International Journal of Heat and Mass Transfer 99 (2016) 532–540

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer

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

Effective thermal conductivity of nanofluids – A new model taking into

consideration Brownian motion
Kedar N. Shukla a, Thomas M. Koller a, Michael H. Rausch a,b, Andreas P. Fröba a,b,⇑
a
Erlangen Graduate School in Advanced Optical Technologies (SAOT), Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU), Paul-Gordan-Straße 6, D-91052 Erlangen, Germany
b
Lehrstuhl für Technische Thermodynamik (LTT), Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU), Am Weichselgarten 8, D-91058 Erlangen, Germany

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

Article history: In this study, a new analytical model for the effective thermal conductivity of liquids containing dis-
Received 3 August 2015 persed spherical and non-spherical nanometer particles was developed. In addition to heat conduction
Received in revised form 26 February 2016 in the base fluid and the nanoparticles, we also consider convective heat transfer caused by the
Accepted 30 March 2016
Brownian motion of the particles. For nanoparticle suspensions, the latter mechanism has significant
influence on the effective thermal conductivity, which is reduced compared to a system in which only
conduction is considered. The simple model developed allows for the prediction of the effective thermal
Keywords:
conductivity of nanofluids as a function of volume fraction, diameter, and shape of the nanoparticles as
Effective thermal conductivity
Modeling
well as temperature. Due to the inconsistency of experimental data in the literature, the model has been
Nanofluids compared with the established Hamilton–Crosser model and other empirical models for the systems alu-
Nanoparticles minum oxide (Al2O3) and titanium dioxide (TiO2) suspended in water and ethylene glycol. The theoretical
estimates show no anomalous enhancement of the effective thermal conductivity and agree very well
with the Hamilton–Crosser model within relative deviations of less than 8% for volume fractions of spher-
ical particles up to 0.25. In accordance with the Hamilton–Crosser model for non-spherical particles, our
model reveals that a more distinct increase in the enhancement of the effective thermal conductivity can
be achieved using non-spherical nanoparticles having a larger volume-specific surface area.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction present in suspensions of nanoparticles is necessary, which may

Over the past decade, the dispersions of nanometer-sized parti-
cles with typical diameters ranging from 1 to 100 nm in a liquid
medium, usually called nanofluids, have been reported to possess
substantially higher thermal conductivities than anticipated from
Maxwell’s classical theory [1]. As shown in the review by Tertsini-
dou et al. [2], a large number of experimental results have reported
an anomalous increase in the thermal conductivity of nanoparticle
suspensions. This would make them very attractive as potential
heat transfer fluids for many applications. However, results from
other experiments have not shown any anomalous increase in
thermal conductivity [2–5]. This has triggered controversy regard-
ing the actual value of the thermal conductivity of nanofluids and
the reliability of the experimental methods. To solve this problem,
a fundamental understanding of the heat transfer mechanisms
⇑ Corresponding author at: Erlangen Graduate School in Advanced Optical
Technologies (SAOT), Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU),
Paul-Gordan-Straße 6, D-91052 Erlangen, Germany. Tel.: +49 9131 85 29789; fax:
+49 9131 85 25851.
E-mail address: andreas.p.froeba@fau.de (A.P. Fröba).
lead to a physical model for the description of their effective ther-
mal conductivity.
Debate continues regarding the potential causes of abnormal
behavior ranging, e.g., from the effect of nanoparticle clustering
[6] over the layering of the liquid at the liquid-particle interface
[7] to the role of the Brownian motion of nanoparticles [8]. Regard-
ing the impact of Brownian motion on the effective thermal con-
ductivity, one opinion in the literature [8–10] is that this effect is
the main reason for the high thermal conductivity of nanofluids.
Yet, the molecular dynamics simulations performed by Evans
et al. [11] reveal that the enhancement effect due to Brownian
motion of the nanoparticles is rather insignificant. Therefore, dis-
agreement also exists among researchers on the role of Brownian
motion of nanoparticles.
Maxwell was one of the first to use the effective medium theory
to study the properties of a solid bulk material consisting of one
material distributed as spherical inclusions within a continuous
material [1]. His static model for the effective electrical conductiv-
ity of solid-based systems was used by Hamilton and Crosser [12]
to determine the effective thermal conductivity of two-phase,
two-component solids because of the similar mathematical

http://dx.doi.org/10.1016/j.ijheatmasstransfer.2016.03.129
0017-9310/Ó 2016 Elsevier Ltd. All rights reserved.
 K.N. Shukla et al. / International Journal of Heat and Mass Transfer 99 (2016) 532–540 533

Nomenclature

a empirical parameter in model of Lin and Lu [13] Rp,conv thermal resistance of nanoparticle due to convection
abf thermal diffusivity of base fluid (m2 s1) (K W1)
A surface area of particle (m2) Re Reynolds number
b empirical parameter in model of Lin and Lu [13] T temperature (K)
dp diameter of spherical nanoparticle (m) up velocity of nanoparticle (m s1)
dp,eq volume-equivalent diameter of non-spherical nanopar- V volume (m3)
ticle (m)
h convective heat transfer coefficient (W m2 K1) Greek symbols
k thermal conductivity (W m1 K1) mbf kinematic viscosity of base fluid (m2 s1)
kb Boltzmann constant (1.3806488  1023 J K1) u volume fraction of nanoparticles
L edge length of the cube (m) qp density of nanoparticle (kg m3)
m mass of nanoparticle (kg) w sphericity of nanoparticle
n empirical shape factor of particle
N number of nanoparticles Subscripts
Nu Nußelt number
bf base fluid
Pe Peclet number cond pure conduction
Pr Prandtl number conv convection
Rbf thermal resistance of bulk fluid due to heat conduction
eff effective
(K W1) non-sph non-spherical
Reff effective thermal resistance of nanofluid (K W1) p particle
Rp total thermal resistance of nanoparticle (K W1) sph spherical
Rp,cond thermal resistance of nanoparticle due to heat conduc-
tion (K W1)

