You are on page 1of 12

Heat Mass Transfer

DOI 10.1007/s00231-013-1209-9

ORIGINAL

Effects of walls temperature variation on double diffusive natural


convection of Al2O3–water nanofluid in an enclosure
G. A. Sheikhzadeh • M. Dastmalchi •

H. Khorasanizadeh

Received: 7 September 2012 / Accepted: 26 July 2013


Ó Springer-Verlag Berlin Heidelberg 2013

Abstract The effect of wall temperature variations on JP Particle flux vector, kg m-2 s-1
double diffusive natural convection of Al2O3–water nano- k Thermal conductivity, W m-1 K-1
fluid in a differentially heated square enclosure with con- kB Boltzmann constant, J K-1
stant temperature hot and cold vertical walls is studied L Length, m
numerically. Transport mechanisms of nanoparticles n Normal vector
including Brownian diffusion and thermophoresis that NBT Ratio of Brownian and thermophoretic diffusivities
cause heterogeneity are considered in non-homogeneous Nu Nusselt number
model. The hot and cold wall temperatures are varied, but p Pressure, Pa
the temperature difference between them is always main- Pr Prandtl number
tained 5 °C. The thermophysical properties such as thermal R Gas constant, J kg-1 K-1
conductivity, viscosity and density and thermophoresis Ra Rayleigh number
diffusion and Brownian motion coefficients are considered Re Reynolds number
variable with temperature and volume fraction of nano- ST Thermophoresis parameter
particles. The governing equations are discretized using the T Temperature, K
control volume method. The results show that nanoparticle u,v Dimensional x and y-components of velocity, m s-1
transport mechanisms affect buoyancy force and cause U,V Dimensionless velocities
formation of small vortexes near the top and bottom walls x, y Dimensional coordinates, m
of the cavity and reduce the heat transfer. By increasing the X,Y Dimensionless coordinates
temperature of the walls the effect of transport mechanisms
Greek symbols
decreases and due to enhanced convection the heat transfer
a Fluid thermal diffusivity, m2 s-1
rate increases.
b Thermal expansion coefficient, K-1
u Nanoparticle volume fraction
List of symbols
A Normalized nanoparticle volumetric fraction
cp Specific heat capacity, J kg-1 K-1
l Dynamic viscosity, N m s-1
Cs Non-continuum constant
q Nanofluid density, kg m-3
d Diameter, m
– Dimensionless temperature
O
DB Brownian coefficient, m2 s-1
m Kinematic viscosity, m2 s-1
DT Thermophoresis coefficient, m2 s-1 K-1
w Dimensional stream function, kg m-1 s-1
g Gravitational acceleration, m s-2
W Dimensionless stream function
h Local heat transfer coefficient, W m-2 K-1
Subscripts
0 Reference
G. A. Sheikhzadeh (&)  M. Dastmalchi  H. Khorasanizadeh
b Bulk or overall
Department of Mechanical Engineering, University of Kashan,
87317-51167 Kashan, Iran B Brownian
e-mail: sheikhz@kashanu.ac.ir C Cold

