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

net/publication/267698002

A Review of Thermal Conductivity Data, Mechanisms and Models for

Nanofluids

Article in International Journal of Micro-Nano Scale Transport · December 2010

DOI: 10.1260/1759-3093.1.4.269

CITATIONS READS

232 11,280

5 authors, including:

Seung-Hyun Lee Chul Jin Choi

Korea Aerospace University Institute for Basic Science
22 PUBLICATIONS 1,605 CITATIONS 14 PUBLICATIONS 1,263 CITATIONS

SEE PROFILE SEE PROFILE

Seok Pil Jang

Korea Aerospace University
104 PUBLICATIONS 8,448 CITATIONS

SEE PROFILE

All content following this page was uploaded by Seung-Hyun Lee on 17 November 2014.

The user has requested enhancement of the downloaded file.

 269

A Review of Thermal Conductivity Data,

Mechanisms and Models for Nanofluids
Ji-Hwan Lee1, Seung-Hyun Lee2, Chul Jin Choi1, Seok Pil Jang2,
and Stephen U. S. Choi1,*
1Department of Mechanical and Industrial Engineering, University of Illinois at Chicago,

Chicago, IL, 60607

2School of Aerospace and Mechanical Engineering, Korea Aerospace University, Goyang,

Gyeonggi-do, Korea, 412-791

ABSTRACT
Numerous studies have shown that nanofluids have superb physical properties, among
which thermal conductivity has been studied most extensively but remains controversial.
In this review article, we first present important milestones in experimental studies that
show new features of the thermal conductivity of nanofluids, together with those that
show no such special features. After a brief review of the physical mechanisms proposed
to explain the thermal conductivity of nanofluids we present a critical review of the
classical and new models used to predict the thermal conductivity behavior of nanofluids.
We discuss some controversial issues such as data inconsistencies, the sufficiency and
suitability of classical and new mechanisms, and the discrepancies between experimental
data and model predictions. At the end of our review, we give some directions for future
research in nanofluids and to aid researchers in resolving the controversial issues we are
still facing in developing nanofluids with superior thermal properties and performance
for high heat flux cooling and high efficiency applications.

Keywords: nanofluids, thermal conductivity, classical mechanisms, new mechanisms,

static model, dynamic model

NOMENCLATURE
a Radius
C Constant
Cf Heat capacity per unit volume of the fluid
Ĉ Heat capacity of a nanoparticle
cv Specific heat
Do Diffusion coefficient
d Diameter
ec Electric charge
f Factorial function
geq Equilibrium pair distribution function
h Convective heat transfer coefficient
k Thermal conductivity
kB Boltzmann constant
lat The average distance for a particle to travel along one direction without changing its direction
due to the particle Brownian motion

* Tel: +1-312-355-1788; Fax: +1-312-413-0047; E-mail: [email protected]

Volume 1 · Number 4 · 2010

270 A Review of Thermal Conductivity Data, Mechanisms and Models for Nanofluids

l Mean free path

n Number density of nanoparticles
Pr Prandtl number
q Heat flux
Re Reynolds number
Rg The radius of gyration
RK Kapitza resistance
r Position
T Temperature
t Time
V Volume

Greek symbols
α Thermal diffusivity
δ Nanolayer thickness
φ Volume fraction (or volume concentration)
µ Dynamic viscosity
ρ Density
τ Particle relaxation time
Ψ Interparticle potential

Subscripts
ag Aggregate
bf Base fluid
cl Nanoparticle cluster
e Equivalent
eff Effective
int Particles in the aggregate
K Kapitza
layer Interfacial layer (nanolayer)
p Particle
pB Caused by Brownian motion of the nanoparticles
pe Equivalent particle
peff Effective contribution of the particles
x Along transverse axes
z Along longitudinal axes

Superscripts
eq Equilibrium

1. INTRODUCTION
Cooling technology plays a critical role that controls the performance of industrial, consumer, and medical
devices and systems. For example, very high density devices (i.e., those with 8-nm features) are
anticipated to generate local heat fluxes as high as 105 W/cm2 and the heat fluxes at the die level of
103 W/cm2, 10 times those of present-day CMOS devices [1]. Inefficient dissipation of the immense
amount of heat generated in dense electronics and electronic systems would reduce or limit their
performance. Therefore, the thermal management of electronics and electronic systems has been the most
important contemporary technology driver for thermal scientists and engineers. Similarly, innovations in
the thermal management of the power electronics must be achieved for hybrid electric vehicles [2].
Besides the need for enhanced cooling performance, the increasing demand for energy-savings and
emissions reduction makes energy efficiency a pressing issue in the buildings, transportation, power
generation, manufacturing, and many other sectors. However, the inherently poor thermal properties of

International Journal of Micro-Nano Scale Transport

Ji-Hwan Lee, Seung-Hyun Lee, Chul Jin Choi, Seok Pil Jang, and Stephen U. S. Choi 271

conventional heat transfer fluids, such as water, oil, and ethylene glycol, are a major barrier to innovations
in thermal management and energy efficiency. The development of nanoscale materials provided
enormous opportunities across a wide spectrum of critical technologies. In the area of heat transfer fluids,
nanoparticles can provide new innovative technologies with potential to tailor the heat transfer fluid’s
thermal properties through control over particle size, shape, composition, and others. Realizing the unique
properties of nanoparticles and their potential to overcome the intrinsic limitation of conventional heat
transfer fluids, Choi conceived the concept of nanofluids [3]. The term nanofluids, coined by Choi, is
defined as a new class of nanotechnology-based heat transfer fluids that are engineered by stably
suspending nanosized particles, fibers, sheets, or tubes with average sizes below 100 nm in traditional heat
transfer fluids in low volume concentrations (≤1 vol.%) [4].
Numerous experimental studies have shown that nanofluids significantly enhance thermal
conductivities [5-13], the convective heat transfer coefficient [14-19], and other properties such as
wetting and spreading of nanofluids [20] and heat absorption rate [21]. Because of these unique thermal
transport properties and superior performance that are unavailable in traditional heat transfer fluids or
conventional particle fluid suspensions, nanofluids have been of great scientific interest to researchers
worldwide over the past decade [4]. These novel nanofluids show great promise as next-generation heat
transfer fluids for innovative applications in industries such as nuclear power generation, buildings,
transportation, aerospace, electronics, tribology, and medicine among others [4, 22]. Besides enhanced
thermal properties, nanofluids have other potentially useful properties, such as the formation of
nanoporous structures on heated surfaces. Therefore, today’s nanofluids technology can be useful to
broader applications of nanofluids as nanostructured materials.
In this review paper, we first give an overview of the unique features and major mechanisms of the
thermal conductivity, k, of nanofluids (NFs). We then focus our review on the models proposed for
enhanced heat conduction in NFs. We also discuss some major issues and challenges related to
experimental data, mechanisms and models for NFs from selected works. We present both sides of
controversial issues such as data inconsistencies, the sufficiency and suitability of classical and new
mechanisms, and the discrepancies between experimental data and model predictions. Finally, we
present future research directions in order to address some controversial issues and challenges.

2. EXPERIMENTALLY DISCOVERED FEATURES OF THE THERMAL CONDUCTIVITY OF

NANOFLUIDS
Thermal conductivity is the most studied property of nanofluids because it is of great theoretical and
practical interest to scientists and engineers. Pioneering researchers have discovered that the heat
conduction behavior of nanofluids has novel features that are completely lacking in conventional
suspensions of micro- or millimeter-sized solid particles. Moreover, the unprecedented thermal transport
phenomena in nanofluids surpass the fundamental limits predicted by effective medium theory (EMT),
in which thermal diffusion is the only heat conduction mechanism. In contrast to these pioneering
discoveries, this unexpected heat conduction behaviors are not observed in some experimental studies.
In this section, representative experimental studies from both sides will be presented.

2.1. Anomalously high thermal conductivity at low concentrations

Lee et al. [23] and Wang et al. [24] produced oxide nanofluids using a two-step method and showed that
the k of oxide nanofluids is slightly greater than that of base fluids or than that predicted by EMT.
Therefore, this study did not generate significant interest in oxide nanofluids. Although Eastman et al.
[25] and Xuan and Li [26] were the first investigators to make metallic nanofluids containing Cu
nanoparticles, using a two-step method the k of their nanofluids was not superior to that of oxide
nanofluids. However, researchers took notice when Eastman et al. [5] demonstrated for the first time that
copper nanofluids produced by a single-step method show a dramatic enhancement in thermal
conductivity (up to a 40% at a particle concentration of 0.3 vol.%) compared to oxide nanofluids
produced by a the two-step method. Conventional particle-liquid suspensions require high
concentrations (>10%) of particles to achieve such dramatic enhancement. Furthermore, they showed
that Cu nanofluids have an anomalous enhancement in thermal conductivity far beyond the predictions

Volume 1 · Number 4 · 2010

272 A Review of Thermal Conductivity Data, Mechanisms and Models for Nanofluids

of conventional effective medium theory even at very low volume concentrations (<1 vol.%). Patel et al.
[7] produced gold nanofluids using single-step chemical methods and observed anomalous
enhancements in thermal conductivity at vanishing concentrations. Liu et al. [27] synthesized Cu
nanofluids using a chemical reduction method with no surfactant and observed that the thermal
conductivity was enhanced by up to 23.8% at 0.1 vol.%. Jana et al. [12] observed that Cu-water
nanofluids show a 74% increase in conductivity at 0.3 vol.%, far exceeding the previous record of 40%
[5].
More recently, Garg et al. [28] measured the thermal conductivity of Cu nanoparticles in ethylene
glycol and also found an anomalous increase, i.e., the measured increase in conductivity was twice the
predicted value by the MG theory [29]. These studies show that metallic nanofluids truly stand out as high
quality nanofluids. Schmidt et al. [30] also reported that the thermal conductivity of Al2O3/decane
nanofluids have an anomalous enhancement in thermal conductivity (11% at 1 vol.%) and Philip et al.
[31] reported very exciting results that the thermal conductivity enhancement of Fe3O4 nanofluids under
a magnetic field can reach up to 216% at 4.5 vol.%.
In contrast, no anomalous enhancement in thermal conductivity was observed with nanofluids
produced by other groups [for example, see 32-36]. These contradictory k data highlight the need for more
controlled syntheses and accurate characterization and thermal conductivity measurements of nanofluids.
In Figure 1, the thermal conductivity ratio keff /kbf is plotted for several kinds of nanofluids from
different sources as a function of βφ, where β ≡ [kp − kbf]/[kp + 2kbf] [37] and φ is the volume fraction
of nanoparticles, and compared with the prediction from the Maxwell model for dilute nanofluids of
spherical nanoparticles. The very fact that there are huge differences in k as shown in Figure 1 infers
that there are huge differences in the quality of nanofluids. Although a number of data can be
predictable, some other experimental results [for example, 5, 10, 24, 38-40] show anomalously high
thermal conductivity compared with the predictions of the classical EMT-based model. In general,
nanofluids produced by one-step methods have well dispersed and stably suspended nanoparticles and
higher thermal conductivities compared to those produced by two-step methods, as shown in Figure 2.
Therefore, good dispersion and stable suspension of nanoparticles in the host liquids are prerequisites

Al2O3+ Water Nanofluids

Lee et al. (1999) [23]
Wang et al. (1999) [24]
1.60 Xie et al. (2002) [38]
Das et al. (2003) [8]
Murshed et al. (2006) [39]
1.55 Patel et al. (2010) [40]
Al2O3+ EG Nanofluids
1.50 Lee et al. (1999) [23]
Wang et al. (1999) [24]
1.45 Xie et al. (2002) [38]
Patel et al. (2010) [40]
Al2O3+ Decane Nanofluids
1.40 Schmidt et al. (2008) [30]
CuO + Water Nanofluids
1.35 Lee et al. (1999) [23]
Wang et al. (1999) [24]
keff / kbf

1.30 Das et al. (2003) [8]

Li and Peterson (2006) [10]
CuO + EG Nanofluids
1.25 Lee et al. (1999) [23]
Wang et al. (1999) [24]
1.20 Patel et al. (2010) [40]
CuO + Trans Oil Nanofluids
Patel et al. (2010) [40]
1.15 Cu + Water Nanofluids
Patel et al. (2010) [40]
1.10 Cu + EG Nanofluids
Eastman et al. (2001) [5]
1.05 Patel et al. (2010) [40]
Cu + Trans Oil Nanofluids
Patel et al. (2010) [40]
1.00
Maxwell model
0.95
-0.02 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16
βφ

Figure 1. Experimental data on the ratio of thermal conductivity of nanofluids to that of the base
fluid as a function of βφ, where β ≡ [kp − kbf] /[kp + 2kbf], φ is the volume fraction of
nanoparticles.

International Journal of Micro-Nano Scale Transport

Ji-Hwan Lee, Seung-Hyun Lee, Chul Jin Choi, Seok Pil Jang, and Stephen U. S. Choi 273

1.45
Cu + EG Nanofluids
1.40 <One-step method>
Themral conductivity ratio (keff /kbf)
Eastman et al. (2001) [5]
1.35 Zhu et al. (2004) [136]
<Two-step method>
1.30 Patel et al. (2009) [40]
Maxwell model (1873) [29]
1.25

1.20

1.15

1.10

1.05

1.00
0.000 0.002 0.004 0.006 0.008 0.010
βφ
(a) Experimental results of thermal conductivity ratio of Cu-EG nanofluids

1.25
Cu + Water Nanofluids
<One-step method>
Themral conductivity ratio (keff /kbf)

Liu et al. (2006) [27]

1.20 <Two-step method>
Patel et al. (2009) [40]
Maxwell model (1873) [29]
1.15

1.10

1.05

1.00
0.000 0.002 0.004 0.006 0.008 0.010
βφ

(b) Experimental results of thermal conductivity ratio of Cu-water nanofluids

Figure 2. Comparison of the thermal conductivity ratio for nanofluids produced using one-step
and two-step processes.

for the study of nanofluids properties as well as for their applications. This requirement can be achieved
using various dispersion and stabilization methods. However, it is important to keep in mind that some
nanofluids containing uniformly dispersed and stably suspended nanoparticles do not have any
conductivity enhancement for presently unknown reasons.

2.2. Nonlinear relationship between conductivity and concentration

Choi et al. [6] made the first carbon nanotubes (CNT)-based nanocomposite material with the highest

Volume 1 · Number 4 · 2010

274 A Review of Thermal Conductivity Data, Mechanisms and Models for Nanofluids

thermal conductivity enhancement ever achieved by dispersing CNTs in a liquid matrix of PAO and
discovered that the measured thermal conductivity of the CNT nanofluids is nonlinear with
nanoparticle volume fraction at low volume fractions of CNTs. This is another anomalous feature of
the k of nanofluids, since predictions by classical EMT show a linear relationship. In an attempt to
confirm these results, Xie et al. [41] and Wen and Ding [42] made CNT nanofluids and measured the
thermal conductivity of their CNT nanofluids but found smaller enhancements compared to the data
by Choi et al. [6]. Finally, Shaikh et al. [43] independently confirmed the previous results on both the
large magnitude of enhancements and the nonlinear trend in the thermal conductivity of the CNT-PAO
oil nanofluids Choi et al. [6] reported six years ago as shown in Figure 3. This shows that it is neither
straightforward nor speedy to reproduce high-quality nanofluids and their anomalous thermal
properties. Interestingly, Murshed et al. [44], Hong et al. [45] and Chopkar et al. [11] showed that NFs
containing spherical NPs also have strong nonlinear behavior, thereby demonstrating that nonlinear
relationship is not limited to CNTs with a high aspect ratio, with which the new phenomenon was first
discovered.

2.3. Temperature-dependent thermal conductivity

Das et al. [8] were the first to show that nanofluids containing spherical nanoparticles have strongly
temperature-dependent thermal conductivity. They found that the conductivity enhancement of Al2O3 or
CuO nanofluids is two to four times that of the base fluid over a temperature range between 20°C and
50°C. Li and Peterson [10] found that the temperature effect is stronger compared to the data of Das et al.
[8]. This new discovery was confirmed by other research groups [for example, see 7, 46-53]. Temperature
dependent thermal conductivity of nanofluids is significant for engineering applications of nanofluids
because it raises an exciting possibility to develop “smart” nanocoolants that “sense” their thermal
environment and tune their heat conduction property to prevent hot spots. In contrast, other research
groups [33, 35, 54-56] measured thermal conductivity of nanofluids and showed no dependence on

2.8 2.8
Thermal conductivity ratio for CNT Nanofluids
2.6 CNT-PAO by Choi et al. (2001) [6] 2.6
CNT-Water by Xie et al. (2003) [41]
CNT-EG by Xie et al. (2003) [41]
Themral conductivity ratio (keff /kbf)

2.4 2.4
CNT-Decene by Xie et al. (2003) [41]
CNT-Water by Wen and Ding (2004) [42]
2.2 CNT-PAO by Shaikh et al. (2007) [43] 2.2

2.0 2.0

1.8 1.8

1.6 1.6

1.4 1.4

1.2 1.2

1.0 1.0

0.0 0.2 0.4 0.6 0.8 0.10

Volume concentration (%)

Figure 3. Comparison of the thermal conductivity ratio for CNT nanofluids produced by four
groups. Shaikh et al. [Shaikh et al., 2007] independently confirmed the magnitude of
enhancements and nonlinear trend in the thermal conductivity of the CNT-PAO oil nanofluids
Choi et al. reported six years ago.

International Journal of Micro-Nano Scale Transport

Ji-Hwan Lee, Seung-Hyun Lee, Chul Jin Choi, Seok Pil Jang, and Stephen U. S. Choi 275

temperature.

2.4. Size-dependent thermal conductivity

It has been known that the properties of monodisperse suspensions depend on the particle size [57].
However, Chopkar et al. [11] were the first to show experimentally that the effective thermal
conductivity of Al70Cu30 nanofluids strongly depends on the nanoparticle size. This is a significant
feature of nanofluids. In addition, their study shows a nonlinear relationship between the effective
thermal conductivity and the particle size in the particle diameter range of 10–80 nm. Their data clearly
show a steeper size dependence in the small particle size range. Size-dependent thermal conductivity
of nanofluids also has been observed by Chon et al., [46] and Hong et al. [9]. More recently, Kim et al.
[58] showed that the thermal conductivity of nanofluids increases linearly with decreasing particle size
and stated that no existing empirical or theoretical correlation can explain the linear behavior. Strong
size effects in nanofluids are significant in practical applications of nanofluids. In contrast, several
research groups have shown that the thermal conductivity of nanofluids increases with increasing
particle size [for example, see 59, 60].
More recently, Patel et al. [40] measured the thermal conductivity of nanofluids to obtain a
comprehensive dataset for parameters such as the nanoparticle volume fraction and size and
temperature of nanofluids. Their results reconfirmed that the thermal conductivity of nanofluids
displays abnormal enhancement that cannot be predicted by Maxwell model, strong temperature
dependency, and inverse particle size dependency.

2.5. pH-dependent thermal conductivity

Xie et al. [38] and Lee et al. [61] studied the k of nanofluids under varying pH conditions. Xie et al.
[38] are the first to show that the thermal conductivity of Al2O3 nanofluids increases gradually as the
difference between the pH value of nanofluids and the isoelectric point of Al2O3 nanoparticle increases.
Lee et al. [61] found that the k of CuO nanofluids increased by a factor of 3 as pH decreases from the
point of zero charge (PZC) of 8 to 3, indicating that CuO nanofluids have a strongly pH-dependent k.

2.6. Tunable thermal conductivity under external fields

Philip et al. [62, 63] observed that the thermal conductivity of kerosene-based nanofluids containing
6.3 vol.% of magnetite particles has increased up to four times that of the base fluid under an applied
magnetic field. Most recently Shima et al. [64] showed that the ratio of thermal conductivity to
viscosity can be tuned from 0.725 to 2.35 in magnetically controllable nanofluids containing only 0.078
vol.%. These findings demonstrate that magnetically controllable nanofluids can behave like a
multifunctional “smart” material that functions both as a coolant and a damper, offering exciting
applications in microfluidic devices, MEMS/NEMS, and other nanotechnology-based miniature
devices.