formulations of the two transport phenomena. Their model is also
applicable for systems of liquids containing solid particles with dif-
ferent shapes and predicts their effective thermal conductivity keff
by
keff kp þ ðn  1Þkbf  ðn  1Þuðkbf  kp Þ
¼ : ð1Þ
kbf kp þ ðn  1Þkbf þ uðkbf  kp Þ
The terms kbf and kp denote the thermal conductivities of the
base fluid (bf), that is the continuous phase, and the solid particle
(p), that is the dispersed phase. u is the volume fraction of the dis-
persed particles, while n is an empirical shape factor. The latter is
connected with the sphericity w via
3 and b = 4.51 for kp/kbf ? 1. For non-spherical inclusions, anisotro-
n¼ : ð2Þ
w pic effects have to be considered which results in the formulation
The sphericity of a particle is defined as the ratio of the surface
area of a sphere, Ap,sph, having the same volume as the particle, Vp,
to the surface area of the particle, Ap, according to
p1=3 ð6V p Þ2=3
w¼ : ð3Þ
Ap
Particles with strong deviations from spherical shape show
smaller w and thus larger n values. For spherical particles with
w = 1 and n = 3, Eq. (1) can be simplified to the original Maxwell-
like equation
keff kp þ 2kbf  2uðkbf  kp Þ
¼ ð4Þ
kbf kp þ 2kbf þ uðkbf  kp Þ
for predicting the enhancement of the effective thermal conductiv-
ity of nanofluids containing spherical solid particles.
The effective model of Hamilton and Crosser [12] can be applied
to low particle volume fractions and shows no dependence on the
particle diameter and a very weak dependence on the temperature.
For kp  kbf and small u, Eq. (1) reduces to
keff
¼ 1 þ nu: ð5Þ
kbf
Eq. (5) presents a simple linear relation for the effective thermal
conductivity of diluted suspensions. Based on the continuum
model of Hamilton and Crosser [12], there have been some modi-
fications considering additional effects. For example, Lu and Lin
[13] considered near- and far-field pair interactions applicable to
spherical and non-spherical inclusions and modified Eq. (5) to a
second-order polynomial,
keff
¼ 1 þ au þ bu2 : ð6Þ
kbf
For spherical isotropic inclusions, the constants a and b are
given as a = 2.25 and b = 2.27 for kp/kbf = 10 as well as a = 3.00
of an effective thermal conductivity tensor for an anisotropic med-
ium. Besides the models described above, there are numerous
other predictive methods – most of them empirical or complex –
for the effective thermal conductivity that cannot all be mentioned
here. A review about theoretical studies on the effective
thermal conductivity of nanofluids is given by Kleinstreuer and
Feng [14].
Recently, Tertsinidou et al. [2] evaluated extensive data on the
effective thermal conductivity of nanoparticle suspensions con-
taining aluminum oxide (Al2O3), copper (Cu), copper oxide (CuO),
and titanium dioxide (TiO2) suspended in water (H2O) and ethy-
lene glycol (EG). When results for the same thermodynamic system
are obtained using proven experimental techniques, they con-
cluded that the effective thermal conductivity of nanofluids exhi-
bits no inconsistency with the continuum model of Hamilton and
Crosser [12]. The broad span of values for the enhancement of
effective thermal conductivity in the literature, with relative devi-
ations of several tens of percents for a given system, is rather
attributed to poor characterization of the thermodynamic system
and/or the application of experimental techniques of unproven
validity [2]. A systematic benchmark study [15] on the thermal
conductivity of nanofluids, performed over 30 laboratories world-
wide and using a variety of experimental techniques, has also
534 K.N. Shukla et al. / International Journal of Heat and Mass Transfer 99 (2016) 532–540
Fig. 1. Conceptual three-dimensional sketch of a cube with length L containing a
spherical nanoparticle of diameter dp and bulk fluid molecules.
shown good agreement between the experimental data and the
corresponding data predicted by the Hamilton–Crosser model [12].
Nevertheless, the continuum model does not explicitly account
for effects that are induced by nanoparticles with respect to the
effective thermal conductivity of nanofluids. Nanoparticles may
cause additional energy transport due to Brownian motion because
of their small size, large specific surface area, and morphology. The
motion of nanoparticles and base fluid molecules is affected by the
combined effect of hydrodynamic and Brownian forces that pro-
duce micro-convection in the nanofluids. The aim of the present
work was to develop a new model for the effective thermal con-
ductivity of macroscopically static nanofluids taking into account
the heat transfer mechanisms caused by convection as well as
thermal conduction of the particles and the base fluid. A compar-
ison of the model with the established model of Hamilton and
Crosser [12] and the predictions of Lu and Lin [13] is drawn for
selected nanofluid systems as a function of the parameters volume
fraction, particle diameter, particle shape, and temperature.
2. Description of the model
To develop an analytical model for the thermal conductivity of
liquids with suspended nanoparticles, the thermal resistances of
the base fluid and the nanoparticles as well as of convection
induced in the fluid due to Brownian motion of the nanoparticles
are determined. Other possible heat transfer effects in form of ther-
mal radiation or thermal diffusion of the nanoparticles due to a
temperature gradient were found to be negligible.
In the following, the model is derived on basis of a nanofluid
system containing spherical particles. Furthermore, analogies or
differences in connection with the modeling of nanofluids contain-
ing non-spherical particles such as cylinders, cubes, or plates are
given. It is assumed that N nanoparticles of spherical shape with
diameter dp are uniformly suspended in a volume V of the nano-
fluid. The volume fraction of the nanoparticles u is then defined by
Npdp
3
u¼ : ð7Þ
6V
The volume V is now divided into N equal parts such that each
nanoparticle is located in a cube with a side of length L. Fig. 1
depicts a three-dimensional sketch of such a nanoparticle-cube
system. From Eq. (7) it follows that
pd3p
u¼ : ð8Þ
6L3
Thus, the particle diameter can be related to the side length of
the elementary cube by