123
Heat Mass Transfer

D Dufour in the non-homogeneous model nanoparticles transport,


f Fluid due to small relative velocity between nanoparticles and
fr Freezing point of the base liquid the base fluid, is assumed. This in turn may lead to non-
H Hot, homogeneous homogeneity in nanofluid. In the dispersion model, the
nf Nanofluid relative velocity between nanoparticles and the base fluid is
p Particle treated as perturbation in the energy equation, which
T Thermophoresis enhances turbulence and increases the convective heat
transfer.
Superscripts Many researchers have used homogeneous model [7–
-
Average 12]. Khanafer et al. [7] studied Cu–water nanofluid con-
vection in a two-dimensional rectangular enclosure. They
reported an augmentation in heat transfer with increasing
1 Introduction volume fraction of nanoparticles at any given Grashof
number. Jou and Tzdng [8] and Oztop and Abu-Nada [9]
Ultrahigh cooling performance is one of the most vital showed similar results. Abu-Nada et al. [10] studied
needs in many industrial technologies. However, inherently natural convection in horizontal annuli and showed that
low thermal conductivity is a primary limitation in devel- heat transfer is enhanced by using nanofluids. In another
oping energy-efficient heat transfer fluids. Nanofluids are study, Abu-Nada et al. [11] investigated natural convec-
engineered by suspending nanoparticles with average sizes tion of variable properties Al2O3–water nanofluid in an
below 100 nm in traditional heat transfer fluids such as enclosure. They reported decrease in average Nusselt
water, oil, and ethylene glycol. Several researches [1–23] number at high Rayleigh numbers and increase in that at
have indicated that by adding nanoparticle with low vol- low Rayleigh numbers, as volume fraction of Al2O3
ume fraction (1–5 %), the thermal conductivity can be nanoparticles increased. Sheikhzadeh et al. [12] investi-
increased by about 20 %. Although the higher conductivity gated the buoyancy-driven fluid flow and heat transfer of
is encouraging, it is by no means conclusive evidence of Cu–water nanofluid in a square cavity with partially
the cooling capabilities of such fluids. For that, it is nec- active side walls, numerically. They reported that heat
essary to have definitive proof of the performance of these transfer rate increases with increasing volume fraction of
fluids in convective environments. nanoparticles.
Many experimental studies have been performed in the There are several mechanisms such as Brownian motion
field of nanofluid natural convection. Putra et al. [4] studied and thermophoresis nanoparticles transport in suspensions.
natural convection of Al2O3–water and CuO–water in a Brownian motion, named after Robert Brown, is the pre-
horizontal cylinder, experimentally. They found that the sumably random drifting of particles suspended in a fluid
Nusselt number decreases by increasing volume fraction of [13]. Temperature gradient can cause mass flux by a pro-
nanoparticles. Wen and Ding [5] investigated natural cess called either thermophoresis or thermal diffusion and
convection of TiO2–water between two disks, experimen- the Soret effect. This was first reported by Tyndall [14] in
tally, and obtained the same result as of Putra et al. [4]. Ho 1870. Concentration gradients can produce heat transfer
et al. [6] investigated natural convection of Al2O3–water known as diffusion-thermo, or Dufour effect, which is
nanofluid in square enclosures of three different sizes, usually small and negligible [15]. Buongiorno [16] intro-
experimentally. They measured the thermophysical prop- duced seven transport mechanisms which cause relative
erties of the nanofluid as well. They explained the unusual velocity between nanoparticles and fluid. By comparing the
increase or decrease of heat transfer cannot be explained diffusion time scale of transport mechanisms, he showed
solely based on relative changes in thermophysical prop- that the Brownian motion and thermophoresis are the two
erties of the nanofluid, and other factors such as nanopar- most important mechanisms. He theoretically investigated
ticle transport mechanisms which change the homogeneity the unusual Nusselt number increase for force convection
of the volume fraction of nanoparticles in the domain are in a duct and related this heat transfer enhancement to
also important. reduction of viscosity due to nanoparticle transport within
Numerous analytical and numerical models have been the boundary layer. He assumed that energy transfer by
presented for nanofluids convective heat transfer, which nanoparticle dispersion was negligible; hence he didn’t
can be divided into homogeneous, non-homogeneous and study Dufour effect. Hwang et al. [17] investigated theo-
dispersion models. In the homogeneous model, nanofluid is retically and experimentally laminar force convective heat
assumed as a usual fluid; thus, all traditional equations of transfer characteristics of Al2O3–water nanofluids with low
mass, momentum and energy are used considering effec- volume fractions in a circular tube. Based on their exper-
tive constant properties of the nanofluid. On the other hand, imental results, the Darcy friction factor of nanofluid had a

123
Heat Mass Transfer

good agreement with theoretical results for the single phase to play an important role in the convection process. Mok-
flow model. Based on scale analysis, they showed that the meli and Saffar-Avval [26] used the dispersion model to
flattened velocity profile due to particle transport is a investigate the laminar convective heat transfer of nanofl-
possible mechanism of the convective heat transfer uids in a tube. They compared the results obtained from the
enhancement, which cannot be explained by increase in dispersion model with their experimental results and
thermal conductivity of nanofluids alone. However, with- observed good agreement.
out physical explanation, they didn’t consider the effect of Natural convection in pure fluids is driven only by
convection in nanoparticle transport equation. Kuznetsov density variations due to temperature gradient. However,
and Nield [18] studied natural convective boundary-layer Double-diffusive convection in nanofluids is an important
flow of a nanofluid past a vertical plate, analytically. By fluid dynamics topic describing a form of convection dri-
using similarity solution, they showed that the Brownian ven by two different density gradients with different rates
motion and thermophoresis decreases the Nusselt number. of diffusion [27]. These density variations may be caused
They assumed the volume fraction of nanoparticles at wall by gradients in the volume fraction of the nanofluid or by
is constant, whereas physically zero particle flux at the wall temperature gradient. Nithyadevi and Yang [28] numeri-
will happen. This unrealistic boundary condition was also cally studied the effect of double-diffusive natural con-
used in a similar geometry by Rana and Bhargava [19]. vection in a partially heated enclosure considering
Pakravan and Yaghoubi [20] investigated effects of thermophoresis and Dufour effects. They observed that
Brownian motion, thermophoresis and Dufour in natural increase or decrease of heat and mass transfer is affected by
convection heat transfer, analytically. They showed that formation of bi-cellular structure.
Dufour effect reduces the total heat transfer compared with Based on reviewing the results obtained using different
homogenous model predictions. They reported good models and comparison with experimental results, it is
agreement for Nusselt number by comparing their results clear that homogeneous model is not competent to predict
with various experimental results. However they didn’t the heat transfer features of nanofluids. Thus, it is neces-
study some effects such as double diffusive natural con- sary to use non-homogeneous model by which double
vection and change of properties due to presence of volume diffusivity affects natural convection characteristics. In this
fraction gradient. Aminfar and Haghgoo [21] investigated numerical study, to consider transport mechanisms such as
the effects of Brownian motion and thermophoresis on Brownian motion and thermophoresis in natural convection
natural convection heat transfer of alumina–water nano- of variable properties Al2O3–water nanofluid in a square
fluid in a square vertical cavity using the ‘two-component enclosure, the Buongiorno’s two-component non-homoge-
four equation non-homogeneous equilibrium’ model, neous model has been employed. Since, the viscosity and
numerically. They concluded that the use of single phase thermal conductivity of nanofluid and the thermophoresis
homogeneous method does not seem reasonable for mod- and Brownian coefficients are temperature dependent, in
eling this class of natural convection. Haddad et al. [22] this study the effects of temperature change of the cold and
studied numerically, Cuo–water nanofluid Rayleigh- hot walls on the flow and heat transfer characteristics of
Bénard convection considering the role of Brownian and nanofluid are investigated and the results are physically
thermophoresis effects and made comparison with the case clarified.
where both effects were neglected. They noticed higher
heat transfer when Brownian and thermophoresis effects
were considered. Sheikhzadeh et al. [23] studied combined 2 Nanofluid properties
effects of Brownian motion, thermophoresis and Dufour on
laminar natural convective heat transfer of variable prop- Many studies have evaluated the nanofluids properties
erties Al2O3–water nanofluid in a square enclosure depending on both properties of base fluid and nanoparti-
numerically. They showed Dufour effect can be ignored cles and according to them, several models have been
and their predictions were in better agreement with proposed. In 2010, Ho et al. [6] measured thermophysical
experimental results of Ho et al. [6]. properties of Al2O3–water nanofluid, experimentally. The
The dispersion model was first proposed for nanofluids size of nanoparticles was 33 nm and ultra-pure Milli-Q
by Xuan and Roetzel [24]. As an overall modeling strategy, water was used as the base fluid. The properties were
they introduced an empirical dispersion coefficient to measured for different temperatures as well as volume
describe the heat transfer enhancement. However, for fractions of nanoparticles. Khanafer and Vafai [29]
identifying the mechanisms that are probably responsible developed a correlation for the density of Al2O3–water
behind the dispersive behaviors more complicated experi- nanofluid using the experimental data of Ho et al. [6] as a
ments are needed. Das et al. [25] state that among different function of temperature ranged between 5 and 40 °C and
mechanisms in dispersion model, particle transport seems volume fraction of nanoparticles between 0 and 0.04, as:

