International Journal of Thermal Sciences 49 (2010) 1536e1546

Contents lists available at ScienceDirect

International Journal of Thermal Sciences

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

Heat transfer features of buoyancy-driven nanofluids inside rectangular

enclosures differentially heated at the sidewalls
Massimo Corcione*
Dipartimento di Fisica Tecnica, Sapienza Università di Roma, via Eudossiana 18, 00184 Rome, Italy

a r t i c l e i n f o a b s t r a c t

Article history: The heat transfer features of buoyancy-driven nanofluids inside rectangular enclosures differentially
Received 15 January 2010 heated at the vertical walls, are investigated theoretically. The main idea upon which the present work is
Received in revised form based is that nanofluids behave more like a single-phase fluid rather than a conventional solideliquid
26 April 2010
mixture, which implies that all the convective heat transfer correlations available for single-phase flows
Accepted 3 May 2010
can be extended to nanoparticle suspensions, provided that the thermophysical properties appearing in
Available online 31 May 2010
them are the nanofluid effective properties calculated at the reference temperature. In this connection,
two empirical equations, based on a wide variety of experimental data reported in the literature, are
Keywords:
Nanofluids
developed for the evaluation of the nanofluid effective thermal conductivity and dynamic viscosity,
Natural convection whereas the other effective properties are evaluated by the conventional mixing theory. The heat transfer
Differentially heated enclosures enhancement across the differentially heated enclosure that derives from the dispersion of nano-sized
Theoretical analysis solid particles into a host liquid is calculated for different operating conditions, nanoparticle diameters,
Optimal particle loading combinations of suspended nanoparticles and base liquid, and cavity aspect ratios. The fundamental
result obtained is the existence of an optimal particle loading for maximum heat transfer. Specifically, for
any assigned combination of solid and liquid phases, the optimal volume fraction is found to increase
slightly with decreasing the nanoparticle size, and to increase much more remarkably with increasing
both the nanofluid average temperature and the slenderness of the enclosure.
Ó 2010 Elsevier Masson SAS. All rights reserved.

1. Introduction nano-sized solid particles, whose effective thermal conductivity is

significantly higher than that of the pure base liquid.
Buoyancy-induced heat transfer in enclosed spaces has been Most of the papers available in the literature on convective heat
extensively studied in the past, playing a fundamental role in transfer in nanofluids are related to forced convection flows,
a wide number of engineering applications. This is, e.g., the case of demonstrating that nanoparticle suspensions have unquestionably
the cooling of electronic assemblies, in which a large number of a great potential for heat transfer enhancement, as summarized in
high power dissipating components are often packaged in modular the review-articles recently compiled by Daungthongsuk and
enclosures such that space and external cooling sources are Wongwises [1], Murshed et al. [2], and Kakaç and Pramuanjaroenkij
minimal. Moreover, in several situations heat transfer designers [3]. On the contrary, the relatively few studies that concern buoy-
prefer to avoid the use of mechanical equipment for the coolant ancy-driven heat transfer in nanofluids lead to contradictory
circulation, due to power consumption, excessive operating noise, conclusions, thus leaving still unanswered the question if the use of
or reliability concerns. On the other hand, the low thermal nanoparticle suspensions for natural convection applications is
conductivity of conventional heat transfer fluids, such as water, actually advantageous with respect to pure liquids.
ethylene glycol, and mineral oils, generally results in a serious The typical example of conflicting results, frequently cited in the
limitation to the effectiveness of heat removal from the systems literature, is given by the first two papers dealing with natural
whose temperature control relies on natural convection. A possible convection in enclosures differentially heated at the vertical walls,
solution to this problem may be obtained by the replacement of which were published in 2003 by Khanafer et al. [4], and Putra et al.
traditional coolants with nanofluids, i.e., liquid suspensions of [5]. Specifically, in their two-dimensional numerical study, Kha-
nafer and co-workers showed that, owing to the dispersion of
copper nanoparticles into water, the amount of heat transferred
* Tel.: þ39 (0)6 44 58 54 43; fax: þ39 (0)6 48 80 120. across a square enclosure increased remarkably with increasing the
E-mail address: [email protected] nanoparticle volume fraction at any investigated Grashof number,

1290-0729/$ e see front matter Ó 2010 Elsevier Masson SAS. All rights reserved.
doi:10.1016/j.ijthermalsci.2010.05.005
 M. Corcione / International Journal of Thermal Sciences 49 (2010) 1536e1546 1537

Nomenclature Greek symbols

b coefficient of thermal expansion
cp specific heat at constant pressure 4 nanoparticle volume fraction
D Einstein diffusion coefficient m dynamic viscosity
df equivalent diameter of a base fluid molecule r mass density
dp diameter of the nanoparticle sD time required to cover a distance dp moving at velocity
E heat transfer enhancement uB
H height of the enclosure
h coefficient of convection heat transfer Subscripts
k thermal conductivity c cooled sidewall of the enclosure
kb Boltzmann’s constant ¼ 1.38066  1023 J/K eff effective property
L width of the enclosure f base fluid
M molecular weight of the base fluid fr freezing point of the base liquid
N Avogadro number ¼ 6.022  1023 mol1 H height of the enclosure used as characteristic length
Nu Nusselt number h heated sidewall of the enclosure
Pr Prandtl number max maximum value
Ra Rayleigh number opt optimal value
Re nanoparticle Reynolds number ref reference state for the calculation of the
T temperature thermophysical properties
uB mean Brownian velocity of the nanoparticle s solid nanoparticle