1=3
6u
dp ¼ L: ð9Þ
p
Based on Eq. (9), the model is restricted to dp/L < 1. This imposes
a condition on the applicable particle volume fraction of
u < p/6  0.52, i.e., the maximum packing of spherical particles in
a nanofluid. However, such a dense packing is not realistic for a
nanofluid in terms of its synthesis and the suspension stability.
Thus, only reasonable values of u < 0.25 are considered within this
study.
Regarding non-spherical particles such as cylinders or plates of
volume Vnon-sph, they can be regarded as theoretical spherical par-
ticles having the same volumes as the non-spherical particles. The
volume-equivalent particle diameter of the non-spherical particles,
dp,eq, is

1
6V non-sph 3
dp;eq ¼ : ð10Þ
p
This approach allows for the same formulation of the volume
fraction of non-spherical particles in the bulk fluid as given in Eq.
(8). Furthermore, it can be used to model the heat transfer in such
systems analogously to that described in the following for systems
with spherical particles.
The basic idea of the present modeling approach is to treat the
heat transfer problem in connection with nanofluids by the analy-
sis of the corresponding thermal resistances present in such sys-
tems. Fig. 2 illustrates the corresponding circuit diagram for the
total thermal resistance of the nanofluid Reff based on the cube
shown in Fig. 1.
The thermal resistance of the base fluid, Rbf, is considered to be
parallel to the thermal resistance of the nanoparticle, Rp. Thus, Reff
can be expressed by
1 1u u
¼ þ : ð11Þ
Reff Rbf Rp
Arranging the resistances of the bulk fluid and the particles in
parallel is reasonable because a heat flux can be either conducted
through the base fluid or through the particle along a one-
dimensional temperature gradient. The key for a realistic descrip-
tion of the thermal resistance of the nanofluid is that the base fluid
as a continuum fluid phase is analogously treated as a continuum
resistance. To account for the volumes in which the resistances of
the two phases are present, the inverse values of the thermal resis-
tances Rbf and Rp are weighted in Eq. (11) by the corresponding vol-
ume fractions (1u) and u. Here, the circumstance that the volume
fractions in the elementary cube are the same as those in the total
nanofluid system can be employed.
Fig. 2. Circuit diagram for the thermal resistances of the bulk fluid due to
conduction and of the nanoparticle due to conduction and Brownian convection.
 K.N. Shukla et al. / International Journal of Heat and Mass Transfer 99 (2016) 532–540 535

Reff and Rbf are defined by the corresponding thermal conductiv- depends on the surface area of the particle Ap and the convective
ities keff and kbf and the geometries of the cube in the form of heat transfer coefficient h. Eq. (3) can be used to derive an expres-
sion for Ap for particles of arbitrary shape,
L 1
Reff ¼ ¼ ð12Þ
Ap;sph pdp
2
keff L2 keff L
Ap ¼ ¼ : ð17Þ
and w w

L 1 Decreasing sphericity of the particles (w < 1) results in an

Rbf ¼ 2
¼ : ð13Þ increasing surface area and, according to Eq. (16), in a decreasing
kbf L kbf L
value for Rp,conv.
The thermally induced Brownian motion creates a thermal The heat transfer coefficient for a spherical particle is deter-
boundary layer on the spherical nanoparticles. The thermal resis- mined by
tance caused by heat transfer between the nanoparticles and the
kbf
boundary layer is modeled with the resistance of the sphere due h¼ Nu; ð18Þ
to thermal conduction, Rp,cond, and with that due to convection at dp
its surface, Rp,conv, in series, where dp represents the characteristic length and Nu is the Nußelt
Rp ¼ Rp;cond þ Rp;conv : ð14Þ number. For a flow around a sphere, the latter can be correlated
with the Peclet number Pe by [17]
A circuit in series in connection with the nanoparticle can be