123
Heat Mass Transfer

qnf ¼ 1001:064 þ 2738:6191u  0:2095T ð1Þ and 200 nm, suspended in water, ethylene glycol (EG),
propylene glycol (PG) or ethanol (Eth) with nanoparticle
The specific heat of nanofluid can be determined by volume fraction in the range from 0.0001 to 0.071 and
assuming thermal equilibrium between the nanoparticles temperature in the range between 293 and 333 K. When
and the base fluid [29] as: water is used as the base fluid its viscosity as a function of
     
qcp nf ¼ ð1  uÞ qcp f þu qcp p ð2Þ temperature is:
lf ¼ 562:77ðlnðT þ 62:756ÞÞ8:9137 ð6Þ
Experimental data show that classical models such as those
of Maxwell [30] and Hamilton-Crosser [31] for predicting The thermo-physical properties of the base fluid and Al2O3
thermal conductivity and Einstein [32, 33], Brinkman [34], nanoparticles at 295 K are presented in Table 1 [37]. In
and Batchelor [35] for predicting the viscosity of nanofl- Fig. 1, the viscosity and thermal conductivity of Al2O3–
uids do not lead to accurate results [36]. These models water predicted by Corcione models have been compared
include only the effect of the nanoparticle concentration with the experimental data of Ho et al. [6] and predictions
and don’t consider important mechanisms of heat transfer of classical models. Good agreement is observed in Fig. 1a
such as Brownian motion and don’t employ temperature between Corcione models predictions and experimental
and size of nanoparticles. The heat transfer characteristics data of Ho et al., but the Maxwell model does not predict
of nano mechanisms of fluids are dependent on effective the thermal conductivity of nanofluid correctly, even at
viscosity. Enhancement of natural convection heat transfer room temperature. Figure 1b shows that Brinkman model
reported in references [7–12] is due to the fact that in these predicts the viscosity correctly only for low volume frac-
references the classical models for the viscosity have been tions (up to 0.01).
incorporated; for instance the Brinkman model predicts In this study the Corcione models [36] for conductivity
lower values for the nanofluid viscosity. Abu-Nada et al. and viscosity functions of temperature, volume fraction and
[11] studied the effects of using variable properties on heat nanoparticle size and the Khanafer and Vafaei model [29]
transfer characteristics and showed the weakness of the for density function of temperature and volume fraction are
classical models. used for Al2O3–water nanofluid.
Using regression analysis and based on different valid
experimental data, Corcione [36], proposed the following
empirical correlation for thermal conductivity with a 3 Transport mechanisms
standard deviation of less than 1.86 %:
 10  0:03 3.1 Brownian motion
knf T kp
¼ 1 þ 4:4Re0:4 Pr 0:66 u0:66 ð3Þ
kf Tfr kf
The random motion of nanoparticles within the base fluid is
In Eq. (3) Pr and Re are: called Brownian motion which results from continuous
lf 2q kB T collisions between the nanoparticles and the molecules of
Pr ¼ ; Re ¼ f 2 ð4Þ the base fluid. Brownian motion in microscopic scale leads
qf af plf dp
to macroscopic scale diffusion flux. Nanoparticle flux in
Corcione model for thermal conductivity has been pro- Brownian motion is defined as:
posed for nanofluids consisting of alumina, copper oxide, Jp;B ¼ qp DB ru ð7Þ
titanium and copper nanoparticles with diameter in the
range between 10 and 150 nm suspended in water or eth- where the Brownian diffusion coefficient, DB, is given by
ylene glycol (EG) with volume fraction in the range from the Einstein-Stokes’s equation [16] as:
0.002 to 0.09 and temperature in the range between 294
and 324 K.
Corcione model for viscosity is: Table 1 Thermo-physical properties of base fluid and nanoparticles
[34, 35]
lnf 1
¼  0:3 ð5Þ Physical properties Water Al2O3
lf 1  34:87 dp =df u1:03
cp (J/kg K) 4179 765
where df is the equivalent diameter of a base fluid molecule q (kg/m3) 997.8 3970
[11, 36].
K (W/m K) 0.59 40
Corcione model for viscosity was proposed for nanofl- 4
b 9 10 (1/K) 2.3 0.85
uids consisting of alumina, titanium, silica oxides and
dp 9 109 (m) 0.384 33
copper nanoparticles with diameter ranging between 25