which was ascribed to the effect of the irregular and random viscosity plays a role as crucial as that played by the effective
motion of the suspended nanoparticles. In contrast, the experi- thermal conductivity in determining the heat transfer performance
mental data obtained by Putra and colleagues for a horizontal of nanofluids in natural convection flows. This means that the
cylinder differentially heated at ends, filled with Al2O3 þ H2O or results obtained through any numerical simulation procedure that
CuO þ H2O, revealed that the average Nusselt number of the does not rely on a robust viscosity model, capable to predict as
enclosure decreased with increasing the nanoparticle volume more accurately as possible the nanofluid rheological behaviour,
fraction, with a degree of deterioration depending on both the are unrealistic. Unfortunately, all the numerical studies cited above
nanoparticle material and the cavity aspect ratio, which could not are based on the Brinkman equation [12], whose predictions
be explained. However, the disagreement between these results is severely underestimate the effective dynamic viscosity of nano-
insubstantial, since Khanafer et al. based the Nusselt number on the fluids. As regards the measurements executed by Putra and
thermal conductivity of the base fluid, kf, whilst Putra et al. defined colleagues, they did not find any optimal particle loading simply
the Nusselt number using the effective thermal conductivity of the because their experiments were executed for only two nanoparticle
nanofluid, keff, which may bring to ambiguous interpretations of concentrations, i.e., 0.01 and 0.04.
the data. In fact, a Nusselt number that would describe the thermal The relevance of the nanofluid effective dynamic viscosity in
performance of the enclosure with immediacy should increase with enclosed natural flows was stressed also by Santra et al. [13], and
increasing the coefficient of convection of the enclosure, h, and vice Ho et al. [14], who showed that the effect of the nanoparticle
versa. This is what happens when the Nusselt number is defined volume fraction on the cavity Nusselt number depended signifi-
according to Khanafer et al., i.e., Nu ¼ hL/kf, since the ratio L/kf is cantly on the viscosity model adopted. In both studies, the use of
independent of the nanoparticle volume fraction. On the contrary, the Brinkman model resulted in an increase of the average Nusselt
when the Nusselt number is defined as Nu ¼ hL/keff, Nu may either number. Conversely, a more or less pronounced decrease of the
increase or decrease with increasing h, depending on whether average Nusselt number was observed when the nanofluid effective
vh/vkeff is positive or negative. In particular, it may occur that h dynamic viscosity was calculated through empirical correlations. In
increases with increasing the nanoparticle volume fraction, 4, thus detail, Santra used a correlation constructed by fitting the values of
meaning that the addition of nanoparticles enhances the thermal zero shear viscosity obtained experimentally by Kwak and Kim [15]
performance of the enclosure, but, if vh/vkeff < 0, the average Nus- for ethylene glycol suspensions of CuO nanoparticles having the
selt number decreases with increasing 4, which is exactly what shape of prolate-spheroids with an average diameter of 12 nm and
happens with the experimental data of Putra and colleagues. an aspect ratio of approximately 3. For his part, Ho used the
Actually, if the Nusselt numbers reported by Putra et al. are rear- correlation derived by Maïga et al. [16] on the basis of the
ranged in the same form used by Khanafer et al., that means measurements executed by Wang et al. [17] for water-based
properly multiplied by the ratio keff/kf, the results obtained by Putra nanofluids containing Al2O3 nanoparticles with a diameter of
et al. show the same trend of those derived by Khanafer and co- 28 nm. Interestingly, although the correlations used for the
workers. viscosity evaluation were based on experimental data, neither
Enhanced heat transfer in side-heated cavities, consequent to the Santra and co-workers, nor Ho and colleagues, found an optimal
addition of nanoparticles to water, was also reported in the numerical nanoparticle concentration, which deserves to be explained.
studies executed later by Jou and Tzeng [6], Tiwari and Das [7], Oztop First of all, it must be said that the study performed by Santra
and Abu-Nada [8], Abu-Nada and Oztop [9], and Ög üt [10]. et al. [13] was based on the assumption that the enclosure was
A completely different result was found experimentally by filled with a mixture of water and spherical copper nanoparticles
Nnanna [11], who discovered the existence of an optimal particle with a diameter of 100 nm, whose effective thermal conductivity
loading for maximum heat transfer across the enclosure, which was was evaluated through the model proposed by Patel et al. [18].
explained as the consequence of an excessive increase in viscosity On the other hand, as mentioned earlier, the effective dynamic
occurring above a certain nanoparticle concentration. Indeed, it viscosity was calculated by a correlation based upon the Kwak-
seems absolutely reasonable to suppose that the effective dynamic Kim experimental data relative to a liquid suspension of tiny
1538 M. Corcione / International Journal of Thermal Sciences 49 (2010) 1536e1546

rod-like nanoparticles, which has obviously brought to enclosure filled with a nanoparticle suspension is an issue still far
a remarkable overestimation of the viscosity effects. In fact, for from being completely solved, which can be mainly imputed to
the same nanoparticle concentration, the overall contact surface ambiguous evaluations of the nanofluid effective thermophysical
area between small rod-like nanoparticles and base fluid is properties. Framed on this general background, the aim of the
significantly wider than that existing between relatively large present paper is to undertake a comprehensive theoretical study on
spherical nanoparticles and base fluid, thus implying that the this topic, with the main scope to determine the existence of an
nanofluid containing rod-like nanoparticles is characterized by optimal particle loading for maximum heat transfer across the
a larger amount of friction occurring at the solid/liquid interface enclosure, for different operating conditions, nanoparticle diame-
and, correspondingly, a higher effective dynamic viscosity. Notice ters, solideliquid combinations, and cavity aspect ratios.
that the increase of the effective dynamic viscosity with
decreasing the diameter of the suspended nanoparticles, and 2. Theoretical formulation of the problem
then increasing the area of the solid/liquid contact surface, is
largely demonstrated by a number of experimental studies Although nanofluids are solideliquid mixtures, the approach
readily available in the literature e see, e.g., Prasher et al. [19] conventionally used in most studies handles the nanofluid as
and Chevalier et al. [20]. a single-phase fluid. In fact, since the suspended nanoparticles have
Regarding the work executed by Ho et al. [14], the nanofluid usually small size and concentration, the hypothesis of a solid-
effective thermal conductivity was evaluated through a correlation eliquid mixture statistically homogeneous and isotropic can
originally developed for slurries by Charuyakorn et al. [21], that, reasonably be advanced. This means that, under the further
ignoring any nanoscale phenomenon, necessarily gives under- assumptions that the nanoparticles and base fluid are in local
estimated values, with the consequence that the viscosity effects thermal equilibrium, and no slip motion occurs between the solid
prevail over the thermal conductivity effects. Circumstantially, and liquid phases, to all intents and purposes the nanofluid can be
another central point relative to natural convection flows in treated as a pure fluid. Thus, all the convective heat transfer
nanofluids has been raised, i.e., the fact that the trustworthiness of correlations available in the literature for single-phase flows can
the solutions of any numerical work on this subject relies also on easily be extended to the corresponding nanofluid applications,
the accuracy of the model adopted for predicting the nanofluid provided that the thermophysical properties appearing in them are
effective thermal conductivity. In this connection, it is worth the nanofluid effective properties calculated at the reference
noticing that in numerous studies the effective thermal conduc- temperature.
tivity is calculated by the Maxwell-Garnett model [22], that is For natural convection in rectangular enclosures differentially
known to fail dramatically in predicting the increased thermal heated at the vertical walls of height H, with perfectly insulated
conductivity of nanofluids, at least when the temperature of the bottom and top walls of width L, the BerkovskiePolevikov heat
suspension is higher than “room” temperature. transfer correlations, based on both experimental and numerical
Quite recently, Abu-Nada et al. [23] performed a numerical data, are usually recommended [28e30]:
study on side-heated cavities filled with Al2O3 (dp ¼ 47 nm) þ H2O