justified because any heat flux associated with the particle needs Pe 1 2 Pe 1 3 Pe
Nu ¼ 2 þ þ Pe ln þ 0:2073Pe2 þ Pe ln : ð19Þ
to cross the boundary layer acting as first thermal resistance and 2 4 2 16 2
is conducted through the particle material being a second thermal Eq. (19) holds for Pe < 1 and Reynolds numbers Re  1, which is
resistance. Thus, the sum of these two resistances determines the fulfilled for macroscopically static nanofluids. The first term in Eq.
achievable enhancement of the effective thermal conductivity of (19) represents the minimum Nu number of 2 for stagnant spher-
nanofluids compared to the thermal conductivity of the corre- ical bodies suspended in a fluid, i.e., Re = 0. In this case, the convec-
sponding base fluid. tive heat transfer between particles and fluid is at its minimum and
In Eq. (14), the thermal resistance due to thermal conduction of corresponds to pure conduction heat transfer in the fluid. The
the particle is modeled by dimensionless numbers Pe, Re, and the Prandtl number Pr are given
1 þ 0:5 Lp
d by
Rp;cond ¼ : ð15Þ
2pkp dp Pe ¼ Re  Pr ð20Þ
This expression reflects the thermal resistance for an isothermal with
spherical nanoparticle with thermal conductivity kp buried in a
up dp
semi-infinite medium with insulated surface [16]. Here, it is Re ¼ ð21Þ
assumed that the nanoparticle is submerged in the base fluid under
mbf
the influence of a temperature gradient. Using instead the expres- and
sion for the thermal resistance of an isothermal sphere buried in an mbf
infinite medium without any temperature gradient (Rp,cond = 1/ Pr ¼ : ð22Þ
abf
(2pkpdp)) [16], no significant difference for the modeled effective
thermal conductivity of nanofluids is found compared to the use In Eqs. (21) and (22), up, mbf, and abf are the velocity of the
of Eq. (15). nanoparticles as well as the kinematic viscosity and the ther-
For the modeling of Rp of non-spherical particles, it would be mal diffusivity of the bulk fluid. Also for the calculation of Re
plausible to use corresponding expressions from Ref. [16]. Yet, for a sphere, dp is the characteristic length. The velocity up
the solutions regarding non-spherical particles are only valid for required in Eq. (21) can be calculated from the root mean
certain geometrical restrictions and the value of Rp depends on square velocity of a spherical nanoparticle of mass m due to
how the particles are aligned with respect to the direction of thermal motion by
the heat flux. For example, a long cylindrical particle shows less rffiffiffiffiffiffiffiffiffiffiffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffi
3kb T 18kb T
resistance if perpendicularly oriented to the heat flux compared up ¼ ¼ : ð23Þ
to a parallel configuration. Furthermore, it is impossible to predict m pd3p qp
the distribution of the spatial arrangement of the individual par-
Here, kb (=1.38064881023 JK1) is the Boltzmann constant, T
ticles in the nanofluid. All this impedes an analytical modeling of
the temperature, and qp the density of the nanoparticle. Eq. (23)
the effective value of Rp for non-spherical anisotropic particles in
is commonly used for the calculation of the Brownian velocity of
comparison with that for isotropic spherical particles. Our calcu-
nanoparticles in liquid fluids [9,10,18]. In our model, up is consid-
lations have shown that the thermal conduction resistance of
ered to be independent of the particle volume fraction.
non-spherical particles averaged over a broad range of spatial
To our knowledge, no empirical correlations are available in
positions is comparable to that of a spherical particle with the
literature to model the convective heat transfer coefficient in con-
same volume. Apparently, the volume of the particle itself,
nection with finite non-spherical particles of cylindrical, cubic, or
regardless of its shape, is responsible for the heat transfer due
plate shape. Thus, we follow the analogy of treating the non-
to thermal conduction through the particle in the nanofluid.
spherical particles as volume-equivalent spheres with diameter
Hence, the analogy of using the volume-equivalent diameter dp,
dp,eq. This diameter is used for the calculation of up, Re, Pe, Nu,
eq in Eq. (15) is straightforward to model Rp for non-spherical
and h according to Eqs. (23), (21), (20), (19), and (18),
particles.
respectively.
The second term in Eq. (14), the thermal resistance caused by
By substituting Eqs. (9), (12), (13), (15)–(18) into Eq. (11), we
convective heat transfer
obtain the simple expression for the dimensionless effective ther-
1 mal conductivity keff/kbf of nanofluids containing spherical or
Rp;conv ¼ ; ð16Þ
Ap h non-spherical particles
536 K.N. Shukla et al. / International Journal of Heat and Mass Transfer 99 (2016) 532–540

2  1=3 31

1=3 1 þ 0:5 6u

w
u4=3 6 7
keff 6 p k
¼ ð1  uÞ þ p
bf
4 þ 5 :
kbf p 2 kp Nu

ð24Þ
Three different contributions can be seen in Eq. (24). The first
term (1u) considers the influence of the bulk fluid in form of its
volume fraction while the second term is related to the nanoparti-
cles. In the latter term, the brackets include the contributions from
conduction through the particles (associated with kp) and from
convective heat transfer between the particles and the bulk fluid
(given by w/Nu).