2.7. High k nanofluids containing surface-modified nanoparticles and/or no

dispersants
Because the addition of chemical dispersants and stabilizers could adversely influence the properties of
nanofluids, a few research groups produced nanofluids without any dispersants or with surface-modified
nanoparticles [8, 23, 65, 66]. Recently, Yu et al. [67] made stable nanofluids by dispersing plasma-treated
diamond nanoparticles with no surfactants and showed ~20% increase in thermal conductivity at 0.15
vol.% of diamond nanoparticles. This work shows the importance of particle surface treatment. This
work also illustrates that methods for making nanofluids without using any dispersants are preferred,
especially for reference nanofluids. When dispersants and stabilizers are used, their role in altering the
microstructure and properties of nanofluids should be studied.
In summary, the unique features of the thermal conductivity of nanofluids such as anomalously
enhanced thermal conductivity at low particle concentrations, a nonlinear relationship between thermal
conductivity and particle volume fraction, and temperature- and size-dependent thermal conductivity
have been observed in most of the experimental studies carried out over more than 10 years. Nearly all

Volume 1 · Number 4 · 2010

276 A Review of Thermal Conductivity Data, Mechanisms and Models for Nanofluids

of the anomalous features have been confirmed by multiple groups. However, there are other groups
who have observed no anomalous conductivity with their nanofluids. As a result, whether the
anomalous features of the k of nanofluids are real or not has become an issue that has sparked
controversy in the nanofluids community. Considering the various synthesis methods and process
parameter variations and the experimental difficulties in accurately measuring thermal conductivity of
nanofluids, it is understandable that we have this controversy.
The lack of an agreement between experimental data from different groups can be due to differences
in sample quality, the dependence of thermal conductivity on many factors, and differences in
measurement uncertainties. There are several ways to reduce data inconsistencies due to differences in
sample quality. It appears that the primary causes of the significant discrepancies in all nanofluid k data
are vast differences in quality between samples that are supposedly identical or at least similar in
ingredients and concentrations and uncertainty in characterizing nanoparticles including the size, shape,
surface properties, and agglomeration of nanoparticles. Since these characteristics are determined in the
production process, the first step in investigating the causes of these data discrepancies as well as in
evaluating any physical mechanisms and theoretical models is to produce nanofluids of high quality in
terms of stability and thermal conductivity, containing the same concentrations and ingredients, preferably
without any dispersants or stabilizers especially for reference nanofluids, using the same processes for
making nanofluids. When dispersants and stabilizers are used, their role in altering the microstructure and
properties of nanofluids should be understood. Nanofluids produced in this way for thermal conductivity
measurements should then be characterized, including particle size and size distribution, shape, surface
properties, and structures such as levels of agglomeration/clustering of nanoparticles.
In addition to the differences in sample quality, another major cause of the large discrepancies in the
k data is the fact that the k of nanofluids depends on a great number of parameters, some of which are
coupled. Experimental studies have shown that the k of nanofluids is determined by parameters related
to (1) nanoparticles—e.g., concentration, size [7, 38, 46], shape [5, 7, 68—for spherical; 6, 41—for
nonspherical], agglomeration (fractal-like shapes) [8, 23, 26, 44, 61, 69], surface charge [70] and
thermal conductivity; (2) base fluids—e.g., thermal conductivity and viscosity; (3) nanofluids—e.g.,
temperature [7, 8, 10]; (4) the interfacial chemical/physical effect or interaction between the particles
and base fluid [41, 61]; and others. Therefore, for meaningful comparison between measured data and
critical evaluation of models, it is necessary to measure the k of completely characterized nanofluids
including the suspension stability as a function of one variable while the others are fixed. However, this
is challenging as indicated by Bang and Heo [71], who have applied axiomatic design theory [72] to
the design of nanofluids. Some of the parameters are coupled, and ideas to decouple them are needed.
Another way to reduce data inconsistencies due to differences in sample quality and differences in
measurement uncertainties is to conduct round-robin tests using identical test samples and accurate
methods and apparatuses for measuring the thermal conductivity of nanofluids. Recently, an
international nanofluid property benchmark exercise (INPBE) [73] was launched to validate nanofluid
thermal conductivity measurement methods and to resolve the large discrepancies in the k data. The main
findings of the exercise are that the thermal conductivity enhancement was consistent between various
measurement techniques and that the generalized effective medium theory was in agreement with the
measured thermal conductivity data, suggesting that no anomalous enhancement of thermal conductivity
was achieved in the nanofluids they tested [73]. It appears that the nanofluids tested in the exercise are
not considered high quality. For example, the highest average thermal conductivity ratio of 20% was
achieved at 31 vol.% of silica nanoparticles in water. This is much smaller than the theoretical
predictions of the EMT of Maxwell or Bruggeman [74]. In contrast, Lee et al. [75] produced ethylene
glycol-based ZnO nanofluids containing no surfactant by a one-step physical method, conducted round-
robin tests on thermal conductivity measurements of three samples of surfactant-free nanofluids, and
demonstrated that the measured thermal conductivities are beyond the lower and upper bounds
calculated using the models of the Maxwell and Nan et al. [76], with and without the interfacial thermal
resistance. Therefore, it is highly desirable to make high-quality nanofluids with both suspension
stability and high conductivity for use as reference nanofluids as a way to resolve issues such as the

International Journal of Micro-Nano Scale Transport

Ji-Hwan Lee, Seung-Hyun Lee, Chul Jin Choi, Seok Pil Jang, and Stephen U. S. Choi 277

anomalous k behavior of nanofluids.

3. MECHANISMS FOR ENHANCED HEAT CONDUCTION IN NANOFLUIDS

As shown in Figure 1, the classic Maxwell effective medium model that assumes the thermal diffusion
mechanism underpredicts the magnitude of the thermal conductivity enhancement of most of the
nanofluids. Maxwell’s model shows that the thermal conductivity of nanofluids depends only on the
volume fraction of nanoparticles when the volume fraction is low and the thermal conductivity ratio of
nanoparticles to fluid is high. However, the majority of the experimental data described in the previous
section shows that the thermal conductivity of nanofluids depends on a number of other parameters such
as particle size and fluid temperature and acidity. This implies that classical effective medium theories
cannot predict some of the new features of the thermal conductivity of nanofluids. Regarding the
hypothesis that some fundamental mechanisms that are missing in classical effective medium theories
could change the traditional understanding of how heat is conducted, several investigators proposed new
concepts and mechanisms behind the thermal conductivity behavior of nanofluids.
The proposed heat conduction mechanisms in nanofluids can be categorized into static mechanisms
and dynamic mechanisms. Static mechanisms, which assume that nanoparticles are motionless in
nanofluids, include nanolayer, aggregation and percolation, interface thermal resistance, and fractal
geometry. Dynamic mechanisms, which are based on the assumption of randomly moving
nanoparticles in nanofluids, include Brownian motion and nanoconvection. These mechanisms are
developed on a microscopic level. In this section we present the proposed heat conduction mechanisms
in nanofluids from a broader perspective by including microscopic mechanisms such as nanolayer,
aggregation, collision of Brownian particles, nanoconvection and macroscopic mechanisms such as
dual-phase-lagging conduction. The mechanisms of heat conduction in nanofluids are discussed in
more detail in recent reviews by Chandrasekar and Suresh [77], Yu et al. [78], and Wang and Fan [22].

3.1. Microscopic static mechanisms

3.1.1. Nanolayer of ordered liquid molecules at nanoparticle-liquid interface
Yu et al. [79] showed experimentally the existence of an ordered layer of liquid molecules at the solid-
liquid interface that had been predicted for years [80]. This nanolayer structure was introduced by
Keblinski et al. [81] and Yu and Choi [82] as the first static mechanism to explain the enhanced thermal
conductivity of nanofluids. Figure 4 shows that the basic concept of nanolayer. Keblinski et al. [81]
estimated the upper limit for the thermal conductivity enhancement with nanolayer effect by assuming
that the thermal conductivity of the liquid nanolayer is the same as that of the solid nanoparticle. Yu and
Choi [82] proposed the concept that a nanolayer acts as a thermal bridge between a nanoparticle and a
bulk liquid. Based on this new mechanism of a thermal bridge nanolayer, Yu and Choi developed a
renovated Maxwell model for the effective thermal conductivity of nanofluids containing spherical
nanoparticles with an ordered nanolayer. They extended the concept of the thermal bridge nanolayer to
nonspherical particles and renovated the Hamilton-Crosser model [83].

3.1.2. Aggregation of nanoparticles

Clustering or aggregation is an inherent property of nanoparticles whether they are in liquid or dry
powder form due to van der Waals forces. Keblinski et al. [81] conceptualized clustering of
nanoparticle as a mechanism of enhanced k of nanofluids, assuming that clustered nanoparticles
provide local percolation-like paths for rapid heat transport and increase the effective nanoparticle
volume fraction. Wang et al. [84] developed a fractal model for predicting the k of nanofluids
containing nanoparticle clusters. Prasher et al. [85, 86] studied the effects of aggregation on the k of
nanofluids and showed that the aggregation time constant decreases rapidly with decreasing
nanoparticle size and that the k enhancement increases with increasing level of aggregation, leveling
off after the optimum level of aggregation is reached. However, Xuan et al. [69] simulated Brownian
motion and aggregation of nanoparticles and showed that nanoparticle aggregation reduces the k of
nanofluids because the random motion of aggregates is slower than that of a single nanoparticle.

Volume 1 · Number 4 · 2010

278 A Review of Thermal Conductivity Data, Mechanisms and Models for Nanofluids

Figure 4. Single spherical nanoparticle with nanolayer (or interfacial layer) in the base fluid.

3.2. Microscopic dynamic mechanisms

Effective medium theories assume that particles are stationary in a fluid. Thus, effective medium
theories with the microscopic static mechanisms described above cannot explain the k behavior such as
temperature- and particle-size-dependent thermal conductivities of nanofluids. A suspension of
nanosized particles is different from that of micro- or millimeter-sized particles in that the latter are
static and the former is dynamic because nanoparticles are constantly in random motion, even if the
bulk fluid is stationary. So it is expected that there will be a fundamental difference in the mechanisms
of heat transport in nanofluids due to dynamic effects in nanofluids.
There are two kinds of Brownian motion in nanofluids: collision between Brownian nanoparticles
and convection induced by Brownian nanoparticles. Figure 5 shows the basic concept of Brownian-
particle-induced convection.

3.2.1. Collision between Brownian nanoparticles

Collision between nanoparticles due to Brownian motion of nanoparticles is the first dynamic
mechanism studied for enhanced k of nanofluids. However, Wang et al. [24] and Keblinski et al. [81]
have shown that the k enhancement due to collision between Brownian nanoparticles is negligible
because Brownian nanoparticle diffusion is much slower than thermal diffusion.

3.2.2. Nanoscale convection induced by Brownian nanoparticles

Realizing that nanoparticle diffusion has little effect on the k of nanofluids because it is orders of
magnitude slower that thermal diffusion, Jang and Choi [87] proposed the hypothesis that Brownian
nanoparticles induce convection at the nanoscale level and that Brownian-motion-induced
nanoconvection is a key mechanism to explain the temperature- and size-dependent thermal
conductivity of nanofluids. As discussed in section 4.3.1, the Jang and Choi model [87] captures both
the temperature- and size-dependent features of the thermal conductivity of nanofluids whereas EMT-
based models fail to capture any features. Koo and Kleinstreuer [88] extended the concept of
nanoconvection to include the effects of fluids dragged by a pair of nanoparticles and mixing. Prasher
et al. [89] extended the concept of nanoconvection by considering the effect of Brownian-motion-
induced convection from multiple nanoparticles. Patel et al. [90] used nanoconvection induced by
Brownian nanoparticles and the specific surface area of nanoparticles in their micro-convection model.
Ren et al. [91] considered kinetic-theory-based microconvection and liquid layering in addition to
liquid and particle conduction.
However, the nanoconvection mechanism has been questioned. Evans et al. [92] and Vladkov and
Barrat [93], among others, used MDS to show that the contribution of nanoconvection to the thermal
conductivity enhancement is negligible. However, Sarkar and Selvam [94] also used MDS to show that

International Journal of Micro-Nano Scale Transport

Ji-Hwan Lee, Seung-Hyun Lee, Chul Jin Choi, Seok Pil Jang, and Stephen U. S. Choi 279

Convection due Brownian

motion of particle

Figure 5. Microscale convection effect due to the Brownian motion of the particle.

the thermal conductivity enhancement is mostly due to the increased movement of liquid atoms in the
presence of nanoparticles. Li and Peterson [95] simulated the mixing effect of the base fluid directly
adjacent to the nanoparticles and showed that Brownian-motion-induced microconvection and mixing
significantly enhance the macroscopic heat transfer in nanofluids. Eapen et al. [34] showed
experimentally that microconvection does not enhance the thermal conductivity of silica and Teflon
suspensions. Therefore, the debate over the nanoconvection mechanism will continue until it is
resolved experimentally at the nanoscale.

3.3. Macroscopic mechanisms

3.3.1. Dual-phase-lagging conduction
Wang and Wei [96] and Wang and Fan [22] investigated the nature of heat conduction in nanofluids by
developing a macroscopic heat conduction model from first principles. The Wang group proposed dual-
phase-lagging heat conduction as a macroscopic mechanism that comes from the presence of
nanoparticles in nanofluids and showed the possibility of thermal waves and resonances that could
enhance the thermal conductivity of nanofluids more than those observed so far.
In summary, a number of microscale and macroscale mechanisms have been proposed to explain the
magnitude and trend of the thermal conductivity of nanofluids because effective medium theories based
on the underlying assumption of diffusive conduction and motionless nanoparticles cannot predict new
features of the thermal conductivity of nanofluids. New models based on the proposed heat conduction
mechanisms in nanofluids can predict both the magnitude and trend of the thermal conductivity of
nanofluids. However, they cannot accurately predict experimental data. The controversy regarding the
proposed mechanisms of the thermal conductivity of nanofluids is far from over [for example, see 92,
97-99] The controversy comes primarily from a lack of understanding of the scientific basis for the
mechanisms of enhanced thermal conductivity of nanofluids [77, 100]. Therefore, the proposed
concepts and mechanisms behind the thermal conductivity behavior of nanofluids remain to be
validated. More systematic experiments with well-dispersed, well-characterized nanofluids and a better
understanding of the physics of fluid flow and heat transfer at the nanoscale are needed to establish the

Volume 1 · Number 4 · 2010

280 A Review of Thermal Conductivity Data, Mechanisms and Models for Nanofluids

fundamental mechanisms of heat conduction in nanofluids. Understanding these mechanisms is

essential for the development of models that can accurately predict the k behavior of nanofluids.

4. THERMAL CONDUCTIVITY MODELS FOR NANOFLUIDS

There are a number of different types of models developed to predict the newly observed thermal
conduction behavior of nanofluids. In this paper, we have classified thermal conductivity models into
five groups by the main underlying mechanism employed in the models. Five groups of mechanism-
based k models for NFs include those based on (1) classical effective medium theory; (2) nanoscale
layer; (3) Brownian motion; (4) agglomeration; and (5) other mechanisms. Each group of mechanism-
based k models has been further broken into several subcategories.

4.1. Classical effective medium theory based models

The classical models for the thermal conductivity of composite materials are static models with the
assumptions of motionless particles and diffusive heat transfer in both continuous matrix phase and
dispersed phase. The parameters of these static models include the thermal conductivities of the
components; the volume fraction, shape, and distribution of particles; and particle-particle
interaction in dense suspensions. Whereas the classical EMT based models can predict thermal
conductivity of suspensions of particles with a size of a micrometer or larger, they fail to predict
most thermal conductivity data for nanofluids. Nevertheless, a number of EMT-based models have
been used to compare with the k data for nanofluids as described below. However, all of these
models are essentially equivalent because they predict almost identical k for dilute nanofluids.
Therefore, the Maxwell model has been used widely as representative of all classical EMT-based
models.
The simplest model based on the EMT was first developed by Maxwell for a dilute suspension of
non-interacting spherical particles [29]. The Maxwell model [29] and Maxwell-Garnett model [101]
predict well the thermal conductivity of dilute solid-liquid mixtures of relatively large (micro-and
millimeter-size) particles. The effective thermal conductivity can be expressed as

keff = 
( )
 k p + 2 kbf + 2 k p − kbf φ 
 kbf (1)
( )
 k p + 2 kbf − k p − kbf φ 

This mathematical model shows that the effective thermal conductivity of nanofluids depends on
the thermal conductivities of the spherical particle and the base fluid and the volume fraction of the
solid particles.
Bruggeman [74] developed a model based on the symmetrical EMT. For a binary mixture of
homogeneous spherical inclusions, the symmetrical Bruggeman model can be written as

 k p − keff   kbf − keff 

φ  + (1 − φ )  k + 2 k  = 0 (2)
 k p + 2 keff   bf eff 

The Bruggeman model has no limitations on the concentration of spherical particles. For low
particle concentrations, the predictions of thermal conductivity from the Bruggeman model and the
Maxwell model are identical. But as the particle concentration and thermal conductivity ratio increase,
the discrepancy between the two models increases.
Hamilton and Crosser [102] extended the Maxwell model to take into account irregular particle
geometries by introducing a shape factor. When the thermal conductivity of the particles is over 100
times larger than that of the base fluid (kp/kbf > 100) the thermal conductivity can be expressed as

International Journal of Micro-Nano Scale Transport

Ji-Hwan Lee, Seung-Hyun Lee, Chul Jin Choi, Seok Pil Jang, and Stephen U. S. Choi 281

keff = 
(
 k p + ( n − 1) kb − ( n − 1) kbf − k p φ 
 kbf
) (3)
(
 k p + ( n − 1) kbf + kbf − k p φ  )
where n is the empirical shape factor n = 3/ψ, andψ is the particle sphericity, defined as the ratio of the
surface area of a sphere with the same volume as the particle equal to the surface area of the particle.
When the sphericity of the particles is one, the Hamilton-Crosser model reduces to the Maxwell model
for spherical particle mixtures.
Hashin and Shtrikman [103] derived theoretical bounds for the effective thermal conductivity of
suspensions. The Hashin and Shtrikman (HS) bounds can be written as

k p + 2 kbf + 2φ ( k p − kbf ) keff 3k f + 2φ ( k p − kbf ) k p

≤ ≤ (4)
k p + 2 kbf − φ ( k p − kbf ) kbf 3k p − φ ( k p − kbf ) kbf

When the particle thermal conductivity is higher than that of base fluid, the HS lower bound is
consistent with Maxwell model. In this case, the upper bound is derived by simply exchanging the values
for the properties and volume fraction of the particles and the base fluid. Keblinski et al. used the HS
bounds to show that the well-established effective medium theories are capable of explaining nearly all
the published thermal conductivity data for nanofluids without resorting to new mechanisms [104]. As
they mentioned, the HS upper bound corresponds to large pockets of fluid separated by linked chain-
forming or clustered nanoparticles [104]. It describes suspensions with long wire-like structures of
nanoparticles aligned with the heat propagation direction. Such structures of nanoparticles are seldom
realizable with dilute nanofluids containing well-dispersed nanoparticles. Therefore, the upper bound
given by the HS model is not applicable to dilute nanofluids. More realistic upper and lower bounds for
nanofluids having low concentration of well-dispersed nanoparticles are given by Buongiorno et al. [36].
Using the classical Maxwell model without thermal resistance as an upper bound of the thermal
conductivity of dilute nanofluids, it is clear from Figure 1 that the thermal conductivity enhancements
of most nanofluids are larger than the upper bound of the Maxwell model. Therefore, models based on
new mechanisms as well as the classical diffusion mechanism are necessary to predict thermal
conductivity data for nanofluids.
Jeffrey [37] developed a theoretical model assuming the conduction of heat through a stationary and
random and statistically homogeneous suspension of spherical particles in a matrix of uniform
conductivity. Jeffrey [37] also assumed that the volume fraction of particles is small. His mathematical
model can be written as

  3β 3 9β 3 χ + 2 3β 4 
keff = 1 + 3βφ + φ 2  3β 2 + + + 6 + ⋅ ⋅ ⋅  kbf (5)
  4 16 2 χ + 3 2  

with
kp k p − kbf χ −1
χ= and β= = (6)
kbf k p + 2 kbf χ+2

Davis [105] developed a theoretical model for calculating the effective thermal conductivity of
a composite material containing spherical inclusions. He assumed that the surface of a large heated
body kept at a uniform temperature is in contact with a composite material of infinite extent that
has a lower temperature far from the heated body. He used Green’s theorem to compute the rate of
heat transfer from the heated body to the composite material. The resulting formula is

Volume 1 · Number 4 · 2010

282 A Review of Thermal Conductivity Data, Mechanisms and Models for Nanofluids

 3 (C − 1) 
keff = 1 +
 {C + 2 − (C − 1) φ }
{
φ + f (C ) φ 2 + O φ 3  kbf

( )} (7)

with
∞
( )
f (C ) = ∑ p = 6  B p − 3 Ap / ( p − 3) 2 p − 3  (8)

where C is the ratio of the thermal conductivity of the spherical inclusions to that of base fluid and Ap
and Bp are the known functions of C. This model is more relevant than prior works, since higher-order
terms are taken into consideration [105].
Hasselman and Johnson [106] derived an expression for the effective thermal conductivity of
composites, taking into account the thermal barrier resistance at the interface between the materials
and the relations for insertion shapes for spherical, cylindrical, and flat plate for low concentration
of dispersions. The resulting expression for spherical particles can be arranged as

 k p (1 + 2κ ) + 2 kbf + 2φ ( k p (1 − κ ) − kbf ) 
keff =   kbf (9)
 k p (1 + 2κ ) + 2 kbf − φ ( k p (1 − κ ) − kbf ) 

where κ is a dimensionless parameter defined as

aK
κ= (10)
ap

where aK is the so-called Kapitza radius defined as

aK = RK kbf (11)

where RK is the Kapitza resistance, or thermal boundary resistance. Hasselman and Johnson found that
the effective thermal conductivity depends on the volume fraction of the dispersed phase as well as the
dispersion size.
Nan et al. [76] introduced a methodology for predicting the effective thermal conductivity of
composites of arbitrarily shaped inclusions with the Kapitza resistance, using an effective medium
approach. Their model can be arranged as

keff,11 = keff,22 =  11 11 (
 2 + φ  β (1 − L ) 1 + cos2 θ
 )+ β 33 (1 − L33 ) (1 −cos2 θ  
 k )
 bf (12)

  (
2 − φ  β11 L11 1 + cos2 θ )+ β 33 L33 (1 − cos2 θ 
) 


and

keff,33 =
11 (
1 + φ  β (1 − L ) 1 − cos2 θ
  11 )+ β 33 (1 − L33 ) coss2 θ  
 k
 bf (13)

  (
1 − φ  β11 L11 1 − cos2 θ )+ β 33 L33 cos2 θ 
 


with

International Journal of Micro-Nano Scale Transport

Ji-Hwan Lee, Seung-Hyun Lee, Chul Jin Choi, Seok Pil Jang, and Stephen U. S. Choi 283

∫ ρ (θ ) cos θ ⋅ sin θ dθ
2
kiic − kbf
βii = , cos θ
2
= (14)
(
kbf + Lii kiic − kbf ) ∫ ρ (θ ) sin θ dθ
and

kiic = ks
( )
ks + Lii k p − ks (1 − v ) + v k p − ks ( ), v=
a12 a3
(15)
(
ks + Lii k p − ks (1 − v ) ) ( a1 + δ )2 ( a3 + δ K )

The model is applicable for predicting the effects of particle shape, size, orientation, distribution,
volume fraction, and interfacial thermal resistance on the thermal conductivity of composites
containing arbitrarily shaped inclusions.
Table 1 summarize the classical effective medium theory based models for the thermal conductivity
of nanofluids and shows the model equations and key parameters required for determining thermal
conductivity and remarks.