0:29
0:13
or CuO (dp ¼ 29 nm) þ H2O. The effective thermal conductivity was Pr H
NuH ¼ 0:18 Ra
evaluated through the empirical correlation proposed by Chon et al. 0:2 þ Pr H L
[24], whereas the effective dynamic viscosity was calculated by
a pair of correlations based upon the experimental data of Nguyen
3
H Pr H
et al. [25]. It was found that, for the convection dominated regime, 1  2; 103  Pr  105 ; 103  RaH ; (1)
the average Nusselt number decreased with increasing the nano- L 0:2 þ Pr L
particle volume fraction, with a degree of deterioration depending
on both the nanoparticle material and the cavity aspect ratio.
0:28
0:09
Pr H
However, also in this case, the results were meaningfully affected NuH ¼ 0:22 Ra
0:2 þ Pr H L
by an overestimation of the effective dynamic viscosity. In fact, the
dynamic viscosities measured by Nguyen and co-workers for Al2O3
H
(dp ¼ 47 nm) þ H2O were higher than those measured for Al2O3 2  10; Pr  105 ; RaH  1013 : (2)
(dp ¼ 36 nm) þ H2O, which is in sharp contrast with the main body
L
of the literature results (remind that most investigators have found As an example, the distribution of the Nusselt number that emerge
that the effective dynamic viscosity of nanofluids is inversely from the BerkovskiePolevikov correlation of eq. (1) for a water-
proportional to the size of the suspended nanoparticles). Moreover, filled square cavity, plotted against the Rayleigh number in the
as the data relative to dp ¼ 36 nm are in substantial good agreement range between 103 and 109, is displayed in Fig. 1, where the
with the results obtained by Chevalier et al. [20] for dp ¼ 35 nm, it numerical results recently obtained for a two-dimensional enclo-
follows that the data for Al2O3 (dp ¼ 47 nm) þ H2O tend to over- sure by Corcione and Habib [31] under the assumption of laminar
stimate the actual viscosity values. In addition, also for CuO flow and constant physical properties of the fluid, are also reported
(dp ¼ 29 nm) þ H2O, the Nguyen viscosity values are larger than for comparison.
those available in the literature for nanofluids containing nano- For aspect ratios larger than H/L ¼ 10, the heat transfer perfor-
particles of similar size, which is the case of the data reported by mance can be determined through the following correlation due to
Masuda et al. [26] for dp ¼ 27 nm, Pak and Cho [27] for dp ¼ 27 nm, MacGregor and Emery [32]:
and Wang et al. [17] for dp ¼ 28 nm. The reasons behind the over-
estimated values obtained by Nguyen and colleagues are difficult to
0:05
understand, although a possible explanation may be searched in H
NuH ¼ 0:42Pr0:012 Ra0:25
H
the use of an unknown surfactant which could have unusually L
affected the mechanical behavior of the suspensions prepared for
experiments.
3
H H
The above review of the existing literature shows that the 10   40; 1  Pr  2  104 ; 104  RaH  107 : (3)
problem of natural convection in a differentially heated vertical L L
 M. Corcione / International Journal of Thermal Sciences 49 (2010) 1536e1546 1539

the discrepancies among the data may even be of the order of 50%,
which could be due to the different measurement technique used
in experiments, as well as to the different degree of dispersion/
agglomeration obtained for the suspended nanoparticles, and the
accuracy of evaluation of their shape and size. Therefore, in
deriving the correlation proposed here, some literature data-sets
have been discarded because either the data were in a too sharp
contrast with the main body of the available results without any
convincing physical evidence, or the investigation procedure was
not properly described in detail, or specific chemical dispersants/
surfactants were used in experiments, which could have signifi-
cantly altered the thermo-mechanical behaviour of the suspension.
The experimental data upon which the correlation is based are
extracted from the following sources: Masuda et al. [26] for
TiO2(27 nm) þ H2O; Pak and Cho [27] for TiO2(27 nm) þ H2O; Lee
et al. [48] for CuO(23.6 nm) þ H2O, CuO(23.6 nm) þ ethylene
glycol (EG), Al2O3(38.4 nm) þ H2O, and Al2O3(38.4 nm) þ EG;
Fig. 1. BerkovskiePolevikov correlation of NuH vs. RaH for water-filled square cavities. Eastman et al. [49] for Cu(10 nm) þ EG; Das et al. [50] for CuO
(28.6 nm) þ H2O, and Al2O3(38.4 nm) þ H2O; Chon et al. [24] for
Al2O3(47 nm) þ H2O; Chon and Kihm [51] for Al2O3(47 nm and
The dimensionless heat transfer correlations listed above will be 150 nm) þ H2O; Xuan et al. [45] for TiO2(27 nm) þ H2O; Murshed
used to assess the effect of the nanoparticle volume fraction on the et al. [52] for Al2O3(80 nm) þ H2O, and Al2O3(80 nm) þ EG;
heat transfer enhancement, E, defined as Mintsa et al. [53] for CuO(29 nm) þ H2O; and Duangthongsuk
and Wongwises [54] for TiO2(21 nm) þ H2O. The nanoparticle
heff Nueff keff volume fraction and the nanofluid temperature lie in the ranges
E ¼ 1 ¼   1; (4)
hf Nu kf between 0.002 and 0.09, and between 294 K and 324 K,
respectively.
where hf, kf and Nu are the coefficient of convection, the thermal
By way of regression analysis, the following mean empirical
conductivity and the Nusselt number of the base fluid, respectively,
correlation with a 1.86% standard deviation of error is produced
and heff, keff and Nueff are the corresponding effective quantities of
(see Fig. 2):
the nanoparticle suspension. Recall that Nueff is the outcome of eqs.
(1)e(3), in which the Rayleigh and Prandtl numbers of the pure !10 !0:03
fluid are substituted by the effective Rayleigh and Prandtl numbers keff T ks
¼ 1 þ 4:4Re0:4 Pr0:66 40:66 ; (5)
of the nanofluid calculated at the reference average temperature kf Tfr kf
Tref ¼ (Th þ Tc)/2, where Th and Tc are the temperatures of the heated
and cooled sidewalls of the enclosure, respectively. where Re is the nanoparticle Reynolds number, Pr is the Prandtl
number of the base liquid, T is the nanofluid temperature, Tfr is the
freezing point of the base liquid, ks is the thermal conductivity of
3. Nanofluid effective thermophysical properties

3.1. Effective thermal conductivity

The Maxwell-Garnett model [22], and other traditional mean-

field theories, seem somehow suitable for predicting the effective
thermal conductivity of nanofluids at room temperature, as dis-
cussed in Buongiorno et al. [33], but become absolutely inadequate
when the nanofluid temperature is one or some tens degrees
higher than 20e25  C. This has motivated the development of
a number of new models that basically account for the effects of
the phenomena occurring at the solid/liquid interface and/or the
micro-mixing convection caused by the Brownian motion of the
suspended nanoparticles, which is, e.g., the case of the models
proposed by Yu and Choi [34], Xue [35], Kumar et al. [36], Koo and
Kleinstreuer [37], Jang and Choi [38,39], Xie et al. [40], Patel et al.
[18], Ren et al. [41], Prasher et al. [42,43], Leong et al. [44], Xuan
et al. [45], Prakash and Giannelis [46], and Murshed et al. [47].
However, these new models show large discrepancies among each
other, which clearly represents a restriction to their applicability.
Moreover, most of them include empirical constants of pro-
portionality whose values were often determined on the basis of
a limited number of experimental data, or were not clearly defined.
Hence, an empirical correlation based on a wide variety of
experimental data available in the literature has been developed for
keff/kf. In this regard, it is worth pointing out that a certain
dispersion of the experimental data reported by different authors Fig. 2. Comparison between eq. (5) and the experimental data of keff/kf extracted from
for the same type of nanofluid is unavoidable. Indeed, in some cases literature.
1540 M. Corcione / International Journal of Thermal Sciences 49 (2010) 1536e1546

the solid nanoparticles, and 4 is the nanoparticle volume fraction.