3. Results and discussion

To analyze the quality of our model for the description of the

effective thermal conductivity of nanofluid systems by Eq. (24),
we selected the four simple systems Al2O3/H2O, TiO2/H2O, Al2O3/
EG, and TiO2/EG. These systems are common, inexpensive, and
have been widely tested experimentally [2]. The origin of the ther-
mophysical properties of the nanoparticles and base fluids at
atmospheric pressure required for all subsequent calculations is
summarized in the following.
The thermal conductivity kp and density qp, which were both
assumed to be independent of temperature, are taken taken from
Fig. 3. Peclet number Pe (a) and Nußelt number Nu (b) as a function of the diameter
Tertsinidou [19] and Patel et al. [20] for Al2O3 (kp = 40 Wm1K1, of the spherical nanoparticles dp for the studied nanofluids at a temperature of
qp = 3970 kgm3) and from Tertsinidou [19] for TiO2 T = 300 K.
(kp = 11.8 Wm1K1, qp = 4230 kgm3). The properties kbf, qbf,
mbf, and abf for water (H2O) were taken from the Refprop database
[21]. The corresponding data for ethylene glycol (EG) taken from
Patel et al. [20] were correlated with appropriate fits as a function
of temperature between 283.15 K and 363.15 K. For spherical
particles which are commonly used in nanofluids and are of primary
interest in this study, particle diameters were varied from 5 nm up
to 50 nm. For comparison, also non-spherical nanoparticles such as
cylindrical and plated particles with varying length-to-diameter
ratio as well as cubic particles were investigated.
In Fig. 3, the Peclet number Pe and the corresponding Nußelt
number Nu are depicted as a function of the nanoparticle diameter
dp of spheres for the four studied nanofluids at 300 K. Due to the
use of a volume-equivalent diameter dp,eq for non-spherical parti-
cles, the same results are valid for corresponding systems. With
increasing dp, the Pe number decreases almost asymptotically
due to the decrease in the Re number where Re  d1/2p . For the con-
sidered nanofluids, the type of liquid has a stronger influence on
the Pe number than the type of nanoparticle. The trends for the
Nu number are similar to those for the Pe number because for small Fig. 4. Ratio of thermal resistance due to Brownian convection at the nanoparticles
Pe numbers, Eq. (19) is approximately Nu  2 + Pe/2 and hence also to that due to thermal conduction through the nanoparticles for the studied
nanofluids containing spherical nanoparticles with a diameter of 30 nm at
scales with d1/2
p . No significant variation in both dimensionless
T = 300 K.
numbers is observed for varied temperatures. For example, Nu
decreases from 2.0440 at 283.15 K to 2.0417 at 363.15 K for the
Al2O3/H2O system with dp = 30 nm. For all studied systems, the number is always about 2.1 and thus close to the minimum value
decrease in the Pr values with increasing temperature compen- of 2 for the nanoparticle suspensions, cf. Fig. 3b.
sates for the increase in the Re values at a given particle diameter. In the following, the influence of the thermal resistance for the
It should be noted that Jang and Choi [8] considered Nu = Re2Pr2 convective heat transfer at nanoparticles on the resulting effective
by assuming Reynolds and Prandtl numbers on the order of 1 and thermal conductivity of the nanofluid is considered. The second
10 for the typical nanofluids. While the Pr numbers calculated in term of Eq. (24) consists of the contributions from the thermal con-
this study for the systems investigated are comparable to those duction through the nanoparticles as well as from the Brownian
reported by Jang and Choi [8], their Re values are by almost two convection. The two corresponding resistances are modeled to be
orders of magnitude larger than those calculated here. Their results in series. In Fig. 4, the ratio of the thermal resistances attributed
imply enhanced Brownian convection and hence an anomalous to convection and to conduction through the nanoparticles is
increase in the effective thermal conductivity of the nanofluids. shown as a function of the volume fraction u for all four systems
Contrary to their assumptions, our calculations show that the Rey- containing spherical particles with a diameter of 30 nm at a tem-
nolds number is much smaller than 1 and the corresponding Nu perature of 300 K.
 K.N. Shukla et al. / International Journal of Heat and Mass Transfer 99 (2016) 532–540
The thermal resistance due to micro-convection in the bound-
ary layer at the nanoparticle for all systems is by at least one order
of magnitude larger than that due to conduction through the
nanoparticle itself. Increasing particle volume fractions result in
only a slight decrease in the thermal resistance ratio. The same
trends are also found for varied particle size and temperature. This
result can be related to the fact that Nu is close to 2 for all nanopar-
ticle systems studied here. Thus, the convective resistance caused
by Brownian motion and being proportional to 1/Nu  1/2 is more
significant in Eq. (24) than the conduction resistance term propor-
tional to kbf/kp. Depending on the liquid-nanoparticle combination,
kbf/kp values between about 1/20 for TiO2/H2O and 1/160 for Al2O3/
EG can be found. In consequence, it can be deduced that the con-
vective heat transfer resistance is the limiting factor for the
enhancement of the effective thermal conductivity of nanofluids
and may not be neglected in corresponding models.
For non-spherical particles, the value for Rp,conv/Rp,cond is smaller
than for spherical particles at a given particle volume, particle vol-
ume fraction, and nanofluid system. While Rp,cond is considered to
be constant in our model, Rp,conv of non-spherical particles is smal-
ler than that of spherical particles due to the larger surface area
and thus lower sphericity w. From this, it can already be concluded
here that the effective thermal conductivity of nanofluids with
non-spherical particles should be larger than that of nanofluids
with spherical particles with equal particle volume. More details
will be given later on.
In Fig. 5, the percentage enhancement factor (keff/kbf1) calcu-
lated according to Eq. (24) is shown as a function of the volume
fraction u of spherical particles for the two systems Al2O3/EG and
TiO2/H2O at a temperature of 300 K. As described above, these
two systems differ most strongly regarding the ratio kbf/kp, which
allows for a better visualization of the influence of the volume frac-
tion, particle diameter, and convection contribution. Regarding the
influence of convection, the enhancement found on basis of our
model (Eq. (24)), which includes the effect of Brownian convection,
is compared with a ‘‘theoretical” enhancement which does not
account for this effect. The latter system would consist of static
particles suspended in the liquid, neglecting any thermal resis-
tance in the boundary layer where heat is transferred from the liq-
uid to the solid particle. In this case, Rp,conv = 0 in Eq. (14) which is
equivalent to neglecting the term w/Nu = 1/Nu in Eq. (24).
Fig. 5. Percentage enhancement factor 100  (keff/kbf  1) on the basis of Eq. (24) as
a function of the volume fraction u for different diameters dp of spherical particles
at T = 300 K for the nanofluids Al2O3/EG and TiO2/H2O.
537
It is obvious that neglecting the thermal resistance contribution
from the convective heat transfer between particles and base fluid
results in much larger enhancement factors because the thermal
resistance of the nanoparticle is significantly reduced to only the
resistance of conduction through the particle Rp,cond. (see Fig. 5).
Consequently, the additionally considered thermal resistance
between contacting particles and fluid molecules reduces the
effective thermal conductivity of the nanofluid compared to a the-
oretical system featuring no convective resistance between parti-
cles and fluid. Of course, an enhancement of the convective heat
transfer with, e.g., increasing Re, Pe, and thus Nu numbers
decreases the corresponding resistance, but this effect is rather
small with respect to keff. This is caused by the almost stagnant
flow behavior of nanoparticles in the fluid. These findings valid
for nanofluids with spherical particles are in contradiction to the
widely spread opinion in the literature [8–10] that Brownian con-
vection is mainly responsible for the enhancement of the effective
thermal conductivity of nanofluids.
Fig. 5 shows that the effective thermal conductivity increases
with increasing u, with increasing slope for increasing u values.
Mainly due to the about eight times larger kbf/kp ratio for Al2O3/
EG compared to TiO2/H2O, enhancement in the effective thermal
conductivity is stronger for Al2O3/EG for a given particle size. For
u = 0.25 and dp = 5 nm, a difference in the enhancement of about
11% is found for the two nanofluids. For all the systems studied,
keff/kbf decreases with increasing particle size. This can be attribu-
ted to the decrease in the Re, Pe, and Nu numbers as shown in
Fig. 3. Since Nu  d1/2
p approximately holds, the effective thermal
conductivity of the nanofluids does not change significantly for
dp > 25 nm. The influence of temperature on the effective thermal
conductivity enhancement of nanofluids is even less pronounced
for various u and dp values. The reason for this has been given in
the discussion on the Pe and Nu numbers. For example, considering
the system Al2O3/EG containing nanoparticles with dp = 5, 25, and
50 nm at u = 0.25, the differences in the enhancement factors in
the temperature range between 283.15 K and 363.15 K are 1.3%,
0.43%, and 0.30%, respectively.
To test the reliability of the present model, a comparison with
experimental results seems straightforward. However, the current
situation of experimental data in the literature shows an obscure
picture of thermal conductivity enhancement for nanofluids in
general [2]. For the system Al2O3/EG at 298 K, for instance, Oh
et al. [22] measured the effective thermal conductivity of nanoflu-
ids containing spherical nanoparticles with a diameter of 45 nm by
using a 3x method with an uncertainty of 2% and obtained an
enhancement of about 7.5% with respect to the thermal conductiv-
ity at u = 0.03. In contrast, the measurement results of Xie et al.
[23] from a transient short hot wire method specified with an
uncertainty of less than 0.5% show an enhancement of about 27%
for the same system and particle volume fraction, only using smal-
ler spherical particles with a diameter of 26 nm. The enhancement
of 9.0% predicted by the Hamilton–Crosser model [12] agrees with
the result of Oh et al. [22] within their given uncertainty. Given this
data discrepancy, we preferred to check our model by comparing
with other common models in the literature, where the effective
medium model of Hamilton and Crosser [12] is considered to be
one of the most reliable ones [2].
At first, nanofluids containing spherical particles are investi-
gated. In Fig. 6, the enhancement of the thermal conductivity
(keff/kbf  1) predicted from Eq. (24) is given as a function of the
volume fraction u for all four nanofluids at a particle diameter of
dp = 25 nm and a temperature of T = 300 K. In addition to our
calculations, we also applied the continuum model of Hamilton
and Crosser [12] given by Eq. (4) and the empirical model of
Lu and Lin [13] according to Eq. (6) with the reported values for
a and b for the cases kp/kbf = 10 as well as kp/kbf ? 1.
538 K.N. Shukla et al. / International Journal of Heat and Mass Transfer 99 (2016) 532–540

