;
L

y
Y ;
L

j
;
a f0

T  TC
;
DT

4
4b

(28)

5. Numerical procedure and validation

Fig. 2. Schematic of the porous enclosure.

The stream function equation and governing equations with the

associated boundary conditions are numerically solved using the

M. Dastmalchi et al. / International Journal of Thermal Sciences 95 (2015) 88e98

Table 2
Mean Nusselt number for different grid numbers
used in grid independence study (4b 0.02,
Ra 103, dbed 5 mm).
Grid size
81  81
91  91
101  101
111  111

Nu
15.87
16.11
16.21
16.23

nite difference and nite volume method [46] respectively. A nonuniform grid, with a very ne spacing near wall is tted to the
enclosure. High density of grids was provided near the walls in
order to resolve the volume fraction boundary layer properly. To
discretize the combined convection and diffusion ux in the governing equations, the power low scheme is used. The thermophoresis diffusion term in the nanoparticles transport has been
discretized using a second order central difference scheme. The
thermophysical properties such as thermal conductivity, viscosity,
density, thermophoresis diffusion and Brownian motion coefcients, which are variable with temperature and volume fraction
of nanoparticles, are solved concurrently with ow, temperature
and volume fraction in the whole solution domain. The effective
thermal conductivity is computed from Eq. (9) using the Newton's
method. On the control volume faces these properties are averaged
linearly using the calculated values on the grids. The coupled sets of
discretized equations have been solved iteratively using a line-byline procedure, combining the tri-diagonal matrix algorithm
(TDMA) [46]. Optimum value of the relaxation parameter was
chosen on the basis of computing experiments. To obtain
converged solutions, under-relaxation coefcient of 0.8 was used
for the stream function and the energy equations and an underrelaxation coefcient between 0.05 and 0.4 was used for the

93

nanoparticles transport equation. It should be noted that with

increasing volume fraction of nanoparticles a smaller value of
under-relaxation coefcient for transport equation was used.
5.1. Grid independence study
Numerical solutions obtained showed that close to the walls
small vortexes form and there is sharp volume fraction gradient,
thus the results showed sensitive to the grid numbers and expansion coefcients. After testing several grid expansion coefcients,
the value of 1.12 was used in horizontal direction and 1.07 was used
in vertical direction. The mean Nusselt number obtained using
different grid numbers is presented in Table 2. From Table 2 it is
seen that with change of grid numbers from 111111 to 121121
the change of local Nusselt number is not signicant, thus a grid
system of 111111 was used for all simulations.
5.2. Results validation for square enclosure
In order to validate the homogeneous model of this study, solutions were obtained for pure water at different Rayleigh numbers
for conguration and conditions of Ho et al. [14] who carried out an
experimental work in an enclosure with 25 mm width, 25 mm
height and 60 mm length for various bulk volume fractions of
alumina (Al2O3) nanoparticles ranging from 0 to 0.04 dispersed in
pure water. It should be noted that the bulk volume fraction refers
to overall volumetric percentage of nanoparticles added to the base
uid. In Fig. 3, the predicted mean Nusselt numbers by the homogeneous model show good agreement with the experimental results of Ho et al. for 4b 0.0.
Based on the previous discussions in Section 2, incorporation of
homogeneous model is not suitable for nanouid. To validate the
non-homogeneous model one needs to calibrate thermophoresis

Fig. 3. The mean Nusselt number obtained using non-homogeneous and homogeneous models compared with the experimental results of Ho et al. [6].

94

M. Dastmalchi et al. / International Journal of Thermal Sciences 95 (2015) 88e98

!
Table 3
Variations of the average Nusselt number khH for a square cavity lled with a
eff
porous medium in comparison with results from Sheremet and Pop [28].
Authors

Ra

Sheremet and Pop [28]

Present results
Error %

10

100

1000

10000

1.079
0.98
9

3.115
3.587
15

13.667
14.598
7

48.823
49.661
2

parameter rst. This has been achieved by using the experimental

results of Ho et al. [14]. For the thermophoresis parameter of
nanouids, ST, values have not ever been reported in the literature.
However, for large size particles it is only function of base uid and
nanoparticles thermal conductivities as presented by Refs.
[16,20,21] shown by Eq. (15). In this study similar to [20,21] and as a
constant coefcient, it has been calibrated using experimental data
of Ho et al. [14]. To nd ST, numerical results were obtained
considering different values for ST and then the ratio of nanouid
average heat transfer coefcient to that of the base uid were
compared with the same ratio for experimental results of Ho et al.
[14]. By considering the minimal relative difference between the
numerical results and experimental results, the optimal value of ST
was found 0.036 [15]. To validate the non-homogeneous model,
results have been obtained considering different temperature differences and bulk volume fractions similar to experimental work of
Ho et al. [14]. Fig. 3 presents mean Nusselt numbers predicted by
the non-homogeneous as well as the homogeneous models