The Reynolds number of the suspended nanoparticles is defined as
Re ¼ (rf uB dp)/mf, where rf and mf are the mass density and dynamic
viscosity of the base fluid, respectively, and dp and uB are the
nanoparticle diameter and the nanoparticle Brownian velocity,
respectively. Once uB is calculated as the ratio between dp and the
time required to cover such a distance sD ¼ (dp)2/6D e see Keblinki
et al. [55] e, in which D stands for the Einstein diffusion coefficient,
the nanoparticle Reynolds number is given by

2rf kb T
Re ¼ ; (6)
pm2f dp
where kb ¼ 1.38066  1023 J/K is the Boltzmann’s constant. Notice
that in eqs. (5) and (6) all the physical properties are calculated at
the nanofluid temperature T.
It is apparent that, assigned the solid and liquid phases, the
dimensionless effective thermal conductivity of the nanofluid,
Fig. 4. Distributions of keff/kf vs. 4 for Al2O3 þ H2O, with dp and T as parameters.
keff/kf, increases as 4 and T increase, and dp decreases. The distri-
butions of keff/kf vs. T that emerge from eq. (5) for Al2O3 þ H2O, with
dp and 4 as parameters, are displayed in Fig. 3, where the corre-
sponding literature data are also reported. In Fig. 4, the distributions also to the dynamic viscosity. The experimental data used to
of keff/kf vs. 4 for Al2O3 þ H2O, with dp and T as parameters, are construct the correlation proposed here are taken out of the
compared with the prediction of the Maxwell equation, showing following sources: Masuda et al. [26] for TiO2(27 nm) þ H2O, Pak
that the degree of failure of the MaxwelleGarnett model applied to and Cho [27] for TiO2(27 nm) þ H2O, Wang et al. [17] for
nanofluids increases as the temperature increases and the nano- Al2O3(28 nm) þ H2O, Putra and co-workers [5,58] for
particle size decreases. Al2O3(38 nm) þ H2O, Prasher et al. [19] for Al2O3(27 nm, 40 nm,
and 50 nm) þ propylene glycol (PG), He et al. [59] for
TiO2(95 nm) þ H2O, Chen et al. [60,61] for TiO2(25 nm) þ ethylene
3.2. Effective dynamic viscosity glycol (EG), Chevalier et al. [20] for SiO2(35 nm, 94 nm, and
190 nm) þ ethanol, Lee et al. [62] for Al2O3(30 nm) þ H2O, and
Although the Brinkman equation [12], and other traditional Garg et al. [63] for Cu(200 nm) þ EG. The nanoparticle volume
theories, notoriously underpredict the effective dynamic viscosity fraction and the nanofluid temperature lie in the ranges between
of nanofluids, only few models have recently been proposed for 0.0001 and 0.071, and between 293 K and 323 K, respectively.
describing the rheological behaviour of nanofluids, such as those By way of regression analysis, the following mean empirical
developed by Koo [56], and Masoumi et al. [57]. However, as these correlation with a 1.84% standard deviation of error is obtained
models contain empirical correction factors based on an extremely (see Fig. 5):
small number of experimental data, their region of validity is
someway limited.
Therefore, an empirical correlation based on a large number of
experimental data selected from literature has been developed
for meff/mf, where meff is the effective dynamic viscosity of the
nanofluid, and mf is the dynamic viscosity of the base fluid. Of
course, the considerations stated earlier about the dispersion of
the thermal conductivity data available in the literature apply

Fig. 3. Correlation distributions of keff/kf vs. T for Al2O3 þ H2O, with dp and 4 as Fig. 5. Comparison between eq. (7) and the experimental data of meff/mf extracted from
parameters, and the corresponding experimental data extracted from literature. literature.
 M. Corcione / International Journal of Thermal Sciences 49 (2010) 1536e1546 1541

meff 1 and the nanoparticles, respectively. Accordingly, the effective

¼  0:3 ; (7)
mf specific heat at constant pressure of the nanofluid, (cP)eff, is calcu-
1  34:87 dp =df 41:03 lated as

where df is the equivalent diameter of a base fluid molecule,

 
1  4 ðrcP Þf þ4ðrcP Þs
given by ðcP Þeff ¼   : (11)
1  4 rf þ 4rs
" #1=3
6M The effective coefficient of thermal expansion of the nanofluid,
df ¼ ; (8)
N prf 0 beff, is defined by

in which M is the molecular weight of the base fluid, N is the dr

ðrbÞeff ¼  eff : (12)
Avogadro number, and rf0 is the mass density of the base fluid dT
calculated at temperature T0 ¼ 293 K. If we substitute eq. (9) into eq. (12), and replace the temperature
It may be observed that, once the base fluid is assigned, the derivatives of rf and rs by the products (rb)f and (rb)s for the base
dimensionless effective dynamic viscosity of the nanofluid, meff /mf, fluid and the nanoparticles, respectively, we have
increases as dp decreases and 4 increases. Another interesting feature
is that meff /mf is independent of temperature, at least for particle ðrbÞeff ¼ ð1  4ÞðrbÞf þ4ðrbÞs ; (13)
concentrations not too high and temperatures not too far from room
temperature (obviously, the evaluation of the effective dynamic thus obtaining
viscosity meff at a given temperature T requires that mf is calculated at  
1  4 ðrbÞf þ4ðrbÞs
such temperature T). Moreover, meff /mf is also independent of the beff ¼   : (14)
solideliquid combination. As an example, the distributions of meff /mf 1  4 rf þ 4rs
vs. 4 that emerge from eq. (7) for SiO2 þ ethanol, with dp as
As an example, the distributions of the dimensionless effective
a parameter, are displayed in Fig. 6, where the corresponding litera-
mass density, reff /rf, the dimensionless effective specific heat at
ture data are also reported. In the same figure, the prediction of the
constant pressure, (cp)eff /(cp)f, and the dimensionless effective
Brinkman equation is additionally delineated, pointing out that the
coefficient of thermal expansion, beff /bf, plotted against the volume
error deriving from its application to nanofluids increases remarkably
fraction for Al2O3 þ H2O, are shown in Fig. 7, where the distribu-
with decreasing the nanoparticle size.
tions of the dimensionless effective heat capacity per unit volume,
(rcp)eff /(rcp)f, and the dimensionless effective temperature deriv-
3.3. Other effective physical properties ative of the mass density, (rb)eff /(rb)f, are also represented.