4.2. Nanolayer based models

We categorize nanolayer based models into three submodels: theoretical models, combined models, and
computational models.

4.2.1. Theoretical models

Perhaps the earliest example of a nanoscale structure-based model for k of nanofluids is found in
the work of Yu and Choi [82]. Accounting for the concept of the nanolayer, Yu and Choi [82]
modified the Maxwell model to predict the effective thermal conductivity of nanofluids. Two
essential features of the Yu and Choi model are (1) consideration of a new nanoscale structure that
has not been considered in classical EMT models and (2) a new assumption that the solid-like
nanolayer acts as a thermal bridge between a solid nanoparticle and a bulk liquid in nanofluids, an
idea contrary to that in nanocomposite solids, has been made. In order to include the effect of the
nanolayer, Yu and Choi assumed that (1) the solid-like layer of thickness δ around the particles is
more ordered than that of the bulk liquid; (2) the thermal conductivity of ordered layer klayer is thus
higher than that of the bulk liquid; (3) the nanolayer around each particle could be combined with
the particle to form an equivalent particle; and (4) the particle volume fraction is so small that there
is no overlap of equivalent particles. Based on the effective medium theory and assumptions (3) and
(4), the equivalent thermal conductivity kpe of the equivalent particles can be expressed as

 2 (1 − γ ) + (1 + β )3 (1 + 2γ )  γ
k pe =  k (16)
− (1 − γ ) + (1 + β ) (1 + 2γ )
3 p

where γ = klayer /kp is the ratio of nanolayer thermal conductivity to particle thermal conductivity, and
β = δ/ap is the ratio of the nanolayer thickness to the original particle radius. With Equation (16),
the Maxwell model can be modified into

keff =
( )
k pe + 2 kbf + 2 k pe − kbf (1 − β ) φ
3

kbf (17)
k pe + 2 kbf − ( k pe − kbf ) (1 + β ) φ
3

This model reduces to the Maxwell model when γ = 1 or β = 0. The modified Maxwell model

Volume 1 · Number 4 · 2010

 Table 1. Summary of classical effective medium theory based models
284

Investigator Formulation Key parameter Remarks

Maxwell (1873),  k + 2k + 2 k − k φ  φ, kp, kbf Initial study of effective

p bf p bf ( )
Maxwell-Garnett keff =  k medium theory (EMT)
 k + 2 k − k − k φ  bf
(1904)  p bf p bf (
 )
Bruggeman  k p − keff   kbf − keff  φ, kp, kbf EMT with interparticle
(1935) φ  + (1 − φ )  k + 2 k  = 0 interaction
 k p + 2 keff   bf eff 

φ, kp, kbf
Hamilton and  k p + ( n − 1) kb − ( n − 1) kbf − k p φ  ( ) EMT with shape factor
Crosser (1962)
keff =  kbf 3
 k p + ( n − 1) kbf + kbf − k p φ  ( ) n=
ψ

Hashin and k p + 2 kbf + 2φ ( k p − kbf ) keff 3k f + 2φ ( k p − kbf ) k p Theoretical bounds

Shtrikman ≤ ≤ φ, kp, kbf of EMT
k p + 2 kbf − φ ( k p − kbf ) kbf 3k p − φ ( k p − kbf ) kbf
Effective (1962)
medium Jeffrey (1973) Second order
models   3β 3 9β 3 χ + 2 3β 4  φ, kp, kbf term, EMT with
keff = 1 + 3βφ + φ 2  3β 2 + + + 6 + ⋅⋅ ⋅  kbf
  4 16 2 χ + 3 2   interparticle
interaction,
Davis (1986) φ, kp, kbf EMT with interparticle
 3 C −1 ( ) 
keff = 1 + {φ + f ( C φ 2 + O φ 3  kbf
) ( )} interaction
 C + 2 − C −1 φ
{ ( )} 
 
Hasselman and φ, kp, kbf EMT with interfacial
Johnson (1987)  k p (1 + 2κ ) + 2 kbf + 2φ ( k p (1 − κ ) − kbf ) 
kbf RK thermal resistance
keff =  kbf κ=
 k p (1 + 2κ ) + 2 kbf − φ ( k p (1 − κ ) − kbf )  ap (ITR)
Nan et al. (1997) φ, kp, kbf Generalized form of
 2 + φ β 1 − L 1 +
 ( 11 ( Lii: geometrical factor EMT with ITR and
)( cos 2 θ + β33 1 − L33 1 − cos 2 θ  
) )( )
 11  k
keff ,11 = keff , 22 =   bf shape factor
 
2 − φ β11 L11 1 + ( cos 2 θ + β33 L33 1 − cos 2 θ 
) (  )
   
1 + φ  β 1 − L 1 − cos 2 θ 2  
 ( 11 ( )
)( ) + β33 1 − L33 cos θ 
 11  k
keff ,33 =   bf

International Journal of Micro-Nano Scale Transport

A Review of Thermal Conductivity Data, Mechanisms and Models for Nanofluids

 1 − φ  β11 L11 1− cos 2 θ

( ) + β33 L33 cos 2 θ  
   
Ji-Hwan Lee, Seung-Hyun Lee, Chul Jin Choi, Seok Pil Jang, and Stephen U. S. Choi 285

including the effect of nanolayer predicts that the very thin nanolayer has a significant impact on the
k of nanofluids, particularly when the particle diameter is less than 10 nm. However, the nanolayer
impact is small, and the modified Maxwell formula reduces to the original Maxwell equation in the
case of large particles (ap >> δ, β → 0). However, the thickness and thermal conductivity of nanolayer
were not obtained theoretically.
Yu and Choi [82, 83] extended Equation (17), the nanoscale structural model for spherical
nanoparticles, to nonspherical particles and renovated the Hamilton-Crosser model. The nanolayer is
expressed as a confocal ellipsoid with a solid particle. They chose ellipsoidal particles since the
geometric parameters characterizing them can be changed to represent the various shapes of particles
actually used in composites. They also developed a generalized empirical shape factor for anisotropic
complex ellipsoids. Their renovated Hamilton-Crosser model is expressed in terms of the equivalent
thermal conductivity and equivalent volume fraction of anisotropic complex ellipsoids and includes an
empirical shape factor. It can be written as

 mφe A 
keff =  1 + kbf (18)
 1 − φe A 

where the parameter A is defined by

1 ( k pj − kbf )
A=
3
∑ k pj + ( m − 1) kbf
(19)
j = a, b,c

and

φe =
( a2 + δ ) ( b2 + δ ) (c2 + δ ) φ (20)
abc

is the equivalent volume concentration of complex ellipsoids. The equivalent thermal conductivities
along the axes of the complex ellipsoid kpj can be expressed as

 k p − ks 
k pj = 1 +  ks (21)
 k p rd ( j, 0 ) − d ( j, t )  − ks rd ( j, 0 ) − d ( j, t ) − r  

where j(= a, b, and c) is along the semiaxis directions of the ellipsoid, kp and ks are the thermal
conductivities of the solid ellipsoid and its surrounding layer, r is the volume ratio defined by

r=
( a2 + δ ) ( b2 + δ ) (c2 + δ ) (22)
abc

and d(j, v) is the depolarization factor defined by

Volume 1 · Number 4 · 2010

286 A Review of Thermal Conductivity Data, Mechanisms and Models for Nanofluids

d ( j, v ) =
( a2 + v ) ( b2 + v ) (c2 + v )
2
∞ dw (23)
d ( j, v ) = × ∫0
( j 2 + v + w ) ( a2 + v + w ) + ( b2 + v + w ) + (c2 + v + w )
The generalized empirical shape factor m (m = 3ψ −ε, where ε is an empirical parameter and ψ is the
particle sphericity of the complex prolate spheroid particle) is only one adjustable parameter in this
model. This model was used to predict the effective thermal conductivity of nanotube-in-oil
suspensions. With the empirical factor m, this model showed that the solid and liquid interfacial layers
play an important role in correctly predicting the magnitude of the thermal conductivity enhancement
of nanotube-in-oil nanofluids.
However, the two static models developed by Yu and Choi are not able to predict the nonlinear
behavior of the effective thermal conductivity of nanofluids. Xue [107] is the first to model that aspect
of nanofluid thermal conductivity. Although the models based on liquid layering theory predict well the
measured thermal conductivity data, both the thickness and conductivity of the liquid layer have to be
assumed.
Xue [107] developed a model to include the effect of a nanolayer, based on Maxwell theory and average
polarization theory. He assumed that the complex nanoparticle is composed of an elliptical nanopaticle of
thermal conductivity kp with half-radii of (a, b, c) and an elliptical shell of thermal conductivity klayer with
a thickness δ. The resulting formula for the effective thermal conductivity is

 φ keff − kbf φ keff − kc, x keff − kc, y 

9  1 −  +  +4 =0
( ) )( )
(24)
 λ  2 keff + kbf κ  keff + B2, x kc, x − keff
 2 keff + (1 − B2, x kc, y − keff 

where λ = abc/[(a + δ)(b + δ)(c + δ)], kc,j is the effective dielectric constant, and B2,x is the
depolarization factor along the x-symmetrical axis, which is derived from the average polarization
theory. This model is notable in that it can predict for the first time the nonlinear behavior of the
nanotube-in-oil nanofluids. The model predictions agree well with the experimental data using
empirical parameters for liquid layer thickness and conductivity value. However, Yu and Choi [83] and
Kim et al. [108] have shown that the values of the model parameters such as the depolarization factor
are incorrect.
Xue and Xu [109] developed a new equation for the effective thermal conductivity of nanofluids by
modifying the Bruggeman model with the effective thermal conductivity of the so-called “complex
nanoparticles” consisting of a nanoparticle and its surrounding shell. Their modified Bruggeman model
can be expressed as

 φ  keff − kbf ( )( ) ( )(
φ keff − klayer 2 klayer + k p − χ k p − klayer 2 klayer + keff )
 1 − χ  2 k + k + χ =0 (25)
eff bf ( )( ) ( )(
2 keff + klayer 2 klayer + k p + 2 χ k p − klayer klayer − keff )
where χ is the volume ratio of spherical nanoparticles to complex nanoparticles, defined as

3
 ap 
χ=  (26)
 ap + δ 

International Journal of Micro-Nano Scale Transport

Ji-Hwan Lee, Seung-Hyun Lee, Chul Jin Choi, Seok Pil Jang, and Stephen U. S. Choi 287

where ap is the radius of a particle and δ is the thickness of interfacial shell. In Equation (25), kp and
klayer are the thermal conductivity of the nanoparticle and interfacial shell, respectively. The modified
model is in good agreement with the experimental data on the effective thermal conductivity of
CuO/water and CuO/EG nanofluids.
Xie et al. [110] derived a new expression for calculating the effective thermal conductivity of
nanofluids considering a linear thermal conductivity distribution across the interfacial nanolayer. Their
model is given as

 3Θ 2φT2 
keff =  1 + 3ΘφT + kbf (27)
 1 − ΘφT 

with

 klayer − kbf  ( )(
 k p − klayer kbf + 2 klayer ) 
 (1 + γ ) − 
3
k
 layer + 2 kbf   ( )(
 k p + 2 klayer kbf − klayer ) 
Θ= (28)
 k −k  k −k 
(1 + γ )3 + 2  layer bf   p layer 
 klayer + 2 kbf   k p + 2 klayer 

where φT is the total volume fraction of the original nanoparticle and nanolayer defined as φT = φ(1 + γ)3
and γ is the ratio of the nanolayer thickness to the original particle radius defined as γ = δ/ap. Their model
shows the effects of nanolayer thickness, nanoparticle size, volume fraction, and thermal conductivity
ratio of particle to fluid on the thermal conductivity. Moreover, for smaller particles, the effects of
particle size and nanolayer thickness become much more conspicuous than for larger ones, which shows
that controlling nanolayer structure might be an effective method to yield high thermal conductivity of
nanofluids.
Leong et al. [111] proposed a new model for the effective thermal conductivity of nanofluids by
considering the effect of the interfacial layer at the solid-liquid interface. Their new model can be
written as

( k p − klayer ) φ klayer  2β13 − β 3 + 1 + ( k p + 2 klayer ) β13 φβ 3 ( klayer − kbf ) + kbf 
keff = (29)
β13 ( k p + 2 klayer ) − ( k p − klayer ) φ  β13 + β 3 − 1

where β is defined as β = 1 + γ, β1 is defined as β1 = 1 + γ /2, and here γ is ratio of the nanolayer

thickness to the original particle radius defined as γ = δ /ap. This model includes the effects of
nanoparticle size, nanolayer thickness, volume fraction, and thermal conductivities of the
nanoparticle, nanolayer, and the base fluid. If there is no nanolayer at the solid-liquid interface, i.e.,
klayer = kbf and δ = 0 or β1 = β = 1, this model reduces to Maxwell-Garnett model. It appears that their
model predictions of the thermal conductivity enhancements are more consistent with the
experimental results than those of the classical Maxwell theory and the renovated Maxwell model
[82]. However, Doroodchi et al. [112] have demonstrated that the model of Leong et al. [111] is based
on an incorrect derivation.
Murshed et al. [48] proposed two models for the effective thermal conductivity of nanofluids
containing spherical and cylindrical nanoparticles by considering the effect of the interfacial layer
at the solid particle/liquid interface. For spherical nanoparticles, the model can be expressed as

Volume 1 · Number 4 · 2010

288 A Review of Thermal Conductivity Data, Mechanisms and Models for Nanofluids

( k p − klayer ) φ klayer  2β13 − β 3 + 1 + ( k p + 2 klayer ) β13 φβ 3 ( klayer − kbf ) + kbf 
keff = (30)
β13 ( k p + 2 klayer ) − ( k p − klayer ) φ  β13 + β 3 − 1

and for cylindrical nanoparticles

( k p − klayer ) φ klayer  β12 − β 2 + 1 + ( k p + klayer ) β12 φβ 2 ( klayer − kbf ) + kbf 
keff =
β12 ( k p + klayer ) − ( k p − klayer ) φ  β12 + β 2 − 1
(31)

where β is defined as β = 1 + γ, β1 is defined as β1 = 1 + γ / 2, and here γ is ratio of the nanolayer

thickness to the original particle radius, defined as γ = δ /ap. Equation (30) is the same as the model
developed by Leong et al. [111]. These two models include that the effect of the nanoparticle size,
nanolayer thickness, volume fraction, and thermal conductivities of the nanoparticle, nanolayer, and
the base fluid. For the calculation of the thickness of the nanolayer at the nanoparticle/liquid
interface, they used the following model by Hashimoto et al. [113]

δ = 2πσ (32)

where σ is a parameter that characterizes the diffuseness of interfacial boundary whose typical value is
within 0.2 ∼ 0.8 nm. However, this model is not able to predict the nonlinear behavior of the effective
thermal conductivity of nanotube-based nanofluids.
Doroodchi et al. [112] derived mathematical models for the effective thermal conductivity of
nanofluids using spatial-averaging and point-source methods. The model derived using the point-
source method of Maxwell [29] is given as

keff =

(
 38φ a3p δ 3 − 6a 4p δ 2 − 2a3p δ 3 + 6φ a5p δ − 6a5p δ + 24φ a 4p δ 2 + 12φ a p δ 5 + 30φ a 2p δ 4 + 2φδ 6 kbf k p

) 

 (
 − 3a 4 δ 2 + 6φ a6 + 3a5 δ + 30φ a 2 δ 4 + a3 δ 3 + 44φ a3 δ 3 + 2φ a6 + 12φ a δ 5 + 24φ a5 δ + 42φ a 4 δ 2 + 3a6 k
p p p p p p p p p p p )
layer k p


 
(
 − 4 a p δ − 24φ a p δ + 12a p δ − 6φ a p + 6a p + 12a p δ − 82φ a p δ − 66φ a p δ − 4φδ − 60φ a p δ − 30φ a p δ klayer kbf 
3 3 5 5 6 6 4 2 3 3 4 2 6 2 4 5
)
 
(
 − 12φ a p δ + 60φ a p δ + 24φ a p δ + 76φ a p δ + 6a p δ + 6a p δ + 2a p δ + 4φδ + 48φ a p δ klayer
5 2 4 5 3 3 4 2 5 3 3 6 4 2 2
) 
kbf

(
 −19φ a3p δ 3 − 6a 4p δ 2 − 2a3p δ 3 − 3φ a5p δ − 6a5p δ − 12φ a 4p δ 2 − 6φ a p δ 5 − 15φ a 2p δ 4 − φδ 6 kbf k p ) 

 (
 + −3a 4 δ 2 + 3φ a6 − 3a5 δ + 15φ a 2 δ 4 − a3 δ 3 + 22φ a3 δ 3 + φ a6 + 6φ a δ 5 + 12φ a5 δ + 21φ a 4 δ 2 − 3a6 k
p p p p p p p p p p p )
layer k p


 
 + (−12 φ a p δ 5
− 12 a 5
p δ − 4 a p δ
3 3
− 3 φ a 6
p − 6 a 6
p − 12 a p δ
4 2
− 41φ a p δ
3 3
− 33 φ a p δ
4 2
− 2φδ 6
− 30φ a p δ
2 4
− 1 5φ a 5
p δ k)layer k bf 
 
(
 + 38φ a p δ − 6a p δ − 2a p δ + 6φ a p δ − 6a p δ + 24φ a p δ + 12φ a p δ + 30φ a p δ + 2φδ klayer
3 3 4 2 3 3 5 5 4 2 5 2 4 6 2
) 

(33)

Doroodchi et al. [112] then used their models to evaluate the nanolayer based models of Yu and Choi
[82] and Leong et al. [111] and demonstrated that the renovated Maxwell model of Yu and Choi [82]
is identical to their models, but the model of Leong et al. [111] is based on an incorrect derivation.
Therefore, the renovated Maxwell model accounts for the effects of nanolayering accurately, whereas
the Leong et al. [111] model overstates them. However, their models and the renovated Maxwell model
[82] predict effective thermal conductivity enhancements that are not significantly greater than those

International Journal of Micro-Nano Scale Transport

Ji-Hwan Lee, Seung-Hyun Lee, Chul Jin Choi, Seok Pil Jang, and Stephen U. S. Choi 289

predicted by classical Maxwell theory. This implies that nanolayering by itself is unable to account for
the effective thermal conductivity enhancements in nanofluids.