compared with the experimental results of Ho et al. It is seen that

for all of the cases studied a better agreement exist between the
non-homogeneous model predictions and the experimental results
of Ho et al.
5.3. Results validation for porous square enclosure
The present models, in the form of a computational uid dynamics (CFD) code, have been validated successfully against the
works of Sheremet and Pop [29] for steady-state natural convection
in a square porous cavity. Table 3 shows the values of the Nueff
computed for various Rayleigh numbers in the range 10e104 in
comparison with Sheremet and Pop [29].
6. Results and discussion
In this section the effects of temperature difference between the
hot and cold walls, porosity, bulk volume fractions and nanoparticles size on ow eld, isotherms, local and mean Nusselt
number and mass boundary layer are investigated. Since in this
study variable properties are considered for nanouid, simulations
have been performed for various actual physical conditions such as
DT, 4b and porosity.
6.1. Effects of temperature difference
Fig. 4 presents mean Nusselt numbers predicted by the nonhomogeneous as well as the homogeneous models. It means that

Fig. 4. The mean Nusselt number obtained using non-homogeneous and homogeneous models; 0.5; db 5 mm; dp 33 nm.

M. Dastmalchi et al. / International Journal of Thermal Sciences 95 (2015) 88e98

Fig. 5. Ratio of non-homogeneous model convection heat transfer coefcient to that of

water; 0.5; db 5 mm; dp 33 nm.

the homogeneous model always predicts higher heat transfer coefcient which becomes even higher at larger temperature differences and volume fractions.
For a better quantitative comparison, the obtained ratio of
convective heat transfer coefcient using non-homogeneous model
to that of water versus temperature difference is presented in Fig. 5
for different bulk volume fractions. It is seen that this ratio is always
less than one and decreases with increasing bulk volume fraction of


nanoparticles or temperature difference from 1 C to 5
C and then
remains constant.
The local Nusselt number on the left and right walls obtained
using non-homogeneous and homogeneous models are presented
in Fig. 6 for temperature difference of 1 and 20
C and 4b 0.03. It is
observed that the homogeneous model predicts higher Nusselt
number than non-homogeneous model and the difference is signicant on the right wall close to the bottom wall and on the left
wall close to the top wall. Also, the difference is enhanced as the
temperature difference increases.
The hydrodynamic ow and thermal elds inside the square
porous enclosure are demonstrated by means of streamlines and

95

isotherms in Fig. 7 for different DT and 4b 0.03; 0.5;

db 5 mm; dp 33 nm as representative cases. As expected, with
increasing temperature difference the temperature gradient at the
corners decreases compared to the homogeneous model; thus, the
local Nu number decreases at the corners. Comparing the values of
Jmax for transport model in Fig. 7 at different temperature differences, it is evident that the strength of vortexes increases with
increasing temperature difference. The ow exhibits a circulating
ow pattern rising along the heater wall, gets blocked at the
top adiabatic wall and travels toward the cold outer wall. Then, the
ow moves down along the cold wall before occupying the
entire enclosure with the center of rotation at the middle of the
enclosure.
The constant volume fraction lines (F1) are shown in Fig. 8 for
4b 0.03 and temperature differences of 1 and 20
C. It is seen that
a thin mass boundary layer exists close to the vertical hot wall
which grows by moving from bottom toward top and on the top
wall it grows by moving from left toward right. Also, on the cold
wall another mass boundary layer exists which grows by moving
from the top toward bottom and on bottom wall it grows by moving
from right toward left. The mass boundary layers close to the top
and bottom walls are thinner than those close to the hot and cold
walls. Increasing the temperature difference decreases the mass
boundary layer thickness. It should be noted that the top and
bottom walls are insulated; hence, the volume fraction gradient
close to these walls is not due to temperature gradient but owes to
the effect of convection on volume fraction distribution. The
nanoparticles migrate owing to Brownian diffusion caused by
movement of nanouid close to the horizontal walls causing the
growth of mass boundary layer.
The volume fraction of nanoparticles on the hot and cold walls
for two temperature differences of 1 and 20
C are shown in Fig. 9.
With increasing temperature difference, the volume fraction on the
hot wall increases but it decreases on the cold wall. By moving from
the bottom toward the top the volume fraction increases slightly on
the hot wall, but on the cold wall the volume fraction decreases
rst, then increases and nally decreases again. Therefore, except
for low temperature differences, the volume fraction on the cold
and hot walls is not constant.
6.2. Effects of porosity
In this study, the effect of porosity on heat transfer has been
investigated. In Fig. 10, the values of mean Nusselt for nonhomogeneous models are presented for ve porosity of 0.1e0.5,