The other effective physical properties of the nanofluid are

calculated according to the mixing theory, as typically done in the 4. Results and discussion
majority of the studies mentioned above.
The effective mass density of the nanofluid, reff, is given by 4.1. Effective Rayleigh and Prandtl numbers

reff ¼ ð1  4Þrf þ 4rs ; (9) The effective Rayleigh and Prandtl numbers, Raeff and Preff, that
must be substituted into eqs. (1)e(3) in order to obtain the effective
where rf and rs are the mass densities of the base fluid and the solid Nusselt number Nueff, are calculated by simply replacing the ther-
nanoparticles, respectively. mophysical properties of the pure fluid with the effective proper-
The heat capacity per unit volume of the nanofluid, (rcP)eff, is ties of the nanofluid, which gives:
ðrcP Þeff ¼ ð1  4ÞðrcP Þf þ4ðrcP Þs ; (10)  2  h   i
reff =rf beff =bfcp eff = cp f
where cp is the specific heat at constant pressure, and (rcp)f and Raeff ¼    Ra ¼ FðRaÞ  Ra (15)
(rcp)s are the heat capacities per unit volume of the base fluid
meff =mf keff =kf

Fig. 6. Correlation distributions of meff/mf vs. 4 for SiO2 þ ethanol, with dp as a param- Fig. 7. Distributions of the other dimensionless physical properties vs. 4 for
eter, and the corresponding experimental data extracted from literature. Al2O3 þ H2O at T ¼ 309 K.
1542 M. Corcione / International Journal of Thermal Sciences 49 (2010) 1536e1546

h    i 
cp eff = cp f meff =mf
Preff ¼   Pr ¼ FðPrÞ  Pr; (16)
keff =kf

where Ra and Pr are the Rayleigh and Prandtl numbers of the base
fluid, and F(Ra) and F(Pr) are the corresponding factorial functions
depending only on the relevant dimensionless effective properties
of the nanofluid.
In order to evaluate in what measure Raeff and Preff change with
increasing the nanoparticle concentration for different values of the
nanofluid average temperature and the nanoparticle diameter, the
distributions of F(Ra) and F(Pr) vs. 4 that emerge from eqs. (15) and
(16), in which the effective thermophysical properties are calculated
according to eqs. (5), (7), (9), (11) and (14), are displayed in Figs. 8
and 9 for Al2O3 þ H2O. In both figures, the results obtained by
using the MaxwelleGarnett and Brinkman models for the evalua-
tion of the effective thermal conductivity and dynamic viscosity,
respectively, are also reported for comparison. As expected, F(Ra)
Fig. 9. Distributions of F(Pr) vs. 4 for Al2O3 þ H2O, with dp and T as parameters.
decreases with increasing 4. In fact, the numerator of F(Ra), which is
equal to the dimensionless effective heat capacity per unit volume
multiplied by the dimensionless effective temperature derivative of In fact, as said earlier, the decrease of F(Ra) with 4 is determined
the mass density, is almost independent of the nanoparticle volume substantially by the growth of keff /kf and meff /mf. On the other hand,
fraction, since both (rcp)eff /(rcp)f and (rb)eff /(rb)f remain substan- according to eqs. (5) and (7), keff /kf depends very few on the
tially unaltered with increasing 4 (see Fig. 7). In contrast, the nanoparticle material, as denoted by the extremely small exponent
denominator of F(Ra) increases significantly with increasing 4, of the particle-fluid thermal conductivity ratio, whereas meff /mf is
which is due to the fact that both keff /kf and meff /mf increase with 4 completely independent of the solid phase. As regards F(Pr), notice
(see Figs. 3, 4 and 6). In particular, since keff /kf increases as T that, since the ratio between meff /mf and keff /kf depends on the
increases, and both keff /kf and meff /mf increase as dp decreases, the nanoparticle material only marginally, its distribution is a primary
diminution of F(Ra) occurs with a slope that increases with function of (cp)eff /(cp)f, where, according to eq. (11), (cp)eff is equal
increasing T and decreasing dp. As far as F(Pr) is concerned, both its to the ratio between (rcp)eff and reff. Therefore, owing to the small
numerator and denominator increase with 4, which tends to keep changes that (rcp)eff undergoes with varying 4 (see Fig. 7), it follows
F(Pr) almost unaltered at small volume fractions, i.e., up to 4 z that the modifications of F(Pr) with 4 are determined essentially by
0.015e0.020 (actually, at higher temperatures the increase of the how much reff changes with 4, which is given by eq. (9). With
denominator prevails slightly upon the increase of the numerator, reference to Fig. 11, if we address our attention to the nanofluids
and vice versa at lower temperatures). However, at larger volume that can be prepared by dispersing either Al2O3 or TiO2 nano-
fractions the growth of the product between (cp)eff /(cp)f and meff /mf particles in water or ethylene glycol, it is apparent that, since the
outweighs the growth of keff /kf, which results in a definite increase percent difference between the mass densities of Al2O3 and TiO2 is
of F(Pr) with 4, especially for the smaller nanoparticle diameters. smaller than the percent difference between the mass densities of
The effects of the nanoparticle material and the base fluid on water and ethylene glycol, F(Pr) depends mainly on the base fluid.
Raeff and Preff are highlighted in Figs. 10 and 11, where the distri- On the contrary, if we compare what happens when Al2O3 or Cu
butions of F(Ra) and F(Pr) vs. 4 are reported for three different nanoparticles are suspended in water or ethylene glycol, we can
nanoparticle materials, i.e., copper, alumina and titania, and two observe that F(Pr) depends markedly on the nanoparticle material,
different base fluids, i.e., water and ethylene glycol, for dp ¼ 25 nm and much less on the base liquid, which is a consequence of the fact
and T ¼ 324 K. It may be seen that F(Ra) depends only on the base that the percent difference between the mass densities of Al2O3 and
fluid, being practically independent of the nanoparticle material.

Fig. 10. Distributions of F(Ra) vs. 4 for different nanofluids, assumed dp ¼ 25 nm and
Fig. 8. Distributions of F(Ra) vs. 4 for Al2O3 þ H2O, with dp and T as parameters. T ¼ 324 K.
 M. Corcione / International Journal of Thermal Sciences 49 (2010) 1536e1546 1543

Fig. 11. Distributions of F(Pr) vs. 4 for different nanofluids, assumed dp ¼ 25 nm and Fig. 13. Distributions of E(%) vs. 4 for a square cavity filled with Al2O3 (dp ¼ 25 nm) þ
T ¼ 324 K. H2O, with T as a parameter.