123
Heat Mass Transfer

a 0.74 Jp;T ¼ qp DT rT ð9Þ


Maxwell Model [27]
0.72 Corcione Model [3 33]
0.02 Ho et al. [6]
where thermophoresis coefficient [16, 20] is:
0.7 0.04 Ho et al. [6]
ϕ=0.04 l
D T ¼ ST f u ð10Þ
0.68 qf T
knf (W/m.K)

0.66 in which ST is thermophoresis parameter. For relatively


0.64 large particles (dp C 1 lm) it is reported to be [16, 20]:
0.62 1
ST ¼ Cs ð11Þ
0.6
1 þ kp =2kf

0.58 ϕ=0.02 Thermophoresis parameter depends only on thermal con-


ductivity of both particles and base fluid. Unfortunately, data
0.56
295 30
00 305 310 for thermophoresis parameter of nanofluids are not available
T (K) at this time. In this study ST is obtained by a trial and error
2.4
method using experimental results of Ho et al. [6].
b 0.01 Brinkman [3 31] Nanoparticles flux can be written as combination of the
0.04
2.2 Corcione Model [333] effects of Brownian motion and thermophoresis [16, 20] as:
0
2 0.001 Jp ¼ Jp;B þ Jp;T ¼ qp DB ru  qp DT rT ð12Þ
0.003
1.8 0.01 6]
Ho et al. [6
0.02
μ (cp)

0.03
1.6 0.04 4 Governing equations
1.4
A schematic view of the differentially heated square
1.2
enclosure is shown in Fig. 2 (L = H). The left wall is
1 heated at a constant temperature (TH) and the right wall is
0.8 cooled at a constant temperature (TC). The horizontal walls
are adiabatic.
0.6 The nanofluid is treated as a two-component, continuous
280 290
2 300 310
T (K) and dilute Newtonian mixture which has variable physical
properties. The compression work, dispersion and viscous
Fig. 1 Al2O3–water properties calculated using Corcione model
compared with the experimental data of Ho et al. [6] and classical
dissipation are assumed negligible in the energy equation
models. a thermal conductivity and b viscosity and heat conduction is represented by Fourier law. Fur-
thermore, nanoparticles are in thermal equilibrium with the
base fluid and there are not any external force, heat source,
kB T chemical reaction and radiative heat transfer in the
DB ¼ ð8Þ
3plf dp problem.
3.2 Thermophoresis

The mechanism by which nanoparticles transport due to


temperature gradient under the influence of the thermoph-
oretic force is called thermophoresis [38]. Aitken [39]
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 1–100 nm, the
Knudsen number is relatively small, thus the continuum
assumption is reasonable. The particles flux due to ther-
mophoresis is: Fig. 2 Schematic of the enclosure

123
Heat Mass Transfer

There are two approaches to investigate the effects of thermophoresis (second term on the right-hand side).By
particle transport in suspensions. The first approach is considering no slip condition and zero flux of nanoparticles
based on particle mass conservation and the other is based (Jp.n = 0) at the solid walls, the boundary conditions for
on the momentum balance. The second approach is for Eqs. (13)–(17) are:
two phase mixture model in which the nanofluid is con- ou DT oT
sidered to be a single fluid with two phases, but each v ¼ u ¼ 0; T ¼ TH ; ¼ at x ¼ 0
ox DB ox
phase has its own velocity and within any computational
and 0yH
cell there exists a certain volume fraction of each phase.
However, the strength of diffusivity in general and the ou DT oT
v ¼ u ¼ 0; T ¼ TC ; ¼ at x ¼ L
subsequent slip velocity for nanoparticles is not as much ox DB ox ð18Þ
to create noticeable momentum exchange. Hence, in this and 0yH
study a two-component four-equation non-homogeneous oT ou
equilibrium model of Buongiorno [16] that is based on v ¼ u ¼ 0; ¼ 0; ¼ 0 at y ¼ 0; H
oy oy
particle mass conservation for mass, momentum and and 0  x  L
energy is used.
According to the aforementioned assumptions, govern- After determining the flow and temperature fields, stream
ing equations including the continuity, the momentum, the function and Nusselt number can be calculated. The stream
energy and the nanoparticle transport equations are the function value is calculated as:
followings.
ZH
The continuity equation is [40]: ow
qnf u ¼  ) wðx; yÞ ¼  qnf udy þ w0 ð19Þ
oðqnf uÞ oðqnf vÞ oy
þ ¼0 ð13Þ 0
ox oy
Local convective heat transfer coefficient on either the left
The momentum equation [40] is: or right wall is:
     