Fig. 6. Local Nusselt number on the left and right walls for transport and homogeneous models; 4b 0.03, 0.5; db 5 mm; dp 33 (solid lines for DT 1
C and dashed lines for
DT 20
C).

96

M. Dastmalchi et al. / International Journal of Thermal Sciences 95 (2015) 88e98

Fig. 7. Streamlines and isotherms for 4b 0.03; 0.5; db 5 mm; dp 33 nm; and different DT (solid lines for transport model and dashed lines for homogeneous model; the J
values are for transport model).

Fig. 9. Normalized volume fraction on the left hot and right cold walls for 4b 0.03,
0.5; db 5 mm; dp 33 nm and DT 20
C (solid line) and 1
C (dashed line).
Fig. 8. Constant normalized volume fraction of F1 for 4b 0.03, 0.5; db 5 mm;
dp 33 nm and DT 1
C (solid line) and 20
C (dashed line).

4b 0.03 and temperature difference of 10
C. By increasing the

porosity from 0.1 to 0.5 Nu number decreases almost 80%.
In Fig. 11, the streamlines, isotherms and normalized volume
fraction on the left hot and right cold walls for non-homogeneous
model with two porosity have been compared. As can be seen, at
0.1, transfer of heat from the discrete heaters is mainly
controlled by the conduction-dominated mechanism due to porous
drag, which is also reected through the parallel isotherm pattern
as shown in Fig. 11b. At 0.5, the intensity of the convection
becomes stronger, which implies that the convection heat transfer
begins to dominate the thermal ow eld in the enclosure.

7. Conclusions
A two-component non-homogeneous model of Buongiorno [8]
was employed to analyze the ow, heat and mass transfer of
nanouid in a porous square enclosure numerically. This proposed
non-homogeneous model has been calibrated using experimental

Fig. 10. The mean Nusselt number obtained using non-homogeneous models;
4b 0.03; db 5 mm; dp 33 nm and DT 10
C.

M. Dastmalchi et al. / International Journal of Thermal Sciences 95 (2015) 88e98

97

Fig. 11. (a) Streamlines (b) Isotherms for transport model (c) Normalized volume fraction on the left hot and right cold walls; 4b 0.03, dp 33 nm, 0.5 (solid line) and 0.1
(dashed line).

data of Ho et al. [14 and 15]. The non-homogeneous model predictions are in better agreement with experimental results.
1) The homogeneous model predicts higher heat transfer coefcients compared to non-homogeneous model which become
even higher at larger temperature differences and bulk volume
fractions.
2) The obtained local Nusselt number using non-homogeneous
model is less than that of homogeneous model and the difference is signicant at the cavity corners. With rising temperature
difference, the difference becomes more.
3) A thin mass boundary layer forms close to all of the cavity walls.
The mass boundary layer thickness close to the hot and cold
walls decreases with increased temperature difference but this
layer thickens close to the horizontal walls.
4) By increasing the porosity Nu number decreases almost 80%. At
0.1, transfer of heat from the discrete heaters is mainly
controlled by the conduction-dominated mechanism due to
porous drag. At higher porosity, the intensity of the convection
becomes stronger, which implies that the convection heat
transfer begins to dominate the thermal ow eld in the
enclosure.
For more validation of these results further experimental and
especially some ow visualizations are required to demonstrate the
real results in natural convection of nanouids.
Acknowledgment
The authors wish to thank the Energy Research Institute of the
University of Kashan (grant no. 65473) for their support regarding
this research.
References
[1] S. Lee, S.U.S. Choi, S. Li, J.A. Eastman, Measuring thermal conductivity of uids
containing oxide nanoparticles, ASME Trans. J. Heat Transf. 121 (1999)
280e289.
[2] J.A. Eastman, S.U.S. Choi, W. Li, S. Yu, L.J. Thompson, Anomalously increased
effective thermal conductivities of ethylene glycol-based nanouids containing copper nanoparticles, J. Appl. Phys. Lett. 78 (2001) 718e720.
[3] A. Mojtabi, M.C. Charrier-Mojtabi, Double-diffusive convection in porous
media, in: Kambiz Vafai (Ed.), Handbook of Porous Media, Dekker, New York,
2000.
[4] E. Abu-Nada, Z. Masoud, H.F. Oztop, A. Campo, Effect of nanouid variable
properties on natural convection in enclosures, Int. J. Therm. Sci. 49 (2010)
479e491.
[5] R.F. Probstein, Physicochemical Hydrodynamics, second ed., Wiley Interscience, Hoboken, New Jersey, 2003.