4.2.2. Combined models

Ren et al. [91] derived a new theoretical model for the effective thermal conductivity of nanofluids,
which includes the effects of interfacial nanolayer and microconvection owing to the thermal
motion of nanoparticles. They assumed that there are four heat transfer modes in nanofluids: (1)
heat transfer by the base fluid; (2) heat transfer by the nanoparticles; (3) heat transfer by the
nanolayer; and (4) heat transfer by microconvection. Their new theoretical model can be expressed
as

 3Θ 2φT2 
keff =  1 + F ( Pe ) + 3ΘφT + kbf (34)
 1 − ΘφT 

where Θ is the function defined by [110] as

 klayer − kbf  ( )(
 k p − klayer kbf + 2 klayer ) 
 (1 + γ ) − 
3
k
 layer + 2 kbf   ( )(
 k p + 2 klayer kbf − klayer ) 
Θ= (35)
 k −k  k −k 
(1 + γ )3 + 2  layer bf   p layer 
 klayer + 2 kbf   k p + 2 klayer 

and F(Pe) is given as

F ( Pe ) = 0.0556 ( Pe ) + 0.1649 ( Pe )2 − 0.0391 ( Pe )3 + 0.00034 ( Pe )4 (36)

where Pe is defined as

uL 0.75
Pe = φ (37)
α bf T

In Equation (37), u– is the mean velocity of the complex nanoparticle, L is the specific length, αbf is
the thermal diffusivity of the base fluid, and φT is the total volume fraction of the original nanoparticle
and nanolayer defined by Xie et al. [110]. The mean velocity of the complex nanoparticle u– is defined
as

ρ  
3
 
u=
3k B T 4  layer
where mc = ρ pπ a3p   1 + δ  − 1 + 1 (38)
mc 3  ρ p 

ap   
 

where mc is the mass of the complex nanoparticle, rp is the density of the nanoparticle, rlayer is the
density of the nanolayer, and δ is the thickness of the interfacial nanolayer. Now, the specific length L
is defined as

Volume 1 · Number 4 · 2010

290 A Review of Thermal Conductivity Data, Mechanisms and Models for Nanofluids

L = ap + δ( ) 3 34φπ (39)
T

This model implies that the parameters of a nanofluid system, such as the nanoparticle size,
nanolayer thickness, temperature, volume fraction, and thermal conductivities of the nanoparticle and
the base fluid, play important roles in the enhanced thermal conductivity ratios. However, they did not
mention a methodology for determining the value of the density of the nanolayer, ρlayer.
Sabbaghzadeh and Ebrahimi [114] proposed a theoretical model for the effective thermal
conductivity of nanotubes (cylinder-shaped particles) for use in nanotubes-in-fluid suspensions. They
considered five modes of energy transport in nanofluids: (1) collision among the base fluid molecules;
(2) thermal diffusion in nanoparticles, which are covered by the nanolayer in the fluid; (3) thermal
diffusion in the nanolayer in the fluid; (4) thermal interaction of dynamic complex nanoparticles
(intrinsic nanoparticles and nanolayers) with the base fluid molecules; and (5) collision between
nanoparticles due to Brownian motion. They neglected the last term because it is a very slow process.
Their theoretical model can be expressed as

(
keff = kbf [1 − φ (1 + M ′ )] + φ k p + klayer M ′ )
(40)
( 0.35 + 0.56 Re 0bf.52 ) Prbf0.3 kbf
abf
+ φ (1 + M ′ )
(
Prbf 2a p +δ)

where

 δ 
2

M ′ =  + 1 − 1 (41)
 a p  
 

Their theoretical model shows that the effect of nanolayer thickness will be more important for small
diameter of nanotubes.
Murshed et al. [115] proposed a combined static and dynamic mechanisms-based model for the
effective thermal conductivity of nanofluids. Their model included the effects of particle size,
nanolayer, Brownian motion of nanoparticle, particle surface chemistry, and interaction potential. The
model can be expressed as

keff = 
(  )  (  )
 φω k p − ω kbf  2β13 − β 3 + 1 + k p + 2ω kbf β13 φβ 3 (ω − 1) + 1 
k

 ( ) ( 
β1 k p + 2ω kbf − k p − ω kbf φ  β1 + β − 1
3 3 3
)
 

bf

  2 3Λ 2 9 Λ3 kcp + 2 kbf 3Λ 4 
+ φ 2 β 6 kbf  3 Λ + + + 6 + ⋅ ⋅ ⋅ 
  4 16 2 kcp + 3kbf 2  

 1
+ ρcpc p,cp Ls 
B (
 3k T 1 − 1.5β 3φ

+
GT



) (42)
2 2πρcp β a p
3 3
6πµβ a L
  p s
 

with

International Journal of Micro-Nano Scale Transport

Ji-Hwan Lee, Seung-Hyun Lee, Chul Jin Choi, Seok Pil Jang, and Stephen U. S. Choi 291

klayer = ω kbf , Λ=
kcp − kbf
, kcp =
( ) (
2 k p − klayer + β 3 k p + 2 klayer )k (43)
kcp + 2 kbf ( klayer − k p ) + β ( k p + 2klayer ) layer
3

where ω is an empirical parameter that depends on the orderliness of fluid molecules in the interface
and the nature and surface chemistry of nanoparticles; kcp is the thermal conductivity of the complex
particle; β is defined as β = 1 + γ, β1 is defined as β1 = 1 + γ /2, where γ is the ratio of the nanolayer
thickness to the original particle radius defined as γ = δ /ap; and GT is the total interparticle potential,
or DLVO interaction potential. The significant features of their model are as follows: (1) The model is
developed by considering nanoparticles with a thin interfacial layer together with their static and
dynamic mechanisms in the base fluid. The particle size effect is also included; (2) The second term on
the righthand side of Equation (42) represents the interactions between pairs of spherical nanoparticles
in a stationary suspension; (3) The third term on the righthand side of Equation (42), which takes into
account the effect of particle Brownian motion, particle surface chemistry, and interparticle
interactions, is applicable for φ > 0.005. As mentioned before, this is because at such a small volume
fraction, i.e. φ < 0.005, the interparticle separation distance is too large to cause any interactions
through the Brownian and potential forces of particles; (4) If there are no interaction between pairs of
nanoparticles and the interfacial layer, the static part of the model reduces to the Maxwell model and
when φ = 0, the entire model reduces to kbf. The analysis of their results shows that the major
contribution to the enhanced thermal conductivity of nanofluids arises from static mechanisms.

4.2.3. Computational models

Xue et al. [97] investigated the effect of the liquid structure at the solid interface on thermal transport
by using non-equilibrium molecular dynamics simulations of a simple monoatomic liquid system. They
demonstrated that the layering of the liquid atoms at the liquid-solid interface, both in the directions
normal and parallel to the solid–liquid interface, shows minor effect on thermal transport properties.
However, the results of their MD simulations for a simple monoatomic liquid system may not apply to
complex systems like nanofluids.
Li et al. [116] investigated the effect of the molecular layering at liquid/solid interface on the
thermal conductivity of nanofluids using an equilibrium molecular dynamics simulation. They used
the Green-Kubo relationships to calculate the thermal conductivity of nanofluids. Focusing on particle
movement and surrounding liquid atoms [116], they found that a solidlike, thin liquid layer is formed
at the interface between the nanoparticle and liquid, and this nanolayer moves with the nanoparticles
due to Brownian motion. They also found that more liquid atoms were attracted to form the liquid
layer around the nanoparticle when the size of the nanoparticle is larger. This implies that the number
of the atoms in the liquid layer is related to the nanoparticle size. Thus, as shown by these phenomena,
the nanolayer can contribute to the anomalous enhancement of the thermal conductivity of nanofluids.
Based on their simulation results, they conclude that the thermal conductivity of nanofluids can be
significantly enhanced when the thermal conductivity of the nanolayer is larger than that of liquid and
the nanoparticle size increases.
The deficiency of nanolayer-based models is that there exists no method to estimate exactly the
nanolayer thickness and thermal conductivity. Lee [70] introduced the concept of an electrical double
layer (EDL), which is formed on the particle surface in solution and developed equations for the
thickness and effective thermal conductivity of the nanolayer. Tillman and Hill [117] reported a
method to determine the nanolayer thickness and thermal conductivity profile within the nanolayer.
They used a function form to derive the thermal conductivity of the nanolayer, but the function cannot
accurately determine the thermal conductivity. Thus, a method of determining
the nanolayer thickness accurately should be settled first in the future to build a successful theory for
the effective thermal conductivity of nanofluids using the nanolayer.
Table 2 provides a summary of the nanolayer based models for the thermal conductivity of

Volume 1 · Number 4 · 2010

292 A Review of Thermal Conductivity Data, Mechanisms and Models for Nanofluids

nanofluids.

4.3. Brownian motion based models

We categorize Brownian motion based models into three sub-models: theoretical models,
computational models, and empirical models.

4.3.1. Theoretical models

Xuan et al. [69] are the first to develop a dynamic model that takes into account the effects of Brownian
motion of nanoparticles. However, their model did not predict the linear increase of conductivity with
temperature, as obtained by Das et al. [8]. Furthermore, their thermal conductivity equation is
dimensionally incorrect.
Jang and Choi [87] developed a theoretical model that takes into account nanoconvection. They
considered four modes of energy transport in nanofluids: (1) collision between base fluid
molecules; (2) thermal diffusion in nanoparticles in fluids; (3) collision of nanoparticles with each
other by translational motion due to the Brownian motion with long wavelength; (4) Brownian
motion of nanoparticles with short wavelength, which results from collisions between base fluid
molecules and nanoparticles by thermally induced fluctuations. They performed an order-of-
magnitude analysis and neglected the third mode because it is much smaller than the other modes.
Thus, they derived a general expression for the effective thermal conductivity from modes (1), (2),
and (4). The resulting model can be expressed as

d bf
keff = kbf (1 − φ ) + β k pφ + C1 kbf Re d2 p Pr φ (44)
dp

where β is a constant for considering the Kapitza resistance per unit area, C1 is a proportional constant,
and the Reynolds number is defined by

C R.M.d p (45)
Re d p =
ν
–
where C R.M. and ν are the random motion velocity of a nanoparticle and kinematic viscosity of base
fluid, respectively. The random motion velocity can be defined as

Do
C R.M. = (46)
lbf

where Do is the diffusion coefficient and lbf is the liquid mean free path. New model calculations agree
with temperature-dependent data of Das et al. [8] while Maxwell-type theories fail to capture the
temperature-dependent conductivity. Furthermore, their new model predicts size-dependent
conductivity. However, the hypothesis of the existence of convection at the nanoscale has not been
established experimentally or theoretically and the random motion velocity of a single nanoparticle has
not been measured to confirm that it can be expressed as a function of the diffusion coefficient Do and
the liquid mean free path lbf.
Kumar et al. [118] developed a unique model for effective thermal conductivity of nanofluids
composed of two submodels: (1) a stationary particle model and (2) a moving particle model. The
stationary particle model was developed based on Fourier’s law of diffusion. In the moving particle
model, they adapted the concept of the kinetic theory of gases for effective thermal conductivity of
particles and the Stokes-Einstein formula. The proposed combined model can be expressed as

International Journal of Micro-Nano Scale Transport

 Table 2. Summary of models considering nanolayer effect (1/2).

Investigator Formulation Key parameter Remarks

Yu & Choi (2003) 3 φ, kp, kbf, klayer, δ Modified Maxwell
k pe + 2 kbf + 2 k pe − kbf
( ) (1 − β ) φ k
keff = 3 bf β = δ/ap model with nanolayer
k pe + 2 kbf − ( k pe − kbf ) (1 + β ) φ
(NL) effect
Yu & Choi (2004) φ, kp, kbf, klayer, δ Modified HC model
 mφe A 
keff =  1 + kbf φe: equivalent vol. % with NL effect and
 1 − φe A 
shape factor
Xue and Xu φ, kp, kbf, klayer, δ Modified Bruggeman

Volume 1 · Number 4 · 2010

 φ  keff − kbf
(2005) model with NL effect
 1 − χ  2 k + k +
eff bf and shape factor
φ keff − klayer 2 klayer + k p − χ k p − klayer 2 klayer + keff
( )( ) ( )( ) 3
=0  ap 
χ 2 keff + klayer 2 klayer + k p − 2 χ k p − klayer klayer − keff
( )( ) ( )( ) χ= 
 ap + δ 

Xue (2003) φ, kp, kbf, klayer, δ Maxwell model with

 φ keff − kbf φ keff − kc,x
Theoretical 9  1 −  +  NL and depolarization
 
λ 2 keff + kbf κ  keff + B2,x kc,x − keff
models  ( ) B2,x: depolarization factor
keff − kc, y  factor
+4 =0
2 keff + 1 − B2,x
( ) ( kc,y − keff ) 

Xie et al. (2005) φ, kp, kbf, klayer, δ Linear thermal
 3Θ 2 φT2 
keff =  1 + 3ΘφT + kbf conductivity
 1 − ΘφT  φT = φ(1 + δ/ap)3 distribution assumption
with NL
Leong et al. (2006) For sphere, φ, kp, kbf, klayer, δ Multi-medium EMT
with NL
keff =
( k p − klayer ) ( ) (
φ klayer  2β13 − β 3 + 1 + k p + 2 klayer β13 φβ 3 klayer − kbf + kbf )
( ) ( )
β13 k p + 2 klayer − k p − klayer φ  β13 + β 3 − 1
β = 1 + δ/ ap
Sabbaghzadeh φ, kp, kbf, klayer, δ Superposition of each
and Ebrahimi (2007) keff = kbf [1 − φ (1 + M ′ )] + φ k p + klayer M ′
( ) heat transfer mode
Ji-Hwan Lee, Seung-Hyun Lee, Chul Jin Choi, Seok Pil Jang, and Stephen U. S. Choi

abf 2
 δ   with NL effect and
+ φ (1 + M ′ ) ( 0.35 + 0.56 Re bf0.52 ) Prbf0.3 kbf
Prbf 2a p + δ M ' =  + 1 − 1 dynamic consideration
( )  a p  
  for nanotubes
293

Table 2. (Continued )
 294

Table 2. Summary of models considering nanolayer effect (1/2). (Continued)

Investigator Formulation Key parameter Remarks

Murshed et al. For cylinder, φ, kp, kbf, klayer, δ Additional study of
(2008) β = 1 + δ/ap Leong et al. with
( k p − klayer ) φ klayer 2β13 − β 3 + 1 + ( k p + 2klayer ) β13 φβ 3 ( klayer − kbf ) + kbf 
keff = cylindrical shape
β13 ( k p + 2 klayer ) − ( k p − klayer ) φ  β13 + β 3 − 1

Theoretical Droodchi keff = Correct B.C of Leong

models et al. (2009)  5 6  φ, kp, kbf, klayer, δ et al. study with multi-
( −
38φ a3p δ 3 −
6a 4p δ 2 +
2a3p δ 3 −
6φ a5p δ +
6a5p δ 24φ a 4p δ 2
+ 12φ a p δ + 30φ a 2p δ 4
+ 2φδ kbf k p )
 
 − 3a 4 δ 2 + 6φ a6 + 3a5 δ + 30φ a 2 δ 4 + a3 δ 3 + 44φ a3 δ 3 + 2φ a6 + 12φ a δ 5 + 24φ a5 δ + 42φ a 4 δ 2 + 3a6 k  medium EMT with
 ( p p p p p p p p p p p layer kp
 )
 3 3 5 5 6 6 4 2 3 3 4 2 6 2 4 5  nanolayer
 − 4 a p δ − 24φ a p δ + 12a p δ − 6φ a p + 6a p + 12a p δ − 82φ a p δ − 66φ a p δ − 4φδ − 60φ a p δ − 30φ a p δ klayer kbf 
( )
 5 2 4 5 3 3 4 2 5 3 3 6 4 2 2 
(
 − 12φ a p δ + 60φ a p δ + 24φ a p δ + 76φ a p δ + 6a p δ + 6a p δ + 2a p δ + 4φδ + 48φ a p δ klayer ) 
kbf
 −19φ a3p δ 3 − 6a 4p δ 2 − 2a3p δ 3 − 3φ a5p δ − 6a5p δ − 12φ a 4p δ 2 − 6φ a p δ 5 − 15φ a 2p δ 4 − φδ 6 kbf k p
( ) 
 
 + −3a 4 δ 2 + 3φ a6 − 3a5 δ + 15φ a 2 δ 4 − a3 δ 3 + 22φ a3 δ 3 + φ a6 + 6φ a δ 5 + 12φ a5 δ + 21φ a 4 δ 2 − 3a6 k 
 ( p p p p p p p p p p p layer k p )
 5 5 3 3 6 6 4 2 3 3 4 2 6 2 4 5 
 + −12φ a p δ − 12a p δ − 4 a p δ − 3φ a p − 6a p − 12a p δ − 41φ a pδ − 33φ a p δ − 2φδ − 30φ a p δ − 15φ a p δ klayer kbf 
( )
 3 3 4 2 3 3 5 5 4 2 5 2 4 6 2 
(
 + 38φ a p δ − 6a p δ − 2a p δ + 6φ a p δ − 6a p δ + 24φ a p δ + 12φ a p δ + 30φ a p δ + 2φδ klayer ) 

Ren et al. (2005) φ, kp, kbf, klayer, δ Thermal conductivity

 3Θ 2 φT2 
keff =  1 + F ( Pe ) + 3ΘφT + kbf Peclet number (TC) model with NL and
 1 − ΘφT 
Brownian Motion (BM)
Combined Sabbaghzadeh and φ, kp, kbf, klayer, δ Superposition of each
models Ebrahimi (2007) heat transfer mode with
keff = kbf [1 − φ (1 + M ′ )] + φ k p + klayer M ′
( ) 2 NL effect and dynamic
 δ  
abf 0.52
+ φ (1 + M ′ ) ( 0.35 + 0.56 Re bf Prbf0.3 kbf
) M ' = 
 a p
+ 1 − 1
 consideration for
Prbf 2a p + δ
( )    nanotubes

International Journal of Micro-Nano Scale Transport

A Review of Thermal Conductivity Data, Mechanisms and Models for Nanofluids
 Table 2. Summary of models considering nanolayer effect (1/2). (Continued)

Volume 1 · Number 4 · 2010

Murshed et al.  φω k p − ω kbf  2β13 − β 3 + 1 + k p + 2ω kbf β13 φβ 3 (ω − 1) + 1  φ, kp, kbf, klayer, δ Effect of BM, surface
(  )  ( )  
(2009) keff =  3
kbf chemical properties chemistry, interaction

 β1 k p +
( 2ω k bf − k
) (p − ω k bf φ
)  β13 + β 3 − 1 
 potential
  3Λ 2 9 Λ3 kcp + 2 kbf 3Λ 4 
+ φ 2 β 6 kbf  3Λ 2 + + + 6 + ⋅ ⋅ ⋅ 
  4 16 2 kcp + 3kbf 2  
  3k T 1 − 1.5β 3φ 
1  B ( ) GT 
+  ρcp c p,cp Ls  3 3
+ 
2  2πρcp β a p 6πµβ a p L s 
  

Computational Xue et al. (2004) Molecular Dynamics (MD) The NL has no

models significant effect on TC
Li et al. (2010) MD The NL has a
significant role in TC
Ji-Hwan Lee, Seung-Hyun Lee, Chul Jin Choi, Seok Pil Jang, and Stephen U. S. Choi
295
296 A Review of Thermal Conductivity Data, Mechanisms and Models for Nanofluids

  2k T  φ abf 
keff = 1 + C  B 2   kbf (47)
  πµ d p  kbf (1 − φ ) a p 

where C is a constant, kB is the Boltzmann constant, µ is the dynamic viscosity, dp is the diameter of a
particle, abf is the radius of the base fluid, and ap is the radius of a particle. This model accounts for the
dependence of thermal conductivity on particle size, volume fraction, and temperature. Predictions from
the combined model agree with experimental data for nanofluids with an extremely small particle
concentration. However, an order-of-magnitude estimate for the mean free path of a nanoparticle in a base
fluid is 10−2 m, which is physically unlikely. Moreover, this model has an inverse cubic dependence on
the particle size, 1/d2pap. This is inconsistent with the authors’ comment that the effective thermal
conductivity enhancement is inversely proportional to the radius of the particle, 1/ap.
Patel et al. [119] proposed a cell model for the effective thermal conductivity of nanofluids considering
the effects of the high specific surface area of the monodispersed nanoparticles and the microconvective
heat transfer due to Brownian motion. This is a modified model of Patel et al. [90]. Patel et al. [119]
developed a new cell model with following assumptions: (1) Nanofluids with low-particle concentration
would be used; (2) Interparticle interactions are neglected; (3) Particle/fluid heat transfer is much more
significant than particle/particle heat transfer. With these three assumptions, they considered a small control
volume, or cell, of the homogenously distributed suspension. The cell consists of a particle surrounded by
the liquid interacting with the particle. They categorized a new cell model into two parallel paths, i.e., one
related to the heat conduction through the base liquid without considering the suspended particles, and the
other a series of heat transfer which includes heat transfer from the liquid to the moving particle, heat
propagation by conduction within the particle, and finally heat transfer to the liquid from the particle. In
their analysis, they expressed the convective resistance between the liquid and the nanoparticle as the
particle interfacial heat transfer coefficient h, which depends on particle velocity. They considered the
thermal resistances in parallel and serial configurations appropriately and developed a cell model given by

  1
 qeff  kp  π   6φ  3
(48)
 q − 1 = k  
α ⋅ πµ a p   π 
 bf  bf  6 + bf
 C ⋅ 2 k B T 

where qeff is the effective heat flux of nanofluids, qbf is the heat flux of the base fluid, αbf is the thermal
diffusivity of the base fluid, and C is an empirical constant introduced to account for a one-dimensional
approximation of the heat flow in the neighborhood of the spherical particle. This model can account
for the nonlinear dependence of thermal conductivity of nanofluids on particle concentration at low
particle concentrations.
Koo and Kleinstreuer [88] postulated that the enhancement of thermal conductivity of nanofluids
is due mainly to Brownian motion. They developed a thermal conductivity model by adding
Brownian motion effect to a conventional conductivity model, such as the Maxwell model. Taking
into consideration factors such as Brownian motion, temperature, particle size, volume
concentration, and the properties of the base fluid and the particle, they developed the model
described by

( )
 k p + 2 kbf + 2 k p − kbf φ   
keff = 
( ) ( )
 kbf +  5 × 10 4 βφρ pcv , p
 k p + 2 kbf − k p − kbf φ 
kBT
ρ pd p
f ( T , φ , etc.)  (49)
 