Fig. 6. Percentage enhancement factor 100  (keff/kbf  1) as a function of volume fraction u at a temperature of T = 300 K for the nanofluids (a) Al2O3/H2O, (b) Al2O3/EG, (c)
TiO2/H2O, and (d) TiO2/EG containing spherical nanoparticles calculated from the present model (Eq. (24), dp = 25 nm), the Hamilton-Crosser model [12] (Eq. (4)), and the
Lu-Lin model [13] (Eq. (6) for the ratios kp/kbf = 10 and kp/kbf ? 1).

For all the nanofluid systems tested, very good agreement
between our model and the continuum model of Hamilton and
Crosser [12] for various particle volume fractions is found. Regard-
ing the enhancement factors, the relative deviations between the
continuum model and ours are less than 8% for volume fractions
between 0 and 0.25 for the studied conditions. Also for varied tem-
peratures and particle diameters, this deviation is not exceeded
due to the very weak influences of T and dp on keff/kbf in our model
and the fact that the Hamilton–Crosser model [12] also shows only
very weak temperature dependence and depends on the particle
diameter. While our model provides lower values than the contin-
uum model for low volume fractions, there is crossover for particle
volume fractions that seems to depend on the kp/kbf value of the
nanofluid. The lower this thermal conductivity ratio, the lower is
the u value at which our model exceeds the Hamilton–Crosser
model [12].
In this context, a comparison of our model with that of Lu and
Lin [13] provides further information. Their correlations show that
an increasing kp/kbf value results in a larger enhancement of the
effective thermal conductivity due to the addition of particles with
larger thermal conductivity. Our model shows the same trend
when the four nanofluids with different kp/kbf ratios are compared,
see Fig. 6. While for the TiO2/H2O system with kp/kbf = 19.3 our
model fits better with the Lu-Lin model [13] for kp/kbf = 10,
there is better agreement between our model in the case of the
Al2O3/EG system with kp/kbf = 159.1 and their model for
kp/kbf ? 1. In conclusion, the very good agreement of the
established model of Hamilton and Crosser [12] as well as other
empirical models [13] with our model indicates that the convective
heat transfer caused by the Brownian motion of the nanoparticles
needs to be considered for the heat transfer in nanofluids.
Further evidence for this can be given by the investigation of
nanofluids with non-spherical particles. In Fig. 7, the enhancement
of the thermal conductivity (keff/kbf  1) modeled according to our
prediction from Eq. (24) is exemplarily shown as a function of the
sphericity of the particle w for the nanofluid Al2O3/H2O containing
particles with dp = dp,eq = 30 nm at a temperature of T = 300 K and
volume fractions u of 0.05, 0.10, 0.15, and 0.20. Spherical particles
are compared with cubic particles as well as seven cylindrical par-
ticles with varying aspect ratios, i.e., the ratio of length to diameter,
which are specified in Fig. 7. While prolate cylindrical particles
have aspect ratios larger than 1, they are smaller for oblate parti-
cles. A broad range of w values is covered ranging from about
0.22 for prolate cylinders with an aspect ratio of 200:1 to 1 for
spheres. To check our modeled data, we use again the continuum
model of Hamilton and Crosser [12] in its general form given by
Eq. (1).
Also for systems with non-spherical nanoparticles, our modeled
data and those predicted by Hamilton and Crosser [12] agree well
for the various particle geometries. Both models show that
 K.N. Shukla et al. / International Journal of Heat and Mass Transfer 99 (2016) 532–540 539

For static nanofluids considered in this study, this could be rather

realized by increasing the volume-specific surface area than by
increasing the Nu number. Thus, it would be preferable to use
nanofluids containing particles with low sphericity such as elon-
gated thin and/or porous nanoparticles, which would be repre-
sented ideally by, e.g., carbon nanotubes. Here, a much larger
surface area per volume of the particles is provided for convective
heat transfer compared to spherical nanoparticles. In case of mov-
ing or pumped nanofluids, also an increased Nu number may
improve the heat transfer. Yet, this increase is limited by the
increasing viscosity of nanofluids with increasing particle volume
fraction, in particular using non-spherical particles [24].