Cu is much higher than the percent difference between the mass It may be seen that, owing to the dispersion of a progressively
densities of water and ethylene glycol. larger amount of solid nanoparticles into the base liquid, the heat
transfer enhancement increases up to a point, which is due to the
4.2. Heat transfer enhancement increased effective thermal conductivity of the nanofluid. Notice
that the impact of the increased effective thermal conductivity is
The effect of the nanoparticle volume fraction on the heat higher when the nanofluid average temperature is higher and the
transfer enhancement is calculated for different average tempera- diameter of the suspended nanoparticles is smaller. The value of 4
tures of the nanofluid, nanoparticle diameters, combinations of corresponding to the peak of E is defined as the optimal particle
suspended nanoparticles and base liquid, and aspect ratios of the loading 4opt. As the volume fraction is further increased above 4opt,
differentially heated enclosure. the heat transfer enhancement decreases, which is due to the
The results obtained for a square cavity filled with Al2O3 þ H2O excessive growth of the nanofluid effective viscosity. In other
are displayed and discussed first; subsequently, emphasis is given words, we can say that the nanofluid behaviour is a strict conse-
to the effects of both the nanoparticle material and the base liquid; quence of the two opposite effects that originate from the increase
finally, the role of the geometry of the enclosure is analysed. of both the effective thermal conductivity and the effective
The effects of the size of the suspended nanoparticles and the dynamic viscosity with increasing the nanoparticle concentration.
nanofluid average temperature are pointed out in Figs. 12 and 13, According to Figs. 4 and 6, the first effect, which tends to enhance
where the distributions of the heat transfer enhancement, E, are the heat transfer performance, prevails at small volume fractions,
plotted versus the volume fraction, 4, for different nanoparticle whilst the second effect, which tends to degrade the heat transfer
diameters, and different temperatures, respectively. In the same performance, prevails at large volume fractions. Obviously, when
figures, the distributions of E vs. 4 obtained by using the Max- the increased viscosity effect outweighs the increased thermal
welleGarnett and Brinkman models for calculating the effective conductivity effect, the heat transfer enhancement becomes
thermal conductivity and dynamic viscosity are also reported, negative, which means that the use of the nanofluid brings to
confirming the weakness of these models in capturing the actual a deterioration in the convective heat transfer performance.
thermo-mechanical features of nanofluids.

Fig. 12. Distributions of E(%) vs. 4 for a square cavity filled with Al2O3 þ H2O at Fig. 14. Distributions of 4opt (%) vs. T for a square cavity filled with Al2O3 þ H2O, with
T ¼ 309 K, with dp as a parameter. dp as a parameter.
1544 M. Corcione / International Journal of Thermal Sciences 49 (2010) 1536e1546

Fig. 15. Distributions of Emax (%) vs. dp for a square cavity filled with Al2O3 þ H2O, with Fig. 17. Distributions of E (%) vs. 4 for a square cavity filled with different nanofluids,
T as a parameter. assumed dp ¼ 25 nm and T ¼ 324 K.

It is worth observing that the optimal particle loading increases function g[F(Ra)]. However, since it can be easily demonstated that
slightly with decreasing the size of the suspended nanoparticles, f [F(Pr), Pr] z 1 (in accordance with the fact that for liquids
whilst increases remarkably when the nanofluid temperature is Pr > > 0.2), it follows that Nueff /Nu depends substantially on
increased, as clearly shown in Fig. 14, where some distributions of F(Ra), that, in turn, is a function of only the base fluid (see Fig. 10).
4opt versus T are reported for different values of dp. In fact, both On the other hand, as discussed earlier, keff /kf depends very few
keff /kf and meff /mf increase as dp is reduced, which implies that the on the nanoparticle material.
effect of the nanoparticle size on 4opt is almost imperceptible. Finally, the effects of the cavity aspect ratio are displayed in
Conversely, since keff /kf enhances significantly when T is increased, Fig. 18, where the distributions of E versus 4 are plotted for three
whilst meff /mf keeps constant, the nanoparticle concentration at different ranges of the aspect ratio of the enclosure (i.e., the ranges
which the increase in viscosity becomes excessive increases dras- of H/L corresponding to eqs. (1)e(3) used to derive the three curves
tically with increasing the nanofluid average temperature. Typical depicted in the diagram), showing that the optimal particle loading
distributions of the heat transfer enhancement at the optimal increases with increasing H/L. Actually, once the Rayleigh number
particle loading, Emax, are reported in Fig. 15, that illustrates the RaH is assigned, the increase of the cavity aspect ratio is obtained by
dependence of Emax on dp with T as a parameter, and in Fig. 16, that reducing the cavity width, L. This implies that the resistance
illustrates the dependence of Emax on T with dp as a parameter. encountered by the fluid to flow across the cavity decreases, and,
The effects of the nanoparticle material and the base fluid on consequently, the growth of the nanofluid effective viscosity starts
the heat transfer enhancement are pointed out in Fig. 17, where becoming excessive in comparison with the growth of the effective
the distributions of E versus 4 are plotted for different solid-fluid thermal conductivity at a larger volume fraction.
combinations, showing that the effect of the base fluid is much For the specific case of a square enclosure filled with
more pronounced than that of the nanoparticle material. This can Al2O3 þ H2O (that according to the literature review seems to be the
be justified by considering that, based on eq. (4), E depends on nanofluid most frequently studied), a multiple regression analysis
both Nueff /Nu and keff /kf. According to eq. (1), the ratio Nueff /Nu of the results obtained for the percentage optimal particle loading
can be expressed as a function f [F(Pr), Pr] multiplied by another produces the following empirical dimensional algebraic equation:

Fig. 16. Distributions of Emax (%) vs. T for a square cavity filled with Al2O3 þ H2O, with Fig. 18. Distributions of E (%) vs. 4 for rectangular cavities of different aspect ratios
dp as a parameter. filled with Al2O3 þ H2O, assumed dp ¼ 25 nm and T ¼ 309 K.
 M. Corcione / International Journal of Thermal Sciences 49 (2010) 1536e1546 1545

(b) The optimal particle loading increases slightly with decreasing

the size of the suspended nanoparticles, whilst increases
remarkably with increasing the nanofluid average tempera-
ture; the corresponding maximum value of the heat transfer
enhancement increases as the nanoparticle diameter decreases
and the average temperature of the nanofluid increases.
(c) When different combinations of solid and liquid phases, i.e.,
different nanofluids, are considered, the effects of the base fluid
on both the heat transfer enhancement and the optimal
particle loading are much more pronounced than those of the
nanoparticle material.
(d) The heat transfer enhancement and the optimal particle
loading increase with increasing the aspect ratio of the differ-
entially heated enclosure.