ou ou o ou 2 ou ov knf oT 
ox x¼0;L
qnf u þ v ¼ 2l  lnf þ h¼ ð20Þ
ox oy ox  nf ox  3 ox
 oy DT
o ou ov op
þ l þ  ð14Þ Average convective heat transfer coefficient is obtained by
oy nf oy ox ox
     integrating the local convective heat transfer coefficient
ov ov o ov 2 ou ov along these walls as:
qnf u þ v ¼ 2l  lnf þ
ox oy oy  nfoy 3   ox oy ZH
o ou ov op 1
þ lnf þ   qnf g h ¼ hdy ð21Þ
ox oy ox oy L
0
ð15Þ
The local and the mean Nusselt numbers, respectively, are:
The energy equation [40] is:
      hH 
hH
  oT oT o oT o oT Nu ¼ ; Nu ¼ ð22Þ
qcp nf u þv ¼ knf þ knf kf kf
ox oy ox ox oy oy
ð16Þ To compare the effects of the Brownian and the
thermophoresis diffusivities, ratio of the Brownian
The nanoparticle transport equation [16] is: diffusivity to the thermophoretic diffusivity is defined as
     [16]:
ou ou o ou o ou
u þv ¼ DB þ DB ub DB KB qf T 2
ox oy ox  ox  oy  oy  NBT ¼ ¼ ð23Þ
o oT o oT DT DT 3pST l2f dp DT
þ DT þ DT ð17Þ
ox ox oy oy
Rayleigh number is defined as:
Equation (17) states that nanoparticles can be transported gbf 0 DTL3
in nanofluid by convection (term on the left-hand side), but Ra ¼ ð24Þ
af 0 m f 0
they also possess a slip velocity relatively to the fluid with
a slip velocity relative to the fluid by Brownian diffusion The following dimensionless variables are used to present
(first term on the right-hand side) and also by the results.

123
Heat Mass Transfer

x y w u 1.2 9 10-3 and 5.7 9 10-4, respectively. The local Nus-


X ¼ ; Y ¼ ; W¼ ; U¼ ; selt number on the right wall obtained using different grid
L L qf 0 af 0 af 0 =L
v T  TC u numbers (N 9 N) for a particular case is presented in
V¼ ; h¼ ; U¼ ð25Þ Fig. 3. From Fig. 3 it is seen that with change of grid
af 0 =L DT ub
numbers from 201 9 201 to 211 9 211 the change of local
Nusselt number is not significant, thus a grid system of
5 Numerical procedure and validation 201 9 201 was used for all simulations.

The governing equations with the associated boundary 5.2 Results validation
conditions are numerically solved using the finite volume
method [41]. To discretize the combined convection and In order to validate the homogeneous model of this study,
diffusion flux in the governing equations, various schemes solutions were obtained for pure water at different Ray-
including the upwind, hybrid, power low and QUICK were leigh numbers for configuration and conditions of Ho et al.
tested and the power low scheme seemed suitable [42], [6] who carried out an experimental work in an enclosure
hence used. The thermophoresis diffusion term, the second with 25 mm width, 25 mm height and 60 mm length for
term on the right hand side of the Eq. (17), has been dis- various bulk volume fractions of alumina (Al2O3) nano-
cretized using a second order central difference scheme. particles ranging from 0 to 0.04 dispersed in pure water. It
The thermophoresis term in the nanoparticles transport should be noted that the bulk volume fraction refers to
equation and other added terms due to variation of prop- overall volumetric percentage of nanoparticles added to the
erties with temperature in the momentum equation, have base fluid. In Fig. 4, the predicted mean Nusselt numbers
been added to the corresponding source terms. The ther- by the homogeneous model show good agreement with the
mophysical properties such as thermal conductivity, vis- experimental results of Ho et al. for ub ¼ 0:0.
cosity and density and thermophoresis diffusion and Based on the previous discussions in Sect. 2, incorpo-
Brownian motion coefficients, which are variable with ration of homogeneous model is not suitable for nanofluid.
temperature and volume fraction of nanoparticles, are To validate the non-homogeneous model one needs to
solved concurrently with flow, temperature and volume calibrate thermophoresis parameter first. This has been
fraction in the whole solution domain. On the control achieved by using the experimental results of Ho et al. [6].
volume faces these properties are averaged linearly using For the thermophoresis parameter of nanofluids, ST, values
the calculated values on the grids. A staggered grid system have not ever been reported in the literature. However, for
together with the SIMPLER algorithm has been adopted to large size particles it is only function of base fluid and
solve for the pressure and the velocity components. The nanoparticles thermal conductivities as presented by [16,
coupled sets of discretized equations have been solved 20, 21] shown by Eq. (11). In this study similar to [20, 21]
iteratively using a line-by-line procedure, combining the and as a constant coefficient, it has been calibrated using
tri-diagonal matrix algorithm (TDMA) [41]. To obtain experimental data of Ho et al. [6]. To find ST, numerical
converged solutions, under-relaxation coefficient of 0.9
was used for the momentum and the energy equations and a
under-relaxation coefficient between 0.05 to 0.4 was used
16 N=181
for the nanoparticle transport equation. It should be noted
that with increasing volume fraction of nanoparticles a N=191
14
N=201
smaller value of under-relaxation coefficient for transport N=211
12
equation was used.
10
5.1 Grid independence study
Nu