[6] J. Tyndall, On dust and disease, Proc. R. Inst. 6 (1870) 1e14.

[7] R.B. Bird, W.E. Stewart, E.N. Lightfoot, Transport Phenomena, second ed.,
Wiley, New York, 1960.
[8] J. Buongiorno, Convective transports in nanouids, ASME Trans. J. Heat Transf.
128 (2006) 240e250.
[9] A.V. Kuznetsov, D.A. Nield, Natural convective boundary-layer ow of a
nanouid past a vertical plate, Int. J. Therm. Sci. 49 (2010) 243e247.
[10] H.A. Pakravan, M. Yaghoubi, Combined thermophoresis, Brownian motion and
Dufour effects on natural convection of nanouids, Int. J. Therm. Sci. 50 (2011)
394e402.
[11] H. Aminfar, M.R. Haghgoo, Brownian motion and thermophoresis effects on
natural convection of aluminaewater nanouid, J. Mech. Eng. Sci. 6 (2012)
1e11.
[12] H.A. Pakravan, M. Yaghoubi, Analysis of nanoparticles migration on natural
convective heat, 2011, Int. J. Therm. Sci. 68 (June 2013) 79e93.
[13] Z. Haddad, E. Abu-Nada, H.F. Oztop, A. Mataoui, Natural convection
in nanouids: are the thermophoresis and Brownian motion effects signicant in nanouid heat transfer enhancement? Int. J. Therm. Sci. 57 (2012)
1e11.
[14] C.J. Ho, W.K. Liu, Y.S. Chang, C.C. Lin, Natural convection heat transfer of
Alumina-water nanouid in vertical square enclosures: an experimental
study, Int. J. Therm. Sci. 49 (2010) 1345e1353.
[15] G.A. Sheikhzadeh, M. Dastmalchi, H. Khorasanizadeh, Effects of nanoparticles
transport mechanisms on Al2O3-water nanouid natural convection in a
square enclosure, Int. J. Therm. Sci. 66 (2013) 51e62.
[16] G.A. Sheikhzadeh, M. Dastmalchi, H. Khorasanizadeh, Effects of walls temperature variation on double diffusive natural convection of Al2O3ewater
nanouid in an enclosure, Heat Mass Transf. 49 (2013) 1689e1700.
[17] P. Cheng, W.J. Minkowycz, Free convection about a vertical at plate
embedded in a porous medium with application to heat transfer from a dike,
J. Geophys. Res. 82 (1977) 2040e2044.
[18] A. Bijan, Convection Heat Transfer, third ed., Wily, NewYork, 1984.
[19] O.V. Trevisan, A. Bejan, Natural convection with combined heat and mass
transfer buoyancy effects in a porous medium, Inl. J. Hear Mass Transf. 28
(1985) 1597e1611.
[20] C. Beckermann, S. Ramadhyani, R. Viskanta, Natural convection ow and heat
transfer between a uid layer and a. Porous layer inside a rectangular
enclosure, J. Heat Transf. 109 (1987) 363e370.
[21] Tanmay Basak, S. Roy, T. Paul, I. Pop, Natural convection in a square cavity
lled with a porous medium: effects of various thermal boundary conditions,
Int. J. Heat Mass Transf. 49 (2006) 1430e1441.
[22] D.A. Nield, A.V. Kuznetsov, The ChengeMinkowycz problem for natural
convective boundary layer ow in a porous medium saturated by a nanouid,
Int. J. Heat Mass Transf. 52 (2009) 5792e5795.
[23] W.A. Khan, A. Aziz, Double-diffusive natural convective boundary layer ow
in a porous medium saturated with a nanouid over a vertical plate: prescribed surface heat, solute and nanoparticle uxes, Int. J. Therm. Sci. 50
(2011) 2154e2160.
[24] D.A. Nield, A.V. Kuznetsov, The ChengeMinkowycz problem for the doublediffusive natural convective boundary layer ow in a porous medium saturated by a nanouid, Int. J. Heat Mass Transf. 54 (2011) 374e378.
[25] Dhananjay Yadav, R. Bhargava, G.S. Agrawal, Boundary and internal heat
source effects on the onset of DarcyeBrinkman convection in a porous layer
saturated by nanouid, Int. J. Therm. Sci. 60 (2012) 244e254.
[26] A.V. Kuznetsov, D.A. Nield, The ChengeMinkowycz problem for natural
convective boundary layer ow in a porous medium saturated by a nanouid:
a revised model, Int. J. Heat Mass Transf. 65 (2013) 682e685.
[27] L. Tham, R. Nazar, I. Pop, Mixed convection ow from a horizontal circular
cylinder embedded in a porous medium lled by a nanouid: Buongiornoe
Darcy model, Int. J. Therm. Sci. 84 (2014) 21e33.