4. Conclusions

A new analytical model to determine the effective thermal con-

ductivity in fluids containing well-dispersed spherical and non-
spherical nanoparticles was presented. The model takes thermal
resistances in connection with the base fluid, the nanoparticles,
and the micro-convection between the nanoparticles and the fluid
Fig. 7. Percentage enhancement factor 100  (keff/kbf  1) as a function of the
sphericity of the particle w at a temperature of T = 300 K for the nanofluid Al2O3/ due to Brownian motion of the particles into account. Furthermore,
H2O containing volume-equivalent (dp = dp,eq = 30 nm) nanoparticles of different the model is based on well-defined properties and does not include
shape calculated from the present model (Eq. (24)) and the Hamilton-Crosser model any empirical constants. It has revealed the significant role of the
[12] (Eq. (1)) for various volume fractions u. The ratio ‘‘length:diameter” is given for considered convective heat transfer resistance in reducing the
prolate (ratio > 1) and oblate (ratio < 1) particles.
effective thermal conductivity of a nanofluid compared with a the-
oretical nanofluid showing no thermal contact resistance between
particle and base fluid. The convective heat transfer resistance
decreasing sphericity of the nanoparticles goes along with a dis-
turned out to control the achievable thermal conductivity
tinct increase in the effective thermal conductivity. For example,
enhancement. Our model points out that the enhancement of the
our model predicts that the enhancement for cylinders with an
effective thermal conductivity of the nanofluid related to the ther-
aspect ratio of 10:1 (w = 0.58) is already twice as large as that for
mal conductivity of the base fluid is strongly dependent on the par-
spherical particles (w = 1). The reduced convective heat transfer
ticle volume fraction, but very weakly dependent on the diameter
resistance caused by the larger specific surface area of non-
of the nanoparticle and the temperature. For four exemplary sys-
spherical particles compared to spherical particles seems to rea-
tems containing spherical Al2O3 or TiO2 nanoparticles suspended
sonably account for the increased enhancement factors in our
in water or ethylene glycol, very good agreement was found
developed model.
between our calculation results and the commonly recommended
For all volume fractions, the slopes for the increasing keff values
model of Hamilton and Crosser [12]. In accordance with this effec-
with decreasing w values are larger for our model compared to that
tive continuum model, our model does not show any anomalous
calculated from the Hamilton-Crosser model [12]. This deviation is
enhancement of the effective thermal conductivity of nanofluids
especially pronounced for small w values below 0.5. The reason for
for low volume fractions of spherical nanoparticles as it is often
this discrepancy might be related to the limited applicability range
reported in the literature. The presented model also suggests that
with respect to w in the Hamilton–Crosser model [12]. It was
a stronger enhancement in the effective thermal conductivity of
developed based on experimental results for mixtures of rubber
nanofluids is found using cubic and especially prolate or oblate
and dispersed alumina particles with sphericities only between
cylindrical particles due to their larger volume-specific surface
0.58 and 1, resulting in the empirical correlation between w and
areas. A further reduction of the convective heat transfer restric-
the shape factor n given in Eq. (2). Yet, it is also possible that our
tions can be expected by the formation of rows of nanotubes hav-
approach of adopting volume-equivalent diameters in our model
ing a large surface-to-volume ratio.
for non-spherical particles is not fully transferable to particle
shapes which differ more and more from the spherical shape.
Acknowledgements
Changing the particle–fluid combination leads to the same gen-
eral dependencies as discussed above for the system Al2O3/H2O.
This work was supported by the German Research Foundation
The main difference is that for increasing kp/kbf for the nanofluid
(Deutsche Forschungsgemeinschaft, DFG) by funding the Erlangen
systems, better agreement between the present model and the
Graduate School in Advanced Optical Technologies (SAOT) within
Hamilton–Crosser model as a function of the sphericity of the par-
the German Excellence Initiative. K.N. Shukla acknowledges sup-
ticles is found, also for low w values. Furthermore, the trends
port from the Alexander von Humboldt Foundation in sponsoring
regarding the distinct influence of the volume fraction and the neg-
a renewed research stay at the Friedrich-Alexander-University
ligible influences of temperature as well as particle diameter on
Erlangen-Nuremberg (FAU).
the enhancement of the effective thermal conductivity found for
systems containing spherical particles can also be observed for
the corresponding systems containing non-spherical particles. References
Based on the above results, we can conclude that the heat trans-