References

[1] W. Daungthongsuk, S. Wongwises, A critical review of convective heat

transfer in nanofluids. Renew. Sustain. Energy Rev. 11 (2007) 797e817.
[2] S.M.S. Murshed, K.C. Leong, C. Yang, Thermophysical and electrokinetic
properties of nanofluids e a critical review. Appl. Thermal Eng. 28 (2008)
2109e2125.
[3] S. Kakaç, A. Pramuanjaroenkij, Review of convective heat transfer enhance-
ment with nanofluid. Int. J. Heat Mass Transfer 52 (2009) 3187e3196.
[4] K. Khanafer, K. Vafai, M. Lightstone, Buoyancy-driven heat transfer enhance-
ment in a two-dimensional enclosure utilizing nanofluids. Int. J. Heat Mass
Fig. 19. Comparison between eq. (17) and the theoretical data of 4opt (%). Transfer 46 (2003) 3639e3653.
[5] N. Putra, W. Roetzel, S.K. Das, Natural convection of nano-fluids. Heat Mass
   0:19 Transfer 39 (2003) 775e784.
4opt ð%Þ ¼ 5  104 ½tð CÞ2:335 dp ðnmÞ ; (17) [6] R.-Y. Jou, S.-C. Tzeng, Numerical research of nature convective heat transfer
enhancement filled with nanofluids in rectangular enclosures. Int. Comm.
with a 2.05% standard deviation of error, as shown in Fig. 19, in Heat Mass Transfer 33 (2006) 727e736.
[7] R.K. Tiwari, M.K. Das, Heat transfer augmentation in a two-sided lid-driven
which t( C) is the average temperature of the nanofluid in degrees differentially heated square cavity utilizing nanofluids. Int. J. Heat Mass
centigrade, and dp(nm) is the nanoparticle diameter in nm. Transfer 50 (2007) 2002e2018.
[8] H.F. Oztop, E. Abu-Nada, Numerical study of natural convection in partially
heated rectangular enclosures filled with nanofluids. Int. J. Heat Fluid Flow 29
5. Conclusions (2008) 1326e1336.
[9] E. Abu-Nada, H.F. Oztop, Effects of inclination angle on natural convection in
enclosures filled with Cuewater nanofluid. Int. J. Heat Fluid Flow 30 (2009)
The adoption of inadequate models for predicting the effective 669e678.
thermal conductivity and dynamic viscosity of nanofluids can lead [10] E.B. Ög üt, Natural convection of water-based nanofluids in an inclined
enclosure with a heat source. Int. J. Thermal Sci. 48 (2009) 2063e2073.
to erroneous conclusions about the advantage or disadvantage in [11] A.G.A. Nnanna, Experimental model of temperature-driven nanofluid. J. Heat
using nanoparticle suspensions for enclosed natural convection Transfer 129 (2007) 697e704.
applications with respect to traditional heat transfer liquids. [12] H.C. Brinkman, The viscosity of concentrated suspensions and solutions.
J. Chem. Phys. 20 (1952) 571.
In the present paper, assumed that nanofluids behave more like
[13] A.K. Santra, S. Sen, N. Chakraborty, Study of heat transfer characteristics of
a single-phase fluid rather than a conventional solideliquid mixture, copper-water nanofluid in a differentially heated square cavity with different
the heat transfer correlations available in the literature for single- viscosity models. J. Enhanced Heat Transfer 15 (2008) 273e287.
[14] C.J. Ho, M.W. Chen, Z.W. Li, Numerical simulation of natural convection of
phase natural convection in enclosures have been extended to
nanofluid in a square enclosure: effects due to uncertainties of viscosity and
nanoparticle suspensions, by simply replacing the thermophysical thermal conductivity. Int. J. Heat Mass Transfer 51 (2008) 4506e4516.
properties appearing in them with the nanofluid effective properties [15] K. Kwak, C. Kim, Viscosity and thermal conductivity of copper oxide nanofluid
calculated at the reference average temperature. In this connection, dispersed in ethylene glycol. Korea-Aust. Rheol. J. 17 (2005) 35e40.
[16] S.E.B. Maiga, C.T. Nguyen, N. Galanis, G. Roy, Heat transfer behaviours of nano-
two empirical equations, based on a wide variety of experimental fluids in a uniformly heated tube. Superlatt. Microstruct. 35 (2004) 543e557.
data reported in the literature, have been developed for the evalua- [17] X. Wang, X. Xu, S.U.S. Choi, Thermal conductivity of nanoparticle-fluid
tion of the nanofluid effective thermal conductivity and dynamic mixture. J. Thermophys. Heat Transfer 13 (1999) 474e480.
[18] H.E. Patel, T. Sundararajan, T. Pradeep, A. Dasgupta, N. Dasgupta, S.K. Das, A
viscosity, whereas the mixing theory has been used for the calcula- micro-convection model for the thermal conductivity of nanofluids. Pramana
tion of the other effective properties. The heat transfer enhancement e J. Phys. 65 (2005) 863e869.
deriving from the dispersion of solid nanoparticles into a base liquid [19] R. Prasher, D. Song, J. Wang, P. Phelan, Measurements of nanofluid viscosity and
its implications for thermal applications. Appl. Phys. Lett. 89 (2006) 133108.
has been calculated for different values of the average temperature of [20] J. Chevalier, O. Tillement, F. Ayela, Rheological properties of nanofluids
the nanofluid in the range between 294 K and 324 K, the nanoparticle flowing through microchannels. Appl. Phys. Lett. 91 (2007) 233103.
diameter in the range between 25 nm and 150 nm, the height-to- [21] P. Charuyakorn, S. Sengupta, S.K. Roy, Forced convection heat transfer in
microencapsulated phase change material slurries. Int. J. Heat Mass Transfer
width aspect ratio of the enclosure in the range between 1 and 40, as 34 (1991) 819e833.
well as three different nanoparticle materials (i.e., Cu, Al2O3, and [22] J.C. Maxwell, A Treatise on Electricity and Magnetism, third ed. Dover,
TiO2) and two different base fluids (i.e., water and ethylene glycol). New York, 1954.
[23] E. Abu-Nada, Z. Masoud, H.F. Oztop, A. Campo, Effects of nanofluid variable
The main results obtained may be summarized as follows:
properties on natural convection in enclosures. Int. J. Thermal Sci. 49 (2010)
479e491.
(a) The heat transfer enhancement increases with increasing the [24] C.H. Chon, K.D. Kihm, S.P. Lee, S.U.S. Choi, Empirical correlation finding the
nanoparticle volume fraction up to an optimal particle loading role of temperature and particle size for nanofluid (Al2O3) thermal conduc-
tivity enhancement. Appl. Phys. Lett. 87 (2005) 153107.
at which the amount of heat transferred across the enclosure [25] C.T. Nguyen, F. Desgranges, G. Roy, N. Galanis, T. Maré, S. Boucher, H. Angue
has a peak. Mintsa, Temperature and particle-size dependent viscosity data for water-
1546 M. Corcione / International Journal of Thermal Sciences 49 (2010) 1536e1546