Numerical solutions obtained showed that close to the 6


walls small vortexes form and there is sharp volume frac-
4
tion gradient, thus the results showed sensitive to the grid
numbers and expansion coefficients. After testing several 2
grid expansion coefficients, the value of 1.12 was used in
0
horizontal direction and 1.05 was used in vertical direction. 0 0.2 0.4 0.6 0.8 1
Y
Smallest grid steps near the walls in the X and Y directions
are 1.8 9 10-8 and 4.8 9 10-6, respectively. The largest Fig. 3 Local Nusselt number on the right wall in grid independence
ones at the cavity center in the X and Y directions are study (ub = 0.02 and DT = 5 K)

123
Heat Mass Transfer

12.5 13 13
12 Ho et al. Ho et al.
12 Non-homogenous 12 Non-homogenous
11.5 Homogenous Homogenous
11 11 11
10.5
| 10 10
Nu

10

Nu

Nu
| |
| | |
9.5 9 9
9
8 8
8.5 Homogenous
8 Ho et al.
7 7
7.5
1E+06 2E+06 3E+06 6 6
2 4 6 8 10 2 4 6 8 10
Ra
ΔΤ (Κ) ΔΤ (Κ)
ϕb = 0
ϕb = 0.01 ϕb = 0.02

13 13
Ho et al. Ho et al.
12 Non-homogenous 12 Non-homogenous
Homogenous Homogenous
11 11

10
10
Nu

Nu
| |
|
9
9

8
8

7
7

6
2 4 6 8 10 6
2 4 6 8 10
ΔΤ (Κ)
ΔΤ (Κ)
ϕb = 0.03 ϕb = 0.04

Fig. 4 The mean Nusselt number obtained using non-homogeneous and homogeneous models compared with the experimental results of Ho
et al. [6]

results were obtained considering different values for ST cold wall was increased from 295 to 320 K while the
and then the ratio of nanofluid average heat transfer coef- temperature difference between the walls was kept 5 K and
ficient to that of the base fluid were compared with the volume fraction was kept 0.02. Figure 5 shows the varia-
same ratio for experimental results of Ho et al. [6]. By tion of average Nusselt number with change of the cold
considering the minimal relative difference between the wall temperature for homogenous model as well as non-
numerical results and experimental results, the optimal homogeneous model for volume fraction of 0.03. As seen,
value of ST was found 0.036 [42]. with increasing the cold wall temperature the average
To validate the non-homogeneous model, results have Nusselt number increases for pure water as well as nano-
been obtained considering different temperature differ- fluid. By increasing the cold wall temperature from 295 to
ences and bulk volume fractions similar to experimental 320 K this increase for pure water is 21 % but for nanofluid
work of Ho et al. [6]. Figure 4 presents mean Nusselt using homogenous model is 14 % and using non-homo-
numbers predicted by the non-homogeneous as well as the geneous model is 21 %. This increase is due to enhanced
homogeneous models compared with the experimental Rayleigh number with increasing the cold wall temperature
results of Ho et al. It is seen that for all of the cases studied and so convection strength. In the specified temperature
a better agreement exist between the non-homogeneous range the mean Rayleigh number for pure water and
model predictions and the experimental results of Ho et al. nanofluid increases 183 and 165 %, respectively, as
observed in Fig. 6.
Variation of the local Nusselt number on the left and
6 Results and discussion right walls for TC = 295 and 320 K is shown in Fig. 7 for
ub =0.02 for homogenous as well as non-homogeneous
In order to study the effect of changing walls temperatures, model. It is seen that the local Nusselt number for non-
a case study was performed in which the temperature of the homogeneous model is less than homogenous model

123
Heat Mass Transfer

12 everywhere; however the difference between them is more


on the upper and lower corners of the right wall and the
11.5
upper corner of the left wall. Also it is observed that with
11 increasing the cold wall temperature this difference
decreases.
10.5
In Fig. 8 stream lines are shown for different values of
|
the cold wall temperature. For the non-homogeneous
Nu

| 10
|
model and lower cold wall temperatures small weak vor-
9.5
texes are observed close to the upper wall at the right
9 corner as well as the bottom left corner and the distance
Water
Hemogenous Model between the streamlines is more at the corners.
8.5 Non-homogenous Model As observed from isotherms in Fig. 9, at lower cold wall
8 temperature and close to the corners of the cavity the
295 300 305 310 315 320
temperature gradient is less for non-homogeneous model
TC(k)
compared with homogenous model but the difference
Fig. 5 Mean Nusselt number verses the cold wall temperature for become less as the cold wall temperature rises.
pure fluid and nanofluid with ub = 0.02 for homogenous and non- In a detailed study about the effect of change of prop-
homogeneous models erties due to volume fraction gradient in non-homogeneous
model on heat transfer it has been shown that the most
effective parameter is change of density [42]. Variable
density affects the buoyancy and advection terms in the
4E+06

Raf momentum and energy equations. The concept of double


Ranf diffusive natural convection was mentioned in introduc-
tion. In order to study the importance of this concept and
the effects of variation of density on buoyancy force the
3E+06

constant density lines are shown in Fig. 10 for non-


homogeneous as well as homogenous models. As observed
Ra

the constant density regions in the cavity particularly at the


2E+06

corners have different form for non-homogeneous model


compared with homogenous model. This changes the
buoyancy force and is the cause of formation of secondary
flow at the upper right as well as the lower left corners as
1E+06

noticed in Fig. 8. In other words, the flow is driven by


295 300 305 310 315 320 opposite thermal and Solutal buoyancies. This secondary
T(K)
flow is the cause of reduction of heat and mass transfer in
Fig. 6 Nanofluid and base fluid mean Ra number verses the cold wall non-homogeneous model compared with homogenous
temperature model as observed in Fig. 7. In fact, the transport