[1] J.C. Maxwell, A Treatise on Electricity and Magnetism, third ed., vol. I, Oxford,
fer resistance due to convection between the particles and the base Clarendon, 1892.
fluid seems to dominate the effective thermal conductivity. Hence, [2] G. Tertsinidou, M.J. Assael, W.A. Wakeham, The apparent thermal conductivity
a strong increase in the effective thermal conductivity of nanoflu- of liquids containing solid particles of nanometer dimensions: a critique, Int. J.
Thermophys. 36 (7) (2015) 1367–1395.
ids seems to be achievable only if the convective heat transfer [3] P. Keblinski, J.A. Eastman, D.G. Cahill, Nanofluids for thermal transport, Mater.
resistance between particles and base fluid is strongly reduced. Today 8 (6) (2005) 36–44.
540 K.N. Shukla et al. / International Journal of Heat and Mass Transfer 99 (2016) 532–540
[4] X. Zhang, H. Gu, M. Fujii, Experimental study on the effective thermal
conductivity and thermal diffusivity of nanofluids, Int. J. Thermophys. 27 (2)
(2006) 569–580.
[5] X. Zhang, H. Gu, M. Fujii, Effective thermal conductivity and thermal diffusivity
of nanofluids containing spherical and cylindrical nanoparticles, Exp. Therm.
Fluid Sci. 31 (6) (2007) 593–599.
[6] J.W. Gao, R.T. Zheng, H. Ohtani, D.S. Zhu, G. Chen, Experimental investigation of
heat conduction mechanisms in nanofluids. Clue on clustering, Nano Lett. 9
(12) (2009) 4128–4132.
[7] W. Yu, S.U.S. Choi, The role of interfacial layers in the enhanced thermal
conductivity of nanofluids: A renovated Hamilton–Crosser model, J. Nanopart.
Res. 6 (4) (2004) 355–361.
[8] S.P. Jang, S.U.S. Choi, Role of Brownian motion in the enhanced thermal
conductivity of nanofluids, Appl. Phys. Lett. 84 (21) (2004) 4316–4318.
[9] R. Prasher, P.E. Phelan, P. Bhattacharya, Effect of aggregation kinetics on the
thermal conductivity of nanoscale colloidal solutions (nanofluid), Nano Lett. 6
(7) (2004) 1529–1534.
[10] J. Koo, C. Kleinstreuer, A new thermal conductivity model for nanofluids, J.
Nanopart. Res. 6 (6) (2004) 577–588.
[11] W. Evans, J. Fish, P. Keblinski, Role of Brownian motion hydrodynamics on
nanofluid thermal conductivity, Appl. Phys. Lett. 88 (9) (2006) 093116.
[12] R.L. Hamilton, O.K. Crosser, Thermal conductivity of heterogeneous two-
component systems, Ind. Eng. Chem. Fundam. 1 (3) (1962) 187–191.
[13] S.-Y. Lu, H.-C. Lin, Effective conductivity of composites containing aligned
spheroidal inclusions of finite conductivity, Appl. Phys. 79 (9) (1996) 6761–
6769.
[14] C. Kleinstreuer, Y. Feng, Experimental and theoretical studies of nanofluid
thermal conductivity enhancement: a review, Nanoscale Res. Lett. 6 (2011)
629.
[15] J. Buongiorno, D.C. Venerus, N. Prabhat, T. McKrell, J. Townsend, R.
Christianson, Y.V. Tolmachev, P. Keblinski, L.-W. Hu, J.L. Alvarado, I. Cheol
Bang, S.W. Bishnoi, M. Bonetti, F. Botz, A. Cecere, Y. Chang, G. Chen, H. Chen, S.
Jae Chung, M.K. Chyu, S.K. Das, R. Di Paola, Y. Ding, F. Dubois, G. Dzido, J. Eapen,
W. Escher, D. Funfschilling, Q. Galand, J. Gao, P.E. Gharagozloo, K.E. Goodson, J.
G. Gutierrez, H. Hong, M. Horton, K. Sik Hwang, C.S. Iorio, S. Pil Jang, A.B.
Jarzebski, Y. Jiang, L. Jin, S. Kabelac, A. Kamath, M.A. Kedzierski, L. Geok Kieng,
C. Kim, J.-H. Kim, S. Kim, S. Hyun Lee, K. Choong Leong, I. Manna, B. Michel, R.
Ni, H.E. Patel, J. Philip, D. Poulikakos, C. Reynaud, R. Savino, P.K. Singh, P. Song,
T. Sundararajan, E. Timofeeva, T. Tritcak, A.N. Turanov, S. Van Vaerenbergh, D.
Wen, S. Witharana, C. Yang, W.-H. Yeh, X.-Z. Zhao, S.-Q. Zhou, A benchmark
study on the thermal conductivity of nanofluids, J. Appl. Phys. 106 (9) (2009)
094212.
[16] J.E. Sunderland, K.R. Johnson, Shape factors for heat conduction through bodies
with isothermal or convective boundary conditions, ASHRAE Trans. 70 (1964)
237–241.
[17] E.M. Efstathios, Nanofluidics: Thermodynamic and Transport Properties,
Springer, 2014.
[18] R. Azizian, E. Doroodchi, B. Moghtaderi, Effect of micro-convection caused by
Brownian motion on the enhancement of thermal conductivity in nanofluids,
Ind. Eng. Chem. Res. 51 (4) (2012) 1782–1789.
[19] G. Tertsinidou, personal communication, 14.05.2015.
[20] H.E. Patel, T. Sundararajan, S.K. Das, An experimental investigation into the
thermal conductivity enhancement in oxide and metallic nanofluids, J.
Nanopart. Res 12 (3) (2010) 1015–1031.
[21] E.W. Lemmon, M.L. Huber, M.O. McLinden, REFPROP Reference Fluid
Thermodynamic and Transport Properties, Standard Reference Database 23,
Version 9.0, National Institute of Standards and Technology, United States,
2010.
[22] D.W. Oh, A. Jain, J.K. Eaton, K.E. Goodson, J.S. Lee, Thermal conductivity
measurement and sedimentation detection of aluminum oxide nanofluids by
using the 3x method, Int. J. Heat Fluid Flow 29 (5) (2008) 1456–1461.
[23] H.Q. Xie, J.C. Wang, T.G. Xi, Y. Liu, F. Ai, Q.R. Wu, Thermal conductivity
enhancement of suspensions containing nanosized alumina particles, J. Appl.
Phys. 91 (7) (2002) 4568–4572.
[24] E.V. Timofeeva, J.L. Routbort, D. Singh, Particle shape effects on thermophysical
properties of alumina nanofluids, Appl. Phys. 106 (1) (2009) 014304.