International Journal of Micro-Nano Scale Transport

Ji-Hwan Lee, Seung-Hyun Lee, Chul Jin Choi, Seok Pil Jang, and Stephen U. S. Choi 297

where ρp is the density of the particle, cv,p is the specific heat of the particle, kB is the Boltzmann
constant, dp is the particle diameter, β represents the fraction of the liquid volume that moves with
a particle, and f is the factorial function. In Equation (49), the first bracket is the Maxwell model,
while the second bracket represents a dynamic model of thermal conductivity due to Brownian
motion of a large portion of surrounding liquid traveling with randomly moving nanoparticles.
However, it is difficult to obtain the function f theoretically; it can only be determined from
experimental data. Furthermore, the average traveling distance without changing its direction, lat,
has to be determined by, for example, molecular dynamics simulations. They assumed that lat is the
same as the diameter of the nanoparticle.
Prasher et al. [89] considered three possible mechanisms of thermal energy transfer in
nanofluids: (1) translational Brownian motion; (2) the existence of an interparticle potential; and (3)
convection in the liquid due to the Brownian movement of the particles. They performed an order-
of-magnitude analysis on these three possible mechanisms and deduced that local convection due
to the Brownian movement of the nanoparticles is the only mechanism that could explain the
anomalous enhancement of thermal conductivity. Prasher et al. [89] also reasoned that, if the
observed exceptional enhancements of thermal conductivity are due to small particle size, then
thermal conductivity for large particle sizes should be explained based on the conventional effective
medium theory such as the Maxwell-Garnett model. Therefore, they modified the Maxwell-Garnett
model by including the Brownian-motion-induced convection from multiple nanoparticles. Their
semi-empirical model can be written as

( )
 k p + 2 kbf + 2 k p − kbf φ 
(
keff = 1 + A Re m Pr 0.333 φ ) ( )
 kbf
 k p + 2 kbf − k p − kbf φ 
(50)

where the Reynolds number Re is defined as

1 18 k B T
Re = (51)
ν πρ p d p

and A and m are empirical constants. In Equation (51), ν is the kinematic viscosity of the fluid, kB is
the Boltzmann constant, ρp is the density of particle, and dp is the particle diameter. In Equation (50),
the first bracket represents a general correlation for convective heat transfer coefficient h for the
Brownian motion-induced convection from multiple nanoparticles proposed by Prasher et al. [89]. The
new model can predict the temperature-dependent thermal conductivity data for nanofluids and shows
that the Brownian-motion-induced convection from multiple nanoparticles is the main reason for the
observed thermal conductivity enhancement and trend. The model reduces to the Maxwell-Garnett
model as Reynolds number becomes zero for larger particles. However, constants A and m can only be
determined by experiment. Prasher et al. [89] explained that any model will be semi-empirical in nature
because of the complexities involved with the interaction in the convective currents due to multiple
nanoparticles.
Shukla and Dhir [120] proposed a microscopic model that takes into account the dependence of
nanoparticle characteristics (size, volume fraction, interparticle potential) and liquid properties
(viscosity, temperature), based on the theory of Brownian motion of nanoparticles in a fluid. To
develop a model for the effective thermal conductivity of nanofluids, they used ensemble averaging
techniques assuming the existence of small departures from equilibrium and the presence of
pairwise-additive interaction potential between various nanoparticles. The model can be expressed
as

Volume 1 · Number 4 · 2010

298 A Review of Thermal Conductivity Data, Mechanisms and Models for Nanofluids

keff = 
( )
 k p + 2 kbf + 2 k p − kbf φ 
 kbf +
ˆ T
nCk B
( )
 k p + 2 kbf − k p − kbf φ  6πµ a p

( ) ∂2 Ψ ( r T ) 
n 2 k B ∞ 2 eq ˆ  ∞ 3 eq
n 2CT
( ) ( ) ( )
6 µ a p ∫0 6 µ a p  ∫0
+ r g r Ψ r T dr − r g r dr  (52)
∂T ∂r 
n 2Cˆ  ∞ 3 eq ∂Ψ ( r T ) 
r g (r ) Ψ (r T )
6 µ a p k B T  ∫0
− dr 
∂r 

where n is the number density of nanoparticles, Ĉ is the heat capacity of a nanoparticle, kB is the
Boltzmann constant, µ is the dynamic viscosity of the base fluid, ap is the radius of the nanoparticles,
r is an arbitrary point, geq is the equilibrium pair distribution function, and Ψ is the interparticle
potential. In Equation (52), the first term represents macroscopic contribution given by the Hamilton-
Crosser model, and the remaining four terms represent contribution from Brownian motion of
nanoparticles. In the model, the kinetic contribution to the effective thermal conductivity was
neglected. They selected a specific form of the repulsive DLVO potential, which accounts for the
electrostatic repulsion between charged spherical nanoparticles in order to design the interparticle
interaction between various nanoparticles. The potential function is assumed to be a function of the
surface potential, the permittivity of the fluid, and the Debye length. They analyzed the interparticle
potential effect on thermal conductivity through calculations involving DLVO interaction between the
electric double layers on spherical nanoparticles. These calculations show the importance of long-range
repulsive potentials for the enhancement of thermal conductivity of nanofluids. Shukla and Dhir’s
model can be used to estimate the thermal conductivity of various nanofluids with different
nanoparticles, which may interact with each other through a specific interparticle potential such as
DLVO, steric, or van der Waals forces. However, their analysis considered only two-body interactions
and ignored the effects of higher multibody interactions. Moreover, their model is restricted to
relatively large nanoparticles whose size is significantly larger than the molecular dimensions.
Yang [121] developed a thermal conductivity model derived form the kinetic theory of particles in
the fluids under relaxation time approximations. This model takes into account convective heat transfer
due to the Brownian movement of nanoparticles. Yang’s effective thermal conductivity model is
expressed as

keff = 
(
 k p + 2 kbf + 2 k p − kbf φ )
 kbf + 157.5φC f u 2pτ (53)
( )
 k p + 2 kbf − k p − kbf φ 

where Cf is the heat capacity per unit volume of the fluid, τ is the particle relaxation time, and up is the
Brownian velocity of the particle expressed as

3k B T
up = (54)
mp

where kB is the Boltzmann constant and mp is the mass of particle. The first term in Equation (53) is the
Maxwell model, which represents diffusive conduction, and the second term corresponds to convection
due to the Brownian motion of nanoparticles. In the second term, 157.5 is not a fitting constant, but an
analytically obtained value from integration of the fluid velocity over the hydrodynamic boundary layer
around the Brownian particle. From this model, Yang found that the relaxation time of particle
Brownian motion, which provides a measure of the time required for a particle to forget its initial
velocity, could be significantly affected by the long-time tail in Brownian motion, which indicates that
particle velocity is astonishingly persistent. However, this particle relaxation time τ is difficult to
obtain, both experimentally and theoretically.

International Journal of Micro-Nano Scale Transport

Ji-Hwan Lee, Seung-Hyun Lee, Chul Jin Choi, Seok Pil Jang, and Stephen U. S. Choi 299

4.3.2. Computational models

Keblinski et al. [81] performed molecular dynamics (MD) simulations to calculate the thermal
conductivity of a simple model of the liquid and solid by using the Green-Kubo relationships. Their
simulation results demonstrated that collision between nanoparticles due to Brownian motion does not
affect the thermal conductivity and the microscopic heat transport mechanism of the simple model while
particle/liquid interfaces play an important role in the high overall conductivity of the nanofluids.
Bhattacharya et al. [122] developed a technique to predict the effective thermal conductivity of
nanofluids using Brownian dynamics simulation coupled with the equilibrium Green-Kubo method.
The combined parallel model of the effective thermal conductivity is given as

keff = φ k peff + (1 − φ ) kbf (55)

where kpeff is the effective contribution of the particles towards the overall thermal conductivity of the
system. For the calculation of kpeff, they used the Green-Kubo relation as

n
1
k peff =
k B T 2V
∑ Q ( 0 ) Q ( j ∆t ) ∆t. (56)
j=0

where T is the temperature, V is the volume of the domain, n is the number of time steps used in the
simulation, ∆t is the time step, and 〈Q(0)Q(j ∆t)〉 is the time-autocorrelation function of Q(t). Based on
their simulation results, they conclude that their model can predict the effective thermal conductivity of
nanofluids with high accuracy.
Evans et al. [92] carried out both an analysis with kinetic theory and the simulation to investigate
the effect of the Brownian motion on the effective thermal conductivity of nanofluids. For the
simulation, they used the molecular dynamics simulations of heat flow in a model nanofluid with well-
dispersed particles. Based on their kinetic theory analysis and results of MD simulation, they concluded
that the enhancements of the thermal conductivity of nanofluids are not affected by hydrodynamic
effects due to the Brownian motion and that the effective medium theory, such as the Maxwell-Garnett
model, can be used for predicting the effective thermal conductivity of nanofluids with well-dispersed
nanoparticles.
Vladkov and Barrat [93] simulated the thermal properties of nanofluids by using the molecular
dynamics (MD) simulations. Based on their simulation results, they conclude that the Brownian
motion of the particle does not affect the cooling process and that the Maxwell-Garnett model can
predict the effective thermal conductivity of nanofluids. They also concluded that the essential
parameter that influences the effective thermal conductivity is the ratio of the Kapitza length to the
particle radius and that the large heat transfer enhancements observed in nanofluids comes from
aggregation effects, such as particle clustering and percolation or cooperative heat transfer modes.
Using computer simulation, Li and Peterson [95] analyzed the mixing effect of the base fluid
directly adjacent to the nanoparticles due to the Brownian motion of nanoparticles. They used CFX
5.5.1 software (computational fluid dynamics software of British AEA Technology, Inc.) and a finite-
volume method for simulating the corresponding temperature, pressure, and velocity fields. They
investigated the effects of single, adjacent, and multiple nanoparticles, respectively. Their results
imply that Brownian motion induced microconvection and that mixing significantly enhances the
macroscopic heat transfer in nanofluids. Their results also indicated that Brownian motion is one of
the important factors for anomalous enhancement of the effective thermal conductivity of nanofluids.
Sarkar and Selvam [94] developed the nanofluids systems that consist of a base fluid model of
argon and a nanofluid model of copper particles in argon with various nanoparticles concentrations.
They used an equilibrium molecular dynamics simulation to model this nanofluid system and used
the Green-Kubo relation to calculate the thermal conductivity of the base fluid and nanofluids. They
found that the effective thermal conductivity of copper-argon nanofluids was much greater than
predicted by the Hamilton-Crosser model at both very low volume concentration (up to 0.4%) and

Volume 1 · Number 4 · 2010

300 A Review of Thermal Conductivity Data, Mechanisms and Models for Nanofluids

high volume concentration (up to 8%). From mean square displacement computation by using
molecular dynamics simulation, they also found that the movement of liquid atoms in nanofluids
increases significantly (1.41 times in 1% nanofluid) compared to the movement of liquid atoms in
base fluid, while the nanoparticle movement was 28 times slower than that of the liquid phase in
1% nanofluids. This implies that the Brownian motion of the nanoparticles is far too slow to
transport the heat; on the other hand, localized fluid movement around nanoparticles is induced by
much faster liquid atoms in nanofluids. They concluded that these phenomena are the main
mechanism for enhanced thermal conductivity of nanofluids. However, the simluation cell
considered only single nanoparticles and excluded the effects of aggregation.
Jain et al. [123] calculated the effective thermal conductivity of nanofluids using Brownian
dynamic simulation and the thermal conductivity of nanofluids computed by using the Green-Kubo
relation. They considered various parameters, such as particle concentrations ranging from 0.5 to 3
vol.%, particle size raging from 15 to 150 nm, and temperature ranging from 290 to 320 K. Their
Brownian dynamic simulations exclude the fluid molecules and include the effect of hydrodynamic
interactions related to the base fluid, through a position-dependent interparticle friction tensor. They
commented that the simulation based on N-coupled Langevin equations, though very fundamental
in its formulation, was able to simulate the effects of parameters such as particle concentration,
particle size, and temperature of the fluid on the effective thermal conductivity of nanofluids. They
assumed that thermal conduction caused by the motion of nanoparticles and the base fluid
molecules occurs in parallel, and thus they used the combined parallel model for the calculation of
the effective thermal conductivity as

keff = φ k pB + (1 − φ ) kbf (57)

where kpB is the thermal conductivity owing to the Brownian motion of the nanoparticles. For the
calculation of kpB, they used the Green-Kubo relation as

 n− j
1 n
1 
k pB =
k B T 2V
∑  3 ( n − j ) ∑ Q ( i ∆t ) ⋅ Q ( i + j ) ∆t   ∆t.

(58)
j=0 i=0

where T is the temperature, V is the volume of the domain, n is the number of time steps used in the
simulation, ∆t is the time step. With this simulation, they showed the effect of particle size, temperature,
and volume concentration on the effective thermal conductivity of nanofluids. Based on their
simulation results, they conclude that their model can predict the effective thermal conductivity of
nanofluids and that the Brownian motion of the particle is the most important phenomenon, the key
mechanism for the enhancement in the thermal conductivity of nanofluids.
A summary of Brownian motion based models for the thermal conductivity of nanofluids is listed in
Table 3.

4.3.3. Empirical models

Chon et al. [46] and Vasu et al. [124] have developed empirical correlations for the effective thermal
conductivity of nanofluids. Although they can predict the effective thermal conductivity of nanofluids
very well, they are valid only for the nanofluids whose data were used to formulate the correlation and
in the range of the data.

4.4. Aggregation based models

We categorize aggregation based models into three submodels: modified models, combined effects
models, and computational models.

4.4.1. Modified models

International Journal of Micro-Nano Scale Transport

Ji-Hwan Lee, Seung-Hyun Lee, Chul Jin Choi, Seok Pil Jang, and Stephen U. S. Choi 301

Wang et al. [84] developed a fractal model of the effective thermal conductivity of nanofluids based
on the effective medium theory and the fractal theory for the description of nanoparticle cluster and
its radial distribution. This model involves the application and improvement of the multicomponent
Maxwell-Garnett model by substituting the effective thermal conductivity of the particle clusters,
kcl, which comes from Bruggeman model. Their fractal model can be expressed as

 ∞ kcl ( acl ) n ( acl ) 

 (1 − φ ) + 3φ ∫0 dacl 
  kcl ( acl ) + 2 kbf  
keff =  kbf
∞ kbf ( acl ) n ( acl )
(59)
 (1 − φ ) + 3φ
 ∫0  k ( a ) + 2k  cl  da
  cl cl bf  

where acl is the equivalent radius of cluster, kcl(acl) is the effective thermal conductivity of nanoparticle clusters,
and n(acl) is the radius distribution function. The fractal model predicts well for CuO-water nanofluids when
adsorption is included in the analysis, but it underpredicts the enhancement when the adsorption effect is
excluded. This model can only predict the effective thermal conductivity of nonmetallic nanofluids.

4.4.2. Combined effects models

Xuan et al. [69] developed a theoretical model for the effective thermal conductivity of nanofluids
considering Brownian motion and aggregation structures. They assumed that the random motion of the
nanoparticles is the main reason for the difference between nanofluids and conventional fluids.
Although the suspended nanoparticles are in Brownian motion, some particles may aggregate because
of collisions with other particle. Thus, they simulated random motion and the aggregation process of
the nanoparticles by using the theory of Brownian motion and the diffusion-limited aggregation model.
Their thermal conductivity model can be written as

keff = 
(
 k p + 2 kbf + 2 k p − kbf φ  ) ρ φc
 kbf + p v , p
kBT
(60)
(
 k p + 2 kbf − k p − kbf φ ) 2 3π acl µ

where acl denotes the apparent radius of the nanoparticle cluster and depends on the fractal dimension
of the cluster structure. In Equation (60), the first term is the Maxwell model and the second term
corresponds to the component of thermal conductivity enhancement due to the Brownian motion of the
suspended nanoparticles and/or clusters. This model indicates that the effective thermal conductivity of
nanofluids can be affected by the Brownian motion of the suspended nanoparticles and/or clusters and
the structure of the nanoparticles and/or clusters.
Prasher et al. [85] combined the aggregation kinetics of nanofluids based on colloidal chemistry
with the physics of thermal transport (microconvective effects due to Brownian motion) to study the
effects of aggregation on the effective thermal conductivity of nanofluids. The particles well
dispersed at initial time (t = 0) agglomerate to form multiple aggregates as time goes on. These
aggregates are considered to be new “particles” with an effective radius (aag) that enhance the
thermal conductivity of nanofluids. However, this enhancement will decrease as the aggregates
continue to agglomerate to form much bigger aggregates. As t → ∞, all nanoparticles will
agglomerate to form one large aggregate, at which time enhancement of thermal conductivity may
not occur because one large aggregate will be settled. The suggested relation is

φ p = φintφag (61)

where φp is the volume fraction of the primary particles, φint is the volume fraction of the particles
in the aggregates, and φag is the volume fraction of the aggregates in the entire fluid. The relation

Volume 1 · Number 4 · 2010

 Table 3. Summary of models considering Brownian motion effect (1/2)
302

Investigator Formulation Key parameter Remarks

Jang and Choi (2004) φ, kp, kbf, BM induced

d bf kB T nanoconvection
keff = kbf (1 − φ ) + β k p φ + C1 kbf Re 2d p Pr φ up =
dp 3πµ d p lbf

Kumar et al. (2004) φ, kp, kbf, Consider the effect of

Theoretical   2k T  φ abf  2kB T BM of particle with
keff = 1 + C  B 2   kbf up =
  πµ d p  kbf (1 − φ ) a p  (πηd p2 )
models kinetic theory
Patel et al. (2008)   Thermal electric approach
1 φ,kp, kbf,
 qeff  kp  π  6φ 3
 − 1 =     2kB T with BM induced
 qbf  kbf  6 + α bf ⋅ πµ a p   π  up = nanoconvection effect
 C ⋅ 2 k B T  ( πηd p2 )
Koo and Kleinstreuer φ, kp, kbf, BM induced
(2004) nanoconvection with
 k p + 2 kbf + 2 k p − kbf φ    EMT
( ) kB T 18 k B T
keff =   kbf +  5 × 10 4 βφρ p cv , p
( ) f ( T , φ , etc.)  up =
 k p + 2 kbf − k p − kbf φ 
( ) ρ d
p p  πρ p d p3
 

Prasher et al. (2005) φ, kp, kbf BM induced

 k p + 2 kbf + 2 k p − kbf φ  ( ) 18 k B T nanoconvection with
keff = 1 + A Re m Pr 0.333 φ 
( )  kbf up =
 k p + 2 kbf − k p − kbf φ  ( ) πρ p d p3
  EMT consider ITR
Shukla and Dhir (2008) φ, kp, kbf, Consider BM effect
 k p + 2 kbf + 2 k p − kbf φ 
( ) nCk ˆ T
B Ψ: interparticle potential based on interparticle
keff =  kbf +
(
 k p + 2 kbf − k p − kbf φ  ) 6 πµ ap potential with EMT
n 2 k B ∞ 2 eq n 2 CT
ˆ  ∞ 3 eq ∂2 Ψ ( r T ) 
+ ∫ r g ( r ) Ψ ( r T ) dr −  ∫ r g (r ) dr 
6µ a p 0( )
6µ a p  0 ∂T ∂r 
n 2 Cˆ  ∞ 3 eq ∂Ψ ( r T ) 
− r g (r ) Ψ (r T ) dr 
6 µ a p k B T  ∫0 ∂r 

International Journal of Micro-Nano Scale Transport

A Review of Thermal Conductivity Data, Mechanisms and Models for Nanofluids

Table 3. (Continued )
 Table 3. Summary of models considering Brownian motion effect (1/2)

Investigator Formulation Key parameter Remarks

Yang (2008) φ, kp, kbf, BM induced

nanoconvection under
 k p + 2 kbf + 2 k p − kbf φ 
( )
keff =   kbf + 157.5φC f u 2p τ relaxation time
 k p + 2 kbf − k p − kbf φ 
( ) 3k B T
  up = approximation with
mp
Maxwell model
Keblinski et al. (2002) MD The BM is negligible
keff = φ k peff + (1 − φ ) kbf
n
effect on nanofluids
1

Volume 1 · Number 4 · 2010

k peff = ∑ Q ( 0 ) Q ( j ∆t ) ∆tt.
kB T 2V j=0

Bhattacharya et al. (2004) Brownian Dynamics The BM is significant

Simulation (BD) effect on NFs
Evans et al. (2006) MD The BM is negligible
effect and well matched
with EMT
Computational Vladkov and Barrat (2006) MD The BM is negligible
models effect and well
matched with EMT
with ITR
Li and Peterson (2007) CFX The BM induced
(Commercial software) nanoconvectin is
significant effect
on NFs
Sarkar and Selvam (2007) , MD The BM induced
nanoconvectin is
significant effect on NFs
Jain et al. (2009) BD The BM induced
keff = φ k pB + (1 − φ ) kbf
, nanoconvectin is
Ji-Hwan Lee, Seung-Hyun Lee, Chul Jin Choi, Seok Pil Jang, and Stephen U. S. Choi

significant effect on
n  n− j
1 1 
k pB = Q ( i ∆t ) ⋅ Q ( i + j ) ∆t   ∆t. NFs
kB T 2V 
∑  3 ( n − j ) ∑
j=0 i=0
303
304 A Review of Thermal Conductivity Data, Mechanisms and Models for Nanofluids

in Equation (61) shows that, in the case of a well-dispersed system, φint = 1 because there is only
one particle in each aggregate and thus φag = φp, while for a medium composed wholly of
aggregates, φag = 1, and thus φint = φp. They assumed that the maximum thermal conductivity owing
to conduction will lie between these two limits. Considering aforementioned phenomena, they
modified the Maxwell-Garnett model by substituting the thermal conductivity of the aggregates kag
and the volume fraction of the aggregates φag as

( )
 kag + 2 kbf + 2 kag − kbf φag 
(
keff = 1 + A Re m Pr 0.333 φ ) ( )
 kbf
 kag + 2 kbf − kag − kbf φag 
(62)

They stated that the effective thermal conductivity of nanofluids depends on parameters such as the
Hamaker constant, the zeta potential, pH, and ion concentration. They also mentioned that the
conductive component of the thermal conductivity ratio can also increase with temperature, depending
on the chemistry of the solution. However, they excluded the effects of thermal boundary resistance
between the particles and the fluid for simplicity. The modified Maxwell-Garnett model reduces to the
original Maxwell-Garnett model for well-dispersed media.
Feng et al. [125] developed a new model for effective thermal conductivity of nanofluids that
accounts for the effect of the nanolayer and the aggregation effect of nanoparticles. They divided the
aggregation model into two parts: the coherent fluid and a quarter of the column of radius (ap + δ). They
further divided the column into two parts: the touching particles and the base fluid. Their model can be
expressed as

( )
 k pe + 2 kbf + 2 k pe − kbf (1 − η )3 φ 
keff = (1 − φe )   kbf
( )
 k pe + 2 kbf − k pe − kbf (1 + η ) φ 
3

 3φ  1 ap + δ   (63)

+ φe  1 − φe  + e  ln
3
− 1  kbf


2  η η
 ( )
a p + δ (1 − η )  


where φe is the equivalent volume fraction defined as φe = φ(1 + β)3 and β = δ /ap is the ratio of the
nanolayer thickness to the original particle radius; kpe is the equivalent thermal conductivity of the
equivalent particles defined by Yu and Choi [82]; and η = 1 – kbf /kpe. It can be shown that if φe →
2/3, the aggregation model dominates the heat transfer and all particles are in the touching state,
while if φe → 0, the Maxwell model dominates the heat transfer and all particles are in the
nontouching state. They found that the contributions from clusters owing to nanoparticle aggregation
increase with decreasing nanoparticle size at a constant volume fraction. This phenomenon is caused
by decreasing the average interparticle distance as the particle size decreases, which increases the
van der Waals forces, resulting in increased probability of aggregation. Based on this explanation,
they concluded that the aggregation of suspended nanoparticles is one of the crucial mechanisms for
enhancing the thermal conductivity of nanofluids. However, this model was derived with
approximation of two-dimensional lattice model. Thus, the disordering effect and the defect
tolerance of nanoparticles in the fluids may be considered for more realistic prediction.