98

M. Dastmalchi et al. / International Journal of Thermal Sciences 95 (2015) 88e98

[28] M.A. Sheremet, T. Grosan, I. Pop, Free convection in shallow and slender
porous cavities lled by a nanouid using Buongiorno's model, ASME J. Heat
Transf. 136 (2014) 082501.
[29] M.A. Sheremet, I. Pop, Conjugate natural convection in a square porous cavity
lled by a nanouid using Buongiorno's mathematical model, Int. J. Heat Mass
Transf. 79 (2014) 137e145.
[30] M.A. Sheremet, I. Pop, M.M. Rahman, Three-dimensional natural convection in
a porous enclosure lled with a nanouid using Buongiorno's mathematical
model, Int. J. Heat Mass Transf. 82 (2015) 396e405.
[31] M. Hajipour, A. Molaei Dehkordi, Mixed-convection ow of Al2O3eH2O
nanouid in a channel partially lled with porous metal foam: experimental
and numerical study, Exp. Therm. Fluid Sci. 53 (2013) 49e56.
[32] M. Nazari, M. Ashouri, M.H. Kayhani, A. Tamayol, Experimental study of
convective heat transfer of a nanouid through a pipe lled with metal foam,
Int. J. Therm. Sci. 88 (2015) 33e39.
[33] K. Khanafer, K. Vafai, A critical synthesis of thermophysical characteristics of
nanouids, Int. J. Heat Mass Transf. 54 (2011) 4410e4428.
[34] J.C. Maxwell, A Treatise on Electricity and Magnetism, third ed., Dover, New
York, 1954.
[35] R.L. Hamilton, O.K. Crosser, Thermal conductivity of heterogeneous two
component systems, Ind. Eng. Chem. Fundam. 1 (1962) 187e191.
[36] A. Einstein, Eine neue bestimmung der molekul-dimension (A new determination of the molecular dimensions), Ann. Phys. 19 (1906) 289e306.

[37] A. Einstein, Berichtigung zu meiner arbeit: Eine neue bestimmung der

molekul-dimension (Correction of my work: a new determination of the
molecular dimensions), Ann. Phys. 34 (1911) 591e592.
[38] H.C. Brinkman, The viscosity of concentrated suspensions and solutions,
J. Chem. Phys. 20 (1952) 571.
[39] G. Batchelor, The effect of Brownian motion on the bulk stress in a suspension
of spherical particles, J. Fluid Mech. 83 (1977) 97e117.
[40] M. Corcione, Empirical correlating equations for predicting the effective
thermal conductivity and dynamic viscosity of nanouids, Energy Convers.
Manage 52 (2011) 789e793.
[41] Z. Alloui, P. Vasseur, M. Reggio, Natural convection of nanouids in a shallow
cavity heated from below, Int. J. Therm. Sci. 50 (2010) 1e9.
[42] S. Ergun, Fluid ow through packed columns, Chem. Eng. Prog. 48 (1952)
89e94.
[43] K. Vafai, Handbook of Porous Media, second ed., Taylor & Francis, New York,
2005.
[44] F. Zheng, Thermophoresis of spherical and non-spherical particles: a review of
theories and experiments, Adv. Colloid Interface Sci. 97 (2002) 255e278.
[45] J. Aitken, On the formation of small clear spaces in dusty air, R. Soc. Edinb. 32
(1884) 239e272.
[46] S.V. Patankar, Numerical Heat Transfer and Fluid Flow, second ed., Hemisphere, McGraw-Hill, Washington DC, 1980.