based nanofluids e hysteresis phenomenon. Int. J. Fluid Flow 28 (2007) [46] M. Prakash, E.P. Giannelis, Mechanism of heat transport in nanofluids.
1492e1506. J. Computer-Aided Mater. Des. 14 (2007) 109e117.
[26] H. Masuda, A. Ebata, K. Teramae, N. Hishinuma, Alteration of thermal [47] S.M.S. Murshed, K.C. Leong, C. Yang, A combined model for the effective
conductivity and viscosity of liquid by dispersing ultra-fine particles thermal conductivity of nanofluids. Appl. Thermal Eng. 29 (2009)
(dispersion of g-Al2O3, SiO2, and TiO2 ultra-fine particles). Netsu Bussei 4 2477e2483.
(1993) 227e233. [48] S. Lee, S.U.S. Choi, S. Li, J.A. Eastman, Measuring thermal conductivity of fluids
[27] B.C. Pak, Y.I. Cho, Hydrodynamic and heat transfer study of dispersed fluids containing oxide nanoparticles. J. Heat Transfer 121 (1999) 280e289.
with submicron metallic oxide particles. Exp. Heat Transfer 11 (1998) [49] J.A. Eastman, S.U.S. Choi, S. Li, W. Yu, L.J. Thompson, Anomalously increased
151e170. effective thermal conductivity of ethylene glycol-based nanofluids containing
[28] I. Catton, Natural convection in enclosures, , In: Proceedings of the 6th copper nanoparticles. Appl. Phys. Lett. 78 (2001) 718e720.
International Heat Transfer Conference, Toronto, vol. 6 (1978) pp. 13e43. [50] S.K. Das, N. Putra, P. Thiesen, W. Roetzel, Temperature dependence of thermal
[29] A. Bejan, Convection Heat Transfer, third ed. John Wiley & Sons, Inc., Hoboken, conductivity enhancement for nanofluids. J. Heat Transfer 125 (2003)
New Jersey, 2004, p. 270. 567e574.
[30] F.P. Incropera, D.P. Dewitt, T.L. Bergman, A.S. Lavine, Fundamentals of Heat [51] C.H. Chon, K.D. Kihm, Thermal conductivity enhancement of nanofluids by
and Mass Transfer, sixth ed. John Wiley & Sons, Inc., Hoboken, New Jersey, Brownian motion. J. Heat Transfer 127 (2005) 810.
2007, p. 589. [52] S.M.S. Murshed, K.C. Leong, C. Yang, Investigations of thermal conductivity
[31] M. Corcione, E. Habib, Buoyant heat transport in fluids across tilted square and viscosity of nanofluids. Int. J. Thermal Sci. 47 (2008) 560e568.
cavities discretely heated at one side. Int. J. Thermal Sci. 49 (2010) 797e808. [53] H.A. Mintsa, G. Roy, C.T. Nguyen, D. Doucet, New temperature dependent
[32] R.K. MacGregor, A.F. Emery, Free convection through vertical plane layers e thermal conductivity data for water-based nanofluids. Int. J. Thermal Sci. 48
moderate and high Prandtl number fluids. J. Heat Transfer 91 (1969) 391e403. (2009) 363e371.
[33] J. Buongiorno, et al., A benchmark study on the thermal conductivity of [54] W. Duangthongsuk, S. Wongwises, Measurement of temperature-dependent
nanofluids. J. Appl. Phys. 106 (2009) 094312. thermal conductivity and viscosity of TiO2ewater nanofluids. Exp. Thermal
[34] W. Yu, S.U.S. Choi, The role of interfacial layers in the enhanced thermal Fluid Sci. 33 (2009) 706e714.
conductivity of nanofluids: a renovated Maxwell model. J. Nanopart. Res. 5 [55] P. Keblinski, S.R. Phillpot, S.U.S. Choi, J.A. Eastman, Mechanisms of heat flow in
(2003) 167e171. suspensions of nano-sized particles (nanofluids). Int. J. Heat Mass Transfer 45
[35] Q.-Z. Xue, Model for effective thermal conductivity of nanofluids. Phys. Lett. A (2002) 855e863.
307 (2003) 313e317. [56] J. Koo, Computational nanofluid flow and heat transfer analyses applied to
[36] D.H. Kumar, H.E. Patel, V.R.R. Kumar, T. Sundararajan, T. Pradeep, S.K. Das, micro-systems, Dissertation Thesis, North Carolina State University, Rayleigh,
Model for heat conduction in nanofluids. Phys. Rev. Lett. 93 (2004) 144301. NC (2005).
[37] J. Koo, C. Kleinstreuer, A new thermal conductivity model for nanofluids. [57] N. Masoumi, N. Sohrabi, A. Behzadmehr, A new model for calculating the
J. Nanopart. Res. 6 (2004) 577e588. effective viscosity of nanofluids. J. Phys. D: Appl. Phys. 42 (2009) 055501.
[38] S.P. Jang, S.U.S. Choi, Role of Brownian motion in the enhanced thermal [58] S.K. Das, N. Putra, W. Roetzel, Pool boiling characteristics of nano-fluids. Int. J.
conductivity of nanofluids. Appl. Phys. Lett. 84 (2004) 4316e4318. Heat Mass Transfer 46 (2003) 851e862.
[39] S.P. Jang, S.U.S. Choi, Effects of various parameters on nanofluid thermal [59] Y. He, Y. Jin, H. Chen, Y. Ding, D. Cang, H. Lu, Heat transfer and flow
conductivity. J. Heat Transfer 129 (2007) 617e623. behaviour of aqueous suspensions of TiO2 nanoparticles (nanofluids) flowing
[40] H. Xie, M. Fujii, X. Zhang, Effect of interfacial nanolayer on the effective upward through a vertical pipe. Int. J. Heat Mass Transfer 50 (2007)
thermal conductivity of nanoparticle-fluid mixture. Int. J. Heat Mass Transfer 2272e2281.
48 (2005) 2926e2932. [60] H. Chen, Y. Ding, C. Tan, Rheological behaviour of nanofluids. N. J. Phys. 9
[41] Y. Ren, H. Xie, A. Cai, Effective thermal conductivity of nanofluids containing (2007) 367.
spherical nanoparticles. J. Phys. D: Appl. Phys. 38 (2005) 3958e3961. [61] H. Chen, Y. Ding, Y. He, C. Tan, Rheological behaviour of ethylene glycol based
[42] R. Prasher, P. Bhattacharya, P.E. Phelan, Thermal conductivity of nanoscale titania nanofluids. Chem. Phys. Lett. 444 (2007) 333e337.
colloidal solutions (nanofluids). Phys. Rev. Lett. 94 (2005) 025901. [62] J.-H. Lee, K.S. Hwang, S.P. Jang, B.H. Lee, J.H. Kim, S.U.S. Choi, C.J. Choi, Effective
[43] R. Prasher, P. Bhattacharya, P.E. Phelan, Brownian motion-based convective- viscosities and thermal conductivities of aqueous nanofluids containing low
conductive model for the effective thermal conductivity of nanofluids. J. Heat volume concentrations of Al2O3 nanoparticles. Int. J. Heat Mass Transfer 51
Transfer 128 (2006) 588e595. (2008) 2651e2656.
[44] K.C. Leong, C. Yang, S.M.S. Murshed, A model for the thermal conductivity of [63] J. Garg, B. Poudel, M. Chiesa, J.B. Gordon, J.J. Ma, J.B. Wang, Z.F. Ren, Y.T. Kang,
nanofluids e the effect of interfacial layer. J. Nanopart. Res. 8 (2006) 245e254. H. Ohtani, J. Nanda, G.H. McKinley, G. Chen, Enhanced thermal conductivity
[45] Y. Xuan, Q. Li, X. Zhang, M. Fujii, Stochastic thermal transport of nanoparticle and viscosity of copper nanoparticles in ethylene glycol nanofluid. J. Appl.
suspensions. J. Appl. Phys. 100 (2006) 043507. Phys. 103 (2008) 074301.