4.4.3. Computational models

Prasher et al. [86] demonstrated that the aggregation of nanoparticles can significantly enhance the
thermal conductivity of nanofluids. They assumed that a fractal cluster is embedded within a sphere

International Journal of Micro-Nano Scale Transport

Ji-Hwan Lee, Seung-Hyun Lee, Chul Jin Choi, Seok Pil Jang, and Stephen U. S. Choi 305

with a radius equal to the radius of gyration Rg and is composed of several approximately linear
chains, which span the entire cluster and side chains. These linear chains are called the backbone of
the cluster, while the other particles are called dead ends [126, 127]. Figure 6 represents the
backbone of the cluster and dead ends. The backbone can play an important role in thermal
conductivity due to its connectivity. They used the modified Maxwell-Garnett model, Equation (60)
for calculating the effective thermal conductivity of the whole system. They showed that predictions
of their modified Maxwell-Garnett model are in excellent agreement with detailed numerical
calculations on model nanofluids involving fractal clusters. The results also showed the importance
of cluster morphology on thermal conductivity enhancements. Based on the results of the numerical
simulation, they concluded that conduction-phenomenon-based thermal conductivity of nanofluids
can be significantly enhanced as a result of aggregation of the nanoparticles. This aggregation is a
function of the chemical dimension of the aggregates and the radius of gyration of the aggregates.
However, they excluded the effect of thermal interfacial resistance in their analysis.
Evans et al. [128] analyzed and simulated the effect of the aggregation and interfacial thermal
resistance on the effective thermal conductivity of nanofluids and nanocomposites. This was an
extension of the study of Prasher et al. [86], including the effect of interfacial thermal resistance.
They also used the modified Maxwell-Garnett model, Equation (60), derived by Prasher et al. [85],
for calculating the effective thermal conductivity of the whole system, as well as the three-level
homogenization theory validated by Monte Carlo simulations. They demonstrated that the
aggregation of high conductive nanoparticles in a liquid or solid low-conductivity matrix affects the
thermal conductivity enhancement significantly. The key factor is the high aspect ratio backbone of
the fractal-like aggregates. Based on the results of their analysis and simulations, they obtained the
same results of Prasher et al. [86], and they also showed that there was no enhancement in nanofluid
thermal conductivity when the particle radius is equal to the Kapitza radius. Thus any enhancement
in the thermal conductivity would be degraded.
Table 4 summarizes the aggregation based models for the thermal conductivity of nanofluids.

4.5. Other mechanism based models

Nan et al. [129] developed a simple formula for the effective thermal conductivity of carbon
nanotube composites, based on Maxwell-Garnett type effective medium theory. They assumed that
the thermal conductivities of the carbon nanotubes along transverse axes kx and longitudinal axes kz
are much higher than the thermal conductivity of the base fluid kbf. Moreover, in the case of the dilute
limit case (φ < 0.02), the effective thermal conductivity of carbon nanotube composites can be
expressed as

 φk 
keff =  1 + cnt  kbf (64)
 3kbf 

where kcnt denotes the thermal conductivity of carbon nanotubes. They found that this simple formula
predicts much higher thermal conductivity enhancement even in the dilute limit case because of the
ultrahigh thermal conductivity and aspect ratio of the carbon nanotubes. Comparing this model to
experimental data, Nan et al. [129] conclude that the simple model shows good agreement with the
nanotube-based composites.
Nan et al. [130] also developed a simple equation for predicting the effective thermal conductivity
of nanotube-based nanofluids in terms of an effective medium approach accounting for interfacial
thermal resistance. The effective thermal conductivity of the nanotube composite with carbon
nanotubes randomly dispersed in a matrix can be expressed as

Volume 1 · Number 4 · 2010

306 A Review of Thermal Conductivity Data, Mechanisms and Models for Nanofluids

Figure 6. Schematic of ‘backbone cluster’ and ‘dead ends’.

3 + φ ( β x + βz )
keff = (65)
3 − φβ x

with

βx =
(
2 k xx − kbf ), βz =
kzz
−1 (66)
k xx + kbf kbf

where kxx and kzz are the equivalent thermal conductivities along transverse axes and longitudinal axes
of composite unit cells, respectively, i.e., a nanotube surrounded by a very thin interfacial thermal
barrier layer and can be expressed as

kcnt kcnt
k xx = , kzz = (67)
2aK kcnt 2aK kcnt
1+ 1+
dcnt kbf Lcnt kbf

where dcnt and Lcnt are the diameter and length of the nanotubes, respectively; and aK is the Kapitza
radius defined as

aK = RK kbf (68)

International Journal of Micro-Nano Scale Transport

 Table 4. Summary of models considering aggregation effect

Investigator Formulation Key parameter Remarks

Modified Wang et al. (2003)  ∞ kcl ( acl ) n ( acl )  φ, kp, kbf, Modified Bruggeman
models  (1 − φ ) + 3φ ∫0 dacl  Df :fractal dimension model with aggregation
  kcl ( acl ) + 2 kbf  
keff =  kbf (Agg) effect
 1 − φ + 3φ ∞ kbf ( acl ) n ( acl ) da 
( ) ∫0 cl 
 kcl ( acl ) + 2 kbf 
 

Xuan et al. (2003)  k p + 2 kbf + 2 k p − kbf φ  ρ φc φ, kp, kbf, EMT approach with
( ) kB T
keff =   kbf + p v , p Agg and BM effect

Volume 1 · Number 4 · 2010

 k p + 2 kbf − k p − kbf φ 
( ) 2 3π acl µ
  Df :fractal dimension

Combined Prasher et al. (2006) φ, kp, kbf, EMT approach based

1
effects  kag + 2 kbf + 2 kag − kbf φag   Df on the Agg kinetics and
( ) Rg t
models keff = 1 + A Re m Pr 0.333 φ 
( )  kbf = 1 +  BM effect
 kag + 2 kbf − kag − kbf φag 
 (  ) ap  t p

Feng et al. (2007)  k pe + 2 kbf + 2 k pe − kbf (1 − η )3 φ  φ, kp, kbf, EMT approach with NL
keff = (1 − φe ) 
( )  kbf φT = φ(1 + δ/ap)3 and Agg effect
 k pe + 2 kbf − k pe − kbf (1 + η )3 φ 
( )
 
 3 3φ  1 ap + δ  

+φe  1 − φe  + e  ln − 1  kbf
  2  η η ( )
a p + δ (1 − η )  
  

Prasher et al. (2006, APL) Monte Carlo Agg with 3 level EMT
simulation (MC) theory, this model can
be well matched with
Computational MC simulation results

models Evans et al. (2008) MC Additional study of

Prasher et al. 06 APL
with ITR
Ji-Hwan Lee, Seung-Hyun Lee, Chul Jin Choi, Seok Pil Jang, and Stephen U. S. Choi
307
308 A Review of Thermal Conductivity Data, Mechanisms and Models for Nanofluids

where RK is the Kapitza resistance. For the case of the dilute limit case (φ < 0.01), the effective thermal
conductivity of carbon nanotube composites can be simplified as

keff = 1 +
(
φ p kcnt / kbf ) (69)
3 p + 2aK kcnt
dcnt kbf

where p is the aspect ratio of nanotubes (p = Lcnt /dcnt). This simple expression demonstrates that
interface thermal resistance diminishes the thermal conductivity enhancement. Thermal
conductivity of carbon nanotubes without interface thermal resistance would be much higher than
that with the interface thermal resistance. If a carbon nanotube has no interface thermal resistance
and p >> 1, the effective thermal conductivity of carbon nanotube composites can further be
simplified as

 φk  (70)
keff =  1 + cnt  kbf
 3kbf 

which is independent of geometric parameters of nanotubes and is same as the expression by Nan et al.
[129]. With the results of their analysis, they conclude that the effective thermal conductivity of carbon
nanotube-based nanofluids is limited by interface thermal resistance. If a carbon nanotube has high
interface thermal resistance across the nanotube-matrix interface, this causes a significant degradation
in thermal conductivity enhancement.
Xue [131] developed a new model for the effective thermal conductivity of carbon nanotube-based
nanofluids accounting for orientation distribution. They derived two formulae for the effective thermal
conductivity of CNTs-based nanofluids, based on Maxwell theory. The first formula is

(
 (1 − φ ) + ( 4φ / π ) kcnt / kbf arctan π / 4 kcnt / kbf
keff = 
)  k (71)
(
 (1 − φ ) + ( 4φ / π ) kbf / kcnt arctan π / 4 kcnt / kbf
 )  bf

and the second formula is

{ (
 (1 − φ ) + 2φ kcnt / kcnt − kbf
keff = 
)} ln {( kcnt + kbf ) / 2kbf }  k
{ )} ln {( kcnt + kbf ) / 2kbf }  bf
(72)
(
 (1 − φ ) + 2φ kbf / kcnt − kbf


The formation of these two models depends on determination of distribution function. These models
imply that dispersion of very small amount of CNTs can cause a remarkable enhancement in the
effective thermal conductivity of CNTs-based nanofluids.
Murshed et al. [39] developed a new model for predicting the effective thermal conductivity of
nanofluids accounting for the geometric structure of uniformly dispersed nanoparticles in base fluids
along with the particle size. They mentioned that in order to achieve better stability of nanofluids, it
is important to ensure homogeneous distribution of nanoparticles in the base fluid. They considered
body-centered cubic arrays to be ideal for their study: these arrays have the maximum number of
points for heat flow analysis (eight), and they show slightly higher heat flow rate compared to other
periodic arrays, such as simple cubic and face-centered cubic. They considered two thermal circuits,
each of which was in one-dimensional conduction with a two-phase system. One circuit included
thermal resistances offered by the solid and fluids phases in parallel (upper bound), and thus the

International Journal of Micro-Nano Scale Transport

Ji-Hwan Lee, Seung-Hyun Lee, Chul Jin Choi, Seok Pil Jang, and Stephen U. S. Choi 309

effective thermal conductivity of two-phase medium can be expressed as

keff = φ k p + (1 − φ ) kbf (73)

The other circuit included thermal resistances in series (lower bound) to the direction of heat flow,
and the effective thermal conductivity of the two-phase medium can be written as

φ (1 − φ )
keff = + (74)
kp kbf

Their new model for the effective thermal conductivity of nanofluids is given as

  kp  0.52φ  k p 
kbf 1 + 0.27φ 4 / 3  − 1  1 + 1/ 3  − 1 
  kbf    1 − φ  kbf  
keff = (75)
 kp   0.52φ 
1 + φ4/3  − 1  + 0.27φ1 / 3 + 0.27
 bf
k   1 − φ 1/ 3


This model can predict the effective thermal conductivity of nanofluids more accurately than the
classical models of Maxwell, Hamilton-Crosser, and Bruggeman. The model shows that in
homogeneously dispersed nanofluids, the particle size affects the thermal conductivity implicitly
through the volume fraction and the particle distance.
Gao and Zhou [132] presented differential effective medium theory for the effective thermal
conductivity of nanofluids by simultaneously taking into account the geometric anisotropy and the
physical anisotropy. The geometric anisotropy arises from the large aspect ratio of carbon nanotubes,
and the physical anisotropy comes from the interfacial thermal resistance [130]. Their model is given
as

3A 3C1 3C2
 kbf   kbf + B1   kbf + B2 
1− φ =   k +B  k +B  (76)
 keff   bf 1  eff 2

where

A=
(
2 Lz −5 − 8 Lz − 3 L2z ) (77)
( 2 + 6 Lz ) ( 3 Lz − 5 )
and

kz − 3k x ( −1 + Lz ) − 3kz Lz ± N
B1, 2 = (78)
−10 + 6 Lz

and

Volume 1 · Number 4 · 2010

310 A Review of Thermal Conductivity Data, Mechanisms and Models for Nanofluids

( ) (
N 1 + 8 Lz − 9 L2z ±  k x ( −1 + Lz ) (13 + 21Lz ) + kz 1 + 5 Lz + 47 L2z − 21L3z 

2
 ) (79)
C1, 2 =
N ( −10 + 6 Lz ) (1 + 3 Lz )

with

2 2
(
N = kz2 (1 − 3 Lx ) + 9 k x2 ( −1 + Lz ) + 2 k x kz 13 + 12 Lz − 9 L2z ) (80)

where Lz [Lx ≡ (1 − Lz)/2] is the depolarization factor of spheroidal particles and z denotes the rotational
axis; and kx and kz are the thermal conductivities along transverse axes and longitudinal axes, respectively.
They found that the adjustment of the particle shape is helpful to achieve appreciable enhancement of the
effective thermal conductivity of nanofluids. For randomly isotropic spherical inclusions, their formula
reduces to the Bruggeman model. Their model can predict the nonlinear relationship between the effective
thermal conductivity and the volume fraction, even for very low volume concentrations. However, their
model cannot explain the strong dependence of enhancement on temperature, and to achieve such a
dependence, Brownian motion must be considered for nanoparticles with small sizes.
Xue [133] developed a new formula for predicting the effective thermal conductivity of carbon
nanotube-based nanofluids accounting for the interface thermal resistance with an average polarization
theory. In developing the procedure, their approach is very similar to Nan et al. [130]. Their model can
be written as

9 (1 − φ )
keff − kbf 
+φ
keff − kzz
+
(
4 keff − k xx ) 
=0
( ) ( )
(81)
2 keff + kbf  keff + 0.14 ( 2acnt / Lcnt ) kzz − keff 2 keff + 0.5 k xx − keff 

where
kcnt kcnt
k xx = , kzz = (67)
2aK kcnt 2aK kcnt
1+ 1+
dcnt kbf Lcnt kbf

Their model predicts that the thermal conductivity enhancement of nanofluids increases quickly as
the thermal conductivity of the base fluid decreases and increases as the thermal conductivity of the
carbon nanotube increases [133].
Minnich and Chen [134] introduced a modified effective medium approach formulation for
nanocomposites and assumed that the characteristic length of the inclusion is equal to or less than the
phonon mean free path. Their modified effective medium approach formulation can be expressed as

1
keff =  cv , bf v pho, bf
1

{
  k p (1 + 2κ ) + 2 kbf + 2φ k p (1 − κ ) − kbf }  (82)
 3 ( ) {
1 / lbf + ( Φ / 4 )   k p (1 + 2κ ) + 2 kbf − φ k p (1 − κ ) − kbf } 

where cv,bf is the volume specific heat of the base fluid, vpho,bf is the phonon group velocity of the base
fluid, lbf is the phonon mean free path of the base fluid, and Φ is the interface density for spherical
particles defined as

International Journal of Micro-Nano Scale Transport

Ji-Hwan Lee, Seung-Hyun Lee, Chul Jin Choi, Seok Pil Jang, and Stephen U. S. Choi 311

Φ=
(
4π d p / 2 )2 = 6φ (83)
Lav dp

where Lav is the average length of a side of a cube that encloses one nanoparticle. In Equation (82),
κ is a dimensionless parameter defined as

aK (84)
κ=
( d p / 2)
where aK is the Kapitza radius defined as

aK = RK kbf (68)

where RK is the Kapitza resistance (thermal boundary resistance). Their results show good agreement
with Monte Carlo (MC) simulations and solutions to the Boltzmann equation [134]. They also found
that the thermal conductivity in nanocomposites is governed by the interface density and that thermal
boundary resistance (TBR) plays an important role in determining the effective thermal conductivity.
Jung and Yoo [135] proposed a novel model for the effective thermal conductivity of nanofluids.
They adopted the kinetic theory to show the effect of Brownian motion with a mean free path and
the interparticle interaction due to the electrical double layer (EDL). The effect of interparticle
interaction due to the existence of electrical double layer was considered by Jung and Yoo for the
first time. Their model consists of the stationary mode (based on conventional theory), the single
particle motion mode (based on thermal conductivity of nanofluids considering the kinetic theory
with Brownian motion), and the interparticle interaction mode (based on thermal conductivity of
nanofluids and originating from the kinetic theory regarding the particle motions induced by
electrical double layer). Their model can be expressed as

keff = 
( )
 k p + 2 kbf + 2 k p − kbf φ  φl cˆ
 kbf + p v
kBT
+
φl pcˆv Aec2l p
(85)
( )
 k p + 2 kbf − k p − kbf φ  3 3πµ d plbf 3 (
ε bf m p a pφ −1 / 3 )
2

where lp is the mean free path of the nanoparticle, lbf is the liquid mean free path, ĉ v is the specific heat
per particle as estimated by the Debye model per particle, A is the Coulomb constant, ec is the electrical
charge, εbf is the dielectric constant of the base fluid, and mp is the mass of the nanoparticle. The mean
free path of the nanoparticle based on the assumption that all particles move about with a Maxwellian
velocity distribution is given as

1 (86)
lp =
2φπ d p2

Their model shows that the interparticle interaction due to the electrical double layer contributes
significantly to the enhancement of the thermal conductivity of nanofluids.
Table 5 summarizes other mechanism based models for the thermal conductivity of nanofluids and
shows the formulation, key parameters for determining thermal conductivity, and remarks.

Volume 1 · Number 4 · 2010

312 A Review of Thermal Conductivity Data, Mechanisms and Models for Nanofluids

In summary, classical EMT-based static models cannot predict the magnitude and trend of the
thermal conductivity of nanofluids, although predictions of EMT-based models agree with some data
that do not show anomalously enhanced or temperature- or particle-size-dependent conductivities.
New static models developed to improve classical models are much better than effective medium
based classical models in narrowing the gap between thermal conductivity enhancement data and
model predictions. However, new static models are unable to capture other features such as the
dependence of the thermal conductivity of nanofluids on nanoparticle size and fluid temperature.
However, new models based on a combination of static and dynamic mechanisms appear to predict
not only the magnitude of the enhanced thermal conductivity but also other features of the thermal
conduction behavior of nanofluids. In the future, comprehensive models that can predict all the novel
features of nanofluids, including size- and temperature-dependent thermal conductivity should be
developed based on new mechanisms as well as the classical diffusion mechanism.

5. CONCLUDING REMARKS
Pioneering scientists and engineers have discovered novel properties in nanofluids that are utterly
unexpected in conventional suspensions of micro- or millimeter-sized solid particles. They have
proposed mechanisms of energy transport in nanofluids based on the links between nanoparticle
structure and/or dynamics and thermal properties and developed various models for k of nanofluids.
Although there have been numerous experimental, theoretical, analytical, and numerical studies, most
of the issues related to the data, mechanisms, and models are still unresolved and pose many formidable
challenges. Therefore, much research is still required in the future in order to resolve critical issues such
as data inconsistencies, the sufficiency and suitability of classical and new mechanisms, and the
discrepancies between experimental data and model predictions.

5.1. Data inconsistencies

Despite the problem of conflicting experimental data, it is generally agreed that thermal conductivity
behavior of nanofluids has features that cannot be explained by the classical effective medium theories
alone. The classical effective medium theories significantly underpredict the magnitude of most
thermal conductivity enhancement data. Moreover, they are inherently unable to predict the size- and
temperature-dependent thermal conductivity of nanofluids. As shown in Figure 2, nanofluids
produced by one-step processes set themselves apart from those produced by two-step methods.
Furthermore, nearly all nanofluids used so far for the studies of their k are polydisperse systems of
nanoparticles and aggregated nanoparticles. However, it is highly desirable to make monodisperse
nanofluids for discovery of mechanisms and development of new models for the thermal conductivity
of nanofluids. Therefore, it is recommended to produce high-quality nanofluids with well-dispersed
monosize nanoparticles using a one-step process [5, 27, 136] with little or no dispersants. It is crucial
to generate a thermal conductivity database using such high-quality nanofluids for the discovery and
validation of mechanisms behind all the features of the thermal conductivity of nanofluids.