Fig. 7 Local Nusselt number Left wall Right wall


on the left and right wall for 1 1
non-homogeneous and Non-homogeneous
homogenous models Homogeneous
0.8 0.8
(ub = 0.02) TC=295 K
TC=320 K

0.6 0.6 TC=320 K


Y
Y

0.4 0.4
TC=295 K

0.2 0.2 Non-homogeneous


Homogeneous

0 0
0 5 10 15 20 0 5 10 15 20
Nu Nu

123
Heat Mass Transfer

Fig. 8 Streamlines for


ub = 0.03 and different cold
wall temperatures (solid lines
for non-homogeneous model
and dashed lines for
homogeneous model)

Fig. 9 Isotherms for ub = 0.02


and different cold wall
temperatures (solid lines non-
homogeneous model and
dashed lines homogeneous
model)

0.8

0.6
Left wall Right wall
Y

0.4

0.2 TC=320 K
TC=295 K

0
Fig. 10 Constant density lines ub = 0.02 (solid lines non-homoge- 0.8 1 1.2 1.4 1.6
Φ
neous model and dashed lines homogeneous model)
Fig. 11 Normalized volume fraction on cold and hot wall;
mechanisms such as thermophoresis and Brownian affect ub = 0.02, DT = 5 K, Tc = 295 K (solid line) and (dashed line)
the opposite buoyancy force which leads to reduced heat
transfer. non-homogeneous and homogenous models decreases. In
With increasing the cold wall temperature and based on this study the difference in heat transfer was 13 % for
Eq. (23) the ratio of the Brownian diffusivity to the ther- TC = 295 K which decreased to 4 % for TC = 320 K.
mophoresis diffusivity increases and as shown in Fig. 11
the volume fraction gradient between the cold and hot
walls reduces. This in turn decreases the effect of the 7 Conclusions
Brownian diffusivity. At the other hand, increasing the cold
wall temperature reduces the Brownian and thermophoresis A two-component non-homogeneous model of Buongiorno
diffusivity effects. As a result, the secondary flow weakens [16] was employed to analyze the flow, heat and mass
and the small vortexes close to the top and bottom transfer of nanofluid in a square enclosure numerically.
walls vanish and the heat transfer difference between the This proposed non-homogeneous model has been