5.2. Mechanisms
Effective medium theories based on the underlying assumption of diffusive conduction and motionless
nanoparticles are not sufficient to predict the new features of the thermal conductivity of nanofluid. A
number of microscale and macroscale mechanisms have been proposed to explain the magnitude and trend
of the thermal conductivity of nanofluids. New models based on the proposed heat conduction mechanisms
in nanofluids can predict both the magnitude and trend of the thermal conductivity of nanofluids. However,
they cannot accurately predict experimental data. Therefore, the proposed concepts and mechanisms behind
the thermal conductivity behavior of nanofluids remain to be validated. Therefore, more systematic
experiments with well-dispersed, well-characterized nanofluids and a better understanding of the physics
of fluid flow and heat transfer at the nanoscale are needed to establish the fundamental mechanisms of heat
conduction in nanofluids. Understanding the underlying mechanisms of heat conduction in nanofluids is
the essential requirement for the development of models that can accurately predict the k behavior of
nanofluids.

International Journal of Micro-Nano Scale Transport

 Table 5. Summary of models considering other effects

Investigator Formulation Key parameter Remarks

Nan et al. (2003) φ, kbf, kcnt The CNT NFs results
 φk 
keff =  1 + cnt  kbf can be well predicted
 3kbf 
by conventional EMT
Nan et al. (2004) φ, kbf, kcnt EMT approached model
φ p kcnt / kbf
( )
keff = 1 + kxx, kzz: equivalent TC with ITC
along with each axis
3 p + 2aK kcnt
dcnt kbf

Volume 1 · Number 4 · 2010

Xue (2005) φ, kbf, kcnt EMT approach with
 (1 − φ ) + ( 4φ / π ) kcnt / kbf arctan π / 4 kcnt / kbf
( )  k Distribution function orientation distribution
keff = bf
bf cnt cnt bf
 (1 − φ ) + ( 4φ / π ) k / k arctan π / 4 k / k (
 ) 
cnt + kbf bf
 (1 − φ ) + 2φ kcnt / kcnt − kbf
{ ( )} ln {( k ) / 2k }  k
keff =  bf
bf cnt bf cnt + kbf bf
 (1 − φ ) + 2φ k / k − k
{ ( )} ln {( k
 ) / 2k } 
Murshed et al.
  kp  0.52φ  k p  φ, kp, kbf, Thermal resistance
kbf 1 + 0.27φ 4 / 3  − 1  1 + 1/ 3 
− 1 
Other (2006)   kbf    1 − φ  kbf   approach with
keff =
mechanism  k p   0.52φ  geometrical structure of
1/ 3
1 + φ 4/3  − 1  + 0.27φ1/ 3 + 0.27
based  kbf  1− φ  nanoparticle
models
3A 3C1 3C2
Gao and Zhou  kbf   kbf + B1   kbf + B2  φ, kp, kbf, Differential EMT
(2006) 1− φ =   k +B  k +B  with geometrical structure
 keff   bf 1  eff 2
Lz: depolarization factor

Xue (2006) keff − kbf  keff − kzz 4 keff − k xx  φ, kp, kbf,

+φ +
( ) =0
9 (1 − φ ) kxx, kzz: equivalent TC Similar with
2 keff + kbf  keff + 0.14 ( 2acnt / Lcnt ) kzz − keff
 ( 2 keff + 0.5 k xx − keff
) ( ) 
 along with each axis Nan et al. 04
Minnich and Chen φ, kp, kbf EMT approach with
(2007) 1 1   k p (1 + 2κ ) + 2 kbf + 2φ k p (1 − κ ) − kbf  Φ: Interface density surface phonon
Ji-Hwan Lee, Seung-Hyun Lee, Chul Jin Choi, Seok Pil Jang, and Stephen U. S. Choi

keff =  cv ,bf v pho,bf 

{ }  scattering
 3 ( )
1 / lbf + ( Φ / 4 )   k p (1 + 2κ ) + 2 kbf − φ k p (1− κ ) − kbf { } 

Jung and Yoo  k p + 2 kbf + 2 k p − kbf φ  φ l cˆ φ l p cˆv Aec2 l p φ, kp, kbf EMT with BM
313

( ) kB T
keff =   kbf + p v +
(2009)  k p + 2 kbf − k p − kbf φ  3 3πµ d p lbf 3 −1/ 3 2 Zeta potential and EDL effects
 (  ) ε bf m p a p φ( )
314 A Review of Thermal Conductivity Data, Mechanisms and Models for Nanofluids

5.3. Models
Classical EMT-based static models fail to predict the magnitude and trend of the thermal conductivity
of nanofluids, although there are some data supporting EMT. New static models are much better than
classical models based on the effective medium theories in narrowing the gap between thermal
conductivity enhancement data and model predictions. However, they are unable to capture other
features such as the dependence of the thermal conductivity of nanofluids on nanoparticle size and fluid
temperature. However, new models based on a combination of static and dynamic mechanisms appear
to predict not only the magnitude of the enhanced thermal conductivity, but also other features of the
thermal conduction behavior of nanofluids.

5.4. Future research directions

The inconsistencies in the reported thermal conductivity data from different groups can be due to
differences in sample quality, dependence of thermal conductivity on many factors, and differences in
measurement uncertainties. The first step to resolve data inconsistencies is to produce high-quality
nanofluids having both suspension stability and high conductivity, preferably without any dispersants
or stabilizers especially for reference nanofluids, using one-step processes known for making such
nanofluids. Nanofluids produced in this way for thermal conductivity measurements should then be
characterized, including their detailed microstructures such as particle size, shape, distribution, the
levels of agglomeration/clustering, and surface properties and detailed dynamics of nanoparticles such
as particle motion and interactions with base fluids. The next step to reduce data inconsistencies due to
differences in sample quality and differences in measurement uncertainties is to conduct round-robin
tests using identical high-quality reference samples and proven accurate methods and apparatuses for
measuring the thermal conductivity of nanofluids.
The controversy regarding the proposed mechanisms of the thermal conductivity of nanofluids
comes primarily from a lack of understanding of the scientific basis for these mechanisms. Therefore,
the proposed concepts and mechanisms behind the thermal conductivity behavior of nanofluids remain
to be validated. More systematic experiments with well-dispersed, well-characterized nanofluids and a
better understanding of the physics of fluid flow and heat transfer at the nanoscale are needed to
establish the underlying mechanisms of heat conduction in nanofluids.
The classical EMT-based models with the assumptions of motionless particles and the diffusive
heat transfer in both continuous matrix phase and dispersed phase can predict thermal conductivity
of suspensions of micrometer- or larger-sized particles but fail to predict the thermal conductivity
data for most nanofluids. New models based on a combination of static and dynamic mechanisms
appear to predict not only the magnitude of the enhanced thermal conductivity, but also other
features of the thermal conduction behavior of nanofluids. However, the mechanisms used in the
new models remain to be validated. In the future, comprehensive science-based models should be
developed based on new mechanisms as well as the classical diffusion mechanisms. Development
of comprehensive science-based models that are capable of accurately predicting the new novel
features including the size and temperature dependent thermal conductivity of nanofluids will
further advance the new field of nanofluids.

ACKNOWLEDGEMENTS
This work was supported by the National Research Foundation of Korea Grant funded by the Korean
Government (NRF-2010-013-D00006).

REFERENCES
[1] ITRS, “Emerging Research Devices,” International Technology Roadmap for Semiconductors,
2008. (http://www.itrs.net/Links/2008ITRS/Update/2008_Update.pdf)
[2] FreedomCAR Power Electronics and Electrical Machines, “Electrical and Electronics Technical
Team Roadmap,” Washington, D.C., Nov. 2006.

International Journal of Micro-Nano Scale Transport

Ji-Hwan Lee, Seung-Hyun Lee, Chul Jin Choi, Seok Pil Jang, and Stephen U. S. Choi 315

[3] Choi, S. U. S., “Enhancing thermal conductivity of fluids with nanoparticles,” in: D. A. Singer
and H. P. Wang (Eds.), Developments and Applications of Non-Newtonian Flows, FED-vol.
231/MD-vol. 66, ASME, New York, pp. 99–105, 1995.
[4] Choi, S. U. S., “Nanofluids: From Vision to Reality Through Research,” ASME Journal of Heat
Transfer, vol. 131, no. 3, pp. 033106-1–033106-9, 2009.
[5] Eastman, J. A., Choi, S. U. S., Yu, W., and Thompson, L. J., “Anomalously increased effective
thermal conductivity of ethylene glycol-based nanofluids containing copper nanoparticles,”
Applied Physics Letters, Vol. 78, no. 6, pp. 718–720, 2001.
[6] Choi, S. U. S., Zhang, Z. G., Yu, W., Lockwood, F. E., and Grulke, E. A., “Anomalous thermal
conductivity enhancement in nanotube suspensions,” Applied Physics Letters, Vol. 79, no. 14,
pp. 2252–2254, 2001.
[7] Patel, H. E., Das, S. K., Sundararajan, T., Nair, A. S., George, B., and Pradeep, T., “Thermal
conductivities of naked and monolayer protected metal nanoparticle based nanofluids:
Manifestation of anomalous enhancement and chemical effects,” Applied Physics Letters, vol.
83, no. 14, 2931–2933, 2003.
[8] Das, S. K., Putra, N., Thiesen, P., and Roetzel, W., “Temperature Dependence of Thermal
Conductivity Enhancement for Nanofluids,” Journal of Heat Transfer, Transactions of the
ASME, vol. 125, no. 4, pp. 567–574, 2003.
[9] Hong, K. S., Hong, T. K., and Yang, H. S., “Thermal conductivity of Fe nanofluids depending
on the cluster size of nanoparticles,” Applied Physics Letters, vol. 88, no. 3, 031901-1–031901-
3, 2006.
[10] Li, C. H. and Peterson, G. P., “Experimental investigation of temperature and volume fraction
variations on the effective thermal conductivity of nanoparticle suspensions (nanofluids),”
Journal of Applied Physics, vol. 99, no. 8, 084314-1–084314-8, 2006.
[11] Chopkar, M., Das, P. K., and Manna, I., “Synthesis and characterization of nanofluid for
advanced heat transfer applications,” Scripta Materialia, vol. 55, no. 6, pp. 549–552, 2006.
[12] Jana, S., Salehi-Khojin, A., and Zhong, W.-H., “Enhancement of fluid thermal conductivity by
the addition of single and hybrid nano-additives,” Thermochimica Acta, vol. 462, no. 1–2, pp.
45–55, 2007.
[13] Lee, J.-H., Hwang, K.S., Jang, S.P., Lee, B.H., Kim, J.H., Choi, S.U.S., and Choi, C.J.,
“Effective viscosities and thermal conductivities of aqueous nanofluids containing low volume
concentrations of Al2O3 nanoparticles,” International Journal of Heat and Mass Transfer, vol.
51, no. 11-12, pp. 2651-2656, 2008.
[14] Xuan, Y. and Li, Q., “Investigation on Convective Heat Transfer and Flow Features of
Nanofluids,” Journal of Heat Transfer, Transactions of the ASME, vol. 125, no. 1, pp. 151–155,
2003.
[15] Wen, D. and Ding, Y., “Experimental investigation into convective heat transfer of nanofluids at
the entrance region under laminar flow conditions,” International Journal of Heat and Mass
Transfer, vol. 47, no. 24, pp. 5181–5188, 2004.
[16] Ding, Y., Alias, H., Wen, D., and Williams, R. A., “Heat transfer of aqueous suspensions of
carbon nanotubes (CNT nanofluids),” International Journal of Heat and Mass Transfer, vol. 49,
no. 1–2, pp. 240–250, 2006.
[17] Nguyen, C. T., Roy, G., Gauthier, C., and Galanis, N., “Heat transfer enhancement using
Al2O3–water nanofluid for an electronic liquid cooling system,” Applied Thermal Engineering,
vol. 27, no. 8–9, pp. 1501–1506, 2007.
[18] Hwang, K.S., Jang, S.P., and Choi, S.U.S., “Flow and convective heat transfer characteristics of
water-based Al2O3 nanofluids in fully developed laminar flow regime,” International Journal
of Heat and Mass Transfer, vol. 52, no. 1-2, pp. 193-199, 2009.

Volume 1 · Number 4 · 2010

316 A Review of Thermal Conductivity Data, Mechanisms and Models for Nanofluids

[19] Xie, H., Li, Y, and Yu, W., “Intriguingly high convective heat transfer enhancement of nanofluid
coolants in laminar flows,” Physics Letters A, vol. 374, no. 25, pp. 2566–2568, 2010.
[20] Wasan, D. T. and Nikolov, A. D., “Spreading of nanofluids on solids,” Nature, vol. 423, no.
6936, pp. 156–159, 2003.
[21] Kim, J.-K., Jung, J. Y., and Kang, Y. T., “The effect of nano-particles on the bubble absorption
performance in a binary nanofluid,” International Journal of Refrigeration, vol. 29, no. 1, pp.
22–29, 2006.
[22] Wang, L. and Fan, J., “Nanofluids Research: Key Issues,” Nanoscale Research Letters, vol. 5,
no. 8, pp. 1241–1252, 2010.
[23] Lee, S., Choi, S. U. S., Li, S., and Eastman, J. A., “Measuring Thermal Conductivity of Fluids
Containing Oxide Nanoparticles,” Journal of Heat Transfer, Transactions of the ASME, vol. 121,
no. 2, pp. 280–289, 1999.
[24] Wang, X., Xu, X., and Choi, S. U. S., “Thermal Conductivity of Nanoparticle-Fluid Mixture,”
Journal of Thermophysics and Heat Transfer, vol. 13, no. 4, pp. 474–480, 1999.
[25] Eastman, J. A., Choi, S. U. S., Li, S., Thompson, L. J., Lee, S., “Enhanced thermal conductivity
through the development of nanofluids,” In: Komarneni, S., Parker, J.C., Wollenberger, H.J.
(Eds.), Nanophase and Nanocomposite Materials II. MRS, Pittsburg, PA, pp. 3–11, 1997.
[26] Xuan, Y. and Li, Q., “Heat transfer enhancement of nanofuids,” International Journal of Heat
and Fluid Flow, vol. 21, no. 1, pp. 58–64, 2000.
[27] Liu, M.-S., Lin, M. C.-C., Tsai, C. Y., and Wang, C.-C., “Enhancement of thermal conductivity
with Cu for nanofluids using chemical reduction method,” International Journal of Heat and
Mass Transfer, vol. 49, no. 17–18, pp. 3028–3033, 2006.
[28] Garg, J., Poudel, B., Chiesa, M., Gordon, J. B. Ma, J. J., Wang, J. B., Ren, Z. F. Kang, Y. T.,
Ohtani, H., Nanda, J., McKinley, G. H., and Chen, G., “Enhanced thermal conductivity and
viscosity of copper nanoparticles in ethylene glycol nanofluid,” Journal of Applied Physics, vol.
103, no. 7, pp. 074301-1–074301-6, 2008.
[29] Maxwell, J. C., “ATreatise on Electricity and Magnetism,” 1st Edition, vol. 1, Clarendon Press,
Oxford, U.K., pp. 360–366, 1873.
[30] Schmidt, A. J., Chiesa, M., Torchinsky, D. H., Johnson, J. A., Nelson, K. A., and Chen, G.,
“Thermal conductivity of nanoparticle suspensions in insulating media measured with a transient
optical grating and a hotwire,” Journal of Applied Physics, vol. 103, no. 8, pp. 083529-
1–083529-5, 2008.
[31] Philip, J., Shima, P. D., and Raj, B., “Nanofluid with tunable thermal properties,” Applied
Physics Letters, vol. 92, no. 4, pp. 043108-1–043108-3, 2008.
[32] Putnam, S. A., Cahill, D. G., Braun, P. V., Ge, Z., and Shimmin, R. G., “Thermal conductivity
of nanoparticle suspensions,” Journal of Applied Physics, vol. 99, no. 8, 084308-1–084308-6,
2006.
[33] Zhang, X., Gu, H., and Fujii, M., “Effective thermal conductivity and thermal diffusivity of
nanofluids containg spherical and cylindrical nanoparticles,” Journal of Applied Physics, vol.
100, no. 4, pp. 044325-1–044325-5, 2006.
[34] Eapen, J., Williams, W. C., Buongirono, J., Hu, L-w., Yip, S., Rusconi, R., and Piazza, R.,
“Mean-Field Versus Microconvection Effects in Nanofluid Thermal Condition,” Physical
Review Letter, Vol. 99, no. 9, pp. 095901-1–095901-4, 2007.
[35] Timofeeva, E. V., Gavrilov, A. N., McCloskey, J. M., and Tolmachev, Y. V., Sprunt, S., Lopatina,
L. M., Selinger, J. V., “Thermal conductivity and particle agglomeration in alumina nanofluids:
Experiment and theory,” Physical Review E, vol. 76, no. 6, pp. 061203-1–061203-15, 2007.
[36] Buongiorno et al., “A benchmark study on the thermal conductivity of nanofluids,” Journal of
Applied Physics, vol. 106, no. 9, pp. 094312-1–094312-14, 2009.

International Journal of Micro-Nano Scale Transport

Ji-Hwan Lee, Seung-Hyun Lee, Chul Jin Choi, Seok Pil Jang, and Stephen U. S. Choi 317

[37] Jeffrey, D. J., “Conduction Through a Random Suspension of Spheres,” Proceedings of the
Royal Society of London, Series A, Mathematical and Physical Sciences, vol. 335, no. 1602, pp.
355–367, 1973.
[38] Xie, H., Wang, J., Xi, T., Liu, Y., Ai, F., and Wu, Q., “Thermal conductivity enhancement of
suspensions containing nanosized alumina particles,” Journal of Applied Physics, vol. 91, no. 7,
pp. 4568–4572, 2002.
[39] Murshed, S. M. S., Leong, K. C., and Yang, C., “A Model for Predicting the Effective Thermal
Conductivity of Nanoparticle–Fluid Suspensions” International Journal of Nanoscience, Vol. 5,
no. 1, pp. 23–33, 2006.
[40] Patel, H. E., Sundararajan, T., and Das, S. K., “An experimental investigation into the thermal
conductivity enhancement in oxide and metallic nanofluids,” Journal of Nanoparticle Research,
vol. 12, no. 3, pp. 1015–1031, 2010.
[41] Xie, H., Lee, H., Youn, W., and Choi, M., “Nanofluids containing multiwalled carbon nanotubes
and their enhanced thermal conductivities,” Journal of Applied Physics, vol. 94, no. 8, pp.
4967–4971, 2003.
[42] Wen, D. and Ding, Y., “Effective Thermal Conductivity of Aqueous Suspensions of Carbon
Nanotubes (Carbon Nanotube Nanofluids),” Journal of Thermophysics and Heat Transfer, vol.
18, No. 4, pp. 481–485, 2004.
[43] Shaikh, S., Lafdi, K., and Ponnappan, R., “Thermal conductivity improvement in carbon
nanoparticle doped PAO oil: An experimental study,” Journal of Applied Physics, vol. 101, No.
6, pp. 064302-1–064302-7, 2007.
[44] Murshed, S. M. S., Leong, K. C., and Yang, C., “Enhanced thermal conductivity of TiO2–water
based nanofluids,” International Journal of Thermal Sciences, vol. 44, no. 4, pp. 367–373, 2005.
[45] Hong, T. K., and Yang, H. S., and Choi, C. J., “Study of the enhanced thermal conductivity of
Fe nanofluids,” Journal of Applied Physics, vol. 97, no. 6, 064311-1–064311-4, 2005.
[46] Chon, C. H., Kihm, K. D., Lee, S. P. and Choi, S.U.S., “Empirical correlation finding the role of
temperature and particle size for nanofluid (Al2O3) thermal conductivity enhancement,”
Applied Physics Letter, vol. 87, no.15, pp. 153107-1–153107-3, 2005.
[47] Li, C. H. and Peterson, G. P., “The effect of particle size on the effective thermal conductivity
of Al2O3-water nanofluids,” Journal of Applied Physics, vol. 101, no. 4, 044312-1–044312-5,
2007.
[48] Murshed, S. M. S., Leong, K. C., and Yang, C., “Investigations of thermal conductivity and
viscosity of nanofluids,” International Journal of Thermal Sciences, vol. 47, no. 5, pp. 560–568,
2008.
[49] Jha, N. and Ramaprabhu, S., “Synthesis and Thermal Conductivity of Copper Nanoparticle
Decorated Multiwalled Carbon Nanotubes Based Nanofluids,” The Journal of Physical
Chemistry C, vol. 112, no. 25, pp. 9315–9319, 2008.
[50] Mintsa, H. A., Roy, G., Nguyen, C. T. and Doucet, D., “New temperature dependent thermal
conductivity data for water-based nanofluids,” International Journal of Thermal Sciences, vol.
48, no.2, pp. 363–371, 2009.
[51] Vajjha, R. S. and Das, D. K., “Experimental determination of thermal conductivity of three
nanofluids and development of new correlations,” International Journal of Heat and Mass
Transfer, vol. 52, no. 21–22, pp. 4675–4682, 2009.
[52] Yu, W., Xie, H., Chen, L., and Li, Y., “Investigation on the thermal transport properties of
ethylene glycol-based nanofluids containing copper nanoparticles,” Powder Technology, vol.
197, no. 3, pp. 218–221, 2010.
[53] Paul, G., Pal, T., and Manna, I., “Thermo-physical property measurement of nano-gold dispersed
water based nanofluids prepared by chemical precipitation technique,” Journal of Colloid and
Interface Science, vol. 349, no. 1, pp. 434–437, 2010.