123
Heat Mass Transfer

calibrated using experimental data of Ho et al. [6]. To study 10. Abu-Nada A, Masoud Z, Hijazi A (2008) Natural convection heat
the effect of changing walls temperatures, the temperature transfer enhancement in horizontal concentric annuli using
nanofluids. Int Commun Heat Mass Transf 35:657–665
of the cold wall increased while the temperature difference 11. Abu-Nada E, Masoud Z, Oztop HF, Campo A (2010) Effect of
remained constant. By increasing the cold wall temperature nanofluid variable properties on natural convection in enclosures.
the average Nusselt number increased for pure water as Int J Therm Sci 49:479–491
well as nanofluid. This is due to enhanced Rayleigh number 12. Sheikhzadeh GA, Arefmanesh A, Kheirkhah MH, Abdollahi R
(2011) Natural convection of Cu–water nanofluid in a cavity with
and so convection strength. The local Nu number for non- partially active side walls. Eur J Mech B Fluids 30:166–176
homogeneous model was less than homogenous model 13. Probstein RF (2003) Physicochemical hydrodynamics, 2nd edn.
everywhere and the difference was more on the upper and Wiley Interscience, Hoboken, NJ
lower corners of the right wall and the upper corner of the 14. Tyndall J (1870) On dust and disease. Proc R Inst 6:1–14
15. Bird RB, Stewart WE, Lightfoot EN (1960) Transport phenom-
left wall. Also as the cold wall temperature increases this ena, 2nd edn. Wiley, New York
difference decreased. As for the non-homogeneous model 16. Buongiorno J (2006) Convective transports in nanofluids. ASME
and at low cold wall temperatures small weak vortexes Trans J Heat Transf 128:240–250
were observed close to the upper wall at the right corner 17. Hwang KS, Jang SP, Choi SUS (2009) Flow and convective heat
transfer characteristics of water-based Al2O3 nanofluids in fully
and the distance between the streamlines at the corners was developed laminar flow regime. Int J Heat Mass Transf
more. Studying the constant density regions in the cavity 52:193–199
particularly at the corners revealed that the flow is driven 18. Kuznetsov AV, Nield DA (2010) Natural convective boundary-
by opposite buoyancy caused by gradients in the volume layer flow of a nanofluid past a vertical plate. Int J Therm Sci
49:243–247
fraction of the nanofluid or by temperature gradient. At the 19. Rana P, Bhargava R (2012) Flow and heat transfer of a nanofluid
other hand increased cold wall temperature enhanced over a nonlinearly stretching sheet: a numerical study. Commun
convection thus reduced the Brownian and thermophoresis Nonlinear Sci Numer Simul 17:212–226
diffusivity effects. As a result the secondary flow weak- 20. Pakravan HA, Yaghoubi M (2011) Combined thermophoresis,
Brownian motion and Dufour effects on natural convection of
ened, the small vortexes close to the top and bottom walls nanofluids. Int J Therm Sci 50:394–402
vanished and the heat transfer difference decreased. 21. Aminfar H, Haghgoo MR (2012) Brownian motion and thermo-
phoresis effects on natural convection of alumina–water nano-
Acknowledgments The authors wish to thank the Energy Research fluid. J Mech Eng Sci 6:1–11
Institute of the University of Kashan for their support regarding this 22. Haddad Z, Abu-Nada E, Oztop HF, Mataoui A (2012) Natural
research (Grant No. 65473). convection in nanofluids: are the thermophoresis and Brownian
motion effects significant in nanofluid heat transfer enhancement?
Int J Therm Sci 57:1–11
23. Sheikhzadeh GA, Dastmalchi M, Khorasanizadeh H (2013)
References Effects of nanoparticles transport mechanisms on Al2O3–water
nanofluid natural convection in a square enclosure. Int J Therm
1. Lee S, Choi SUS, Li S, Eastman JA (1999) Measuring thermal Sci 66:51–62
conductivity of fluids containing oxide nanoparticles. ASME 24. Xuan Y, Roetzel W (2000) Conceptions for heat transfer corre-
Trans J Heat Transf 121:280–289 lation of nanofluids. Int J Heat Mass Transf 43:3701–3707
2. Eastman JA, Choi SUS, Li W, Yu S, Thompson LJ (2001) 25. Das SK, Choi SUS, Yu W, Pradeep T (2008) Nanofluids science
Anomalously increased effective thermal conductivities of eth- and technology. Wiley, New Jersey
ylene glycol-based nanofluids containing copper nanoparticles. 26. Mokmeli A, Saffar-Avval M (2010) Prediction of nanofluid
J Appl Phys Lett 78:718–720 convective heat transfer using the dispersion model. Int J Therm
3. Xuan Y, Li Q (2000) Heat transfer enhancement of nanofluids. Int Sci 49:471–478
J Heat Fluid Flow 21:58–64 27. Mojtabi A, Charrier-Mojtabi MC (2000) Double-diffusive con-
4. Putra N, Roetzel W, Das SK (2003) Natural convection of nano- vection in porous media. In: Vafai K (ed) Handbook of porous
fluids. Heat Mass Transf 39:775–784 media. New York, Dekker
5. Wen D, Ding Y (2004) Experimental investigation into convec- 28. Nithyadevi N, Yang RJ (2009) Double diffusive natural con-
tive heat transfer of nanofluids at the entrance region under vection in a partially heated enclosure with Soret and Dufour
laminar flow conditions. Int J Heat Mass Transf 47:5181–5188 effects. Int J Heat Fluid Flow 30:902–910
6. Ho CJ, Liu WK, Chang YS, Lin CC (2010) Natural convection 29. Khanafer K, Vafai K (2011) A critical synthesis of thermo-
heat transfer of Alumina-water nanofluid in vertical square physical characteristics of nanofluids. Int J Heat Mass Transf
enclosures: an experimental study. Int J Therm Sci 54:4410–4428
49:1345–1353 30. Maxwell JC (1954) A treatise on electricity and magnetism, 3rd
7. Khanafer K, Vafai K, Lightstone M (2003) Buoyancy-driven heat edn. Dover, New York
transfer enhancement in a two dimensional enclosure utilizing 31. Hamilton RL, Crosser OK (1962) Thermal conductivity of het-
nanofluids. Int J Heat Mass Transf 46:3639–3653 erogeneous two component systems. Ind Eng Chem Fundam
8. Jou R, Tzdng S (2006) Numerical research of nature convective 1:187–191
heat transfer enhancement filled with nanofluids in rectangular 32. Einstein A (1906) Eine neue bestimmung der molekul-dimension
enclosures. Int Commun Heat Mass Transf 33:727–736 (A new determination of the molecular dimensions). Ann Phys
9. Oztop HF, Abu-Nada E (2008) Numerical study of natural con- 19:289–306
vection in partially heated rectangular enclosure filled with 33. Einstein A (1911) Berichtigung zu meiner arbeit: Eine neue
nanofluids. Int J Heat Fluid Flow 29:1326–1336 bestimmung der molekul-dimension (Correction of my work: a

123
Heat Mass Transfer

new determination of the molecular dimensions). Ann Phys 38. Zheng F (2002) Thermophoresis of spherical and non-spherical
34:591–592 particles: a review of theories and experiments. Adv Colloid
34. Brinkman HC (1952) The viscosity of concentrated suspensions Interface Sci 97:255–278
and solutions. J Chem Phys 20:571 39. Aitken J (1884) On the formation of small clear spaces in dusty
35. Batchelor G (1977) The effect of Brownian motion on the bulk air. R Soc Edinb 32:239–272
stress in a suspension of spherical particles. J Fluid Mech 40. Bijan A (1984) Convection heat transfer, 3rd edn. Wily, New
83:97–117 York
36. Corcione M (2011) Empirical correlating equations for predicting 41. Patankar SV (1980) Numerical heat transfer and fluid flow, 2nd
the effective thermal conductivity and dynamic viscosity of edn. Hemisphere, McGraw-Hill, Washington, DC
nanofluids. Energy Convers Manage 52:789–793 42. Dastmalchi M (2011) Numerical investigation of nanoparticles
37. Alloui Z, Vasseur P, Reggio M (2010) Natural convection of transport in natural convection of nanofluids with variable
nanofluids in a shallow cavity heated from below. Int J Therm Sci properties. MSc thesis, University of Kashan, Kashan, Iran
50:1–9

123

You might also like