Volume 1 · Number 4 · 2010

318 A Review of Thermal Conductivity Data, Mechanisms and Models for Nanofluids

[54] Venerus, D. C., Kabadi, M. S., Lee, S. and Perez-Luna, V., “Study of thermal transport in
nanoparticle suspension using forced Rayleigh scattering,” Journal of Applied Physics, vol.100,
no. 9, pp. 094310-1–094310-5, 2006.
[55] Yang, B. and Han, Z. H., “Temperature-dependent thermal conductivity of nanorods-based
nanofluids,” Applied Physics Letters, vol. 89, no. 8, pp. 083111-1–083111-3, 2006.
[56] Beck, M. P., Yuan, Y., Warrier, P., and Teja, A. S., “The thermal conductivity of alumina
nanofluids in water, ethylene glycol, and ethylene glycol __water mixtures,” Journal of
Nanoparticle Research, vol.12, no. 4, pp. 1469–1477, 2010.
[57] Frens. G., “Controlled Nucleation for the Regulation of the Particle Size in Monodisperse Gold
Suspensions,” Nature: Nature Physical Science, vol. 241, no. 105, pp. 20–22, 1973.
[58] Kim, S. H., Choi, S. R., and Kim, D., “Thermal Conductivity of Metal-Oxide Nanofluids:
Particle Size Dependence and Effect of Laser Irradiation,” ASME Journal of Heat Transfer, vol.
129, no. 3, pp. 298–307, 2007.
[59] Chen, G., Yu, W., Singh, D., Cookson, and D., Routbort, J., “Application of SAXS to the study
of particle-size-dependent thermal conductivity in silica nanofluids,” Journal of Nanoparticle
Research, vol. 10, no. 7, pp.1109–1114, 2008.
[60] Beck, M.P., Yuan, Y., Warrier, P., and Teja, A.S., “The effect of particle size on the thermal
conductivity of alumina nanofluids,” Journal of Nanoparticle Research, vol. 11, no. 5, pp. 1129-
1136, 2009.
[61] Lee, D., Kim, J.-W., and Kim, B. G., “A New Parameter to Control Heat Transport in
Nanofluids: Surface Charge State of the Particle in Suspension,” The Journal of Physical
Chemistry B, vol. 110, no. 9, pp. 4323–4328, 2006.
[62] Philip, J., Shima, P. D., and Raj, B., “Enhancement of thermal conductivity in magnetite based
nanofluid due to chainlike structures,” Applied Physics Letters, vol. 91, no. 20, pp. 203108-
1–203108-3, 2007.
[63] Philip, J., Shima, P. D., and Raj, B., “Evidence for enhanced thermal conduction through
percolating structures in nanofluids,” Nanotechnology, vol. 19, no. 30, pp. 305706-1–305706-7,
2008.
[64] Shima, P. D., Philip, J., and Raj, B., “Magnetically controllable nanofluid with tunable thermal
conductivity and viscosity,” Applied Physics Letters, vol. 95, no. 13, pp. 133112-1–133112-3,
2009.
[65] Das, S. K., Putra, N., and Roetzel, W., “Pool boiling characteristics of nano-fluids,”
International Journal of Heat and Mass Transfer, vol. 46, no. 5, pp. 851–862, 2003.
[66] Das, S. K., Putra, N., and Roetzel, W., “Pool boiling of nano-fluids on horizontal narrow tubes,”
International Journal of Multiphase Flow, vol. 29, no. 8, pp. 1237–1247, 2003.
[67] Yu, Q., Kim, Y. J., and Ma, H., “Nanofluids with plasma treated diamond nanoparticles,” Applied
Physics Letters, vol. 92, no. 10, pp. 103111-1–103111-3, 2008.
[68] Xie, H., Wang, J., Xi, T., Liu, Y., “Thermal Conductivity of Suspensions Containing Nanosized
SiC Particles,” International Journal of Thermophysics, vol. 23, no. 2, pp. 571–580, 2002.
[69] Xuan, Y., Li, Q., and Hu, W., “Aggregation Structure and Thermal Conductivity of Nanofluids,”
AIChE Journal, vol. 49, no. 4, pp. 1038–1043, 2003.
[70] Lee, D., “Thermophysical Properties of Interfacial Layer in Nanofluids,” Langmuir, vol. 23, no.
11, pp. 6011–6018, 2007.
[71] Bang, I. C. and Heo, G., “An axiomatic design approach in development of nanofluid coolants,”
Applied Thermal Engineering, vol. 29, no. 1, pp. 75–90, 2009.
[72] Suh, N. P., “Axiomatic Design: Advances and Applications,” Oxford University Press, New
York, NY, USA, 2001.

International Journal of Micro-Nano Scale Transport

Ji-Hwan Lee, Seung-Hyun Lee, Chul Jin Choi, Seok Pil Jang, and Stephen U. S. Choi 319

[73] Buongiorno, J., Hu, L-W., McKrell, T., and Prabhat, N. Report and analysis of INPBE results—
thermal conductivity. International nanofluid properties benchmark exercise INPBE, 2009.
http://mit.edu/nse/nanofluids/benchmark/workshop/mit.pdf
[74] Bruggeman, D. A. G., “BerechnungVerschiedener Physikalischer Konstanten von Heterogenen
Substanzen, I. Dielektrizitatskonstanten und Leitfahigkeiten der Mischkorper aus Isotropen
Substanzen,” Annalen der Physik. Leipzig, vol. 24, pp. 636–679, 1935.
[75] Lee, W.-H. et al., “Round-Robin Test on Thermal Conductivity Measurement of ZnO Nanofluids
and Comparison of Experimental Results with Theoretical Bounds,” Nanoscale Research
Letters, vol. 6, p. 258, 2011.
[76] Nan, C.-W., Birringer, R., Clarke, D.R., and Gleiter, H. “Effective thermal conductivity of
particulate composites with interfacial thermal resistance,” Journal of Applied Physics, Vol. 81,
no. 10, pp. 6692–6699, 1997.
[77] Chandrasekar, M. and Suresh, S., “A Review on the Mechanisms of Heat Transport in
Nanofluids,” Heat Transfer Engineering, vol. 30, no.14, pp. 1136–1150, 2009.
[78] Yu, W., France, D. M., Singh, D., Timofeeva, E. V., Smith, D. S., and Routbort, J. L.,
“Mechanisms and Models of Effective Thermal Conductivities of Nanofluids,” Journal of
Nanoscience and Nanotechnology, vol. 10, no. 8, pp. 4824–4849, 2010.
[79] Yu, C.-J., Richter, A. G., Datta, A., Durbin, M. K., and Dutta, P., “Molecular layering in a liquid
on a solid substrate: an X-ray reflectivity study,” Physica B: Condensed Matter, vol. 283, no.
1–3, pp. 27–31, 2000.
[80] Henderson, J. R. and van Swol, F., “Theory and simulations of a hard sphere fluid at a hard
wall,” Molecular Physics, vol. 51, no. 4, pp. 991–1010, 1984.
[81] Keblinski, P., Phillpot, S. R., Choi, S. U. S., and Eastman, J. A., “Mechanisms of heat flow in
suspensions of nano-sized particles (nanofluids),” International Journal of Heat and Mass
Transfer, vol. 45, no. 4, pp. 855–863, 2002.
[82] Yu, W. and Choi, S. U. S., “The role of interfacial layers in the enhanced thermal conductivity
of nanofluids: Arenovated Maxwell model,” Journal of Nanoparticle Research, vol. 5, no. 1–2,
pp. 167–171, 2003.
[83] Yu, W. and Choi, S. U. S., “The role of interfacial layers in the enhanced thermal conductivity
of nanofluids: A renovated Hamilton–Crosser model,” Journal of Nanoparticle Research, vol. 6,
no. 4, pp. 355–361, 2004.
[84] Wang, B. X., Zhou, L. P., and Peng, X. F., “A fractal model for predicting the effective thermal
conductivity of liquid with suspension of nanoparticles,” International Journal of Heat and
Mass Transfer, vol. 46, no. 14, pp. 2665–2672, 2003.
[85] Prasher, R., Phelan, P. E., and Bhattacharya, P., “Effect of aggregation kinetics on the thermal
conductivity of nanoscale colloidal solutions (nanofluid),” Nano Letters, vol. 6, no. 7, pp.
1529–1534, 2006.
[86] Prasher, R., Evans, W., Meakin, P., Fish, J., Phelan, P., and Keblinski, P., “Effect of aggregation
on thermal conduction in colloidal nanofluids,” Applied Physics Letters, vol. 89, no. 14, pp.
143119-1–143119-3, 2006.
[87] Jang, S. P. and Choi, S. U. S., “Role of Brownian motion in the enhanced thermal conductivity
of nanofluids,” Applied Physics Letters, Vol. 84, no. 21, pp. 4316–4318, 2004.
[88] Koo, J. and Kleinstreuer, C., “A new thermal conductivity model for nanofluids,” Journal of
Nanoparticle Research, vol. 6, no. 6, pp. 577–588, 2004.
[89] Prasher, R., Bhattacharya, P., and Phelan, P. E., “Thermal conductivity of nanoscale colloidal
solutions (nanofluids),” Physical Review Letters, vol. 94, no. 2, pp. 025901-1–025901-4, 2005.
[90] Patel, H. E., Sundararajan, T., Pradeep, T., Dasgupta, A., Dasgupta, N., and Das, S. K., “A
microconvection model for thermal conductivity of nanofluids,” PRAMANA—Journal of
physics, vol. 65, no. 5, pp. 863–869, 2005.

Volume 1 · Number 4 · 2010

320 A Review of Thermal Conductivity Data, Mechanisms and Models for Nanofluids

[91] Ren, Y., Xie, H., and Cai, A., “Effective thermal conductivity of nanofluids containing spherical
nanoparticles,” Journal of Physics D: Applied Physics, vol. 38, no. 21, pp. 3958–3961, 2005.
[92] Evans, W., Fish, J., and Keblinski, P., “Role of Brownian motion hydrodynamics on nanofluid
thermal conductivity,” Applied Physics Letters, Vol. 88, no. 9, pp. 093116-1–093116-3, 2006.
[93] Vladkov, M. and Barrat, J.-L., “Modeling transient absorption and thermal conductivity in a
simple nanofluid,” Nano Letters, vol. 6, no. 6, pp. 1224–1228, 2006.
[94] Sarkar, S. and Selvam, R. P., “Molecular dynamics simulation of effective thermal conductivity
and study of enhanced thermal transport mechanism in nanofluids,” Journal of Applied Physics,
vol. 102, no. 7, pp. 074302-1–074302-7, 2007.
[95] Li, C. H. and Peterson, G. P., “Mixing effect on the enhancement of the effective thermal
conductivity of nanoparticle suspensions (nanofluids),” International Journal of Heat and Mass
Transfer, vol. 50, no. 23–24, pp. 4668–4677, 2007.
[96] Wang, L. and Wei, X., “Nanofluids: synthesis, heat conduction, and extension,” Journal of Heat
Transfer, Transactions of the ASME, vol. 131, no. 3, pp. 033102-1–033102-7, 2009.
[97] Xue, L., Keblinski, P., Phillpot, S. R., Choi, S. U.-S., and Eastman, J. A., “Effect of liquid
layering at the liquid–solid interface on thermal transport,” International Journal of Heat and
Mass Transfer, vol. 47, no. 19–20, pp. 4277–4284, 2004.
[98] Kumar S. and Murthy, J. Y., “A numerical technique for computing effective thermal
conductivity of fluid–particle mixtures,” Numerical Heat Transfer, Part B: Fundamentals, vol.
47, no. 6, pp. 555–572, 2005.
[99] Evans, W., Fish, J., and Keblinski, P., “Thermal conductivity of ordered molecular water,” The
Journal of Chemcal Physics, Vol. 126, no. 15, pp. 154504-1–1545046-4, 2007.
[100] Das, S. K., Choi, S. U. S., and Patel, H. E., “Heat Transfer in Nanofluids – A Review,” Heat
Transfer Engineering, vol. 27, no.10, pp. 3–19, 2006.
[101] Maxwell Garnett, J. C., “Colours in metal glasses and in metallic films,” Philosophical
Transactions of the Royal Society of London, Series A, Containing Papers of a Mathematical of
Physical Character, vol. 203, pp. 385–420, 1904.
[102] Hamilton, R. L. and Crosser, O. K., “Thermal conductivity of heterogeneous two-component
system,” Industrial & Engineering Chemistry Fundamentals, vol. 1, no. 3, pp. 187–191, 1962.
[103] Hashin, Z. and Shtrikman, S., “A variational approach to the theory of the effective magnetic
permeability of multiphase materials,” Journal of Applied Physics, Vol. 33, no. 10, pp.
3125–3131, 1962.
[104] Keblinski, P., Prasher, R., and Eapen, J., “Thermal conductance of nanofluids: is the controversy
over?” Journal of Nanoparticle Research, vol. 10, no. 7, pp. 1089–1097, 2008.
[105] Davis, R. H., “The effective thermal conductivity of a composite material with spherical
inclusions,” International Journal of Thermophysics, vol. 7, no. 3, pp. 609–620, 1986.
[106] Hasselman, D. P. H. and Johnson, L. F., “Effective thermal conductivity of composites with
interfacial thermal barrier resistance,” Journal of Composite Materials, vol. 21, no. 6, pp.
508–515, 1987.
[107] Xue, Q. Z., “Model for effective thermal conductivity of nanofluids,” Physics Letters A, vol.
307, no. 5–6, pp. 313–317, 2003.
[108] Kim, J., Kang, Y. T., and Choi, C. K., “Analysis of convective instability and heat transfer
characteristics of nanofluids,” Physics of Fluids, vol. 16, no. 7, pp. 2395–2401, 2004.
[109] Xue, Q. and Xu, W. M., “A model of thermal conductivity of nanofluids with interfacial shells,”
Materials Chemistry and Physics, vol. 90, no. 2–3, pp. 298–301, 2005.
[110] Xie, H., Fujii, M., and Zhang, X., “Effect of interfacial nanolayer on the effective thermal
conductivity of nanoparticle-fluid mixture,” International Journal of Heat and Mass Transfer,
vol. 48, no. 14, pp. 2926–2932, 2005.

International Journal of Micro-Nano Scale Transport

Ji-Hwan Lee, Seung-Hyun Lee, Chul Jin Choi, Seok Pil Jang, and Stephen U. S. Choi 321

[111] Leong, K. C., Yang, C., and Murshed, S. M. S., “A model for the thermal conductivity of
nanofluids – the effect of interfacial layer,” Journal of Nanoparticle Research, vol. 8, no. 2, pp.
245–254, 2006.
[112] Doroodchi, E., Evans, T. M., and Moghtaderi, B., “Comments on the effect of liquid layering on
the thermal conductivity of nanofluids,” Journal of Nanoparticle Research, vol. 11, no. 6, pp.
1501–1507, 2009.
[113] Hashimoto, T., Fujimura, M., and Kawai, H., “Domain-Boundary Structure of Styrene–Isoprene
Block Copolymer Films Cast from Solutions. 5. Molecular-Weight Dependence of Spherical
Microdomains,” Macromolecules, vol. 13, no. 6, pp. 1660–1669, 1980.
[114] Sabbaghzadeh, J. and Ebrahimi, S., “Effective thermal conductivity of nanofluids containing
cylindrical nanoparticles,” International Journal of Nanoscience, Vol. 6, no. 1, pp. 45–49, 2007.
[115] Murshed, S. M. S., Leong, K. C., and Yang, C., “A combined model for the effective thermal
conductivity of nanofluids,” Applied Thermal Engineering, vol. 29, no. 11–12, pp. 2477–2483,
2009.
[116] Li, L., Zhang, Y., and Ma, H., “Molecular dynamics simulation of effect of liquid layering
around the nanoparticle on the enhanced thermal conductivity of nanofluids,” Journal of
Nanoparticle Research, vol.12, no. 3, pp. 811–821, 2010.
[117] Tillman, P., and Hill, J. M., “Determination of nanolayer thickness for a nanofluid,”
International Communications in Heat and Mass Transfer, vol. 34, no. 4, pp. 399–407, 2007.
[118] Kumar, D. H., Patel, H. E., Kumar, V. R. R., Sundararajan, T., Pradeep, T., and Das, S. K.,
“Model for heat conduction in nanofluids,” Physical Review Letters, vol. 93, no. 14, pp. 144301-
1–144301-4, 2004.
[119] Patel, H. E., Sundararajan, T., and Das, S. K., “A cell model approach for thermal conductivity
of nanofluids,” Journal of Nanoparticle Research, vol. 10, no. 1, pp. 87–97, 2008.
[120] Shukla, R. K. and Dhir, V. K., “Effect of Brownian motion on thermal conductivity of
nanofluids,” Journal of Heat Transfer, vol. 130, no. 4, pp. 042406-1–042406-13, 2008.
[121] Yang, B., “Thermal conductivity equations based on Brownian motion in suspensions of
nanoparticles (nanofluids),” Journal of Heat Transfer, vol. 130, no. 4, pp. 042408-1–042408-5,
2008.
[122] Bhattacharya, P., Saha, S. K., Yadav, A., Phelan, P. E., and Prasher, R. S., “Brownian dynamics
simulation to determine the effective thermal conductivity of nanofluids,” Journal of Applied
Physics, vol. 95, no. 11, pp. 6492–6494, 2004.
[123] Jain, S., Patel, H. E., and Das, S. K., “Brownian dynamics simulation for the prediction of
effective thermal conductivity of nanofluid,” Journal of Nanoparticle Research, vol. 11, no. 4,
pp. 767–773, 2009.
[124] Vasu, V., Krishna, K. R., and Kumar, A. C. S., “Analytical prediction of thermophysical
properties of fluids embedded with nanostructured materials,” International Journal of
Nanoparticles, vol. 1, no. 1, pp. 32–49, 2008.
[125] Feng, Y., Yu, B., Xu, P., and Zou, M., “The effective thermal conductivity of nanofluids based
on the nanolayer and the aggregation of nanoparticles,” Journal of Physics D: Applied Physics,
vol. 40, no. 10, pp. 3164–3171, 2007.
[126] Shih, W.–H., Shih, W. Y., Kim, S.–I., Liu, J., and Aksay, I. A., “Scaling behavior of the elastic
properties of colloidal gels,” Physical Review A: Atomic, molecular, and optical physics, vol. 42,
no. 8, pp. 4772–4779, 1990.
[127] de Rooji, R., Potanin, A. A., van den Ende, D., and Mellema, J., “Steady shear viscosity of
weakly aggregating polystyrene latex dispersions,” Journal of Chemical Physics, vol. 99, no. 11,
pp. 9213–9223, 1993.

Volume 1 · Number 4 · 2010

 322 A Review of Thermal Conductivity Data, Mechanisms and Models for Nanofluids

[128] Evans, W., Prasher, R., Fish, J., Meakin, P., Phelan, P., and Keblinski, P., “Effect of aggregation
and interfacial thermal resistance on thermal conductivity of nanocomposites and colloidal
nanofluids,” International Journal of Heat and Mass Transfer, vol. 51, no. 5–6, pp. 1431–1438,
2008.
[129] Nan, C.-W., Shi, Z., and Lin, Y., “A simple model for thermal conductivity of carbon nanotube-
based composites,” Chemical Physics Letters, vol. 375, no. 5–6, pp. 666–669, 2003.
[130] Nan, C.-W., Liu, G., Lin, Y., and Li, M., “Interface effect on thermal conductivity of carbon
nanotube composites” Applied Physics Letters, vol. 85, no. 16, pp. 3549–3551, 2004.
[131] Xue, Q. Z., “Model for effective thermal conductivity of carbon nanotube-based composites,”
Physica B, vol. 368, no. 1–4, pp. 302–307, 2005.
[132] Gao, L. and Zhou, X. F., “Differential effective medium theory for thermal conductivity in
nanofluids,” Physics Letters A, vol. 348, no. 3–6, pp. 355–360, 2006.
[133] Xue, Q. Z., “Model for the effective thermal conductivity of carbon nanotube composites,”
Nanotechnology, vol. 17, no. 6, pp. 1655–1660, 2006.
[134] Minnich, A. and Chen, G., “Modified effective medium formulation for the thermal conductivity
of nanocomposites,” Applied Physics Letters, Vol. 91, no. 7, pp. 073105-1–073105-3, 2007.
[135] Jung, J.-Y., and Yoo, J. Y., “Thermal conductivity enhancement of nanofluids in conjunction with
electrical double layer (EDL),” International Journal of Heat and Mass Transfer, vol. 52, no.
1–2, pp. 525–528, 2009.
[136] Zhu, H.-t., Lin, Y.-s., and Yin, Y.-s., “A novel one-step chemical method for preparation of
copper nanofluids,” Journal of Colloid and Interface Science, vol. 277, no.1, pp. 100–103, 2004.

International Journal of Micro-Nano Scale Transport

View publication stats